alt link

alt link
BEHAVIORAL AND BRAIN SCIENCES (2004) 27, 000–000
Printed in the United States of America
Toward a mechanistic psychology
of dialogue
Martin J. Pickering
University of Edinburgh, Department of Psychology, Edinburgh EH8 9JZ,
United Kingdom
[email protected]
http://www.psy.ed.ac.uk/Staff/academics.html#PickeringMartin
Simon Garrod
University of Glasgow, Department of Psychology, Glasgow G12 8QT, United
Kingdom.
[email protected]
http://staff.psy.gla.ac.uk/~simon/
Abstract: Traditional mechanistic accounts of language processing derive almost entirely from the study of monologue. Yet, the most
natural and basic form of language use is dialogue. As a result, these accounts may only offer limited theories of the mechanisms that underlie language processing in general. We propose a mechanistic account of dialogue, the interactive alignment account, and use it to derive a number of predictions about basic language processes. The account assumes that, in dialogue, the linguistic representations employed by the interlocutors become aligned at many levels, as a result of a largely automatic process. This process greatly simplifies
production and comprehension in dialogue. After considering the evidence for the interactive alignment model, we concentrate on three
aspects of processing that follow from it. It makes use of a simple interactive inference mechanism, enables the development of local dialogue routines that greatly simplify language processing, and explains the origins of self-monitoring in production. We consider the need
for a grammatical framework that is designed to deal with language in dialogue rather than monologue, and discuss a range of implications of the account.
Keywords: common ground; dialogue; dialogue routines; language comprehension; language production; monitoring, perception-behavior link
1. Introduction
Psycholinguistics aims to describe the psychological processes underlying language use. The most natural and basic form of language use is dialogue: Every language user,
including young children and illiterate adults, can hold a
conversation, whereas reading, writing, preparing speeches
and even listening to speeches are far from universal skills.
Therefore, a central goal of psycholinguistics should be to
provide an account of the basic processing mechanisms that
are employed during natural dialogue.
Currently, there is no such account. Existing mechanistic accounts are concerned with the comprehension and
production of isolated words or sentences, or with the processing of texts in situations where no interaction is possible, such as in reading. In other words, they rely almost entirely on monologue. Hence, theories of basic mechanisms
depend on the study of a derivative form of language processing. We argue that such theories are limited and inadequate accounts of the general mechanisms that underlie
processing. In contrast, this paper outlines a mechanistic
theory of language processing that is based on dialogue, but
that applies to monologue as a special case.
Why has traditional psycholinguistics ignored dialogue?
There are probably two main reasons, one practical and one
theoretical. The practical reason is that it is generally assumed to be too hard or impossible to study, given the de© 2004 Cambridge University Press
0140-525X/04 $12.50
gree of experimental control necessary. Studies of language
comprehension are fairly straightforward in the experimental psychology tradition – words or sentences are stimuli that can be appropriately controlled in terms of their
characteristics (e.g., frequency) and presentation conditions (e.g., randomized order). Until quite recently it was
also assumed that imposing that level of control in many
language production studies was impossible. Thus, Bock
(1996) points to the problem of “exuberant responsing” –
how can the experimenter stop subjects from saying whatever they want? However, it is now regarded as perfectly
possible to control presentation so that people produce the
appropriate responses on a high proportion of trials, even
in sentence production (e.g., Bock 1986a; Levelt & Maassen 1981).
Contrary to many people’s intuitions, the same experimental control is possible with dialogue. For example,
Branigan et al. (2000) showed effects of the priming of syntactic structure during language production in dialogue that
were exactly comparable to the priming shown in isolated
sentence production (Bock 1986b) or sentence recall (Potter & Lombardi 1998). In Branigan et al.’s study, the degree
of control of independent and dependent variables was no
different from in Bock’s study, even though the experiment
involved two participants engaged in a dialogue rather than
one participant producing sentences in isolation. Similar
control is exercised in studies by Clark and colleagues (e.g.,
1
Pickering and Garrod: Toward a mechanistic psychology of dialogue
Brennan & Clark 1996; Wilkes-Gibbs & Clark 1992; also
Brennan & Schober 2001; Horton & Keysar 1996). Wellcontrolled studies of language production in dialogue may
require some ingenuity, but such experimental ingenuity
has always been a strength of psychology.
The theoretical reason why psycholinguistics has ignored
dialogue is that psycholinguistics has derived most of its
predictions from generative linguistics, and generative linguistics has developed theories of isolated, decontextualized sentences that are used in texts or speeches, that is, in
monologue. In contrast, dialogue is inherently interactive
and contextualized: Each interlocutor both speaks and
comprehends during the course of the interaction; each interrupts both others and himself; on occasion two or more
speakers collaborate in producing the same sentence
(Coates 1990). So it is not surprising that generative linguists commonly view dialogue as being of marginal grammaticality, contaminated by theoretically uninteresting
complexities. Dialogue sits ill with the competence/performance distinction assumed by most generative linguistics
(Chomsky 1965), because it is hard to determine whether a
particular utterance is “well-formed” or not (or even
whether that notion is relevant to dialogue). Thus, linguistics has tended to concentrate on developing generative
grammars and related theories for isolated sentences; and
psycholinguistics has tended to develop processing theories
that draw upon the rules and representations assumed by
generative linguistics. So far as most psycholinguists have
thought about dialogue, they have tended to assume that
the results of experiments on monologue can be applied to
the understanding of dialogue, and that it is more profitable
to study monologue because it is “cleaner” and less complex
than dialogue. Indeed, they have commonly assumed that
dialogue simply involves chunks of monologue stuck together.
The main advocate of the experimental study of dialogue
is Clark. However, his primary focus is on the nature of the
Martin J. Pickering bio goes here
Simon Garrod bio goes here
2
BEHAVIORAL AND BRAIN SCIENCES (2004) 27:2
strategies employed by the interlocutors rather than basic
processing mechanisms. Clark (1996) contrasts the “language-as-product” and “language-as-action” traditions. The
language-as-product tradition is derived from the integration of information-processing psychology with generative
grammar and focuses on mechanistic accounts of how people compute different levels of representation. This tradition has typically employed experimental paradigms and
decontextualized language; in our terms, monologue. In
contrast, the language-as-action tradition emphasizes that
utterances are interpreted with respect to a particular context and takes into account the goals and intentions of the
participants. This tradition has typically considered processing in dialogue using apparently natural tasks (e.g.,
Clark 1992; Fussell & Krauss 1992). Whereas psycholinguistic accounts in the language-as-product tradition are
admirably well-specified, they are almost entirely decontextualized and, quite possibly, ecologically invalid. On the
other hand, accounts in the language-as-action tradition
rarely make contact with the basic processes of production
or comprehension, but rather present analyses of psycholinguistic processes purely in terms of their goals (e.g.,
the formulation and use of common ground; Clark 1985;
1996; Clark & Marshall 1981).
This dichotomy is a reasonable historical characterization. Almost all mechanistic theories happen to be theories
of the processing of monologue; and theories of dialogue
are almost entirely couched in intentional non-mechanistic
terms. But this need not be. The goals of the language-asproduct tradition are valid and important, but researchers
concerned with mechanisms should investigate the use of
contextualized language in dialogue.
In this paper we propose a mechanistic account of dialogue and use it to derive a number of predictions about
basic language processing. The account assumes that in
dialogue, production and comprehension become tightly
coupled in a way that leads to the automatic alignment of
linguistic representations at many levels. We argue that the
interactive alignment process greatly simplifies language
processing in dialogue. It does so (1) by supporting a
straightforward interactive inference mechanism, (2) by
enabling interlocutors to develop and use routine expressions, and (3) by supporting a system for monitoring language processing.
The first part of the paper presents the main argument
(sects. 2–6). In section 2 we show how successful dialogue
depends on alignment of representations between interlocutors at different linguistic levels. In section 3 we contrast the interactive alignment model developed in section
2 with the autonomous transmission account that underpins
current mechanistic psycholinguistics. Section 4 describes
a simple interactive repair mechanism that supplements
the interactive alignment process. We argue that this repair
mechanism can reestablish alignment when interlocutors’
representations diverge without requiring them to model
each other’s mental states. Thus, interactive alignment and
repair enable interlocutors to get around many of the problems normally associated with establishing what Stalnaker
(1978) called common ground. The interactive alignment
process leads to the use of routine or semi-fixed expressions. In section 5 we argue that such “dialogue routines”
greatly simplify language production and comprehension
by short-circuiting the decision making processes. Finally,
in section 6 we discuss how interactive alignment enables
Pickering and Garrod: Toward a mechanistic psychology of dialogue
interlocutors to monitor dialogue with respect to all levels
at which they can align.
The second part of the paper explores implications of the
interactive alignment account. In section 7 we discuss implications for linguistic theory. In section 8 we argue for a
graded distinction between dialogue and monologue in
terms of different degrees of coupling between speaker and
listener. In section 9 we argue that the interactive alignment
account may have broader implications in terms of current
developments in areas such as social interaction, language
acquisition, and imitation more generally. Finally, in section
10 we enumerate the differences between the interactive
alignment model developed in the paper and the more traditional autonomous transmission account of language processing.
2. The nature of dialogue and the alignment of
representations
Table 1 shows a transcript of a conversation between two
players in a cooperative maze game (Garrod & Anderson
1987). In this extract one player A is trying to describe his
position to his partner B, who is viewing the same maze on
a computer screen in another room. The maze is shown in
Figure 1.1
At first glance the language looks disorganized. Many of
the utterances are not grammatical sentences (e.g., only
one of the first six contains a verb). There are occasions
when production of a sentence is shared between speakers,
as in (7–8) and (43–44). It often seems that the speakers
do not know how to say what they want to say. For instance,
A describes the same position quite differently in (4) “two
along from the bottom one up,” and (46) “two along, two
up.”
Figure 1. Schematic representation of the maze being described
in the conversation shown in Table 1. The crossed bars indicate
closed paths. The arrow points to the position being described by
the utterances marked in bold in the table.
In fact the sequence is quite orderly so long as we assume
that dialogue is a joint activity (Clark 1996; Clark & WilkesGibbs 1986). In other words, it involves cooperation between interlocutors in a way that allows them to sufficiently
understand the meaning of the dialogue as a whole; and this
meaning results from these joint processes. In Lewis’
(1969) terms, dialogue is a game of cooperation, where both
participants “win” if both understand the dialogue, and neither “wins” if one or both do not understand.
Table 1. Example dialogue taken from Garrod and Anderson (1987)
1— —B: . . . Tell me where you are?
2— —A: Ehm : Oh God (laughs)
3— —B: (laughs)
4— —A: Right : two along from the bottom one up:*
5— —B: Two along from the bottom, which side?
6— —A: The left : going from left to right in the second box.
7— —B: You’re in the second box.
8— —A: One up (1 sec.) I take it we’ve got identical mazes?
9— —B: Yeah well : right, starting from the left, you’re one along:
10— —A: Uh-huh:
11— —B: and one up?
12— —A: Yeah, and I’m trying to get to . . .
[ 28 utterances later ]
41— —B: You are starting from the left, you’re one along, one up? (2 sec.)
42— —A: Two along : I’m not in the first box, I’m in the second box:
43— —B: You’re two along:
44— —A: Two up (1 sec.) counting the : if you take : the first box as being one up:
45— —B: (2 sec.) Uh-huh:
46— —A: Well : I’m two along, two up (1.5 sec.)
47— —B: Two up ? :
48— —A: Yeah (1 sec.) so I can move down one:
49— —B: Yeah I see where you are:
* The position being described in the utterances shown in bold is identified with an arrow in Figure 1. Colons mark noticeable pauses
of less than 1 second.
BEHAVIORAL AND BRAIN SCIENCES (2004) 27:2
3
Pickering and Garrod: Toward a mechanistic psychology of dialogue
Conversational analysts argue that dialogue turns are
linked across interlocutors (Sacks et al. 1974; Schegloff &
Sacks 1973). A question, such as (1) “Tell me where you
are?”, calls for an answer, such as (4) “Two along from the
bottom one up.” Even a statement like (4) “Right, two along
from the bottom one up,” cannot stand alone. It requires either an affirmation or some form of query, such as (5) “Two
along from the bottom, which side?” (Linnell 1998). This
means that production and comprehension processes become coupled. B produces a question and expects an answer of a particular type; A hears the question and has to
produce an answer of that type. For example, after saying
“Tell me where you are?” in (1), B has to understand “two
along from the bottom one up” in (4) as a reference to A’s
position on the maze; any other interpretation is ruled out.
Furthermore, the meaning of what is being communicated
depends on the interlocutors’ agreement or consensus
rather than on dictionary meanings (Brennan & Clark
1996) and is subject to negotiation (Linnell 1998, p. 74).
Take for example utterances (4–11) in the fragment shown
above. In utterance (4), A describes his position as “Two
along from the bottom one up,” but the final interpretation
is only established at the end of the first exchange when
consensus is reached on a rather different description by B
(9–11) “You’re one along . . . and one up?” These examples
demonstrate that dialogue is far more coordinated than it
might initially appear.
At this point we should distinguish two notions of coordination that have become rather confused in the literature.
According to one notion (Clark 1985), interlocutors are coordinated in a successful dialogue just as participants in any
successful joint activity are coordinated (e.g., ballroom
dancers, lumberjacks using a two-handed saw). According
to the other notion, coordination occurs when interlocutors
share the same representation at some level (Branigan et al.
2000; Garrod & Anderson 1987). To remove this confusion,
we refer to the first notion as coordination and the second
as alignment. Specifically, alignment occurs at a particular
level when interlocutors have the same representation at
that level. Dialogue is a coordinated behavior (just like ballroom dancing). However, the linguistic representations that
underlie coordinated dialogue come to be aligned, as we
claim below.
We now argue six points: (1) Alignment of situation models (Zwaan & Radvansky 1998) forms the basis of successful dialogue; (2) the way that alignment of situation models
is achieved is by a primitive and resource-free priming
mechanism; (3) the same priming mechanism produces
alignment at other levels of representation, such as the lexical and syntactic; (4) interconnections between the levels
mean that alignment at one level leads to alignment at other
levels; (5) another primitive mechanism allows interlocutors to repair misaligned representations interactively; and
(6) more sophisticated and potentially costly strategies that
depend on modeling the interlocutor’s mental state are only
required when the primitive mechanisms fail to produce
alignment. On this basis, we propose an interactive alignment account of dialogue in the next section.
2.1. Alignment of situation models is central to
successful dialogue
A situation model is a multi-dimensional representation of
the situation under discussion (Johnson-Laird 1983; San4
BEHAVIORAL AND BRAIN SCIENCES (2004) 27:2
ford & Garrod 1981; van Dijk & Kintsch 1983; Zwaan &
Radvansky 1998). According to Zwaan and Radvansky, the
key dimensions encoded in situation models are space,
time, causality, intentionality, and reference to main individuals under discussion. They discuss a large body of research that demonstrates that manipulations of these dimensions affect text comprehension (e.g., people are faster
to recognize that a word has previously been mentioned
when that word refers to something that is spatially, temporally, or causally related to the current topic). Such models are assumed to capture what people are “thinking about”
while they understand a text, and therefore are in some
sense within working memory (they can be contrasted with
linguistic representations on the one hand and general
knowledge on the other).
Most work on situation models has concentrated on comprehension of monologue (normally, written texts) but they
can also be employed in accounts of dialogue, with interlocutors developing situation models as a result of their interaction (Garrod & Anderson 1987). More specifically, we
assume that in successful dialogue, interlocutors develop
aligned situation models. For example, in Garrod and Anderson, players aligned on particular spatial models of the
mazes being described. Some pairs of players came to refer
to locations using expressions like right turn indicator, upside down T shape, or L on its side. These speakers represented the maze as an arrangement of patterns or figures.
In contrast, the pair illustrated in the dialogue in Table 1
aligned on a spatial model in which the maze was represented as a network of paths linking the points they described to prominent positions on the maze (e.g., the bottom left corner). Pairs often developed quite idiosyncratic
spatial models, but both interlocutors developed the same
model (Garrod & Anderson 1987; Garrod & Doherty 1994;
see also Markman & Makin 1998).
Alignment of situation models is not necessary in principle for successful communication. It would be possible to
communicate successfully by representing one’s interlocutor’s situation model, even if that model were not the same
as one’s own. For instance, one player could represent the
maze according to a figure scheme but know that his partner represented it according to a path scheme, and vice
versa. But this would be wildly inefficient as it would require maintaining two underlying representations of the situation, one for producing one’s own utterances and the
other for comprehending one’s interlocutor’s utterances.
Even though communication might work in such cases, it is
unclear whether we would claim that the people understood the same thing. More critically, it would be computationally very costly to have fundamentally different representations. In contrast, if the interlocutors’ representations
are basically the same, there is no need for listener modeling.
Under some circumstances storing the fact that one’s interlocutors represent the situation differently from oneself
is necessary (e.g., in deception, or when trying to communicate to one interlocutor information that one wants to
conceal from another). But even in such cases, many aspects of the representation will be shared (e.g., I might lie
about my location, but would still use a figural representation to do so if that was what you were using). Additionally,
it is clearly tricky to perform such acts of deception or concealment (Clark & Schaefer 1987). These involve sophisticated strategies that do not form part of the basic process
Pickering and Garrod: Toward a mechanistic psychology of dialogue
of alignment, and are difficult because they require the
speaker to concurrently develop two representations.
Of course, interlocutors need not entirely align their situation models. In any conversation where information is
conveyed, the interlocutors must have somewhat different
models, at least before the end of the conversation. In cases
of partial misunderstanding, conceptual models will not be
entirely aligned. In (unresolved) arguments, interlocutors
have representations that cannot be identical. But they
must have the same understanding of what they are discussing in order to disagree about a particular aspect of it
(e.g., Sacks 1987). For instance, if two people are arguing
the merits of the Conservative versus the Labour parties for
the U.K. government, they must agree about who the
names refer to, roughly what the politics of the two parties
are, and so on, so that they can disagree on their evaluations. In Lewis’ (1969) terms, such interlocutors are playing a game of cooperation with respect to the situation
model (e.g., they succeed insofar as their words refer to the
same entities), even though they may not play such a game
at other “higher” levels (e.g., in relation to the argument itself). Therefore, we assume that successful dialogue involves approximate alignment at the level of the situation
model at least.
2.2. Achieving alignment of situation models
In theory, interlocutors could achieve alignment of their
models through explicit negotiation, but in practice they
normally do not (Brennan & Clark 1996; Clark & WilkesGibbs 1986; Garrod & Anderson 1987; Schober 1993). It is
quite unusual for people to suggest a definition of an expression and obtain an explicit assent from their interlocutor. Instead, “global” alignment of models seems to result
from “local” alignment at the level of the linguistic representations being used. We propose that this works via a
priming mechanism, whereby encountering an utterance
that activates a particular representation makes it more
likely that the person will subsequently produce an utterance that uses that representation. (On this conception,
priming underpins the alignment mechanism and should
not simply be regarded as a behavioral effect.) In this case,
hearing an utterance that activates a particular aspect of a
situation model will make it more likely that the person will
use an utterance consistent with that aspect of the model.
This process is essentially resource-free and automatic.
This was pointed out by Garrod and Anderson (1987) in
relation to their principle of output/input coordination.
They noted that in the maze game task speakers tended to
make the same semantic and pragmatic choices that held
for the utterances that they had just encountered. In other
words, their outputs tended to match their inputs at the
level of the situation model. As the interaction proceeded,
the two interlocutors therefore came to align the semantic
and pragmatic representations used for generating output
with the representations used for interpreting input.
Hence, the combined system (i.e., the interacting dyad) is
completely stable only if both subsystems (i.e., speaker A’s
representation system and speaker B’s representation system) are aligned. In other words, the dyad is only in equilibrium when what A says is consistent with B’s currently active semantic and pragmatic representation of the dialogue
and vice versa (see Garrod & Clark 1993). Thus, because
the two parties to a dialogue produce aligned language, the
underlying linguistic representations also tend to become
aligned. In fact, the output/input coordination principle applies more generally. Garrod and Anderson also assumed
that it held for lexical representations. We argue that alignment holds at a range of levels, including the situational
model and the lexical level, but also at other levels, such as
the syntactic, as discussed in section 2.3, and that alignment
“percolates” between levels, as discussed in section 2.4.
Other work suggests that specific dimensions of situation
models can be aligned. With respect to the spatial dimension, Schober (1993) found that interlocutors tended to
adopt the same reference frame as each other. When interlocutors face each other, terms like on the left are ambiguous depending on whether the speaker takes what we can
call an egocentric or an allocentric reference frame.
Schober found that if, for instance, A said on the left meaning on A’s left (i.e., an egocentric reference frame), then B
would subsequently describe similar locations as on the
right (also taking an egocentric frame of reference). Other
evidence for priming of reference frames comes from experiments conducted outside dialogue (which involve the
same priming mechanism in our account). Thus, CarlsonRadvansky and Jiang (1998) found that people responded
faster on a sentence-picture verification task if the reference frame (in this case, egocentric vs. intrinsic to the object) used on the current trial was the same as the reference
frame used on the previous trial.2
So far we have assumed that the different components of
the situation model are essentially separate (in accord with
Zwaan & Radvansky 1998), and that they can be primed individually. But in a particularly interesting study, Boroditsky (2000) found that the use of a temporal reference frame
can be primed by a spatial reference frame. Thus, if people
had just verified a sentence describing a spatial scenario
that assumed a particular frame of reference (in her terms,
ego moving or object moving), they tended to interpret a
temporal expression in terms of an analogous frame of reference. Her results demonstrate priming of a structural aspect of the situation model that is presumably shared between the spatial and temporal dimensions at least. Indeed,
work on analogy more generally suggests that it should be
possible to prime abstract characteristics of the situation
model (e.g., Gentner & Markman 1997; Markman & Gentner 1993), and that such processes should contribute to
alignment in dialogue.
There is some evidence for alignment of situation models in comprehension. Garrod and Anderson (1987) found
that players in the maze game would query descriptions
from an interlocutor that did not match their own previous
descriptions (see sect. 4). Recently, Brown-Schmidt et al.
(in press) have provided direct and striking evidence for
alignment in comprehension. Previous work has shown that
eye movements during scene perception are a strong indication of current attention, and that they can be used to index the rapid integration of linguistic and contextual information during comprehension (Chambers et al. 2002;
Tanenhaus et al. 1995). Brown-Schmidt et al. monitored
eye movements during unscripted dialogue, and found that
the entities considered by the listener directly reflected the
entities being considered by the speaker at that point. For
example, if the speaker used a referring expression which
was formally ambiguous but which the speaker used to refer to a specific entity (and hence regarded as disambiguated), the listener also looked at that entity. Hence,
BEHAVIORAL AND BRAIN SCIENCES (2004) 27:2
5
Pickering and Garrod: Toward a mechanistic psychology of dialogue
whatever factors were constraining the speaker’s situation
model, were also constraining the listener’s situation model.
2.3. Achieving alignment at other levels
Dialogue transcripts are full of repeated linguistic elements
and structures indicating alignment at various levels in
addition to that of the situation model (Aijmer 1996;
Schenkein 1980; Tannen 1989). Alignment of lexical processing during dialogue was specifically demonstrated by
Garrod and Anderson (1987), as in the extended example
in Table 1 (see also Garrod & Clark 1993; Garrod & Doherty 1994), and by Clark and colleagues (Brennan & Clark
1996; Clark & Wilkes-Gibbs 1986; Wilkes-Gibbs & Clark
1992). These latter studies show that interlocutors tend to
develop the same set of referring expressions to refer to
particular objects, and that the expressions become shorter
and more similar on repetition with the same interlocutor
and are modified if the interlocutor changes.
Levelt and Kelter (1982) found that speakers tended to
reply to “What time do you close?” or “At what time do you
close?” (in Dutch) with a congruent answer (e.g., “Five o’clock” or “At five o’clock”). This alignment may be syntactic (repetition of phrasal categories) or lexical (repetition of
at). Branigan et al. (2000) found clear evidence for syntactic alignment in dialogue. Participants took it in turns to describe pictures to each other (and to find the appropriate
picture in an array). One speaker was actually a confederate of the experimenter and produced scripted responses,
such as “the cowboy offering the banana to the robber” or
“the cowboy offering the robber the banana.” The syntactic structure of the confederate’s description strongly influenced the syntactic structure of the experimental subject’s
description. Branigan et al.’s work extends “syntactic priming” to dialogue. Bock (1986b) showed that speakers
tended to repeat syntactic form under circumstances in
which alternative non-syntactic explanations could be excluded (Bock 1989; Bock & Loebell 1990; Bock et al. 1992;
Hartsuiker & Westenberg 2000; Pickering & Branigan
1998; Potter & Lombardi 1998; cf. Smith & Wheeldon
2001, and see Pickering & Branigan 1999, for a review).
Branigan et al.’s (2000) results support the claim that
priming activates representations and not merely procedures that are associated with production (or comprehension) – in other words, that the explanation for syntactic
priming effects is closely related to the explanation of alignment in general. This suggests an important “parity” between the representations used in production and those
used in comprehension (see sect. 3.2). Interestingly, Branigan et al. (2000) found very large priming effects compared
to the syntactic priming effects that occur in isolation.
There are two reasons why this might be the case. First, a
major reason why priming effects occur is to facilitate alignment, and therefore they are likely to be particularly strong
during natural interactions. In the Branigan et al. (2000)
study, participants responded at their own pace, which
should have made processing “natural,” and hence conducive to strong priming. Second, we would expect interlocutors to have their production systems highly activated
even when listening, because they have to be constantly
prepared to become the speaker, whether by taking the
floor or simply making a backchannel contribution.
If syntactic alignment is due, in part, to the interactional
nature of dialogue, then the degree of syntactic alignment
6
BEHAVIORAL AND BRAIN SCIENCES (2004) 27:2
should reflect the nature of the interaction between
speaker and listener. As Clark and Schaeffer (1987; see also
Schober & Clark 1989; Wilkes-Gibbs & Clark 1992) have
demonstrated, there are basic differences between addressees and other listeners. So we might expect stronger
alignment for addressees than for other listeners. To test for
this, Branigan et al. (2002) had two speakers take turns describing cards to a third person, so the two speakers heard
but did not speak to each other. Priming occurred under
these conditions, but it was weaker than when two speakers simply responded to each other. Hence, syntactic alignment is affected by speaker participation in dialogue. Although, we would claim, the same representations are
activated under these conditions as during dyadic interaction, the closeness of dyadic interaction means that it leads
to stronger priming. For instance, we assume that the production system is active (and hence is ready to produce an
interruption) when the addressee is listening to the speaker.
By contrast, Branigan et al.’s (2002) side participant is not
in a position to make a full contribution, and hence does not
need to activate his production system to the same extent.
Alignment also occurs at the level of articulation. It has
long been known that as speakers repeat expressions, articulation becomes increasingly reduced (i.e., the expressions
are shortened and become more difficult to recognize when
heard in isolation; Fowler & Housum 1987). However,
Bard et al. (2000) found that reduction was just as extreme
when the repetition was by a different speaker in the dialogue as it was when the repetition was by the original
speaker. In other words, whatever is happening to the
speaker’s articulatory representations is also happening to
his interlocutor’s. There is also evidence that interlocutors
align accent and speech rate (Giles & Powesland 1975;
Giles et al. 1992).
Finally, there is some evidence for alignment in comprehension. Levelt and Kelter (1982, Experiment 6) found that
people judged question-answer pairs involving repeated
form as more natural than pairs that did not; and that the
ratings of naturalness were highest for the cases where
there was the strongest tendency to repeat form. This suggests that speakers prefer their interlocutors to respond
with an aligned form.
2.4. Alignment at one level leads to alignment
at another
So far, we have concluded that successful dialogue leads to
the development of both aligned situation models and
aligned representations at all other linguistic levels. There
are good reasons to believe that this is not coincidental, but
rather that aligned representations at one level lead to
aligned representations at other levels.
Consider the following two examples of influences between levels. First, Garrod and Anderson (1987) found that
once a word had been introduced with a particular interpretation it was not normally used with any other interpretation in a particular stretch of maze-game dialogue. For instance, the word row could refer either to an implicitly
ordered set of horizontal levels of boxes in the maze (e.g.,
with descriptions containing an ordinal like “I’m on the
fourth row”) or to an unordered set of levels (e.g., with descriptions that do not contain ordinals like “I’m on the bottom row”).3 Speakers who had adopted one of these local
interpretations of row and needed to refer to the other
Pickering and Garrod: Toward a mechanistic psychology of dialogue
would introduce a new term, such as line or level. Thus, they
would talk of the fourth row and the bottom line, but not the
fourth row and the bottom row (see Garrod & Anderson
1987, p. 202). Aligned use of a word seemed to go with a
specific aligned interpretation of that word. Restricting usage in this way allows dialogue participants to assume quite
specific unambiguous interpretations for expressions. Furthermore, if a new expression is introduced they can assume that it has a different interpretation from a previous
expression, even if the two expressions are “dictionary synonyms.” This process leads to the development of a lexicon
of expressions relevant to the dialogue (see sect. 5). What
interlocutors are doing is acquiring new senses for words or
expressions. To do this, they use the principle of contrast
just like children acquiring language (e.g., Clark 1993).
Second, it has been shown repeatedly that priming at one
level can lead to more priming at other levels. Specifically,
syntactic alignment (or “syntactic priming”) is enhanced
when more lexical items are shared. In Branigan et al.’s
(2000) study, the confederate produced a description using
a particular verb (e.g., the nun giving the book to the clown).
Some experimental subjects then produced a description
using the same verb (e.g., the cowboy giving the banana to
the burglar), whereas other subjects produced a description
using a different verb (e.g., the cowboy handing the banana
to the burglar). Syntactic alignment was considerably enhanced if the verb was repeated (as also happens in monologue; Pickering & Branigan 1998). Thus, interlocutors do
not align representations at different linguistic levels independently. Likewise, Cleland and Pickering (2003) found
people tended to produce noun phrases like the sheep that’s
red as opposed to the red sheep more often after hearing the
goat that’s red than after the book that’s red. This demonstrates that semantic relations between lexical items enhance syntactic priming.
These effects can be modeled in terms of a lexical representation outlined in Pickering and Branigan (1998). A
node representing a word (i.e., its lemma; Levelt et al. 1999;
cf. Kempen & Huijbers 1983) is connected to nodes that
specify its syntactic properties. So the node for give is connected to a node specifying that it can be used with a noun
phrase and a prepositional phrase. Processing giving the
book to the clown activates both of these nodes and therefore makes them both more likely to be employed subsequently. However, it also strengthens the link between
these nodes, on the principle that coactivation strengthens
association. Thus, the tendency to align at one level, such
as the syntactic, is enhanced by alignment at another level,
such as the lexical. Cleland and Pickering’s (2003) finding
demonstrates that exact repetition at one level is not necessary: the closer the relationship at one level (e.g., the semantic), the stronger the tendency to align at the other
(e.g., the syntactic). Note that we can make use of this tendency to determine which specific levels are linked.
In comprehension, there is evidence for parallelism at
one level occurring more when there is parallelism at another level. Thus, pronouns tend to be interpreted as coreferential with an antecedent in the same grammatical role
(e.g., “William hit Oliver and Rod slapped him” is interpreted as Rod slapping Oliver; Sheldon 1974; Smyth 1994).
Likewise, the likelihood of a gapping interpretation of an
ambiguous sentence is greater if the relevant arguments are
parallel (e.g., “Bill took chips to the party and Susan to the
game” is often given an interpretation where Susan took
chips to the game; Carlson 2001). Finally, Gagné and
Shoben (2002; cf. Gagné 2001) found evidence that interpreting a compound as having a particular semantic relation
(e.g., type of doctor in adolescent doctor) was facilitated by
prior interpretation of a compound containing either the
same noun or adjective that used the same relation (e.g.,
adolescent magazine or animal doctor). These effects have
only been demonstrated in reading, but we would also expect them to occur in dialogue.
The mechanism of alignment, and in particular the percolation of alignment between levels, has a very important
consequence that we discuss in section 5. Interlocutors will
tend to align expressions at many different levels at the
same time.4 When all levels are aligned, interlocutors will
repeat each others’ expressions in the same way (e.g., with
the same intonation). Hence, dialogue should be highly
repetitive, and should make extensive use of fixed expressions. Importantly, fixed expressions should be established
during the dialogue, so that they become dialogue routines.
2.5. Recovery from misalignment
Of course, these primitive processes of alignment are not
foolproof. For example, interlocutors might align at a “superficial” level but not at the level of the situation model
(e.g., if they both refer to John but do not realize that they
are referring to different Johns; cf. Garrod & Clark 1993).
In such cases, interlocutors need to be able to appeal to
other mechanisms to establish or reestablish alignment.
The account is not complete until we outline such mechanisms, which we do in section 4 below. For now, we simply
assume that such mechanisms exist and are needed to supplement the basic process of alignment.
3. The interactive alignment model of dialogue
processing
The interactive alignment model assumes that successful
dialogue involves the development of aligned representations by the interlocutors. This occurs by priming mechanisms at each level of linguistic representation, by percolation between the levels so that alignment at one level
enhances alignment at other levels, and by repair mechanisms when alignment goes awry. Figure 2 illustrates the
process of alignment in fairly abstract terms. It shows the
levels of linguistic representation computed by two interlocutors and ways in which those representations are
linked. Critically, Figure 2 includes links between the interlocutors at multiple levels.
In this section, we elucidate the figure in three ways.
First, we contrast it with a more traditional “autonomous
transmission” account, as represented in Figure 3, where
multiple links between interlocutors do not exist. Second,
we interpret these links as corresponding to channels
whereby priming occurs. Finally, we argue that the bi-directional nature of the links means that there must be parity between production and comprehension processes.
3.1. Interactive alignment versus autonomous
transmission
In the autonomous transmission account, the transfer of information between producers and comprehenders takes
BEHAVIORAL AND BRAIN SCIENCES (2004) 27:2
7
Pickering and Garrod: Toward a mechanistic psychology of dialogue
Figure 2. A and B represent two interlocutors in a dialogue in this schematic representation of the stages of comprehension and production processes according to the interactive alignment model. The details of the various levels of representation and interactions between levels are chosen to illustrate the overall architecture of the system rather than to reflect commitment to a specific model.
place via decoupled production and comprehension
processes that are “isolated” from each other (see Fig. 3).
The speaker (or writer) formulates an utterance on the basis of his representation of the situation. Crudely, a non-linguistic idea or “message” is converted into a series of linguistic representations, with earlier ones being syntactic,
and later ones being phonological. The final linguistic representation is converted into an articulatory program,
which generates the actual sound (or hand movements)
(e.g., Levelt 1989). Each intermediate representation
serves as a “way station” on the road to production – its significance is internal to the production process. Hence,
there is no reason for the listener to be affected by these intermediate representations.
In turn, the listener (or reader) decodes the sound (or
movements) by converting the sound into successive levels
of linguistic representation until the message is recovered
(if the communication is successful). He then infers what
the speaker (or writer) intended on the basis of his autonomous representation of the situation. So, from a processing point of view, speakers and listeners act in isolation.
The only link between the two is in the information conveyed by the utterances themselves (Cherry 1956). Each
act of transmission is treated as a discrete stage, with a particular unit being encoded into sound by the speaker, being
transmitted as sound, and then being decoded by the listener. Levels of linguistic representation are constructed
during encoding and decoding, but there is no particular association between the levels of representation used by the
speaker and listener. Indeed, there is even no reason to assume that the levels will be the same, nor that the levels involved in comprehension should constrain those in production or vice versa. Hence, Figure 3 could just as well
8
BEHAVIORAL AND BRAIN SCIENCES (2004) 27:2
involve different levels of representation for speaker and
listener.
The autonomous transmission model is not appropriate
for dialogue because, in dialogue, production and comprehension processes are coupled (Garrod 1999). In formulating an utterance the speaker is guided by what has just been
said to him and in comprehending the utterance the listener is constrained by what the speaker has just said, as in
the example dialogue in Table 1. The interlocutors build up
utterances as a joint activity (Clark 1996), with interlocutors
often interleaving production and comprehension tightly.
They also align at many different levels of representation,
as discussed in section 2. Thus, in dialogue each level of representation is causally implicated in the process of communication and these intermediate representations are retained implicitly. Because alignment at one level leads to
alignment at others, the interlocutors come to align their
situation models and hence are able to understand each
other. This follows from the interactive alignment model
described in Figure 2, but is not reflected in the autonomous transmission account in Figure 3.
3.2. Channels of alignment
The horizontal links in Figure 2 correspond to channels by
which alignment takes place. The communication mechanism used by these channels is priming. Thus, we assume
that lexical priming leads to the alignment at the lexical
level, syntactic priming leads to alignment at the syntactic
level, and so on. Although fully specified theories of how
such priming operates are not available for all levels, sections 2.2 and 2.3 described some of the evidence to support
priming at these levels, and detailed mechanisms of prim-
Pickering and Garrod: Toward a mechanistic psychology of dialogue
Figure 3. A and B represent two interlocutors in a dialogue in this schematic representation of the stages of comprehension and production processes according to the autonomous transmission account. The details of the various levels of representation and interactions
between levels are chosen to illustrate the overall architecture of the system rather than to reflect commitment to a specific model.
ing are proposed in many of the papers referred to there.
As an example, Branigan et al. (2000) provided an account
of syntactic alignment in dialogue that involved priming of
syntactic information at the lemma stratum. Because channels of alignment are bi-directional, the model predicts that
if evidence is found for alignment in one direction (e.g.,
from comprehension to production) it should also be found
for alignment in the other (e.g., from production to comprehension). Of course, the linguistic information conveyed
by the channels is encoded in sound.
Critically, these channels are direct and automatic (as implied by the term “priming”). The activation of a representation in one interlocutor leads to the activation of the
matching representation in the other interlocutor directly.
There is no intervening “decision box” where the listener
makes a decision about how to respond to the “signal.” Although such decisions do of course take place during dialogue (see sect. 4 below), they do not form part of the basic
interactive alignment process, which is automatic and
largely unconscious. We assume that such channels are similar to the direct and automatic perception-behavior link
that has been proposed to explain the central role of imitation in social interaction (Bargh & Chartrand 1999; Dijksterhuis & Bargh 2001).
Figure 2 therefore indicates how interlocutors can align
in dialogue via the interactive alignment model. It does not
of course provide an account of communication in monologue, but the goal of monologue is not to get to aligned
representations. Instead, the listener attempts to obtain a
specific representation corresponding to the speaker’s message, and the speaker attempts to produce the appropriate
sounds that will allow the listener to do this. Moreover, in
monologue (including writing), the speaker’s and the lis-
tener’s representations can rapidly diverge (or never align
at all). The listener then has to draw inferences on the basis of his knowledge about the speaker, and the speaker has
to infer what the listener has inferred (or simply assume
that the listener has inferred correctly). Of course, either
party could easily be wrong, and these inferences will often
be costly. In monologue, the automatic mechanisms of
alignment are not present (the consequences for written
production are demonstrated in Traxler & Gernsbacher
1992; 1993). It is only when regular feedback occurs that
the interlocutors can control the alignment process.
The role of priming in dialogue is very different from
monologue. In monologue, it can largely be thought of as
an epiphenomenal effect, which is of considerable use to
psycholinguists as a way of investigating representation and
process, but of little importance in itself. However, our
analysis of dialogue demonstrates that priming is the central mechanism in the process of alignment and mutual understanding. Thus, dialogue indicates the important functional role of priming. In conclusion, we regard priming as
underlying the links between the two sides of Figure 2, and
hence the mechanism that drives interactive alignment.
3.3. Parity between comprehension and production
On the autonomous transmission account, the processes
employed in production and comprehension need not draw
upon the same representations (see Fig. 3). By contrast, the
interactive alignment model assumes that the processor
draws upon the same representations (see Fig. 2). This parity means that a representation that has just been constructed for the purposes of comprehension can then be
used for production (or vice versa). This straightforwardly
BEHAVIORAL AND BRAIN SCIENCES (2004) 27:2
9
Pickering and Garrod: Toward a mechanistic psychology of dialogue
explains, for example, why we can complete one another’s
utterances (and get the syntax, semantics, and phonology
correct; see sect. 7.1). It also serves as an explanation of why
syntactic priming in production occurs when the speaker
has only heard the prime (Branigan et al. 2000; Potter &
Lombardi 1998), as well as when he has produced the
prime (Bock 1986b; Pickering & Branigan 1998).
The notion of parity of representation is controversial but
has been advocated by a wide range of researchers working
in very different domains (Calvert et al. 1997; Liberman &
Whalen 2000; MacKay 1987; Mattingly & Liberman 1988).
For example, Goldinger (1998) demonstrated that speech
“shadowers” imitate the perceptual characteristics of a
shadowed word (i.e., their repetition is judged acoustically
more similar to the shadowed word than to another production of the same word by the shadower). Goldinger argued that this vocal imitation in shadowing strongly suggests an underlying perception-production link at the
phonological level.
Parity is also increasingly advocated as a means of explaining perception/action interactions outside language
(Hommel et al. 2001). We return to this issue in section 9.
Note that parity only requires that the representations be
the same. The processes leading to those representations
need not be related (e.g., there is no need for the mapping
between representations to be simply reversed in production and comprehension).
4. Common ground, misalignment, and interactive
repair
In current research on dialogue, the key conceptual notion
has been “common ground,” which refers to background
knowledge shared between the interlocutors (Clark & Marshall 1981). Traditionally, most research on dialogue has assumed that interlocutors communicate successfully when
they share a common ground, and that one of the critical
preconditions for successful communication is the establishment of common ground (Clark & Wilkes-Gibbs 1986).
Establishment of common ground involves a good deal of
modeling of one’s interlocutor’s mental state. In contrast,
our account assumes that alignment of situation models follows from lower-level alignment, and is therefore a much
more automatic process. We argue that interlocutors align
on what we term an implicit common ground, and only go
beyond this to a (full) common ground when necessary. In
particular, interlocutors draw upon common ground as a
means of repairing misalignment when more straightforward means of repair fail.
4.1. Common ground versus implicit common ground
Alignment between interlocutors has traditionally been
thought to arise from the establishment of common, mutual, or joint knowledge (Lewis 1969; McCarthy 1990;
Schiffer 1972). Perhaps the most influential example of this
approach is Clark and Marshall’s (1981) argument that successful reference depends on the speaker and the listener
inferring mutual knowledge about the circumstances surrounding the reference. Thus, for a female speaker to be
certain that a male listener understands what is meant by
“the movie at the Roxy,” she needs to know what he knows
and what he knows that she knows, and so forth. Likewise,
10
BEHAVIORAL AND BRAIN SCIENCES (2004) 27:2
for him to be certain about what she means by “the movie
at the Roxy,” he needs to know what she knows and what
she knows that he knows, and so forth. However, there is no
foolproof procedure for establishing mutual knowledge expressed in terms of this iterative formulation because it requires formulating recursive models of interlocutors’ beliefs (see Barwise 1989; Clark 1996, Ch 4; Halpern & Moses
1990; Lewis 1969). Therefore, Clark and Marshall (1981)
suggested that interlocutors instead infer what Stalnaker
(1978) called the common ground. Common ground reflects what can reasonably be assumed to be known to both
interlocutors on the basis of the evidence at hand. This evidence can be non-linguistic (e.g., if both know that they
come from the same city they can assume a degree of common knowledge about that city; if both admire the same
view and it is apparent to both that they do so, they can infer a common perspective) or can be based on the prior
conversation.
Even though inferring common ground is computationally more feasible than inferring the iterative formulation of
mutual knowledge, it still requires the interlocutor to maintain a very complex situation model that reflects both his
own knowledge and the knowledge that he assumes to be
shared with his partner. To do this, he has to keep track of
the knowledge state of the interlocutor in a way that is separate from his own knowledge state. This is a very stringent
requirement for routine communication, in part because he
has to make sure that this model is constantly updated appropriately (e.g., Halpern & Moses 1990).
In contrast, the interactive alignment model proposes
that the fundamental mechanism that leads to alignment of
situation models is automatic. Specifically, the information
that is shared between the interlocutors constitutes what
we call an implicit common ground. When interlocutors are
well aligned, the implicit common ground is extensive. Unlike common ground, implicit common ground does not
derive from interlocutors explicitly modeling each other’s
beliefs. Implicit common ground is therefore built up automatically and is used in straightforward processes of repair. Interlocutors do of course make use of (full) common
ground on occasion, but it does not form the basis for alignment.
Implicit common ground is effective because an interlocutor builds up a situation model that contains (or at least
foregrounds) information that the interlocutor has processed (either by producing that information or comprehending it). But because the other interlocutor is also present, he comprehends what the first interlocutor produces
and vice versa. This means that both interlocutors foreground the same information, and therefore tend to make
the same additions to their situation models. Of course,
each interlocutor’s situation model will contain some information that he is aware of but the other interlocutor is not,
but as the conversation proceeds and more information is
added, the amount of information that is not shared will be
reduced. Hence, the implicit common ground will be extended. Notice that there is no need to infer the situation
model of one’s interlocutor.
This interactive alignment account predicts that speakers only automatically adapt their utterances when the information can be accessed from their own situation model.
However, because access is from aligned representations,
which reflect the implicit common ground, these adaptations will normally be helpful incidentally for the listener.
Pickering and Garrod: Toward a mechanistic psychology of dialogue
This point was first made by Brown and Dell (1987), who
noted that if speaker and listener have very similar representations of a situation, then most utterances that appear
to be sensitive to the mental state of the listener may in fact
be produced without reference to the listener. This is because what is easily accessible for the speaker will also be
easily accessible for the listener. In fact, the better aligned
speaker and listener are, the closer such an implicit common ground will be to the full common ground, and the less
effort need be exerted to support successful communication.
Hence, we argue that interlocutors do not need to monitor and develop full common ground as a regular, constant
part of routine conversation, as it would be unnecessary and
far too costly. Establishment of full common ground is, we
argue, a specialized and non-automatic process that is used
primarily in times of difficulty (when radical misalignment
becomes apparent). We now argue that speakers and listeners do not routinely take common ground into account
during initial processing. We then discuss interactive repair,
and suggest that full common ground is only used when
simpler mechanisms are ineffective.
4.2. Limits on common ground inference
Studies of both production and comprehension in situations where there is no direct interaction (i.e., situations
that do not allow feedback) indicate that language users do
not always take common ground into account in producing
or interpreting references. For example, Horton and
Keysar (1996) found that speakers under time pressure did
not produce descriptions that took advantage of what they
knew about the listener’s view of the relevant scene. In
other words, the descriptions were formulated with respect
to the speaker’s current knowledge of the scene rather than
with respect to the speaker and listener’s common ground.
Keysar et al. (1998) found that, in visually searching for a
referent for a description, listeners are just as likely to initially look at things that are not part of the common ground
as things that are, and Keysar et al. (2000) found that listeners initially considered objects that they knew were not
visible to their conversational partner. In a similar vein,
Brown and Dell (1987) showed that apparent listener-directed ellipsis was not modulated by information about the
common ground between speaker and listener, but rather
was determined by the accessibility of the information for
the speaker alone (though cf. Lockridge & Brennan 2002,
and Schober & Brennan, 2003, for reservations). Finally,
Ferreira and Dell (2000) found that speakers did not try to
construct sentences that would make comprehension easy
(i.e., by preventing syntactic misanalysis on the part of the
listener).
Even in fully interactive dialogue it is difficult to find evidence for direct listener modeling. For example, it was
originally thought that articulation reduction might reflect
the speaker’s sensitivity to the listener’s current knowledge
(Lindblom 1990). However, Bard et al. (2000) found that
the same level of articulation reduction occurred even after
the speaker encountered a new interlocutor. Degree of reduction seemed to be based only on whether the reference
was given information for the speaker, and not on whether
it was part of the common ground. Additionally, speakers
will sometimes use definite descriptions (to mark the referent as given information; Haviland & Clark 1974) when
the referent is visible to them, even when they know it is
not available to their interlocutor (Anderson & Boyle 1994).
Nevertheless, under certain circumstances interlocutors
do engage in strategic inference relating to (full) common
ground. As Horton and Keysar (1996) found, with less time
pressure speakers often do take account of common ground
in formulating their utterances. Keysar et al. (1998) argued
that listeners can take account of common ground in comprehension under circumstances in which speaker/listener
perspectives are radically different (see also Brennan &
Clark 1996; Schober & Brennan 2003), though they proposed that this occurs at a later monitoring stage, in a
process that they called perspective adjustment. More recently, Hanna et al. (2003) found that listeners looked at an
object in a display less if they knew that the speaker did not
know of the object’s existence (see Nadig & Sedivy 2002, for
a related study with 5–6 year old children). These differences emerged during the earliest stages of comprehension, and therefore suggest that the strongest form of perspective adjustment cannot be correct. However, their task
was repetitive and involved a small number of items, and
listeners were given explicit information about the discrepancies in knowledge. Under such circumstances, it is not
surprising that listeners develop strategies that may invoke
full common ground. During natural dialogue, we predict
that such strategies will not normally be used.
In conclusion, we have argued that performing inferences about common ground is an optional strategy that interlocutors employ only when resources allow. Critically,
such strategies need not always be used, and most “simple”
(e.g., dyadic, non-didactic, non-deceptive) conversation
works without them most of the time.
4.3. Interactive repair using implicit common ground
Of course, the automatic process of alignment does not always lead to appropriately aligned representations. When
interlocutors’ representations are not properly aligned, the
implicit common ground is faulty. We argue that they employ an interactive repair mechanism that helps to maintain
the implicit common ground. The mechanism relies on two
processes: (1) checking whether one can straightforwardly
interpret the input in relation to one’s own representation,
and (2) when this fails, reformulating the utterance in a way
that leads to the establishment of implicit common ground.
Importantly, this mechanism is iterative, in that the original
speaker can then pick up on the reformulation and, if alignment has not been established, reformulate further.
Consider again the example in Table 1. Throughout this
section of dialogue A and B assume subtly different interpretations for two along. A interprets two along by counting the boxes on the maze, whereas B is counting the links
between the boxes (see Fig. 1). This misalignment arises
because the two speakers represent the meaning of expressions like two along differently in this context. In other
words, the implicit common ground is faulty.
Therefore, the players engage in interactive repair, first
by determining that they cannot straightforwardly interpret
the input, and then by reformulation. The reformulation
can be a simple repetition with rising intonation (as in 7), a
repetition with an additional query (as when B says “two
along from the bottom, which side?” in 5), or a more radical restatement (as when A reformulates “two along” as
“second box” in 6). Such reformulation is very common in
BEHAVIORAL AND BRAIN SCIENCES (2004) 27:2
11
Pickering and Garrod: Toward a mechanistic psychology of dialogue
conversation and is described by some linguists as clarification request (see Ginzburg 2001). None of these reformulations requires the speaker to take into account the listener’s situation model. They simply reflect failures to
understand what the speaker is saying in relation to the listener’s own model. They serve to throw the problem back to
the interlocutor who can then attempt a further simple reformulation if he still fails to understand the description. For
example, B says “you’re one along, one up?” (41), which A
reformulates as “Two along” (42). Probably because of this
reformulation, B then asks the clarification request “You’re
two along.” The cycle continues until the misalignment has
been resolved in (44) when A is able to complete B’s utterance without further challenge (for discussion of such embedded repairs see also Jefferson 1987). This repair process
can be regarded as involving a kind of dialogue inference,
but notice that it is externalized, in the sense that it can only
operate via the interaction between the interlocutors. It contrasts with the kind of discourse inference that occurs during text comprehension (or listening to a speech), where the
reader has to mentally infer the writer’s meaning (e.g., via a
bridging inference; Haviland & Clark 1974).
4.4. Interactive repair using full common ground
Interactive repair using implicit common ground is basic
because it only relies on the speaker checking the conversation in relation to his own knowledge of the situation. Of
course there will be occasions when a more complicated
and strategic assessment of common ground may be necessary, most obviously when the basic mechanism fails. In
such cases, the listener may have to draw inferences about
the speaker (e.g., “She has referred to John; does she mean
John Smith or John Brown? She knows both, but thinks I
don’t know Brown, hence she probably means Smith.”).
Such cases may of course involve internalized inference, in
a way that may have more in common with text comprehension than with most aspects of everyday conversation.
But interlocutors may also engage in explicit negotiation or
discussion of the situation models. This appears to occur in
our example when A says “I take it we’ve got identical
mazes” (8).
Use of full common ground is particularly likely when
one speaker is trying to deceive the other or to conceal information (e.g., Clark & Schaefer 1987), or when interlocutors deliberately decide not to align at some level (e.g.,
because each interlocutor has a political commitment to a
different referring expression; Jefferson 1987). Such cases
may involve complex (and probably conscious) reasoning,
and there may be great differences between people’s abilities (e.g., between those with and without an adequate “theory of mind”; Baron-Cohen et al. 2000). For example, Garrod and Clark (1993) found that younger children could not
circumvent the automatic alignment process. Seven-yearold maze game players failed to introduce new description
schemes when they should have done so, because they
could not overcome the pressure to align their description
with the previous one from the interlocutor. By contrast,
older children and adults were twice as likely to introduce
a new description scheme when they had been unable to
understand their partner’s previous description. Whereas
the older children could adopt a strategy of non-alignment
when appropriate, the younger children seemed unable to
do so. Our claim is that these strategic processes are over12
BEHAVIORAL AND BRAIN SCIENCES (2004) 27:2
laid on the basic interactive alignment mechanism. However, such strategies are clearly costly in terms of processing resources and may be beyond the abilities of less skilled
language users.
The strategies discussed above relate specifically to
alignment (either avoiding it or achieving it explicitly), but
of course many aspects of dialogue serve far more complicated functions. A speaker can attempt to produce a particular emotional reaction in the listener by an utterance, or
persuade the listener to act in a particular way or to think
in depth about an issue (e.g., in expert-novice interactions).
Likewise, the speaker can draw complex inferences about
the mental state of the listener and can try to probe this
state by interrogation. Thus, it is important to stress that we
are proposing interactive alignment as the primitive mechanism underlying dialogue, not a replacement for the more
complicated strategies that conversationalists may employ
on occasion.
Nonetheless, we claim that normal conversation does not
routinely require modeling the interlocutor’s mind. Instead, the overlap between interlocutors’ representations is
sufficiently great that a specific contribution by the speaker
will either trigger appropriate changes in the listener’s representation, or will bring about the process of interactive repair. Hence, the listener will retain an appropriate model of
the speaker’s mind, because, in all essential respects, it is the
listener’s representation as well.
Processing monologue is quite different in this respect.
Without automatic alignment and interactive repair the listener can only resort to costly bridging inferences whenever
he fails to understand anything. And, to ensure success, the
speaker will have to design what he says according to what
he knows about the audience (see Clark & Murphy 1982).
In other words, he will have to model the mind or minds of
the audience. Interestingly, Schober (1993) found that
speakers in monologue were more likely to adopt a listeneroriented reference frame than speakers in dialogue, and
that this was costly. Because adopting the listener’s perspective can be very complex (e.g., if different members of
the audience are likely to know different amounts), it is not
surprising that people’s skill at public speaking differs enormously, in sharp contrast to everyday conversation.
5. Alignment and routinization
The process of alignment means that interlocutors draw
upon representations that have been developed during the
dialogue. Thus, it is not always necessary to construct representations that are used in production or comprehension
from scratch. This perspective radically changes our accounts of language processing in dialogue. One particularly
important implication is that interlocutors develop and use
routines (set expressions) during a particular interaction.
Most of this section addresses the implications of this perspective for language production, where they are perhaps
most profound. We then turn more briefly to language comprehension.
5.1. Speaking: Not necessarily from intention
to articulation
The seminal account of language production is Levelt’s
(1989) book Speaking, which has the informative subtitle
Pickering and Garrod: Toward a mechanistic psychology of dialogue
From intention to articulation. Chapter by chapter, Levelt
describes the stages involved in the process of language
production, starting with the conceptualization of the message, through the process of formulating the utterance as a
series of linguistic representations (representing grammatical functions, syntactic structure, phonology, metrical
structure, etc.), through to articulation. The core assumption is that the speaker necessarily goes through all of these
stages in a fixed order. The same assumption is common to
more specific models of word production (e.g., Levelt et al.
1999) and sentence production (e.g., Bock & Levelt 1994;
Garrett 1980). Experimental research is used to back up
this assumption. In most experiments concerned with understanding the mechanisms underlying language production, the speaker is required to construct the word or utterance from scratch, or from a pre-linguistic level at least.
For example, a common method is picture description (e.g.,
Bock 1986b; Schriefers et al. 1990). These experiments
therefore employ methods that reinforce the ideomotor
tradition of action research that underlies Levelt’s framework (see Hommel et al. 2001).
It appears to be generally agreed that this exhaustive
process is logically necessary because speakers have to articulate the words. Indeed, a common claim in work on language production is that, although comprehenders can
sometimes “short-circuit” the comprehension process by
taking into account the prior context (e.g., guessing thematic roles without actually parsing), producers always have
to go through each step from beginning to end. To quote
Bock and Huitema (1999):
there may be times when just knowing the words in their contexts is enough to understand the speaker, without a complete
syntactic analysis of the sentence. But in producing a sentence,
a speaker necessarily assigns syntactic functions to every element of the sentence; it is only by deciding which phrase will
be the subject, which the direct object, and so on that a grammatical utterance can be formed – there is no way around syntactic processing for the speaker. (p. 385)
In fact, this assumption is wrong: It is logically just as possible to avoid levels of representation in production as in
comprehension. Although we know that a complete output
normally occurs in production, we do not know what has
gone on at earlier stages. Thus, it is entirely possible, for example, that people do not always retrieve each lexical item
as a result of converting an internally generated message
into linguistic form (as assumed by Levelt et al. 1999, for
example), but rather that people draw upon representations that have been largely or entirely formed already.
Likewise, sentence production need not go through all the
representational stages assumed by Garrett (1980), Bock
and Levelt (1994), and others. For instance, if one speaker
simply repeated the previous speaker’s utterance, the representation might be taken “as a whole,” without lexical access, formulation of the message, or computation of syntactic relations.
Repetition of an utterance may seem unnatural or uncertainly related to normal processing, but in fact, as we
have noted, normal dialogue is highly repetitive (e.g., Tannen 1989). This is of course different from carefully crafted
monologue where – depending to some extent on the genre
– repetition is regarded as an indication of poor style (see
Amis 1997, pp. 246–50). In our example dialogue in Table
1, 82% of the 127 words are repetitions; in this paragraph
only 25% of the 125 words are repetitions. (Ironically, we –
the authors – have avoided repetition even when writing
about it.) In fact, the assumption that repetition is unusual
or special is a bias probably engendered by psychologists’
tendency to spend much of their time reading formal prose
and designing experiments using decontextualized “laboratory” paradigms like picture naming.
So it is possible that people can short-circuit parts of the
production process just as they may be able to short-circuit
comprehension. Moreover, this may be a normal process
that occurs when engaged in dialogue. We strongly suspect
(see sect. 5.2.2) that phrases (for instance) are not simply
inserted as a whole, but that the true picture is rather more
complicated. But it is critical to make the logical point that
the stages of production are not set in stone, as previous
theories have assumed.
5.2. The production of routines
A routine is an expression that is “fixed” to a relatively great
extent. First, the expression occurs at a much higher frequency than the frequency of its component words would
lead us to expect (e.g., Aijmer 1996). (In computational linguistics this corresponds to having what is called a high
“mutual information” content; Charniak 1993.) Second, it
has a particular analysis at each level of linguistic representation. Thus, it has a particular meaning, a particular syntactic analysis, a particular pragmatic use, and often particular phonological characteristics (e.g., a fixed intonation).
Extreme examples of routines include repetitive conversational patterns such as How do you do? and Thank you very
much. Routines are highly frequent in dialogue: Aijmer estimates that up to 70% of words in the London-Lund
speech corpus occur as part of recurrent word combinations (see Altenberg 1990). However, different expressions
can be routines to different degrees, so actual estimates of
their frequency are somewhat arbitrary. Some routines are
idioms, but not all (e.g., I love you is a routine with a literal
interpretation in the best relationships; see Nunberg et al.
1994; Wray & Perkins 2001).
Most discussion of routines focuses on phrases whose
status as a routine is pretty stable. Although long-term routines are important, we also claim that routines are set up
“on the fly” during dialogue. In other words, if an interlocutor uses an expression in a particular way, it may become a routine for the purposes of that conversation alone.
We call this process routinization. Here we consider why
routines emerge and why they are useful. The next section
considers how they are produced (in contrast to non-routines). This, we argue, leads to a need for a radical reformulation of accounts of sentence production. Finally, we
consider how the comprehension of routines causes us to
reformulate accounts of comprehension.
5.2.1. Why do routines occur? Most stretches of dialogue
are about restricted topics and therefore have quite a limited vocabulary. Hence, it is not surprising that routinization occurs in dialogue. But monologue can also be about
restricted topics, and yet all indications suggest it is much
less repetitive and routinization is much less common. The
more interesting explanation for routinization in dialogue is
that it is due to interactive alignment. A repeated expression (with the same analysis and interpretation) is of course
aligned at most linguistic levels. Thus, if interlocutors share
highly activated semantic representations (what they want
BEHAVIORAL AND BRAIN SCIENCES (2004) 27:2
13
Pickering and Garrod: Toward a mechanistic psychology of dialogue
to talk about), lexical representations (what lexical items are
activated), and syntactic representations (what constructions are highlighted), they are likely to use the same expressions, in the same way, to refer to the same things. The
contrast with most types of monologue occurs (in part, at
least) because the producer of a monologue has no one to
align his representations with (see sect. 2). The use of routines contributes enormously to the fluency of dialogue in
comparison to most monologue – interlocutors have a
smaller space of alternatives to consider and have ready access to particular words, grammatical constructions, and
concepts.
Consider the production of expressions that keep being
repeated in a dialogue, such as “the previous administration” in a political discussion. When first used, this expression is presumably constructed by accessing the meaning of
“previous” and combining it with the meaning of “administration.” The speaker may well have decided “I want to refer to the Conservative Government, but want to stress that
they are no longer in charge, etc., so I’ll use a circumlocution.” He will construct this expression by selecting the
words and the construction carefully. Likewise, the listener
will analyze the expression and consider alternative interpretations. Both interlocutors are therefore making important choices about alternative forms and interpretations.
But if the expression is repeatedly used, the interlocutors
do not have to consider alternatives to the same extent. For
example, they do not have to consider that the expression
might have other interpretations, or that “administration” is
ambiguous (e.g., it could refer to a type of work). Instead,
they treat the expression as a kind of name that refers to the
last Conservative government. Similar processes presumably occur when producing expressions that are already
frozen (Pinker & Birdsong 1979; see also Aijmer 1996).
Generally, the argument is that people can “short-circuit”
production in dialogue by removing or drastically reducing
the choices that otherwise occur during production (e.g.,
deciding which synonym to use, or whether to use an active
or a passive form).
Why might this happen? The obvious explanation is that
routines are in general easier to produce than non-routines.
Experimental work on this is lacking, but an elegant series
of field studies by Kuiper (1996) suggests that this explanation is correct. Kuiper investigated the language of sports
commentators and auctioneers, who are required to speak
extremely quickly and fluently. For example, radio horseracing commentators have to produce a time-locked and
accurate monologue in response to rapidly changing events.
This monologue is highly repetitive and stylized, but quite
remarkably fluent. He argued that the commentators achieve
this by storing routines, which can consist of entirely fixed
expressions (e.g., they are coming round the bend) or expressions with an empty slot that has to be filled (e.g., X is
in the lead), in long-term memory, and then accessing these
routines, as a whole, when needed. Processing load is
thereby greatly reduced in comparison to non-routine production. Of course, this reduction in load is only possible
because particular routines are stored; and these routines
are stored because the commentators repeatedly produce
the same small set of expressions in their career.
Below, we challenge Kuiper’s assumption that routines
are accessed “as a whole,” and argue instead that some linguistic processing is involved. But we propose a weaker version of his claims, namely that routines are accessed tele14
BEHAVIORAL AND BRAIN SCIENCES (2004) 27:2
graphically, in a way that is very different from standard assumptions about language production (as in, e.g., Levelt
1989). Moreover, we argue that not all routines are learned
over a long period, but that they can instead emerge “on the
fly,” as an effect of alignment during dialogue.
5.2.2. Massive priming in language production. Contrary
to Kuiper (1996), some compositional processes take place
in routines, as we know from the production of idiom
blends (e.g., that’s the way the cookie bounces; Cutting &
Bock 1997). However, there are good reasons to assume
that production of idioms and other routines may be highly
telegraphic. The normal process of constructing complex
expressions involves a large number of lexical, syntactic,
and semantic choices (why choose one word or form rather
than another, for instance). In contrast, when a routine is
used, most of these choices are not necessary. For example,
speakers do not consider the possibility of “passivizing” an
idiom that is normally active (e.g., The bucket was kicked),
so there is no stage of selection between active and passive.
Likewise, they do not consider replacing a word with a synonym (e.g., kick the pail), as the meaning would not be preserved. Similarly, a speech act like I name this ship X is
fixed, insofar as particular illocutionary force depends on
the exact form of words (cf. I give this ship the name X).
Also, flat intonation suggests that no choices are made
about stress placement (Kuiper 1996).
Let us expand this by extending some of the work of Potter and Lombardi to dialogue (Lombardi & Potter 1992;
Potter & Lombardi 1990; 1998). They address the question
of how people recall sentences (see also Bock 1986b; 1996).
Recall differs from dialogue in that (1) the same sentences
are perceived and produced; and (2) there is only one participant, acting as both comprehender and producer. Potter and Lombardi had experimental subjects read and then
recall sentences while performing concurrent tasks. They
found that a “lure” word sometimes intruded into the recalled sentence, indicating that subjects did not always
store the surface form of the sentence; that these lure words
caused the surface syntax of the sentence to change if they
intruded and did not fit with the sentence that was read; and
that other clauses could syntactically prime the target sentence so that it was sometimes misremembered as having
the form of the prime sentence. They argued that people
did not remember the surface form of the sentence but
rather remembered its meaning and had the lexical items
and syntactic constructions primed during encoding. Recall
therefore involved converting the meaning into the surface
form using the activation of lexical items and syntax to cause
a particular form to be regenerated. In normal sentence recall, this is likely to be the form of the original sentence.
This suggests that language production can be greatly enhanced by the prior activation of relevant linguistic representations (in this case, lexical and syntactic representations). In dialogue, speakers do not normally aim simply to
repeat their interlocutors’ utterances. However, production
will be greatly enhanced by the fact that previous utterances will activate their syntactic and lexical representations. Hence, they will tend to repeat syntactic and lexical
forms, and therefore to align with their interlocutors. These
arguments suggest why sentence recall might actually present a reasonable analogue to production in naturalistic dialogue; and why it is probably a better analogue than, for
example, isolated picture description. In both sentence re-
Pickering and Garrod: Toward a mechanistic psychology of dialogue
call and production in dialogue, very much less choice
needs to be made than in monologue. The decisions that
occur in language production (e.g., choice of word or structure) are to a considerable extent driven by the context and
do not need to be a burden for the speaker. Thus, they are
at least partly stimulus-driven rather than entirely internally
generated, in contrast to accounts like Levelt (1989).
However, our account differs from Potter and Lombardi’s in one respect. They assume no particular links between the activation of syntactic information, lexical information, and the message. In other words, the reason that
we tend to repeat accurately is that the appropriate message
is activated, the appropriate words are activated, and the
appropriate syntax is activated. But we have already argued
that alignment at one level leads to more alignment at other
levels (e.g., syntactic priming is enhanced by lexical overlap; Branigan et al. 2000). The alignment model assumes interrelations between all levels, so that a meaning, for instance, is activated at the same time as a word. This explains
why people not only repeat words but also repeat their
senses in a dialogue (Garrod & Anderson 1987). In other
words, what actually occurs in dialogue is lots of lexical, syntactic, and semantic activation of various tokens at each
level, and activation of particular links between the levels.
This leads to a great deal of alignment, and hence the production of routines. It also means that the production of a
word or utterance in dialogue is only distantly related to the
production of a word or utterance in isolation.
Kuiper (1996) assumes that most routines are stored after repeated use, in a way that is not directly related to dialogue. However, he considers an example of how an auctioneer creates a “temporary formula” by repeating a phrase
(p. 62). He regards this case as exceptional and does not employ it as part of his general argument. In contrast, we assume that the construction of temporary formulae is the
norm in dialogue. Many studies show how new descriptions
become established for the dialogue (e.g., Brennan & Clark
1996; Clark & Wilkes-Gibbs 1986; Garrod & Anderson
1987). In general, it is striking how quickly a novel expression can be regarded as entirely normal, whether it is a genuine neologism or a novel way of referring to an object
(Gerrig & Bortfeld 1999).
In situations in which a community of speakers regularly
discusses the same topic we might expect the transient routines that they establish to eventually become fixed within
that community. In fact, Garrod and Doherty (1994)
demonstrated that an experimentally established community of maze-game players quickly converged on a common
description scheme. They also found that the scheme established by the community of players was used more consistently than schemes adopted by isolated pairs of players
over the same period. This result points to the interesting
possibility that the interactive alignment process can be responsible for fixing routines in the language or dialect spoken by a community of speakers (see Clark 1998).
5.3. Producing words and sentences
Most models of word production assume that the apparent
fluency of production hides a number of stages that lead
from conceptual activation to articulation. In Levelt et al.
(1999) a lexical entry consists of sets of nodes at different
levels (or strata): a semantic representation, a syntactic (or
lemma) representation, a phonological representation, a
phonetic representation, and so on. Each level is connected
to the one after it, so that the activation of a semantic representation (e.g., for cat) leads to the activation of its syntactic representation (the “cat” lemma plus syntactic information specifying that it is a singular count noun), which in
turn leads to the access of the phonological representation
/k//æ//t/. Evidence for the sequential nature of activation
comes from time-course data (Schriefers et al. 1990; van
Turennout et al. 1998), “tip-of-the-tongue” data (Vigliocco
et al. 1997), and so on. Alternative accounts question the
specific levels assumed by Levelt et al. and the mechanisms
of activation, but do not question the assumption that earlier levels become activated before later ones (Caramazza
1997; Dell 1986). Notice that the data used to derive these
accounts is almost entirely based on paradigms that require
generation from scratch (e.g., picture naming) or from linguistic information with a very indirect relationship to the
actual act of production required (e.g., responding with the
object of a definition).
We do not contend that the dialogical perspective leads
us to a radically different view of word production. More
specifically, we have no reason to doubt that the same levels of representation are accessed in the same order during
production in dialogue (though this question has not been
addressed by mainstream psycholinguistic research). For
example, Potter and Lombardi’s (1990) data suggest that
even in repetition of a word, it is likely that lexical access occurs (and that there is no direct access of the word-form, for
example). However, contextual activation is likely to have
some effects on the time-course of production, particularly
in relation to the decisions at different stages in the production process. For example, a choice between two synonyms might normally involve some processing difficulty,
but if one has been established in the dialogue (e.g., by lexical entrainment), no meaningful process of selection is
needed.
The situation is very different with isolated sentence production. Models of production assume that a speaker initially constructs a message, then converts this message into
a syntactic representation, then into a phonological representation, and then into sound (Bock & Levelt 1994; Garrett 1980; Levelt 1989). Normally, they also assume that the
syntactic level involves at least two stages: a functional representation, and a constituent-structure representation. It
is accepted that cascading may happen, so that the complete message does not need to be computed before syntactic encoding can begin (e.g., Meyer 1996). But ordering
is assumed, so that, for instance, a word cannot be uttered
until it is assigned a functional role and a position within a
syntactic representation.
However, we propose that it may be possible to break this
rigid order of sentence production, and instead to build a
sentence “around” a particular phrase if that phrase has
been focused in the dialogue. In accord with this, context
can affect sentence formulation in monologue, so that a focused phrase is produced first (Bock 1986a; Prat-Sala &
Branigan 2000). Prat-Sala and Branigan, in particular,
found effects of focus on word order that were not due to
differences in grammatical role. Hence it may be possible
to utter a phrase before assigning it a grammatical role. For
example, in Pictures, I think you like, and Pictures, I think
please you, the meaning of Pictures does not vary but its
grammatical role (subject or object) does vary. Assuming
that production is at least partially incremental, people can
BEHAVIORAL AND BRAIN SCIENCES (2004) 27:2
15
Pickering and Garrod: Toward a mechanistic psychology of dialogue
therefore utter Pictures before deciding which role it
should be given. This would of course not be possible within
traditional models where phonological representations and
acoustic form cannot be constructed before grammatical
role is assigned (e.g., Bock & Levelt 1994). So the effects of
strong context, in either dialogue or monologue, may be to
change the process of sentence production quite radically.
5.4. Alignment in comprehension
The vast literature on lexical comprehension is almost entirely concerned with monologue (e.g., reading words in
sentential or discourse contexts) or isolated words. But the
alignment model suggests that lexical comprehension in dialogue is very different from monologue. A major consequence of alignment at a lexical level is that local context
becomes central. Listeners, just like speakers, should be
able to select words from a set that have been central to that
dialogue – a “dialogue lexicon.”
One of the most universally accepted phenomena in experimental psychology, which is enshrined in all classic
models (e.g., Morton 1969), is the word frequency effect:
More frequent words are understood and produced faster
than less frequent words. Of course, processing is affected
by repetition but this is normally regarded as only modulating the underlying frequency effect. However, in dialogue, local context is so central that the frequency of an expression (or, e.g., its age of acquisition) should become far
less important. To a large extent, frequency is replaced by
accessibility with respect to the dialogue context. In contrast, the analogous context in monologue does not lead to
alignment and there is a strong tendency to avoid repetition
in many genres (e.g., formal writing) so the value of local
context will be much less. Frequency is central to comprehension of monologue because it is what people fall back on
if they have no strong context. So a prediction of our account is that frequency effects will be dramatically reduced
in dialogue.
With respect to lexical ambiguity, we predict that context
will have a very strong role, so that effects of meaning frequency can be overridden. Most current theories of lexical
ambiguity resolution follow Swinney (1979) in assuming
that multiple meanings of an ambiguous word are accessed
in a bottom-up manner, largely irrespective of context. Similarly, differences in frequency do not affect access, unless
perhaps one meaning is highly infrequent (see Balota et al.
1999 and Moss & Gaskell 1999, for discussion). But in dialogue, only the contextually relevant meaning may be activated (or, in a modular account, the irrelevant meaning may
always be suppressed rapidly). Hence, an interlocutor will
straightforwardly adopt the appropriate meaning. An implication is that dialogue context should allow “subordinate
bias effect” to be overridden (Duffy et al. 1988). According
to Duffy et al., context can support the less frequent meaning and make it as accessible as the more frequent meaning, but it cannot cause the less frequent meaning to become more accessible than the more frequent meaning
(Binder & Rayner 1998; cf. Kellas & Vu 1999; Rayner et al.
1994). Although this may be true for reading (and monologue processing generally), it may not hold for dialogue.
The comprehension of routines is in a sense like lexical
comprehension, in that their “frequency” and interpretation is set by the dialogue. However, this effect is in fact so
strong that it appears to occur in monologue comprehen16
BEHAVIORAL AND BRAIN SCIENCES (2004) 27:2
sion as well. A great deal of work is concerned with the comprehension of novel compounds in isolation (e.g., Murphy
1988; Wisniewski 1996), and the interpretations assigned
depend on specific aspects of the words combined. Strong
discourse contexts appear to enable direct access to infrequent interpretations of compound nouns such as baseball
smile in reference to the smile of a boy given a baseball
(Gerrig & Bortfeld 1999). This would indicate that people
can also “short-circuit” the normal access to the individual
nouns in a compound when there is a restricted meaning
available from the immediate context.
6. Self-monitoring
The autonomous transmission model assumes that the
speaker constructs a message, formulates an utterance as a
series of linguistic representations and then articulates it as
sound; and the listener then hears the message, converts it
into linguistic representations and then comprehends it.
The interlocutors (ideally) end up with the same semantic
representation, and alignment at other levels is a derivative
process (if it ever occurs at all). In contrast, Figure 2 proposes that interlocutors align themselves at different levels
simultaneously via the automatic channels, and the parity
assumption insures that the same representations are used
in production and comprehension. Self-monitoring uses
the same mechanism of alignment, but within the speaker.
All models assume that speakers monitor their own output, so that, for instance, they are able to interrupt their
productions in order to change what they say (Hartsuiker &
Kolk 2001; Levelt 1983; 1989). This can occur either before
or after they start to produce a word. According to Levelt,
speakers monitor their own productions by using the comprehension system (cf. Postma 2000, for discussion of alternatives). They can monitor their actual outputs, in which
case comprehension proceeds in an essentially normal way.
According to a model that only contained this outer loop,
monitoring would fit straightforwardly into the autonomous
transmission model shown in Figure 3. The only difference
would be that both interlocutors are the same person. However, Levelt assumed the existence of an inner loop as well,
which acts upon the phonological representation according
to Wheeldon and Levelt (1995). Additionally, Levelt assumes that monitoring can occur within conceptualization,
to make so-called “appropriateness repairs,” for example. It
is impossible to include “inner” monitoring straightforwardly within the autonomous transmission model, because the monitor acts upon a representation that the interlocutor cannot act upon. From another perspective, it is
unclear how the inner loop or the loop within the conceptualizer should have developed, given that they bear no relationship to any process involved in comprehending one’s
interlocutor. The postulation of a monitor that uses the
comprehension system is parsimonious (and it is easy to see
how it could have evolved), but the postulation of special
routes from production to comprehension that serve no
other purpose, is not.
In contrast, the inner loop and the loop within the conceptualizer fit straightforwardly into the interactive alignment model. Interlocutors are affected by each other’s semantic and phonological representations via the channels
of alignment represented in Figure 2. Hence a speaker can
also be affected by his own representations at these levels.
Pickering and Garrod: Toward a mechanistic psychology of dialogue
Self-monitoring is therefore compatible with Figure 2, except that A and B now refer to the same person (regarded
as producer and comprehender). However, there is an important difference between interacting with oneself and interacting with an interlocutor. When interacting with an interlocutor, the information conveyed by the channels is
encoded as sound. But when interacting with oneself, there
is no need to encode the information as sound (indeed, the
existence of internal monitoring proves that this is not necessary).
Given the existence of such levels of representation,
there is no reason why the speaker should not automatically
monitor at these levels. We propose that the speaker performs monitoring at these different levels in a way that leads
to self-alignment. When the speaker produces an error at
(say) the syntactic level (e.g., by selecting the wrong
lemma), the result is a lack of alignment between the intended representation and the representation available to
the monitor. This will become apparent as the levels of representation are traversed. For example, if a speaker accesses the semantic and syntactic forms of “dog” in order to
utter it but wrongly accesses the phonological form of “cat,”
he will monitor this form, and then access its syntactic and
semantic representations. Because these do not match the
representation that he has accessed during production, the
speaker will realize his error and (normally) attempt to correct himself. If he detects the mismatch and begins to correct himself before articulation begins, the repair will be
covert; if not, some or all of “cat” will be produced. Self-correction involves a repair process that is essentially similar to
the straightforward repair process used during interaction
(see sect. 4.3). As the speaker’s production and comprehension systems draw upon the same implicit common
ground, this repair process will tend to be successful, and
hence there is normally no need to make reference to full
common ground in self-monitoring.5
The interactive alignment model makes the very interesting prediction that monitoring can occur at any level of
linguistic representation that can be aligned. For example,
we predict the existence of syntactic monitoring. Consider
the misassignment of syntactic gender and its subsequent
detection. Speakers clearly can begin to say Le tête and then
correct to La tête. This detection could occur externally or
via the phonological channel. But an important prediction
of this account is that monitoring (and the correction of errors) can also occur at the syntactic level (e.g., correcting
gender, count/mass errors, errors of auxiliary selection, or
errors of subcategorization), and at other levels as well. One
reason for suspecting that this might be correct is that
“other monitoring” (i.e., detecting errors in others’ speech)
appears faster for phonological than syntactic errors
(Oomen & Postma 2002). If self-monitoring of syntax occurred via the phonological loop, we would predict that it
would be slow in comparison to self-monitoring of phonological errors. But we know of no evidence for this claim.
More generally, the existence of monitoring appears to
be a consequence of dialogue. In dialogue, interlocutors
have to switch between speaking and listening rapidly and
repeatedly, and interlocutors have to be able to listen and
plan their next utterance at the same time (otherwise the
lack of pauses, for instance, could not be explained). The
obvious way in which this can occur is for interlocutors to
be listening at all times, with that listening involving aligning one’s representations with the input. If interlocutor A is
speaking, then B is listening to A and thus aligning with A.
But if A is speaking, then A listens to himself through monitoring and thus aligns with himself. In other words, monitoring is a by-product of a language processing system that
is sufficiently flexible to allow comprehension and production to occur to some extent simultaneously in dialogue.
This means that monitoring should tend to be hard during
periods of overlapping speech. Furthermore, monitoring is
a key part of the checking and interactive repair process discussed in section 4.3. As a speaker you have to monitor your
own contributions with respect to the implicit common
ground and as a listener you have to monitor your partner’s
contributions with respect to the same implicit common
ground.
7. Dialogue and linguistic representation
In the introduction, we noted that the main theoretical reason why mechanistic psycholinguistics has largely ignored
dialogue is that formal linguistics has largely failed to address dialogue. We cannot of course rectify this situation
here, but it is important to provide some sketch of how linguistic theory could support the study of dialogue, just as it
has so far provided support for the study of monologue.
Rather than attempt to address all relevant phenomena, we
restrict ourselves to the discussion of two important general
issues: the analysis of linked utterances and the architecture
of the language system.
7.1. Dealing with linked utterances
As noted in section 2, dialogue turns are not isolated utterances, but are linked across interlocutors. However, traditional linguistics is based on monologue, and therefore
treats the contribution of a single speaker as the unit of
analysis. Even when the contributions are linked fragments,
each contribution is treated on its own.6 However, this is
clearly wrong. As long ago as 1973, Morgan demonstrated
that there were syntactic restrictions on well-formed exchanges between interlocutors. For example, in A: What
does Tricia enjoy most? B: Being called “your highness”/*To
be called “your highness”, the grammatical form of the answer is constrained by the subcategorization requirements
of the verb in the question (Morgan 1973; see also Ross
1969). Likewise, if A utters Is Jack in town? and B replies
Jack?, B’s clarification request can only be analyzed with respect to A’s utterance (Ginzburg 2001). The syntactic form
of such elliptical requests is determined by the context (e.g.,
Who? is also a possible response because it is a noun phrase
like Jack). Hence, this demonstrates a syntactic parallelism
constraint between turns in dialogue.
The meaning of dialogue turns is also heavily constrained
by context. If produced in isolation, the meaning of Jack?
would be unclear; as a reply to Is Jack in town?, it means
either “are you asking if Jack is in town?” (the clausal reading) or “who is the person named Jack you were referring
to?” (the constituent reading). On both readings, some syntactic parallelism is required (e.g., he but not him can be
used to clarify Is he in town?). The constituent reading
employs phonological (or perhaps phonetic) parallelism,
as it actually requires “echoing” of the exact form used
(Ginzburg 2001). A satisfactory linguistic account of dialogue should provide an account of how the form and interpretation of such short answers is constrained by the linBEHAVIORAL AND BRAIN SCIENCES (2004) 27:2
17
Pickering and Garrod: Toward a mechanistic psychology of dialogue
guistic context. In part, this is because they are very common: According to Fernández and Ginzburg (2002), nonsentential utterances constitute more than 11% of dialogue
turns in their sample of the British National Corpus
(Burnard 2000), and clarification ellipses constitute nearly
9% of these. Ginzburg and Sag (2001) offer a linguistic account of such phenomena by incorporating context into linguistic representations.7 The interactive alignment model
predicts parallelism in general and hence it is not surprising that parallelism emerges as a linguistic constraint in
linked dialogue turns. Thus, Goldinger’s (1998) finding of
phonological echoing and the phonological restriction on
the constituent reading of clarification ellipsis may not be
coincidental. Note that an adequate theory of language production also needs to be able to account for the contextual
dependency of such utterances. It is not clear that current
theories can do this, because they are designed to account
for the production of isolated (and “complete”) sentences
(e.g., Bock & Levelt 1994; Garrett 1980).
The linguistic analysis of linked contributions as a single
unit means that the mechanisms used to produce and comprehend them can be narrowly linguistic, in the sense that
there is no need to appeal to “bridging” inference. Let us
consider this in relation to a particularly extreme example
of joint construction, when one interlocutor completes the
other’s fragment. For example, Clark and Wilkes-Gibbs
(1986) cite the following exchange: A: That tree has, uh,
uh, . . . B: Tentworms. A: Yeah. B: Yeah. Here, A appears
unable to utter the appropriate expression, and B helps out
by making a suggestion (which is then accepted). Of course,
B’s response is only felicitous because it is syntactically congruent with A’s fragment (has can take a noun-phrase complement such as Tentworms, but could not take a prepositional-phrase complement such as Of tentworms).
According to the orthodox (monological) view, B would
have to parse A’s utterance and assign it a semantic interpretation. Presumably, the parser can interpret an input
(That tree has, uh, uh,) that is ungrammatical and not even
a traditional constituent (though how this can be done is
rarely specified). Then B would have to access its syntax and
semantics (at least) but suppress production of these words.
Next B must “fill in” the missing noun phrase by accessing
and producing Tentworms. A will in turn have to interpret
B’s “degenerate” utterance, and then integrate these two
fragments via a bridging inference (though note that neither fragment has a propositional interpretation). This
should cause processing difficulty (Haviland & Clark 1974),
but does not appear to. If things are this complicated it is
unclear why interruptions should occur at all,8 why they can
occur so rapidly, or why producing language in such contexts is not manifestly harder, say, than monologue. It also
predicts that elliptical responses to questions should be
harder than non-elliptical ones. This is clearly incorrect
(e.g., Clark, 1979, showed that full responses are complex
and have special implicatures).
Contrast this with the claim of the interactive alignment
model, in which B, as listener, activates the same representations as A. These representations can be used in production in just the same way as in comprehension. Thus, we predict that it should be more-or-less as easy to complete
someone else’s sentence as one’s own, and this does appear
to be the case. Similarly, interlocutors should be able to complete each other’s words (e.g., if one speaker has difficulty)
by making use of shared phonological representations. One
18
BEHAVIORAL AND BRAIN SCIENCES (2004) 27:2
prediction is that speech errors could be induced through
perception as well as production (e.g., if B finishes off A’s
tongue twister, then B should be liable to produce errors).
The existence of non-sentential turns in dialogue suggests that any appropriate grammatical account needs to be
able to deal with such fragments, and allow their interpretations to be integrated into the dialogue context (as in, e.g.,
Poesio & Traum 1997). A reasonable assumption is that the
grammar should treat all well-formed dialogue turns as constituents, with a semantic interpretation, so that their
meaning can be combined with the meanings of other participants’ turns in a compositional manner. This would require a “flexible” notion of constituency, where many fragments that are traditionally not constituents are treated as
constituents (e.g., The tree has). One linguistic approach
that accords with this is Combinatorial Categorial Grammar (Steedman 2000; cf. Ades & Steedman 1982; Pickering
& Barry 1993). It allows most (but not all) fragments to be
constituents, and is therefore a plausible candidate for analyzing the syntax of dialogue (and can also deal with monologue). It also provides a natural account of routines, because these may be constituents within flexible categorial
grammar but not traditional linguistics (e.g., He’s overtaking; Kuiper 1996; for other linguistic treatments, see
Kempson et al. 2001; Phillips, 2003). Such linguistic proposals have already had some impact on psycholinguistic
accounts concerned primarily with monologue comprehension (e.g., Altmann & Steedman 1988; Pickering &
Barry 1991), in part because they provide a natural account
of incremental interpretation (e.g., Just & Carpenter 1980;
Marslen-Wilson 1973). Of course, any appropriate account
also has to treat some dialogue utterances as ill-formed, for
example, when a speaker simply stops mid-utterance (Levelt 1983). In general, we need a linguistic account of wellformed dialogue utterances, and this account cannot be derived straightforwardly from linguistic theories based on
monologue or citation speech.
7.2. The architecture of the language system
The interactive alignment model assumes independent but
linked representations for syntax, semantics, and phonology
(at least), where each level of representation plays a causal
role via alignment channels (see Fig. 2). This sits ill with a
Chomskyan “transformational” theory, with a central generative syntactic component and peripheral semantic and
phonological systems that are purely “interpretative.” In
Chomskyan approaches (whether Standard Theory, Government and Binding Theory, or Minimalism), syntax creates sentence structure, and sound and meaning are “read
off’ this structure (Chomsky 1965; 1981; 1995). Instead, the
interactive alignment model is compatible with constraintbased grammar approaches in which syntax, semantics, and
phonology form separate but equal parts of a multidimensional sign (Gazdar et al. 1985; Kaplan & Bresnan 1982;
Pollard & Sag 1994).
Within this tradition, Jackendoff’s (1997; 1999; 2002)
framework forms a particularly appropriate linguistic basis
for the interactive alignment model. He assumes that
phonological, syntactic, and semantic formation rules generate phonological, syntactic and semantic structures respectively, and are brought into correspondence by interface rules, which encode the relationship between different
systems.9 In our terms, the alignment channels can affect
Pickering and Garrod: Toward a mechanistic psychology of dialogue
the application of the formation rules, whereas the interface rules are encoded in the links between the levels.10
Jackendoff’s framework also provides a natural account of
idioms and other routines, because the lexicon includes
complex expressions (2002, Ch. 6).
In contrast, it is much more difficult to see why alignment should occur at phonological and semantic levels if no
generative component underlies these levels. Moreover,
the correspondence between the Chomskyan architectures
and models of production and comprehension has always
been difficult to sustain (e.g., Bock et al. 1992; Fodor et al.
1974; Pickering & Barry 1991). Thus, we see the integration
of a framework incorporating multiple generative components with a grammar that has a flexible approach to constituency as forming the linguistic basis for a psycholinguistic account of dialogue.
8. Distinguishing between dialogue and
monologue
In this target article we have argued that dialogue is the primary setting for language use and, hence, that dialogue processing represents the basic form of language processing.
Throughout, we have treated dialogue and monologue as
distinct kinds of language use. But is there a clear-cut distinction between dialogue and monologue or do they range
along a dialogic continuum?
8.1. Degree of coupling defines a dialogic continuum
Interactive activities vary according to the degree of coupling between the interacting agents. Whereas a tightly
coupled activity such as ballroom dancing requires continuous coordination between partners, a loosely coupled activity such as golf only requires intermittent coordination
(one may have to wait until one’s partner has struck the ball,
quality of play may be affected by how close the scores are,
etc.). Similarly, different styles of communication vary in
the degree of coupling between communicators. Whereas
holding a one-to-one intimate conversation may require
precise and continuous coordination (e.g., interruption,
joint construction of utterances, back-channeling), giving a
lecture only requires intermittent coordination (e.g., altering one’s style according to visual or vocal feedback from the
audience, or responding to an occasional question).
The interactive alignment model was primarily developed to account for tightly coupled processing of the sort
that occurs in face-to-face spontaneous dyadic conversation
between equals with short contributions. We propose that
in such conversation, interlocutors are most likely to respond to each other’s contributions in a way that is least affected by anything apart from the need to align. Hence, it
is not surprising that such language use in such situations is
often regarded as primitive or basic (Clark 1996; Linnell
1998). As the conversational setting deviates from this
“ideal,” the process of alignment becomes less automatic.
For example, video-mediated conversation, ritualized interactions, multi-party discussions, tutorials, and speeches
during debates each deviate in different ways from the
ideal. In such cases, interlocutors will be less able to rely on
automatic alignment and repair, and will need to spend
more time constructing models of their interlocutors’ mental states if they are to be successful.
For example, Doherty-Sneddon et al. (1997) found that
interlocutors in a collaborative problem-solving task were
more efficient when they could see and hear each other
than when they could only hear each other or when they interacted via a high-quality video link. Specifically, face-toface participants employed fewer words and checked their
interlocutors’ comprehension less often than participants in
the other conditions. Likewise, Fay et al. (2000) compared
discussions involving five- or ten-member groups. In the
small groups, the pattern of interruptions and turn-taking
were similar to those in dyadic dialogue. Most interestingly,
speakers tended to align with the immediately preceding
speaker (with respect to their opinions about what was most
important). But in the large groups, speakers did not align
with the preceding speaker, but rather with the dominant
speaker in the group. Hence, the interactive alignment
model predicted behavior in small groups but not large
groups, where speakers appeared to use “serial monologue.”
Whereas the prototypical form of dialogue involves
tightly coupled contributions by interlocutors, the prototypical form of monologue involves one communicator
making a single presentation without receiving any feedback. Good examples of this are speeches where there is no
possibility of audience reaction (e.g., when speaking on the
radio), and traditional written communication. In such
cases, the communicator has to formulate everything on his
own. He receives no help about what to produce, and cannot make use of an interlocutor’s contributions, because
nothing from the addressee comes in through the alignment channels. (The only information that comes through
the channels is via self-monitoring, and this is of much more
limited use.) Hence, true monologue is very difficult, with
successful communication often requiring very considerable planning (as in planning and rehearsing speeches) or
use of very routinized speech (as in Kuiper’s sportscasters
and auctioneers). However, much narrative is not as difficult as this, because the audience provides a considerable
amount of feedback via backchannel and non-linguistic
contributions (e.g., Bavelas et al. 2000). In cases where an
interchange moves between highly interactive interchanges
and long speeches by one interlocutor, we predict dynamic
shifts in the difficulty of production.
In the comprehension of monologue, the listener will
have to bring to bear appropriate inference skills. For example, he will often have to draw costly bridging inferences
to help understand what the writer or speaker had really
meant with a definite reference (Garrod & Sanford 1977;
Haviland & Clark 1974), though again the difficulty is reduced if the listener can give feedback (Schober & Clark
1989). But in “passive” comprehension, there is no opportunity to call on aligned linguistic representations and no
opportunity to resolve ambiguities using interactive alignment. Instead people have to fall back on the frequency of
words, syntactic forms, and meanings in making comprehension decisions, as no other useful information is available.
Therefore, language users need to develop a whole range
of elaborate strategies to become competent processors of
monologue. Of course much of education involves training
in writing essays and producing speeches, and the like, and
a smaller part involves comprehension of monologue (e.g.,
in being able to identify the important arguments in a text).
In contrast, people are very rarely taught how to hold conBEHAVIORAL AND BRAIN SCIENCES (2004) 27:2
19
Pickering and Garrod: Toward a mechanistic psychology of dialogue
versations (except in some clinical circumstances). Without
training in monologue, people are very likely to go off track
during comprehension and production. Even after these
strategies have been developed, people still find monologue far more difficult than dialogue.
9. Implications
The interactive alignment model is designed to account for
the processing of dialogue, but we have already suggested
that monologue can be regarded as an extreme case of noninteractive language use. This means that it can be harnessed into accounts of monologue processing as well. We
shall briefly suggest its relevance to a range of other issues
that extend beyond dialogue.
One interesting possibility is that it can serve as the basis
for predominantly automatic accounts of social interaction
more generally. There is considerable evidence that people
imitate each other in non-linguistic ways, and hence alignment is presumably not purely linguistic. For example,
Chartrand and Bargh (1999) demonstrated non-conscious
imitation of such bodily movements as foot rubbing. Such
findings, together with findings of the effects of the automatic activation of stereotypes on behavior, have led to the
postulation of an automatic perception-behavior link that
underlies such imitation (Bargh & Chartrand 1999; Bargh
et al. 1996; Dijksterhuis & Bargh 2001; Dijksterhuis & Van
Knippenberg 1998). According to these researchers, the
strength of this link means that the great majority of social
acts do not involve a decision component. Our contention
is somewhat related, in that we argue that the process of
alignment allows the reuse of representations that are constructed during comprehension, in a way that removes the
need to make complex decisions about how to represent the
mental state of the interlocutor. Of course, there are still
some conscious decisions about what one wants to talk
about, but the computational burden is greatly reduced by
making the process as automatic as possible. The social-psychological literature is fairly vague about precisely what is
imitated; in contrast, our account assumes that people align
on well-defined linguistic representations.
Indeed, the interactive-alignment account of dialogue
meshes well with recent proposals about the central role of
imitation within psychological and neuroscientific theorizing more generally (Heyes 2001; Hurley & Chater, in
press). The discovery of mirror neurons provides a reason
to expect certain forms of imitation to be straightforward,
and the finding that the same areas of the brain (Brodmann’s Areas 44 and 45) are involved in imitation as in language use (Iacoboni et al. 1999; Rizzolatti & Arbib 1998)
provides support for the assumption that alignment constitutes a fundamental aspect of language use. To make these
links more explicit, it would probably be necessary to perform the very difficult task of investigating brain activity
during dialogue.
An obvious application of our account is to language acquisition, because alignment underlies imitative processes
that occur as children acquire language. For instance,
Brooks and Tomasello (1999) showed that 2–3 year olds
could be trained to use passives by being presented with
other passives. A prediction of the interactive-alignment
model is that children will tend to repeat a construction that
is novel to them to a greater extent when they also repeat
20
BEHAVIORAL AND BRAIN SCIENCES (2004) 27:2
lexical items.11 From a rather different perspective, work
on atypical language development might provide evidence
for the circumstances under which the propensity for alignment might be disrupted. One would predict that this
would be most likely when social functioning was impaired,
and indeed there is evidence that imitation in general is impaired in autism (Williams et al. 2001). However, it is important to stress that alignment is unlikely to require a complete “theory of mind,” because it is not dependent on the
modeling of the interlocutor’s mental state. Indeed, findings such as Brooks and Tomasello’s speak against this account, on the grounds that such alignment occurs before
most children pass “false belief” tasks (e.g., Baron-Cohen
et al. 2000).
However, the model does not claim that assumptions
about the mental state of one’s interlocutor are irrelevant to
alignment. Presumably, one can decide whether one is interacting with an agent with which it is appropriate to align.
Thus, we can consider the interesting case of human-computer interaction, where people may or may not align with
computers’ utterances. If the conscious ascription of a mental state is necessary for alignment, then people will only
align if they perform such ascriptions. But if people behave
toward computers as “social agents,” whatever they consciously believe about their mental states, then we predict
unimpaired alignment will occur with computers, just as
many other aspects of social behavior do (Reeves & Nass
1996).
10. Summary and conclusion
This article has presented a mechanistic model of language
processing in dialogue, the interactive alignment model.
The model assumes that as dialogue proceeds, interlocutors
come to align their linguistic representations at many levels
ranging from the phonological to the syntactic and semantic. This interactive alignment process is automatic and only
depends on simple priming mechanisms that operate at the
different levels, together with an assumption of parity of
representation for production and comprehension. The
model assumes that alignment at one level promotes alignment at other levels including the level of the discourse
model and hence acts as a mechanism to promote mutual
understanding between interlocutors.
The interactive alignment model was contrasted with an
autonomous transmission account that represents the traditional psycholinguistic framework for language processing applied to dialogue. The main points of contrast between the two models are summarized in Table 2.
First, according to the interactive alignment account, the
interaction between interlocutors supports direct channels
between the linguistic representations that they use for language processing. In effect, the sounds come to directly encode words, meanings and even aspects of the situation
model. Alignment occurs at different levels of representation and alignment at one level leads to further alignment
at other levels. One of the mechanisms for this direct encoding is what we call routinization (see Table 2[3]): the setting up of semi-fixed complex expressions that directly
encode specific meanings. A second contrast with the autonomous processing account relates to the nature of the inference processes associated with establishing the common
ground in dialogue. Whereas inference in the traditional ac-
Pickering and Garrod: Toward a mechanistic psychology of dialogue
Table 2. Contrasts between autonomous transmission account of language processing in dialogue
and the interactive alignment account
Autonomous transmission account
Interactive alignment account
1. Linkage between interlocutors
Via sound alone – no direct links across other
levels of representation.
1. Linkage between interlocutors
Links across multiple levels of representation via “alignment
channels.” Sound comes to encode words, linguistic information,
and aspects of situational models.
2. Inference
Externalized in the interaction between interlocutors via a
basic interactive repair mechanism.
2. Inference
Internalized in the mind of speaker/listener: Speaker
in terms of audience design; Listener in terms of bridging
inference process.
3. Routines
Special case of language largely associated with idioms.
4. Self-monitoring
Inner loop monitoring requires a special internal route from
production to comprehension.
5. Repair mechanisms
Distinct repair mechanisms for self-repair and other-repair
in dialogue.
6. Linguistic representations
Only need to account for the structure of isolated and
complete sentences.
count is internalized in the minds of the speaker and listener, in the interactive alignment account it is externalized
through an interactive repair mechanism that makes use of
clarification requests. A third set of contrasts derives from
the nature of the monitoring process assumed in the interactive alignment account. Whereas in the traditional account internal self-monitoring leads to the stipulation of a
special mechanism in addition to the normal comprehension process, in the interactive alignment account it arises
directly from the parity assumption. Monitoring output can
occur at any level at which there is interactive alignment.
Furthermore, there is a direct and simple relationship between self-repair processes and other repair processes in
dialogue because the self-monitoring process is directly
comparable to the other-monitoring process (see Table 2
[4]). Finally, the interactive alignment account challenges
linguists to come up with a more flexible account of grammar capable of capturing linguistic constraints on linked
sentence fragments.
ACKNOWLEDGMENTS
Order of authorship is arbitrary. We wish to thank Ellen Bard,
Holly Branigan, Nick Chater, Herb Clark, Jonathan Ginzburg, Art
Glenberg, Rob Hartsuiker, Bernhard Hommel, Gerard Kempen,
Art Markman, Keith Rayner, Tony Sanford, Philippe Schyns,
Mark Steedman, Patrick Sturt, and two anonymous reviewers for
valuable comments, criticisms, and helpful suggestions on earlier
versions of this paper.
NOTES
1. In more detail, the procedure is as follows. Two players are
confronted with two computer-controlled mazes that do not differ in relevant ways. They are seated in different rooms but com-
3. Routines
Arise out of the application of the interactive alignment process.
A high proportion of dialogue uses routines, which simplify
both production and comprehension.
4. Self-monitoring
Monitoring occurs at any level of representation that is subject to
alignment as a consequence of the account.
5. Repair mechanisms
The same basic repair mechanism for self-repair and other-re.
pair
6. Linguistic representations
Needed to deal with linked utterances in dialogue, including
non-sentential “fragments.”
municate via an audio link. The players each have a token representing their current position in their maze, which is only visible
to them, and they take turns to move the tokens through the maze
one position at a time until both players have reached their respective goal positions. At any time approximately half of the paths
in each maze are closed. The closed paths are in different positions for each player and are only visible to that player. What
makes the game collaborative is that the mazes are linked in such
a way that when one player lands in a position where the other
player’s maze has a “switch” box, all of his closed paths open and
open paths close. This means that the players have to keep track
of each other’s positions to successfully negotiate their mazes. The
dialogue shown in Table 1 is taken from a conversation that occurred at the beginning of a game. Garrod and Anderson (1987)
analyzed transcripts from 25 pairs of players to see how location
descriptions developed over the course of each game. Some of the
results of this analysis are considered in more detail in section 2.2.
2. Actually, Carlson-Radvansky and Jiang only found inhibition
if the two trials used the same axis of the reference frame (e.g., the
up-down axis). This limitation may be related to the fact that priming was assessed outside a dialogue situation. An interesting prediction is that interlocutors would align on reference frames, not
just axes.
3. Critically, ordinals such as 4th can only quantify over ordered sets of items, whereas locative adjectives such as top or bottom usually modify unordered sets of items. Therefore when
speakers say 4th row, they either have to give a post modifying
phrase such as from the bottom, which imposes a particular ordering on the set of rows, or they have to assume that row denotes
an element in an implicitly ordered set of rows. In other words,
they assume that row in the bare 1st row is to be interpreted like
storey of a building in 1st storey. (Notice that it is odd to talk of
the 2nd storey from the bottom or even the bottom storey of a
building, but fine to talk about the bottom floor.)
4. A very interesting issue occurs when alignment at one level
BEHAVIORAL AND BRAIN SCIENCES (2004) 27:2
21
Pickering and Garrod: Toward a mechanistic psychology of dialogue
conflicts with alignment at another. Perhaps the most obvious
cases of this are when alignment at the situation model requires
nonalignment at the lexical level. For example, in Schober’s (1993)
example, two interlocutors who are facing each other use different terms to refer to similar locations (on the left vs. on the right)
to maintain the same egocentric frame of reference. Likewise,
Markman and Gentner (1993) show that successful use of analogy
can require lexical misalignment. In Garrod and Anderson’s
(1987) maze game, if one player uses second row to refer to the
second row from the top in a five-row maze, then the other player
will tend to use fourth row to refer to the second row from the bottom. The player could lexically align by using second row in this
way, but of course this would involve misalignment of situation
models, and would therefore be misleading. The implication is
that normally alignment at the situation level overrides alignment
at lower levels.
5. We assume that a case, for example, where the speaker could
not remember who he meant by John (while speaking) would be
pathological.
6. Most theories accept that a few dialogue phenomena do
need to be explained. For example, “binding” theory (Chomsky
1981) can be evoked to explain why himself is coreferential with
John in A: Who does John love? B: Himself; though see Ginzburg
(1999) for evidence against an account in such terms. Rather than
think of question-answer pairs as a marginal phenomenon that
needs special explanation in a monological account, we regard
them as a particularly orderly aspect of dialogue.
7. Roughly, Ginzburg and Sag assume feature structures taken
from Head-Driven Phrase Structure Grammar (Pollard & Sag
1994), in which context is incorporated into the representation of
the fragments using the critical notion of QUDs (“questions under discussion”).
8. Estimates from small group dialogues indicate that as many
as 31% of turns are interrupted by the listener (Fay et al. 2000).
9. Jackendoff uses the term conceptual structures instead of semantic structures, for reasons that we shall ignore for current purposes.
10. Note that Jackendoff (2002) assumes interface rules between semantic (conceptual) structures and phonological structures (p. 127, Fig. 5.5). If this is correct, it suggests that Fig. 2
should incorporate such a link as well. He also suggests that the
lexicon should be regarded as part of the interface components
(p. 131).
11. The tendency might even be stronger for young children
than adults, at least when it is the verb that is repeated. According
to the “verb island hypothesis,” syntactic information is more
strongly associated with individual verbs in young children than it
is in adults (e.g., children are often able to use a particular construction with some verbs but not others; Tomasello 2000).
22
BEHAVIORAL AND BRAIN SCIENCES (2004) 27:2
Open Peer Commentary
Commentary submitted by the qualified professional readership of this
journal will be considered for publication in a later issue as Continuing
Commentary on this article. Integrative overviews and syntheses are especially encouraged.
Is language processing different in dialogue?
Dale J. Barra and Boaz Keysarb
aDepartment
of Psychology, University of California, Riverside, CA 92521;
of Psychology, University of Chicago, Chicago, IL 60637.
[email protected]
[email protected]
bDepartment
Abstract: Pickering & Garrod (P&G) claim that the automatic mechanisms that underlie language processing in dialogue are absent in monologue. We disagree with this claim, and argue that dialogue simply provides a different context in which the same basic processes operate.
Pickering & Garrod (P&G) call for closer attention to the mechanisms underlying coordination in dialogue. There are good empirical grounds for accepting many of the basic assumptions of the
interactive alignment model. Specifically, the strong egocentrism
of speakers and listeners that we have uncovered in our own studies (cf. Barr & Keysar [in press], for a recent review) makes much
sense within a context of strong representational overlap, and the
interactive alignment model provides an appealing explanation for
how such overlap comes about.
A central assumption of P&G is a categorical distinction between language processing in monologue and in dialogue. Clearly,
monologue is different from dialogue because in monologue there
is no feedback and no opportunity for interactive repair. But P&G
go further than this, asserting that in monologue, “the automatic
mechanisms of alignment are not present” (sect. 3.2, para. 3) and
that “there is no opportunity to call on aligned linguistic representations” (sect. 8.1, para. 5). Under this view, dialogue involves
processes that are fundamentally distinct from those present in
monologue. We disagree. We suggest that only the strategic mechanisms of feedback and interactive repair are absent in monologue, not the automatic mechanisms of alignment. Just as there
is a “dialogic continuum” defining different kinds of interactive activities, we argue that there is a continuum of processing and that
alignment will be observed even under non-interactive circumstances.
Differences in processing may simply be a matter of degree, not
of kind, especially given that monologue-like episodes are common even in naturalistic conversation. Consider, for example, the
first utterance of a conversation, which in many cases is a minimonologue before a full-fledged interactive exchange develops.
The speaker will need to go through the various stages to produce
the first utterance (e.g., Levelt 1989), and the comprehender must
parse the utterance in order to appreciate its significance. Later
on in the exchange, enough of shared but nonmutual information
(what P&G unfortunately term “implicit common ground”)1 may
have built up to short-circuit many aspects of these processes. But
there is no reason to believe that different theories would be required to explain the processing of the first utterance of a dialogue
versus the hundred-and-first. Indeed, P&G claim that the alignment processes are automatic and resource-free, and processes of
these sorts cannot simply be switched off. Finally, the “dialogic
continuum” cited by the authors does not just provide a means for
classifying whole conversations but actually represents a domain
of activity that can be fully traversed even within the span of a single conversation. A conversationalist can at any point secure turn
space in order to engage in an extended monologue (Sacks et al.
1974) – for example, about her trip to India – followed by close
coordination with her interlocutor in order to arrange a time for a
Commentary/Pickering and Garrod: Toward a mechanistic psychology of dialogue
future meeting. There is no reason to expect alignment processes
and bridging inferences to toggle on and off as the interactivity of
the discourse changes.
Studies that find differences between the quality of comprehension of interlocutors and of overhearers (e.g., Schober & Clark
1989; Wilkes-Gibbs & Clark 1992) might be taken as evidence to
support the idea of radically different cognitive processes in interactive discourse. They need not be. In such studies interlocutors had opportunities for feedback and repair that the overhearers lacked. Because different people will misunderstand different
things, those who can ask for clarification will receive feedback
that is relevant to them and consequently might understand better. Therefore any difference between such noninteractive and interactive comprehension could be fully attributable to strategic,
effortful feedback but not necessarily to automatic alignment. In
fact, Barr and Keysar (2002) found that even when such feedback
is removed, listeners who believed themselves to be overhearers
automatically aligned their semantic representations with the
speaker’s to the same degree as listeners who believed themselves
to be addressees.
In closing, far from qualitatively changing the nature of processing, it is likely that dialogue provides a radically different context in which the same processes operate. The context includes an
interlocutor and mechanisms for feedback and interactive repair.
For us there is no question that it is important to study conversation in vivo, but it remains to be seen whether this would reveal
automatic processes that are truly unique to dialogue.
NOTE
1. In their seminal work on common ground, Clark and Marshall
(1981) clearly make the case that common ground is a form of metaknowledge that is conceptually distinct from shared knowledge. What
P&G are referring to by “implicit common ground” is really just shared
knowledge, not common ground, because interlocutors need not represent the fact that their representations are shared. Such usage is certain to
contribute to the legacy of confusion that has plagued discussions of mutual knowledge and common ground (see Keysar 1997 and Lee 2001 for
discussion).
Full alignment of some but not all
representations in dialogue
Holly P. Branigan
Department of Psychology, University of Edinburgh, Edinburgh EH8 9JZ,
United Kingdom. [email protected]
http://www.psy.ed.ac.uk/Staff/ra/hbranigan/publications.html
Abstract: I argue that alignment of linguistic representations and situation models in dialogue are qualitatively distinct. By virtue of the isomorphy between interlocutors’ linguistic representations, interlocutors align
their linguistic representations fully. However, evidence about situation
models is indirect and mediated through language, with the result that
alignment of situation models is only partial.
Pickering & Garrod (P&G) provide a plausible and very welcome
account of language processing in dialogue. Their account assigns
a central importance to the notion of alignment. Here I consider
the nature of alignment and, in particular, whether alignment of
strictly linguistic representations and alignment of situation models are qualitatively distinct.
P&G suggest that successful dialogue arises from the alignment
of representations between interlocutors, and particularly alignment of situation models. Alignment is defined as interlocutors
having the same representation at a particular level of structure.
It is uncontroversial that language is represented in the same way
in different speakers. Linguists and psycholinguists assume a common competence grammar in adult speakers (Chomsky 1965),
even though they may differ as to whether this grammar is innate
(e.g., Pinker 1989) or to some extent constructed through experi-
ence (e.g., Tomasello 2000). Crucially, adult speakers’ internalised
knowledge of the syntax, lexicon, and morpho-phonology of a language is held to be identical, such that there is an isomorphic mapping from any one speaker’s internalised representation of the language to any other’s. In a dialogue, then, interlocutors necessarily
make use of identical representations in producing their utterances.
Under P&G’s maximally parsimonious assumption of parity of
representations, interlocutors also necessarily draw upon identical representations in both producing and comprehending utterances. Note also that speakers’ utterances provide direct linguistic evidence to the listener. So when a listener hears an utterance,
he receives direct evidence (except in cases of mishearing or unresolved structural ambiguities) about the syntactic, lexical, and
morpho-phonological representations that the speaker has employed. If a speaker produces an utterance like I am in row two,
for example, the listener has direct evidence that she has used the
words I, am, and so on (and their relevant inflectional markings),
that she has used a pronoun and a verb and so on, and that she has
used a noun phrase, a verb phrase, a prepositional phrase, and so
on. Taken together, the combination of isomorphy of representations and direct evidence strongly supports P&G’s contention
(summarised in their Figure 2) that linguistic representations used
by interlocutors in dialogue act directly upon one another, and
that, in a very real sense, when we talk about interlocutors having
aligned linguistic representations, we mean that those representations are identical. In summary, P&G’s arguments for full alignment at linguistic levels of representation seem well founded.
But are situation models aligned in the same way as linguistic
representations? It is unclear that this is the case. In P&G’s model,
interlocutors’ situation models act directly upon one another (see
the authors’ Figure 2), in the same way as syntactic, lexical, and
morpho-phonological representations do, and alignment of situation models is taken as critical for successful communication. But
situation models differ qualitatively from strictly linguistic representations. A speaker’s utterances do not give direct evidence of
the situation model that the speaker holds, only indirect evidence
encoded in linguistic representations, from which the listener has
to infer the speaker’s situation model. So whereas an utterance like
I am in row two gives direct evidence about the speaker’s syntactic, lexical, and morpho-phonological representations, it gives only
indirect evidence about the speaker’s situation model. The listener
must construct a situation model based upon his or her interpretation of the speaker’s meaning – which may or may not be correct. Of course, as P&G note, misunderstandings may come to
light, and interlocutors may initiate repairs to bring about situation models that are aligned in the relevant aspects. But as they
also note, some misunderstandings may not be repaired. In fact,
it seems likely that interlocutors quite frequently have situation
models that are misaligned in major respects. Communication will
be (apparently) successful as long as the misalignment is not apparent to the interlocutors. To take P&G’s example of interlocutors using John to refer to different people, it is quite possible for
them to have a mutually satisfying dialogue concerning this person without ever realising that they are discussing different people; unless one of them says something that is inconsistent with
the other’s knowledge, they can successfully (for their purposes)
complete a dialogue with quite radically different situation models. Equally, a doctor and a patient may have a dialogue concerning the patient’s chronic back problem that appears to be successful, in that they are both satisfied that they understand each other
well; yet their situation models may differ considerably because of
unresolved (and unapparent) differences in their interpretation of
chronic. Situation models need only be aligned sufficiently for the
current communicative goal to be (apparently) met.
So it seems that alignment of situation models and alignment of
linguistic representations are quite different. With linguistic representations, interlocutors genuinely employ aligned (i.e., identical) representations that act directly upon one another; whereas,
because evidence for situation models is only indirect, interlocuBEHAVIORAL AND BRAIN SCIENCES (2004) 27:2
23
Commentary/Pickering and Garrod: Toward a mechanistic psychology of dialogue
tors rarely if ever have identical models. Instead, they have partially aligned models that may differ in many – sometimes important – respects. And because evidence for situation models is mediated through language, it seems highly unlikely that they can act
directly upon one another (contra P&G’s Figure 2).
One interesting result of the distinction between alignment of
linguistic and situation models is that alignment of linguistic representations may sometimes lead to misaligned situation models.
Garrod and Clark (1993) found that young children had a tendency to use the same words to describe a maze – that is, showed
lexical alignment – even when their situation models were quite
different. Similarly, in the case of the doctor-and-patient scenario,
one speaker’s use of the term chronic may well reinforce the
other’s use of the same term, leading to more misunderstanding
than if a different term were used. In both examples, full alignment at the linguistic level misleads interlocutors into believing
that they also have alignment at the level of situation models.
ACKNOWLEDGMENTS
This research was supported by a British Academy Postdoctoral Fellowship and by ESRC grant R000239363.
Two steps forward, one step back:
Partner-specific effects in a psychology
of dialogue
Susan E. Brennan and Charles A. Metzing
Department of Psychology, State University of New York, Stony Brook, NY
11794-2500. [email protected]
[email protected]
http://www.psychology.sunysb.edu/psychology/personnel/Susan.htm
Abstract: Pickering & Garrod’s (P&G’s) call to study language processing
in dialogue context is an appealing one. Their interactive alignment model
is ambitious, aiming to explain the converging behavior of dialogue partners via both intra- and interpersonal priming. However, they ignore the
flexible, partner-specific processing demonstrated by some recent dialogue studies. We discuss implications of these data.
In human language processing, the whole is greater than the sum
of the parts; therefore, those who study the language processing system in dialogue contexts are poised to make different sorts of discoveries than those who study the parts working alone. Pickering &
Garrod (P&G) present a convincing argument that psycholinguists
should pay attention to dialogue. In fields such as artificial intelligence and human-computer interaction, where the goal is often to
build a fully working dialogue partner, many will find this a worthy
enterprise as well. After presenting evidence for phonological, lexical, and syntactic convergence between dialogue partners and for
representations shared between comprehension and production,
P&G make a strong claim that is far less convincing: “normal conversation does not routinely require modeling the interlocutor’s
mind” (sect. 4.4, para. 4). They support this position with evidence
from studies that fail to meet the very standards they seek to advance, while ignoring evidence that complicates matters for their interactive alignment model. Thus, their position on the importance
of studying language in dialogue does not go far enough.
This position assumes that interlocutors achieve aligned mental representations without having to track anything specific about
each other’s knowledge because both have evolved with the same
cognitive architecture; what is easiest for speakers is easiest for addressees (Brown & Dell 1987). It further assumes that there is no
need to track common ground, as interlocutors each use their own
memory of the conversation as a proxy. By this argument, what appears to be partner-specific or “audience design” is actually inflexible and unavoidable, at least in the earliest moments of processing. P&G propose a two-stage model (similar to that of Horton
& Keysar 1996), arguing that interlocutors “do not routinely take
common ground into account during initial processing . . . full
24
BEHAVIORAL AND BRAIN SCIENCES (2004) 27:2
common ground is only used when simpler mechanisms are ineffective” (sect. 4.1). This (circular) view relegates any aspect of production or interpretation that displays flexibility or sensitivity to an
interlocutor’s needs (as distinct from one’s own) to the status of a
relatively late adjustment, managed as a kind of repair or pragmatic garden path.
Granted, it is difficult to design a good experiment on audience
design. A good experiment must distinguish one interlocutor’s
perspective from another’s, avoid confounding individual perspectives with common ground (Keysar 1997), and allow interlocutors to interact naturally or contingently (Schober & Brennan
2003). But we are surprised that studies succeeding in all this (and
finding partner-specific effects early in processing, e.g., Hanna et
al. 2003; Nadig & Sedivy 2002) are dismissed by P&G: “their task
was repetitive and involved a small number of items, and listeners
were given explicit information about the discrepancies in knowledge” (target article, sect. 4.2, para. 3). Then follows a very broad
claim: “Under such circumstances, it is not surprising that listeners develop strategies that may invoke full common ground. During natural dialogue, we predict that such strategies will not normally be used.”
Paradoxically, evidence to support this position comes mainly
from studies that did not allow any potential for interaction. These
include Brown and Dell (1987), Ferreira and Dell (2000), Horton
and Keysar (1996), and others in which partners did not interact
naturally or provide contingent feedback. Sometimes this matters;
for example, Brown and Dell (1987) concluded that speakers did
not take addressees’ specific needs into account when retelling
stories; but their addressees had no needs (they were confederates who knew the stories better than the speakers did). When we
ran a similar study using spontaneously interacting speakers and
addressees (Lockridge & Brennan 2002), speakers’ early syntactic
choices indeed showed sensitivity to addressees’ needs.
There is additional good evidence of rapid, partner-specific effects from the comprehension side. Hanna and Tanenhaus (2004)
asked addressees to follow a (confederate) speaker’s directions in
a cooking task (e.g., Hand me the cake mix); the addressees’ eye
fixations showed that they restricted candidate referents for ambiguous expressions (e.g., when two cake mixes were present) depending on what the speaker was holding and what she could not
reach; they did this from the earliest moments of processing.
And we have demonstrated that addressees interpret the same
utterance differently when it is spoken by different speakers with
whom the addressees have different dialogue histories (Metzing
& Brennan 2001; 2003). In our experiment, addressees were instructed by (confederate) speakers to reposition objects among a
relatively large set; they did this several times, evolving shared
perspectives and terms for critical objects (e.g., the shiny cylinder). Then the speaker left the room and either returned or else a
new confederate speaker entered. In the final trial, the new or old
speaker used either the familiar term or a new, equally good term
(e.g., the silver pipe) for the same critical object (amid many other
references that did not use different terms). Addressees gazed immediately at the object when either speaker used the old term.
However, when the old speaker used a new term (inexplicably
breaking a conceptual pact), addressees experienced interference,
delaying gazing at the target object. There was no such delay when
the new speaker used the new term (in fact, resolving this was just
as fast as the old term spoken by the new speaker). This partnerspecific interference suggests that the pragmatic force of breaking a conceptual pact has impact immediately, rather than just as
a late adjustment or repair.
Such immediate effects provide evidence of impressive agility
and potential for partner-specific processing in the language processing system, which the interactive alignment proposal fails to
address. Pragmatic and partner-specific knowledge is implemented by basic mechanisms of memory and does not rely on special processes or exhaustive partner models. Audience design –
truly partner-specific processing – can occur immediately and effortlessly as well as more slowly and deliberately, depending on
Commentary/Pickering and Garrod: Toward a mechanistic psychology of dialogue
how activated relevant information is. P&G’s strong separation between “implicit common ground” (automatic but excluding any
partner specificity) and “full common ground” (requiring reprocessing) is unconvincing.
Yes, the potential for interaction matters! But the interactive
alignment model has farther to go before it can accommodate the
flexible and adaptive processing that these data support.
ACKNOWLEDGMENTS
This material is based upon work supported by the National Science Foundation under grants No. 0082602 and No. 9980013. Any opinions, findings, and conclusions or recommendations expressed in this material are
those of the authors and do not necessarily reflect the views of the National
Science Foundation.
Priming and alignment: Mechanism
or consequence?
Sarah Brown-Schmidt and Michael K. Tanenhaus
Department of Brain and Cognitive Sciences, University of Rochester, River
Campus, Rochester, NY 14627. [email protected]
[email protected]
http://www.bcs.rochester.edu/people/sschmidt/
http://www.bcs.rochester.edu/people/mtan/mtan.html
Abstract: We agree with Pickering & Garrod’s (P&G’s) proposal that dialogue is an important empirical and theoretical test bed for models of language processing. However, we offer two cautionary notes. First, the enterprise will require explicit computational models. Second, such models
will need to incorporate both joint and separate speaker and hearer commitments in ways that go beyond priming and alignment.
We applaud and second Pickering & Garrod’s (P&G’s) call to psycholinguists to include dialogue as an empirical and theoretical
test bed for models of language processing. There is much to be
gained by combining the tools for studying real-time processing
developed within the language-as-product tradition with the more
natural interactive situations typically used within the action tradition (Tanenhaus et al. 2004). And we believe that many of the
basic processes in language comprehension, including spokenword recognition, syntactic processing, and reference resolution,
can be studied with the same precision in dialogue settings as in
more traditional controlled experiments (e.g., Brown-Schmidt et
al. 2002; in press). However, we would place a somewhat different emphasis on why it is important to study dialogue.
First, unlike P&G, who suggest that theories of language processing within the product tradition are admirably well-specified,
we think that there has been a dearth of explicit mechanistic models of language processing. This is especially true within language
comprehension, where, with the exception of some limited models developed within the neural network tradition, most computational models make only tenuous contact with behavioral data, and
vice versa (Christiansen & Chater 2001). However, we believe that
the emergence of interactive dialogue systems within computational linguistics offers an opportunity to develop explicit computational models in domains that can also be used to study human
language processing, thus creating a synergistic feedback loop between modeling and experimentation. Although the field has not
yet reached this stage, the opportunity is on the horizon, as computational linguists strive to implement systems that can engage
in continuous generation and understanding.
The motivation for continuous understanding and generation in
dialogue systems is instructive (Allen et al. 2001). During utterance generation, a speaker needs to have the capacity to monitor
feedback from an addressee, both verbal and nonverbal, and plan
or adjust the continuation of her utterance accordingly. Consider,
for example, an utterance which begins, “Now, take thee,1 uh,
Phillips head screwdriver. . . .” If the addressee nods or says “uh
huh,” the speaker can continue with “and tighten the bolt.” How-
ever, if the addressee says “huh,” looks perplexed, or begins to
reach for the wrong tool, the speaker is likely to continue with,
“the one with the blue handle, the one closest to your wrench” or
say “No, that one” (while pointing). This example illustrates one
of the potential benefits of studying interactive dialogue: It can
shed different light on some basic assumptions. For example, most
psycholinguists use the fact that comprehension occurs more or
less continuously as a reason for why they focus on response measures that are closely time-locked to the input. Yet, if pressed to
answer the question of why comprehension is so incremental, they
would likely appeal to the fact that memory capacity is limited.
But, working memory constraints do not explain why processing
is as relentlessly continuous as it is. In fact, the early theories of
language comprehension, which were largely driven by workingmemory assumptions, assumed that comprehension was a catchup game, with many delays so that listeners could avoid making
premature commitments. However, incremental processing is
necessary for efficient use of feedback from an interlocutor.
In confronting the challenges involved in integrating comprehension and production, as well as speech recognition, intention
recognition, and utterance planning, psycholinguists will be required to go beyond the boxes and arrows and consider more detailed models of what needs to be computed. Here, the study of dialogue provides a useful paradigm by allowing psycholinguists and
computational linguists to study and model these processes in relatively constrained domains where it is possible to be explicit about
the relevant components of the system and how they interact.
We believe that when P&G consider the problem of dialogue
from this perspective, they will need to rethink two assumptions
that guide their approach. The first is that interlocutors do not
need to take into account differences between speaker and hearer
knowledge and perspectives, except for repair strategies. We
agree that many aspects of language use can be egocentric, though
we think that P&G overstate the case, mistaking evidence that
common ground does not fully constrain referential domains, with
evidence for the stronger claim that it is ignored in initial processing (cf. Hanna et al. 2003). Moreover, natural language is rife
with constructions that depend crucially on differences between
assumed speaker and hearer knowledge. To take a simple example, the declarative question “You used the Phillips tool?” can be
uttered only when the speaker believes the addressee is committed to the presupposed assertion that is being questioned (Gunlogson, 2003). If the addressee had in fact used the tool, she might
respond with an explanation; if she had not, she would need to
contradict the presupposed commitment by saying something like
“No, of course not, I. . . .”
These kinds of phenomena will emerge as psycholinguists consider richer dialogue situations in which the participants have reasons to make choices between alternative forms of utterances, and
more complex interactions where it is important for interlocutors
to track each other’s attention and intentions. Finally, note that
both intention and attention, which are crucial components of
monitoring an interlocutor, do not necessarily appeal to high-level,
resource-demanding processes. Eye gaze, for example, is a powerful source of information about attention. Moreover, the aspect
of an interlocutor’s knowledge that has to be monitored can often
be circumscribed by goal structures. Cast in this light, perspective
monitoring can be seen as one of the basic components of communication, rather than a special purpose, resource-intensive, revision process. In many of the examples that P&G focus on, the
goal structures are completely defined by the task and the choice
of utterances is limited. Under these conditions, it is easy to view
alignment and priming as the primary mechanisms of dialogue,
rather than as interesting phenomena that can be used to provide
insight into the representations and processes underlying interactive conversation.
ACKNOWLEDGMENT
This work was supported by a National Institutes of Health grant, No.
HD-27206, to Michael K. Tanenhaus.
BEHAVIORAL AND BRAIN SCIENCES (2004) 27:2
25
Commentary/Pickering and Garrod: Toward a mechanistic psychology of dialogue
NOTE
1. We use the word “thee” to indicate a disfluent pronunciation of the
word “the” (and not in the old English usage as a pronoun).
A call for more dialogue and more details
J. Cooper Cutting
Department of Psychology, Illinois State University, Normal, IL 61790.
[email protected]
http://main.psy.ilstu.edu/faculty/cutting/index.html
Abstract: Pickering & Garrod (P&G) argue that contemporary models of
language use are inadequate. This has resulted largely because of an experimental focus on monologue rather than dialogue. I agree with the
need for increased experimentation that focuses on the interplay between
production and comprehension. However, I have some concerns about the
Interactive Alignment model that the authors propose.
Pickering & Garrod (P&G) make an excellent argument for the
need to increase the study of the interplay between production
and comprehension in mainstream psycholinguistic research. Historically, most research on language has treated the processes of
comprehension and production as largely independent systems.
This separation may reflect the different starting points of the two
processes. Input for comprehension is easy to manipulate and
control experimentally. Input for production is thought, something that is much harder to control experimentally (Bock 1996).
As a result, language comprehension research has a long history
relative to that of language production research. Furthermore,
theories of comprehension are rarely informed by research on
production, and vice versa; however, the two processes are clearly
intertwined. A complete model of language use needs to consider
how production and comprehension processes work in conjunction. The Interactive Alignment model is explicitly designed to do
just that. Rather than focus on how we process language as a
monologue, as most mainstream psycholinguistic research does,
the model focuses on the processes involved in dialogue between
two (or more) language users.
Although I support the idea that much more research focusing
on the interplay of production and comprehension is needed, I believe that the Interactive Alignment model that the authors propose is underspecified in several important respects.
1. What is alignment? The central feature of the model is the
alignment of multiple levels of linguistic representations between
interlocutors. This is achieved via “channels of alignment” through
which “the activation of a representation in one interlocutor leads
to the activation of the matching representation in the other interlocutor directly” (sec. 3.2). In P&G’s Figure 2, these channels
are depicted as bidirectional links between representations within
both dialogue participants. Although it is easy to imagine direct
links between representations within a person (see the parity assumption below), it is much more difficult to determine what
these “direct” links between different individuals correspond to.
The authors propose that the mechanism of alignment is a “primitive and resource-free priming mechanism” both within and between levels of representation. Although this may be different
from strict serial models of language use, it seems to be a feature
of existing interactive models of language use (for recent reviews,
see Dell et al. 1999; Pickering et al. 2000). Within these models,
the impact of recent use (of representations) in dialogue could be
modeled with residual activation of representations from earlier
parts of the conversation. However, this priming mechanism does
not seem to be very “direct” but is instead indirect.
2. Is all priming alike? The authors suggest that alignment of
multiple levels of representation (including phonological, syntactic, lexical, semantic, and situational) is achieved by priming. However, “priming” phenomena in language may not all share the same
underlying mechanisms. For example, Bock and Griffin (2000)
have suggested that syntactic persistence is a reflection of learn-
26
BEHAVIORAL AND BRAIN SCIENCES (2004) 27:2
ing rather than spreading of activation. It is unclear how the interactive alignment model distinguishes between activation-based
and learning-based priming mechanisms.
3. Parity of representations. The authors propose that there is
parity between the representation used by production and that
used by comprehension. In other words, although production and
comprehension processes may be different, they share the same
set of linguistic representations. However, the issue of shared versus distinct representations needs to be considered for each level
of representation (e.g., Balota 1990; Caramazza 1991; Levelt
1989). For example, Cutting (1998) used a word-picture priming
technique in which two words were presented in the prime trial.
The participant produced one word while the other was ignored.
Cutting found that both produced and ignored semantically related prime words interfered with picture naming. However, only
produced phonologically related primes influenced picture naming; ignored phonologically related primes had no effect. Cutting
interpreted this pattern of results as support for a model in which
semantic representations are shared by production and comprehension whereas phonological representations are separate. The
authors refer to Hommel et al.’s (2001) Theory of Event Coding
(TEC) as a model for language representations. Interestingly,
Hommel et al.’s proposal also states that TEC is most appropriate
for abstract distal coding. In other words, because phonological
representations are the linguistic representations “nearest” to the
sensory code and muscular innervation patterns, it should not be
surprising that they are different for production and comprehension (cf. Martin et al. 1999; Zwitserlood 1994).
4. Routinized language. The authors propose that interlocutors use routines which are developed “on the fly” during dialogue
(similar to idioms like “kick the bucket”). These routines result
from the processes of alignment in dialogue. As speakers and listeners use particular lexical, semantic, and syntactic representations, some of these representations become bound into routines.
The use of these routines may allow a speaker to “bypass” some of
the early stages of processing assumed by traditional models of
language production (e.g., Bock & Levelt 1994; Levelt 1989; Levelt et al. 1999). However, the process of routinization in the interactive alignment account is underdeveloped. It is unclear how
these representations are bound (concurrent activation, learned
connections, etc.), how they are represented (in either the shortterm or the long-term; Cutting & Bock 1997), or how they actually “bypass” stages of language production (i.e., do routines bypass stages or just grease the wheels a bit?). Without further
specification of these issues it is difficult to evaluate the model’s
predictions.
In conclusion, the interactive alignment model as currently proposed is underspecified with respect to how processes of alignment, priming, and routines work. As currently stated, it is difficult to determine specific testable predictions that would
distinguish this model from currently available models of production and comprehension. However, the point of taking a dialogue
perspective to these models (and the creation of new ones) is an
excellent one. Although there is no shortage of monologue use of
language (television, radio, print, lecture, etc.), clearly dialogue
should be a central aspect of research. It is time to tear down some
of the historically designed barriers between models of language
comprehension and production and examine what impact a dialogue perspective has on these models.
Commentary/Pickering and Garrod: Toward a mechanistic psychology of dialogue
Situation alignment and routinization in
language acquisition
Peter F. Dominey
Institut des Sciences Cognitives, CNRS UMR 5015, 69675 BRON Cedex,
France. [email protected]
http://www.isc.cnrs.fr
Abstract: Pickering & Garrod (P&G) describe a mechanism by which the
situation models of dialog participants become progressively aligned via
priming at different levels. This commentary attempts to characterize how
alignment and routinization can be extended into the language acquisition
domain by establishing links between alignment and joint attention, and
between routinization and grammatical construction learning.
Pickering & Garrod (P&G) describe a mechanism by which the
situation models of dialog participants become progressively
aligned via priming at different levels, including lexical, syntactic,
semantic, and situational representations. An essential interest
and novelty of this approach is that, instead of requiring a complex
and effortful mechanism for explicitly constructing a common
ground, it offers a rather straightforward mechanism that operates
largely automatically via priming.
It is of potential interest that this type of alignment can be seen
to be useful in other communicative contexts besides dialogue.
Two such contexts can be considered, both of which extend the
situation alignment mechanism into the domain of language acquisition. The first concerns the alignment of situation models in
which one of the interlocutors is in a prelingual, acquisition phase.
This emphasizes the suggestion that alignment can take place via
nonverbal influences. Second, in the current formulation, the
process of alignment and the formation of routines takes place on
the time scale of single dialogues; however, these mechanisms can
also be considered to span time frames that greatly exceed a single dialogue, particularly in the case of familiar repeated situations
(feeding, bathing, playing), yielding “virtual dialogues” that can
span a time period of several months. In such a situation, we can
consider the formation of routines in the context of language acquisition to be analogous to the development of grammatical constructions.
Language acquisition can be functionally defined as the process
of establishing the relation between sentences/discourses and
their meanings. A significant part of this problem concerns the issue that before these relations can be established, the speaker and
listener should be aligned with respect to the target meaning. If
the meaning for the target utterance is not established both for the
speaker and the listener, then construction of the mapping from
utterance to meaning is indeterminate. This suggests the required
existence of extra- or prelinguistic alignment mechanisms. Interestingly, there is indeed a significant body of research indicating
that by 6 months of age, human infants achieve prelinguistic situation alignment by exploiting joint attention cues (e.g., gaze direction, postural orientation) in order to identify intended referents (e.g., Morales et al. 2000; Tomasello 2003). This indicates that
P&G’s Figure 2 could be modified to include nonlinguistic inputs
at the semantic and situation model levels. Such a modification
will allow both the “alignment bootstrapping” in which initial situation model alignment will play a crucial role in language acquisition as well as the influence of extralinguistic inputs in adult
alignment contexts.
In a related extension of the alignment model into the acquisition domain, we can consider the relation between the development of production and comprehension routines in the time
frame of a single dialogue and the development of grammatical
constructions in the time frame of the first years of language acquisition. As specified by P&G, the creation of routines requires
a coherent context in which the routines are applicable, and so,
stretching this time frame to the scale of months and years is a
non-negligible issue. Interestingly, Tomasello (2003) notes that
repetitive events such as feeding, bathing, playing, and so on are
relatively similar from episode to episode, and thus provide ap-
propriate contexts that coherently span significant time periods.
Given a temporally extended “virtual dialogue” domain, we can
consider the development of routines as facilitatory not only
within the context of a single dialogue but also in the more fundamental role of the development of communicative conventions
that span significant time periods, thus forming the basis for language acquisition. In this context, routines take on the alternative
identity of grammatical constructions (see Goldberg 1995), with
all of their processing advantages. In particular, as described by
P&G, the use of routines significantly eliminates the need for syntactic derivation of the appropriate grammatical structural forms,
both for production and comprehension. When this approach is
applied at the acquisition time scale, it is remarkably similar to the
usage-based developmental approach to language acquisition advocated by Tomasello (2003).
In this framework, relatively fixed grammatical forms are linked
to their corresponding meanings in the context of repetitive events
(e.g., feeding, playing, etc.). These constructions/routines are
then progressively opened to allow generalization within a given
construction (e.g., variable replacement) to form new instances,
and subsequent generalization to new constructions. Again, in
both P&G’s dialogue context and Tomasello’s development context, highly functional communicative form-meaning constructions/routines are developed without reliance on a heavy initial investment in generative syntactic capabilities.
I have recently performed a series of simulation (Dominey
2000) and robotic (Dominey 2003a; 2003b) experiments to determine the feasibility of this type of approach to language acquisition in a restricted context. The underlying assumptions in the
model are (1) that grammatical constructions correspond to the
learned mapping between a given sentence type and its corresponding meaning frame (see Goldberg 1995), and (2) that grammatical constructions are uniquely identified by a limited set of
cues that include word order and grammatical morphology including free and bound morphemes (Bates & MacWhinney 1987).
The model is provided with sentence, meaning pairs as input
and should learn the Word-to-Referent and Sentence-to-Meaning
mappings. For the current discussion, we assume that a limited set
of concrete, open-class elements have been learned and will consider how this knowledge allows the learning of simple grammatical constructions. When a sentence, meaning pair is presented,
the configuration of closed-class (function) elements is extracted
and used as an index to “look up” the corresponding construction
(routine) in the construction inventory. The construction corresponds to the learned mapping of open-class element positions in
the sentence onto their thematic and event roles in the meaning
representation. If there is no entry in the construction inventory
(i.e., the current sentence type has never been previously encountered), then the construction is built on the fly by matching
the referents for the open-class words with their respective roles
in the meaning representation. The construction is then stored for
future use. The developmental aspects of this learning are presented in more detail in Dominey (2000).
Thus, similar to P&G’s routines, constructions are built by pairing the grammatical form with the aligned meaning (situation)
representation. The interesting suggestion is that, at least to a certain degree, P&G’s proposed situation alignment and routine construction capabilities provide a mechanism for language acquisition (at least the learning of fixed grammatical constructions that
can generalize to new instances of the same constructions) which
avoids the enlistment of generative grammar mechanisms. If a situation alignment priming mechanism could be demonstrated to
perform in both the dialogue and acquisition time scales, this
would be evidence for an ingenious economy of functional mechanisms for language processing in the context of dialogue.
BEHAVIORAL AND BRAIN SCIENCES (2004) 27:2
27
Commentary/Pickering and Garrod: Toward a mechanistic psychology of dialogue
Production-comprehension asymmetries
Fernanda Ferreira
Department of Psychology, Michigan State University, East Lansing, MI
48824-1117. [email protected]
http://eyelab.msu.edu/people/ferreira/fernanda.html
Abstract: Pickering & Garrod’s (P&G’s) mechanistic theory of dialogue is
a major advance for psycholinguistics. But the commitment to representational parity in production and comprehension is problematic. Recent
research suggests that speakers frequently produce a structure that listeners find ungrammatical and have trouble understanding. If the grammars
of the two systems are different, then the assumption of representational
parity must be relaxed.
The field of psycholinguistics has needed precisely what Pickering & Garrod (P&G) provide: a mechanistic theory of dialogue.
P&G’s core idea is that priming is more than just a tool cognitive
psychologists use to learn about the structure or processing characteristics of some cognitive system; in dialogue, priming is itself
a mechanism for producing alignment between interlocutors, and
from it ultimately emerges the common ground that is critical for
successful communication. But their views seem to commit them
to the idea that the production and comprehension systems use
the same representations, and this is one problematic aspect of
their approach.
If the same representations are used for comprehension and
production, speakers should not make use of a syntactic construction that comprehenders find unacceptable and hard to understand. Yet recent work from our laboratory demonstrates that they
do (Ferreira & Swets 2003). We invented a paradigm for eliciting
sentences such as “This is a dog that I don’t know what it eats,”
which is ungrammatical in English. In the experiments, people
first see a picture of a cat (for example) combined with a short verbal label such as “eats fish.” Then they see a picture of another cat
with a minimally different label – “eats chicken,” for example. The
next picture in the series consists of a cat and a question mark in
the same spot where the labels occurred in the other examples.
About 70% to 75% of the time, speakers will describe this third
picture with a sentence containing a relative clause that violates
island constraints (Chomsky 1973) and a resumptive pronoun in
place of the illegal trace (Creswell 2002). (Henceforth we refer to
these as IRPs: Island Resumptive Pronoun sentences.) This experiment, then, provided evidence that speakers can reliably be
induced to produce a structure that is ungrammatical. But perhaps this result is not surprising, as such sentences occur fairly often in natural situations (Prince 1990).
More interesting were the results from a follow-up study in
which speakers performed the same task but under time pressure.
Some views of the IRP structure assume that these forms result
from a failure to plan adequately. The idea is that speakers paint
themselves into a syntactic corner and at the last moment try to
salvage the utterance by inserting the resumptive pronoun. Yet,
we observed that this form was no more likely to occur when
speakers were under time pressure, and alternate, more acceptable forms were actually more frequent (e.g., “This is a dog and I
don’t know what it eats,” or “This dog, I don’t know what it eats”).
It appears, then, that the production system plans this form and
considers it part of its expressive repertoire. After all, the meaning that the utterance attempts to convey is not rare, and the IRP
form is really the only way to verbalize it succinctly (neither of the
paraphrases provided here as examples of alternate forms are as
semantically accurate).
Now, the interesting question is: How does the comprehension
system feel about these same structures? We knew from the start
that they are generally viewed as ungrammatical (McDaniel &
Cowart 1999). But, to be certain, we gave our listener-subjects
written versions of the sentences that the subjects from the production experiments had uttered. Because this form is generally
encountered only in spoken language and therefore is perhaps
more easily processed with appropriate prosody, we conducted an
28
BEHAVIORAL AND BRAIN SCIENCES (2004) 27:2
auditory version of the same grammaticality judgment task. We
obtained the same results. We were not shocked at these findings,
though, because our speaker-subjects often giggled nervously after producing IRPs, betraying their awareness that the utterances
were off in some way. We then conducted two more comprehension studies, one showing that comprehenders do not answer
questions about IRPs as accurately as questions about matched
controls (e.g., “This is a dog that doesn’t know what it eats”), and
the other demonstrating that IRPs elicited more regressive eye
movements, launched from the clause containing the resumptive
pronoun and aimed at the head noun of the relative clause.
It appears that the production system makes reliable use of a
structure that the comprehension system views as ungrammatical
and has a hard time understanding. We believe these results suggest a disconnect between the two systems. Now, P&G might respond that this asymmetry could be because of processing differences rather than representational nonparity. There are three
problems with this argument. First, P&G invoked parity to explain
how interlocutors are able to complete each other’s utterances;
but certainly, if there are forms that only one system likes, then
these predictive feats will not be possible. Second, it is hard to
imagine that hearing an IRP sentence could prime a speaker to
produce one, given that the interpretive system balks at the form.
And third, and most important, the syntactic representations used
by the two systems appear to be nonoverlapping: The IRP form is
part of the grammar that the production system uses, but it is not
available to the comprehension system.
The authors might respond that these studies were obtained in
production and comprehension experiments conducted entirely
independently. Perhaps if these structures occurred as part of an
interactive dialogue, they would either fail to emerge (which is unlikely given that they are attested in natural conversations) or they
would be treated as licit and understood by listeners. This tack is
not unreasonable, but it does make me uncomfortable and it
should make the authors a bit nervous as well. For if we accept this
line of argument, then all of the findings from basic psycholinguistic experiments that have been done in monologue situations
are suspect. Moreover, this response would undermine the authors’ fundamental goal, which is to provide a mechanistic theory
of dialogue – a theory that is informed by decades of systematic
research which has isolated the mechanisms of language processing through experiments that could only be done in monologue
contexts. As a result, the authors’ theory would become less attractive, because one major reason for its appeal is that it brings
together two important traditions in linguistics. Better to question
the assumption of representational parity than to cast doubt on
three decades of significant results, many of which critically informed the authors’ approach.
Visual copresence and conversational
coordination
Susan R. Fussell and Robert E. Kraut
Human-Computer Interaction Institute, Carnegie Mellon University,
Pittsburgh, PA 15213. [email protected]
[email protected]
http://www.andrew.cmu.edu/~sfussell/Susan_Fussell.html
http://www-2.cs.cmu.edu/~kraut/
Abstract: Pickering & Garrod’s (P&G’s) theory of dialogue production
cannot completely explain recent data showing that when interactants in
referential communication tasks have different views of a physical space,
they accommodate their language to their partner’s view rather than mimicking their partner’s expressions. Instead, these data are consistent with
the hypothesis that interactants are taking the perspective of their conversational partners.
We applaud Pickering & Garrod’s (P&G’s) attempt to explain one
of the most basic features of human language – its dialogue struc-
Commentary/Pickering and Garrod: Toward a mechanistic psychology of dialogue
ture. They provide a thought-provoking theory of dialogue in
which coordination in message production occurs when interactants generate their messages from similar situation models and
mimic their partner’s production at the syntactic, semantic, lexical, phonological, and phonetic levels, based on primitive priming
mechanisms. They argue that these alignment processes plus
techniques for repairing misalignments are sufficient to explain
most cases of what others have considered evidence of a deeper
type of perspective-taking, in which speakers take their partners’
mental states into account in forming their own speech.
We believe, however, that P&G’s theory cannot completely explain recent data about language production. In our own work, for
example, we find evidence across several experiments that when
interactants in referential communication tasks have different
views of a physical space, they accommodate their language to
their partner’s view rather than mimicking their partner’s expressions (e.g., Fussell et al. 2000; 2003; Kraut et al. 2002; 2003).
These data are consistent with the hypothesis that interactants are
taking the perspective of their conversational partners.
Consider, for example, the case of deictic reference in a bicycle
repair task (Kraut et al. 2003). In this task, one person (the
“worker”) performs a series of repair tasks under the guidance of
a second person (the “helper”). The helper is located either beside the worker, where both can see and interact with the work
area, or in a separate room connected only by an audio link. In a
third condition, they are connected by an audio/video link through
which the helper can see what the worker is doing but cannot interact with the work area. The conversations typically consist of
helper’s instructions followed by worker’s actions, questions, or acknowledgments of understanding. Interactants can refer to task
objects and locations with either extended linguistic expressions
(e.g., “take the long dangling piece and put it in where the two
large screws are”) or shorter deictic references (e.g., “take this
piece and put it there”).
As Figure 1 shows, the ways in which workers refer to parts,
tools, and other task objects depend on their partners’ ability to
see the work area. In the side-by-side condition, both helpers and
workers can view one another and task objects, and both use a
large number of deictic expressions. In the audio-only condition,
the remote helpers cannot see the work space, and neither workers nor helpers use deictic expressions. The interesting case, from
an alignment point of view, is the video condition. Here, the
helpers can see the workers and work space but cannot point to
objects in it. Under these conditions, helpers rarely use deixis.
However, workers can point to task objects and they know that
helpers can see them do so through the video link. They use deixis
instead of matching the helpers’ nondeictic expressions. If conversational alignment were driven by primitive priming mechanisms, then the workers should use nondeictic references in the
video condition, after hearing helpers’ uttering many of these expressions. (Because the helper could not be seen, he or she would
have no way of using deictic expressions to match the worker’s utterances.) In short, the way workers referred to task objects and
locations depended upon what their partners could see, not the
language their partners previously used to refer to these same objects and locations.
We believe these results demonstrate that one type of deep
common ground – visual copresence – is assessed during message
production, at least for lexical selection processes. Indeed, in experiments where the views can change, interactants often explicitly exchange information about what each can see, with phrases
such as: “Can you see the table?” (Kraut et al. 2002; 2003).
P&G might argue that video-mediated communication is a
nonprototypical dialogue setting and hence may elicit special
processes of assessing deep common ground. Note, however, that
the audio-only discourse from Garrod and Anderson’s (1987)
maze study is similarly nonprototypical. Rather than demonstrating that people in face-to-face dialogues use processes of verbal
alignment in lieu of deeper considerations of common ground,
Garrod and Anderson’s results may indicate that people verbally
Figure 1 (Fussell & Kraut). Percentage of references to task objects containing verbal deixis, by media condition and participant
role (from Kraut et al. 2003).
align primarily when the context has been stripped of all other indicators of common ground.
In their discussion of deep common ground versus automatic
alignment, P&G in essence take a straw-man approach to describing the processes involved in conversational grounding. As
Clark and Marshall (1981) have discussed, common ground can
be determined using heuristics based on community comembership, linguistic copresence, and physical copresence. Some calculations of common ground (e.g., a helper trying to determine
whether a worker on the bicycle task knows what a derailleur is)
may be difficult; others (e.g., a worker trying to determine
whether the helper can see the work space) may be relatively easy.
Ruling out deep common ground as a fundamental process in dialogue production would require a series of carefully controlled
studies that have not been performed to date. P&G’s paper is valuable for the detail with which it specifies an alternative model that
would need to be included in such experiments.
We conclude by observing that calculations of deep common
ground are essential for determining when to speak and what to
say. For example, in the bicycle repair studies, in the audio condition, when the helpers cannot see them, workers describe what
they are doing. Helpers rely on these verbal reports to determine
when to provide new instructions or clarify preceding ones. In the
video and side-by-side conditions, workers do not bother describing what they are doing because they know that the helpers are
watching their activities. If common ground is available to communicators for these processes of message timing and content, it
should not require much additional effort for them to incorporate
it into the messages themselves.
Intrinsic misalignment in dialogue: Why there
is no unique context in a conversation
Jonathan Ginzburg
Dept of Computer Science, King’s College London, London WC2R 2LS,
United Kingdom. [email protected]
http://www.dcs.kcl.ac.uk/staff/ginzburg
Abstract: Pickering & Garrod’s (P&G’s) claim that conversationalists do
not explicitly keep track of their interlocuters’ information states is important. Nonetheless, via alignment, they seem to create a virtually symmetrical view of the information states of speaker and addressee – a key
component of their accounts of collaborative utterances and of self-monitoring. As I show, there is significant evidence for intrinsic contextual misalignment between conversationalists that can persist across turns.
BEHAVIORAL AND BRAIN SCIENCES (2004) 27:2
29
Commentary/Pickering and Garrod: Toward a mechanistic psychology of dialogue
I am very sympathetic with Pickering & Garrod’s (P&G’s) central
message: namely, that there is a need to develop a detailed, computationally oriented account of the mechanisms underlying language use in dialogue, and moreover, that these – not mechanisms
developed for monologue – should be regarded as the primary
mechanisms of the language faculty.1 On a more technical front,
I think that a key claim of the authors, that dialogue participants
do not explicitly keep track of their interlocuters’ information
states, but rather, that this is emergent from the dynamic alignment of each other’s information states, is an important one.
Nonetheless, in seeking intrinsically to couple dialogue participants via alignment, they seem to create a virtually symmetrical
view of the information states of speaker and addressee. Thus, a
key component of P&G’s accounts of collaborative utterances as
well of self-monitoring is the claim that who is speaking at a given
point does not, in some sense, make a difference given alignment:
The addressee can take over or the speaker can “change voice” and
self-correct. As I show in this commentary, this claim is incorrect:
There is significant evidence that the contexts available to the conversationalists are not identical. Hence, there is actually intrinsic
contextual misalignment between conversationalists that can persist across turns.
That a common context (cf. the common ground prominent in
work by Clark Marshall 1981; Lewis 1979; Stalnaker 1978)
emerges in dialogue is an important insight. It yields a better picture of, for instance, querying, than classical speech-act views provide (Searle 1969). So in asking a question, to take one example, a
speaker puts up a question for discussion and whoever takes over
the turn can address it – either the original asker or the original
addressee:
1. A: Whom should we invite to the conference?
A or B: Would Phil be a good idea?
And yet, the actual situation is not as symmetrical as this: The
speaker’s options for self-repair or, indeed, other follow-up are
quite distinct from the addressee’s options. This can be illustrated
succinctly by a phenomenon I have dubbed the Turn Taking Puzzle (Ginzburg 1997a; 1997b). Questions of the form “Why?” involve radical context dependence – pre-theoretically, the context
supplies a propositional referent of some kind (Moore 1995). Interestingly, examples 2a and 2b (below) show that the resolution
accorded to the bare “why” changes according to who keeps or
takes over the turn. The resolution that can be associated with
“Why?” if A keeps the turn is unavailable to B if he or she had
taken over, and vice versa:
2a. A: Which members of our team own a parakeet?
2a. A: Why? ( Why own a parakeet?)
2b. A: Which members of our team own a parakeet?
2b. B: Why? ( Why are you asking which members of our
team own a parakeet?)
2c. Which members of our team own a parakeet? Why am I
asking this question?
Example 2c shows that these facts cannot be reduced to coherence or plausibility – the resolution unavailable to A in example
2a yields a coherent follow-up to A’s initial query if it is expressed
by means of a nonelliptical form. In other words, the context is responsible for these interpretational asymmetries or, rather, they
are a consequence of the fact that distinct contexts are associated
with the conversationalists.
Similarly, a common strategy for requesting a clarification is by
means of a reprise fragment – a word or constituent of the previous utterance (see Purver et al. 2002; 2003 for corpus and experimental evidence on clarification requests, particularly reprise
fragments). Reprise fragments have two prominent understandings (Ginzburg & Cooper, in press), exemplified in 3a (below).
However, it is quite strange for a speaker to follow up her utterance with a reprise fragment. This becomes felicitous only if followed up by an additional correction, such as “Wait, did I say Bo,
no I mean Lou,” or some such. However, even then the readings
that arise in example 3a, whose resolution is radically context-dependent, are not manifested:
30
BEHAVIORAL AND BRAIN SCIENCES (2004) 27:2
3a. A: Did Bo leave?
3a. B: Bo? ( Either: Are you asking if BO of all people left?
Or: Who were you referring to as “Bo”?)
3b. A: Did Bo leave?
3b. A: #Bo?2
It is worth noting that contextual asymmetries of this kind can
persist for quite a number of turns – essentially as long as a given
discourse topic remains under discussion. Example 4 is an extract
from the British National Corpus in which Chris’s “Why?” is naturally understood to refer to Norrine’s utterance five turns back,
an utterance which seems to be viewed as grounded (Clark 1996):
4. Norrine (1): When is the barbecue, the twentieth? (pause)
Something of June.
Chris (2): Thirtieth.
Norrine (3): A Sunday.
Chris (4): Sunday.
Norrine (5): Mm.
Chris (6): Why? ( Why do you ask when is the barbecue?)
Norrine (7): Because I forgot (pause) That was the day I was
thinking of having a proper lunch party but I won’t do it if
you’re going out.
Note that the resolution associated with Chris’s “Why?” is simply
unavailable to Norrine at all subsequent points, as illustrated in
the constructed variant of statement (4) in example 5a. As with
previous examples, this cannot be explained on “pragmatic”
grounds because the speaker can fairly coherently express the requisite reading in nonelliptical fashion, as in example 5b:
5a. Norrine (1): When is the barbecue, the twentieth? (pause)
Something of June.
5a. Chris (2): Thirtieth.
5a. Norrine (3): A Sunday.
5a. Chris (4): Sunday.
5a. Norrine (5): # Why? (Cannot mean: Why is Norrine asking
when is the barbecue?)
5b. Norrine (5): Mm. Why am I asking? Can you guess?
5a. Chris (6): No idea.
Phenomena such as this suggest that the different roles conversationalists play with respect to a given utterance (speaker vs.
addressee) are not something that gets neutralized in the utterance’s aftermath. The contextual possibilities available for one
conversationalist differ from those of the other conversationalist
(I am referring to dialogue here; as P&G point out, multilogue is
a genre with various distinct properties from two-person dialogue). In other words, a single context is not fully adequate to describe dialogue, even when talking about “public” context, which
results from overtly registered conversational actions. Instead,
one needs to view dialogue as involving updates by each conversationalist of some type of a publicly accessible domain which is
relative to each conversationalist and so is parametrizable by unpublicized factors such as individual goals and intentions (cf.
Hamblin’s individual commitment slate, Hamblin 1970). A framework named KOS spells out this view and develops theoretical accounts as well as computational implementations of illocutionary
and metacommunicative acts, including a detailed account of puzzles such as the Turn Taking Puzzles exemplified above (cf.
Cooper et al. 2000; Ginzburg 1996; 2002; forthcoming; Larsson
2002).
ACKNOWLEDGMENTS
The research described here is funded by grant number RES-000-23-0065
from the Economic and Social Research Council of the United Kingdom
and by grant number GR/R04942/01 from the Engineering and Physical
Sciences Research Council of the United Kingdom.
NOTES
1. The authors point out the dearth of work in mechanistic psychology
and theoretical linguistics (primarily by syntacticians) on dialogue. Since
the late 1990s, however, there has been work by formal and computational
semanticists precisely on developing theories of information states and
their dynamics in dialogue – see, for example, work within the EU
TRINDI project (Consortium 2000) and the annual series of conferences
Commentary/Pickering and Garrod: Toward a mechanistic psychology of dialogue
on the semantics and pragmatics of dialogue. For those interested,here is
some further identifying information on the conferences: MUNDIAL
(Munich 1997, http://www.cis.uni-muenchen.de/sil/workshop/dialogwsh
.html); TWENDIAL (Twente 1998, http://parlevink.cs.utwente.nl/
Conferences/twlt13.html); AMSTELOGUE (Amsterdam 1999, http://www
.hum.uva.nl/computerlinguistiek/amstelog); GOTALOG (Gothenburg
2000, http://www.ling.gu.se/gotalog); BIDIALOG (Bielefeld 2001,
http://www.uni-bielefeld.de/BIDIALOG); EDILOG (Edinburgh 2002,
http://www.ltg.ed.ac.uk/edilog/); DIALBRUCK (Saarbrücken 2003,
http://www.coli.uni-sb.de/diabruck/).
2. The # symbol is used here, as standard in linguistics, to mark an infelicitous utterance.
Dialogue: Can two be cheaper than one?
Sam Glucksberg
Psychology Department, Princeton University, Princeton, NJ 08544.
[email protected]
http://www.princeton.edu
Abstract: Pickering & Garrod (P&G) argue that language processing in
dialogue is in principle easier than in monologue. Although dialogue situations may provide more opportunities for facilitative priming, those priming mechanisms are also available in monologue situations. In both cases,
the interactive alignment model calls strict modular accounts of language
processing into serious question.
Pickering & Garrod’s (P&G’s) argument for the primacy of dialogue is a refreshing alternative to the isolated sentence as the unit
of analysis in psycholinguistics. Others, such as Krauss, Clark, and
their students and colleagues, have made similar arguments (cf.
Clark 1996; Fussell & Krauss 1992), but none have made the additional argument that much of dialogue production and comprehension is automatic, requiring minimal cognitive or linguistic resources. The details of the priming processes have yet to be
worked out in detail, but P&G provide enough empirical evidence
to warrant optimism about an eventual fleshing out of their interactive alignment proposal. Beyond such details, their account has
several intriguing implications, not the least of which is the claim
that language production and comprehension are easier in interactive dialogue than in monologue. This flies in the face of the consensus that dialogue requires complex inferences about the mental state of one’s interlocutor that are not required when one is,
say, talking to oneself, or processing utterances under monologue
conditions.
Is language processing in dialogues inevitably easier (more fluent, requires less processing, etc.) than in monologues? I do not
think this question can be answered unequivocally. I can think of
contexts in which dialogue language production would be quite
difficult and contexts in which monologue production would be
quite easy. An example of the former might be trying to hold a conversation with a sullen teenaged son or daughter on the merits of
atonal music. Examples of the latter might include the think-aloud
paradigm used in research on reasoning and problem-solving, or
young children talking to themselves or to their dolls when no one
else is around. So, as usual, the answer is: It depends.
A more useful question addresses the ways in which dialogue
might be more likely than monologue to facilitate language processing. Such ways might in principle be available only in dialogue.
Alternatively, it might just be a matter of likelihood, rather than a
principled difference between monologue and dialogue. Consider
the stages of speech production that could be facilitated in monologues and in dialogues. Following Levelt (1989), production
could be facilitated at any or all of the following stages: the message (what to talk about); grammatical encoding (syntactic representation and lexical access); phonological encoding; and articulation. One of the more obvious ways to facilitate speech production
is to provide something to talk about. In dialogue, the situation (including one’s interlocutor) usually provides this, but of course
there are (sometimes painful) situations where one does not know
what to talk about next. In monologue, including the occasional
classroom lecture (or a BBS commentary, for that matter), one can
be at a loss as to what to say or write, but this is not a principled
characteristic of monologues, just an unfortunate occasional one.
Grammatical and phonological encoding, on the other hand,
may well be advantaged in dialogue via the mechanisms of priming and the opportunities for imitation. P&G make a convincing
argument that such priming not only occurs but may well be automatic. However, such priming could in principle operate in
monologue as well, although the opportunities are undoubtedly
less likely. In monologue, as in dialogue, opportunities to repeat
specific syntactic forms should facilitate successive productions.
Similarly, opportunities to repeat specific word senses should facilitate lexical access and sense selection. What’s missing from
monologue, of course, are the contributions of an interlocutor,
which can provide opportunities for imitation, as well as for syntactic, lexical, and phonological priming (but see Ferreira & Griffin 2003 for evidence of potential sources of priming in monologue).
Finally, P&G note that contextual constraints might operate
more efficiently in dialogue than in monologue, but, again, this is
probably not a principled difference. They write “With respect to
lexical ambiguity, we predict that context will have a very strong
role, so that effects of meaning frequency can be overridden”
(sect. 5.4, para. 3). To my knowledge, this prediction has not yet
been tested under dialogue conditions. However, it has been confirmed under monologue conditions, most recently by Sereno et
al (2003), who found very early context effects on word recognition that overrode frequency effects.
To the extent that automatic priming can facilitate language
comprehension and production, dialogue may well afford more
opportunities for such priming than monologue. Nevertheless, the
same mechanisms should be available for both. P&G’s contribution is thus more profound than the observation that language processing in dialogue might be easier than in monologue. Instead,
the contribution is in the insight that there may be mechanisms of
language production and comprehension that can circumvent the
need for costly cognitive processing in interactive conversation.
People may not need to use a theory of mind or make inferences
about common ground and mutual knowledge. They may, instead,
rely on automatic priming both within and between interlocutors
to produce alignment of representations, which, in turn, makes for
successful and relatively effortless speech processing. The theoretical implication for theories of language processing is clear: A
strictly modular, bottom-up language processing model simply
will not do the job.
ACKNOWLEDGMENT
I thank Boaz Keysar for helpful comments on an earlier draft of this commentary.
Resonance within and between linguistic
beings
Stephen D. Goldingera and Tamiko Azumab
aDepartment of Psychology, Arizona State University, Tempe, AZ
85287-1104; bDepartment of Speech and Hearing Science, Arizona State
University, Tempe, AZ 85287-0102. [email protected]
[email protected]
Abstract: Pickering & Garrod (P&G) deserve appreciation for their cogent argument that dialogue merits greater scientific consideration. Current models make little contact with behaviors of dialogue, motivating the
interactive alignment theory. However, the theory is not truly “mechanistic.” A full account requires both representations and processes bringing
those representations into harmony. We suggest that Grossberg’s (1980)
adaptive resonance theory may naturally conform to the principles of dialogue.
BEHAVIORAL AND BRAIN SCIENCES (2004) 27:2
31
Commentary/Pickering and Garrod: Toward a mechanistic psychology of dialogue
In their target article, Pickering & Garrod (P&G) present a clear
and compelling case: Despite its being the predominate form of
linguistic behavior, dialogue has been unduly ignored. Moreover,
once dialogue is studied, it displays properties of entrainment at
all levels (semantic, syntactic, phonetic, and others), which few
psycholinguistic theories can address. To fill this void, the authors
propose the interactive alignment model, wherein interlocutors
share information at all linguistic levels, quickly arriving at states
near equilibrium, allowing people to communicate with tremendous ease and efficiency. A by-product of alignment is that people
tend to converge in dialogue, using similar expressions, words, and
phrase structures. Beyond such linguistic constructs, interlocutors
also tend to converge in their manner of speaking and even their
physical postures (Capella 1981; Giles et al. 1991; Newtson 1994;
Shockley et al. 2003).
The interactive alignment model is advanced to explain these
coupling dynamics, using standard priming mechanisms. This approach is representation-driven, wherein one person’s utterances
prime ideas, words, and syntactic structures in his or her interlocutor. Mutual priming persists as a loosely coupled feedback
loop (Van Gelder & Port 1995), so the same forms tend to occur
repeatedly in conversation. At a general level, we believe this idea
must be correct. However, the proffered model is underspecified,
stated without explicit processes that act on the activated representations. Thus, the model nicely characterizes the problem
space for a psychology of dialogue, but leaves many open questions. For example, how does the priming of extant representations predict new entities, such as routines, new jargon, or momentary changes in articulatory habit? How does speech
perception lead to cascaded priming across linguistic levels? In
short, what are the processes that explain the emergence of new
psychological structures when people engage in dialogue?
In both language perception and production, models based on
interactive activation have enjoyed great success (e.g., Dell 1986;
Gaskell & Marslen-Wilson 1997; Levelt et al. 1999; McClelland &
Elman 1986; McClelland et al. 1989). However, as P&G observe,
these are all monologue models. Moreover, all generally assume
fixed networks of nodes and connections, with weighting schemes
that develop over extensive periods of training. Given their architecture, it is difficult to imagine how such models could communicate at multiple “levels,” as suggested in Figure 2 of the target
article. Of greater importance is that the core processes of interactive activation are not amenable to rapid transformation of representations or combinatorial rules. If one interlocutor can rapidly
change another’s internal weights, people would be subject to
“catastrophic unlearning” of language.
As an alternative approach, we believe that interactive alignment may naturally “fall out” of processing in Grossberg’s adaptive resonance theory (ART; Grossberg 1980; 1999; Grossberg &
Myers 2000; Grossberg & Stone 1986; Grossberg et al. 1997; Vitevitch & Luce 1999). In speech perception, a basic hypothesis of
ART is that conscious percepts are emergent products of resonant
brain states. This is an interdependent system: Perception occurs
when bottom-up and top-down knowledge sources bind into stable states. Processing begins when feature input activates items
(feature clusters) in working memory. Items, in turn, activate list
chunks in memory. Chunks are products of prior learning (perhaps prototypes) corresponding to feature combinations, such as
phonemes, words, or common expressions.
Once items activate list chunks, a feedback cycle begins. Items
feed activation upward through synaptic connections, and inputconsistent chunks return activation. If items receive sufficient topdown confirmation, they continue sending activation upward.
Within limits, this feedback loop (a resonance) is self-perpetuating, binding the respective activation patterns into a coherent
whole. The bottom-up pattern that initiates interactive activation
need not perfectly match its resultant feedback for resonance to
occur. Cooperation between “levels” and competition within “levels” smooth out small mismatches, but large mismatches prohibit
resonance. This stipulation seems necessary for dialogue, as an-
32
BEHAVIORAL AND BRAIN SCIENCES (2004) 27:2
other person’s utterances (even their manner of producing phonemes) may never perfectly match one’s stored representations.
Rather than assuming tiered representations, chunks in ART
are attractor states. Of particular importance is that resonance can
momentarily bind attractors in original combinations, allowing the
creation of new mental structures (Carpenter & Grossberg 1987).
This is the natural basis of communication, wherein novel utterances are constructed from familiar parts. In ART, however, those
“parts” are not relegated to modular levels of representation: Although familiar patterns (e.g., common words) will cohere quickly
in processing, all structures – both familiar and novel – must selforganize through competitive dynamics. Hence, in a dialogue, different structures, such as nascent jargon or momentary routines,
can naturally emerge, as P&G describe.
As we have argued elsewhere (Goldinger & Azuma, in press),
the self-organizing nature of ART makes it uniquely suited to
model context-sensitive perception, as dialogue seems to require.
For example, when ART recognizes a word spoken by Mary, its
emergent percept will reflect the unique aspects of her utterance,
allowing token-specific encoding (Goldinger 1998). Hence, the
model has a natural basis to engage in dialogue, where contextspecific cognition is pervasive. However, we are not suggesting
that ART can currently explain the wide-ranging behaviors described by P&G. ART has successfully simulated many perceptual
and memorial behaviors (among others; see Carpenter & Grossberg 1991). However, full sentence processing, even in monologue, remains a serious challenge.
As befits the study of dialogue, we envision a two-way street
with equal benefits to students of dialogue and students of self-organizing networks. The properties of ART may give substance to
the metaphoric theory described by P&G. By the same token, the
unique challenges of dialogue may force changes in ART: To study
dialogue, it would be ideal to have separate, self-contained ART
simulations that communicate with each other through one or
more channels, working together toward some shared goal (as in
the authors’ maze task). To our knowledge, no such study has ever
been conducted using ART. However, a conceptually similar study
by de Boer (2000) suggests that dialogues between autonomous
models can evoke self-organization. De Boer created speech-perceiving “agents” equipped with articulatory synthesizers, perceptual systems that calculated the differences between signals, and
associative memories for storing signals. Each agent was able to
interact with others in an “imitation game.” The agents’ task was
to imitate other agents accurately with a large repertoire of vowels; they could update their vowel repertoires based on their interactions. In many simulations, de Boer found that stable vowel
systems consistently self-organized from these simple “dialogues.”
Moreover, these emergent vowel systems were remarkably similar to those found in human languages. This suggests that self-organizing structures may partially drive the basic forms of language.
It also suggests that dialogue is a powerful force that should be
studied both empirically and computationally.
ACKNOWLEDGMENTS
Support for this commentary was provided by grants R01-DC04535-03
(SDG) and 1-R03-DC4231-0182 from the National Institute of Deafness
and Communicative Disorders (NIH). We thank Paul Luce, Greg Stone,
and Guy Van Orden for many helpful discussions on the topics of resonance and speech units.
Commentary/Pickering and Garrod: Toward a mechanistic psychology of dialogue
Dialogue in the degenerate case?
Patrick G. T. Healey
Interaction, Media, and Communication Research Group, Department of
Computer Science, Queen Mary, University of London, London E7 0QY,
United Kingdom. [email protected]
http://www.dcs.qmul.ac.uk/~ph
Abstract: The interactive alignment model treats within- and between-individual co-ordination as essentially equivalent. It is argued that this leads
to a conservative account of alignment that presupposes high levels of linguistic and conceptual co-ordination. Data from the maze task are used to
argue that this approach is not sufficient to account for important co-ordination phenomena.
Pickering & Garrod (P&G) bring together a large body of empirical evidence that highlights an important phenomenon. Participants in dialogue display higher levels of phonological, lexical, syntactic, semantic, and situational co-ordination than is expected
from considering processes of production and comprehension in
isolation. The authors propose that the central mechanism driving
this level of communicative alignment is priming. This is envisaged as an automatic, bidirectional process operating in parallel
on several different levels of representation. Importantly, they
suggest that this process operates in essentially the same way both
within and between speakers.
The treatment of dialogue participants as effectively interchangeable embodies a strong claim. It presupposes a high degree
of equivalence between the representational systems underpinning
communicative co-ordination. In places this is qualified as equivalent “in all essential respects” or “basically the same.” However, in
cases where priming is the mechanism of alignment, the residual
differences between individual’s representations are irrelevant.
What could underwrite the requisite similarity in mental representation? Priming is itself a conservative mechanism that alters
the accessibility but not the form or content of the associated mental representations. For example, the account of co-ordination in
the maze game distinguishes a basic repertoire of situational models of the maze (figure scheme, path scheme, line scheme, and matrix scheme) and a variety of referring expressions built up from
the lexicon, syntax, and semantics of a fragment of English (Garrod & Anderson 1987). By hypothesis, the same basic repertoire
of mental models and fragment of English are available to both
participants prior to co-ordination. Priming does not change the
repertoire; rather, it promotes alignment by favouring one particular model and its associated referring expressions.
Interactive alignment predicts that when maze game participants co-ordinate on a particular description and situational
model, it should become more strongly primed, and they should
stick with it. However, several aspects of the experimental data do
not fit this prediction. Evolution and development of description
schemes are common. As P&G observe, maze game pairs often
develop their own idiosyncratic description schemes. Pairs switch
description schemes frequently; for example, the isolated pairs in
Garrod and Anderson (1987) and the non-community group in
Garrod and Doherty (1994). Moreover, pairs do not reliably coordinate on the scheme that is most highly primed initially. For example, in Garrod and Doherty’s (1994) community group, the line
scheme is most frequent initially, but they subsequently co-ordinate on a matrix scheme.
In order to account for phenomena of this kind and to provide
for co-ordination where participants are not effectively interchangeable, P&G propose iterative interactive repair as a key additional co-ordination mechanism. If a participant cannot find an
interpretation for an utterance, he or she either shifts perspectives
until an interpretation can be found or reformulates. This suggestion is not developed in detail, but it appears P&G envisage this
basic form of repair as a shift between different co-ordination
equilibria in the sense of Lewis (1969).
Data from the maze task suggest that this is inadequate. The results reported by Garrod and Anderson (1987), Garrod and Doherty (1994), and Healey (1997; in preparation) indicate underly-
ing patterns of migration from figure and path schemes to more
abstract line and matrix schemes. In addition, when maze game
pairs have trouble co-ordinating, there is evidence that they show
a reliable preference for shifting to a more basic (figure or path)
scheme. The direction of these shifts is not predicted by frequency
of prior exposure (Healey 1997; in preparation).
These observations show that the choice between different situational models is not neutral in the way that a co-ordination game
or priming mechanism presumes. It is sensitive to the particular
properties of the different situational models and, presumably,
this relates to their implications for co-ordination. The pattern of
preferences for shifts in description scheme and/or reformulations needs to be accounted for by the co-ordination mechanisms.
P&G aim to provide a model of dialogue co-ordination that
avoids implausible assumptions about interlocutors constructing
elaborate models of each other’s context and mental states. This is
surely right. Explicit negotiation and repair are relatively rare and
do not provide a general account of co-ordination. However, the
mechanisms of the interactive alignment model do not seem to do
the work required.
Ironically, the idealisation of speaker and hearer as interchangeable reproduces one of the problems with treating dialogue
as a form of monologue: the implication that participants are linguistically and conceptually transparent to one another. Interactive alignment focuses on dialogue in the degenerate case: interactions in which people are, in a sense, already co-ordinated. As
P&G note, dialogue is important partly because it is the primary
context for exposure to and acquisition of language. But this is also
the situation in which the assumptions embodied in the interactive alignment model are least likely to be satisfied.
Interactive alignment:
Priming or memory retrieval?
Michael Kaschaka and Arthur Glenbergb
aDepartment
of Psychology, Florida State University, Tallahassee, FL 32310;
of Psychology, University of Wisconsin, Madison, WI 53706.
[email protected]
[email protected]
bDepartment
Abstract: Pickering & Garrod’s (P&G’s) interactive alignment model explains the existence of alignment between speakers via an automatic priming mechanism. We propose that it may be preferable to explain alignment
through processes of memory retrieval. Our discussion highlights how
memory retrieval can produce the same results as the priming mechanism
and presents data that favor the memory-based view.
Pickering & Garrod (P&G) claim that dialogue is marked by a
great degree of alignment between interlocutors. The alignment
is taken to arise from the operation of a largely automatic priming
mechanism. We suggest that much of what P&G ascribe to the operation of a priming mechanism can be captured by a memorybased view of language processing.
The memory-based view is rooted in the assumption that language processing is performed against the background of one’s experience. The comprehension of an incoming utterance is guided
by experience of utterances with similar structures and meanings,
just as experience communicating via speech shapes the production of utterances. This assumption forms the basis of many extant
psycholinguistic theories: constraint-based approaches to sentence processing (e.g., MacDonald et al. 1994), memory-based
models of text processing (Myers & O’Brien 1998), and theories
of lexical access (Goldinger 1998). The memory retrieval mechanism is what has been called a “global matching” process (e.g.,
Hintzman 1986). That is, the memory retrieval process involves a
global, parallel search of memory that is influenced by both contextual and temporal factors (e.g., the retrieval process shows a recency effect when recently processed language is treated as a separate context from previously processed language).
BEHAVIORAL AND BRAIN SCIENCES (2004) 27:2
33
Commentary/Pickering and Garrod: Toward a mechanistic psychology of dialogue
The memory-based view explains why processing appears different in dialogue and monologue situations. It also provides an
alternative to the priming mechanism around which P&G build
their theory. For monologue experiments, which involve processing language in a decontextualized and artificial situation, there is
little in the way of specific prior experience that might constrain
the memory retrieval process. The structure of the experiments
(e.g., randomized trials, random selection of materials) also makes
it unlikely that recent experience within the session will exert a
systematic influence on task performance. Hence, monologuebased experiments produce memory retrieval conditions that are
sensitive to general properties of one’s linguistic experience, such
as the frequency of particular syntactic structures or word meanings. On the other hand, dialogue situations foster highly constrained memory retrieval. For example, the perceptual similarity
between utterances within a dialogue can constrain memory retrieval so that it most strongly resonates with previous experiences
with that particular speaker (see Goldinger 1998). All other things
being equal, memory retrieval is also weighted toward more recent experience (see Bock & Griffin 2000). Together, these factors
can produce the kinds of repetition and alignment that P&G discuss.
Support for the learning- and memory-based view comes from
experiments reported in Kaschak (2003). The participants in these
experiments were trained on a syntactic construction with which
they were not familiar, called the Needs construction. In the first
part of the experiments, participants in the Needs training condition were exposed to the novel construction in sentences such as:
“The meal needs cooked given that dinner is in an hour.” Participants in the control training condition had an identical training
phase in which they read the standard version of the Needs construction for their dialect (“The meal needs to be cooked given that
dinner is in an hour”). By the end of the training phase (which included 10 to 12 exposures to the Needs or standard constructions,
depending on the experiment), participants in the Needs training
condition read the novel construction as quickly as participants in
the control training condition read the standard construction. This
is consonant with P&G’s claim that recent experience has a strong
influence on language processing. Indeed, it fits nicely with the
claim that recent experience can override global frequency effects
in determining the speed of processing for particular kinds of language.
One of the questions addressed in these experiments was how
training on the Needs construction would affect the processing of
another construction with which the participants were already familiar: the modifier construction (“The meal needs cooked vegetables given that the guests are vegetarian”). Training on the Needs
construction sets up an ambiguity between the novel structure and
the existing modifier construction, such that the sentence is ambiguous at the word “cooked.” The priming mechanism favored by
P&G (and discussed earlier in Pickering & Branigan 1998) makes
a straightforward prediction about the influence of Needs training on the processing of the modifier construction. Recent exposure to Needs sentences will prime the syntactic features associated with the construction, leading readers to initially choose the
Needs interpretation of the ambiguous sentences (thereby slowing processing of the modifier construction).
Kaschak’s (2003) results showed that the effect of Needs training on the reading of the modifier construction depends on the
particular nature of one’s experiences during training. If participants reading, “The meal needs cooked . . .” during training initially interpreted the sentence as the modifier construction before
figuring out that the sentence was an example of the Needs construction, Needs training facilitated the participants’ reading of
the modifier construction. If the participants were given instructions such that they did not initially misinterpret the Needs sentences as examples of the modifier construction, the Needs training did not facilitate processing of the modifier construction.
Kaschak (2003) discusses how these data can be explained by assuming that the participants in these experiments remembered
34
BEHAVIORAL AND BRAIN SCIENCES (2004) 27:2
not only the outcome of processing the initial examples of the
Needs construction, but also the processing work that went into
comprehending those sentences. This leads to the conclusion that
it is the memory that one has for particular linguistic experiences,
rather than a priming mechanism, that may better explain how recent experience with language shapes the way subsequent utterances are processed.
We resonate with many of the arguments and claims advanced
by P&G, but we think their explanation can be substantially improved by replacing the notion of alignment-by-priming with
alignment based on retrieval from memory of recent processing
episodes.
Grammars with parsing dynamics:
A new perspective on alignment
Ruth Kempson
Philosophy Department, King’s College London, London WC2R 2LS, United
Kingdom. [email protected]
http://www.kcl.ac.uk/kis/schools/hums/philosophy/top.html
Abstract: This commentary argues that dialogue alignment can be explained if parsing-directed grammar formalisms are adopted. With syntax
defined as monotonic growth of semantic representations as each word is
parsed, alignment between interlocutors is shown to be expected. Hence,
grammars can be evaluated according to relative success in characterizing
dialogue phenomena.
Although Pickering & Garrod (P&G) suggest grammars should be
compatible with alignment patterns in dialogue, commending
multilevel constraint-based formalisms, they could have posed a
stronger challenge. Given that dialogues constitute primary data,
can linguistic theory provide a basis for explaining alignment?
Does some grammar-external mechanism determine alignment,
or is it a consequence of the architecture of the language system
and the way it is put to use in dialogue? The authors opt for the
first alternative. However, with grammar formalisms being promoted which reflect the dynamics of a parser (Kempson et al.
2001; Phillips 1996), alignment can be more strongly buttressed
by natural-language grammars than they anticipated.
In Dynamic Syntax (Kempson et al. 2001), a grammar formalism is defined that provides such underpinnings. A constraintbased architecture is defined, with syntax as the progressive
growth of semantic representations, specifying possible parse routines. That syntactic and semantic alignment go hand in hand, is
thus ensured. Semantic representations (logical forms) are in tree
format, with the propositional formula assigned to a string decorating the top node and the dominated nodes in the tree decorated
with sub-terms of this formula (quantification is expressed in the
form of names denoting witness sets, so resulting representations
resemble the situational models of Johnson-Laird 1983). Growth
of logical forms, central to this concept of syntax, is goal-directed
– first setting out the overall goal to establish a propositional structure, with subgoals introduced for predicate-argument structure.
The partial structure induced by such goals is enriched by updates
which the words in sequence provide for these skeletal structures;
and these, together with construal of pronouns relative to context
(context construed as a sequence of semantic representations), determine the building up of a semantic representation which, when
complete, meets the goal (a goal tree).
Lexical specifications are also defined in terms of tree growth,
with each word associated with some input it provides to progressive articulation of semantic representations. That lexical alignment should go along with syntactic/semantic alignment thus follows. The apparent syntax-specific alignment across interlocutors
relative to constant semantic content, for example, repeating double-object constructions rather than their full dative equivalent –
a putative counter-example to such collapse of syntactic/semantic/
Commentary/Pickering and Garrod: Toward a mechanistic psychology of dialogue
lexical distinctions – follows from this definition of lexical content.
A word with both double-object and full-dative variants provides
two distinct forms of update onto a single structure, hence a processing ambiguity. Repeating words across interlocutors will retrieve the lexical specification just used, not some alternant. By
collapsing what would in other systems be discrete syntactic/semantic/lexical levels, the multilevel nature of alignment is transformed into a single phenomenon.
We can now see perception/production correspondences and
self-monitoring of production as essential, not merely functionally
convenient. The consequences of taking parsing as basic when designing a production model are twofold. First, in using the same
grammar formalism, the production mechanism must be using the
same structure-inducing growth process as is articulated for parsing, the major difference being that in production, the goal tree is
known. Following general tree development steps defined by the
parser, which in the initial steps of the parse define only a skeletal
partial tree, the task is to retrieve from the lexicon some appropriate first word that can take the partial tree so far defined and
update it to provide a more developed structure that subsumes the
goal tree. If such a subsumption relation holds, then utterance of
that word is licensed. This task repeats itself to induce a sequence
of words until the goal tree is obtained through the parse mechanism – hence the correspondence between comprehension and
production and essential self-monitoring in production.1
Alignment between interlocutors is now explicable. The major
challenge of production, given some conceptual structure to be
communicated, is selecting words to express that content, imposing a general lexicon search. The incrementality of the tree growth
mechanisms suggests that this search is activated successively,
which is a huge task. However, production, like parsing, is contextdependent; and part of the context is that small subset of the
lexicon already accessed. Repeating words accordingly avoids a
general lexicon search. Pronouns equally, being place-holders
substituted by contextually available representations, sidestep full
lexicon access. And with elliptical fragments, lexical search is restricted to the fragment. So the high rate of alignment, as displayed by these forms, arises because production, following the
parsing dynamics, is context-dependent, which is essential for
minimizing the word-retrieval burden.
Finally, shared utterances are expected. Given the use of parsing tools to induce production steps, in successful communication,
interlocutors must coincide on constructing some particular sequence of structures. The shift from parser to producer is thus
straightforward if the parser, having constructed some partial tree,
makes an abduction step to determine what is needed to complete
it. The shift from producer to parser is equally natural: it is a shift
into the task of processing lexical input to complete the partial
tree, which up to that point was constructed as the means of making production choices.
Hence, there is a stronger conclusion than the authors’ modest
challenge to linguists. Rather, we might adopt methodologies in
which linguistic theories are evaluated by their potential for expressing coordination of comprehension and production in dialogue. Grammar formalisms defined in terms of parsing meet this
challenge in a particularly direct way.
ACKNOWLEDGMENTS
This commentary is a reflection of the work of a team. Thanks are particularly due to Masayuki Otsuka and Matthew Purver for extensive discussions on generation modeling. However, nobody other than myself can be
blamed for the particular dialogue perspective suggested here.
NOTES
1. Modeling generation using the dynamic syntax framework constitutes ongoing research (Purver & Otsuka 2003).
Is alignment always the result of automatic
priming?
Robert M. Krauss and Jennifer S. Pardo
Psychology Department, Columbia University, New York, NY 10027.
[email protected]
[email protected]
http://www.columbia.edu/cu/psychology/commlab/
http://www.columbia.edu/~jsp2003/
Abstract: Pickering & Garrod’s (P&G’s) mechanistic theory of dialogue
attempts to detail the psychological processes involved in communication
that are lacking in Clark’s theory. By relying on automatic priming and
alignment processes, however, the theory falters when it comes to explaining much of dialogic interaction. We argue for the inclusion of less
automatic, though not completely conscious and deliberate, processes to
explain such phenomena.
In his influential book Using Language, Clark (1996) argued
against a conceptualization of communicators as autonomous information processors, contending that language use is intrinsically
a joint activity and examining communicators’ practices from this
perspective. His account provides a compelling description of
some of the things talkers accomplish in dialogue, but it is weak
on details of the psychological processes on which these accomplishments rest. Pickering & Garrod’s (P&G’s) mechanistic theory
of dialogue is an attempt to provide such an account. In many respects, it is quite successful, providing a glimmer of light at the
end of a long psycholinguistic tunnel. However, there are some respects in which the theory falls short of its authors’ goal of providing a mechanistic explanation for the phenomena Clark described.
There are two key propositions in the target article’s argument:
first, that communication entails the alignment of participants’ situation models; second, that priming is the principle mechanism
by which this is accomplished. We find the first proposition more
convincing than the second. By stressing the automaticity of the
process, P&G’s mechanistic theory appears incapable of accounting for the way interlocutors use information in what appears to
be a more reflective fashion. Because of space limitations, we shall
confine ourselves to just a few instances where the theory is deficient.
1. The automatic priming mechanism appears to leave no room
for addressee accommodation in the absence of a misunderstanding, yet there are many examples of interlocutors taking their partners’ informational needs into account that are incompatible with
automatic priming. For example, Kingsbury (1968) found that
Bostonians who were asked, “I’m from out of town, can you tell
me how to get to Jordan Marsh?” gave more detailed directions
than those simply asked, “Can you tell me how to get to Jordan
Marsh?” When asked the latter question in an exotic (nonlocal) dialect, Bostonians also gave more detailed directions. Fussell and
Krauss (1992) found that the number of words used in the initial
reference to a photo of a landmark in an interactive coordination
task was a function of the landmark’s perceived identifiability – the
more identifiable the landmark was thought to be, the fewer the
words used to refer to it. It is not clear how priming could account
for these results or those of a host of similar studies (see Krauss &
Fussell 1996 for a review). We believe that such “audience design”
effects (Clark & Murphy 1982; Fussell & Krauss 1989) occur prior
to referent selection, and not just as an attempt to remedy an
emergent misunderstanding.
2. Representational alignment requires that two or more entities be identical in some way. Assertions of identity may work for
descriptions of relatively abstract syntactic and lexical levels of
representation, but not for representations at the phonological
level, because the phonological level is graded and repeated phonetic elements (even within the same talker) are not physically
identical. Although different instances of a phoneme may be perceived as members of the same phoneme category, perception
preserves some phonetic distinctions. Hence, the very level that is
the point of contact between talkers, the acoustic-phonetic level,
BEHAVIORAL AND BRAIN SCIENCES (2004) 27:2
35
Commentary/Pickering and Garrod: Toward a mechanistic psychology of dialogue
cannot support an automatic alignment/imitation-based model
because it is impossible to produce a perfect imitation – the monitoring system would be reporting continual error. Although the
mechanism allows for degrees of alignment, we lack a rule for determining how much alignment is required.
Interestingly, the strongest evidence cited by P&G for phonetic
alignment comes from Goldinger’s (1998) study of lexical shadowing. However, Goldinger’s procedure assessed perceived imitation, which is not equivalent to phonetic similarity. Imitations of
the voices of well-known figures by vocal impressionists are caricatures that exaggerate particularly salient features rather than
produce acoustically accurate reproductions. The remainder of
the published evidence for phonological imitation is mainly of increased similarity in speech rate and pitch (Giles et al. 1991; Natale 1975a; 1975b) and voice onset timing (Sancier & Fowler
1997). In a continuously variable system, what degree of similarity constitutes an imitation?
3. Interlocutors’ speech does not always become more similar
over the course of their interaction; in some cases, interaction
yields divergence rather than convergence. Moreover, the speech
of different participants may change to different degrees; convergence can be radically asymmetrical. It would be little more than
an annoyance if such departures from symmetrical convergence
were random, but frequently they reflect social processes that are
fundamental to the interlocutors’ interpersonal relationship and
the ways in which they define the interaction situation. For example, Bourhis and Giles (1977) found divergence in accentedness
when a talker’s ethnic identity was devalued. Gregory and Webster (1996) found that the symmetry of pitch convergence between a talk-show host and his guests depended on the guest’s status relative to that of the host – not surprisingly, higher-status
guests changed less than their lower-status counterparts. Again, it
is difficult to reconcile such phenomena with an automatic priming explanation. It seems more plausible to suppose that they derive from a prior assessment that sets up the system to evoke particular kinds of priming.
Although our commentary is directed at what we see as deficiencies in P&G’s theory, we applaud their attempt to move beyond participants’ goals and intentions and focus on the psychological mechanisms that make dialogue possible. Their thoughtful
paper is admirable in both its scope and depth, and offers much
to contemplate. A complete account, we believe, will require a hybrid model in which alignment or imitation derives from both the
kinds of automatic processes they describe and processes that are
more directed or reflective. Hybrid models of this sort may be less
tidy (although not necessarily less mechanistic) than the one P&G
propose, but they do seem necessary to capture the subtlety and
richness of dialogic phenomena. We are reminded of an anecdote
about French President François Mitterand, who, when asked by
an acquaintance if she might address him using the personal tu
form, responded, “Si vous voulez.” Even in cooperative settings
without misunderstanding, alignment may be used strategically –
language is used in the pursuit of individual goals. An elaboration
of how a situation model incorporates key aspects of social and interpersonal dynamics would increase the explanatory power of a
mechanistic theory of dialogue.
36
BEHAVIORAL AND BRAIN SCIENCES (2004) 27:2
One alignment mechanism or many?
Arthur B. Markman, Kyungil Kim, Levi B. Larkey, Lisa
Narvaez, and C. Hunt Stilwell
Department of Psychology, University of Texas, 1 University Station, A8000,
Austin, TX 78712. [email protected]
[email protected]
[email protected]
[email protected]
[email protected]
http://www.psy.utexas.edu/psy/FACULTY/Markman/index.html
Abstract: Pickering & Garrod (P&G) suggest that communicators synchronize their processing at a number of linguistic levels. Whereas their
explanation suggests that representations are being compared across individuals, there must be some representation of all conversation participants
in each participant’s head. At the level of the situation model, it is important to maintain separate representations for each participant. At other levels, it seems less crucial to have a separate representation for each participant. This analysis suggests that different mechanisms may synchronize
representations at different linguistic levels.
Introduction. The core of Pickering & Garrod’s (P&G’s) article
is illustrated in their Figure 2. Participants in a conversation are
attempting to achieve alignment between linguistic representations at the phonetic, lexical, semantic, and discourse levels simultaneously. Two key questions arise from this figure. First, what
sorts of representations are being aligned; and second, how is this
alignment achieved?
What is being aligned? The target article’s Figure 2 illustrates
that linguistic representations at a variety of levels are aligned. The
authors suggest that representations from one participant’s head
are aligned directly with those from the other participant’s head.
This notation is a convenient shorthand, but people cannot directly access each other’s mental states. To make this model work
as it is drawn, participants must keep track separately of their own
knowledge at these levels, as well as the knowledge of other participants.
For some levels of discourse, this separation is more critical
than others. For the situation model, it is important that speakers
know what information is possessed by the listener. This information is crucial for ensuring that given new conventions are followed when generating sentences, for ensuring that new utterances are relevant to the discourse, and for maintaining common
ground (Clark 1996; Sperber & Wilson 1986). A key question is
the degree of information participants must have about their partner’s knowledge (Keysar 1994; Keysar et al. 1998).
At other linguistic levels, the distinction between one’s own
knowledge and that of a partner may be less crucial. It may not be
necessary to distinguish between one’s own grammatical constructions and those of a partner. Similarly, representations of
phonology and prosody of speech need not be kept separate in order to process a discourse. This distinction in the knowledge required to process and use language for the situation model as opposed to grammatical or phonetic representations suggests that
these levels may differ in the degree to which people are aware of
the effects of alignment. In particular, people may recognize that
they have designed their utterances to convey particular kinds of
information that their partner does not have. In contrast, they may
be unaware that the grammatical, phonological, or prosodic form
of one sentence has been influenced by the form of a sentence
spoken previously by a partner.
How are representations aligned? The target article refers to
the process that synchronizes participants’ linguistic representations as alignment. The discussion of the roles of linguistic representations in the previous section suggests that there may not be
a single alignment mechanism at work. In particular, the discourse-level representation (e.g., the situation model) is the only
one that really seems to require a separate representation of what
is known by each conversational participant. In contrast, the lower
levels of representation need not have distinct representations for
each participant.
When there are separate representations for each conversa-
Commentary/Pickering and Garrod: Toward a mechanistic psychology of dialogue
tional participant, a structural alignment process like the one
posited to be involved in analogical reasoning and similarity can
be used to help synchronize representations (Gentner 1983; Gentner & Markman 1997). The structural alignment process takes
pairs of relational representations and places them in correspondence. Structural alignment requires that the relations and objects
that are matched be seen as identical (or can be decomposed into
structures that are partially identical). Thus, this process provides
a method for creating semantic parallels between people’s situation models and their representations of conversational partners’
situation models.
One advantage of thinking about the alignment of situation
models as a structural alignment process is that there are several
established empirical benchmarks that are hallmarks of analogical
reasoning (Markman & Gentner 2000). It is possible to test for the
presence of these benchmarks in discourse-level processing. For
example, there is substantial interest in discourse on the role of inferences in comprehension (e.g., Graesser et al. 2001; McKoon &
Ratcliff 1992). The structural alignment process suggests that inferences are made on the basis of systematicity. Specifically, a
piece of information is inferred when there is a correspondence
between two representations and there is information in one representation that is connected to the correspondence that can be
carried over to the second representation (Clement & Gentner
1991; Markman 1997). Other information is not carried over as an
inference. Thus, this view makes a testable prediction about what
kinds of information are likely to be inferred by conversational
participants. Similarly, experiments could be devised to explore
the other published benchmarks.
It is not clear that the structural alignment process should be
involved at other linguistic levels. If people do not maintain separate representations for themselves and their partners for grammatical and phonetic processing, then some other process must be
involved in synchronizing these representations across individuals. The authors suggest one way this could occur. If the same representations are used for both production and comprehension at
the level of grammar and phonology, then these representations
will be synchronized by virtue of the nature of the language
process. Representations at these levels would be influenced by
factors such as priming (as suggested in the target article).
This analysis suggests that factors such as systematicity that
should affect the formation of situation models should not have an
influence on phonetic and grammatical processing. It would be interesting to carry out studies to explore whether systematicity and
other benchmark phenomena of structural alignment affect other
levels of linguistic processing. If they did, this would suggest that
people are maintaining separate representations of their own processing and that of their partners at a variety of linguistic levels.
Beyond linguistic alignment
Allan Mazur
Center for Environmental Policy and Administration, Maxwell School,
Syracuse University, Syracuse, NY 13244. [email protected]
Abstract: Dialogue requires ability beyond the production and comprehension of word strings. The interactive alignment account is good as far
as it goes, but it must be embedded in a broader model encompassing
alignment of paralinguistic representations.
Dialogue is not only the natural arena for language; it requires
abilities that are different from the production and comprehension of sentences and phrases. Children with the variant of autism
known as Asperger’s Syndrome have good sentence skills but cannot carry on a normal conversation because they have difficulty
with eye-to-eye gaze, facial expression, and body gestures that regulate the exchange.
Pickering & Garrod’s (P&G’s) ingenious account of interactive
alignment illustrates gains that can be had in broadening the study
of language from monologue to dialogue. I urge upon the authors
a further expansion, embedding spoken dialogue within a broader
model of human (and ape) turn-taking communication.
Conversations allow interlocutors to share meaning conveyed by
body postures, facial cues, and intonation – apart from the particular strings of words that are used. Sometimes the words spoken
are irrelevant to the primary representation that is to be aligned. A
conversation between two knowledgeable people, ostensibly about
Chaucer, may be more importantly a dominance contest, or a seduction, or bonding chatter between friends – though words manifestly about these meanings may never be exchanged.
We often recognize the relationship between conversing foreigners and the general intent of their communication, even
though we may not understand their language.
As long as actors can see one another, they can communicate
with remarkable efficiency even when distance or a barrier precludes speech. I recall standing with a handsome male colleague
on a busy street corner. He whispered aside to me that he was visually flirting with a woman seated on a bus some distance away;
he predicted (correctly) that as the bus drove off, she would turn
toward him with a broad smile.
Paralinguistic signaling is especially effective in aligning social
relationships within a dyad. This enables Ego and Alter to agree
on who is the leader and who the follower; to communicate affiliation, affection, or hostility; and to sympathize emotionally. Consider, as an example, how dialogue can establish the relative status
of interlocutors, independently of the words that are used.
Spoken or nonverbal “dialogue” is governed by mutually understood rules (Mazur 1985). Some rules are asymmetrical, specifying different actions for a high-status actor and for a low-status
one. Two asymmetrical rules are:
1. The high-status person sets the pace and mood of the conversation, and the low-status person follows. Pace and mood may
be set with smiles, jokes, frowns, exclamations, and volume, rapidity, or intonation of speech. If Ego tells a loud joke, Alter can
deferentially comply with a loud laugh or can challenge by substituting an inappropriate response.
2. The high-status person introduces and terminates major topics of conversation. This rule, like the previous one, indicates that
the high-status person can take control of the conversation, which
is the essence of having high status. If both interlocutors attempt
to set the conversational agenda, there is a dominance contest.
Other rules are symmetrical, applying without regard for the status of the actor. It is the violation of symmetrical rules that signals
a dominant act, whereas strict conformity to them signals deference or politeness. Important symmetrical rules are:
3. If one individual is speaking, the other should remain quiet.
If Ego interrupts Alter’s speech, Ego has acted dominantly.
4. A listener who is offered the floor should speak. A speaker
can pass the floor by asking a question of the listener, or by directing his eyes to the listener after concluding a speech. If Ego
remains silent after Alter offers the floor, Ego has acted dominantly.
5. Do not look into another individual’s eyes when no one is
speaking (unless in a romantic context). The violation of this rule,
silent staring, is a common dominant act among primates, whereas
rule-following eye aversion indicates deference.
6. Look at the speaker’s face, especially if the speaker is looking
at you. To look away, suggesting inattention, is hard to do if you
respect the person speaking to you. If the speaker is of minor consequence, it is easy to violate the rule, thus showing your dominance. (This rule is inoperative when averted eyes overtly signal
submission, as in looking down while being scolded.)
7. Do not speak loudly, sternly, or angrily. Shouting matches
and arguments are obviously dominance contests.
8. The speaker should direct the listener’s actions by request
rather than command, and should avoid a stern or stubborn tone.
To speak in a commanding or inflexible way implies that the listener is of lower status.
BEHAVIORAL AND BRAIN SCIENCES (2004) 27:2
37
Commentary/Pickering and Garrod: Toward a mechanistic psychology of dialogue
These rules operate within a context of linguistic interaction.
The words spoken may or may not be an important part of the
whole display. When we deferentially compliment someone, we
speak in strict accordance with the rules, whereas our verbal insults gain emphasis when violating the rules of dialogue. We accompany our speech with appropriate gestures, perhaps glaring
for dominance or smiling for deference. This full array of actions
– words, gestures, and rules – constitutes the status display.
A natural model of conversation must go beyond the interactive
alignment of word strings. P&G have briskly stepped onto the
road of dialogue. I hope they soon take another step forward.
Correspondences between the interactive
alignment account and Skinner’s in
Verbal Behavior
Joseph J. Pear
Department of Psychology, University of Manitoba, Winnipeg, Manitoba R3T
2N2, Canada. [email protected]
http://www.umanitoba.ca/cgi-bin/psychology/hpg_main.cgi?data/
pear.txt
Abstract: Pickering & Garrod’s (P&G’s) interactive alignment account
corresponds directly with the account Skinner (1957) gave in his book Verbal Behavior. This correspondence becomes evident when “properties of
verbal stimuli” substitutes for “channels of alignment.” Skinner’s account
appears to have the dual advantages of requiring fewer basic terms and integrating the field of verbal behavior with the whole field of human behavior.
There are strong correspondences between Pickering & Garrod’s
(P&G’s) interactive alignment (IA) account and Skinner’s (1957)
verbal behavior (VB) account. Similar to P&G’s assumption that
dialogue is the basic form of language processing, Skinner took the
interaction between speaker and listener as fundamental (see
Skinner 1957, Figures 1– 6, pp. 38, 39, 57, 84, 85). I explain fur-
ther correspondences with reference to Table 1 below, which is
based on Table 2 in the target article.
Regarding row 1, the situation models in the IA account correspond in the VB account with variables controlling each interlocutor’s behavior, some of which may be private (Skinner 1957,
pp. 130–146; also see Skinner 1953, Ch. 17). The alignment channels in the IA account correspond with properties of verbal stimuli. Because of reinforcement of many different instances of
echoic behavior, which is the direct imitation of the properties of
verbal stimuli, echoic behavior generalizes widely (Skinner 1957,
pp. 55–56), just as the tendency to align generalizes across alignment channels in the IA account.
Regarding row 2, sustained dialogue occurs only if the interlocutors have implicit common ground in the IA account or emit
similar verbal behavior in the VB account. In the VB account, effective verbal interaction occurs between the extremes of identical and completely dissimilar verbal behavior (Skinner 1957,
pp. 271–272). In the IA account, there is a repair mechanism for
preventing dialogue from breaking down or for restoring it if it
does. In the VB account there are several reinforced responsestrengthening techniques that maintain dialogue, such as prompts
and probes (Skinner 1957, Ch. 10, pp. 253–292; see also p. 58 for
examples of clarification and expansion requests involving echoic
behavior).
What in the IA account is called full common ground – a complex model shared by both interlocutors – is not common in ordinary dialogue according to either account. Both accounts maintain
that constructing and using these models are derived (e.g.,
learned) abilities or behaviors.
Both accounts also agree that we obtain a distorted view of
grammar by looking at it only in monologue. In the VB account,
grammar is the expression of responses called autoclitics that
modify the effect on the listener of responses that Skinner (1957,
p. 312) called “the raw material out of which sustained verbal behavior is manufactured.” When listeners cannot frequently or immediately prompt or probe the speaker’s verbal behavior, there is
often pressure on speakers to incorporate a high density of autoclitics according to constraints embodied in formal rules of gram-
Table 1 (Pear). Correspondences between the interactive alignment account of language processing in dialogue
and the verbal behavior account
Interactive alignment account (IA)
Verbal behavior account (VB)
1. Linkage between interlocutors
Links across multiple levels of representation via
“alignment channels.” Sound comes to encode
words, linguistic information, and aspects of
situational models.
2. Inference
Externalized in the interaction between
interlocutors via a basic interactive repair mechanism.
3. Routines
Arise out of the application of the interactive alignment
in specific situations. A high proportion of dialogue
contains these units, which facilitate both production
and comprehension.
4. Self-monitoring
Monitoring occurs at any level of representation that is
subject to alignment as a consequence of the account.
5. Repair mechanisms
The same basic repair mechanism for self-repair and
other-repair.
1. Linkage between interlocutors
Control across multiple properties of verbal stimuli.
Sound comes to bring the listener’s behavior under
the control of variables controlling the speaker’s behavior.
38
BEHAVIORAL AND BRAIN SCIENCES (2004) 27:2
2. Inference
Externalized in the interaction between interlocutors
via reinforced techniques of strengthening verbal behavior.
3. Routines
Functional verbal units that are conditioned or strengthened
process. A high proportion of dialogue uses routines, which
simplify both production and comprehension.
4. Self-monitoring
Occurs for all aspects of verbal behavior because speakers
typically are also listeners.
5. Repair mechanisms
The same basic principles apply for strengthening one’s own verbal behavior and strengthening that of others in specific situations.
Commentary/Pickering and Garrod: Toward a mechanistic psychology of dialogue
mar. When these pressures are weak, as in much of dialogue as
well as certain literary styles (Skinner 1957, p. 356), verbal behavior is more fluid. Skinner likely would have endorsed the call in
the last sentence of the target article for “a more flexible account
of grammar capable of capturing linguistic constraints on linked
sentence fragments.”
Regarding row 3, a routine in the IA account appears to be, in
the VB account, a functional verbal unit (Skinner 1957, pp. 21,
116) that has been conditioned or strengthened in a specific situation. In the IA account, repetition of “the previous speaker’s utterance” appears to be important in this process (sect. 5.1 of target article). According to Skinner, “a verbal response of a given
form sometimes seems to pass easily from one type of operant to
another” (Skinner 1957, p. 188). Hence, a response emitted as an
instance of echoic behavior may simultaneously or subsequently
appear in other categories of verbal behavior (for examples, see
Skinner 1957, pp. 188–189, 360 – 362).
Like routines in the IA account, functional verbal units in the
VB account may be larger than a single word. Similar to the
process of routinization in the IA account (sect. 5.2 of the target
article), a process called composition in the VB account generates
large verbal patterns that can come to function as units. Skinner
(1957, Ch. 14, pp. 344 – 367) proposed that composition consists
primarily of adding autoclitics (including ordering) to the raw verbal material mentioned above. “Formal evidence alone will not
show whether sentences [or other large segments of verbal behavior] have been composed” (Skinner 1957, p. 346) as opposed
to being emitted as units. Once a composed utterance has been
reinforced several times, it may begin to function as a unit. Given
the right conditions, however, a unit may break into smaller units
(Skinner 1957, pp. 116 –117). This seems very close to the dynamics of routines as described in the IA account.
Regarding row 4, in the VB account an “important fact about
verbal behavior is that speaker and listener may reside within the
same skin” (Skinner 1957, p. 163). As in the IA account, in the VB
account speakers monitor their own verbal behavior and edit it after, during, or even before its emission (Skinner 1957, Ch. 15, 16,
pp. 369–402). Both accounts agree that there is no difference in
principle between self- and other-monitoring. Both accounts also
agree that when interacting with oneself, the stimuli need not be
in the form of an external medium.
Regarding row 5, the techniques for strengthening one’s own
weak verbal behavior are in principle the same as those for
strengthening another’s verbal behavior (Skinner 1957, Ch. 17,
pp. 403–417). These include manipulating stimuli and changing
the level of editing.
Given the correspondences between the two accounts, it may
be impossible to distinguish them empirically. However, the VB
account appears to require fewer terms, “and the terms created
are derived from a few prior technical terms common to the whole
field of human behavior” (Skinner 1957, p. 456).
Putting the interaction back into dialogue
Emanuel A. Schegloff
Department of Sociology, University of California at Los Angeles, Los
Angeles, CA 90095-1551. [email protected]
http://www.sscnet.ucla.edu/soc/faculty/schegloff/
Abstract: I share the authors’ stance on the dialogic or interactional character of language. The authors, however, have left actual interaction out of
their conception of dialogue. I sketch a number of organizations of practices of talking and understanding that supply the basic arena for talk-ininteraction. It is by reference to these that mechanisms for speech production and understanding need to be understood.
I write as a conversation analyst. I have spent nearly 40 years
studying the sorts of events which Pickering & Garrod (P&G) now
take to be the fundamental premise of language. I am, of course,
predisposed to take the same view. Indeed I have already done so
in a number of publications (Schegloff 1979; 1989; 1996a, inter
alia). The arguments of the target article aside, in the choice between a fundamentally monologic, “individualist” stance and a dialogic, interactional one, there are compelling reasons for preferring the latter. For now, one line will have to do.
For most humans on the planet since the species developed
“language,” the overwhelmingly most common ecological niche
for its use has been (1) the turn at talk, (2) as part of a coherent
sequence of turns, (3) through which a course or trajectory of action is jointly pursued by some or all of the participants (not necessarily cooperatively, but jointly), (4) in an episode of interaction,
(5) between two or more persons, (6) organized into two or more
parties, (7) the occasion of interaction being composed of one or
more such episodes. If that is where language as a publicly deployed resource and utility resides, it is plausible to expect that it
has been designed and fashioned by its users and uses in a manner adapted to the contingencies of its “environment” – that is, by
the contingencies of talk-in-interaction (of which the foregoing
are but several aspects) and its virtually omnipresent bodily companions – gesture, posture, gaze deployment, facial expression,
and so on. Such an expectation is not merely plausible; detailed
and repeated examination of recorded episodes of naturally occurring talk-in-interaction shows it to be so – indeed, at a thoroughly implausible (and yet demonstrable) level of detail. The
mechanisms of production and comprehension being addressed
by P&G need to be understood in this context. It is this context
that is missing from P&G’s treatment.
By “context” here I do not mean the ordinary characterizations
of settings as domestic or public, intimate or formal, and others
drawing on different genres of social and cultural diction (important as these may be). I mean the various organizations of practice
that deal with the various generic organizational contingencies of
interaction without which it cannot proceed in an orderly way: (1)
The “turn-taking” problem: Who should talk next and when
should they do so? How does this affect the construction and understanding of the turns themselves? (2) The “sequence-organizational” problem: How are successive turns formed up to be “coherent” with the prior turn (or some prior turn), and what is the
nature of that coherence? (3) The “trouble” problem: How should
one deal with trouble in speaking, hearing, and/or understanding
the talk such that the interaction does not freeze in place, that intersubjectivity is maintained or restored, and that the turn and sequence and activity can progress to possible completion? (4) The
word selection problem: How do the components that get selected
as the elements of a turn get selected, and how does that selection
inform and shape the understanding achieved by the turn’s recipients? (5) The overall structural organization problem: How does
the overall structural organization of an occasion of interaction get
structured, what are those structures, and how does placement in
the overall structure inform the construction and understanding
of the talk as turns, as sequences, and so on?
The organizations of practice addressed to these issues – turn
organization (Goodwin 1979; Schegloff 1996a), turn-taking organization (Jefferson 1986; Sacks et al. 1974; Schegloff 1987a;
2000a; 2001), sequence organization (Schegloff 1990; 1995; forthcoming), the organization of repair (Drew 1997; Jefferson 1974;
1987; Schegloff 1979; 1987b; 1991; 1992; 1997a; 1997b; 2000b;
Schegloff et al. 1977), the organization of word selection (Sacks
1972a; 1972b; 1992; Sacks & Schegloff 1979; Schegloff 1972;
1996b), overall structural organization (Schegloff 1986; Schegloff
& Sacks 1973), and others – constitute, in the options that they
shape and the practices made available, a spate of interaction recognizable as “conversation,” as “interview,” as “meeting,” as “lecturing,” as “giving a speech,” as “interrogation,” and so on. These
are what we call “speech-exchange systems” (Sacks et al. 1974,
pp. 729–731), and can be seen as particular, here-and-now-withthese-participants instances of these.
What makes an interaction is not just the juxtaposition of bodBEHAVIORAL AND BRAIN SCIENCES (2004) 27:2
39
Commentary/Pickering and Garrod: Toward a mechanistic psychology of dialogue
ies. What mediates and organizes the conduct of the parties is not
a structureless, featureless, transparent medium. The composition of a turn at talk – whether it is made up of one or more component units; whether these are sentences or sub-sentential – its
syntactic construction and choice of lexicon are shaped in part by
the contingencies of turn production imposed by a turn-taking organization that will have others empowered or required or allowed
to talk next, at points in the turn’s development not wholly under
the speaker’s control. Particular courses of action implemented
through turns at talk (such as request sequences, complaint sequences, storytelling sequences, news-conveying sequences, etc.)
implicate certain ways of understanding what is being said that
render meaningful and consequential selection between apparently equivalent expressions, the delay of a turn’s start by twotenths of a second or less, and the like. How one says what one says
can depend on who the other is; and, of all the persons and categories which could be used to characterize “the other,” depend on
which ones have been made relevant at that moment in the talk,
or can be made relevant by constructing the same “sayable” in this
way or that. And so on.
A very high proportion of the matters discussed by P&G as if
they were unrelated to anything but the mechanisms the authors
are concerned to develop, are not interactionally random. They
are part of the fabric of some organization of practices for talk-ininteraction. Many of them have been given quite detailed and systematic treatment in the literature – things like “routines” (target
article, sect. 5.2.1, cf. Schegloff 1986) and “how are you” routines
in particular (Jefferson 1980; Sacks 1975); things like “joint constructions” (sect. 7.1, para. 3; cf. Lerner 1991; 1996; Sacks 1992,
vol. I, pp. 144–147 et passim); things like “non-sentential turns”
(sect. 7.1, para.6; cf. Sacks et al. 1974; Schegloff 1996a): things like
“monitoring during overlapping speech (sect. 6, para. 6; cf. Schegloff 2000a; 2001); and so on and on.
Most striking is P&G’s treatment of “repair”; the discussion
rests on a terminology (“repair,” “other-repair,” “self-repair”) which
they neither explicate nor cite but the latter two of which they treat
as discrete sets of things, not an organization of practices. This
leads them – incorrectly, in my view – to treat the basic mechanisms of self-repair and other-repair as the same (see Table 2 of the
target article) when, interactionally speaking, they are not the same
in either execution or interactional import (Schegloff 1979,
pp. 267–69; Schegloff et al. 1977, inter alia). I believe the analysis
of talk-in-interaction along such lines has much to contribute not
only to our understanding of the mechanisms addressed by P&G,
but to work in the neurobiology of behavior more generally – precisely the remit of this journal. But that is another matter.
Some notes on priming, alignment,
and self-monitoring
Niels O. Schillera,b and Jan Peter de Ruiterb
aMaastricht University, Faculty of Psychology, Department of Cognitive
Neurocognition, 6200 MD Maastricht, The Netherlands; bMax Planck Institute
for Psycholinguistics, 6500 AH Nijmegen, The Netherlands.
[email protected]
[email protected]
http://www.psychology.unimaas.nl/Base/Personal/Niels/Schiller.htm
http://www.mpi.nl/world/persons/profession/schiller.html
Abstract: Any complete theory of speaking must take the dialogical function of language use into account. Pickering & Garrod (P&G) make some
progress on this point. However, we question whether their interactive
alignment model is the optimal approach. In this commentary, we specifically criticize (1) their notion of alignment being implemented through
priming, and (2) their claim that self-monitoring can occur at all levels of
linguistic representation.
The primary way of language use is dialogue, not monologue. We
want to acknowledge the authors’ effort to stress this important
40
BEHAVIORAL AND BRAIN SCIENCES (2004) 27:2
point, which needs to be addressed explicitly in empirical and
modeling work in speech production and comprehension research. We believe that these issues are especially relevant for syntactic processing. For instance, one wonders how syntactically incomplete (dialogue) utterances can be syntactically encoded in
more traditional models, if there is no overt verb present in the
generated utterance. Take, for example, the following extract from
the dialogue transcript in section 2 of the target article:
1——B: . . . Tell me where you are?
[Utterances 2 and 3 omitted]
4——A: Right: {I am} two along from the bottom one up:* [our addition in curly brackets]
In this example, speaker A does not produce the appropriate verb
form of “to be” (i.e., “I am”) but nevertheless gives an acceptable
and cooperative answer to speaker B’s question. This type of ellipsis can only be correctly produced if the syntax generator has
access to previously stored discourse information, allowing the
speaker to omit “I am,” even though the original question containing the verb occurred several utterances earlier in the discourse (see also Levelt 1989, p. 89, for a similar analysis).
Although we agree in principle with the authors’ assessment
that the dialogical structure of language should receive more attention in accounts of language processing, we are not convinced
that adopting the interactive alignment model is the right way to
do so. For instance, it is unclear to us exactly how priming can account for alignment, and, in particular, we fail to see in what way
priming is more than “a behavioral effect” (see target article, sect.
2.2). We believe that “priming” does not explain or implement interactive alignment. Real interactive alignment necessarily involves storing selected fragments from previous utterances. Priming can raise the probability of certain linguistic structures being
selected, but this is not sufficient for the strong and explicit type
of alignment the authors want to incorporate in models of language processing. Also, syntactic priming effects are weak effects.
It is hard to see how an elaborate mechanism such as interactive
alignment could be realized by only raising the probability of selecting a certain syntactic construct by roughly 10% to 20% (see,
e.g., Pickering & Branigan 1998).
Our second critical note concerns one of the few testable predictions from the interactive alignment model, namely, that selfmonitoring by the speaker occurs at all levels of linguistic representation (see sect. 6). While other researchers (e.g., Wheeldon &
Levelt 1995) have claimed that internal self-monitoring works on
abstract phonological form representations, Pickering & Garrod
(P&G) propose that self-monitoring can occur at any level of linguistic representation that can be aligned (i.e., semantic, syntactic, lexical, phonological, and phonetic representations) – and not
only at the phonological level.
For example, the authors explicitly claim that speakers can correct gender errors, such as le tête instead of la tête (“the head”) in
French or de been instead of het been (“the leg”) in Dutch not only
after they have been articulated but even before their overt production. This is an interesting claim that needs to be investigated
in the future. However, we are somewhat skeptical about this
claim because to our knowledge there is no evidence that selfmonitoring of gender features (or any other syntactic features) is
possible. For example, Desrochers and his collaborators (Desrochers & Paivio 1990; Desrochers et al. 1989; Muller-Gass et al.
2000) found that selecting a gender label (e.g., feminine or masculine) took about 200 msec longer than selecting the indefinite
article in French gender decision. Furthermore, Tucker et al.
(1977) provided empirical evidence suggesting that French speakers implicitly construct a noun phrase including the article and the
noun to determine a noun’s gender. However, if speakers can selfmonitor abstract gender information at the level of syntactic representation, as suggested by P&G, why would they go through the
trouble of generating the gender-marked article of a noun to determine its gender?
In contrast to these findings about syntactic representations, re-
Commentary/Pickering and Garrod: Toward a mechanistic psychology of dialogue
cent evidence from our own laboratory as well as from other laboratories demonstrated that self-monitoring does occur at the
level of phonological encoding. We have empirical data about the
monitoring of phonological segments (Schiller, in press; Wheeldon & Morgan 2002), word stress (Schiller 2001; Schiller et al., in
press), syllable boundaries ( Jansma & Schiller 2004), and syllables
(Morgan & Wheeldon 2003). However, we also have evidence that
participants are unlikely to monitor a phonetic-acoustic representation of the respective utterances. Although gender decision as a
task is widely used (Müller & Hagoort 2001; Schiller et al. 2003;
Schmitt et al. 2001a; Schmitt et al. 2001b; Van Turennout et al.
1998), it remains to be shown whether or not this task actually taps
syntactic processing, because abstract gender information may not
be directly available to the speaker. Rather, gender information
may be available only via its phonological realization, for example,
an article or gender-marking suffix. Interestingly, effects of gender congruency have recently been re-interpreted as determiner
congruency effects occurring at the phonological level and not at
the gender feature level (see overview in Schiller & Caramazza
2003).
To conclude, we believe that internal self-monitoring does not
occur at every single level of linguistic representation (as claimed
by P&G) or at every processing level in models of speech production (e.g., Levelt et al. 1999). Rather, there is abundant evidence
that internal self-monitoring works on phonological representations, which are created during phonological encoding in speaking, for example, when segments are prosodified into phonological words. It is at this level that information about segments,
syllables, syllable boundaries, and word stress is available to the
speaker. Although the abundance of evidence for phonologicallevel monitoring does not necessarily exclude other monitoring
levels (e.g., at the conceptual level; see Levelt 1989), we are unaware of any evidence for self-monitoring at earlier or later levels
of grammatical and form encoding.
ACKNOWLEDGMENTS
Niels O. Schiller is supported by a grant from the Netherlands Organization for Scientific Research (NWO). Jan Peter de Ruiter is supported by
the European Union Project Conversational Multi-modal Interaction with
Computers (COMIC), grant IST-2001-32311. The authors thank Pim
Levelt for his comments on an earlier draft.
Just how aligned are interlocutors’
representations?
Michael F. Schober
Department of Psychology, F330, New School for Social Research, New
York, NY 10003. [email protected]
http://homepage.newschool.edu/~schober
Abstract: Conversational partners’ representations may be less aligned
than they appear even when interlocutors believe they have successfully
understood each other, as data from a series of experiments on surveys
about facts and behaviors suggest. Although the goal of a mechanistic psychology of dialogue is laudable, the ultimate model is likely to require far
greater specification of individual and contextual variability.
When conversational partners come to agree that they are talking
about the same thing, how aligned are their conceptualizations?
The interactive alignment account holds that when dialogue is
successful, interlocutors’ linguistic representations are aligned at
multiple levels. Although Pickering & Garrod (P&G) observe that
alignment is sometimes only approximate and that evident misalignments can be interactively repaired, the general thrust of
their approach is that successful communication requires representations to be the same.
I contend that interlocutors’ using the same words can actually
mask a surprising degree of undetected misalignment. Fred Con-
rad and I, with other colleagues, have carried out a series of laboratory and field studies examining how people interpret ordinary
words in ongoing U.S. government surveys about facts and behaviors, words like “job,” “bedroom,” “smoking,” and “cigarettes”
(Conrad & Schober 2000; Schober & Conrad 1997; Schober et al.
2004; Suessbrick et al. 2000). Because the agencies that carry out
these surveys have thorough definitions for the terms, answers to
the survey questions provide evidence about the extent to which
respondents’ conceptions match the survey designers’. Our finding is that people’s representations are frequently wildly misaligned with the survey designers’ – and with each other’s – without anyone’s noticing.
For example, in one study (Suessbrick et al. 2000), survey respondents interpreted terms like “smoking” and “cigarettes” in a
question such as “have you smoked at least one hundred cigarettes
in your entire life?” differently enough (tobacco, cloves, marijuana? Finished or just a puff? Bought or borrowed?) that 10%
of the respondents subsequently presented with a definition
changed their answer to the question from yes to no or from no to
yes. In a national telephone sample (Conrad & Schober 2000),
more than 40% of reported purchases did not fit the survey designers’ definitions, even though the questions had been widely
pretested. And this is not just because the official definitions failed
to match the population consensus about the meaning of terms;
respondents’ interpretations differed from each other’s as much as
they differed from the survey designers’.
Across our various studies, respondents are quite surprised at
the thought that someone else might interpret the same words differently from the way they do; when given the opportunity to request clarification about the meanings of survey terms, they
choose to do it a very small percentage of the time. People seem
to follow a “presumption of interpretability” (Clark & Schober
1991): It should be the questioner’s responsibility to forestall misinterpretation.
These data suggest a far more Quinian view of successful referring than the P&G account encompasses: Seemingly successful referring can mask conceptual misalignments that reflect deep
underlying indeterminacies. The point is that people can believe
they have understood each other well enough for current purposes (as proposed in Clark & Wilkes-Gibbs 1986) and yet never
actually discover that their conceptions were misaligned.
An important contention in the P&G article is that seemingly
complex interactional processes can be modeled largely with simple individual mechanistic processes. The proposal is that conversational partners, following a principle of parsimony, do not ordinarily model each other’s mental states or make inferences about
common ground, except when there is evidence that not doing so
has led to obvious misunderstanding and when cognitive resources allow.
I would argue that none of the current data actually allow us to
distinguish this position from an alternative: that the ordinary case
is that conversational partners do model each other, and that they
fail to do so only when they are under heavy cognitive load or when
circumstances weigh heavily against doing so. Why should we assume that the ordinary case is one where the interlocutor does not
need to be modeled and the speaker is under heavy cognitive
load? As far as I can tell, no one knows the level of load encountered in the range of ordinary interactive situations. As Susan
Brennan and I have argued (Schober & Brennan 2003), the evidence for egocentric processing is far from conclusive; the experiments purported to show egocentric processing as basic rely on
null results and experimental methods that are far removed from
ordinary processing situations. When such studies are carried out
in more realistic settings, the findings can look rather less egocentric.
Not to overstate the case, but one could argue that modeling
one’s partner only when it is needed may require a level of situational monitoring that leads to a paradox: How can one know exactly when one needs to model one’s partner without already
knowing what the partner model would clarify? (This seems akin
BEHAVIORAL AND BRAIN SCIENCES (2004) 27:2
41
Commentary/Pickering and Garrod: Toward a mechanistic psychology of dialogue
to Fodor’s frame problem or the notorious conundrums about selective attentional filtering). Arguably, it might be simpler to
model the partner as the ordinary case. P&G assume that representing common ground requires unusually complex mental representations; this may be so, but it is also possible that representations of common ground are ordinary memory representations
with the right content.
In fact, what is parsimonious may well vary across individuals
and conversational contexts. Our notions of what is effortful in dialogue are not well worked out (Schober 1995; 1998a); people’s
knowledge of each other’s perspectives, interest in modeling their
partners, and abilities for doing so probably vary substantially
more than current theories allow. For example, people who have
very poor mental rotation abilities may be unable to conceive of a
partner’s point of view on a scene, leading to a different sort of
partner modeling than that carried out by mental rotation whizzes
(Schober 1998b). And different abilities may be relevant in different conversational contexts; mental rotation ability may have little
to do with the consideration of lexical alternatives in dialogues that
are not about physical scenes, whereas working memory capacity
differences or different conversational agendas (Russell &
Schober 1999) might. P&G are to be applauded for furthering the
attempts to bridge the “language as product” and “language as action” traditions, but they have also opened a can of worms; the ultimate mechanistic account of dialogue is likely to require far
greater specification of individual and contextual variability.
Dialogue processing: Automatic alignment or
controlled understanding?
Hadas Shintel and Howard C. Nusbaum
Department of Psychology, The University of Chicago, Chicago, Illinois
60637. [email protected]
[email protected]
Abstract: Pickering & Garrod’s (P&G’s) mechanistic account of dialogue
assumes that linguistic alignment between interlocutors takes place automatically, without using cognitive resources. However, even the most basic processes of speech perception depend on resource use. The lack of invariant mapping between input patterns and interpretations in dialogue,
as in speech perception, may require controlled, rather than automatic,
processing.
In their target article, Pickering & Garrod (P&G) challenge current psycholinguistic theory by proposing the study of dialogue as
the appropriate paradigm for understanding language processing.
They argue that the main problem facing interlocutors is the alignment of their respective situation models, and they propose a
mechanistic account of alignment in dialogue, based on an interactive and resource-free priming mechanism. Linguistic structures produced on the surface by one interlocutor putatively
prime corresponding structures in the other interlocutor. While
we agree that conversation, rather than decontextualized, isolated
sentences, should be taken as basic form of language use and as
the foundation upon which language has evolved, it seems less
plausible that much of the cognitive work in conversation takes
place automatically.
P&G’s model of automatic alignment assumes that a mapping
can be established directly between the representations uttered
by one interlocutor and those activated in the head of the other
interlocutor. But as they themselves note, the linguistic information is encoded in sound. Hence, for the listener the spoken utterance must be recognized antecedent to any alignment occurring. However, even findings regarding the basic processes of
speech perception argue against the assumption of automatic processing. Automaticity implies a passive process in which the input
is processed in an invariant, inflexible manner, regardless of the
beliefs and expectations of the listener (Nusbaum & Schwab
1986). For a process to be automatized, there must be a consis-
42
BEHAVIORAL AND BRAIN SCIENCES (2004) 27:2
tent mapping between input patterns and responses, the benefit
of which is a process that does not impose demands on cognitive
and attentional resources (Shiffrin & Schneider 1977).
Of course the hallmark of speech is the lack of consistent, invariant mapping between acoustic patterns and linguistic categories across phonetic contexts (Liberman et al. 1967) and across
talkers (Peterson & Barney 1952). As a result of variable mapping,
speech perception depends on controlled, active processing requiring attention and working memory (Nusbaum & Magnuson
1997). For example, variation between talkers (as in the circumstance of a conversation among three interlocutors) requires
talker normalization, the process by which listeners compensate
for acoustic-phonetic variability in vocal characteristics (see, e.g.,
Nearey 1989). Nusbaum and Morin (1992) found that talker variability slowed speech recognition (see also Mullennix & Pisoni
1990), and this slowing was because of increased demand on working memory.
Furthermore, talker normalization is contingent on listeners’
expectations regarding the interpretation of acoustic patterns –
which should not happen for an automatic direct-mapping
process. Magnuson and Nusbaum (1994) demonstrated that when
listeners expected a pitch difference to signal a talker difference,
they showed talker normalization, but if the same pitch difference
was expected as a way one talker accented speech, no normalization occurred. Similarly, the expectation that a talker was male or
female significantly changed the interpretation of vowel tokens
(Johnson et al. 1999). The acoustic patterns of speech are
processed differently depending on listeners’ expectations, arguing against invariant automatic processing. Expectation effects
suggest that alignment may not be a result of a direct, automatic,
causal link between the activation of a representation in one interlocutor and the activation of an objectively matching representation in the other interlocutor. Rather, the process is mediated by
listeners’ expectations. For example, alignment at the level of articulation, where interlocutors converge on each other’s speaking
style, may be mediated by interlocutors’ stereotypic expectations
about the other interlocutor’s accent and speech rate, resulting in
subjective but not objective alignment (Thakerar et al. 1982).
Listeners may need to use controlled active processing because
the one-to-many mapping (one pattern may have multiple interpretations) in speech represents a nondeterministic computational problem that cannot be solved, in principle, by a deterministic system such as an automatic process (cf. Nusbaum &
Manguson 1997). This problem of one-to-many relationships between linguistic patterns and interpretations occurs across all levels of linguistic analysis (Nusbaum & Henly 1989). Indeed, research has shown that the same spoken sentence can be differently
interpreted in different visual referential contexts (Tanenhaus et
al. 1995) and that the same indirect requests can be processed differently as a result of speaker status (Holtgraves 1994). Although
P&G focused on utterance-level matches that can occur within
stretches of dialogue, real conversation is less predictable and routinized – otherwise communication would be unnecessary.
This raises several deep questions facing dialogue research:
How do interlocutors cope with variability in linguistic and perceptual context? How are diverse sources of information integrated to constrain production and comprehension? And how do
interlocutors flexibly adapt to different conversational circumstances? Although in real conversations these questions are often
subjectively resolved quickly and without apparent effort, this is
seldom an accurate barometer of the demands on cognitive resources such as attention and working memory.
P&G propose that the process of automatic alignment bypasses
the need for modeling the interlocutor’s mental states and for distinct, conscious decision stages. We agree that a constant monitoring of common ground would be unnecessary and costly, but
the dichotomy between automatic processes and conscious decision processes involving complex inferences about the interlocutor’s mental state does not represent the full range of possible processing alternatives. In clarifying this point, it is worth considering
Commentary/Pickering and Garrod: Toward a mechanistic psychology of dialogue
what is meant by automatic, as distinct from controlled, processing. Automatic processes have been defined by different criteria
such as being unintentional, occurring outside of conscious awareness, not requiring cognitive resources, and being autonomous;
but these criteria do not necessarily hold simultaneously (Bargh
1989). Speech perception involves controlled active processes that
are not resource-free or autonomous, yet they do occur largely
outside of awareness, do not require a conscious intention on the
part of the listener, and are subjectively experienced as effortless.
Likewise, the subjective ease and speed that subjectively characterize language use in dialogue may not reflect the complexity of
the underlying processing. The variability and flexibility shown in
the processes of language comprehension and production in dialogue call for dynamic adaptation rather than a passive automatic
mechanism.
Top-down influences in the interactive
alignment model: The power of the
situation model
Tessa Warrena and Keith Raynerb
a607 LRDC (Learning, Research, and Development Center) and Department
of Psychology, University of Pittsburgh, Pittsburgh, PA 15260; bDepartment
of Psychology, University of Massachusetts, Amherst, Amherst, MA 01003.
[email protected]
[email protected]
Abstract: Pickering & Garrod’s (P&G’s) model is an innovative and important step in the study of naturalistic language. However, the simplicity
of its mechanisms for dialogue coordination may be overstated and the hypothesized direct priming channel between interlocutors’ situation models is questionable. A complete specification of the model will require
more investigation of the role of top-down inhibition among representations.
Pickering & Garrod’s (P&G’s) new model of linguistic interaction
in dialogue is an important contribution to the study of psycholinguistics. This model breaks new ground between two traditionally
disparate areas of language research and combines the mechanistic detail characteristic of sentence processing research with the
emphasis on language as a cooperative process characteristic of dialogue research.
The central mechanism of P&G’s interactive alignment model
is a process they term “alignment,” whereby dialogue participants’
linguistic and discourse representations become more similar over
the course of an interaction. Alignment is an automatic process
that results from priming between linguistic representations.
Priming can occur among different levels of representation within
a single individual, as when increased lexical or semantic overlap
between a prime and a target causes increased syntactic priming,
or between the representations of different individuals, as when a
speaker who has just comprehended a particular syntactic form
produces the same form in a subsequent utterance. The resourcefree characterization of the system relies on the assumption that
alignment at one level of representation increases alignment at
other levels of representation, and therefore essentially complete
alignment can be obtained through simple and automatic
processes.
Though the strength of this model is in its elegance and simplicity, it is not clear whether these characteristics will survive a
more explicit specification of the system. According to the characterization of the model as a network of linked representations
where priming and alignment in one representation causes increased alignment in all others, alignment at the level of the situation model is simply an epiphenomenon of alignment at lower
levels. Yet P&G acknowledge that in some cases priming at one
level will decrease alignment at other levels and specifically note
that alignment of the situation model takes priority over alignment
at other levels of representation. This complicates the system, as
lower-level priming is constrained by the very alignment it drives.
In fact, lower-level priming, supposedly a driving force in alignment, can be inhibited by extremely subtle nuances of local context. For example, P&G discuss Garrod and Anderson’s (1987)
findings that participants used different words for horizontal
groupings of boxes depending on whether the grouping was modified with an ordinal adjective or not. So participants who used the
phrase “the second row” later spoke of “the bottom line,” even
though “the bottom row” was an acceptable, unambiguous, and
more lexically primed alternative. From examples like this it is
clear that priming and alignment at the level of the situation model
are very powerful mechanisms in the system, and that they have
significant inhibitory power over the automatic, lower-level priming that is hypothesized to drive alignment at all levels of the system. It remains to be seen whether bottom-up mechanisms are robust enough to drive alignment, as P&G claim, or whether
top-down inhibition directs the system.
The interactive alignment model relies heavily on the assumption that the situation model of one individual can directly prime
the situation model of another individual through the same automatic mechanisms that are responsible for phonological, lexical,
semantic, and syntactic priming between individuals. This assumption seems hasty. Phonological, lexical, and syntactic priming
are similar in that the representation that will eventually be
primed is an inalienable part of the structure of the message that
causes the priming. The words, sounds, and structural patterns
that make up an input string necessarily activate representations
for exactly those words, sounds, and patterns. Semantic priming is
not as direct, as semantic representations cannot be directly read
off an input string. However, semantic representations are generally similar across individuals. For example, it is safe to assume that
the word “dog” will activate the concept “cat” more strongly than
the concept “book” for the wide majority of individuals. This similarity of representation may allow for what P&G represent as a direct priming link between the semantic representations of different individuals. If an individual produces the word “goat,” the
concepts that will be primed in her own semantic representation
are likely similar to the concepts that will be primed in her interlocutor’s representation, because the representations are structured in a similar way. In this indirect way, it could be said that the
semantic representation of one individual can prime the representation of another.
The same thing cannot be said of situation models. Like semantic priming and unlike phonological, lexical, and syntactic
priming, there is no direct priming channel between individuals
through physical aspects of a message such as sounds, words, or
word patterns. Again, priming must be indirect, through the activation of words or phrases that suggest a particular state or property of a situation model. But unlike semantic representations, individuals do not necessarily begin dialogue with similar situation
representations. Therefore the priming link between individuals’
situation models in the interactive alignment model must be of a
different sort from the priming channels between other representations.
The interactive alignment model opens a new and exciting area
of inquiry into language processing. However, more research into
the details of situation model priming and the complex interplay
of priming and inhibition between different levels of representation will be necessary in order to fully specify the operation of the
model and to evaluate its ascribed simplicity. Carrying out this research will not be easy, as it will be difficult to maintain the necessary experimental control in the sorts of experiments that will be
required. But as P&G optimistically point out, “Well-controlled
studies . . . may require some ingenuity, but such experimental ingenuity has always been a strength of psychology” (target article,
sect. 1, para. 4).
BEHAVIORAL AND BRAIN SCIENCES (2004) 27:2
43
Response/Pickering and Garrod: Toward a mechanistic psychology of dialogue
Authors’ Response
The interactive-alignment model:
Developments and refinements
Martin J. Pickeringa and Simon Garrodb
aDepartment of Psychology, University of Edinburgh, Edinburgh EH8 9JZ,
United Kingdom; bDepartment of Psychology, University of Glasgow,
Glasgow G12 8QT, United Kingdom. [email protected]
[email protected]
Abstract: The interactive-alignment model of dialogue provides
an account of dialogue at the level of explanation normally associated with cognitive psychology. We develop our claim that interlocutors align their mental models via priming at many levels of
linguistic representation, explicate our notion of automaticity, defend the minimal role of “other modeling,” and discuss the relationship between monologue and dialogue. The account can be
applied to social and developmental psychology, and would benefit from computational modeling.
The target article set out to show how it would be possible
to develop a theory of interactive language processing at the
level of explanation normally associated with cognitive psychology. In our theory, successful communication involves
the alignment of interlocutors’ representations. We proposed that each level of representation becomes aligned via
an automatic process that we treat as a form of priming, and
that alignment at one level automatically strengthens alignment at other levels. The role of conscious or deliberate
strategies involving explicit reasoning about the mental
states of one’s interlocutor is comparatively small in our account.
Our commentators have raised a number of insightful
points that have caused us to refine our proposals. Many
commentators have focused on the nature of the alignment
process. At a basic level, they consider whether alignment
is the primary mechanism leading to conversational success, to what extent it is automatic, and whether it can be
explained by a single mechanism at all levels and in all contexts. Commentators have also questioned our downplaying
of “other modeling” in ordinary conversation and our claims
about the nature of the difference between monologue and
dialogue. In responding to these and other comments, we
have divided our reply into 11 sections whose order roughly
follows that of the topics raised in our target article.
R1. To what extent do interlocutors align?
Perhaps the most basic issue about our model is whether
interlocutors actually align their situation models, or, less
dramatically, whether they align to the extent that we claim
they do. Schober proposes that interlocutors may be much
less aligned than they appear even when they believe that
they have understood each other. Of course, this would not
matter if it solely concerned rare cases of genuine misunderstanding (e.g., when two interlocutors refer to different
people called John); see also Branigan, who points out that
communication may be “successful” in some sense even
when there is some misunderstanding. But Schober argues
that misalignment is endemic to dialogue. His comments
relate particularly to the interpretation of referring expressions with respect to the discourse model. He draws on ex44
BEHAVIORAL AND BRAIN SCIENCES (2004) 27:2
amples from surveys where respondents interpret terms in
ways that are very different from those intended by the survey compositors. Our response is that such surveys do not
constitute dialogue: The compositors construct the survey,
and the respondents then respond. There is no feedback,
no possibility for repair, and hence no interactive alignment. Schober also raises the important point that people
need not necessarily fully interpret expressions (Clark &
Wilkes-Gibbs 1986). In fact, full interpretation probably
does not always occur in the comprehension of monologue
(Barton & Sanford 1993; Frazier & Rayner 1990; Frisson &
Pickering 1999; 2001; cf. Sanford & Sturt 2002), with people often not determining the precise sense of referring expressions (e.g., newspaper meaning an object vs. a day’s edition), and there is no reason to assume that dialogue is any
different. We suspect that both producers and comprehenders determine meaning to the extent necessary for current
purposes, and that one way in which interlocutors align is
by both processing referring expressions to equal depth.
R2. What precisely are they aligning?
Several commentators appear concerned with the question
of what exactly is being aligned within our model. At the
“lower levels” of phonology, syntax, the lexicon, and so on,
interlocutors presumably align the representational content of each of those levels (phonemes, syntactic structures,
lexical items, etc.), but it is perhaps less clear what they
align at the level of the situation model. In the target article, our intention was to argue for alignment of structural
aspects of the situation model, as exemplified by our example of reference frames. Some of our commentators assume
that we are referring to the content of the situation model.
The questions about alignment of content are much more
difficult, and we shall try to explain the issues below.
In our account, interlocutors align on representations
relevant to the dialogue. These include lexical, semantic,
and syntactic representations, but also the situation model.
So if, at a given point in a conversation, one interlocutor has
a situation model containing two individuals, Mary and
John, with Mary in focus, with each at different locations,
and so on, then the conversation will be successful to the
extent that the other interlocutor constructs the same situation model. Of course, one interlocutor can now introduce
another character (or a new relation between the existing
characters) – indeed, introducing new information is central to any conversation that is not entirely repetitive. To do
this, the speaker draws upon his knowledge (typically using
long-term memory) and adds information to his situation
model. The effect of the alignment is that the listener updates his modelso that it remains similar to that of the
speaker. For example, the listener will interpret ambiguous
words and utterances in the way that the speaker has employed them.
A much bolder claim is that the choice of new topics is
affected by alignment. We did not make this claim in the
target article, although we believe that it is true to some extent. For example, if one interlocutor refers to the couch,
then the other is more likely to refer to the couch as well
(Brennan & Clark 1996; Garrod & Anderson 1987). As a result of this, the use of couch presumably activates knowledge about couches, and hence makes it more likely that
the interlocutor will talk about couches rather than some
Response/Pickering and Garrod: Toward a mechanistic psychology of dialogue
other topic. To this extent, alignment is surely unsurprising
(and simply amounts to the claim that interlocutors will persist with particular topics).
It may also be that interlocutors align on particular styles
of reasoning or accessing of knowledge. For instance, if one
interlocutor is engaged in a careful search of long-term
memory, then the other will tend to behave similarly (e.g.,
if you play a general-knowledge game seriously, then I am
likely to do so too). Alignment on style of reasoning is relevant to the construction of the situation model (cf. Gentner
& Markman 1997), but takes us beyond the scope of the target article, just as nonlinguistic imitation more generally
does (e.g., Chartrand & Bargh 1999). For now, our goals are
limited to understanding linguistic factors that assist in the
alignment of situation models.
Cutting questions our characterization of interactive
alignment as reflecting a direct link between interlocutors.
He suggests that it only has an indirect effect on the language processes themselves. To answer this comment we
need to clarify how interactive alignment relates to language processing. As we have said, our contention is that interactive alignment (and in particular the automatic alignment channels) affects the structures used in production
and interpretation rather than directly determining the
content of production and interpretation. In other words,
we assume that alignment provides an explanation of the
manner in which interlocutors produce and interpret contributions. So we propose that alignment channels only produce a direct link between the structures that the interlocutors use in language processing. Hence, the alignment
process is automatic and direct, even though it does not determine exactly what the speaker produces (as this depends
on his long-term knowledge) or how the addressee interprets what he hears “beyond” the level of the situation
model.
Other commentators also query whether we have specified the appropriate mechanism for alignment of situation
models. They raise this concern in relation to two more specific topics: whether there is one alignment mechanism or
several (Branigan, Glucksberg, and Markman, Kim,
Larkey, Narvaez, & Stilwell [Markman et al.]), and
whether (or in what sense) alignment is automatic (Krauss
& Pardo). All of these commentaries, in some sense, are
concerned with the issue of how alignment could affect the
content of situation models. We address these in the following two sections.
R3. The mechanisms of alignment
Several commentators question the details of the interactive alignment mechanism itself and point out that we have
not fully specified a mechanistic account. Brown-Schmidt
& Tanenhaus make a general plea for modeling, which we
fully agree with (see sect. R11). Goldinger & Azuma argue that we do not give a detailed characterization of the
process by which alignment comes about. We have no commitment to interactive-activation models and are open to
the suggestion that Grossberg’s (1980) adaptive-resonance
theory may provide an appropriate framework for the interactive-alignment account.1
Beyond this, two somewhat different issues are raised.
Some commentators argue that we assume alignment is
based on transient activation, and they propose instead that
it is based on facilitated memory retrieval or implicit learning. Others claim that we are wrong to assume a unified account for all levels of alignment.
Kaschak & Glenberg argue that alignment is not due
to priming but rather to a facilitated memory retrieval
mechanism. In response, we note that the interactive-alignment model is specified at a functional level and makes no
commitment to specific mechanisms, and that we use the
term “priming” to refer to both transient activation and facilitation in memory-based accounts. Our model attempts
to capture the way in which representations used for both
production and comprehension automatically become
aligned as a consequence of the process of interaction.
These representations may be subject to transient activation or, instead, there may be enhancement of the mechanisms underlying their retrieval from memory (as envisaged
by Kashak & Glenberg).
Perhaps more likely, there may be two separate mechanisms involved in alignment. For example, some recent accounts of syntactic priming are based on implicit learning
(Bock & Griffin 2000; Chang et al. 2000), whereas some are
based on activation of grammatical nodes (Hartsuiker et al.,
in press; Pickering & Branigan 1998). Some experimental
research finds clear evidence for long-term priming that is
largely unaffected by intervening material (Bock & Griffin
2000; Hartsuiker & Westenberg 2000), whereas others
shows rapid decay (Branigan et al. 1999; Levelt & Kelter
1982; Wheeldon & Smith 2003). Most likely, different tasks
and sentence types lead to very different time-courses of
priming. Although most of this work does not involve dialogue (except Levelt & Kelter 1982), under our account we
would expect similar patterns of results to occur in dialogue. We therefore suggest that transient activation explains some aspects of alignment, and memory-based
mechanisms explain other aspects of alignment. In section
R9 below, we suggest that alignment due to routinization is
likely to involve the establishment of memory traces for
semi-fixed expressions.
Schiller & de Ruiter argue that interactive alignment
involves storing and re-using selected fragments from previous utterances (see sect. R9); this constitutes a specific
version of a memory-based account. However, their argument is based on the claim that priming is insufficient to account for interactive alignment because syntactic priming
effects are too weak. In fact, the 10–20% effects that they
refer to, occur in monologue. In dialogue, our studies have
shown 55% priming effects when the verb is repeated
(Branigan et al. 2000) and up to 47% with a rare structure
when the noun is repeated (Cleland & Pickering 2003).
Likewise, lexical entrainment almost always occurs for ambiguous words (Brennan & Clark 1996; Garrod & Anderson 1987). In our model, percolation effects between levels
also increase the degree of alignment, and extended dialogue iteratively reinforces alignment.
A number of commentators question whether alignment
operates in the same way at all levels in our model. Markman et al. argue that there are different requirements on
alignment at the different levels. In particular, they separate the situation model from lower levels of linguistic representation. We agree that the structural alignment process
they identify may well be appropriate at the level of the situation model, because models reflect complex higher order
relations between elements (see sect 2.2 of the target article). However, we disagree with their argument that, unlike
BEHAVIORAL AND BRAIN SCIENCES (2004) 27:2
45
Response/Pickering and Garrod: Toward a mechanistic psychology of dialogue
lower level representations, situation models have to be
partially misaligned either to ensure that given-new conventions are followed, or for the maintenance of common
ground. We propose that these requirements can be fulfilled through the implicit common ground which does not
differentiate between the speaker’s and listener’s situation
models.
Branigan also separates the situation model from other
levels, but for reasons that differ from those of Markman
et al. In our terms, she accepts channels of alignment at
syntactic, lexical, and morpho-phonological levels but not at
the level of the situation model, because she believes that
utterances do not provide direct evidence about the situation model. She claims that I am in row two provides direct
evidence about lower levels, whereas the listener has to interpret the utterance (presumably, by using background
knowledge) in order to construct the situation model. We
disagree with this, because all levels of analysis require a
combination of top-down and bottom-up information. For
example, resolving phonemes, ambiguous words, or syntactically ambiguous utterances requires the use of context. It
is therefore wrong to assume that only the level of the situation model is “abstract.” We therefore see no reason to assume that channels of alignment are used only at lower levels, nor do we see any reason to alter our assumption that
alignment at lower levels leads to alignment at the level of
the situation model.
Warren & Rayner argue that the priming link between
individuals’ situation models must be different from that for
lower levels. This is because interlocutors do not necessarily begin dialogues with similar situation representations
and so alignment has to be built up over a period of interaction. Again, we see no fundamental difference between
situation models and lower levels in this respect – alignment at all levels is built up, though the rate of alignment
may differ at different levels. Additionally, Warren &
Rayner question how conflicts in alignment at different levels are resolved (e.g., when aligning on the same name,
“John” might produce a semantic misalignment in contexts
where there are two Johns present). In fact, the issue was
briefly discussed in the target article where we argued that
alignment at the level of the situation model would override
alignment at lower levels (target article, footnote 4). Adopting a particular situation model will influence the way a
speaker frames almost everything he says, whereas adopting a particular word or syntactic structure will only affect
the subsequent choice of that word in preference to another or influence the subsequent use of that particular syntactic structure. Because the situation model is so pervasive, it will be constantly reinforced in implicit common
ground, and misalignment at this level will be more likely
to trigger interactive repair. This suggests that the timecourse of priming at the level of the situation model may be
long-lasting, whereas priming at low levels, such as phonology, may be much more short-lived. Priming at the syntactic level might be intermediate in duration, or depend more
on its precise context (as suggested above). It therefore
might be the case that priming of the situation model depends primarily on memory representations, whereas priming at low levels might depend primarily on transient activation. All of this, however, requires detailed modeling.
We also believe that routinization plays an important role
in reinforcing the links between lower and higher levels of
representation. We take this up in section R9.
46
BEHAVIORAL AND BRAIN SCIENCES (2004) 27:2
R4. What it means for interactive alignment to be
an automatic process
The commentators raise two important issues about automaticity that require discussion. Krauss & Pardo argue
against the idea that alignment can be accounted for in
terms of automatic priming between interlocutors. Shintel
& Nusbaum argue that speech comprehension processes
may be far from automatic in dialogue. To answer these
concerns we need to first explicate our notion of automaticity, and, second, indicate what we assume to be automatic.
Our notion of automaticity is derived from the perspective of perception-action relationships (e.g., Hommel et al.
2001) and, more particularly, social cognition and social
cognitive neuroscience (e.g., Dijksterhuis & Bargh 2001;
Hurley & Chater, in press). Just as Dijksterhuis and Bargh
argue for an automatic perception-behavior expressway, we
propose that the alignment channels are automatic (see
sect. 3.2) – they operate without any intermediary decision
process. Hence, the alignment process is automatic. To be
more explicit, we propose that the automaticity of alignment may take place at what Bargh (1989) calls the postconscious level, whereby automaticity requires awareness
of the stimulus when it originally occurred. This means that
interlocutors have to attend to what the other is saying in
order for automatic alignment to occur. Dijksterhuis and
Bargh (2001, p. 29) also argue that automatic social influences can be inhibited when they conflict with current goals
and purposes. We suggest that the same is true for interactive alignment (see Garrod & Pickering 2004). For example, if a maze game player wants to try a new description
scheme because he has failed to understand the last description from his interlocutor (see sect. 2.1 of the target article), then this high level goal of introducing a new scheme
may inhibit low level alignment arising from what his interlocutor has just said. However, in a similar vein to Dijksterhuis and Bargh, we predict that overriding alignment is
going to be more difficult (or effortful) than adopting alignment. Additionally, this postconscious notion of automaticity can explain why alignment is affected by partner-specific
factors (e.g., Branigan et al. 2003; Metzing & Brennan
2003), without invoking additional mechanisms such as
“other modeling.” It is also presumably relevant to many of
the factors that affect the extent of speech accommodation
(Giles et al. 1992). In general, we expect that rate of alignment may be affected by social factors even when the interlocutors are unaware that they are aligning. There is evidence for such alignment outside language (Epley &
Gilovich 1999; Lakin & Chartrand 2003), and we expect it
also to occur in language.
Krauss & Pardo agree with our claim that communication entails the alignment of situation models, but suggest
that it does not principally take place via automatic priming. For example, they point to evidence that speakers accommodate to their listeners. This presents no problem according to the above conception of automaticity, which
allows inhibition or facilitation by social factors. Glucksberg raises an interesting case, involving a difficult dialogue with a non-cooperative teenage son, in which degree
of alignment may be reduced.
Shintel & Nusbaum argue that speech comprehension
processes may be far from automatic in dialogue. We are
quite happy to accept this general point but see no prob-
Response/Pickering and Garrod: Toward a mechanistic psychology of dialogue
lems for our proposal. In our account, the process of aligning the structures used in comprehension (and production)
is automatic, but other aspects of comprehension (and production) are presumably not automatic. Additionally, their
conception of automaticity is that it “implies a passive
process in which the input is processed in an invariant, inflexible manner, regardless of the beliefs and expectations
of the listener.” This is not the notion of automaticity that
we intend, and we hope that the above discussion of Bargh
(1989) helps to make this clear.
Our conception of automaticity also differs from a Skinnerian one, as suggested by Pear. Crucially, we assume that
alignment is not due to reinforcement, just as Dijksterhuis
and Bargh (2001) assume for the perception-behavior expressway. Instead, alignment follows from a primitive tendency to imitate that does not appear to be learned (e.g.,
Metzoff & Decety 2003). However, our account does share
certain features with Skinner’s (1957) account, in particular
that alignment implicates low-level learning mechanisms.
R5. Parity
One concern is whether there is true representational parity between production and comprehension. Ferreira describes experiments in which participants plan to produce
utterances that they know to be ungrammatical (i.e., participants do not simply make errors). She assumes that interlocutors use and understand such utterances during dialogue (which is almost certainly correct) and suggests that
comprehenders in dialogue would regard them as illicit.
Whereas it is possible that there are differences between
monologue and dialogue with respect to judgments of
grammaticality, we accept that such differences are unlikely. In her experiments, we suggest that speakers realize
they are producing something ungrammatical, but do so
anyway because they cannot think of any other way of saying what they want to say. As long as this realization takes
place within the production system (i.e., does not purely occur during self-monitoring), there does not appear to be a
problem for the parity assumption. Compare sports commentators who sometimes cannot identify a player at the
point when they need to produce the utterance (“About to
kick the ball, Smith”), which listeners might well regard as
illicit. This account seems more likely than a real disconnect
between grammars in comprehension and production.
However, if there is a disconnect for some highly specific
constructions, it merely leads to a very slight weakening of
the parity assumption, not its abandonment.
Ginzburg argues that the interpretation of the same sequence of words can change according to whether it represents a single contribution from one speaker or two contributions from different speakers:
1. A: Which members of our team own a parakeet? A:
Why? ( Why own a parakeet?)
2. A: Which members of our team own a parakeet? B:
Why? ( Why are you asking which members of our team
own a parakeet?)
He suggests that our interactive alignment mechanism cannot account for the fact that Why? has a different interpretation in interactions (1) and (2). This is an interesting observation, but the difference in interpretation between (1)
and (2) hinges on the dialogue move (e.g., questioning, answering, checking, informing) being performed at that
point. Because dialogue moves are generally associated
with particular speakers, it is obviously crucial that interlocutors monitor the source of an utterance when interpreting it (as also follows from results like those of Metzing
& Brennan 2003). For example, the speaker treats a question from his interlocutor differently from the way he would
treat a question from himself. We accept that interlocutors
can monitor the source of a contribution (i.e., they can differentiate between what they are saying and what their
partner is saying) and can take this into account in their interpretation at the level of the dialogue move.
Cutting proposes parity for semantic but not phonological representations on the basis of picture-word interference experiments. From his brief description, we suggest
that participants process the words that they actually produce both semantically and phonologically, but that they
process the words that they are told to ignore semantically
but not phonologically (or at least not to a sufficient depth
to affect priming). Krauss & Pardo also question evidence
for phonological alignment (and by implication for phonological parity). Although we accept that Goldinger (1998)
does not directly demonstrate phonological alignment, recent evidence does support parity between production and
comprehension at this level (Fowler et al. 2003).
Kempson defends a more radical proposal that parity
comes from the symmetry between production and parsing
processes. In her Dynamic Syntax account of parsing, syntactic information is combined with lexical information,
which define semantic interpretations that are built up
word-by-word. Production is assumed to work in essentially
the reverse order. Hence, she sees interactive alignment as
operating at the level of the production and parsing
processes themselves. This is a challenging linguistic proposal, but it would need explicit modeling before it could
be incorporated into a mechanistic account of language
processing in dialogue.
R6. Is it only misunderstanding that drives
interactive repair?
One concern is whether interactive repair is driven primarily by comprehension failure, as we proposed in section 4.3
of the target article. Healey points out that even in the context of Garrod and Anderson’s (1987) maze-game dialogues,
interlocutors change their description scheme in a systematic fashion (e.g., shifting from a path or figural scheme to
a line or matrix scheme). He argues that it is unlikely that
this systematic shift can be accounted for only in terms of
an interactive repair mechanism based on comprehension
failure. Of course we recognize (see sect. 4.4 of the target
article) that alignment does not depend only upon this
process. There are many things that determine what people choose to say and even how they do so which go beyond
the simple automatic mechanisms discussed in the target
article. For example, the shift in description scheme that
Healey mentions probably reflects two opposing pressures.
Whereas the abstract line and matrix descriptions are more
efficient over a period of time than figural or path descriptions (e.g., a line or matrix description involves few words
and is not influenced by whether the position is near a
salient point in the maze or lies in a salient pattern), they
are more difficult to align (e.g., matrix descriptions depend
upon alignment of the origin and of the counting convenBEHAVIORAL AND BRAIN SCIENCES (2004) 27:2
47
Response/Pickering and Garrod: Toward a mechanistic psychology of dialogue
tions used). So they can often not be used securely until
there is a richer implicit common ground (e.g., repeated
use of path descriptions which begin at one corner of the
maze can lead to this corner being adopted as the origin for
matrix description). We suspect that once the implicit
common ground has become sufficiently rich to support the
more abstract description, a speaker is more likely to adopt
that scheme when he encounters a position that is particularly awkward to describe even when it requires a violation
of alignment.
Note that this shift occurs without the speaker having to
take account of the listener’s knowledge. Healey therefore
brings up an important general point, that interlocutors can
go beyond interactive alignment and repair in ways that do
not require other-modeling or the establishment of a full
common ground. For example, a speaker can decide that a
representational scheme is unnecessarily complex or a referring expression is unnecessarily long even if the interlocutors have aligned on that scheme or expression. Similarly, in preparing lectures, I might change how I am
speaking on the basis of my knowledge of the audience (full
common ground), but I might also do it on the basis that
“Hang on, I’m not doing this efficiently given my own resources – I am trying to remember too much and can’t manage it.” This might be argued to involve access to a second
model of one’s own mental state, which is therefore costly,
but less costly than keeping track of full common ground.
Such decisions require there to be some inhibition of the
basic alignment process in light of a conflicting goal (see
sect. R4). In conclusion, Healey’s point reflects something
that is additional to our account rather than in conflict with
it (cf. Krauss & Pardo, who point out that not only misunderstanding drives accommodation).
R7. Other modeling
Although interlocutors undeniably do pay attention to each
others’ mental states on occasion, our contention is that
such other-modeling is resource-intensive, essentially because it involves storing two representations: a representation of one’s own state of knowledge, and a separate
representation of one’s partner’s state of knowledge. We
therefore believe that most of the process of alignment occurs via the interactive-alignment mechanism where othermodeling is not required. But we stress that other-modeling is not purely used for “difficult” cases of interactive
repair when automatic processes fail to work. When a boy
decides to tell his mother what happened at school today,
he presumably realizes that his mother does not know about
the event in question, and therefore that he knows something that his mother does not know. This explicit modeling
of the difference between knowledge states leads to him
running to tell his mother about the event, and does not follow from the failure of interactive alignment and interactive repair. Similarly, a bilingual decides which language to
speak on the basis of his assumptions about which language
his listener knows. However, the undeniable use of such
“broad-brush” other-modeling does not mean that othermodeling is employed in a fine-grained way to explain detailed decisions about one’s individual contributions to an
ongoing dialogue.
In this context, Krauss & Pardo point to evidence that
speakers sometimes modulate their utterances to take into
48
BEHAVIORAL AND BRAIN SCIENCES (2004) 27:2
account their knowledge of the listener: They produce
more informative contributions when they perceive their
addressees to be less knowledgeable about the relevant
topic (see also Isaacs & Clark 1987). The evidence from
Kingsbury (1968) shows that speakers do not simply pay attention to what they believe about their specific interlocutor but make inferences about how much such a person is
likely to know on the basis of the evidence at hand, which
is, in this case, made quite deliberately apparent to the
speaker (e.g., the questioner frames a question to stress his
ignorance of the city). In general, we suspect that speakers
make a one-off decision based on such issues as the perceived expertise of their addressees about how to frame
their contributions (e.g., the decision not to make any assumptions about local geographic knowledge). A teacher
can be much less explicit in the common room than in the
classroom, and a mother does not speak motherese to her
friends. Such decisions need not remain fixed for the whole
conversation (e.g., they might change when the speaker
guesses that his addressee is not a local but then realizes he
is mistaken). But such a change is very different from a continuous, dynamic process of utterance accommodation
based on full common-ground inference, which we argue
to be implausible for reasons of resource limitations (see
sect. 4.1 of the target article). We are therefore grateful to
Krauss & Pardo for stressing that explicit modeling does not
only occur when automatic processes fail to produce alignment, but we see no concern for our assumption that automatic mechanisms underlie alignment.
Fussell & Kraut argue that speakers with different
views of a spatial scene take into account the listener’s perspective, in effect modeling the listener’s mental state. They
describe a collaborative bicycle repair situation in which an
expert helper guides a novice repairer. They note that when
the repairer knows that he can be seen even when he cannot see his remote helper, he will use deictic expressions to
describe the things in front of him (e.g., See this piece, while
pointing at a cycle component), whereas the remote helper
will not (e.g., See the derailleur). They argue that this is inconsistent with alignment and provides further evidence of
other modeling. We are not convinced. We suspect that
speakers in this situation prefer to use deictic expressions
because they are shorter, do not require word finding, and
so on. But deixis is not an option for the remote helper because he cannot point to anything. Instead, he has to fall
back on more complex nondeictic descriptions. (One remote helper is quoted as saying in frustration, “If I could
point to it, it’s right there”; Kraut et al. 2003, p. 36.) So the
circumstances may force the speaker to use a more complex
nonaligned utterance. It is of course reasonable that alignment is broken under such circumstances, because it simply would not work. One important point this raises is that
the tendency toward alignment is likely to be stronger under conditions where two interlocutors are placed in comparable environments. Presumably this reflects nonlinguistic contributions to linguistic alignment (see also the
discussion of Dominey in sect. R11).
Nevertheless, we certainly agree with the general point
that when communicators share a physical situation they
take situational awareness into account in formulating utterances. But is this evidence for listener modeling? In the
“side-by-side” situation described by Kraut et al. (2003,
communicators use direction of gaze to establish joint attention, but the effect of one partner’s point of gaze on the
Response/Pickering and Garrod: Toward a mechanistic psychology of dialogue
other partner’s focus of attention reflects low-level mechanisms which do not depend on inferences about the partner’s mental state (see Schuller & Rossion 2001). And, because in this situation what is accessible to the speaker will
usually be equally accessible to the partner (see sect. 4.1),
an essentially egocentric approach will generally support
successful communication without requiring speakers to
model their listeners.
Schober suggests that current evidence cannot distinguish two possibilities: that interlocutors only model each
other’s mental states under exceptional circumstances, and
that interlocutors normally model each other’s mental
states and only fail to do so when under great cognitive load
or when circumstances weigh heavily against doing so. We
accept that current evidence does not distinguish between
these two positions. However, our account assumes the use
of fewer resources and is parsimonious (obviously, an account containing two mental models is harder to falsify than
an account limited to one, just as a parallel account is harder
to falsify than a serial account). Moreover, Schober’s proposal cannot hold for multiparty dialogues containing more
than a small number of people, because it must become impossible to retain and regularly update a different mental
model for each person. In general, cognitive psychology
teaches us that constructing mental models is hard and
holding onto different models at the same time is especially
hard (e.g., Johnson-Laird 1983). We suggest that the paradox of how one can know when to model one’s partner is
easy to accommodate: Contributions to the dialogue will
make it clear that alignment is breaking down, and if interactive repair does not solve the problem, the interlocutor is
forced to assume that what his partner knows is likely to be
different from what he knows. Even in such cases, it may
be that interlocutors only model those differences between
themselves and their partners that need to be modeled in
order to allow the recovery of alignment.
Brennan & Metzing also criticize our assumption that
interlocutors do not routinely employ full common ground.
A fast-growing body of literature suggests that interlocutors
sometimes do pay attention to each others’ knowledge in
comprehension and production (e.g., Hanna et al. 2003;
Lockridge & Brennan 2002; Nadig & Sedivy, 2002) and
sometimes do not (e.g., Brown & Dell 1987; Ferreira &
Dell 2000; Keysar et al. 2003). It is too early to say precisely
when such knowledge can affect processes of production
and comprehension, but current evidence suggests both
that interlocutors can immediately draw upon knowledge
about differences between their own knowledge and their
beliefs about their partner’s knowledge, and that interlocutors can make egocentric decisions about production and
comprehension. Most of these studies involve a fairly artificial situation in which the experimental subject is informed that his interlocutor may have knowledge about the
situation that differs from his in quite specific ways. Experiments like that of Hanna et al. (2003) show that it can be
straightforward to add one fact about your interlocutor –
namely, that he does not have access to a particular piece of
information that you have. Even in such cases, some egocentric behavior remains, as Hanna et al. acknowledge and
Keysar et al. (2003) demonstrate. But adding one fact about
your interlocutor’s knowledge is quite different from maintaining a full representation of the interlocutor’s situation
model, and performing reasoning based on that model.
Available resources do not normally allow interlocutors to
constantly update models of each others’ mental states.
However, this does not lead to communicative breakdown
because aligned interlocutors develop the same situation
models.
In response to Brennan & Metzing, we stress that it
was not our intention to commit to a two-stage account
(e.g., Horton & Keysar 1996), in which other modeling occurs during revision but not during initial processing
(whether production or comprehension). We note that
Krauss & Pardo and Brown-Schmidt & Tanenhaus also
interpret us as making this proposal, and accept that we did
not make this very clear. Rather, we claim that “performing
inferences about common ground is an optional strategy
that interlocutors employ only when resources allow” (target article, sect. 4.2, para. 4).
We do not regard Metzing and Brennan’s (2003) demonstration of partner-specific effects as problematic, and assume they can be explained in similar ways to Branigan et
al.’s (2003) demonstration that syntactic alignment is sensitive to participant status (see sect. 2.3 of the target article).
As we have pointed out in section R6 of this response, we
assume that alignment is automatic at a postconscious level
(Bargh 1989) and, hence, can be affected by a range of social factors from stereotype activation to participant status.
A particular speaker is associated with a particular form,
and breaking that association causes disruption. There is no
need for other modeling to occur in this process of partnerspecific lexical entrainment. The term conceptual pact appears to suggest that other modeling is used in lexical entrainment. If so, we would question whether it is generally
appropriate.
R8. Routines
Schiller & de Ruiter propose that interactive alignment
necessarily involves selecting stored fragments from previous utterances. This corresponds to our notion of routinization (see sect. 5 of the target article). We suspect that
routinization comes about as a result of a longer lasting
alignment mechanism based on memory retrieval rather
than transient activation. This is because routines reflect
multiple links between different levels of representation
(e.g., they fix the relation between a word and its meaning,
its syntactic form, and even its interpretation within a situation model) and it is difficult to imagine how this could be
captured and routinized through purely transient activation. Rather than assume that routinization is the sole explanation of alignment, we suggest that it is a consequence
of implicit learning but that transient activation also promotes alignment (see sect. R3). It may of course be that
routines emerge from a resonance process, as Goldinger
& Azuma suggest. In addition, because routinization works
by linking levels of representation, it may explain how alignment percolates up from lower to higher levels (cf. issues
raised by Warren & Rayner and Branigan, as discussed
in sect. R6).
Within the interactive-alignment account, we regard
routines as an extreme case of alignment, involving a fixed
form and interpretation. It may be best to think of routinization as falling on a continuum, with expressions that
contain some fixed elements (as in many of Kuiper’s 1996
examples) being more or less “semi-routinized.” Assuming
that it is correct to regard alignment as a mixture of tranBEHAVIORAL AND BRAIN SCIENCES (2004) 27:2
49
Response/Pickering and Garrod: Toward a mechanistic psychology of dialogue
sient activation and implicit learning, we propose that the
more routinized an expression is, the more it is best explained in terms of implicit learning – for the purposes of
the conversation at least, the expression and its interpretation are stored and retrieved. Of course, if an expression becomes sufficiently entrenched, it may survive that conversation. Although other frameworks are no doubt possible,
we regard Jackendoff’s (2002) account of fixed and semifixed expressions as an appropriate representational scheme
for semi- and completely routinized expressions (see Pickering & Garrod 2004, for discussion)
R9. Self-monitoring
Schiller & de Ruiter question our claim that self-monitoring can occur at any level of linguistic representation that
can be aligned. We did not claim that there is conclusive evidence for this hypothesis and we believe that careful empirical work is needed to distinguish our proposal from the
proposal that monitoring works externally on sound and internally on phonological representations alone. However,
we would query whether the reported evidence provides
strong support for this alternative proposal. First, the comparative slowness of selecting a gender label in comparison
to selecting the indefinite article in French may have many
explanations, perhaps most likely that selecting between
genders is a more abstract and difficult task than selecting
between (very common) words. Second, the strong evidence for monitoring of various aspects of phonological
representations is completely compatible with monitoring
of other linguistic representations. Although some or all
gender-congruency effects in picture-word interference
tasks may really be determiner congruency effects (Schiller
& Caramazza 2003), there is also considerable evidence
that grammatical gender can be accessed when phonological form is not available (Badecker et al. 1995; Vigliocco et
al. 1997). Therefore, it is at least plausible that people can
directly monitor for errors of grammatical gender and indeed for other aspects of syntactic representations. If an utterance is ill-formed at different levels of representation simultaneously, we suspect that there may be a race between
monitoring processes at these different levels, in which case
it might not always be possible to detect monitoring that
takes place at the “slower” level.
R10. On the difference between dialogue and
monologue
A number of commentators argue that language processing
in dialogue is not fundamentally different from that in
monologue. For example, both Barr & Keysar and
Glucksberg point out that the same basic language processes operate in monologue and dialogue so there is no
principled difference between the two. We agree in the
sense that the actual production and comprehension mechanisms are the same (at what we might term a “microlevel”).
However, the radically different contexts in which they operate lead to very different results. For example, a speaker’s
utterances are dramatically affected by the presence of the
interlocutor – the speaker aligns with the interlocutor’s utterances via the mechanisms we have described. In this respect we argue that the language processing system is designed for dialogue rather than monologue. As a result,
50
BEHAVIORAL AND BRAIN SCIENCES (2004) 27:2
speakers have to learn special strategies to deal with monologue which are not required during dialogue processing.
We agree with Glucksberg that dialogue is not necessarily easier than monologue, and accept that contextual effects can be very strong in monologue. We propose that the
priming mechanisms are ideally suited for dialogue. Presumably they have developed from imitation (Arbib 2003)
and it may be that the organization of dialogue (e.g., time
between turns) is optimal for the mechanisms of priming.
Therefore, dialogue does not need to rely on nonautomatic
inference. In contrast, monologue cannot use priming between interlocutors (by definition) and therefore has to rely
on inference, other-modeling, and so on. Priming is of
course present in monologue, but we contend that it is far
less useful than in dialogue (e.g., repetition is much rarer in
monologue than in dialogue; see sect. 5.1 of the target article). So we concur that there is not a principled distinction
between dialogue and monologue, but at the same time
maintain that dialogue will usually but not always be easier
than monologue.
Barr & Keysar appear to disagree with us more than we
think they actually do. They are mistaken in assuming that
we propose a categorical distinction between monologue
and dialogue. In section 8 of the target article, we refer to
a “dialogic continuum” with monologue at one end, and
fully interactive dialogue at the other. We assume that the
same mechanisms are present in dialogue and monologue
(i.e., people do not set some processing “switch”). In true
monologue, the speaker has no interlocutor to align with.
He can of course align with himself and certainly does so
(e.g., re-using the same word with the same meaning). We
completely agree that dialogues go through various stages,
with some involving rapid turn-taking (e.g., question answering) and some involving much more limited feedback
(e.g., during narratives). Boden (1994) distinguishes between conversational phases and presentational phases in
group discussion. These presentational phases are not
monologues, as even minimal feedback affects them considerably (Bavelas et al. 2000).
Hence, we stress that monologue and dialogue lie on a
continuum, and we predict that the degree of alignment
will be affected by the position on the continuum. One important area for research is to consider the effects of dialogue genre on alignment (in which context we can regard
monologue as particular genre). For example, Schegloff
points to the importance of different speech-exchange systems (conversation, interview, giving a speech, etc.) in affecting the characteristics of the dialogue (e.g., turn-taking
behavior, routinization). We predict that the rate and characteristics of alignment are not constant for all forms of dialogue, but will depend on the speech-exchange system.
For example, forms of interaction that do not allow unconstrained feedback and where turn-taking is externally managed (e.g., interviews) will fail to employ the interactive repair mechanism to the extent that is possible in casual
conversation.
R11. Extensions and discussion
Schegloff argues that our mechanistic account fails to consider the richness of the interaction afforded by dialogue.
Although Schegloff’s sociological starting point (i.e., in
terms of organizational practice and interaction contingen-
Response/Pickering and Garrod: Toward a mechanistic psychology of dialogue
cies) is somewhat different from ours, we certainly agree
that there are additional specific details of dialogue organization that must enter into any complete mechanistic account. We also recognize the considerable contribution that
Schegloff and colleagues have made in mapping out the details of these organizational practices and the contingencies
they afford. However, our mechanistic aspiration goes beyond mapping out such practices and contingencies. Like
Brown-Schmidt & Tanenhaus we believe that a mechanistic account should make it possible to formulate a computational model of the processes involved in the comprehension and production of dialogue and how these take
advantage of the interactional nature of dialogue. We also
recognize that any complete model will have to take account of both self and other commitments in dialogue processing (see our response to Ginzburg in sect. R5). We
stress that our paper is entitled “Toward a mechanistic psychology of dialogue”!
Two commentators argue for a broadening of the interactive alignment account to include other interactive situations. Mazur proposes that interactive alignment needs to
be embedded in a broader theory of communication that
pays attention to a range of social conventions. We agree
that a full theory of interactive alignment will make reference to nonlinguistic as well as linguistic information, and
believe that our suggestions about the relations between
our account and implicit social cognition is a step in this direction.
Dominey draws interesting parallels between the interactive alignment process in adult dialogue and certain features of language acquisition. Language learning depends
upon extralinguistic or prelinguistic alignment mechanisms
(e.g., establishing joint attention on intended referents
through gaze direction or postural orientation) . Also, there
is evidence that routinization of utterances associated with
repeated action scenarios (feeding, bathing, etc.) may play
an important role in the acquisition of syntax (Tomasello
2003). These suggestions help reinforce the claim that nonlinguistic alignment may lead to linguistic alignment, just as
linguistic alignment at one level leads to linguistic alignment at other levels (see our discussion of Fussell & Kraut
in sect. R7). In fact, Dominey suggests that such linguistic/
nonlinguistic links are necessary to explain the process of
language acquisition, where one partner (the infant) does
not initially have linguistic abilities. A full theory of how interactive-alignment might explain acquisition would be fascinating. In particular, we are intrigued by the suggestion
that learning by alignment might avoid the enlistment of
generative grammar mechanisms, perhaps in a way similar
to that envisaged by Tomasello.
Language acquisition is a good example of how it may be
possible to extend our account into new domains. Other areas that we have highlighted at various points in the target
article and this response include social psychology and human-computer interaction. A recurring theme is that it may
be sensible to include nonlinguistic alignment into developments of our model; interlocutors who are aligned in
nonlinguistic (e.g., body posture) or paralinguistic (e.g.,
tone of voice) ways may be more likely to align linguistically.
We emphasize that our use of the term “priming” is at a
fairly abstract functional level, as our notion of automaticity makes clear (sect. R4). It allows nonconscious mediation by factors that may originate in distinctions that interlocutors are aware of (e.g., participant status, social status,
cooperativeness). We also note that “priming” may employ
transient activation or implicit learning or both. To be more
speculative, we suspect that interactive alignment may
work by two distinct mechanisms: a brief activation-based
process that may not be affected by intentional distinctions,
and a longer-lasting memory-based process that is intentionally mediated. The effects of these two processes will
depend on precise timing, and will therefore be differentially affected by aspects of the conversation that affect
timing. For example, a high-engagement face-to-face dialogue between intimate friends may result in timing that is
precisely attuned to increasing alignment, whereas a dialogue between strangers that depend on external factors
such as rules of engagement (e.g., in an interview) or technology (e.g., walkie-talkies) may not. We suspect that the
longer-lasting process will not be affected but the activation
process might be impaired in low-involvement dialogue.
These speculative comments could inform an extensive
program of empirical research concerned with the conditions that lead to alignment in dialogue (e.g., its time
course).
The other obvious area for development is explicit computational modeling, as highlighted by Brown-Schmidt &
Tanenhaus in particular. To perform such modeling, it
would of course be necessary to explicate many assumptions of our account that are currently vague or implicit, for
instance by developing interactive alignment, interactive
repair, and other-modeling components. It would be necessary to model the process whereby alignment at one level
leads to alignment at other levels, and to understand how
conflicts of alignment are resolved (see Warren &
Rayner). We need to know whether transient activation
and implicit learning should be distinguished, and if so, how
they interact. Finally, any such account should explain the
process of routinization and describe its effects on alignment.
NOTE
1. Note that the uses of “interactive” in interactive alignment
and interactive activation are unrelated.
References
Letters “a” and “r” appearing before authors’ initials refer to target article
and response respectively.
Ades, A. & Steedman, M. J. (1982) On the order of words. Linguistics and
Philosophy 4:517–58. [aMJP]
Aijmer, K. (1996) Conversational routines in English: Convention and creativity.
Longman. [aMJP]
Allen, J. F., Byron, D. K., Dzikovska, M., Ferguson, G., Galescu, L. & Stent, A.
(2001) Towards conversational human-computer interaction. AI Magazine
22:27–35. [SB-S]
Altenberg, B. (1990) Speech as linear composition. In: Proceedings from the
Fourth Nordic Conference for English Studies, ed. G. Caie, K. Haastrup, A. L.
Jakobsen, J. E. Nielsen, J. Sevaldsen, H. Specht & A. Zettersten, pp. 133–43.
University of Copenhagen. [aMJP]
Altmann, G. T. & Steedman, M. J. (1988) Interaction with context during human
sentence processing. Cognition 30:191–238. [aMJP]
Amis, K. (1997) The King’s English: A guide to modern English usage. Harper
Collins. [aMJP]
Anderson, A. H. & Boyle, E. (1994) Forms of introduction in dialogues: Their
discourse contexts and communicative consequences. Language and
Cognitive Processes 9:101–22. [aMJP]
Arbib, M. A. (2003) From monkey-like action recognition to human language: An
evolutionary framework for neurolinguistics. Unpublished manuscript.
[rMJP]
BEHAVIORAL AND BRAIN SCIENCES (2004) 27:2
51
References/Pickering and Garrod: Toward a mechanistic psychology of dialogue
Badecker, W., Miozzo, M. & Zanuttini, R. (1995) The dissociation of form-based
lexical retrieval from access to word-specific aspects of grammar: Evidence for
two-stage lexical processing. Cognition 57:193 –216. [rMJP]
Balota, D. A. (1990) The role of meaning in word recognition. In: Comprehension
processes in reading, ed. D. A. Balota, G. B. Flores d’Arcais & K. Rayner.
Erlbaum. [JCC]
Balota, D. A., Paul, S. T. & Spieler, D. H. (1999) Attentional control of lexical
processing pathways during word recognition and reading. In: Language
processing, ed. S. Garrod & M. Pickering. Psychology Press. [aMJP]
Bard, E. G., Anderson, A. H., Sotillo, C., Aylett, M., Doherty-Sneddon, G. &
Newlands, A. (2000) Controlling the intelligibility of referring expressions in
dialogue. Journal of Memory and Language 42:1–22. [aMJP]
Bargh, J. A. (1989) Conditional automaticity: Varieties of automatic influence in
social perception and cognition. In: Unintended thoughts, ed. J. S. Uleman &
J. A. Bargh, pp. 3 – 51. Guilford Press. [rMJP, HS]
Bargh, J. A. & Chartrand, T. L. (1999) The unbearable automaticity of being.
American Psychologist 54:462–79. [aMJP]
Bargh, J. A., Chen, M. & Burrows, L. (1996) Automaticity of social behavior:
Direct effects of trait construct and stereotype activation on action. Journal of
Personality and Social Psychology 71:230 – 44. [aMJP]
Baron-Cohen, S., Tager-Flusberg, H. & Cohen, D. (2000) Understanding other
minds: Perspectives from developmental neuroscience. Oxford University
Press. [aMJP]
Barr, D. J. & Keysar, B. (2002) Anchoring comprehension in linguistic precedents.
Journal of Memory and Language 46:391– 418. [DJB]
(in press) Making sense of how we make sense: The paradox of egocentrism in
language use. In: Figurative language comprehension: Social and cultural
influences, ed. H. L. Colston & A. N. Katz. Erlbaum. [DJB]
Barton, S. & Sanford, A. J. (1993) A case-study of pragmatic anomaly-detection:
Relevance-driven cohesion patterns. Memory and Cognition 21:477–87.
[rMJP]
Barwise, J. (1989) Three views of common knowledge. In: The situation in logic,
ed. J. Barwise. CSLI. [aMJP]
Bates, E. & MacWhinney, B. (1987) Competition, variation and language learning.
In: Mechanisms of language acquisition, ed. B. MacWhinney & E. Bates.
Erlbaum. [PFD]
Bavelas, J. B., Coates, L. & Johnson, T. (2000) Listeners as co-narrators. Journal of
Personality and Social Psychology 79:941– 52. [arMJP]
Binder, K. S. & Rayner, K. (1998) Context strength does not modulate the
subordinate bias effect: Evidence from eye fixations and self-paced reading.
Psychonomic Bulletin and Review 5:271–76. [aMJP]
Bock, J. K. (1986a) Meaning, sound, and syntax: Lexical priming in sentence
production. Journal of Experimental Psychology: Learning, Memory, and
Cognition 12:575 – 86. [aMJP]
(1986b) Syntactic persistence in language production. Cognitive Psychology
18:355 – 87. [aMJP]
(1989) Closed class immanence in sentence production. Cognition 31:163–86.
[aMJP]
(1996) Language production: Methods and methodologies. Psychonomic
Bulletin and Review 3:395 – 421. [JCC, aMJP]
Bock, J. K. & Griffin, Z. M. (2000) The persistent of structural priming: Transient
activation or implicit learning? Journal of Experimental Psychology: General
129:177– 92. [JCC, MK, rMJP]
Bock, J. K. & Huitema, J. (1999) Language production. In: Language processing,
ed. S. Garrod & M. Pickering. Psychology Press. [aMJP]
Bock, J. K. & Levelt, W. J. M. (1994) Language production: Grammatical
encoding. In: Handbook of psycholinguistics, ed. M. A. Gernsbacher,
pp. 945 – 84. Elsevier/Academic Press. [JCC, aMJP]
Bock, J. K. & Loebell, H. (1990) Framing sentences. Cognition 35:1–39. [aMJP]
Bock, J. K., Loebell, H. & Morey, R. (1992) From conceptual roles to structural
relations: Bridging the syntactic cleft. Psychological Review 99:150–71.
[aMJP]
Boden, D. (1994) The business of talk. Polity Press. [rMJP]
Boroditsky, L. (2000) Metaphorical structuring: Understanding time through
spatial metaphors. Cognition 75:1–28. [aMJP]
Bourhis, R. Y. & Giles, H. (1977) The language of intergroup distinctiveness. In:
Language, ethnicity and intergroup relations, ed. H. Giles, pp. 119–35.
Academic Press. [RMK]
Branigan, H. P., Pickering, M. J. & Cleland, A. A. (1999) Syntactic priming in
written production: Evidence for rapid decay. Psychonomic Bulletin and
Review 6:635 – 40. [rMJP]
(2000) Syntactic coordination in dialogue. Cognition 75:B13 –25. [arMJP]
(2003) Syntactic alignment and participant status in dialogue (submitted).
[arMJP]
Brennan, S. E. & Clark, H. H. (1996) Conceptual pacts and lexical choice in
conversation. Journal of Experimental Psychology: Learning, Memory, and
Cognition 22:1482– 93. [arMJP]
Brennan, S. E. & Schober, M. F. (2001) How listeners compensate for dysfluencies
52
BEHAVIORAL AND BRAIN SCIENCES (2004) 27:2
in spontaneous speech. Journal of Memory and Language 44:274 – 96.
[aMJP]
Brooks, P. & Tomasello, M. (1999) Young children learn to produce passives with
nonce verbs. Developmental Psychology 35:29–44. [aMJP]
Brown, P. M. & Dell, G. S. (1987) Adapting production to comprehension: The
explicit mention of instruments. Cognitive Psychology 19:441–72. [SEB,
arMJP]
Brown-Schmidt, S., Campana, E. & Tanenhaus, M. K. (2002) Reference resolution
in the wild: How addresses circumscribe referential domains in a natural,
interactive problem-solving task. Paper presented at the Annual Meeting of
the Cognitive Science Society, Fairfax, VA, August 2002. [SB-S]
(in press) Real-time reference resolution by naïve participants during a taskbased unscripted conversation. In: World-situated language processing:
Bridging the language as product and language as action traditions, ed. J. C.
Trueswell & M. K. Tanenhaus. MIT Press. [SB-S, aMJP]
Burnard, L. (2000) Reference guide for the British National Corpus (World
Edition). Oxford University Computing Services. [aMJP]
Calvert, G. A., Bullmore, E. T., Brammer, M. J., Campbell, R., Williams, S. C. R.,
McGuire, P. K., Woodruff, P. W. R., Iversen, S. D. & David, A. S. (1997)
Activation of auditory cortex during silent lipreading. Science 276:593 – 96.
[aMJP]
Capella, J. (1981) Mutual influence in expressive behavior: Adult-adult and infantadult dyadic interaction. Psychological Bulletin 89:101–32. [SDG]
Caramazza, A. (1991) Issues in reading, writing, and speaking. A
neuropsychological perspective. Kluwer Academic. [JCC]
(1997) How many levels of processing are there in lexical access? Cognitive
Neuropsychology 14:177–208. [aMJP]
Carlson, K. (2001) The effects of parallelism and prosody in the processing of
gapping structures. Language and Speech 44:1–26. [aMJP]
Carlson-Radvansky, L. A. & Jiang, Y. (1998) Inhibition accompanies reference
frame selection. Psychological Science 9:386–91. [aMJP]
Carpenter, G. & Grossberg, S. (1987) A massively parallel architecture for a selforganizing neural recognition machine. Computer Vision, Graphics, and
Image Processing 37:54–115. [SDG]
Carpenter, G. & Grossberg, S., eds. (1991) Pattern recognition by self-organizing
neural networks. MIT Press. [SDG]
Chambers, C. G., Tanenhaus, M. K., Eberhard, K. M., Filip, H. & Carlson, G. N.
(2002) Circumscribing referential domains in real-time sentence
comprehension. Journal of Memory and Language 47:30–49. [aMJP]
Chang, F., Dell, G. S., Bock, J. K. & Griffin, Z. M. (2000) Structural priming as
implicit learning: A comparison of models of sentence production. Journal of
Psycholinguistic Research 29:217–29. [rMJP]
Charniak, E. (1993) Statistical language learning. MIT Press. [aMJP]
Chartrand, T. L. & Bargh, J. A. (1999) The chameleon effect: The perceptionbehavior link and social interaction. Journal of Personality and Social
Psychology 76:893–910. [arMJP]
Cherry, E. C. (1956) On human communication. MIT Press. [aMJP]
Chomsky, N. (1965) Aspects of the theory of syntax. MIT Press. [HPB, aMJP]
(1973) Conditions on transformations. In: A Festschrift for Mossis Halle, ed.
S. R. Anderson & R. Kiparsky, pp. 232–86. Holt, Rinehart and Winston.
[FF]
(1981) Lectures on government and binding. Foris. [aMJP]
(1995) The minimalist program. MIT Press. [aMJP]
Christiansen, M. H. & Chater, N. (2001) Connectionist psycholinguistics:
Capturing the empirical data. Trends in Cognitive Sciences 5:82–8. [SB-S]
Clark, E. V. (1993) The lexicon in acquisition. Cambridge University Press.
[aMJP]
Clark, H. H. (1979) Responding to indirect speech acts. Cognitive Psychology
11:430–77. [aMJP]
(1985) Language and language users. In: The handbook of social psychology, 3rd
edition, ed. G. Lindzey & E. Aronson. Harper Row. [aMJP]
(1992) Arenas of language use. University of Chicago Press. [aMJP]
(1996) Using language. Cambridge University Press. [JG, SG, RMK, ABM,
aMJP]
(1998) Communal lexicons. In: Context in language learning and language
understanding, ed. K. Malmkjoer & J. Williams. Cambridge University Press.
[aMJP]
Clark, H. H. & Marshall, C. R. (1981) Definite reference and mutual knowledge.
In: Elements of discourse understanding, ed. A. K. Joshi, I. A. Sag & B. L.
Webber. Cambridge University Press. [DJB, SRF, JG, aMJP]
Clark, H. H. & Murphy, G. L. (1982) Audience design in meaning and reference.
In: Language and comprehension, ed. J. F. Le Ny & W. Kintsch. NorthHolland. [RMK, aMJP]
Clark, H. H. & Schaefer, E. F. (1987) Concealing one’s meaning from overhearers.
Journal of Memory and Language 26:209–25. [aMJP]
Clark, H. H. & Schober, M. F. (1991) Asking questions and influencing answers.
In: Questions about questions: Inquiries into the cognitive bases of surveys,
ed. J. M. Tanur, pp. 15–48. Russell Sage Foundation. [MFS]
References/Pickering and Garrod: Toward a mechanistic psychology of dialogue
Clark, H. H. & Wilkes-Gibbs, D. (1986) Referring as a collaborative process.
Cognition 22:1– 39. [arMJP, MFS]
Cleland, A. A. & Pickering, M. J. (2003) The use of lexical and syntactic
information in language production: Evidence from the priming of nounphrase structure. Journal of Memory and Language 49:214 – 30. [arMJP]
Clement, C. A. & Gentner, D. (1991) Systematicity as a selection constraint in
analogical mapping. Cognitive Science 15:89 –132. [ABM]
Coates, J. (1990) Modal meaning: The semantic-pragmatic interface. Journal of
Semantics 7:53 – 64. [aMJP]
Conrad, F. G. & Schober, M. F. (2000) Clarifying question meaning in a household
telephone survey. Public Opinion Quarterly 64:1–28. [MFS]
Consortium, T. T. (2000) The TRINDI Book. University of Gothenburg. Available
at: http://www.ling.gu.se/research/projects/trindi. [JG]
Cooper, R., Larsson, S., Hieronymus, J., Ericsson, S., Engdahl, E. & Ljunglof, P.
(2000) GODIS and questions under discussion. In: The TRINDI Book.
University of Gothenburg. Available at: http://www.ling.gu.se/research/
projects/trindi. [JG]
Creswell, C. (2002) Resumptive pronouns, wh-island violations, and sentence
production. In: Proceedings of the Sixth International Workshop on Tree
Adjoining Grammar and Related Frameworks (TAG6), pp. 101–109.
Universita di Venezia. [FF]
Cutting, J. C. (1998) The production and comprehension lexicons: What is shared
and what is not. Doctoral dissertation, University of Illinois, 1997. Dissertation
Abstracts International: Section B: The Sciences and Engineering 58:10–B.
[JCC]
Cutting, J. C. & Bock, J. K. (1997) That’s the way the cookie bounces: Syntactic
and semantic components of experimentally elicited idiom blends. Memory
and Cognition 25:57–71. [JCC, aMJP]
de Boer, B. (2000) Self-organization in vowel systems. Journal of Phonetics 28:441–
65. [SDG]
Dell, G. S. (1986) A spreading-activation theory of retrieval in sentence
production. Psychological Review 93:283 – 321. [SDG, aMJP]
Dell, G. S, Chang, F. & Griffin, Z. M. (1999) Connectionist models of language
production: Lexical access and grammatical encoding. Cognitive Science
23:517– 42. [JCC]
Desrochers, A. & Paivio, A. (1990) Le phonem initial des noms inanimes et son
effet sur l’identification du genre grammatical [The initial phoneme of
inanimate nouns and its effect for the identification of grammatical gender].
Canadian Journal of Psychology 44:44 – 57. [NOS]
Desrochers, A., Paivio, A. & Desrochers, S. (1989) L’effet de la frequence d’usage
des noms inanimes et de la valeur predictive de leur terminaison sur
l’identification du genre grammatical [The effect of frequency of usage of
inanimate nouns and of the predictive value of their endings on the
identification of grammatical gender]. Canadian Journal of Psychology 43:62–
73. [NOS]
Dijksterhuis, A. & Bargh, J. A. (2001) The perception-behavior expressway:
Automatic effects of social perception on social behavior. In: Advances in
experimental social psychology, vol. 33, ed. M. P. Zanna, pp. 1– 40. Academic
Press. [arMJP]
Dijksterhuis, A. & van Knippenberg, A. (1998) The relation between perception
and behavior or how to win a game of Trivial Pursuit. Journal of Personality
and Social Psychology 74:865 –77. [aMJP]
Doherty-Sneddon, G., Anderson, A. H., O’Malley, C., Langton, S., Garrod, S. &
Bruce, V. (1997) Face-to-face and video-mediated communication: A
comparison of dialogue structure and task performance. Journal of
Experimental Psychology: Applied 3:105 –25. [aMJP]
Dominey, P. F. (2000) Conceptual grounding in simulation studies of language
acquisition. Evolution of Communication 4(1):57– 85. [PFD]
(2003a) Learning grammatical constructions in a miniature language from
narrated video events. In: Proceedings of the 25th Annual Meeting of the
Cognitive Science Society. (in press). Available at: http://www.isc.cnrs.fr/wp/
DomineyISC-WP0.pdf. [PFD]
(2003b) Learning grammatical constructions from narrated video events for
human-robot interaction. In: Proceedings of the 2003 IEEE Humanoid Robot
Conference, Karlsruhe, Germany. [PFD]
Drew, P. (1997) “Open” class repair initiators in response to sequential sources of
troubles in conversation. Journal of Pragmatics 28:69 –101. [EAS]
Duffy, S. A., Morris, R. K. & Rayner, K. (1988) Lexical ambiguity and fixation
times in reading. Journal of Memory and Language 27:429–46.
[aMJP]
Epley, N. & Gilovich, T. (1999) Just going along: Nonconscious priming and
conformity to social pressure. Journal of Experimental Social Psychology
35:578 – 89. [rMJP]
Fay, N., Garrod, S. & Carletta, J. (2000) Group discussion as interactive dialogue
or as serial monologue: The influence of group size. Psychological Science
11:481– 86. [aMJP]
Fernández, R. & Ginzburg, J. (2002) Non-sentential utterances: Grammar and
dialogue dynamics in corpus annotation. In: Proceedings of the 19th
International Conference on Computational Linguistics (CoLing), pp. 253–
59. Morgan Kaufman. [aMJP]
Ferreira, F. & Swets, B. (2003) The production and comprehension of resumptive
pronouns in relative clause “island” contexts. Paper presented at the Max
Planck Institute’s Four Corners Workshop Series. Workshop 2: The
relationship between language comprehension and language production.
Nijmegen, the Netherlands, April 5, 2003. [FF]
Ferreira, V. S. & Dell, G. S. (2000) Effect of ambiguity and lexical availability on
syntactic and lexical production. Cognitive Psychology 40:296–340. [SEB,
arMJP]
Ferreira, V. S. & Griffin, Z. M. (2003) Phonological influences on lexical
(mis)selection. Psychological Science 14:86–90. [SG]
Fodor, J. A., Bever, T. G. & Garrett, M. F. (1974) The psychology of language.
McGraw Hill. [aMJP]
Fowler, C. A., Brown, J. M., Sabadini, L. & Weihing, J. (2003) Rapid access to
speech gestures in perception: Evidence from choice and simple responsetime tasks. Journal of Memory and Language 49:396–413. [rMJP]
Fowler, C. & Housum, J. (1987) Talkers’ signaling “new” and “old” words in speech
and listeners’ perception and use of the distinction. Journal of Memory and
Language 26:489–504. [aMJP]
Frazier, L. & Rayner, K. (1990) Taking on semantic commitments: Processing
multiple meanings vs. multiple senses. Journal of Memory and Language
29:181–200. [rMJP]
Frisson, S. & Pickering, M. J. (1999) The processing of metonymy: Evidence from
eye-movements. Journal of Experimental Psychology: Learning, Memory, and
Cognition 25:1366–83. [rMJP]
(2001) Obtaining a figurative interpretation of a word: Support for
underspecification. Metaphor and Symbol 16:149–71. [rMJP]
Fussell, S. R. & Krauss, R. M. (1989) Understanding friends and strangers: The
effects of audience design on message comprehension. Journal of
Experimental Social Psychology 19:509–26. [RMK]
(1992) Coordination of knowledge in communication: Effects of speakers’
assumptions about what others know. Journal of Personality and Social
Psychology 62:378–91. [SG, RMK, aMJP]
Fussell, S. R., Kraut, R. E. & Siegel, J. (2000) Coordination of communication:
Effects of shared visual context on collaborative work. In: Proceedings of the
CSCW 2000 Conference on Computer-Supported Cooperative Work. ACM
Press. [SRF]
Fussell, S. R., Setlock, L. D. & Kraut, R. E. (2003) Effects of head-mounted and
scene-oriented video systems on remote collaboration on physical tasks. In:
Proceedings of the CHI 2003 Conference on Human Factors in Computing
Systems. ACM Press [SRF]
Gagné, C. L. (2001) Relation and lexical priming during the interpretation of
noun-noun combinations. Journal of Experimental Psychology: Learning,
Memory, and Cognition 27:236–54. [aMJP]
Gagné, C. L. & Shoben, E. J. (2002) Priming relations in ambiguous noun-noun
combinations. Memory and Cognition 30:637–46. [aMJP]
Garrett, M. (1980) Levels of processing in speech production. In: Language
production vol. 1, ed. B. Butterworth. Academic Press. [aMJP]
Garrod, S. (1999) The challenge of dialogue for theories of language processing.
In: Language processing, ed. S. Garrod & M. Pickering. Psychology Press.
[aMJP]
Garrod, S. & Anderson, A. (1987) Saying what you mean in dialogue: A study in
conceptual and semantic co-ordination. Cognition 27:181–218. [SRF,
PGTH, arMJP, TW]
Garrod, S. & Clark, A. (1993) The development of dialogue co-ordination skills in
schoolchildren. Language and Cognitive Processes 8:101–26. [HPB, aMJP]
Garrod, S. & Doherty, G. (1994) Conversation, co-ordination and convention: An
empirical investigation of how groups establish linguistic conventions.
Cognition 53:181–215. [PGTH, aMJP]
Garrod, S. & Pickering, M. J. (2004) Why is conversation so easy? Trends in
Cognitive Sciences 8:8–11. [rMJP]
Garrod, S. & Sanford, A. J. (1977) Interpreting anaphoric relations: The integration
of semantic information while reading. Journal of Verbal Learning and Verbal
Behavior 16:77–90. [aMJP]
Gaskell, M. G. & Marslen-Wilson, W. D. (1997) Integrating form and meaning: A
distributed model of speech perception. Language and Cognitive Processes
12:613–56. [SDG]
Gazdar, G., Klein, E., Pullum, G. & Sag, I. A. (1985) Generalized phrase structure
grammar. Blackwell. [aMJP]
Gentner, D. (1983) Structure-mapping: A theoretical framework for analogy.
Cognitive Science 7:155–70. [ABM]
Gentner, D. & Markman, A. B. (1997) Structural alignment in analogy and
similarity. American Psychologist 52(1):45–56. [ABM,arMJP]
Gerrig, R. J. & Bortfeld, H. (1999) Sense creation in and out of discourse contexts.
Journal of Memory and Language 41:457–68. [aMJP]
Giles, H., Coupland, N. & Coupland, J. (1991) Accommodation theory:
Communication, context, and consequence. In: Contexts of accommodation:
BEHAVIORAL AND BRAIN SCIENCES (2004) 27:2
53
References/Pickering and Garrod: Toward a mechanistic psychology of dialogue
Developments in applied sociolinguistics, ed. H. Giles, J. Coupland, & N.
Coupland, pp. 1– 68. Cambridge University Press. [SDG, RMK, arMJP]
Giles, H. & Powesland, P. F. (1975) Speech styles and social evaluation. Academic
Press. [aMJP]
Ginzburg, J. (1996) Interrogatives: Questions, facts, and dialogue. In: Handbook of
contemporary semantic theory, ed. S. Lappin. Blackwell. [JG]
(1997a) On some semantic consequences of turn taking. In: Proceedings of the
11th Amsterdam Colloquium on Formal Semantics and Logic, ed. P. Dekker,
M. Stokhof & Venema, pp. 145 –150. ILLC. Available at:
http://www.dcs.kcl.ac.uk/staff/ginzburg. [JG]
(1997b) Structural mismatch in dialogue. In: Proceedings of MunDial 97
(Technical Report 97–106), ed. G. Jaeger & A. Benz, pp. 59 –80. Universität
München Centrum für Informations- und Sprachverarbeitung. Available at:
http://www.dcs.kcl.ac.uk/staff/ginzburg. [JG]
(1999) Ellipsis resolution with syntactic presuppositions. In: Computing meaning
1: Current issues in computational semantics, ed. H. Bunt & R. Muskens.
Kluwer. [aMJP]
(2001) Fragmenting meaning: Clarification ellipsis and nominal anaphora. In:
Computing meaning 2: Current issues in computational semantics, ed. H.
Bunt. Kluwer. [aMJP]
(2002) Disentangling public from private meaning. In: Advances in discourse
and dialogue, ed. J. van Kuppevelt & R. Smith. Kluwer. Available at:
http://www.dcs.kcl.ac.uk/staff/ginzburg. [JG]
(forthcoming) Semantics and interaction in dialogue. CSLI Publications and
University of Chicago Press. Draft chapters available at:
http://www.dcs.kcl.ac.uk/staff/ginzburg. [JG]
Ginzburg, J. & Cooper, R. (in press) Clarification, ellipsis, and the nature of
contextual updates. Linguistics and Philosophy. Available at:
http://www.dcs.kcl.ac.uk/staff/ginzburg. [JG]
Ginzburg, J. & Sag, I. A. (2001) Interrogative investigations. CSLI. [aMJP]
Goldberg, A. (1995) Constructions. University of Chicago Press. [PFD]
Goldinger, S. D. (1998) Echoes of echoes? An episodic theory of lexical access.
Psychological Review 105:251–79. [SDG, MK, RMK, arMJP]
Goldinger, S. D. & Azuma, T. (in press) Puzzle-solving science: The quixotic quest
for units in speech perception. Journal of Phonetics. [SDG]
Goodwin, C. (1979) The interactive construction of a sentence in natural
conversation. In: Everyday language: Studies in ethnomethodology, ed. G.
Psathas, pp. 97–121. Irvington. [EAS]
Graesser, A. C., Weimer-Hastings, P. & Weimer-Hastings, K. (2001) Constructing
inferences and relations during text comprehension. In: Text representation:
Linguistic and psycholinguistic aspects, ed. T. Sanders, J. Schilperoord & W.
Spooren, pp. 249 –272. Benjamins. [ABM]
Gregory, S. W & Webster, S. (1996) A nonverbal signal in voices of interview
partners effectively predicts communication accommodation and social status.
Journal of Personality and Social Psychology 70:1231– 40. [RMK]
Grossberg, S. (1980) How does a brain build a cognitive code? Psychological
Review 87:1– 51. [SDG, rMJP]
(1999) The link between brain learning, attention, and consciousness.
Consciousness and Cognition 8:1– 44. [SDG]
Grossberg, S., Boardman, I. & Cohen, M. (1997) Neural dynamics of variable-rate
speech categorization. Journal of Experimental Psychology: Human
Perception and Performance 23:483 – 503. [SDG]
Grossberg, S. & Myers, C. W. (2000) The resonant dynamics of speech perception:
Interword integration and duration-dependent backward effects.
Psychological Review 107:735 – 67. [SDG]
Grossberg, S. & Stone, G. O. (1986) Neural dynamics of word recognition and
recall: Priming, learning, and resonance. Psychological Review 93:46–74.
[SDG]
Gunlogson, C. (2003) True to form: Rising and falling declaratives as questions in
English. Routledge. [SB-S]
Halpern, Y. & Moses, Y. (1990) Knowledge and common knowledge in a distributed environment. Journal of the ACM 37:549 – 87. [aMJP]
Hamblin, C. L. (1970) Fallacies. Methuen. [JG]
Hanna, J. E. & Tanenhaus, M. K. (2004) Pragmatic effects on reference resolution
in a collaborative task: Evidence from eye movements. Cognitive Science
28:105 –115. [SEB]
Hanna, J. E., Tanenhaus, M. K. & Trueswell, J. C. (2003) The effects of common
ground and perspective on domains of referential interpretation. Journal of
Memory and Language 49(1):43 – 61. [SEB, arMJP, SB-S]
Hartsuiker, R. J. & Kolk, H. H. J. (2001) Error monitoring in speech production: A
computational test of the perceptual loop theory. Cognitive Psychology
42:113 – 57. [aMJP]
Hartsuiker, R. J. & Westenberg, C. (2000) Persistence of word order in written and
spoken sentence production. Cognition 75:B27– 39. [arMJP]
Haviland, S. E. & Clark, H. H. (1974) What’s new? Acquiring new information as a
process in comprehension. Journal of Verbal Learning and Verbal Behavior
13:512–21. [aMJP]
Healey, P. G. T. (1997) Expertise or expertese?: The emergence of task-oriented
54
BEHAVIORAL AND BRAIN SCIENCES (2004) 27:2
sub-languages. In: Proceedings of the 19th Annual Conference of the Cognitive
Science Society, 7th–10th August, Stanford University, California, ed. M. G.
Shafto & P. Langley, pp. 301–306. Erlbaum. [PGTH]
(in preparation) Semantic co-ordination in dialogue: Communication as a special
case of misunderstanding. Unpublished manuscript. [PGTH]
Heyes, C. M. (2001) Causes and consequences of imitation. Trends in Cognitive
Sciences 5:253–61. [aMJP]
Hintzman, D. L. (1986) “Schema-abstraction” in a multiple trace model.
Psychological Review 93:411–28. [MK]
Holtgraves, T. (1994) Communication in context: Effects of speaker status on the
comprehension of indirect requests. Journal of Experimental Psychology:
Learning, Memory, and Cognition 20:1205–18. [HS]
Hommel, B., Müsseler, J., Aschersleben, G. & Prinz, W. (2001) The theory of event
coding (TEC): A framework for perception and action planning. Behavioral
and Brain Sciences 24:849–937. [JCC, arMJP]
Horton, W. S. & Keysar, B. (1996) When do speakers take into account common
ground? Cognition 59:91–117. [SEB, arMJP]
Hurley, S. & Chater, N., eds. (in press) Perspectives on imitation: From mirror
neurons to memes, vols. 1 & 2. MIT Press. [arMJP]
Iacoboni, M., Woods, R. P., Brass, M., Bekkering, H., Mazziotta, J. C. & Rizzolatti,
G. (1999) Cortical mechanisms of human imitation. Science 286:2526 –28.
[aMJP]
Isaacs, E. A. & Clark, H. H. (1987) References in conversations between experts
and novices. Journal of Experimental Psychology: General 116:26 – 37.
[rMJP]
Jackendoff, R. (1997) The architecture of the language faculty. MIT Press.
[aMJP]
(1999) Parallel constraint-based generative theories of language. Trends in
Cognitive Sciences 3:393–400. [aMJP]
(2002) Foundations of language. Oxford University Press. [arMJP]
Jansma, B. M. & Schiller, N. O. (2004) Monitoring syllable boundaries during
speech production. Brain and Language 90:311–17. [NOS]
Jefferson, G. (1974) Error correction as an interactional resource. Language in
Society 2:181–99. [EAS]
(1980) On “trouble-premonitory” response to inquiry. Sociological Inquiry
50(3–4):153–85. [EAS]
(1986) Notes on “latency” in overlap onset. Human Studies 9:153–83. [EAS]
(1987) On exposed and embedded corrections in conversation. In: Talk and
social organization, ed. G. Button & J. R. E. Lee. Multilingual Matters.
[aMJP, EAS]
Johnson, K., Strand, E. A. & D’Imperio, M. (1999) Auditory-visual integration of
talker gender in vowel perception. Journal of Phonetics 27(4):359–84. [HS]
Johnson-Laird, P. N. (1983) Mental models: Toward a cognitive science of language,
inference and consciousness. Harvard University Press. [RK, arMJP]
Just, M. A. & Carpenter, P. A. (1980) A theory of reading: From eye fixations to
comprehension. Psychological Review 87:329–54. [aMJP]
Kaplan, R. & Bresnan, J. (1982) Lexical-functional grammar: A formal system for
grammatical representation. In: The mental representation of grammatical
relations, ed. J. Bresnan. MIT Press. [aMJP]
Kaschak, M. P. (2003) This syntax needs learned: Adult acquisition of novel
syntactic constructions. Unpublished doctoral dissertation, University of
Wisconsin–Madison. [MK]
Kellas, G. & Vu, H. (1999) Strength of context does modulate the subordinate bias
effect: A reply to Binder and Rayner. Psychonomic Bulletin and Review
6:511–17. [aMJP]
Kempen, G. & Huijbers, P. (1983) The lexicalization process in sentence
production and naming: Indirect election of words. Cognition 14:824 – 43.
[aMJP]
Kempson, R., Meyer-Viol, W. & Gabbay, D. (2001) Dynamic syntax. Blackwell.
[RK, aMJP]
Keysar, B. (1994) The illusory transparency of intention: Linguistic perspective
taking in text. Cognitive Psychology 26:165–208. [ABM]
(1997) Unconfounding common ground. Discourse Processes 24:253 –70.
[DJB, SEB]
Keysar, B., Barr, D. J. & Balin, J. A. (1998) Definite reference and mutual
knowledge: Process models of common ground in comprehension. Journal of
Memory and Language 39:1–20. [ABM]
Keysar, B., Barr, D. J., Balin, J. A. & Brauner, J. S. (2000) Taking perspective in
conversation: The role of mutual knowledge in comprehension. Psychological
Science 11:32–38. [aMJP]
Keysar, B., Barr, D. J., Balin, J. A. & Paek, T. S. (1998) Definite reference and
mutual knowledge: Process models of common ground in comprehension.
Journal of Memory and Language 39:1–20. [aMJP]
Keysar, B., Lin, S. H. & Barr, D. J. (2003) Limits on theory of mind use in adults.
Cognition 89:25–41. [rMJP]
Kingsbury, D. (1968) Manipulating the amount of information obtained from a
person giving directions. Unpublished honors thesis, Department of Social
Relations, Harvard University. [RMK, rMJP]
References/Pickering and Garrod: Toward a mechanistic psychology of dialogue
Krauss, R. M. & Fussell, S. R. (1996) Social psychological models of interpersonal
communication. In: Social psychology: A handbook of basic principles, ed. E.
T. Higgins & A. Kruglanski, pp. 655 –701. Guilford. [RMK]
Kraut, R. E., Fussell, S. R. & Siegel, J. (2003) Visual information as a
conversational resource in collaborative physical tasks. Human-computer
interaction 18:13 – 49. [SRF, rMJP]
Kraut, R. E., Gergle, D. & Fussell, S. R. (2002) The use of visual information in
shared visual spaces: Informing the development of virtual co-presence. In:
Proceedings of the CSCW 2002 Conference on Computer-Supported
Cooperative Work. ACM Press. [SRF]
Kuiper, K. (1996) Smooth talkers: The linguistic performance of auctioneers and
sportscasters. Erlbaum. [arMJP]
Lakin, J. L. & Chartrand, T. L. (2003) Using nonconscious behavioral mimicry to
create affiliation and rapport. Psychological Science 14:334 – 39. [rMJP]
Larsson, S. (2002) Issue based dialogue management. Doctoral dissertation,
Gothenburg University. [JG]
Lee, B. P. H. (2001) Mutual knowledge, background knowledge and shared beliefs:
Their roles in establishing common ground. Journal of Pragmatics 33:21–44.
[DJB]
Lerner, G. H. (1991) On the syntax of sentences-in-progress. Language in Society
20:441– 58. [EAS]
(1996) On the “semi-permeable” character of grammatical units in conversation:
Conditional entry into the turn space of another speaker. In: Interaction and
grammar, ed. E. Ochs, E. A. Schegloff & S. A. Thompson, pp. 238 –76.
Cambridge University Press. [EAS]
Levelt, W. J. M. (1983) Monitoring and self-repair in speech. Cognition 14:41–104.
[aMJP]
(1989) Speaking: From intention to articulation. MIT Press.
[DJB,JCC,SG,aMJP,NOS]
Levelt, W. J. M. & Kelter, S. (1982) Surface form and memory in question
answering. Cognitive Psychology 14:78 –106. [arMJP]
Levelt, W. J. M. & Maassen, B. (1981) Lexical search and order of mention in
sentence production. In: Crossing the boundaries in linguistics: Studies
presented to Manfred Bierwisch, ed. W. Klein & W. J. M. Levelt. Riedel.
[aMJP]
Levelt, W. J. M., Roelofs, A. & Meyer, A. S. (1999) A theory of lexical access in
speech production. Behavioral and Brain Sciences 22:1–75.
[JCC,SDG,aMJP,NOS]
Lewis, D. K. (1969) Convention: A philosophical study. Basil Blackwell/Harvard
University Press. [PGTH,aMJP]
(1979) Score keeping in a language game. In: Semantics from different points of
view, ed. R. Bauerle, pp. 172–187. Springer. [JG]
Liberman, A. M., Cooper, F. S., Shankweiler, D. P. & Studdert-Kennedy, M. (1967)
Perception of the speech code. Psychological Review 74:431– 61. [HS]
Liberman, A. M. & Whalen, D. H. (2000) On the relation of speech to language.
Trends in Cognitive Sciences 4:187– 96. [aMJP]
Lindblom, B. (1990) Explaining variation: A sketch of the H and H theory. In:
Speech production and speech modeling, ed. W. Hardcastle & A. Marchal.
Kluwer. [aMJP]
Linnell, P. (1998) Approaching dialogue: Talk, interaction, and contexts in a
dialogical perspective. Benjamins. [aMJP]
Lockridge, C. B. & Brennan, S. E. (2002) Addressees’ needs influence speakers’
early syntactic choices. Psychonomic Bulletin and Review 9:550–57.
[SEB,arMJP]
Lombardi, L. & Potter, M. C. (1992) The regeneration of syntax in short term
memory. Journal of Memory and Language 31:713 – 33. [aMJP]
MacDonald, M. C., Pearlmutter, N. J. & Seidenberg, M. S. (1994) Lexical
nature of syntactic ambiguity resolution. Psychological Review 101:676–703.
[MK]
MacKay, D. (1987) The organization of perception and action. Springer. [aMJP]
Magnuson, J. & Nusbaum, H. C. (1994) Some acoustic and nonacoustic conditions
that produce talker normalization. In: Proceedings of the Spring Meeting of
the Acoustical Society of Japan, Tokyo, March 1994, pp. 637– 38. Acoustical
Society of Japan. [HS]
Markman, A. B. (1997) Constraints on analogical inference. Cognitive Science
21(4):373 – 418. [ABM]
Markman, A. B. & Gentner, D. (1993) Structural alignment during similarity
comparisons. Cognitive Psychology 25:431– 67. [aMJP]
(2000) Structure mapping in the comparison process. American Journal of
Psychology 113(4):501– 38. [ABM]
Markman, A. B. & Makin, V. S. (1998) Referential communication and category
acquisition. Journal of Experimental Psychology: General 127:331–54.
[aMJP]
Marslen-Wilson, W. D. (1973) Linguistic structure and speech shadowing at very
short latencies. Nature 244:522–23. [aMJP]
Martin, R. C., Lesch, M. F. & Bartha, M. C. (1999) Independence of input and
output phonology in word processing and short-term memory. Journal of
Memory and Language 41:3 –29. [JCC]
Mattingly, I. G. & Liberman, A. M. (1988) Specialized perceiving systems for
speech and other biologically significant sounds. In: Auditory function, ed.
G. M. Edelman. Wiley. [aMJP]
Mazur, A. (1985) A biosocial model of status in face-to-face primate groups. Social
Forces 64:377–402. [AM]
McCarthy, J. (1990) Formalization of two puzzles involving knowledge. In:
Formalizing common sense: Papers by John McCarthy, ed. V. Lifschitz. Ablex.
[aMJP]
McClelland, J. L. & Elman, J. L. (1986) The TRACE model of speech perception.
Cognitive Psychology 18:1–86. [SDG]
McClelland, J. L., St. John, M. & Taraban, R. (1989) Sentence comprehension: A
parallel distributed processing approach. Language and Cognitive Processes
4:287–336. [SDG]
McDaniel, D. & Cowart, W. (1999) Experimental evidence for a minimalist
account of English resumptive pronouns. Cognition 70:B15–B24. [FF]
McKoon, G. & Ratcliff, R. (1992) Inference during reading. Psychological Review
99(3):440–66. [ABM]
Meltzoff, A. N. & Decety, J. (2003) What imitation tells us social cognition: A
rapprochement between developmental psychology and cognitive
neuroscience. Philosophical Transactions of the Royal Society of London
358:491–500. [rMJP]
Metzing, C. & Brennan, S. E. (2001) When conceptual pacts are broken: Partner
effects in the comprehension of referring expressions. In: Abstracts of the
Psychonomic Society, 42th Annual Meeting, Orlando, FL, November 2001,
p. 29. Psychonomic Society. [SEB]
(2003) When conceptual pacts are broken: Partner-specific effects on the
comprehension of referring expressions. Journal of Memory and Language
49:201–13. [SEB, rMJP]
Meyer, A. S. (1996) Lexical access in phrase and sentence production: Results from
picture-word interference experiments. Journal of Memory and Language
35:477–96. [aMJP]
Moore, J. (1995) Participating in explanatory dialogues. Bradford Books/MIT
Press. [JG]
Morales, M., Mundy, P., Delgado, C. E. F., Yale, M., Messinger, D., Neal, R. &
Schwartz, H. K. (2000) Responding to joint attention across the 6- through 24month age period and early language acquisition. Journal of Applied
Developmental Psychology 21(3):283–98. [PFD]
Morgan, J. L. (1973) Sentence fragments and the notion “sentence”. In: Issues in
linguistics: Papers in honor of Henry and Renée Kahane, ed. B. B. Kachru,
R. B. Lees, Y. Malkiel, A. Pietrangeli & S. Saporta. University of Illinois Press.
[aMJP]
Morgan, J. L. & Wheeldon, L. R. (2003) Syllable monitoring in internally and
externally generated English words. Journal of Psycholinguistic Research
32:269–96. [NOS]
Morton, J. (1969) Interaction of information in word recognition. Psychological
Review 76:165–78. [aMJP]
Moss, H. E. & Gaskell, G. M. (1999) Lexical semantic processing during speech.
In: Language processing, ed. S. Garrod & M. Pickering. Psychology Press.
[aMJP]
Mullennix, J. W. & Pisoni, D. B. (1990) Stimulus variability and processing
dependencies in speech perception. Perception and Psychophysics 47(4):379–
90. [HS]
Müller, O. & Hagoort, P. (2001) Semantic and syntactic properties of a word –
which are first in reading? In: Abstracts of the Annual Meeting of the
Cognitive Neuroscience Society, p. 124. Cognitive Neuroscience Society.
[NOS]
Muller-Gass, A., Gonthier, I., Desrochers, A. & Campbell, K. B. (2000) Multiple
P3 evidence of a two-stage process in word gender decision. NeuroReport
16:3527–31. [NOS]
Murphy, G. L. (1988) Comprehending complex concepts. Cognitive Science
12:529–62. [aMJP]
Myers, J. L. & O’Brien, E. J. (1998) Accessing the discourse representation during
reading. Discourse Processes 26:131–157. [MK]
Nadig, J. S. & Sedivy, J. C. (2002) Evidence of perspective-taking constraints in
children’s on-line reference resolution. Psychological Science 13:329 – 36.
[SEB, arMJP]
Natale, M. (1975a) Convergence of mean vocal intensity in dyadic communication
as a function of social desirability. Journal of Personality and Social
Psychology 32:790–804. [RMK]
(1975b) Social desirability as related to convergence of temporal speech
patterns. Perceptual and Motor Skills 40:827–39. [RMK]
Nearey, T. M. (1989) Static, dynamic, and relational properties in vowel
perception. Journal of the Acoustical Society of America 85(5):2088–113.
[HS]
Newtson, D. (1994) The perception and coupling of behavior waves. In:
Dynamical systems in social psychology, ed. R. Vallacher & A. Nowak,
pp. 139–67. Academic Press. [SDG]
Niedzielski, N. (1999) The effect of social information on the perception of
BEHAVIORAL AND BRAIN SCIENCES (2004) 27:2
55
References/Pickering and Garrod: Toward a mechanistic psychology of dialogue
sociolinguistic variables. Journal of Language and Social Psychology 18(1):62–
85. [HS]
Nunberg, G., Sag, I. A. & Wasow, T. (1994) Idioms. Language 70:491–538.
[aMJP]
Nusbaum, H. C. & Henly, A. S. (1989) Understanding speech from the perspective
of cognitive psychology. Paper presented at the Workshop on Spoken
Language Understanding, Department of Psychology, State University of New
York at Buffalo, 1989. [HS]
Nusbaum, H. C. & Magnuson, J. (1997) Talker normalization. In: Talker variability
in speech processing, ed. K. Johnson & J. W. Mullennix. Academic Press.
[HS]
Nusbaum, H. C. & Morin, T. M. (1992) Paying attention to differences among
talkers. In: Speech production, perception, and linguistic structure, ed. Y.
Tohkura, Y. Sagisaka & E. Vatikiotis-Bateson, pp. 113 – 34. Ohmasha
Publishing. [HS]
Nusbaum, H. C. & Schwab, E. C. (1986) The role of attention and active
processing in speech perception. In: Pattern recognition by humans and
machines: Speech perception, vol. 1, ed. H. C. Nusbaum & E. C. Schwab, pp
113 – 57. Academic Press. [HS]
Oomen, C. C. E. & Postma, A. (2002) Limitations in processing resources and
speech monitoring. Language and Cognitive Processes 17:163–84. [aMJP]
Peterson, G. & Barney, H. (1952) Control methods used in the study of vowels.
Journal of the Acoustical Society of America 24:175 – 84. [HS]
Phillips, C. (1996) Order and structure. Unpublished doctoral dissertation, MIT.
[RK]
(2003) Linear order and constituency. Linguistic Inquiry 34:37–90. [aMJP]
Pickering, M. & Barry, G. (1991) Sentence processing without empty categories.
Language and Cognitive Processes 6:229 – 59. [aMJP]
(1993) Dependency categorial grammar and coordination. Linguistics 31:855–
902. [aMJP]
Pickering, M. J. & Branigan, H. P. (1998) The representation of verbs: Evidence
from syntactic priming in language production. Journal of Memory and
Language 39:633 – 51. [MK, arMJP, NOS]
(1999) Syntactic priming in language production. Trends in Cognitive Sciences
3:136 – 41. [aMJP]
Pickering, M., Clifton, C. & Crocker, M. W. (2000) Architectures and mechanisms
in sentence comprehension. In: Architectures and mechanisms for language
processing, ed. M. W. Crocker & M. Pickering. Cambridge University Press.
[JCC]
Pickering, M. J. & Garrod, S. (2004) Routinization in the interactive-alignment
model of dialogue. Unpublished manuscript. [rMJP]
Pinker, S. (1989) Learnability and cognition: The acquisition of argument
structure. MIT Press/Bradford Books. [HPB]
Pinker, S. & Birdsong, D. (1979) Speakers’ sensitivity to rules of frozen word order.
Journal of Verbal Learning and Verbal Behavior 18:497– 508. [aMJP]
Poesio, M. & Traum, D. R. (1997) Conversational actions and discourse situations.
Computational Intelligence 13:309 – 47. [aMJP]
Pollard, C. & Sag, I. A. (1994) Head-driven phrase structure grammar. University
of Chicago Press and CSLI. [aMJP]
Postma, A. (2000) Detection of errors during speech production: A review of
speech monitoring models. Cognition 77:97–131. [aMJP]
Potter, M. C. & Lombardi, L. (1990) Regeneration in the short-term recall of
sentences. Journal of Memory and Language 29:633 – 54. [aMJP]
(1998) Syntactic priming in immediate recall of sentences. Journal of Memory
and Language 38:265 – 82. [aMJP]
Prat-Sala, M. & Branigan, H. P. (2000) Discourse constraints on syntactic
processing in language production: A cross-linguistic study in English and
Spanish. Journal of Memory and Language 42:168 – 82. [aMJP]
Prince, E. F. (1990) Syntax and discourse: A look at resumptive pronouns. In:
Proceedings of the Sixteenth Annual Meeting of the Berkeley Linguistics
Society, ed. K. Hall, pp. 482– 97. [FF]
Purver, M., Ginzburg, J. & Healey, P. (2002) On the means for clarification in
dialogue. In: Advances in discourse and dialogue, ed. J. van Kuppevelt & R.
Smith. Kluwer. Available at: http://www.dcs.kcl.ac.uk/pg/purver/
publications.html. [JG]
Purver, M., Healey, P. G. T., King, J., Ginzburg, J. & Mills, G. J. (2003) Answering
clarification questions. In: Proceedings of the 4th SIGdial Workshop on
Discourse and Dialogue. Sapporo. Available at: http://www.dcs.kcl.ac.uk/pg/
purver/publications.html. [JG]
Purver, M. & Otsuka, M. (2003) Incremental generation by incremental parsing:
Tactical generation in Dynamic Syntax. In: Proceedings of 9th EACL
(European Association for Computational Linguistics) Workshop in Natural
Language Generation, Budapest, April 2003, ed. E. Reiter, H. Horecek & K.
van Deemter, pp. 79 – 86. Association for Computational Linguistics. [RK]
Rayner, K., Pacht, J. M. & Duffy, S. A. (1994) Effects of prior encounter and
discourse bias on the processing of lexically ambiguous words. Journal of
Memory and Language 33:527– 44. [aMJP]
Reeves, B. & Nass, C. (1996) The media equation: How people treat computers,
56
BEHAVIORAL AND BRAIN SCIENCES (2004) 27:2
television, and new media like real people and places. Cambridge University
Press. [aMJP]
Rizzolatti, G. & Arbib, M. A. (1998) Language within our grasp. Trends in
Neurosciences 21:188–94. [aMJP]
Ross, J. R. (1969) Guess who? In: Papers from the Fifth Regional Meeting of the
Chicago Linguistics Society, ed. R. I. Binnick, A. Davison, G. M. Green &
J. L. Morgan. University of Chicago. [aMJP]
Russell, A. W. & Schober, M. F. (1999) How beliefs about a partner’s goals affect
referring in goal-discrepant conversations. Discourse Processes 27(1):1–33.
[MFS]
Sacks, H. (1972a) An initial investigation of the usability of conversational data for
doing sociology. In: Studies in social interaction, ed. D. N. Sudnow, pp. 31–
74. Free Press. [EAS]
(1972b) On the analyzability of stories by children. In: Directions in
sociolinguistics: The ethnography of communication, ed. J. J. Gumperz & D.
Hymes, pp. 325–55. Holt, Rinehart and Winston. [EAS]
(1975) Everyone has to lie. In: Sociocultural dimensions of language use, ed. M.
Sanches & B. G. Blount, pp. 57–80. Academic Press. [EAS]
(1987) On the preferences for agreement and contiguity in sequences in
conversation. In: Talk and social organization, ed. G. Button & J. R. E. Lee.
Multilingual Matters. [aMJP]
(1992) Lectures on conversation, ed. G. Jefferson, introduction by E. A.
Schegloff, (2 vols.). Blackwell. [EAS]
Sacks, H. & Schegloff, E. A. (1979) Two preferences in the organization of
reference to persons and their interaction. In: Everyday language: Studies in
ethnomethodology, ed. G. Psathas, pp. 15–21. Irvington. [EAS]
Sacks, H., Schegloff, A. E. & Jefferson, G. (1974) A simplest systematics for the
organization of turn-taking for conversation. Language 50:696 –735. [DJB,
aMJP, EAS]
Sancier, M. & Fowler, C. A. (1997) Gestural drift in a bilingual speaker of Brazilian
Portuguese and English. Journal of Phonetics 25:421–36. [RMK]
Sanford, A. J. & Garrod, S. C. (1981) Understanding written language. Wiley.
[aMJP]
Sanford, A. J. & Sturt, P. (2002) Depth of processing in language comprehension:
Not noticing the evidence. Trends in Cognitive Sciences 6:382–86. [rMJP]
Schegloff, E. A. (1972) Notes on a conversational practice: Formulating place. In:
Studies in social interaction, ed. D. N. Sudnow, pp. 75–119. Free Press.
[EAS]
(1979) The relevance of repair for syntax-for-conversation. In: Syntax and
semantics 12: Discourse and syntax, ed. T. Givon, pp. 261–88. Academic
Press. [EAS]
(1986) The routine as achievement. Human Studies 9:111–51. [EAS]
(1987a) Recycled turn beginnings: A precise repair mechanism in conversation’s
turn-taking organisation. In: Talk and social organisation, ed. G. Button &
J. R. E. Lee, pp. 70–85. Multilingual Matters. [EAS]
(1987b) Some sources of misunderstanding in talk-in-interaction. Linguistics
25:201–18. [EAS]
(1989) Reflections on language, development, and the interactional character of
talk-in-interaction. In: Interaction in human development, ed. M. Bornstein &
J. S. Bruner, pp. 139–53. Erlbaum. [EAS]
(1990) On the organization of sequences as a source of “coherence” in talk-ininteraction. In: Conversational organization and its development, ed. B.
Dorval, pp. 51–77. Ablex. [EAS]
(1991) Conversation analysis and socially shared cognition. In: Perspectives on
socially shared cognition, ed. L. Resnick, J. Levine & S. Teasley, pp. 151–71.
American Psychological Association. [EAS]
(1992) Repair after next turn: The last structurally provided place for the
defence of intersubjectivity in conversation. American Journal of Sociology
95(5):1295–345. [EAS]
(1995) Sequence organization. Unpublished manuscript, Department of
Sociology, University of California at Los Angeles. [EAS]
(1996a) Turn organization: One intersection of grammar and interaction. In:
Interaction and grammar, ed. E. Ochs, E. A. Schegloff & S. A. Thompson,
pp. 52–133. Cambridge University Press. [EAS]
(1996b) Some practices for referring to persons in talk-in-interaction: A partial
sketch of a systematics. In: Studies in anaphora, ed. B. A. Fox, pp. 437– 85.
John Benjamins. [EAS]
(1997a) Practices and actions: Boundary cases of other-initiated repair. Discourse
Processes 23:499–545. [EAS]
(1997b) Third turn repair. In: Towards a social science of language: Papers in
honor of William Labov, vol 2: Social interaction and discourse structures, ed.
G. R. Guy, C. Feagin, D. Schiffrin & J. Baugh, pp. 31–40. John Benjamins.
[EAS]
(2000a) Overlapping talk and the organization of turn-taking for conversation.
Language in Society 29(1):1–63. [EAS]
(2000b) When “others” initiate repair. Applied Linguistics 21(2):205–43.
[EAS]
(2001) Accounts of conduct in interaction: Interruption, overlap and turn-taking.
References/Pickering and Garrod: Toward a mechanistic psychology of dialogue
In: Handbook of sociological theory, ed. J. H. Turner, pp. 287– 321. Plenum
Press. [EAS]
(forthcoming) A primer in conversation analysis: Sequence organization.
Cambridge University Press. [EAS]
Schegloff, E. A., Jefferson, G. & Sacks, H. (1977) The preference for selfcorrection in the organization of repair in conversation. Language 53(2):361–
82. [EAS]
Schegloff, E. A. & Sacks, H. (1973) Opening up closings. Semiotica 8:289–327.
[aMJP, EAS]
Schenkein, J. (1980) A taxonomy of repeating action sequences in natural
conversation. In: Language production, vol. 1, ed. B. Butterworth, pp. 21–47.
Academic Press. [aMJP]
Schiffer, S. R. (1972) Meaning. Oxford University Press. [aMJP]
Schiller, N. O. (2001) Metrical encoding during speech production. Abstracts of
the Psychonomic Society 6:29. [NOS]
(in press) Verbal self-monitoring. In: Twenty-first century psycholinguistics:
Four cornerstones, ed. A. Cutler. Erlbaum. [NOS]
Schiller, N. O. & Caramazza, A. (2003) Grammatical feature selection in noun
phrase production: Evidence from German and Dutch. Journal of Memory
and Language 48:169 – 94. [rMJP, NOS]
Schiller, N. O., Jansma, B. M., Peters, J. & Levelt, W. J. M. (in press) Monitoring
metrical stress in polysyllabic words. Language and Cognitive Processes.
[NOS]
Schiller, N. O., Münte, T., Horemans, I. & Jansma, B. M. (2003) Semantic and
phonological factors in syntactic decisions. Psychophysiology 40:869–77.
[NOS]
Schmitt, B. M., Rodriguez-Fornells, A., Kutas, M. & Münte, T. F. (2001a)
Electrophysiological estimates of semantic and syntactic information access
during tacit picture naming and listening to words. Neuroscience Research
41:293 – 98. [NOS]
Schmitt, B. M., Schiltz, K., Zaake, W., Kutas, M. & Münte, T. F. (2001b) An
electrophysiological analysis of the time course of conceptual and syntactic
encoding during tacit picture naming. Journal of Cognitive Neuroscience
13:510 –22. [NOS]
Schober, M. F. (1993) Spatial perspective-taking in conversation. Cognition 47:1–
24. [aMJP]
(1995) Speakers, addressees, and frames of reference: Whose effort is minimized
in conversations about location? Discourse Processes 20(2):219 – 47. [MFS]
(1998a) Different kinds of conversational perspective-taking. In: Social and
cognitive psychological approaches to interpersonal communication, ed. S. R.
Fussell & R. J. Kreuz, pp. 145 –74. Erlbaum. [MFS]
(1998b) How partners with high and low spatial ability choose perspectives in
conversation. Abstracts of the 39th Annual Meeting of the Psychonomic
Society, No. 39. [MFS]
Schober, M. F. & Brennan, S. E. (2003) Processes of interactive spoken discourse:
The role of the partner. In: Handbook of discourse processes, ed. A. C.
Graesser, M. A. Gernsbacher & S. R. Goldman, pp. 123 – 64. Erlbaum.
[SEB, aMJP, MFS]
Schober, M. F. & Clark, H. H. (1989) Understanding by addressees and overhearers. Cognitive Psychology 21:211– 32. [DJB, aMJP]
Schober, M. F. & Conrad, F. G. (1997) Does conversational interviewing reduce
survey measurement error? Public Opinion Quarterly 61:576 – 02. Reprinted
in: Interviewing, vol. 1, ed. N. G. Fielding. SAGE Benchmarks in Social
Science Research Series, 2003. Sage Publications. [MFS]
Schober, M. F., Conrad, F. G. & Fricker, S. S. (2004) Misunderstanding
standardized language in research interviews. Applied Cognitive Psychology
18:169 – 88. [MFS]
Schriefers, H., Meyer, A. S. & Levelt, W. J. M. (1990) Exploring the time course of
lexical access in language production: Picture-word interference studies.
Journal of Memory and Language 29:86 –102. [aMJP]
Schuller, A. M. & Rossion, B. (2001) Spatial attention triggered by eye gaze
increases and speeds up early visual acuity. NeuroReport 12:2381–86.
[rMJP]
Searle, J. (1969) Speech acts. Cambridge University Press. [JG]
Sereno, S. C., Brewer, C. & O’Donnell, P. J. (2003) Context effects in word
recognition: Evidence for early interactive processing. Psychological Science
14:328 – 33. [SG]
Sheldon, A. (1974) The role of parallel function in the acquisition of relative
clauses in English. Journal of Verbal Learning and Verbal Behavior 13:272–
81. [aMJP]
Shiffrin, R. M. & Schneider, W. (1977) Controlled and automatic human
information processing: II. Perceptual learning, automatic attending and a
general theory. Psychological Review 84(2):127– 90. [HS]
Shockley, K., Santana, M-V. & Fowler, C. A. (2003) Mutual interpersonal postural
constraints are involved in cooperative conversation. Journal of Experimental
Psychology: Human Perception and Performance 29:326 – 32. [SDG]
Skinner, B. F. (1953) Science and human behavior. Macmillan. [JJP]
(1957) Verbal behavior. Appleton-Century-Crofts. [JJP, rMJP]
Smith, M. & Wheeldon, L. (2001) Syntactic priming in spoken sentence
production – an online study. Cognition 78:123–64. [aMJP]
Smyth, R. (1994) Grammatical determinants of ambiguous pronoun resolution.
Journal of Psycholinguistic Research 23:197–229. [aMJP]
Sperber, D. & Wilson, D. (1986) Relevance. Blackwell. [ABM]
Stalnaker, R. C. (1978) Assertion. In: Syntax and semantics, vol. 9: Pragmatics, ed.
P. Cole, pp. 315–32. Academic Press. [JG, aMJP]
Steedman, M. (2000) The syntactic process. MIT Press. [aMJP]
Suessbrick, A. L., Schober, M. F. & Conrad, F. G. (2000) Different respondents
interpret ordinary questions quite differently. In: Proceedings of the American
Statistical Association (Section on Survey Research Methods). American
Statistical Association. [MFS]
Swinney, D. A. (1979) Lexical access during sentence comprehension. Journal of
Verbal Learning and Verbal Behavior 18:645–59. [aMJP]
Tanenhaus, M. K., Chambers, C. C. & Hanna, J. E. (2004) Referential domains in
spoken language comprehension: Using eye movements to bridge the product
and action traditions. In: The interface of language, vision, and action: Eye
movements and the visual world, ed. J. M. Henderson & F. Ferreira.
Psychology Press. [SB-S]
Tanenhaus, M. K., Spivey-Knowlton, M. J., Eberhard, K. M. & Sedivy, J. C. (1995)
Integration of visual and linguistic information in spoken language
comprehension. Science 268(5217):632–34. [aMJP, HS]
Tannen, D. (1989) Talking voices: Repetition, dialogue, and imagery in
conversational discourse. Cambridge University Press. [aMJP]
Thakerar, J. N., Giles, H. & Chesire, J. (1982) Psychological and linguistic
parameters of speech accommodation theory. In: Advances in the social
psychology of language, ed. C. Fraser & K. R. Scherer, pp. 205–255.
Cambridge University Press. [HS]
Tomasello, M. (2000) Do young children have adult syntactic competence?
Cognition 74:209–53. [HPB, aMJP]
(2003) Constructing a language: A usage-based theory of language acquisition.
Harvard University Press. [PFD, rMJP]
Traxler, M. J. & Gernsbacher, M. A. (1992) Improving written communication
through minimal feedback. Language and Cognitive Processes 7:1–22.
[aMJP]
(1993) Improving written communication through perspective-taking. Language
and Cognitive Processes 8:311–36. [aMJP]
Tucker, G. R., Lambert, W. E. & Rigault, A. A. (1977) The French speaker’s skill
with grammatical gender: An example of rule-governed behavior. Mouton.
[NOS]
van Dijk, T. A. & Kintsch, W. (1983) Strategies in discourse comprehension.
Academic Press. [aMJP]
van Gelder, T. & Port, R. F. (1995) It’s about time: An overview of the dynamical
approach to cognition. In: Mind as motion, ed. R. F. Port & T. van Gelder,
pp. 1–44. MIT Press. [SDG]
van Turennout, M., Hagoort, P. & Brown, C. M. (1998) Brain activity during
speaking: From syntax to phonology in 40 milliseconds. Science 280:572–74.
[aMJP, NOS]
Vigliocco, G., Antonini, T. & Garrett, M. F. (1997) Grammatical gender is on the
tip of Italian tongues. Psychological Science 8:314–17. [arMJP]
Vitevitch, M. S. & Luce, P. A. (1999) Probabilistic phonotactics and neighborhood
activation in spoken word recognition. Journal of Memory and Language
40:374–408. [SDG]
Wheeldon, L. & Levelt, W. J. M. (1995) Monitoring the time course of
phonological encoding. Journal of Memory and Language 34:311–34.
[aMJP, NOS]
Wheeldon, L. & Morgan, J. L. (2002) Phoneme monitoring in internal and external
speech. Language and Cognitive Processes 17:503–35. [NOS]
Wheeldon, L. R. & Smith, M. C. (2003) Phrase structure priming: A short-lived
effect. Language and Cognitive Processes 18:431–42. [rMJP]
Wilkes-Gibbs, D. & Clark, H. H. (1992) Coordinating beliefs in conversation.
Journal of Memory and Language 31:183–94. [DJB, aMJP]
Williams, J. H. G., Whiten, A., Suddendorf, T. & Perrett, D. I. (2001) Imitation,
mirror neurons, and autism. Neuroscience and Biobehavioral Reviews 25:287–
95. [aMJP]
Wisniewski, E. L. (1996) Construal and similarity in conceptual combination.
Journal of Memory and Language 35:434–53. [aMJP]
Wray, A. & Perkins, M. R. (2001) The functions of formulaic language: An
integrated model. Language and Communication 20:1–28. [aMJP]
Zwaan, R. A. & Radvansky, G. A. (1998) Situation models in language
comprehension and memory. Psychological Bulletin 123:162–85. [aMJP]
Zwitserlood, P. (1994) Access to phonological-form representations in language
comprehension and production. In: Perspectives on sentence processing, ed.
C. Clifton, L. Frazier & K. Rayner, pp. 83–106. Erlbaum. [JCC]
BEHAVIORAL AND BRAIN SCIENCES (2004) 27:2
57
Was this manual useful for you? yes no
Thank you for your participation!

* Your assessment is very important for improving the work of artificial intelligence, which forms the content of this project

Download PDF

advertisement