Reciprocity in Spanish: Two Puzzles of Scope 1.0. Introduction'

Reciprocity in Spanish: Two Puzzles of Scope 1.0. Introduction'
Reciprocity in Spanish: Two Puzzles of Scope
Chip Geffen
1.0. Introduction'
Heim, Lasnik, and May (1991a, henceforth HLMa) note an interesting contrast in the
behavior of the following sentences in English (their 78a -b):
1.a) They look like each other.
b) They look alike.
As HLMa point out, when embedded, the two sentences have distinct properties (their 79a -b):
2.a) John and Mary think they look like each other.
b) John and Mary think they look alike.
Sentence (2a) is ambiguous between broad and narrow scope interpretations. Thus, (2a) can either
mean 'John thinks he looks like Mary, and Mary thinks that she looks like John' (the broad
reading) or 'John and Mary think they (John and Mary) look like each other' (the narrow
reading). In contrast, (2b) can only be construed with narrow scope. For HLMa the ambiguity
of (2a) receives an explanation in terms of the morphological complexity of the reciprocal
expression each other. Specifically, the quantificational distribution element each is adjoined to
an antecedent, which is then subject to QR via the rule move -a at logical form (see May 1977,
1985). Put simply, this allows for different scope interpretations, depending on how far up the
phrase marker each is moved. In contrast, the morphologically simplex alike contains no
detachable distribution element, and, as a result, only the narrow scope reading is available.
Of interest here is the fact that HLMa base their argument on the distinction between
reciprocal meaning that is incorporated within a morphologically simplex versus a
morphologically complex item. In support of this claim, they offer the following minimal pair
of sentences from Italian (attributed to Luigi Rizzi):2
3.a) I due pensano [di essersi
battuti] (contradictory)
the two thought be-each other -clitic beaten
b) I due pensano [di avere prevalso l'uno sull'altro] (ambiguous)
the two thought
have prevailed the one over the other
HLMa note that when taken by themselves, the embedded clauses in (3a -b) are both
'I would like to thank Andy Barss, Molly Diesing, Ander Moina, Pilar Pifiar, and Laura Conway for help with
the issues addressed here. I also thank Pilar Pifiar, Montse Sanz, Rosa García, Raquel Mejía, Jorge Lemus and Esther
Lemus for their native speaker intuitions on the examples used throughout. All errors are, of course, my own.
zIt is not altogether clear exactly what is meant by "minimal pair" here. A significant difference between the two
sentences would seem to be that the embedded verb meaning 'beat' in (3a) subcategorizes for an accusative NP,
while the verb meaning 'prevail' does not take an NP complement. The presence of the reciprocal clitic would thus
be impossible in (3b). The question of whether a non -contradictory construal of (3a) is available with the full
reciprocal remains open.
36
contradictory, but that only (3b) receives a non -contradictory reading in the embedded
construction. In a manner analogous to their treatment of the English data in (1 -2), HLMa claim
that this distinction is attributable to the fact that the clitic in (3a) forms a morphological unit
with the verb to which it is attached and, thus, cannot be moved at LF. In contrast, they follow
Belletti (1982) in arguing that the full form of the Italian reciprocal l'uno...l'altro includes a
distributor l'uno which can be detached and moved at LF. Though no specific analysis is
provided, it is assumed that the broad scope, and hence non -contradictory, construal of (3b) is
attributable to the adjunction of the distributor l'uno to the antecedent I due.
With these facts in mind, I consider the question of scope in Spanish reciprocal
constructions. In sections 2 and 3, I present a surprising scope asymmetry between non -full
(clitic) and full reciprocal constructions, which indicates that unlike English, the full reciprocal
el uno al otro in Spanish does not allow for broad scope interpretations when embedded. In
section 4, I argue that el uno al otro in Spanish is best analyzed as an adjunct, rather than as the
subcategorized argument of the verb. And in section 5, I explore HLM's (1991b) "each- binding"
variant of the movement analysis proposed in HLMa, showing that the asymmetry between full
and non -full reciprocals can be accounted for in terms of the obligatory local A'- binding of the
variable el uno of the adjoined full form.
In section 6, I expand the data, providing evidence of another scope asymmetry.
Specifically, I show that in contrast to the el uno al otro adjunct of the clitic doubled
construction, VP adjuncts such as prepositional phrases with a reciprocal object do allow broad
construals from embedded clauses. I argue that this asymmetry motivates the need to formally
distinguish between at least two types of adjuncts, appositional adjuncts such as the doubled el
uno al otro construction, and standard adjuncts such as PPs. I suggest that a profitable way of
making this distinction can be found in restricting the assignment of referential indexes in the
Relativized Minimality framework (Rizzi 1990). This approach both preserves the account of the
asymmetry between non -full or clitic reciprocals and their doubled counterparts, as allows for
broad construals from standard adjuncts.
2.0 Scope Ambiguities and Spanish se
In looking at Spanish reciprocals, the HLMa analysis predicts that we should encounter
similar scope restrictions for non -full reciprocal forms, that is, constructions containing only the
reciprocal clitic3. Quite simply, if the clitic forms a morphological unit with the verb, no
distribution element such as the English each is available for movement. Thus, embedded non full reciprocals should be limited to narrow scope interpretations. In fact, this is not the case.
Embedded clitic reciprocal constructions in Spanish systematically permit broad scope
interpretations. Consider, for example, the data in 4 -5, where (4b and 5b) can be considered
structurally analogous to the Italian (3a).
4.a) Juan y María creían que pro se habían visto.
and M
thought that
cl. had
seen
'Juan and María thought they had seen each other'
J
3As with other romance languages such as French and Italian, the reciprocal clitic in Spanish is homophonous
with the reflexive clitic. Note that for simplicity I refer only to the 3rd person clitic "se" throughout.
37
b) Juan y María creían PRO haberse visto.
have -cl seen
and M
thought
'Juan and María thought they had seen each other'
J
5.a) Franco y Carrillo juraban que pro se odiaban.
cl- hate
F
swore that
and C
'Franco and Carrillo swore that they hated each other.'
b) Franco y Carrillo juraban PRO odiarse.
F
hate -cl
and C
swore
'Franco and Carrillo swore that they hated each other.'
Both (4) and (5) are clearly ambiguous between broad and narrow construals. Thus, (4)
can be interpreted as follows: 1) Juan thought he saw María, and María thought she saw Juan
(broad scope); 2) Juan and María thought they (Juan and María) saw each other (narrow scope).
Likewise, (5) can either mean: 1) Franco swore that he hated Carrillo, and Carrillo swore that
he hated Franco; or 2) both men swore that they (both of them) hated each other.
More evidence in support of the availability of wide scope interpretations is found in the
Spanish analogue to (3a). HLMa claim that Spanish speakers "spontaneously reject" the Spanish
analogue to this sentence as "somehow deviant ". However, they do not specify which of the
possible analogues they provided for their informants. The data in (6 -7) can both be considered
analogues of (3a), and both are ambiguous between contradictory (narrow scope) and non contradictory (broad scope) interpretations.
6. Emilio y Pedro creían PRO haberse vencido.
E
and P
have -cl defeated
thought
'Emilio and Pedro thought they had defeated each other'
7. Emilio y Pedro creían PRO haberse ganado.
E
and P
thought
have -cl won
'Emilio and Pedro thought they had won (over) each other'
In contrast to Italian, then, it is clear that Spanish systematically allows for broad scope
construals of embedded clitic reciprocals.
3.0 Scope and the Full Reciprocal el uno al otro
The ability of clitic reciprocals to take wide scope is, in fact, recognized by HLMa in a
footnote (n.17). They provide the following example:
8. Juan y María me confesaron secretamente que se gustaban. Ambos piensan que no son
correspondidos.
'Juan and María confessed to me secretly that they (clitic) liked each other. Both think that
their feelings (lit.: they) are unrequited.
Obviously, a broad construal must be available here; otherwise, the assertion that each
believes that s/he is unrequited would be anomalous. Of particular interest, however, is that HLM
38
claim that the circumstances under which clitics can take wide scope are distinct from those
under which wide scope is available for their non -clitic counterparts. As I have argued above,
clitic reciprocals regularly take broad scope. Ironically, however, it is not clear that non -clitic
reciprocals in Spanish do, in fact, permit regular broad scope construals when embedded.
Consider the data in (9). Note that unlike Italian, the reciprocal clitic obligatorily doubles the
reciprocal pronoun el uno al otro in the full form, an issue addressed in section 4 below.
9.a) Juan y María creían que pro se habían visto el uno al otro.
thought that
cl. had
seen the one the other
J and M
'Juan and María thought they had seen each other'
b) Juan y María creían PRO haberse visto el uno al otro.
J and M
to have -cl seen the one the other
thought
'Juan and María thought they had seen each other'
In contrast to the ambiguous readings available for (4a -b) above, (9a -b) can only be
construed with narrow scope. That is, the sentences in (9) can only mean that 'Juan and María
thought that they (Juan and Maria) saw each other.' If, by analogy with English and Italian
(Belletti 1982), we maintain that the full reciprocal contains an autonomous distributor el uno,
the lack of a wide scope interpretation for (9) is surprising. Paradoxically, in fact, the presence
of the full form enforces a narrow interpretation in sentences with embedded reciprocals. This
is seen clearly in the following full forms corresponding to (6 -7) above.
10. Emilio y Pedro creían PRO haberse vencido el uno al otro.
E
and P
thought
have -cl defeated the one the other
'Emilio and Pedro thought they had defeated each other'
11. Emilio y Pedro creían PRO haberse ganado el uno al otro.
have -cl won the one the other
and P
E
thought
'Emilio and Pedro thought they had won (over) each other'
Contrary to what is predicted by HLMa on the basis of the Italian data in (3), only
contradictory interpretations are available in (10 -11). In short, the Spanish reciprocal is doubly
puzzling: clitic reciprocals permit scope ambiguity in embedded sentences, while full reciprocal
forms can only be interpreted with narrow scope.
4.0 The Structure of the Reciprocal
In order to treat the issue of why full reciprocals prohibit broad scope interpretations, it
is useful to consider the structure of the full reciprocal form. As noted above, full reciprocals
require clitic doubling in Spanish, regardless of whether the verb subcategorizes for an accusative
or a dative object. In this sense, full reciprocal (and reflexive) forms pattern with pronouns in
triggering obligatory clitic doubling. Consider, for example, the familiar paradigms for non reciprocal clitics:
39
12. Accusative Clitics
Dative Clitics
a. Juan cortó el pan.
Juan cut the bread.
pegó a Juan.
d. Pedro le
Pedro cl -dat. hit Juan
'Pedro hit Juan'
b. Juan lo
e. Pedro le
pegó e.
Pedro cl -dat. hit
'Pedro hit him'
cortó e.
Juan cl -acc. cut
'Juan cut it'
c. *Juan lo
f. *Pedro pegó a Juan.
'Pedro hit Juan'
cortó el pan.
the bread
'John cut the bread'
John cl -acc cut
With respect to clitic doubling, accusative R- expression objects cannot be doubled, while
dative R- expression objects must be doubled.4 In (12a -c), the verb cortar takes an accusative
object. Thus, the accusative clitic lo cannot double the direct object el pan, as seen in (12c). In
contrast, the verb pegar in (12d -f) subcategorizes for a dative object. Here, the clitic must double
the indirect object, as seen by the ungrammaticality of (12f). In both cases, the clitic surfaces
when the object is an empty category, a fact which has led researchers such as Jaeggli (1986) to
argue that clitics in Spanish can absorb case.5
Interestingly, the distinction between accusative and dative verbs with regard to clitic
doubling is lost when the post -verbal NP is a pronoun. As shown in (12), the accusative -assigning
ver and the dative- assigning pegar must surface with a clitic when they take pronominal objects.
13. Accusative Clitics
Dative Clitics
a. La
vi
a ella.
cl -acc. saw -I her
'I saw her'
d. Le
b. Vi
e. Le
cl -dat. hit -I
him
'I hit him'
a María.
saw -I María
pegué a Juan.
cl -dat. hit -I
John
'I hit John'
'I saw Mary'
c. *Vi
pegué a él.
f. *Pegue a él.
a ella.
saw -I her
hit -I
'I saw her'
him
'I hit him
Of interest is the fact that in accusative forms, the appearance of a clitic is obligatory in
two environments: 1) when there is no overt, post - verbal argument, as in (12b); and 2) when the
4In the description of the data here, I limit myself to standard Peninsular Spanish, which does not allow clitic
doubling of accusative objects. See Jaeggli (1982,1986).
5 See Sutler (1987, 1988) for a different perspective on the role of clitics.
40
argument of the verb is pronominal, as seen in the contrast between (13a) and (13c).
Additionally, it is important to note that both dative and accusative clitic -doubled pronominal
objects are marked constructions in Spanish. The full pronoun is primarily used to mark a
referential contrast (see Jelinek 1984; Piñar 1991), and is thus given a phonological prominence
not found for non- pronominal objects. Bresnan and Mchombo (1987) point out that this type of
phonological prominence is one of the principal criteria for determining the status of such
doubled "arguments ", claiming that languages in which incorporated pronominals are the
arguments of the verb, independent pronouns are used primarily for referential contrast or focus.
This is precisely the case with the full reciprocal in Spanish. It is not unreasonable, then, to
hypothesize that pronominal objects in Spanish are not syntactic arguments, but rather, adjuncts
to the verb phrase, as Jelinek (1984) argues. If this is the case, the seemingly odd fact that
accusative clitics must double post -verbal pronominal objects, but cannot double R- expression
objects, receives a simple explanation. The two environments licensing clitic doubling in
accusatives are collapsed into one: accusative clitics must surface when the subcategorized
argument is not overt. (13a) can thus be represented as in (14), where, following standard
assumptions (e.g. Rizzi 1986), the clitic forms a chain with a coindexed empty category.
14. Lai
vi
cl -acc. saw -I
e, [a ella]
her
e
'I saw her'
Returning to the issue of the reciprocal construction, we recall that the full form el uno
al otro is obligatorily doubled by the reciprocal/reflexive clitic, as shown in (15):
15.a) Juana y Pepe se vieron el uno al otro.
Juana and Pepe cl saw
the one the other
'Juana and Pepe saw each other'
b) *Juana y Pepe vieron el uno al otro.
Juana and Pepe saw the one the other.
The ungrammaticality of (15b) shows that el uno al otro patterns with object pronouns in
triggering obligatory clitic doubling. And, as with the pronouns, the full form of the reciprocal
is marked in Spanish, insofar as its primary role is to disambiguate, that is, to distinguish
between possible reflexive or reciprocal interpretations. (15a) is thus assigned the structure in
(16), where el uno al otro is adjoined to VP.
16. a,[Juana y Pepe, vP[[Sei vieron ei]Ni l uno al otro]]
7
and P
cl saw
the one the other
'Juana and Pepe saw each other'
Interestingly, more support for the analysis of the full reciprocal as an adjunct comes from
Italian and French. Belletti (1982) notes that Italian allows apparent clitic doubling of the
accusative reciprocal clitic by the phrase l'un l'altro 'the one the other'. Though l'un l'altro
differs from Spanish el uno al otro in that the Italian construction never shows gender or number
agreement with its antecedent, it serves the same disambiguating function as the Spanish full
41
form. In the Italian case, there would seem to be little doubt that such a form is an adjunct, since
Italian does not permit doubled accusative objects6. Belletti points out that French (citing Kayne
1975) also allows such doubling in structures with the full reciprocal l'un l'autre 'the one the
other'. Again, French does not double accusative objects, and the French full form is even more
similar to the Spanish el uno al otro in that it inflects for number and gender.
In contrast to el uno al otro, the clitic reciprocal is assigned the structure in (17), (see
Rizzi 1986 and Manzini 1986 for a similar treatment of Italian si):
17. ,p[ Juana y Pepe, vP[se, vieron e,]]
J and P
cl saw
'Juana and Pepe saw each other'
Note that if this analysis is on the right track, the descriptive generalization that emerges
for Spanish is that only R- expressions can appear as overt complements of the verb. Non -RExpressions complements are represented in terms of a chain consisting of an empty element and
either a pronominal or an anaphoric clitic, with the clitic pronoun constituting the overt
instantiation of the subcategorized argument. As I will show, the treatment of the full reciprocal
el uno al otro as an adjunct allows for a straightforward account of the puzzle of scope described
above.
5.0 Accounting for the Scope Asymmetry
As pointed out in 2, the fact that clitic reciprocals permit wide scope readings poses
technical problems for the "each- movement" type of analysis of HLMa. However, responding to
Williams (1991), Heim, Lasnik and May (1991b; henceforth HLMb) propose an alternative
analysis for English reciprocals, in which each is not moved, but rather, A' -bound by a
distribution operator D adjoined to an antecedent of the reciprocal. Different scopes thus arise
as a function of the distance between the variable each and its binder. If each is bound by an
operator in the matrix clause, a wide scope interpretation obtains, while binding by an operator
in the embedded clause accounts for the narrow reading. This is shown in (18): (their 7)
18. a) [[John and Mary], DZ]2 think they2 like [each2 other]3
b) [[John and Mary], D4]4 think [[they,] D2]2 like [each2 other]3
(broad)
(narrow)
Here, (18a) represents the broad construal, as the variable each is bound by the D- operator
adjoined to the matrix subject [John and Mary]. (18b) represents the narrow reading, as each is
bound by the operator adjoined to the subject of the embedded clause.
In looking at the questions raised by the Spanish reciprocal, I will adopt the essential
insight of the HLMb analysis of English. Specifically, I will argue that the constituent el uno
patterns analogously with the English each in functioning as a bound variable which provides the
contrast argument for el otro (see HLMa for details relating to the semantics of this treatment
of the reciprocal construction)'. Unlike English each other, however, the full reciprocal el uno
'Except in clitic left dislocation constructions (cf. Cinque 1990).
'See also Fiengo and Lasnik (1973) and Langendoen (1978) for discussion of reciprocal interpretation.
42
al otro is an adjunct rather than an argument, and, as a consequence, the bound variable el uno
is subject to strict locality conditions. In contrast to the full form, I argue that the
morphologically "simplex" clitic reciprocal contains no bound variable and is analyzed as an
anaphor, subject only to Condition A of the Binding Theory (Chomsky 1981, 1986, etc.). Scope
ambiguities involving the clitic form follow simply from whether the A- binder of the clitic chain
is itself a product of bound variable or coreference anaphora.
5.1 Clitic Scope
Let us first address the case of the non -full reciprocal. Consider the data in (4), here
repeated as (19);
19. a) Juan y María creían que pro se; habían visto e;.
cl. had
seen e
J and M
thought that
'Juan and María thought they had seen each other'
Simplifying somewhat from Chomsky (1986), if the chain consisting of the reciprocal clitic and
its coindexed empty category is an anaphor, it must be bound in the least complete functional
complex (CFC) containing a possible antecedent. In this case, the anaphoric chain (se;,e;) must
be bound within the IP containing pro, which c- commands the empty category and constitutes
a possible antecedent. Recalling the examples in (18), two possible representations are available
for the IP containing pro. These are shown in (20):
20. a) [[Juan y María],D]2 creían que pro2 see habían visto e2.
b) [Juan y María], creían que [[pro,]D2)2 see habían visto e2.
thought that pro
cl had seen e
J and M
'Juan and María thought they had seen each other'
Scope ambiguity is thus accounted for as follows. (20a) exemplifies the broad construal.
Specifically, pro receives its interpretation via bound variable anaphora with the distributed
matrix subject, as it bears the index of the entire distributed NP. Given Condition A, pro must
be coindexed with the anaphoric chain (se2,e2). As a result, the distributed NP of the matrix
clause takes scope over the whole sentence, thus producing the broad reading. In (20b), however,
the interpretation of pro is derived via its coreferential indexation with the non -distributed [Juan
y María]. In turn, pro itself is subject to distribution by the operator D. Finally, the anaphoric
chain, pace Condition A, is coindexed with the entire distributed NP [[pro,]D2]2, thus providing
the narrow construal. In short, broad scope obtains when the anaphoric chain is coindexed with
an antecedent which receives its interpretation via bound variable anaphora. And narrow scope
is accounted for via the coindexation of the anaphor with an antecedent subject to distribution
under D. Note in (20b) that (se2,e2) cannot be coindexed with pro itself, as pro is not an
argument, but rather, a constituent of the argument [[pro,]D2]2, and thus does not constitute a
potential A- binder of the anaphor.
5.2 Full Reciprocal Scope
Given the analysis here, the full reciprocal construction contains an anaphoric chain (se;,e;)
as well as the adjunct el uno al otro. Like the English each other on the HLMb analysis, I argue
that the full form contains a variable, el uno which must be A' -bound by a distribution operator.
43
Recall that unlike each other, however, the NP el uno al otro is an adjunct to VP, rather than an
argument. Consider (9a), here repeated as (21).
21. Juan y María creían que pro se habían visto el uno al otro.
J and M
cl. had
thought that
seen the one the other
'John and Mary thought they had seen each other'
Following HLMb's treatment of each other, (21) can be structurally represented with the
indexation in (22):
22.a) *[[Juan y María]1D2]2 creían que pro2 vp[[se2 habían visto e2]Np[el uno2 al otro]3].
b) [Juan y María]1 creían que [[pro1]D2]2 vp[[se2 habían visto e2)Np[el uno2 al otro]3].
'Juan and María thought they had seen each other'
Note that as in (20), the indexation in (22) satisfies Condition A with respect to the A- binding
of the anaphor (se2,e2) by the subject of the embedded clause. In addition, however, the variable
el uno must also be A' -bound by an operator. In (22a), the closest A'- binder is the D operator
adjoined to the matrix subject. Since the broad scope construal is not possible with embedded
full reciprocals, I conclude that the variable el uno is subject to a strict locality condition.
Specifically, I claim it must be antecedent governed by its binder. Thus, the unavailability of the
wide scope interpretation represented in (22a) is due to the fact that the bound variable is not
antecedent governed by its binder, the D operator adjoined to the matrix subjects.
Interestingly, then, the possibility of wide scope interpretation in the English analogue to
(22a) indicates that English permits long distance binding of the variable each, while Spanish
does not. An explanation for this difference lies in the argument/adjunct distinction. As a variable
within an adjunct it is not unreasonable to expect that el uno should be subject to strict locality
constraints, similar to those found for movement from adjuncts. Consider, for example, parallels
to wh- extraction from adjuncts (cf. Huang 1982; Lasnik and Saito 1984; Chomsky 1986b, etc.):
23. *How do you wonder who fixed the car t
24. *How did John announce a plan to fix the car t
Interestingly, though the embedded Spanish full reciprocal does not involve movement, the
adjunct island effects such as those seen in (23 -24) seem to obtain.
In contrast to (22a), the closest A'- binder in (22b) is adjoined to the NP subject of the
embedded clause. Here, the clitic chain is A -bound by the embedded subject pro under
'Apparently, when there is no c- commanding long distance binder, a broad construal is available. Consider the
following:
i. Sus, entrenadores dijeron que pro, se, ganarían e, [el uno, al otro]
'Their coaches said they would beat each other.'
Here, pro is coindexed with the non -c- commanding possessive 'su', and a non -contradictory reading is
permitted. I leave this question for further research.
44
distribution by D. The variable el uno is A' -bound and antecedent governed by the D adjoined
to the embedded subject pro. And the result is the narrow interpretation.
5.3 Summary
The above data reveal a surprising scope asymmetry between full reciprocal and reduced
or clitic reciprocal constructions in Spanish. Embedded clitic reciprocals permit both broad and
narrow scope interpretations, while embedded full reciprocals are limited to narrow construals.
As I have pointed out, this asymmetry is not predicted by the "each movement" theory of HLMa,
and the fact that the Spanish data are more easily accommodated within the context of the "each binding" framework of HLMb constitutes an argument in favor of the latter approach.
Under the analysis proposed thus far, then, the lack of broad scope in embedded full
reciprocal constructions finds an explanation in the argument/adjunct distinction. The full
reciprocal el uno al otro is analyzed as an adjunct to VP, with a variable el uno providing the
contrast argument for el otro. This variable must be bound by a distribution operator D, under
strict locality conditions. Drawing parallels to wh- extraction from adjuncts, I propose that the D
operator must antecedent govern the variable, thus blocking broad scope construals. The clitic
reciprocal contains no internal bound variable and is analyzed as an anaphor, subject only to
Condition A of the Binding Theory. Scope ambiguities involving clitic reciprocals are derived
straightforwardly from whether the A- binder of the clitic chain is itself interpreted via bound
variable or coreference anaphora.
6.0 A Second Puzzle: Adjuncts
Thus far we have focused exclusively on reciprocal constructions in embedded clauses
with transitive verbs, that is, verbs subcategorizing for an NP argument. Such verbs obligatorily
surface with the clitic reciprocal, while the full reciprocal el uno al otro is optional and analyzed
as an adjunct to VP. In this sense, the account of the scope asymmetry hinges on the distinction
between arguments and adjuncts. Arguments permit scope ambiguities; adjuncts do not. Consider,
however, the following contradictory sentence.
25. #Juan y Pepe habían prevalecido el uno sobre el otro
had
J and P
prevailed the one over the other
'John and Pepe prevailed over each other'
(25) is the Spanish analogue to the embedded Italian sentence in (3b). And like its Italian
counterpart, it affords a non -contradictory reading when embedded.
26. Juan y Pepe le dijeron a María que pro habían prevalecido el uno sobre el otro
her told
J and P
Mary that they -had prevailed the one over the other
'John and Peter told Mary that they had prevailed over each other'
That (26) is ambiguous between contradictory and non -contradictory interpretations
indicates that broad scope must be available. Note the absence of the clitic reciprocal in (25 -26).
This is explained by the fact that prevalecer 'prevail' does not take an NP object. Here, the full
reciprocal construction el uno sobre el otro 'the one over the other' in (26) is analyzable as a
45
prepositional phrase adjunct to the embedded VP.' Interestingly, then, a broad scope reading is
available, despite the fact that the full reciprocal is within an adjunct.
Broad scope from within adjuncts is not restricted to (26). In (27a -b), the intransitive
trabajar 'work' precedes the adjunct [el uno más que el otro] 'the one more than the other'.10
Note that (27a) is contradictory, while (27b) permits a non -contradictory reading. It is worth
noting as well that the English counterpart to (27b) 'John and Pepe told Mary that they had
worked more than each other' is marginally non -contradictory at best. Strangely, nothing in the
HLMb "each binding" analysis of English would seem to rule out a broad construal here, as each
can be bound by a c- commanding D operator adjoined to the matrix subject. In Spanish, the
clear, non -contradictory reading of (27b) shows that a broad reading is available.
27.a) #Juan y Pepe habían trabajado el uno más que el otro
had
worked the one more than the other
J and P
'John and Pepe worked more than each other'
b) Juan y Pepe le dijeron a María que pro habían trabajado el uno más que el otro
had
worked the one more than the other
Mary that
her told
J and P
'Juan and Pepe told María that they had worked harder than each other'
Consider as well the data in (28), where the full reciprocal el uno...el otro surfaces in the
locative PP. Though the question here is not one of contradictory versus non -contradictory
construals, at least one interpretation of (28) indicates that broad scope is again available.
el uno cerca del otro.
28. Juan y Pepe le dijeron a María que pro iban a vivir
Mary that
were going to live the one near the other
her told
J and P
'John and Pepe told Mary they were going to live near each other'
The most salient interpretation of (28) is the narrow reading: 'Juan and Pepe each told María that
they (Juan and Pepe) were going to live near each other.' However, a broad construal obtains as
well. For example, each of Juan and Pepe can have told María that he was going to live in
Tucson, while neither was aware of the fact that the other would be living near him." Proof that
such a reading is possible is found in examples such as (29).
29. Los dos enemigos le dijeron a María que iban a vivir el uno cerca del otro, pero la gracia
está en que ninguno de los dos se ha dado cuenta de que van vivir en el mismo pueblo.
'The two enemies told Mary that they were going to live near each other, but the funny part is
9I follow Belletti (1982) in analyzing constructions such as [el uno sobre el otro] 'the one over the other' as a
prepositional phrase to which [el uno] is adjoined.
10I leave open the question of how to analyze the internal structure of the comparative construction [el uno más
que el otro]. E. Jelinek (p.c.) has suggested that it might be a small clause. Another possibility is to treat it as an
adverb phrase. For the purposes of the present discussion, I argue only that this structure occupies a position adjoined
to the VP headed by the intransitive 'trabajar'.
"I am grateful to Andy Barss for discussion concerning such interpretations.
46
that neither of them has realized that they are going to live in the same town.'
As with (8) above, broad scope must be available in (29). Otherwise, the assertion that neither
realizes that he is going to live in the same town as the other would be anomalous.
Taken together, these examples give rise to a paradox. Specifically, if the lack of broad
scope interpretations in embedded full reciprocal constructions with transitive verbs is due to the
adjunct status of el uno al otro, there is no clear explanation for the availability of broad
construals in (26 -28), where the full reciprocal el uno...el otro surfaces in prepositional phrases
that are adjuncts to VP. The data thus point toward a second asymmetry. Full reciprocals in
embedded VPs headed by transitive verbs do not permit broad scope interpretations, while full
reciprocals in adjuncts such as locatives and comparatives do.
6.1 Adjuncts
There are at least two obvious approaches to dealing with this new puzzle. The first is to
abandon the claim that the full reciprocal el uno al otro is an adjunct in structures such as (16),
here repeated as (30).
30. ,P [Juana y Pepe, Nip{ [se, vieron e,]NPel uno al otro]]
cl saw
J and P
the one the other
'Juana and Pepe saw each other'
Under such an approach, el uno al otro might be analyzed as an argument in a manner
analogous to HLMa&b's treatment of each other in English. This strategy is problematic in at
least three respects. First, English each other systematically permits broad scope readings when
embedded, while el uno al otro fails to allow broad construals. Secondly, such an analysis
provides no obvious explanation for the striking scope asymmetry between clitic reciprocals and
their doubled el uno al otro counterparts within Spanish. Thirdly, we lose the straightforward
explanation for the asymmetry in clitic doubling in accusative constructions in general, that is,
the account for why accusative clitics only double pronominal (but never R- expression)
"arguments ".
Suppose, then, that [el uno al otro] is an adjunct. An obvious consequence of such an
approach is the need to further refine the notion "adjunct" in such a way as to preserve the
analysis of clitic and full reciprocals above, while providing an account of the asymmetry
between the adjuncts in (26 -28) and the doubled full reciprocal in (30). I anticipate the nature
of this refinement by calling the doubled el uno al otro an appositional adjunct or
"adposition "'Z. In contrast, I reserve the standard term "adjunct" for structures such as the
locative PP el uno cerca del otro in (28).
Assuming that adpositions and adjuncts are distinct types of adjuncts, a formal
codification of this distinction is necessary as it pertains to the scope asymmetry outlined in
section 7. That is, if this asymmetry is to receive a syntactic explanation, the representations of
these structures must provide the means for limiting adpositions to narrow construals, while
permitting broad readings for el uno...el otro adjuncts. I suggest that the mechanism for making
the necessary distinction can be found within the theory of Relativized Minimality (Rizzi 1990).
This is in the spirit of Jelinek's (1984) adargument vs. adsentential distinction.
47
Specifically, I will adopt a version of Rizzi's restricted use of referential indexation, arguing that
the difference between these adjunct types lies in the licensing of referential indices for the
reciprocal el uno...el otro in adjuncts but not in adpositions. The consequence of this indexation
will be to allow the long -distance binding of the variable el uno in adjuncts, while limiting
adpositional el uno . to binding via antecedent government.
6.2 Adpositional Scope
Recalling Chomsky (1965), Rizzi (1990) proposes a restricted theory of indexation by
which referential indices are only licensed by a referential theta -role. As a consequence,
"arguments" and "adjuncts" are formally distinguished in that only the former are assigned
referential indices. Relevant here is the fact that in A' dependencies, operators can be connected
to their variables in one of two ways: via binding in the usual sense of referential coindexation
under c- command; or via antecedent government, if the variable lacks a referential index.
In this context, consider the adpositional el uno al otro, which, we recall, is always
doubled by the reciprocal clitic, as in (31).
31. pro se ven
e [el uno al otro]
they cl see -3rd -pl e the one the other
'they see each other'
Following standard assumptions (cf. Chomsky 1981, etc.), the internal theta -role of the verb is
assigned to the chain (se,e) headed by the reciprocal clitic. Now consider the adposition [el uno
al otro]. If it were to have a theta -role, that theta -role would have to be the same one assigned
to the clitic chain. Given the theta- criterion, such assignment is prohibited, and I conclude that
el uno al otro bears no referential index. Interestingly, the lack of such an index accounts
straightforwardly for the inability of the full reciprocal to be interpreted with broad scope when
embedded. This becomes clear if we reconsider sentence (22), here repeated as (32), where el uno
al otro is unindexed.
32.a) *[[Juan y María],D2]2 creían que pro2 vp[[se2 habían visto e2]Np[el uno al otro]].
Jr and M
thought that pro
cl had seen e the one the other
b) [Juan y María], creían que [[pro,]D2]2 vp[[se2 habían visto e2]Np[el uno al otro]].
'Juan and María thought they had seen each other'
Recall that the unacceptable (32a) represents the broad construal. Since the variable el uno
lacks a referential index yet must be bound, the only strategy available within the Relativized
Minimality framework is antecedent government, which Rizzi defines as follows (p.92):
33. Antecedent Government: X antecedent governs Y iff
(i) X and Y are non -distinct
(ii) X c- commands Y
(iii) no barrier intervenes
(iv) Relativized Minimality is respected
Under this definition, the unacceptability of (32a) becomes transparent. The binder of the
48
variable is the D operator adjoined to the matrix subject [Juan y Maria]D], which I assume is
"non- distinct" and thus a possible antecedent for the variable el uno. Although D in (32a) ccommands the variable, the embedded CP would seem to constitute a barrier. Antecedent
government is thus blocked by its intervention between D and el uno.
Note that even without the barrier clause in the definition of antecedent government, (32a)
is further ruled out as a violation of Relativized Minimality, as defined in (34) (Rizzi 1990:7):
34. Relativized Minimality: X a- governs Y only if there is no Z such that
(i) Z is a typical a- governor for Y,
(ii) Z c- commands Y and does not c- command X
In (32a) the subject pro of the embedded CP constitutes a typical a- governor for the variable el
uno, as it is a potential antecedent governor. While pro c- commands el uno, it does not ccommand the matrix subject [NPJuan y Marïa]D]. Thus, by definition, antecedent government by
the D of the matrix subject violates Relativized Minimality, since the subject pro of the
embedded clause is a closer potential governor. Antecedent binding by D in (32a) is again ruled
out.
As Rizzi points out, antecedent government is a strictly local relationship. Given the lack
of movement implicit in HLMb "each binding" theory assumed here, successive cyclic movement
is not an option. That is, el uno remains in situ, and must therefore be bound by a strictly local
D operator. This is the case in (32b). Here, the embedded subject [[proi]D2]Z is under a
distribution D. The variable el uno is antecedent governed by the D operator, as no barrier
intervenes, and Relativized minimality is respected. As we expect, the narrow reading obtains.
Interestingly, this approach is preferable to the analysis of the lack of wide scope construals in
embedded reciprocals in 5.2 above. In 5.2, I simply claimed that the lack of broad scope in clitic
doubled reciprocal constructions could be accounted for if el uno had to be antecedent governed.
Here, however, obligatory antecedent government is a direct consequence of the absence of
referential indices.
6.3 Adjunct Scope
The above analysis accounts nicely for the systematic lack of broad scope in embedded
full reciprocal constructions. Since the full reciprocal is an appositional adjunct, it receives no
referential index. The lack of such an index requires that the dependency between el uno and the
D operator be one of antecedent government. This analysis, however, raises difficult questions
for constructions such as (35).
35) Los dos dijeron que pro habían prevalecido el uno sobre el otro
The two said that pro had
prevailed the one over the other
'The two said that they had prevailed the one over the other'
As noted above, the possibility of a non -contradictory construal indicates that broad scope is
available. Under a strict interpretation of Rizzi's approach to the assignment of referential indices,
the PP adjunct [PP[NPel uno[PPsobre el otro]]] is not assigned a referential theta -role by the verb
and thus should not license the referential indexation of its object. The lack of such indices, as
we have just seen, requires that the variable el uno be antecedent bound. In this case, the
unfortunate prediction is that a non -contradictory reading for (35) should not be available.
49
Though the argument is admittedly circular, I suggest that the availability of broad scope
in these adjunct constructions indicates that the discontinuous NP object el uno...el otro is
referentially indexed. Unfortunately, the issue of theta- assignment itself is quite vague13 (see,
for example, Dowty 1990). But it seems clear that there exists a clear difference between the
adpositional el uno al otro, whose function is to disambiguate between possible reciprocal and
reflexive readings, and adjunct PPs such as el uno sobre el otro 'the one over the other' and el
uno cerca del otro 'the one near the other'. This difference derives, I suggest, from the fact that
the adpositional el uno al otro has no theta -role, while the objects of adjunct PPs are theta -
marked by their respective prepositions. In the case of el uno sobre el otro, we might, for
example, say that the object of sobre expresses the goal of the act of prevailing over. Rizzi
himself recognizes the need to allow prepositions, in conjunction with verbs, to endow selected
prepositional phrases with referential theta- roles. I suggest here that adjunct prepositional phrases
in Spanish with the discontinuous reciprocal NP object el uno...el otro are selected in this
fashion.
Having taken this step, one might attempt to argue that the preposition a of the adposition
el uno al otro can also be said to assign a theta role to its el uno...el otro object and thus license
the assignment of a referential index in these structures. However, it is not clear that a is, in fact,
a preposition. Consider the following data, in which the syntactic behavior of the adpositional
el uno al otro is contrasted with that of true prepositional phrases whose object is the reciprocal
el uno...el otro. In (36a), for example, we see that the preposition a can never precede el uno.
In contrast, in (37a -b) the cerca of the complex preposition cerca de 'near' can either precede
or follow el uno.
36.a) *Juan y María se ven [á1 uno el otro]
J and M
cl see "a" the one the other
b) Juan y María se ven [el uno al otro]
J and M
cl see the one "a" the other
'Juan and María see each other'
37.a) Juan y María duermen [cerca el uno del otro]
J and M
sleep
near the one of the other
b) Juan y María duermen [el uno cerca del otro]
J and M
sleep
the one near of the other
'Juan and María sleep near each other'
In (38a -c), the quantificational adverb mucho 'a lot' cannot appear within the adposition, while
it can precede, follow, and surface within a true reciprocal PP, as seen in (39a -c).
13See also Jackendoff (1990) for a discussion of the complexities of theta- assignment, as well as a discussion
of problematic aspects of the them-criterion.
50
38.a) *Juan y María se quieren [el uno] mucho [al otro].
the other
the one a lot
J and M cl love
b) Juan y María se quieren mucho [el uno al otro].
cl love a lot the one the other
J and M
c) Juan y María se quieren [el uno al otro] mucho.
cl love the one the other a lot
J and M
'Juan y María love each other a lot'
39.a) Juan y María discuten [el uno] mucho [con el otro].
with the other
the one a lot
argue
J and M
b) Juan y María discuten mucho [el uno con el otro].
the one with the other
argue a lot
J and M
c) Juan y María discuten [el uno con el otro] mucho.
the one with the other a lot
argue
J and M
'Juan y María argue with each other a lot'
The same pattern is found with adverbs such as frecuentemente 'frequently.' In (40a -c), the
adverb fails to surface within the adposition, but in the case of prepositional adjuncts, the adverb
can follow el uno.
40.a) *Juan y María se ven [el uno] frecuentemente [al otro].
the other
cl see the one frequently
J and M
b) Juan y Maria se ven frecuentemente [el uno al otro].
the one the other
J and M
cl see frequently
c) Juan y María se ven [el uno al otro] frecuentemente.
J and M
cl see the one the other frequently
'Juan and María see each other frequently'
41.a) Juan y María caminan [el uno] frecuentemente [con el otro].
walk
the one frequently
with the other
J and M
b) Juan y María caminan frecuentemente [el uno con el otro]
walk
the one with the other
J and M
frequently
frecuentemente.
c) Juan y María caminan [el uno con el otro]
walk
the one with the other frequently
J and M
'Juan and María walk with each other frequently.
While (41a) is admittedly the least favored of the three possibilities, it contrasts clearly with the
unacceptable (40a). These examples show that a does not pattern with true prepositions, and I
51
suggest that this is due to the fact that el uno al otro is simply a complex NP. Other researchers
have reached similar conclusions regarding the non - prepositional status of a in Spanish. For
example, Jaeggli (1982) cites Kayne, who argues that the a in dative object constructions is not
a preposition. Suñer (1888) argues that this a is a marker of definiteness and animacy. Of
relevance here is the following: if a is not taken to be a preposition, it does not assign a theta role and cannot, under any circumstances, license referential indices.
Let us assume, then, that in contrast to adpositions, the reciprocal el uno...el otro in PP
adjuncts is referentially indexed, as seen in (42).
42. [[Los dos]1D2]2 dijeron que pro habían prevalecido [e12 uno sobre el otro3]
the two
said
that pro had prevailed
the one over the other
'The two said that they had prevailed over each other'
Crucially, the presence of a referential index on el uno will allow the dependency between the
operator and the variable to be satisfied simply by binding, as binding requires only c- command
and coindexation. In (42), the D operator adjoined to the matrix subject is coindexed with and
c- commands el uno. Neither intervening barriers nor Relativized Minimality are relevant to the
binding relation. Thus, given the indexation in (42), the availability of broad scope is expected.
Consider finally (43), the narrow scope analogue of (42).
43. *[Los dos], dijeron que [prol]D2]2 habían prevalecido [e12 uno sobre el otro3]
the two said that pro
had
prevailed
the one over the other
'The two said that they had prevailed over each other'
Here, el uno is bound by the c- commanding, coindexed D2 of the lower clause. In this case, the
result is the narrow, contradictory construal.
7.0 Conclusions
This paper has explored the question of scope construals in embedded reciprocal
constructions in Spanish within the context of the related "each- movement" and "each- binding"
theories of Heim, Lasnik, and May (1991b). With regard to the Spanish data, two surprising
asymmetries emerge. First, critic reciprocal constructions are shown to permit both broad and
narrow construals when embedded, while their clitic doubles or "full" reciprocal counterparts are
limited to narrow readings. An explanation for this asymmetry is proposed, based on the
argument status of the reciprocal clitic chain, versus the adjunct status of the full el uno al otro.
Under this analysis, el uno is taken to be a variable that must be bound via antecedent
government. The clitic chain, in contrast, is taken to be an anaphoric argument subject only to
Condition A of the Binding Theory.
I then consider the scope properties of the el uno...el otro reciprocal construction in
prepositional phrase adjuncts to VP. In these structures, broad construals are permitted, giving
rise to a second asymmetry. Concretely, if narrow construals with doubled full reciprocals are
a function of the adjunct status of el uno al otro, no clear explanation is available for why broad
construals are possible from reciprocal PP adjuncts. I argue that this asymmetry motivates the
need for a more finely grained notion of adjuncts themselves, and claim that the theory of
Relativized Minimality provides the means for making the necessary distinctions. I make use of
the notion of referential indexation, claiming that the doubled el uno al otro reciprocal is an
52
appositional adjunct or "adposition" which cannot be referentially indexed. Consequently, the
variable el uno in these structures can only be bound via antecedent government, a strictly local
relationship. This accounts for the availability of only narrow construals. In contrast, I claim that
el uno...el otro in true PP adjuncts is referentially indexed. Broad scope is thus licensed through
the inherently non -local binding relation.
References
Belletti, A. (1982) "On the Anaphoric Status of the Reciprocal Construction in Italian ", The
Linguistic Review 2, 101 -137.
Bresnan, J. and S. Mchombo (1987) "Topic, Pronoun, and Agreement in Chiche*a ", Language
63, 741 -782.
Cinque, G. (1990) Types of A' Dependencies. MIT Press, Cambridge.
Chomsky, N. (1965) Aspects of the Theory of Syntax. ma Press, Cambridge.
-- (1981) Lectures on Government and Binding. Foris, Dordrecht.
-- (1986) Knowledge of Language. Praeger, New York.
-- (1986b) Barriers. MIT Press, Cambridge.
Dowty, D. (1990) "Thematic Proto -Roles and Argument Selection" Language 67, 547 -619.
Fiengo, R. and H. Lasnik (1973) "The Logical Structure of Reciprocal Sentences in English ",
Foundations of Language 9, 447 -468.
Heim, I., H. Lasnik, and R. May (1991a) "Reciprocity and Plurality ", Linguistic Inquiry 22, 63101
-- (1991b) "On Reciprocal Scope ", Linguistic Inquiry 22, 173 -192.
Jackendoff, R. (1990) Semantic Structures. MIT Press, Cambridge.
Jaeggli, O. (1982) Topics in Romance Syntax. Foris, Dordrecht.
-- (1986) "Three Issues in the Theory of Clitics: Case, Doubled NPs, and Extraction" Syntax and
Semantics 19, H. Borer, ed., 15 -42.
Jelinek, E. (1984) "Empty Categories, Case, and Configurationality ", Natural Language and
Linguistic Theory 2, 39 -76.
Kayne, R. (1975) French Syntax. MIT Press, Cambridge.
Langendoen, D. T. (1978) "The Logic of Reciprocity" Linguistic Inquiry 9, 177 -197.
Lasnik, H. and M. Saito (1984) "On the Nature of Proper Government ", Linguistic Inquiry 15,
235 -289.
Manzini, M.R. (1986) "On Italian Si ", Syntax and Semantics 19, H. Borer, (ed.), 241 -262.
May, R. (1977) The Grammar of Quantification. Doctoral Dissertation, MIT, Cambridge, MA.
(Distributed by the Indiana University Linguistics Club.)
-- (1985) Logical Form. MIT Press, Cambridge.
Pillar, P. (1991) "Clitic Doubling and Scrambling in Spanish" U. of Arizona ms.
Rizzi, L. (1986) "On Chain Formation ", Syntax and Semantics 19, H. Borer, (ed.), 65 -96.
Suñer, M. (1987) "Dialectal Variation and Clitic -Doubled Direct Objects ", in Studies in Romance
Linguistics: Selected Papers from the Seventeenth Linguistics Symposium on Romance
Languages. C. Kirschner and J. DeCesaris (eds.), (1989), John Benjamins, Amsterdam.
377 -398.
-- (1988) "The Role of Agreement in Clitic- Doubled Constructions ", Natural Language and
Linguistic Theory 6, 391 -434.
Williams, E. (1991) "Reciprocal Scope ", Linguistic Inquiry 22, 159 -173.
53
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