Post on 21-Jan-2023
Processing controlled PROs in Spanish*
Moises Betancorta, Manuel Carreirasa,*, Carlos Acuna-Farinab
aDepartamento de Psicologıa Cognitiva, Universidad de La Laguna, Campus de Guajara, 38205 Tenerife, SpainbUniversidad de Santiago de Compostela, Spain
Received 20 November 2004; accepted 20 April 2005
Abstract
Two experiments were carried out to investigate the processing of the empty category PRO and
the time-course of this in Spanish. Eye movements were recorded while participants read sentences
in which a matrix clause was followed by a subordinate infinitival clause, so that the subject or the
object of the main clause could act as controller of PRO, and therefore as implicit grammatical
subject of the infinitive. In Experiment 1, verb control information was manipulated: The matrix
clause contained either subject-control verbs like prometer (‘promise’) or object-control verbs like
forzar (‘force’). In Experiment 2, the preposition that headed adverbial subordinate clauses was
manipulated: Two different kinds of infinitival adverbial clauses were used, expressing purpose
(preposition para) and reason (preposition por) and in which control information is primarily induced
by the prepositions (para tends to trigger subject-control; por object-control). Experiment 1 showed
that readers make immediate use of verb control information to recover the antecedent of the empty
category PRO in Spanish obligatory control constructions. The data obtained in Experiment 2
suggest that during the processing of the empty category PRO in purpose vs. reason adverbial
subordinate infinitival clauses the control information induced by the prepositions por vs. para is not
initially used as a constraint to guide the selection of the nominal antecedent of PRO. In addition,
both experiments showed that PRO antecedent selection is a very fast process and that, together with
verb control information, recency played an important role. The results are discussed in terms of
current psycholinguistic hypotheses about empty categories and in relation to formal linguistic
hypotheses about PRO control. We argue that obligatory control ties (which are co-referential ties)
Cognition 100 (2006) 217–282
www.elsevier.com/locate/COGNIT
0022-2860/$ - see front matter q 2005 Elsevier B.V. All rights reserved.
doi:10.1016/j.cognition.2005.04.001
* This research was partially supported by grants from the Spanish Ministry of Science and Technology
(BSO2001-3492-C04-03 and BSO2003-01135) and the ESF-EUROCORES-OMLL BFF2002-10379-E to the
second author and from the Xunta de Galicia PGIDIT03PXIA20401PR to the third author. We would like to thank
Chuck Clifton, Enrique Meseguer, Patrick Sturt, and three anonymous reviewers for comments to previous
versions of this manuscript.* Corresponding author. Tel.: C34 922 3175108; fax: C34 922 317461.
E-mail address: mcarreir@ull.es (M. Carreiras).
M. Betancort et al. / Cognition 100 (2006) 217–282218
are processed fast because they are launched from the same lexical platform that launches all fast
syntactic connections (complements, as opposed to modifiers). In short, that such lexically-specified
co-referential ties are regulated by linguistic form.
q 2005 Elsevier B.V. All rights reserved.
Formal theories of grammar, such as the Government & Binding/Minimalist Program
(GB/MP), Lexical Functional Grammar (LFG), Head Driven Phrase Structure Grammar
(HPSG), or Role & Reference Grammar (RRG), have all paid considerable attention to the
question of empty categories (ECs) by giving them first-class consideration, often as
separate modules, within their theoretical frameworks, irrespective of how, or even if such
empty elements figure or not in their syntactic trees. Psycholinguistic research has been
somewhat slower but is quickly catching up in its focus on the processing of linguistic
reality without phonetic realization and there is now empirical evidence pointing to the
representational reality of this. (Bever & McElree, 1988; Featherston, Gross, Munte, &
Clahsen, 2000; Fiebach, Schlesewsky, & Friederici, 2001; MacDonald, 1989; McElree &
Bever, 1989; Nicol, 1988; Nicol & Swinney, 1989). Of the four GB-inspired implicit
nominal entities examined (wh-trace, NP-trace, PRO, and pro), wh-trace (Boland,
Tanenhaus, & Garnsey, 1990; Clifton & Frazier, 1986; Crain & Fodor, 1985; Fiebach et
al., 2001; Frazier, Clifton, & Randall, 1983; Garnsey, Tanenhaus, & Chapman, 1989;
Stowe, 1984; Swinney, Ford, Frauenfelder, & Bresnan, 1988; Tanenhaus, Boland,
Garnsey, & Carlson, 1989) and NP-trace (Bever & McElree, 1988; Cloitre & Bever, 1988;
MacDonald, 1989; McElree & Bever, 1989) have received most psycholinguistic
attention. However, very few research efforts have been devoted to studying pro or PRO,
the object of the present study (Betancort, Meseguer, & Carreiras, 2004; Boland et al.,
1990; Demestre, Meltzer, Garcia Albea, & Vigil, 1999; Garcıa-Albea & Meltzer, 1996;
Mauner & Koenig, 2000; Mauner, Tanenhaus, & Carlson, 1995).
In general, the functional and cognitive motivation of PRO is well-known. As an implicit
subject, its essence relies on easy recoverability, that is, on the ease with which the missing
information can be derived from the linguistic or situational context. For instance, in (1)
(1) Mary tried PRO to be nice
Mary, the subject of tried, is also understood to be the subject of be nice, so grammar
opts for a fused structure in which Mary appears only once as a sister of the matrix verb,
since its latent presence as sister of the lower verb can be grammatically derived. Since, in
compliance with the Theta-criterion (Chomsky, 1981, p. 36), each argument bears only
one theta-role per verb, GB/MP theoreticians postulate a null subject PRO of the infinitive.
PRO receives its theta-role by ordinary structural principles from that lower verb. Given
the fact that PRO does not involve movement (from the gap to the position occupied by the
actual surface form, as in raising; e.g. Maryi seems to (t)i be nice), it is best seen simply as a
covert anaphoric element whose interpretation is controlled by an antecedent NP (the
higher clause subject Mary). Thus, both the antecedent and PRO mean the same thing
(they are co-referential) but, importantly, they are not the same thing, just as in Mary
M. Betancort et al. / Cognition 100 (2006) 217–282 219
thought that she would faint, Mary and she may mean the same thing but occupy two
distinct places in the syntactic tree.
In GB/MP, Binding theory regulates the way PROs come to be controlled by different
kinds of antecedents in different syntactic environments (Chomsky, 1981, 1986; Manzini,
1983). Since Rosenbaum (1967), linguists have tried to regulate syntactically the nature of
the control relation between antecedent and gap. Since that has not proved an entirely
feasible task, they have generally ended up concluding that PRO is very much a lexical
phenomenon emanating from the control properties of complement-taking heads (Bresnan,
1982; Cattell, 1984; Chierchia, 1988; Chomsky, 1981; Jackendoff, 1972; 1974; Sag &
Pollard, 1991). Theories of grammar such as LFG, HPSG or RRG have, in fact, completely
eliminated the representation of the binding relation in syntactic structure (Fodor, 1988). In
these theories PRO, which does not exist phonetically, simply does not exist in syntactic
trees either. So in (1), for instance, the syntactic complement of try is simply the VP be nice
(instead of a clause with a missing subject). Control is instead established in other strata or
levels of grammatical organization where the missing subject of non-finite verbs is made
explicit: f-structure in LFG, or grammatical argument structure in HPSG, for instance.
Fodor (1989) has argued for a similar view of PRO from a psycholinguistic perspective.
In a recent reaction against yet another new attempt, by Hornstein (1999), to regulate
control syntactically by subsuming it under raising (i.e. via movement from a
trace position to the actual matrix subject position), Culicover and Jackendoff (2001)
explain the reasons why this maneuver has traditionally failed. Although it is obviously
impossible to go into the specifics of this question here, a few points they mention may
suffice to illustrate the way lexical specification manifests itself, as well as to project the
nature of PRO into the psycholinguistic dimension that is our concern. As evidence that
control cannot be predicted from structural properties of grammar (unlike raising),
consider (2)–(6):1
(2) An American attempt to PRO invade Vietnam
(3) Johni begged Maryj to PROj leave
(4) Johni promised Maryj to PROi leave
(5) John’s promise/vow/offer/obligation/pledge/oath/commitment to Susan to PRO take
care of himself/*herself
(6) John’s order/instruction/encouragement/reminder/invitation to Susan to PRO take
care of herself/*himself
In (2) the controller is a denominal adjective (American), so it cannot have raised up
(via movement) from a subject position in the infinitival lower clause, as adjective phrases
cannot be subjects. In (3) control is exercised by the higher clause object (Mary) and in (4),
in contrast, by the higher clause subject (John). Since subject-control verbs like promise
are rather exceptional, Rosenbaum (1967) proposed a Minimal Distance Principle (MDP)
according to which, control is assigned to the NP closest in the tree (the object, actually the
controller in the vast majority of cases involving a competition between subject and
1 These examples are taken from Culicover and Jackendoff (2001).
M. Betancort et al. / Cognition 100 (2006) 217–282220
object). However, when one expands the promise class of predicates to nominals, the
exceptionality of the class vanishes, as (5) demonstrates. In all of (5) the controller is John.
In all of (6), however, it is Susan. As Culicover and Jackendoff point out, “the most
plausible basis for the difference in controller between [5] and [6] is thematic structure” (p.
505). This is evident from the fact that all the nominals in (5) receive control, not from a
syntactically-determined subject position but from the thematic role of Source (the giver
of the promise) wherever that may be located syntactically:
(7) The promise to Susan from John to take care of himself/*herself
(8) John gave Susan some sort of promise to take care of himself/*herself
(9) Susan got from John some sort of promise to take care of himself/*herself
If the controller is the lexically-entrenched semantic role of Source, and not the
configurationally-defined function of subject, “the facts emerge elegantly”. However, the
price is that the controller cannot be identified in terms of syntactic position (Culicover &
Jackendoff, 2001, p. 506).
Although there is almost complete unanimity in all syntactic theories that PRO-control is
lexically driven, a psycholinguistic account of how the missing subject is recovered might
yield a different picture. From a processing point of view, the critical issue is the kind of
information (lexical, structural, statistical, etc.) utilized by the parser and, more particularly,
the temporal course of that utilization. In the first place, for structures like (3) and (4), similar
to those used in our first experiment, it is indeed the case that the vast majority of verbs obey
Rosenbaum’s MDP. Since grammatical MDP correlates nicely with the Most Recent Filler
Strategy (MRFS) proposed in Frazier et al. (1983), it will be interesting to see, in these
structures containing subjects and objects followed by PRO, whether PRO is processed
faster when controlled by the object, the most recent filler or whether, alternatively, parsers
make immediate use of their lexical knowledge to anticipate the control properties of each
verb. In their pioneering study, Frazier et al. were primarily concerned withwh-gaps. When,
almost in passing, they examined structures of control like (10) and (11) below:
(10) This is the girl the teacher decided ____ to talk to ____.
(11) This is the girl the teacher forced ____ to talk.
they concluded that control information did not constrain initial filler-gap assignment.
However, using very similar structures in an on-line plausibility monitoring task, Boland
et al. (1990) obtained results that sustained the very opposite conclusion: namely that “verb
control information is used immediately” (p. 427). It is important to keep in mind that
Frazier and her collaborators did not tackle control in depth at the time (indeed they did not
isolate the phenomenon of control from that of movement; see Fodor (1988, p. 141, fn. 17)),
and that since their overall theory of gap-filling has moved away from MRFS into the Active
Filler Strategy (Frazier & Flores D’Arcais, 1989) and the Minimal Chain Principle
(De Vincenci, 1998), there would appear to be no formal theory of gap-filling today which
may be applied to gaps which, like PRO, cannot be ‘actively’ anticipated by the prior
detection of their fillers. Clarifying the role of verb control information and recency/MRF
fully during the on-line processing of Spanish sentences such as (12a) and (12b) below will
M. Betancort et al. / Cognition 100 (2006) 217–282 221
be the first objective of the present study. It is important to notice that (12a) and (12b)
contain only one gap (PRO) and thus allow us to avoid the complex interactions between
wh-gaps and control gaps that have plagued previous discussions of control (Boland et al.,
1990; Fodor, 1978, 1988; Ford & Dalrymple, 1988; Frazier et al., 1983):2
(12)a. Marıai prometio a Pedroj [PROi] ser bastante cauta con los comentarios.
Maryi promised Peter [PROi] to be quite cautious (f) with her comments.
b. Marıai exigio a Pedroj [PROj] ser bastante cauto con los comentarios.
Maryi demanded from Peterj [PROj] to be quite cautious (m) with his comments.
Mary demanded from Peter that he be quite cautious (m) with his comments.
Since, as seen in (2)–(9) above, PRO-control is grammatically diverse, a second
objective will be to compare the way it is processed in verbal structures like (12) with the
kind of PRO that appears in Spanish adjunct clauses like (13a) and (13b):
(13)a. Yolandai se caso con Jorgej para PROi tener dinero en abundancia.
Yolandai married Jorgej in order to PROi have money in excess.
b. Yolandai se caso con Jorgej por PROj tener dinero en abundancia.
Yolandai married Jorgej for PROj having money in excess.
Yolanda married Jorge because he has/had money in excess.
Since in these sentences the prevalent tendency is for the preposition to determine the
interpretation of PRO (para introduces a purpose clause and typically triggers subject
2 Frazier and Clifton found that the relative clause structure in (11) was harder than that in (10). Notice that
these relative structures are very similar to the wh-questions used by Boland et al. (1990) (which outlaw/horse did
the cowboy signal to surrender to the authorities) in that they contain both PROs and wh-gaps at the same time.
The latter involve displaced material and thus require that processors untie the workings of a transformation, in
generative terms (De Vincenci, 1998). As intuition suggests, the interaction of both types of ECs makes them
particularly challenging from a processing point of view, and considerably more complicated than the structures
we will use in our two experiments, which contain PROs only. Fodor (1988, p. 135, 145, and 146) provides a long
and dense linguistic explanation of why the interacting gaps complicate Frazier and Clifton’s conclusions about
the slow recovery of PRO. In essence, it has to do with whether the PRO-filler of structures like (11) is the moved
wh-element itself (a distant filler) or its trace (not distant anymore), and it involves complex accompanying
interactions among various aspects of the submodules of Government, Case theory, and Theta theory. Since our
structures do not have traces (and therefore no such “flurry of filling activity [that] creates a sort of cloud of dust
around the object NP position”, p. 148), we believe we isolate the phenomenon of control more appropriately. It is
also worth noting that the control information that Frazier et al. were examining involved the contrast between
unambiguous control verbs like those in (10) and (11) and ambiguous control verbs like want (see note 5). On the
interaction between different kinds of gaps and the relation between grammars and parsers when dealing with
such complex structures, see Fodor (1978), especially in relation to the XX Extraction Constraint and the Nested
Dependency Constraint. Fodor, however, like Frazier et al. (1983), and Boland et al. (1990), devotes most of her
extensive discussion to dealing with gaps which are the product of movement operations, and she only makes
passing reference to structures of control (to which she refers, using the pre-GB terminology of the time of
writing, as Equi). That merely reinforces our comment, in the introduction, that while movement-related gap-
filling has been extensively studied from very early on, non-movement-related gap-filling (like that involved in
PRO) has received much less attention in the psycholinguistic literature. Ford and Dalrymple (1988) focus on
whether the Frazier et al. results fit more naturally with a GB grammar or a GPSG grammar, discussing that access
to control ties is unproblematic for GPSG, but again, they do not target control alone either.
M. Betancort et al. / Cognition 100 (2006) 217–282222
control, por introduces a reason clause and typically triggers object control),3 it will
likewise be interesting to examine whether these adjunct clauses are easier to process
when the main burden of control falls on the most recent filler.4 So we will be checking the
role of recency in the processing of unannounced gaps using two different constructions.
Another way of approaching this second objective will be to consider to what extent verbs
are really unique both in grammar and in processing, as is often assumed (understandably
so, as verbs really are the core of syntactic structure at the clausal level in virtually every
known theory of grammar). Notice that verb control is fixed in the sense that particular
verbs consistently subcategorize for either subject or object control, but not for the two
kinds (that is, prometer, for instance, never licenses object control, just as exigir never
licenses subject control).5 Conversely, control induced by prepositions in adjunct clauses
is obligatory only in the sense that it must be co-indexed locally within the sentence where
the prepositions occur. Yet, within such a specified local domain, it may shift from subject
to object in pragmatically appropriate circumstances,6 although it is generally true that
there is a very strong propensity for por-clauses to trigger object control and especially for
3 Although the term ‘purpose clause’ is by far the most frequent label used to denote clauses expressing purpose
(Huddleston & Pullum, 2002, p. 727–731 and 1222 ff.), the particular kind of ‘para’ clause that we have used in
our experiment 2 is often referred to as a ‘rationale clause’ in the psycholinguistic literature, and translated into
English by in order to. The latter term is used by Mauner et al. in a number of articles, for instance (see reference
section). We owe this point to an anonymous reviewer.4 A linguistic analysis of the para and por constructions is provided in the introduction to experiment 2. To
anticipate a little of that, it is as well to make it clear now that, unlike their English counterparts, sentences (13a)
and (13b) are absolutely identical from a syntactic point of view in Spanish.5 It is as well to mention that control can shift from subject to object in verbs like expect or want, in the sense
that they exhibit subject control when they do not have objects, and object control when they are used transitively
(I expect to go vs. I expect you to go; in fact, the latter is not even seen as control in the generative literature but as
Exceptional Case Marking; see Culicover & Jackendoff, 2001: 495). By ‘fixed’ we mean that verbs which have
both subjects and objects select either subject or object control consistently, not just any of the two. These latter
predicates are the ones we use in our experiment 1. We owe this clarification to an anonymous reviewer. Verbs
like beg, which even when used transitively may be interpreted as licensing either subject or object control (I beg
you to sing; see Boland et al., 1990) are extremely exceptional (and in fact, many English speakers find the subject
control reading of transitive beg hard to understand).6 Thus, in Fran se caso con Monica por tener tanto dinero (Fran married Monica because she has/had so much
money) and Fran le dijo eso a Monica por fastidiar (Frankie said that to Monica because (he wanted to) tease/bug
her) control shifts from object to subject but is never arbitrary. Likewise, in Fran se enfado con Monica para irse
de la fiesta tan pronto (Fran got angry at Monica in order to leave the party so early) and Juan envio a Monica allı
para hacer ese trabajo (Juan sent Mary there to do that job), control also shifts, from subject to object this time, but
must involve either Juan or Monica, that is, it cannot be arbitrary either. Although control shifts of the kind
illustrated above are more frequent with por than with para, it is not at all difficult to fabricate examples using
purpose clauses: Juan trajo a un especialista para acabar el escenario (Juan brought a specialist to finish the stage),
Marıa necesita a alguien bueno para arreglar eso (Marıa needs somebody (who is) good to fix that), Juan contrato
a Marıa para hacer ese trabajo (Juan hired Marıa to do that job). See Haegeman (1994, p. 262). What these and
other similar examples prove is that even though control in purpose clauses tends to be more fixed than in other
adverbial subordinate clauses, it is not so fixed as in verbs, where its reversibility from subject to object or
viceversa is almost exceptionless (beg may actually be an exception, at least for some speakers, as already noted).
It is difficult to imagine a verb like allow, for instance, licensing subject control occasionally. The difference
between verbs and prepositions as regards the determination of control no doubt reflects the tighter integration
with the matrix of complement clauses than adjunct clauses.
M. Betancort et al. / Cognition 100 (2006) 217–282 223
para-clauses to trigger subject control. In fact, our norming studies suggest that for our
materials at least, this propensity is similar to the one we obtained for fixed, verb-guided
control. This paper does not address a third kind of control, arbitrary control, which occurs
in structures like PRO venir aquı serıa de tontos (PRO coming here would be foolish) or
Gino’s es el mejor lugar para PRO comer eso (Gino’s is the best place PRO to eat that).
In this type of control, PRO is free and usually encompasses all the members in a discourse
scene or all human beings in general, including the speaker. Antecedent recovery is
pragmatic in its very essence and therefore unconstrained by structural requirements such
as locality or c-command (Fernandez Lagunilla & Anula Rebollo, 1995, p. 192;
Huddleston & Pullum, 2002, 1268 ff.).
The third aim of the present study will be to assess the speed with which the right
antecedent of PRO is activated, as there is conflicting evidence about this in the literature.
Using a cross-modal priming technique, Nicol (1988) found no evidence whatsoever of
faster activation of the most recent filler in structures containing PRO (as well as other
ECs) such as (14):
(14) That’s the actressi that the dentistj from the new medical center in town .a. had invited ti PROi to go to the party
b. had hesitated PROj to go to the party with ti
Reaction times of related words to PRO’s most recent antecedent, presented
immediately at the offset of the word “to”, were not faster than those of a control word.
This led Nicol to conclude that, since the relationship between fillers and gaps is a
semantic one of co-reference, ECs are not initially co-indexed with any filler: rather, all
potential antecedents are merely reactivated and choice of gap is delayed until finally
resolved when extra-syntactic information becomes available (about 1000 ms after the
gap, when there is competition between fillers). Nicol further proposed that co-referential
relationships between fillers and gaps are dealt with by a late-acting autonomous module
of the linguistic mind, a suggestion anticipated in Cowart and Cairns (1987). In the same
vein, Sag and Fodor (1995) conclude that, since priming studies have shown fast activation
of wh-trace, but not of PRO (or NP-trace), only the former, but not the latter, appears in
syntactic trees (see also, Fodor, 1989). They also point out that this pattern of results
supports theories of grammar, like HPSG or LFG, which have dispensed with the
representation of control in their syntactic formalisms. Additionally, McElree and Bever
(1989) also report late activation of the antecedent of PRO using a probe recognition task.
Late activation, however, does not seem to fit in well with results obtained with another
probe recognition task (Bever & Sanz, 1997), and with a “stop-making-sense” task
(Mauner & Koenig, 2000, Exp. 2; Mauner et al., 1995, Exps 1, 2 and 3). In these
experiments, Mauner et al. preceded infinitive clauses such as “to win a prize and lots of
cash” with active declarative and full passive sentences that provided an explicit
antecedent to the PRO, as shown in (15a) and (15b), respectively, with short passive
sentences that were hypothesized to introduce an implicit agent (15c), and with
intransitive sentences which introduced neither an implicit nor an explicit agent to be
taken as the PRO antecedent (15d):
M. Betancort et al. / Cognition 100 (2006) 217–282224
(15)a. The contestant spun the game show’s wheel PRO to win a prize and lots of cash
b. The game show’s wheel was spun by the contestant PRO to win a prize and lots of
cash
c. The game show’s wheel was spun PRO to win a prize and lots of cash
d. The game show’s wheel spun PRO to win a prize and lots of cash
They found that anomaly effects emerged in infinitival clauses immediately at the verb
win only when they followed intransitive sentences. In particular, their results indicate that
when readers encountered the verb win, the PRO was immediately linked to its antecedent
when such an antecedent was available-even to an implicit agent in the case of short
passive sentences.7
Finally, late activation of PRO is particularly at odds with an experiment carried out by
Demestre et al. (1999) in which they gathered electrophysiological evidence of gender
agreement violations in Spanish structures such as (16):
(16) Pedro/*Marıa quiere PRO ser rico (masc)
Peter/Mary wants to PRO be rich
Using auditory stimuli, Demestre et al. reported an early negativity with
prominence at anterior and central sites followed by a centro-parietal positivity in
the violating condition (a typical biphasic pattern in syntactic anomalies; cfr. Coulson,
King, & Kutas, 1998; Friederici, Pfeifer, & Hahne, 1993; Haggort, Brown, &
Groothusen, 1993; Osterhout, 1997), which showed that the brain reacts to NP-
adjective gender agreement violations with intervening PROs fairly soon after the
critical stimulus (the adjective). Unfortunately, Demestre et al. only used structures
with one grammatically-defined potential filler (either just one filler, like (16) above,
or object-controlled sentences like (12b), but no subject-controlled ones), which makes
it impossible to pit the MRFS against models that advocate rapid use of lexical
information coming from the (either subject-or object-) control properties of the verbs.
Furthermore, their use of a non-standard onset for the ERP wave analyses makes it
hard to compare them with the standard ERP effects typically associated with
syntactic and semantic operations, as well as to evaluate to what extent there are
either early or late effects.8 Therefore, their results did not allow them to arbitrate
between different parsing models, or to suggest how fast the process of antecedent
assignment is taking place. This, however, is something we hope to be able to do, by
using constructions with two potential fillers while recording eye movements. Eye-
movement methodology has been widely used to search for the roles that different
sources of information play in the time-course of processing.
7 They provide further indirect evidence of the immediacy of PRO processing in other experiments in which
they fronted the infinitive clauses, so that, as in the filler-gap studies, they created an expectancy for a gap to be
satisfied later in the sentence (Mauner and Koenig, 2000, Exp. 3, using a stop-make-sense task; Mauner et al.,
2002, recording eye movements). They found again that readers begin to experience difficulty with only
intransitive sentences soon after they have recognized the main verbs in these sentences, which implies that
readers immediately tried to link an agent to the initial PRO.8 They used the onset of the last, instead of the first, syllable of the adjective to time-lock EEG averages.
M. Betancort et al. / Cognition 100 (2006) 217–282 225
As noted above, a principle-grounded model like Garden Path (Frazier, 1987; Frazier &
Clifton, 1996) would nowadays circumscribe its predictions on gap-filling to the domains
of either the Active Filler Strategy or the Minimal Chain Principle. Since such theories
involve displaced syntactic material that can therefore be detected, a theory like GP would
in principle appear to have no fully explicit prediction to make on the processing of PRO,
which involves no syntactic displacement. Here, we will assume (but we stand to be
corrected) that, in the absence of a new proposal, a formal model of parsing is expected to
invoke the recency effect in the processing of unannounced gaps like PRO. On that basis a
prediction follows that there should be no initial differences in reading times at the
infinitive between subject and object control sentences. The reason is that GP would
stipulate mandatory selection of the most recent filler, the cognitively simplest heuristic.
Since, at the infinitive, only lexical information from the matrix verb might alert the parser
to the likelihood of one or another type of control and since this cannot be consulted yet,
the parser would have to opt for an object-control analysis in every case. This blind choice
will have repercussions, however, in regions after the verb when the object-control
analysis proves to be incompatible with the subject-control sentences. Therefore,
reanalysis is expected only for subject-control sentences after the verb. As an index of
re-processing, it should be observed in late measures, such as second pass and total reading
time. Object-control sentences should not exhibit any trace of reanalysis whatsoever.
Nonetheless, it could be argued that reanalysis processes are very quickly triggered during
the reading of the infinitive region, which could cause longer late processing measures
such as second pass time and/or total reading time at the infinitive for subject-control
sentences.
On the other hand, constraint-satisfaction models (MacDonald, Pearlmutter, &
Seidenberg, 1994; Tanenhaus & Trueswell, 1995) predict that all relevant constraints,
including recency and verb control information, as well as others, are combined in a
parallel competition process from the earliest stages of processing. However, the specific
predictions of these models will mostly depend on how many constraints are being
considered, as well as on their relative strengths. For instance, if control information—
because it is derived from the lexical properties of heads—is seen as the predominant
constraint, then no differences between the two sets of sentences would be expected, since
readers will access the control information when reading the main verb or the preposition,
and they will thus be ready to expect the right antecedent of each upcoming PRO. Given
the enlarged size of the lexical component in these models, this is perhaps the most
obvious reading one can make of their predictions. If, however, recency is considered as
another competing constraint, then it depends on how strong this is as a competitor relative
to the strength of lexically-guided control. For instance, if the recency constraint is
stronger than verb control information, then constraint satisfaction models could actually
be indistinguishable in their predictions from two-stage models (recency acts first, then all
the rest) and predict that object sentences should be read initially faster than subject
sentences. However, later reanalysis processes would be more difficult to accommodate in
a constraint satisfaction framework. If, on the other hand, verb control information is a
stronger constraint than recency, then an initial clash between the two forces should be
expected from the very beginning in early measures. In the absence of a fully implemented
model that allows for explicit and detailed predictions, the best evidence in favour of
M. Betancort et al. / Cognition 100 (2006) 217–282226
a constraint-based approach might come only indirectly from the more specific predictions
of formal models failing to come true.
1. Experiment 1
As already noted, experimental data from English have shown so far (see Boland &
Tanenhaus, 1991; Clifton & Frazier, 1989; Fodor, 1989; Frazier, 1993; Frazier & Clifton,
1989; Nicol & Swinney, 1989, for a review) that the time course of re-activation for the
legal antecedent of an EC is quite different, depending on the type of EC (Wh-trace, NP-
trace, PRO) and the experimental paradigm used. For example, in the cross-modal lexical
priming paradigm (CMLP), the data showed that there is a rapid reactivation of the
antecedent (MacDonald, 1989; McElree & Bever, 1989; Nicol, 1988; Nicol & Swinney,
1989; Swinney et al., 1988) in wh-trace structures. However, the strength of re-activation
for NP-trace and PRO is weak and less consistent than for Wh-trace, which seems to
behave like a referential-expression (Nicol, 1988; Swinney et al., 1988). This pattern of re-
activation has been taken as an index of the different linguistic status they have in GB
theory (Chomsky, 1981, 1986; see, however, Hornstein, 1991, for the opposite view, as
well as this introduction). However, as Featherston et al. (2000) have pointed out, due
mainly to the disparity of the results obtained with the different tasks used, the data
reported up to now leave uncertain at least two central questions: at what point the
referential co-indexing takes place (Osterhout & Swinney, 1993) and, more importantly,
what is the relative contribution from syntactic factors to antecedent reactivation.
The special status that empty categories have makes research in languages other than
English particularly relevant.9 Of particular interest is their behaviour in languages with a
less strict word order. Experiments conducted in German (Fiebach, Schlesewsky, &
Friederici, 2001, 2002), a relatively free word-order language, have shown that, as in
English, in constructions with NP- and Wh-trace readers seem to be able to identify ECs
quickly and refer them to their antecedents. The conclusion of these studies is that when
the processor faces an empty category or recognizes a moved element (as Wh-trace), it
tries to compute its legal antecedent as fast as possible in a way signalled by the so-called
“filled-gap” effect (Crain & Fodor, 1985). Precisely how quickly this assignment is made
and what type of information is initially used are still open questions in the case of the
empty category PRO, and even more so in Spanish (Demestre et al., 1999). Eye-movement
measures should prove helpful in unravelling the role that different sources of information
play in the processing of these structures.
The goal of this experiment is to test the processing time-course of the empty category
PRO in Spanish subordinate infinitival sentences. The central questions to address are: To
what extent does verb control information (thematic control) guide the selection of
the right antecedent of PRO and is this type of information initially consulted?
9 In GB/MP theory, some ECs are formed by movement (wh-traces, NP-traces), while others are base-generated
(PRO). Little pro occurs mainly in pro-drop languages like Italian or Spanish due to their rich inflective
morphology. So ECs seem to vary cross-linguistically depending to some extent on specific language-internal
structure.
M. Betancort et al. / Cognition 100 (2006) 217–282 227
Control constructions are used in which either the subject or the object of the main
sentence acts as controller of PRO, and therefore as implicit grammatical subject of the
infinitive. Sentences will contain subject-control verbs like promise and object-control
verbs like force. In order to test at what point in the sentence the readers have chosen the
NP that controls PRO, in the region immediately following PRO an adjective phrase that
either matches or fails to match the controlling NP in gender is introduced.
The predictions of two models of processing are tested in this experiment. The
garden path model assumes that a single analysis is initially chosen based only on
structural information. On the other hand, according to constraint-based models, structural
information is only one of the multiple constraints (plausibility, context fit, thematic
information, etc.) at work in a unitary parsing process. If the initial analysis is based only
on structural information, readers should take the most recent NP (second NP) as
controller of PRO in all cases, so no differences are predicted in early measures. However,
more difficulty is expected to be observed in late measures, in the condition in which the
most recent NP does not fit with the control information provided by the (subject-control)
verb, which would cause a reanalysis. In sum, the subject-control verb condition should
produce more difficulties and these should be observed late. Reanalysis should therefore
be the key that shows us if recency is obeyed blindly. In addition, a mismatch effect is
expected when the gender of the adjective does not agree with the right antecedent if this
has already been assigned.
On the other hand, if, as constraint satisfaction models may claim, control information
provided by the verb is initially used as the only (or major) constraint to guide the selection
of the right antecedent of PRO, then no differences are expected between subject-and
object-control conditions in the PRO region either in early or late measures, as such
models opt for a unitary one-stage process. Only slower reading times are expected, due to
the mismatch effect between the right antecedent and the adjective when these do not
agree in gender. However, if control information induced by the verb is used promptly
together with recency as concurrent constraints to guide the selection of the right
antecedent of PRO, then more difficulties in the subject-control condition are expected,
and this should be captured from the beginning in very early measures.
2. Method
2.1. Participants
Thirty-two undergraduate students of the University of La Laguna participated in the
experiment for course credit.
2.2. Materials and design
Thirty-two sets of four matched sentences with the structure NP1-VP1-NP2-[PRO]-
Infinitive-AdjP-PP, like those displayed in Table 1, were constructed. The main verb
(VP1) could be either a subject-or an object-control verb, so either NP1 or NP2 will act as
controllers, respectively.
Table 1
Example of materials used in Experiment 1
Subject control match
Marıai/prometio/a Pedroj/[PROi] ser/bastante cauta/con los comentarios
Maryi/promised/Peter/[PROi] to be/quite cautious (f)/with her comments
Subject control mismatch
*Marıai/prometio/a Pedroj/[PROi] ser/bastante cauto/con los comentarios
*Maryi/promised/Peter/[PROi] to be/quite cautious (m)/with his comments
Object control match
Marıai/exigio/a Pedroj/[PROj] ser/bastante cauto/con los comentarios
Maryi /demanded/from Peterj/[PROj] to be/quite cautious (m)/with his comments
Mary demanded from Peter that he be quite cautious (m) with his comments
Object control mismatch
*Marıai/exigio/a Pedroj/[PROj] ser/bastante cauta/con los comentarios
*Maryi/demanded/from Peterj/[PROj] to be/quite cautious (f)/with her comments
*Mary demanded from Peter that she be quite cautious (f) with her comments
*Ungrammatical sentences due to gender mismatch agreement. Slashes indicate how regions were segmented.
However, subjects did not see these during reading.
M. Betancort et al. / Cognition 100 (2006) 217–282228
The two NPs of the main clause were always of different gender, so that the adjective in
the subordinate clause agreed in gender either with one or with the other. Therefore, if
assignment of antecedent for PRO is not delayed, but instead one of the NPs is selected
taking into account verb control information, a match or a mismatch of gender agreement
between the adjective and the PRO antecedent might occur, depending on the type of
matrix verb (subject-vs. object-control). Notice that the infinitive and the adjective phrase
are juxtaposed. In addition, another 92 sentences with different types of structures were
written to serve as fillers, as well as 10 sentences for practice trials. Four versions of the
experiment were created, each containing 32 experimental sentences and 92 fillers. Only
one sentence from each of the sets of four matched experimental sentences appeared in
each version of the experiment. Assignment of these sentences to each of the four
conditions was counterbalanced across versions. Any given subject saw the materials from
one version. The resulting experimental situation was a 2!2 design with type of verb
control (subject vs. object) and gender agreement (match vs. mismatch) as within-subjects
factors. Subject and object control verbs did not differ in printed frequency or in length.
Means of printed frequency were 77 for subject control and 42 for object control verbs;
they were not significantly different [t(34)Z1.17, PZ0.24]. Means of length in number of
characters were 7.7 for subject control verbs and 7.4 for object control verbs; they did not
differ from each other [t(34)Z0.64, PZ0.33]. The sentences used in Experiment 1 are
presented in the Appendix.
2.3. Apparatus and procedure
The sentences were presented in mixed case letters on a video screen interfaced with a PC
compatible computer. The monitor displayed up to 80 characters per line. All the sentences
had less than 80 characters, so they were displayed in one line. Participants were seated 73 cm
away from the monitor and three characters equaled 18 of visual angle. Participants’ eye
M. Betancort et al. / Cognition 100 (2006) 217–282 229
movements were monitored by a Forward Technologies Dual Purkinge Image Eye-tracker
which was interfaced to the computer. The eye-tracker has a resolution of 10 min of arc (half a
character). Viewing was binocular, with eye position recorded from the right eye. The signal
from the eye-tracker was sampled every millisecond by the computer.
When a subject arrived for the experiment he/she was seated in front of the monitor
with his/her head in a chin rest to cancel head movements. The initial calibration of the
eye-tracking system generally required about 5 min. The subject was asked to read silently
the sentences displayed on the monitor and told that he or she would be questioned about
the sentences and should read for comprehension. Questions were asked on approximately
one-third of the trials. Prior to reading each sentence, the subject was instructed to look at a
fixation box which outlined the first character position of the sentence. A red dot within the
square indicated to the subject that he/she was correctly looking at the first square and
ready to start to read, so he/she could press a button to display the sentence. When he/she
pressed the button a sentence appeared immediately on the screen. After reading the
sentence, the subject again pressed a button that could cause the presentation of a question
or of a row of squares. When a question appeared on the screen, participants had to press
one of two buttons to answer yes or no. Each subject initially read 10 practice sentences to
become familiar with the procedure. Then he/she read the 32 experimental sentences
intermingled randomly with the 92 filler sentences.
2.4. Norming studies
The control verbs for the experiment were selected according to evaluation by a group
of linguists of their properties as subject-or object-control in subordinate infinitival
contexts. Additionally, in order to test whether control was unambiguous, sentences
containing two potential antecedents and a control verb were presented in a questionnaire
to a group of 100 students. They were asked to read the sentences and choose one of the
two antecedents, whenever possible, as implicit subject of the infinitival verb that was
embedded in the subordinate sentence. Four versions were created so that position of
antecedents was balanced and crossed with two random orders of items. Both antecedents
were of the same gender. The 32 experimental items were randomized with 128 filler
sentences. When the main verb was a subject-control verb (prometio), 87% of the readers
(range 68–100%; standard errorZ2.08) identified the NP1 as subject of the infinitive,
which was clearly significantly different from chance in the analysis by subjects and by
items [t(99)Z57.8, P!0.001; t(31)Z17.76, P!0.001]. When the main verb was an
object-control verb (obligo), 80% of readers (range 61–99%; standard errorZ3.13)
identified the NP2 as subject of the infinitive (controller of PRO), which was clearly
significantly different from chance in the analysis by subjects and by items [t(99)Z43.13,
P!0.001; t(31)Z9.68, P!0.001]. In addition, to check for the verbs’ sub-categorization
frame, a Spanish data base was checked.10 All the verbs used in our experiment have
10 The corpus used was the BDS (‘Base de Datos del Espanol actual’) which contains an analysis of 160,000
entries taken from the ARTHUS (‘Archivo de textos hispanicos de la Universidad de Santiago de Compostela’.
http://www.bds.usc.es/). The corpus size is 1.5 million words.
M. Betancort et al. / Cognition 100 (2006) 217–282230
a direct object sub-categorization frame as a first option, which allowed us to create an
infinitival subordinate clause either with subject-or object-control. Discounting
prepositional forms of complementation, of the 1007 occurrences of subject-control
verbs, 840 are transitive uses. Of the 461 occurrences of object-control verbs, 445 are
transitive. The intransitive uses of all these verbs are negligible: of the 1007 occurrences of
subject-control verbs only in 70 instances (6.9%) are these verbs used intransitively. Of the
461 occurrences of object-control verbs only 7 (1.5%) show an intransitive use.
The remaining complements are prepositional. The same sentences of the questionnaire
study were used in the eye-tracking experiment.
2.5. Data analysis
We will report results for four different eye-movement measures. First-pass reading
times are the sum of all fixations from when the reader first enters a region from the left to
the time when the region is first exited either to right or left. This measure is usually
considered as reflecting first-pass difficulty. Regression path times are the sum of all
fixations from the first time a region is accessed to the first time the reader leaves the
region to the right. This measure is sometimes, thought not always, considered as an early
measure (Brysbaert & Mitchell, 1996; Duffy, Morris, & Rayner, 1988; Konieczny,
Hemforth, Scheepers, & Strube, 1997; Liversedge, 1994; Traxler, Pickering, & Clifton,
1998). We will interpret it as a late measure, since it includes not only the time spent in the
critical region but also the time in previous regions after regressions from the critical
region, if any, until the reader leaves the critical region to the right. Second-pass times are
the sum of all fixations made in a region after other subsequent regions have already been
fixated. It is the time that readers re-read a region after the eyes leave that region to the
right. Total reading times are the sum of all fixations that occurred in a particular region.
3. Results
For the purpose of the analyses, the texts were segmented as shown with slashes in
Table 1: the first region contained the first NP (“Marıa”); the second comprised the VP1
(“prometio /exigio”); the third the next NP (“a Pedro”);11 the fourth the PRO-infinitive
(“[PRO] ser”); the fifth the adjective (“bastante cauto/a”); and the sixth a PP (“con los
comentarios”). This last region was included in order to capture any spillover effects from
the previous region, as well as to keep away any influence of the last region (e.g. wrap up
effects) on the potential structural effects. The mean first pass reading times, mean
regression path times, mean total reading times and mean second pass times for the NP2,
PRO-infinitive, AdjP and PP regions of the sentences are shown in Table 2a. Deviations
from regression (Ferreira & Clifton, 1986; Trueswell, Tanenhaus, & Garsney, 1994) for
first pass reading times are shown in Table 2b. Reading times calculated according to
11 Technically, a Marıa is not a NP but a PP, as animate direct objects in Spanish are introduced by the
preposition a. ‘NP2’ refers to the NP inside the PP (Marıa).
Table 2a
Means and standard deviations of first pass reading times, regressions path duration, total reading times, and
second pass time for regions 3, 4, 5 and 6 of subject and object control match and mismatch sentences used in
Experiment 1
NP2 PRO-Infinitive AdjP PP
Mean SD Mean SD Mean SD Mean SD
First pass time
c-suj match 580 161.8 298 50.2 620 171.6 931 196.5
c-suj mismatch 572 157.5 297 62.9 620 155 881 245.3
c-obj match 482 105.6 319 57.1 648 167.6 850 242.5
c-obj mismatch 469 108.4 299 61 635 162.7 863 249.8
Regressions path duration
c-suj match 742 173.4 324 89.3 748 197.2 1647 728.7
c-suj mismatch 714 209.7 352 108.3 804 308.7 2069 1170.4
c-obj match 633 198.3 363 160.3 794 267.3 1672 820.9
c-obj mismatch 625 210.2 403 208.8 786 219.2 1963 1066.6
Total reading time
c-suj match 809 225.4 363 95.9 835 219.6 1054 249.3
c-suj mismatch 898 397.7 405 131.4 944 357.4 1152 426.9
c-obj match 720 279.6 403 176 875 252 1044 329.5
c-obj mismatch 794 337 417 118.9 936 317.7 1148 384.6
Second pass time
c-suj cong 214 187 70 70.5 199 151.3 126 137.9
c-suj incong 328 357.9 123 126.9 313 317.8 271 344.4
c-obj cong 240 244.7 84 112.3 212 206.4 187 233.8
c-obj incong 307 295.5 127 104.9 288 249.5 262 283.5
M. Betancort et al. / Cognition 100 (2006) 217–282 231
the measures described above were entered into ANOVAs. The error rate for the
comprehension questions was very low (between 1 and 3%) and did not differ among the
four experimental conditions.
3.1. First pass reading times
ANOVAS by-subjects and by-items carried out separately only showed a type of verb
control effect in the NP2 region (“a Pedro”). This region was read faster in the object
control condition than in the subject control condition [F1(1,31)Z22.54, P!0.001;
Table 2b
Means and standard deviations of deviations from regression for the first pass reading times for regions 3, 4, 5 and
6 of subject and object control match and mismatch sentences used in Experiment 1
First pass time
NP2 PRO-Infinitive AdjP PP
Mean SD Mean SD Mean SD Mean SD
c-suj match K30 151 K119 88 K121 132 K182 377
c-suj mismatch K50 146 K118 98 K154 154 K173 336
c-obj match K90 128 K109 102 K108 163 K174 324
c-obj mismatch K107 108 K112 78 K122 145 K182 322
M. Betancort et al. / Cognition 100 (2006) 217–282232
F2(1,31)Z21.71, P!0.001], whereas no differences were found between conditions in
the rest of the regions. One potential problem is that some verbs had to be repeated across
sentences because of the impossibility of finding enough good control verb exemplars. An
additional analysis including verb repetition as a new factor showed that the interaction
between type of verb control and verb repetition was not significant [F1(1,31)Z2.05,
PO0.05; F2(1,31)Z1,07 PO0,05]. However, the main effect of type of verb remained
significant [F1(1,31)Z18.90, P!0.001; F2(1,31)Z17.31, P!0.001]. A further analysis
including for each subject those sentences that had been seen only for the first time (no
repeated verbs) showed again a significant effect of type of verb [F1(1,31)Z14.44,
P!0.001; F2(1,31)Z19.29, P!0.001]. Thus, repetition of some verbs does not seem to
drive the main effect of type of verb control. Another potential problem is that, in 17 out of
the 32 experimental sentences this region was four characters longer in the subject control
condition than in the object control condition, because a longer preposition was sometimes
required after the matrix verb for the sentences to sound natural. However, new analyses
performed with the same data using a milliseconds per character measure showed the same
effects: the object control condition was read faster than the subject control condition in
the NP2 region [F1(1,31)Z5.39, P!0.05; F2(1,31)Z8.89, P!0.01] (subject control
45.5 ms/char versus object control 41.2 ms/char). In addition, analyses on the same data of
the same region using deviations from regression as a dependent variable were performed.
As pointed out by Ferreira and Clifton (1986), a better way to reduce the variance
produced by comparing regions of different lengths is to predict, through a linear
regression, how fast a region of a given length should be read, and then subtract this from
the observed reading time for that region. Data from deviations from regression analysis
showed the same effect again [F1(1,31)Z6.40, P!0.05; F2(1,31)Z7.10, P!0.05].
Object-control sentences were read faster compared to subject-control ones. Thus, it seems
quite likely that the effects found in the NP2 region cannot be accounted for by differences
in the number of characters. As another complementary analysis, we excluded those items
in which the subject-control condition was of different length than the object-control
condition. The analysis revealed the same patterns as before: subject-control sentences
were read more slowly than object-control ones, even when both conditions were of the
same length [F1(1,31)Z5.38, P!0.05; F2(1,16)Z4.66, P!0.05]. Finally, an ANCOVA
was performed including as covariates the frequency with which each verb co-occurred
with a direct object and the frequency with which they are followed by an infinitive.12
Again, the new analysis revealed the same effect of type of verb [F1(1,27)Z9.29,
P!0.05]. None of the covariates was significant [Fs!1].
3.2. Regression path times
ANOVAS by-subjects and by-items carried out separately showed only a type of verb
control effect in the NP2 region [F1(1,31)Z11.51, P!0.05; F2(1,31)Z6.27, P!0.05].
The NP2 region (“a Pedro”) was read faster in the object control condition than in
12 The frequencies of co-occurrence for the verbs presented in Experiment 1 were computed from the LEXESP,
a corpus of Spanish of five million six thousand entries (Sebastian, Martı, Carreiras, & Cuetos, 2000).
M. Betancort et al. / Cognition 100 (2006) 217–282 233
the subject control condition. The analysis for the PRO-infinitive region (“ser”) showed a
gender agreement effect, although only in the analysis by subjects [F1(1,31)Z5.75,
P!0.05; F2(1,31)Z3.67, PZ0.06]. Gender-matched sentences were read more rapidly
than gender-mismatched ones. It is important to note that the gender manipulation occurs
at the end of the following region, so such differences are unexpected and a bit puzzling,
because it is unlikely that subjects preview the mismatch information at this point.
The interaction between both variables was not significant [Fs!1]. No effects were found
in the next region, the AdjP region (“bastante cauto/a”). However, a gender agreement
effect was found again in the PP region (“con los comentarios”) [F1(1,31)Z8.81,
P!0.05; F2(1,31)Z9.15, PZ0.05]. This region of the gender-matched sentences was
read more rapidly than that of the gender-mismatched sentences. No other effects were
significant.
3.3. Total reading time
ANOVAS by-subjects and by-items carried out separately showed a type of verb
control effect in the NP2 region (“a Pedro”) [F1(1,31)Z10.49, P!0.05; F2(1,31)Z3.90,
P!0.06]. Object-control sentences were read faster than subject-control ones. The gender
agreement effect was significant in this region as well [F1(1,31)Z6,27, P!0.05;
F2(1,31)Z9,18, P!0.01]. Gender-matched sentences were read more rapidly than
gender-mismatched sentences. The interaction was not significant [Fs!1].
No effects were found at the PRO-infinitive region. Finally, only gender agreement
effects were found at the AdjP region [F1(1,31)Z6.89, P!0.05; F2(1,31)Z6.00,
P!0.05], and at the PP region, although only reliable in the analysis by items
Fig. 1. First pass reading times (FPT), regression path times (RPT) and total reading times (TT) over region NP2
in Experiment 1.
Fig. 2. Second pass times (SPT) and total reading times (TT) for subject and object control match and mismatch
sentences over AdjP and PP regions in Experiment 1.
M. Betancort et al. / Cognition 100 (2006) 217–282234
[F1(1,31)Z3.26, PZ0.08; F2(1,31)Z5.11, P!0.05]. Gender-matched sentences were
read more rapidly than gender-mismatched ones.
3.4. Second pass time
ANOVAS by-subjects and by-items carried out separately did not show any effect of
type of verb or of the interaction between both factors in any region. However, there was a
gender agreement effect, indicating that gender-matched sentences were read more rapidly
than gender-mismatched ones in several regions: the NP2 region [F1(1,31)Z8.2, P!0.01;
F2(1,31)Z11.7, P!0.01], the PRO-infinitive region [F1(1,31)Z8.86, P!0.01;
F2(1,31)Z10.8, P!0.01], the AdjP region [F1(1,31)Z9.20, P!0.01; F2(1,31)Z8.43,
P!0.01], and the PP region [F1(1,31)Z10.9, P!0.01; F2(1,31)Z10.6, P!0.01].
See Figs. 1 and 2 for a summary of the more important effects.
4. Discussion
A type of verb control effect and an agreement effect were found with different
measures and at different regions, but, also importantly, no interaction between the two
factors was found in any measure or region. The first pass time analysis showed a reliable
effect at the NP2 region (“a Pedro”), as did the regression path time and total reading time
analyses for this region, which disclosed a type of verb control effect (see Fig. 1). Readers
read more rapidly sentences with object-control verbs than with subject-control verbs.
This suggests that when they enter this region they engage in a kind of differential
processing for subject-vs. object-control sentences. It is important to notice that such
effects suggest a very early use of verb control information. The verb control effect, if any,
M. Betancort et al. / Cognition 100 (2006) 217–282 235
would be expected in the PRO region, since it is at this region where the lack of a
constituent, and therefore the need for an antecedent, manifests itself. However, there is no
effect at the PRO-infinitive region, but there is in the previous one, which seems to indicate
that when readers reach PRO they have already gathered enough information to process its
antecedent NP. It is likely that parafoveal processing of the critical region containing the
gap is taking place.
The milliseconds per character and deviation from regression analyses at the NP2
region suggest that the finding that the object condition is faster than the subject
condition cannot simply be attributed to a length effect. As another complementary
analysis, we conducted a first pass analysis in which we excluded those items in
which the subject-control condition differs in length from the object-control condition.
This analysis revealed the same patterns as before: subject-control sentences were
read more slowly than object-control ones, even when both conditions were of the
same length. So, the most plausible explanation is a kind of parafoveal preview
benefit. But how could we sustain a parafoveal processing account? The Landing
Position measure revealed that readers made an additional fixation on the NP2 region
between characters 5 and 6 (the average of characters in the NP2 region was 10).
Thus, taking into account that the PRO-infinitive region had 4.5 characters (averaged
over all items), it is possible that readers actually processed the PRO-infinitive region
while reading the NP2 region. In addition, the probability of first pass fixations and
the number of fixations at the NP2 “a Pedro”, the PRO-infinitive “[PRO] ser”, and the
adjective “bastante cauto/a” suggest that the infinitive is very often skipped. For each
of these two measures, no differences were found between conditions in each region,
but there was clearly a difference between regions (Probability of first pass fixations,
0.93, 0.62 and 0.94, respectively; number of fixations 1.82; 0.73 and 2.18). Finally, a
new sub-analysis of the PRO region based on a new segmentation that includes the
three last characters of the previous region (NP2) together with the PRO region
indicated that the subject control condition was slower than the object control
condition [F1(1,31)Z3.89, P!0.05; F2(1,31)Z4.21, P!0.05]. Studies conducted on
the perceptual span revealed that readers could extract information from the next
region up to 7–8 characters to the right (McConkie & Zola, 1987; Rayner, Well, &
Pollatsek, 1982; Underwood & McConkie, 1985). In short, data from
landing positions at the NP2 region (“a Pedro”) seem to indicate that at this point
of the sentence readers had a parafoveal preview benefit from the PRO-infinitive
region, which could have triggered the antecedent reactivation process. Even
though we could find no trace of reanalysis at the infinitive, the fact that the
subject-control verb condition was harder to process than the object-control verb
condition seems to show that verb control information is not the only constraint
for PRO-control assignment. If that were the case, we should not have obtained
any differences between the conditions. However, if verb control information and
recency (that is, proximity of antecedent and PRO) are taken into account during the
process of antecedent selection, then we should expect precisely the results we have
found.
The fact that readers took more time to process subject-control verbs indicates that
the distance between the controller NP and the gap plays a role in antecedent selection.
M. Betancort et al. / Cognition 100 (2006) 217–282236
Nonetheless, this conclusion is somewhat at odds with the results of Mauner et al.
(1995). In their third experiment, they used rationale clauses after active declarative,
full passive, short passive, and intransitive sentences, but they did not find differences
between declaratives, full passives and short passives as a function of whether the
controller of PRO immediately preceded the infinitival clause or not. One should bear
in mind, however, that they used a different methodology; a stop-making-sense task.
Later, Manuer, Melinger, Koenig, and Bienvenue (2002) replicated the same results
using an eye-tracking methodology, but they measured the effects of sentence-initial
rational clauses with infinitives that created expectancy for an agent not in the
infinitival clauses themselves but in the verbs of the main clauses, which implies a very
different situation from that of the present experiment. It is also important to note that
the Mauner et al. study did not involve strict, obligatory control but control with
adjuncts.
Another important question is to what extent structural information-recency or
proximity of antecedent-is considered before, or at the same time as, lexically-guided
verb control information during the process of antecedent selection and assignment.
Our data strongly suggest that both types of information are considered very early in
terms of the region where the effects have been found, as well as in terms of the
measures sensitive to the effects. In other words, no trace of delayed verb information
during the processing of this type of ambiguity was observed. In fact, verb control
information and recency were considered very early, which contrasts with other
results on gaps reported using a cross-modal priming paradigm in which the
re-activation effect for the legal antecedent was shown to be delayed and posterior to
the gap region (see Nicol & Swinney, 1989, for a review), but is in line with some
others that have shown an immediate reactivation of PRO antecedents (e.g. Mauner et
al., 1995, 2002). One possible explanation for this discrepancy is that the recording of
eye movements during reading may be more sensitive than cross-modal priming to
capture very early effects of antecedent selection. Data obtained from this experiment
seem to show that readers were not only sensitive to the verb control information and
to the proximity of antecedent and PRO, but they also seem to use these constraints
concurrently in order to quickly assign an antecedent to the PRO. When they reach
the AdjP region, the mismatched sentences showed an additional cost of processing
when the gender of the adjective in the subordinate clause did not agree with the
gender of the previously assigned antecedent (see Fig. 2). Only if antecedent
assignment has occurred is a mismatch gender agreement effect expected. The
mismatch gender effect was obtained in regression path duration, second pass reading
times and total reading times for several regions. This fact seems to suggest that the
recovery of the right antecedent is not delayed at all.
Another possible way of accounting for the lack of effects in the region of the gap
and the effects found in the preceding NP2 region is to contend that the effects
detected reflect the semantic integration of the NP2 and the matrix verb. Seen in this
light, the parafoveal processing account would be unnecessary. This line of reasoning
leads to a negation of the existence of PRO as an entity whose position is involved in
recency computations: there are no effects at the gap because there is no structural
gap. All that is required to explain the object-control advantage is the lexical
M. Betancort et al. / Cognition 100 (2006) 217–282 237
information contained in the verbs’ theta-grid. In short, PRO need not exist.13
Although we believe that such an account is intriguing, we find it hard to maintain for
several reasons. Firstly, it would predict that the frequencies of co-occurrences of
subject and object control verbs with a direct object should influence the reading time
difference at the NP2. However, the ANCOVA showed that none of the co-occurrence
frequency measures used as covariates was significant, while the type of verb control
effects was still significant. Secondly, such an explanation still needs to account for
how the parser assigns the antecedent to the infinitive. The robust effects of gender
mismatch suggest that antecedent assignment has occurred. One might still contend
that such a rapid detection of the gender mismatch reflects the process of building a
sentence interpretation, rather than the more specific result of resolution of a co-
indexing operation mediated by the detection of a PRO gap. However, using the same
structures with an ERP methodology, Demestre et al. (1999) found an effect-left
anterior negativity-which resembled those previously observed in response to a variety
of syntactic, as opposed to semantic, anomalies. Finally, concluding that PRO does
not exist would require a separate explanation for the results of our Experiment 2.14
In summary, we wanted to test how readers process subordinate infinitival sentences
containing gaps-PRO subjects-that appear in the syntactic structure without any previous
warning (beyond the control information provided by the matrix verbs’ subcategoriza-
tion grid). It is important to notice that such structures are different from other
previously-studied empty categories, such as wh-questions or wh-relatives, precisely in
that in these there is a previous signal that something has been moved out of its
original position, an indication that may trigger an ‘active’ search for the upcoming gap
(Frazier & Flores D’Arcais, 1989). The data obtained with eye-movement measures
showed that in the case of Spanish obligatory control constructions, readers seem to
13 We owe this point to an anonymous reviewer, to whom we are grateful.14 Two other factors might stand a better chance of explaining how precisely lexical subcategorization may be at
work when decoding these structures of control (we owe this observation to an anonymous reviewer). Of the 1007
occurrences of subject-control verbs used in our experiment, only 125 take infinitives as complements, while of
the 461 occurrences of object-control verbs, 201 take infinitives. Additionally, when we examined the number of
theta-grids of the two groups of verbs, it turned out that the subject-control group mean was 4.5 argument
structures per verb, but only 2.6 for the object-control group. One should be cautious, however, when considering
these discrepancies, as the role of verb bias in molding processing routines is hotly debated. For instance, Ferreira
and Henderson (1990) found evidence congruent with the claim that verb bias is only used in revision, but
Garnsey, Pearlmutter, Myers, and Lotocky (1997) report early effects of verb bias in guiding syntactic parsing
(see Hare, McRae, & Elman, 2004, for a revision and an explication of why Ferreira & Henderson, 1990, may
have failed to detect the early verb bias effects). On internal theta-grid competition causing a slow down of the
reading process at the verb (not after the verb), see Shapiro and Nagel (1993), and references therein. Schmauder
(1991) presents evidence of the opposite finding. With notable exceptions (e.g. Garnsey et al., 1997; Schmauder,
1991), research into verb bias tends to be done using different forms of cross-modal priming, and it has focused
primarily on the DO vs. Subject status of NPs like the decision in the senator regretted the decision. This
suggests that the issue is yet to be subjected to a thorough and healthy mix of cross-methodological scrutiny, as
well as to a wider variety of structures (including non-ambiguous structures like the ones we used). Our
experiment testifies to the role of lexical specification in determining PRO-linking. Gauging the fine grain of such
verbal lexical biases is, however, out of the scope of the present paper. The literature indicates that it is also an
open issue at present.
M. Betancort et al. / Cognition 100 (2006) 217–282238
make very fast use of the control information stored in the main verb. Thus, the
selection of the controller NP seems to rest on lexical factors. However, at the same
time, our data showed a preference for the recent NP filler, indicating that the selection
and assignment of the NP antecedent is not determined only by lexical factors, but also
by structural factors, such as the proximity of a NP to the empty category. Thus, the
present data could be accounted for by a constraint satisfaction model in which verb
information and recency are acting as important constraints, but are difficult to
accommodate in a two-stage model like Garden Path.
5. Experiment 2
Experiment 1 showed that readers make immediate use of verb control information
to recover the antecedent of the empty category PRO in Spanish obligatory control
constructions. At the same time, the data showed that this selection is not grounded on
lexical information alone. There was a preference for the recent NP filler as compared
to the distant NP filler, concurrent with the influence of verb control information. Thus,
it seems that the parser is influenced by the two types of information very early in the
time course of the processing of verb-guided control. Experiment 2 investigates another
type of structure, one in which control information is primarily induced by the
prepositions por (object control) vs. para (subject control), in order to test for lexical
guidance as well. Verb control is a well-known case of thematic control. Conversely,
control by prepositions is rarely mentioned in psycholinguistic accounts. As noted in the
introduction, an important difference between control dictated by verbal heads and
control induced by prepositional heads is that the latter is less extreme, in the sense that
even though it, too, is obligatory, its co-indexing is more open to the intervention of
interpretative forces. Thus, it is even possible to reverse the tendency of por sentences
to code object control and the even stronger tendency of para sentences to code subject
control. In short, the contrast between Experiment 1 and Experiment 2 is one involving
obligatory fixed control vs. obligatory but not fixed control, respectively (optional,
arbitrary control is out of the scope of the present paper). The contrast reflects the fact
that control by verbs occurs in an environment of tight integration of the infinitival
complement clause and its matrix. Purpose clauses and clauses of reason are not
complements but adjuncts of the matrix and therefore less integrated with it. In
Experiment 2 we used constructions that differ only in the preposition that headed
adverbial subordinate clauses. In addition to defining the kind of adverbial clause, the
preposition determines a semantic context that biases the interpretation of the empty
category either to a subject NP1 or an object NP2 in the matrix. One of these, then, is
the controller. In particular, we manipulated two different kinds of infinitival adverbial
clauses expressing purpose (preposition para) and reason (preposition por).
It is important to realize that the two adverbial clauses are syntactically identical in
Spanish (Marcos-Marın, Satorre, & Viejo, 1998, p. 442 ff.). The simplified tree in (16)
captures this:
M. Betancort et al. / Cognition 100 (2006) 217–282 239
The easy (but far from perfect) association between causal Spanish por and causal
English for at the phrasal level (I could not sleep for the pain in my leg/no podıa dormir
por el dolor de mi pierna; for lack of funds/por falta de fondos) might lead English
speakers to suspect that, since for occurs inside VP while purpose clauses with in order to
or to (which can often be preposed in English) may be seen as sentence adjuncts, we may
be comparing two different structures in Spanish as well.15 This is not the case. In fact,
both clauses are complements of their respective prepositions and the two PPs occupy
exactly the same place in the tree. In Spanish, the two prepositions are transitive and
subcategorise for either an NP complement (por mi padre, para mi padre), a non-finite
clause (por venir, para venir), or a full, finite clause (por-que vengas, para que vengas).
Indeed, it is hard to maintain that por mi padre and para mi padre must be seen as ordinary
VP adjuncts while por venir and para venir (which contain exactly the same prepositional
heads) must occur higher in the tree as sentence adverbials. But, even if one wishes to push
such an analysis, the analysis would apply to the two clauses in the same way. This is
crucial, because we mean to compare the control properties of different lexical heads in an
identical syntactic mold. Obviously, the symmetry evidenced by the complementation
grids licensed by the two prepositions in Spanish is absent in English, where to, and,
especially, in order to and so as to do license non-finite complements, like their Spanish
counterparts, whereas for does not (*I came here for seeing you; one must say I came here
15 The association between causal por and for is far form perfect even at the phrasal level as sometimes por is
translated using for (por falta de fondos/for lack of funds) while on some other occasions it is best translated by
because of (lo hice por mi padre/I did it because of my father). The symmetry between lo hice por mi padre and lo
hice por ayudar a mi padre in Spanish (the preposition expresses cause in the two sentences but does not change;
all that changes is the form of its complement, a NP and an infinitival clause, respectively) is not possible in
English, where causal phrasal for morphs into clausal because (I did it for the money/I did it because I wanted to
get money). So the ‘por’ structures in our experiment, which contain non-finite clauses, sometimes involve the
(finite) because version of the English translation, at least in their clearest translations. To complicate things
more, English for very often also corresponds with Spanish para (I did it for my father/Lo hice para mi padre). All
the materials are in Appendices A and B.
M. Betancort et al. / Cognition 100 (2006) 217–282240
because I wanted to see you, using a finite clause). In Spanish, the two adverbial clauses
may be preposed (Por venir a verte me perdı el tren esta manana (I missed the train this
morning because I came to see you); Para venir hasta aquı lo mejor es que cojas el tren
(To come here, you’d better use the train). The awkwardness of our English translations in
themselves show that the two types of clauses are regulated in syntactically different ways
in English, but not in Spanish, where they are identical.
As in Experiment 1, the predictions of two models of processing are tested in this
experiment. The Garden Path model assumes that a single analysis is initially chosen (see
Frazier, 1987; Frazier & Clifton, 1996; cf. Traxler et al., 1998) taking into account
structural information only. On the other hand, for Constraint-Based approaches
(MacDonald et al., 1994; Tanenhaus & Trueswell, 1995) structural information is only
one of the multiple constraints (plausibility, context, thematic information, etc.) at work in
a unitary-not two-stage-parsing process. If initial analyses are based on structural
information alone, readers should take the most recent NP (NP2) as a controller of PRO in
all cases, so no differences are predicted in early measures. However, more difficulty is
expected to be observed in late measures in the condition in which the most recent NP does
not fit with the control information provided by the preposition, that is, in the subject-
control sentences (preposition para). In these, reanalysis is expected to be visible. In other
words, the subject-control preposition condition should produce more difficulties in late
measures. In addition, a mismatch effect is expected when the gender of the adjective does
not agree with the right antecedent.
On the other hand, if control information induced by the preposition is initially
consulted as the main constraint to guide the selection of the right antecedent, then no
differences are to be expected between subject and object preposition control in the PRO
region either in early or late measures, as constraint satisfaction models suggest a unitary
one-stage process. Only slower reading times will be expected, due to the mismatch effect
between the right antecedent and the adjective when these do not agree in gender.
However, if control information is initially used together with recency as concurrent
constraints to guide antecedent selection, then more difficulties in the subject control
preposition condition are to be expected. These should be captured from the beginning in
very early measures.
6. Method
6.1. Participants
Thirty-two undergraduate students of the University of La Laguna participated in the
experiment for course credit. None of them had participated in the previous experiment.
6.2. Apparatus and procedure
The same apparatus and the same procedure as those in the previous experiment were
used.
M. Betancort et al. / Cognition 100 (2006) 217–282 241
6.3. Norming studies
We presented one hundred students with subordinate adverbial infinitival sentences
headed by the preposition para (subject-control bias) and por (object-control bias) with two
potential antecedents of the same gender in the main sentence. They were asked to read the
sentences and choose one of the two antecedents as implicit subjects of the infinitives
embedded in the subordinate clause. Four versions of the questionnaire were constructed with
48 experimental items and 128 fillers so that the position of the antecedents was balanced and
the order of presentation of items differently randomized. When the preposition that headed
the subordinate sentence was para (subject-control semantic bias), 75% of the readers (range
50–96%; standard errorZ2.1) identified the NP1 as subject of the infinitive, which was clearly
significantly different from chance in the analysis by subjects and by items [t(99)Z26.7,
P!0.001; t(31)Z12.1, P!0.001]. When the preposition that headed the subordinate clause
was por (object-control semantic bias), 78% of readers (range 50–97%; standard errorZ1.8)
identified the NP2 as subject of the infinitive, which was clearly significantly different from
chance in the analysis by subjects and by items [t(99)Z32.5, P!0.001; t(31)Z15.5,
P!0.001]. We take that to mean that in our structures readers identified optional, specific
control almost as clearly as the obligatory verb-guided control we examined in Experiment 1.
In addition, in order to check for the sub-categorization frame of the verbs used in our
experiment, we consulted the same data-base for Spanish verbs as in Experiment 1. All the
verbs used in our experiment subcategorize a direct object complement as a first option.
Identical sentences were used in the questionnaire and eye-tracking studies.
6.4. Materials and design
Forty-eight sets of four matched sentences with the structure NP1-VP1-NP2-
[PrepPRO1 infinitive-NP3-PRO2-infinitive-AdjP-PP] were constructed (see Table 3).
Table 3
Example of materials used in Experiment 2
Subject control bias (preposition “para”) match condition
Yolandai /se caso /con Jorgej /para [PRO1]i tener /dinero /y [PRO2]i ser /heredera /de una fortuna
Yolandai(fem) married Georgej(mas) in order to [PRO1]i have money and [PRO2]i be the heir (fem) to a fortune
Yolanda married George in order to have money and inherit a fortune
Subject control bias (preposition “para”) mismatch condition
*Yolandai/se caso/con Jorgej/para [PRO1]i tener/dinero/y [PRO2]i ser/heredero/de una fortuna
*Yolandai (fem) married Georgej (mas) in order to [PRO1]i have money and [PRO2]i be the heir (masc) to a fortune
Yolanda married George in order (for him) to have money and inherit a fortune
Object control bias (preposition “por”) match condition
Yolandai/se caso/con Jorgej/por [PRO1]j tener/dinero/y [PRO2]j ser/heredero/de una fortuna
Yolandai (fem) married Georgej (mas) for [PRO1]j having money and [PRO2]j for being the heir (masc) to a fortune
Yolanda married George because he has/had money and he is/was the heir to a fortune
Object control bias (preposition “por”) mismatch condition
*Yolandai/se caso/con Jorgej/por [PRO1]j tener/dinero/y [PRO2]j ser/heredera/de una fortuna
*Yolandai (fem) married Georgej (mas) for [PRO1]j having money and [PRO2]j for being the heir (fem) to a fortune
Yolanda married George because she has/had money and because she is /was the heir to a fortune
*Ungrammatical sentences due to the gender mismatch agreement.
Table 4a
Means and standard deviations of first pass reading times, regressions path duration, total reading times, and
second pass times for regions 3, 4, 5, 6, 7 and 8 of subject and object control match and mismatch sentences used
in Experiment 2
NP2 PRO-Infini-
tive
NP3 PRO2 AdjP PP
M. SD M. SD M. SD M. SD M. SD M. SD
First pass time
para-match 400 74 553 135 426 76 397 87 366 63 599 156.4
para-mismatch 405 82 556 111 444 96 413 93 380 108 588 184.6
por-match 400 88 498 121 450 97 407 86 385 89 588 167.2
por-mismatch 401 76 510 138 431 97 416 104 385 94 599 152.3
Regressions path duration
para-match 524 164 673 224 590 186 485 170 467 141 2011 1272.6
para-mismatch 546 185 680 172 620 232 471 128 498 257 2751 2041.8
por-match 534 171 542 459 618 243 527 228 545 413 1901 1090.6
por-mismatch 513 137 615 222 612 282 526 221 487 160 2396 1559
Total reading time
para-match 702 291 927 377 656 269 570 191 567 203 794 299.5
para-mismatch 811 337 1042 415 733 343 654 286 636 303 869 376.8
por-match 672 251 775 269 639 237 540 191 546 187 758 237.2
por-mismatch 743 296 875 392 693 276 630 253 624 248 833 297.4
Second pass time
para-match 254 252 320 298 174 158 155 112 174 159 5 15.7
para-mismatch 357 290 433 383 226 232 243 189 239 237 9 33.7
por-match 232 191 239 207 149 162 112 155 149 122 10 29.2
por-mismatch 301 261 318 288 214 193 189 243 205 179 7 28.8
M. Betancort et al. / Cognition 100 (2006) 217–282242
The prepositions that headed the subordinate sentence could be either para or por. The
lexical features of the preposition para biased the sentence to a subject control, purpose
clause reading in which the NP subject in the main sentence is the antecedent of both
PRO1 and PRO2. The lexical features of the preposition por biased the sentences to an
object control, reason clause reading in which the NP object of the main sentence is the
antecedent of both empty categories PRO1 and PRO2. As in Experiment 1, to ensure that
readers assigned the right NP antecedent to the PROs, as well as to investigate whether this
was an early or a late process, an adjective that could match or mismatch the gender of the
empty categories’ antecedent was inserted in the sentences. To insert the adjective we have
included a second, coordinated PROCinfinitive that was obligatorily linked to the first
one. The main reason for this was to avoid any possible parafoveal preview effects of
gender disambiguation at the first infinitive that could shadow any possible effects of the
preposition control. Therefore, if readers make a decision about which NP is the controller
of the first PRO during the reading of the infinitival verb, the second PRO will inherit this
and thus show up clearly which antecedent is chosen when reaching the adjective region
because the adjective will agree in gender only with one of the two possible NP
antecedents. Thus, for each preposition two versions of each sentence were written. For the
subject-biased preposition para in one version the adjective matched the gender of the
NP1, and in the other the adjective mismatched the gender of the NP1 antecedent.
Table 4b
Means (deviation from regression) and standard deviations of first pass reading times for regions 3, 4, 5, 6, 7 and 8
of subject-and object-control matched and mismatched sentences used in Experiment 2
First pass time
NP2 PRO-Infini-
tive
NP3 PRO2 AdjP PP
M SD M SD M SD M SD M SD M SD
Para-match K35 62 25 87 K32 51 K62 54 K56 56 39.1 109.8
para-mismatch K31 60 30 78 K14 67 K47 54 K43 61 34.6 133.3
por-match K35 66 1 79 K10 58 K48 58 K34 59 28.2 115.3
por-mismatch K35 57 13 80 K28 69 K44 73 K40 57 44.9 97.7
M. Betancort et al. / Cognition 100 (2006) 217–282 243
The same gender match vs. mismatch manipulation was constructed for sentences headed
by the preposition por. Thus, the only difference is that now control points to NP2 as the
antecedent of PRO (see Table 3). In addition, another 96 sentences with different types of
structures were written to serve as fillers, as well as 10 sentences for practice trials. Four
versions of the experiment were created, each containing 48 experimental sentences and
96 fillers. Only one sentence from each of the sets of four matched experimental sentences
appeared in each version of the experiment. Assignment of these sentences to each of the
four conditions was counterbalanced across versions. Any given subject saw the materials
from only one version. The resulting experimental situation was a 2!2 design with type of
preposition control (subject vs. object) and gender agreement (match vs. mismatch) as
within-subjects factors. The sentences used in Experiment 2 are presented in the
Appendix.
7. Results
For the purpose of the analyses, the texts were segmented as is shown with slashes in
Table 3: the first region contained the first NP (“Yolanda”); the second comprised the VP1
(“se caso”); the third the second NP (“con Jorge”);16 the fourth the PRO-infinitive
(“por/para [PRO] tener”); the fifth the NP3 (“dinero”); the sixth a PRO2 region (“y [PRO]
ser”); the seventh the AdjP region (“heredero/a”); the eighth the final PP region (“de una
fortuna”). This last region was included in order to capture any spillover effects from the
previous region, as well as to keep away any influence of wrap-up effects on the potential
structural effects. The mean first pass reading times, mean regression path times, mean
total reading times and mean second pass times for NP2, PRO1-infinitive, NP3, PRO2-
infinitive, AdjP and PP regions of the sentences are shown in Table 4a. Deviations from
regression for first pass reading times are shown in Table 4b. Reading times calculated
according to the measures described above were entered into ANOVAs. The error rate for
the comprehension questions was very low (between 2 and 4%) and did not differ among
the four experimental conditions.
16 Again, technically, ‘con Jorge’ is not a NP but a PP. ‘NP2’ refers to the NP inside the PP (Jorge).
M. Betancort et al. / Cognition 100 (2006) 217–282244
7.1. First pass reading times
Due to the different length in the PRO1-infinitive region between conditions the
deviations from regression in the first pass reading times analysis was used as a dependent
measure to prevent a possible length effect over time of fixation. The ANOVAS by-
subjects and by-items carried out separately showed that the interaction between type of
preposition and gender agreement was significant in the NP3 region (“dinero”)
[F1(1,35)Z4.17, p!0.05; F2(1,47)Z4.9, p!0.05], which seems to be a spurious effect
for three reasons at least: first, none of the simple effects was reliable in the by-subjects
analyses; second, gender disambiguating information appears only at the end of the
seventh region, the AdjP region (“heredero/a”), while this effect is taking place at the fifth
region (“dinero”), which makes quite unlikely any type of parafoveal preview benefit, and
third, it does not fit any theoretical claim. No other reliable effects of first pass times were
found in any of the other regions.
7.2. Regression path times
ANOVAS by-subjects and by-items carried out separately showed a type of sentence
(para vs. por) effect in the PRO1-infinitive region (“para/por tener”) [F1(1,35)Z5.83, P!0.05; F2(1,47)Z5.57, P!0.05]. Sentences headed by the preposition por (object-control
bias) were read faster than sentences headed by the preposition para (subject-control bias).
In contrast, the reverse type of sentence pattern was obtained in the PRO2-infinitive region
(“y ser”), although this was only reliable in the analysis by items [F1(1,35)Z2.21, P!0.08; F2(1,47)Z4.70, P!0.05]. Finally, in the PP region we found an effect of type of
sentence, although reliable only in the analysis by subjects [F1(1,35)Z6.60, P!0.05;
F2!1], and a gender agreement effect [F1(1,35)Z17.98, P!0.05; F2(1,47)Z21.15, P!0.05].
7.3. Total reading time
Reliable effects of type of sentence (para vs. por) were found in the NP2 region
analysis (“Jorge”) [F1(1,35)Z4.47, P!0.05; F2(1,47)Z5.35, P!0.05] as well as in the
PRO1-infinitive region (“para/por tener”) [F1(1,35)Z33.12, P!0.05; F2(1,47)Z34.83,
P!0.05], and in the PP region (“en la sala”) but only in the analysis by items [F1(1,35)Z2.42, P!0.12; F2(1,47)Z10.11, P!0.05]. In the three regions, sentences headed by the
preposition por (object-control bias) were read faster than sentences headed by the
preposition para (subject-control bias).
Reliable effects of gender agreement were found in the NP2 region analysis (“Jorge”)
[F1(1,35)Z14.26, P!0.05; F2(1,47)Z15.97, P!0.05], in the PRO1-infinitive region
(“para/por tener”) [F1(1,35)Z23.15, P!0.05; F2(1,47)Z16.54, P!0.05], in the NP3
region (“dinero”) [F1(1,35)Z6.97, P!0.05; F2(1,47)Z6.67, P!0.05], in the PRO2-
infinitive region (“y ser”) [F1(1,35)Z13.66, P!0.05; F2(1,47)Z13.16, P!0.05], in the
Adj-P region (“de heredero/a”) [F1(1,35)Z6.46, P!0.05; F2(1,47)Z12.14, P!0.05],
and in the PP region (“en la sala”) [F1(1,35)Z6.46, P!0.05; F2(1,47)Z10.11, P!0.05].
M. Betancort et al. / Cognition 100 (2006) 217–282 245
In all these regions, sentences in which there was a mismatch in gender agreement were
read more slowly than those in which there was a match.
7.4. Second pass time
Reliable effects of type of sentence (para vs. por) were found in the PRO1-infinitive
region (“para/por tener”) [F1(1,35)Z15.38, p!0.05; F2(1,47)Z15.48, P!0.05], as well
as in the PRO2-infinitive region (“y ser”). [F1(1,35)Z8.59, P!0.05; F2(1,47)Z11.20,
P!0.05], and finally in the Adj-P region (“heredero/a”) [F1(1,35)Z4.16, P!0.05;
F2(1,47)Z3.23, P!0.05]. In the three cases, sentences headed by the preposition por
(object-control bias) were read faster than sentences headed by the preposition para
(subject-control bias).
Reliable effects of gender agreement were found in the NP2 region analysis (“Jorge”)
[F1(1,35)Z20.72, P!0.05; F2(1,47)Z22.74, P!0.05], in the PRO1-infinitive region
(“para/por tener”) [F1(1,35)Z16.40, P!0.05; F2(1,47)Z17.87, P!0.05], in the NP3
region (“dinero”) [F1(1,35)Z7.97, P!0.05; F2(1,47)Z10.75, P!0.05], in the PRO2-
infinitive region (“y ser”). [F1(1,35)Z15.25, P!0.05; F2(1,47)Z23.73, P!0.05], and
finally in the Adj-P region (“heredero/a”) [F1(1,35)Z8.32, P!0.05; F2(1,47)Z13.86,
P!0.05]. In all these regions, sentences in which there was a mismatch in gender
agreement were read more slowly than those in which there was a match.
See Fig. 3 for a summary of the results.
Fig. 3. Regression path times (RPT), total reading times (TT) and second pass times (SPT) of subject and object
control match and mismatch sentences over PRO-Infinitive region in Experiment 2.
M. Betancort et al. / Cognition 100 (2006) 217–282246
8. Discussion
The data obtained in Experiment 2 suggest that during the processing of the empty
category PRO in purpose vs. reason adverbial subordinate infinitival clauses, the control
information induced by the prepositions por vs. para is not initially used as a constraint to
guide the selection of the nominal antecedent of PRO. The latencies obtained in late
measures, such as total time, regression path duration, and second pass time, show that the
purpose adverbial clauses headed by the preposition para (subject control) are more
difficult to process than the adverbial clauses of reason headed by the preposition por
(object control). These effects indicate that for readers it has been easier to recover the
antecedent of the empty category PRO when that antecedent was in the object position of
the main sentence (recent filler) than when it was in the subject position (distant filler). On
the other hand, the mismatch agreement effects suggest that readers quickly assigned a
controller to the empty category PRO. Sentences in which the gender of the correct NP
antecedent did not match the gender of the adjective exhibited higher reading times. The
gender congruence effect shows again that the process of antecedent assignment for PRO
is a fast one. When subjects reach the adjective they have already made a decision about
the antecedent of PRO, otherwise a gender agreement effect should not have been found
there. Therefore, our data provide evidence against some proposals (Fodor, 1989), and
against some findings obtained with the cross-modal priming methodology (Nicol, 1988;
Nicol & Swinney, 1989), both of which argue for a delay strategy in the assignment of
PRO antecedents.
The present data agree with the predictions one might attribute to a formal model of
parsing like the GP model: lack of initial effects, but more difficulties for subject-control
sentences (para) as compared to object-control sentences (por) because of the need for
reanalysis in the former. If, as the MRFS predicts, the most recent NP is taken as the
antecedent controlling PRO, a reanalysis is expected when the control information
delivered by the preposition does not agree with the initial choice, which happens in the
sentences with the preposition para. However, no additional cost is expected in sentences
with the preposition por, as in these the initial choice of the object NP coincides with the
control information contained in the preposition.
On the other hand, the current data do not support models that predict that syntactic
processing is initially guided by the interaction of different lexical and semantic
constraints. If the lexical information of control stored in the prepositions por and para
were initially consulted, we should not have found differences between sentences with
subject-and object-control prepositions, both in early and in late measures of processing.
The control information should have been quickly consulted and therefore no difficulties
should have arisen. Another possibility is that several constraints, such as control
information and recency, have been simultaneously taken into account. Under such an
assumption, predictions would depend on the relative strength of each constraint. If
preposition control information is a stronger constraint than recency, then an initial clash
between the two should be expected to appear from the very beginning in early measures,
something that our data does not suggest. If the recency constraint had been stronger than
the preposition control information, then predictions could actually be similar to those of a
two-stage model like GP, indicating that object-control sentences should be read initially
M. Betancort et al. / Cognition 100 (2006) 217–282 247
faster than subject-control ones. However, later reanalysis processes are more difficult to
accommodate in a constraint-satisfaction framework, as it is not immediately clear why
reanalysis should occur in a single-stage parse (see Frazier, 1998).
9. General discussion
We are now in a position to address globally the three fundamental objectives we set
out in our introduction. The first question we wanted to examine was whether control
information constrains initial filler-gap assignment (Frazier et al., 1983). When we
examined prototypical cases of verb-guided subject and object control, it became clear
that the answer to that question is yes. In Experiment 1, it was plain to see that even
before reaching the gap, readers made use of the control information provided by the
matrix verb and engaged in differential processing of subject-vs. object-control
structures. Since the gaps we used were not announced by the previous presence of a
dislocated filler (as in wh-traces, for instance), all that readers had at their disposal for
coping with the gap, when suddenly encountered, was the lexical information stored in
each verb’s lexical entry. Given that the verbs used in our experiment subcategorize NP
complements more frequently than infinitival complements (with gaps), the processor’s
rapid use of the right kind of lexical knowledge is all the more impressive. In addition
to lexical specification, our data also revealed the workings of recency as a concurrent
force. Even without any trace of delayed verb-control information (or reanalysis after
the gap), the subject-controlled sentences were harder to process than the object-control
ones (notice however the lack of recency effects in Mauner et al. (1995) with a stop-
making-sense task). Constraint-satisfaction models of processing can easily account
for such a pattern of results. A model like GP, however, would have more difficulty
explaining it.
The second objective we had was to find out whether the results obtained in our first
experiment on verb control could be replicated in other structures where control is not
rigidly determined by the verb. Indeed, most linguistic and psycholinguistic research on
controlled gaps focuses on verbs. However, given the frequency with which adverbial
clauses of reason and purpose, for example, recur in the actual use of language, it is
clearly arbitrary, as well as theoretically dangerous, to draw conclusions about control in
general while ignoring such frequent structures. So the second question we set out to
answer was whether control information constrains filler-gap assignment in structures
where the control is largely determined by prepositions. By examining subject-oriented
purpose clauses of control introduced by the preposition para (‘to/in order to’), and
object-oriented reason clauses of control introduced by the preposition por
(‘for/because’), it was possible to ascertain that the answer to that question is now no.
No trace of initial differential processing was found between the two sets of
constructions. What we did find was that only the subject-biased control sentences
showed clear reprocessing effects. Total reading times, second pass times, and regression
path times were always higher for para-sentences than for por-sentences. The fact that
early measures revealed no significant differences can be accommodated by both GP and
lexicalist models. The former would maintain that the parser opts for the most recent
M. Betancort et al. / Cognition 100 (2006) 217–282248
filler in all cases, but that such a choice can only become visible late, when the distant
filler interpretation becomes unavoidable (subject-controlled induced by para). The latter
would point to the lexical factor: no initial differences emerge because readers make
rapid use of control information granted by the preposition. However, there is only one
model that can explain the reprocessing effects, and that is the GP model. If, as
constraint-satisfaction theoreticians would claim, lexical knowledge guides antecedent
selection from the start, no re-analysis should have been observed.17
Note that there appears to be an interesting logic in the results of our two
experiments. In the first one, we attributed the early NP2 effect to a parafoveal preview
of the infinitive and its subject gap. A good reason to have to cope with the gap even
before fixating it is the fact that the lexical information about control provided by the
verbs makes parafoveal preview a more informative preview than it usually is. This is
because the parser is expecting a gap, which it sees parafoveally at the same time that it
is fixating NP2. A similar parafoveal processing of a gap which is not lexically activated
first is predicted to produce no NP2 effect. That situation obtains in Exp. 2. We suggest
that the interplay between lexical specification and recency provides a unitary account of
both experiments: in the first one, lexical signposts are so luminous that they have an
impact on the resolution of PRO by making the parser predisposed to expect PRO and
thus making the most of a parafoveal preview of it. In the second, there are no such
lexical lanterns illuminating the way, so only recency remains as a facilitating force.
Hence the increased difficulties found for para clauses. Hence also the relative
uninformativity of a parafoveal preview.
The third objective was to measure the speed with which the right antecedent of PRO
was activated. The literature on antecedent activation provides far from conclusive results
(see our introduction). In general, no trace of speedy activation has been found using the
cross-modal priming methodology (Nicol, 1988). Mixed results were obtained in probe
recognition tasks (slow activation in McElree & Bever, 1989, fast activation in Bever &
Sanz, 1997). Finally, other behavioral (Boland et al., 1990; Mauner et al., 1995) and
electrophysiological measures (Demestre et al., 1999; Featherston et al., 2000) have
revealed fast activation. The antecedent reactivation debate lends itself particularly well to
interesting connections with current linguistic theorizing. Thus, in view of the results
obtained in the priming studies (which were prior to the ERP studies), Sag and Fodor
(1995) made the point that, since wh-trace, but not PRO, has shown fast activation, only
the former, but not the latter, is likely to figure in syntactic trees. Since, additionally,
theories of grammar differ in the theoretical status they grant to the different types of EC,
Sag and Fodor conclude that the priming results provide indirect evidence in favor of
theories such as LFG or HPSG, which relegate treatment of PRO to the non-syntactic
domains of grammatical organization (loosely speaking, conceptual structure). Using a
methodology which is particularly suited to reveal the time-course of processing (eye-
tracking), what we have found is that the antecedent of PRO is consistently accessed very
fast. By introducing, immediately after the infinitive, an adjective phrase that either
17 One might reason that the reanalysis effect is not in fact due to reanalysis but to a late resolution of weak
constraints. We owe this point to an anonymous reviewer.
M. Betancort et al. / Cognition 100 (2006) 217–282 249
matched or mismatched the gender of the right antecedent, we could time-lock the binding
of antecedent and PRO. No trace of the slow recovery first reported by Nicol (1988) could
be found. Rather, our experiments lend strong support to the data obtained with
electrophysiological methodology. Thus, the lack of early effects could be just reflecting a
lack of sensitivity of the techniques and/or materials that have been previously used.
Clearly, however, more research is necessary to resolve this question.
Obviously, obtaining neat results does not automatically translate into a coherent
theory, especially if those results point to mutually contradictory conclusions.
We apparently have two important contradictions either in our results themselves or in
the way they relate to data obtained in previous research. The first one should be
obvious: in our first experiment we obtained solid evidence of both fast access to
lexical specification and, concurrently, of the rapid influence of recency. In our
second experiment, however, only recency could account for the facts. The results
from the first experiment fall squarely within the theoretical limits specified by
constraint-satisfaction models of processing, not only because they reveal a role of
lexical strength that is perfectly in line with their lexicalist orientation, but also
because they reveal the workings of more than just one factor in determining initial
interpretations. The results from our second experiment, however, can be better
accommodated by formal models of parsing, like GP, through their usual invocation
of syntactic or geometrical determinism: the geometrically-defined most recent filler
was consistently faster than the more distant one, and reanalysis effects revealed the
parser made a single bet. Of course, constraint satisfaction models could still account
for the results of Experiment 2 if they hypothesized that the recency constraint was
stronger than the lexical information provided by the preposition. However, an
explanation would still be needed for the fact that recency is considered of different
strength in Experiments 1 and 2.
The second contradiction has to do with the speed of activation of the right antecedent
of PRO, and has two ramifications. As the relationship between fillers and gaps is
presumably a semantic one of co-reference, Cowart and Cairns (1987), Fodor (1989),
Nicol (1988) and Nicol and Swinney (1989), among others, have reasoned that the late
activation results obtained in priming experiments reveal just that: If something involves
reference, then it is semantic, and if semantic, then it is late. In other words, the corollary
of the late activation findings is a conclusion that at least PRO is not to be expected in
syntactic trees (Sag & Fodor, 1995). The first side of the contradiction is that our results,
obtained with a methodology that would appear to be better suited to monitor the time
course of activation, sustain precisely the very opposite conclusion, namely, activation of
PRO subjects is unambiguously fast regardless of whether PRO is governed by a verb or
induced by a preposition. The second side of this contradiction has to do with a recent ERP
experiment by Featherston et al. (2000) in which they reveal that raising (as in Mary seems
to be nice, where Mary is raised from the position of subject of be nice to the actual
position of subject of the main verb seem, leaving a trace behind) is harder to process than
control (as in Mary tries to be nice, where the subject of the main verb tries is generated
already there and that of be nice is a co-referential covert PRO, which means there is no
trace left behind and no transformation either). Featherston et al. intriguingly propose that
raising is harder to process because it involves untying the workings of a movement
M. Betancort et al. / Cognition 100 (2006) 217–282250
transformation (see De Vincenci, 1998), whereas control need not do anything of the kind,
as PRO and the surface main clause subject do not move anywhere. The contradiction is:
Why is control, a presumably semantic issue involving (only?) co-reference, easier
and faster than syntactically-governed raising? In the rest of this paper we address each of
these two contradictions in turn
We suggest a possible way out of the first contradiction now. The prima facie facts
are clear: verb control is sufficiently powerful for a parser to feel confident that it will
never lead it down a garden path. That is, control is so grammatically entrenched in
verbs that it is not at all easy to find a subject-control verb like promise licensing object
control. This happens only on rare and specific occasions. Likewise, no object-control
verbs like allow ever occur in particular contexts or circumstances where they can code
subject control. This is the reason why, strangely enough, all theories of grammar
coincide in viewing control as lexically guided (see our introduction for some
grammatical evidence, but especially Culicover & Jackendoff, 2001, for all the evidence;
see also Fodor, 1988). In contrast, prepositions cannot present the parser with the same
kind of exception-free grammaticalization of control. Thus, even though there is a very
significant tendency for adverbial infinitival clauses of reason introduced by por to code
object control (as evidenced by our norming studies), it is possible to manipulate
contexts in order to promote a subject control with the same preposition. In Juan
abandono a Marıa por no atender a sus hijos (‘John left Mary because she did not take
care of his/her/their children’), the matrix verb abandonar (‘leave someone’) may easily
imply a recrimination by the subject of a certain behavior on the part of the object, so
when the infinitive appears it is all too natural to understand that it codes the reason for
the recrimination (the matrix object NP is the implicit subject of the recriminated action)
. However, in Juan le dijo eso a Marıa por fastidiar (‘John said that to Mary in order to
tease/annoy her’), decir (‘say’) does not have the negative connotations of abandonar, so
the causal connection coded by the upcoming infinitive is not so easy to expect, and, in
fact, such a structure uses por and subject control (Juan is the one doing the
teasing/annoying). Even the rather solid tendency of purpose clauses to code subject
control can be reversed as well, as in Juan trajo aquı a Marıa para hacer ese trabajo
(‘Juan brought Mary here to do that job’), where it is the object Mary that controls PRO
(see Haegemann 1991: 262). At least one conclusion is clear: control may be largely a
lexical affair, but if it is determined by verbs, then just as verbs code argument structure
rigidly (and in so doing determine the full structure of the sentence with their primary
and non-primary constituents), they also code grammaticalization of control seamlessly.
When control is induced by prepositions, however, these often need further
accompanying circumstances to cope with the job, so one cannot so properly speak of
‘preposition control’.18 Given our data, it seems that the parser has been opportunistic in
18 Notice that our norming studies do suggest less certainty when it comes to resolving the implicit subjects of
subordinate clauses of purpose and reason than those of complement clauses. Yet, given the rigidity of
complement clause control, it is a little surprising that correct PRO resolution did not reach at least the 90% mark
in these. We attribute this to performance limitations having to do with the relative awkwardness of
decontextualised experimental conditions and to subjects’ strategies that are also present in the questionnaire
situation.
M. Betancort et al. / Cognition 100 (2006) 217–282 251
taking advantage of the fact that verb control in particular is, in its very essence, a well-
regulated lexical affair, and that it is to be trusted, so the parser trusts it. When dealing
with control induced (not ‘dictated’) by prepositions, however, it faces a more blurred
panorama, so it prefers to fall back upon its usual processing reflex. That involves
invoking the most local connections out of sheer computing economy, or because of the
greater accessibility in memory of close constituents. Thus, recency and lexical
subcategorization reveal themselves as processing forces consistently employed by the
parser. The present experiments reveal interesting ways in which these two forces share
the same processing space.
The second contradiction referred to above offers us a new angle from which to see the
processing of PRO. In essence, the contradiction revolves around whether control is an
inherently semantic issue or not (Fodor, 1988). If it is assumed that syntactic
representations are computed more rapidly than semantic representations (Fodor, 1983),
then, under further habitual assumptions in the processing literature, one might expect the
binding of antecedent and co-referential PRO to be somewhat delayed (Cowart & Cairns,
1987; Fodor, 1989; Nicol, 1988; Nicol & Swinney, 1989). But since both
electrophysiological evidence (Demestre et al., 1999; Featherston et al., 2000) and our
own eye-tracking experiments have revealed a rapid resolution of that binding
relationship, we are, in principle, forced to conclude that semantic representations need
not be any slower than syntactic ones. The theoretical implications of this conclusion can
hardly be ignored. Notice that, unlike the case of wh-trace and NP-trace (about which
formal theories of grammar disagree, with GB/MP claiming that the two are syntactically-
governed and, for instance, GPSG insisting that only the former is; see Osterhout &
Swinney, 1993 on interesting processing implications of this), all formal models of
grammar do agree that control cannot be captured structurally alone (through the MDP, for
instance), so there would appear to be no escape from the conclusion that the rapid
resolution of the binding relationship is a rapid resolution of a semantic affair (Fodor,
1988).
An alternative to this view-we suggest-consists in abandoning the assumption that all
co-referential relations are exclusively semantic issues—as suggested by Cowart and
Cairns (1987), Nicol (1988), Nicol and Swinney (1989), or Fodor (1989), among
others-, and maintaining instead that they may be a part of grammatical form—if they
are lexically specified. Viewed in this way, there is an interesting parallelism in the
way lexical heads, especially verbs, project structure. Thus, just as particular verbs
determine which of their expansions are complements and which are adjuncts, they also
determine which of the co-referential relationships they project are coded as rigidly as
complements or as loosely as adjuncts. Lexically-controlled antecedent-PRO ties appear
to be an instance of the former. The different relationships between co-referential NPs
in discourse would appear to be instances of the latter. And just as complements are
expected to be processed quickly, relative to adjuncts, so are lexically-determined
referential constituents like PRO. This would at least partly explain the apparently
controversial finding of Featherston et al. (2000) of easier processing of lexically
controlled PROs relative to raising structures. Control bindings are processed fast
because they are launched from the same lexical platform that launches all fast
connections, be they between a verb and a complement NP, a complement that-clause,
M. Betancort et al. / Cognition 100 (2006) 217–282252
or a complement infinitive. In a complement that-clause, for instance, the matrix verb
determines (and automates) whether the subordinate verb surfaces in the indicative or
the subjunctive mood. In an infinitival clause, the matrix verb determines (and equally
automates) whether the subordinate infinitive is subject-or object-controlled. And the
form of complements is determined by argument structure (not semantic structure).19 In
sum, finding evidence of lexical guidance in control is not evidence of a kind of
processing necessarily ruled by sheer semantic structure. Viewed in this way, our
finding and that of others that PRO antecedent reactivation is fast, need not disturb or
greatly affect the habitual assumption in the processing literature that semantic
representations ought to “limp slightly behind” (Fiebach et al., 2001, p. 328) syntactic
ones. If control were truly only semantically driven, then we would surely expect it to
be processed late.
Appendix A. Sentences used in Experiment 1
We present rather literal translations in italics. Occasionally, whenever we feel that the
literal translation may not be transparent enough, we use single quotes to mark that and
add a second translation that is less literal.
1a. Juan prometio a Marıa mantenerse mas informado del problema de la empresa.
1a. Juan promised Marıa to keep himself more informed about the company’s
problem.
*1b. Juan prometio a Marıa mantenerse mas informada del problema de la empresa.
*1b. Juan promised Marıa to keep *herself* more informed about the company’s
problem.
1c. Juan aconsejo a Marıa mantenerse mas informada del problema de la empresa
1c. Juan advised Marıa to keep herself more informed about the company’s problem.
*1d. Juan aconsejo a Marıa mantenerse mas informado del problema de la empresa
*1d. Juan advised Marıa to keep *himself* more informed about the company’s
problem.
19 To a fairly large extent, the form of complements is shaped by the meaning of the matrix verb. However, as is
the norm in language, such a transparent mapping from semantics to form is far from complete. If it were, all
languages would share the same complement structures with the same verbs (compare, however, suasive
tell/decir in English and Spanish, respectively: John told me to go (infinitive) vs. Juan me dijo que me fuera
(subjunctive that-clause)), and, additionally, we would not expect to find the same predicate licensing
two different complement forms (a. I believe he is innocent b. I believe him to be innocent), or two different
predicates with basically the same meaning subcategorising two different complement forms, even in the same
language (a. Sean told me to leave, b. Sean said to me that I should leave; but note c. *Sean said to me to leave/a.
John was eating b. *John was devouring). Semantics is supposed to be the same in these cases but the
‘grammaticalization’ of the meaning (the argument structure or complement pattern chosen) is arbitrarily
different. That is, it is the product of capricious form.
M. Betancort et al. / Cognition 100 (2006) 217–282 253
2a. Pedro admitio ante Ester vivir demasiado centrado en el trabajo.
2a. Pedro admitted to Esther (to) living too centered (masc) on work.
*2b. Pedro admitio ante Ester vivir demasiado centrada en el trabajo
*2b. Pedro admitted to Esther (to) living too *centered (fem)* on work.
2c. Pedro acuso a Ester de vivir demasiado centrada en el trabajo.
2c. Pedro accused Ester of living too centered (fem) on work.
*2d. Pedro acuso a Ester de vivir demasiado centrado en el trabajo.
*2d. Pedro accused Ester of living too *centered (masc)* on work.
3a. Jose revelo a Eva estar muy desconectado de la polıtica de la empresa.
3a. Jose revealed to Eva being really cut off (masc) from the company’s politics.
*3b. Jose revelo a Eva estar muy desconectada de la polıtica de la empresa
*3b Jose revealed to Eva being really *cut off (fem)* from the company’s politics.
3c. Jose reprocho a Eva estar muy desconectada (fem) de la polıtica de la empresa
3c. Jose reproached Eva for being really cut off (fem) from the company’s politics.
*3d. Jose reprocho a Eva estar muy desconectado de la polıtica de la empresa
*3d. Jose reproached Eva for being really *cut off (masc)* from the company’s politics.
4a. Julian juro ante Marisol ser bastante frıo en el trato con los demas.
4a. ‘Julian swore to Marisol to being rather cold (masc) in the treatment of others’.
Julian swore to Marisol that he was rather cold (masc) in the treatment of others.
*4b. Julian juro ante Marisol ser bastante frıa en el trato con los demas
*4b. ‘Julian swore to Marisol to being rather *cold (fem)* in the treatment of others’.
Julian swore to Marisol that he was rather *cold (fem)* in the treatment of others.
4c. Julian culpo a Marisol de ser bastante frıa en el trato con los demas
4c. ‘Julian blamed/accused Marisol for/of being rather cold (fem) in the treatment of
others’.
Julian accused Marisol of being rather cold (fem) in the treatment of others.
*4d. Julian culpo a Marisol de ser bastante frıo en el trato con los demas
*4d. ‘Julian blamed/accused Marisol for/of being rather *cold (masc)* in the treatment
of others’.
Julian accused Marisol of being rather *cold (masc)* in the treatment of others.
5a. Luıs confeso a Monica ser muy explıcito con los companeros de oficina.
5a. Luıs confessed to Monica to being very explicit (masc) with the co-workers.
*5b. Luıs confeso a Monica ser muy explıcita con los companeros de oficina.
*5b. Luıs confessed to Monica to being very *explicit (fem)* with the co-workers.
5c. Luıs recomendo a Monica ser muy explıcita con los companeros de oficina
5c. ‘Luıs recommended Monica to be very explicit (fem) with the co-workers’.
Luis recommended to Monica that she be very explicit with the co-workers.
M. Betancort et al. / Cognition 100 (2006) 217–282254
*5d. Luıs recomendo a Monica ser muy explıcito con los companeros de oficina
*5d. Luıs recommended Monica to be very *explicit (masc)* with the co-workers.
Luis recommended to Monica that she be very *explicit (masc)* with the co-workers.
6a. Miguel garantizo a Carolina estar mas comprometido con la relacion.
6a. ‘Miguel assured Carolina to be more committed (masc) to the relationship’.
Miguel assured Carolina (that) he was more committed (masc) to the relationship
*6b. Miguel garantizo a Carolina estar mas comprometida con la relacion
*6b. ‘Miguel assured Carolina to be more *committed (fem)* to the relationship’.
Miguel assured Carolina (that) he was more *committed (fem)* to the relationship
6c. Miguel exigio a Carolina estar mas comprometida con la relacion
6c. ‘Miguel demanded Carolina be more committed (fem) to the relationship’.
Miguel demanded that Carolina (should) be more committed (fem) to the relationship.
*6d. Miguel exigio a Carolina estar mas comprometido con la relacion
*6d. ‘Miguel demanded Carolina be more *committed (masc)* to the relationship’.
Miguel demanded that Carolina (should) be more *committed (masc)* to the relationship.
7a. Pedro reconocio ante Luisa ser muy duro con los empleados de la fabrica.
7a. Pedro admitted to Luisa being very hard (masc) with the factory employees.
*7b. Pedro reconocio ante Luisa ser muy dura con los empleados de la fabrica
*7b. Pedro admitted to Luisa being very *hard (fem)* with the factory employees.
7c. Pedro urgio a Luisa ser muy dura con los empleados de la fabrica
7c. Pedro urged Luisa to be very hard (fem) with the factory employees.
7d. Pedro urgio a Luisa ser muy duro con los empleados de la fabrica
*7d. Pedro urged Luisa to be very *hard (masc)* with the factory employees.
8a. Antonio alardeo ante Ines de vivir un poco mas comprometido con la causa.
8a. Antonio boasted to Ines of/about living a little more committed (masc) to the cause.
*8b. Antonio alardeo ante Ines de vivir un poco mas comprometida con la causa.
*8b. Antonio boasted to Ines of/about living a little more *committed (fem)* to the
cause.
8c. Antonio animo a Ines a vivir un poco mas comprometida con la causa.
8c. Antonio encouraged Ines to live a little more committed (fem) to the cause.
*8d. Antonio animo a Ines a vivir un poco mas comprometido con la causa
*8d. Antonio encouraged Ines to live a little more *committed (masc)* to the cause.
9a. Silvia nego ante Andres ser muy estricta con la educacion de los ninos.
9a. Silvia denied to Andres being very strict (fem) in the children’s education.
M. Betancort et al. / Cognition 100 (2006) 217–282 255
*9b. Silvia nego ante Andres ser muy estricto con la educacion de los ninos
*9b. Silvia denied to Andres being very *strict (masc)* in the children’s education.
9c. Silvia encargo a Andres ser muy estricto con la educacion de los ninos
9c. Silvia asked/ordered Andres to be very strict (masc) in the children’s education.
Silvia entrusted Andres with being very strict in the children’s education.
*9d. Silvia encargo a Andres ser muy estricta con la educacion de los ninos
*9d. Silvia asked/ordered Andres to be very *strict (fem)* in the children’s education.
Silvia entrusted Andres with being very *strict (fem)* in the children’s education.
10a. Lucıa amenazo a Esteban con posar totalmente desnuda en la escuela de arte.
10a. ‘Lucıa threatened Esteban to pose totally nude (fem) in art class’.
Lucıa threatened Esteban with posing totally nude (fem) in art class.
*10b. Lucıa amenazo a Esteban con posar totalmente desnudo en la escuela de arte
*10b. ‘Lucıa threatened Esteban to pose totally *nude (masc)* in art class’.
Lucıa threatened Esteban with posing totally *nude (masc)* in art class.
10c. Lucıa obligo a Esteban a posar totalmente desnudo en la escuela de arte
10c. Lucıa obligated Esteban to pose totally nude (masc) in art class.
*10d. Lucıa obligo a Esteban a posar totalmente desnuda enla escuela de arte
*10d. Lucıa obligated Esteban to pose totally *nude (fem)* in art class.
11a. Marıa testifico ante Roberto ser la principal autora del doble asesinato.
11a. Marıa testified to Robert to being the principal perpetrator (fem) of the double
murder.
*11b. Marıa testifico ante Roberto ser el principal autor del doble asesinato
*11b. Marıa testified to Robert to being the principal *perpetrator (masc)* of the double
murder.
11c. Marıa forzo a Roberto a ser el principal autor del doble asesinato
11c. Marıa forced Robert to be the principal perpetrator (masc) of the double murder.
*11d. Marıa forzo a Roberto a ser la principal autora del doble asesinato
*11d. Marıa forced Robert to be the principal *perpetrator (fem)* of the double
murder.
12a. Yolanda acepto ante Luıs estar completamente borracha durante la fiesta.
12a. Yolanda admitted to Luıs (to) being completely drunk (fem) at the party.
*12b. Yolanda acepto ante Luıs estar completamente borracho durante la fiesta
*12b. Yolanda admitted to Luıs (to) being completely *drunk (masc)* at the party.
12c. Yolanda desafio a Luıs a estar completamente borracho durante la fiesta
12c. Yolanda dared/challenged Luıs to be completely drunk (masc) at the party.
*12d. Yolanda desafio a Luıs a estar completamente borracha durante la fiesta
*12d. Yolanda dared/challenged Luıs to be completely *drunk (fem)* at the party.
13a. Alicia acordo con Angel estar callada aproximadamente una hora.
13a. Alicia agreed with/made an agreement with Angel to be quiet (fem) approximately
an hour.
*13b. Alicia acordo con Angel estar callado aproximadamente una hora
*13b. Alicia agreed with/made an agreement with Angel to be *quiet (masc)*
approximately an hour.
13c. Alicia sugirio a Angel estar callado aproximadamente una hora
13c. ‘Alicia suggested Angel be quiet (masc) approximately an hour’.
Alicia suggested to Angel that he be quiet approximately an hour.
*13d. Alicia sugirio a Angel estar callada aproximadamente una hora
*13d. ‘Alicia suggested Angel be *quiet (fem)* approximately an hour’.
Alicia suggested to Angel that she be *quiet (fem)* approximately an hour.
14a. Josefina juro ante Enrique ser muy recatada durante el noviazgo.
14a. ‘Josefina swore to Enrique to being very demure (fem) during the engagement/
courtship’.
Josefina swore to Enrique that she was very demure during the engagement/courtship.
*14b. Josefina juro ante Enrique ser muy recatado durante el noviazgo
*14b. ‘Josefina swore to Enrique to being very *demure (masc)* during the
engagement/courtship’.
Josefina swore to Enrique that she was very *demure (masc)* during the
engagement/courtship.
14c. Josefina acuso a Enrique de ser muy recatado durante el noviazgo
14c. Josefina accused Enrique of being very demure (masc) during the engagement/
courtship.
*14d. Josefina acuso a Enrique de ser muy recatada durante el noviazgo
*14d. Josefina accused Enrique of being very *demure (fem)* during the
engagement/courtship.
15a. Lidia prometio a Francisco ser bastante cauta con los comentarios.
15a. Lidia promised Franciso to be quite cautious (fem) with her comments.
*15b. Lidia prometio a Francisco ser bastante cauto con los comentarios
*15b. Lidia promised Franciso to be quite *cautious (masc)* with her comments.
15c. Lidia exigio a Francisco ser bastante cauto con los comentarios
15c. ‘Lidia demanded Franciso to be quite cautious (masc) with his comments’.
Lidia demanded that Franciso (should) be quite cautious with his comments.
*15d. Lidia exigio a Francisco ser bastante cauta con los comentarios
*15d. ‘Lidia demanded Franciso to be quite *cautious (fem)* with his comments’.
Lidia demanded that Franciso (should) be quite *cautious (fem)* with his comments.
M. Betancort et al. / Cognition 100 (2006) 217–282256
M. Betancort et al. / Cognition 100 (2006) 217–282 257
16a. El juez admitio ante la abogada ser un tanto meticuloso con su trabajo.
16a. The judge (masc) admitted to the lawyer (fem) (to) being a bit meticulous (masc) in
his work.
*16b. El juez admitio ante la abogada ser un tanto meticulosa con su trabajo
*16b. The judge (masc) admitted to the lawyer (fem) (to) being a bit *meticulous (fem)*
in her work.
16c. El juez aconsejo a la abogada ser un tanto meticulosa con su trabajo
16c. The judge (masc) advised the lawyer (fem) to be a bit meticulous (fem) in her work.
*16d. El juez aconsejo a la abogada ser un tanto meticuloso con su trabajo
*16d. The judge (masc) advised the lawyer (fem) to be a bit *meticulous (masc)* in her
work.
17a. El profesor juro ante la directora ser muy benevolo en el trato con los demas.
17a. ‘The professor (masc) swore to the principal (fem) to being very benevolent (masc)
in his treatment of others’.
The professor (masc) swore to the principal (fem) that he was very benevolent in his
treatment of others.
*17b. El profesor juro ante la directora ser muy benevola en el trato con los demas
*17b. ‘The professor (masc) swore to the principal (fem) to being very *benevolent (fem)
* in his treatment of others’.
The professor (masc) swore to the principal (fem) that he was very *benevolent (fem)* in
his treatment of others.
17c. El profesor culpo a la directora de ser muy benevola en el trato con los demas
17c. ‘The professor (masc) blamed/accused the principal (fem) for/of being very
benevolent (fem) in her treatment of others’.
The professor (masc) accused the principal (fem) of being very benevolent (fem) in her
treatment of others.
*17d. El profesor culpo a la directora de ser muy benevolo en el trato con los demas
*17d. ‘The professor (masc) blamed/accused the principal (fem) for/of being very
*benevolent (masc)* in his treatment of others’.
The professor (masc) accused the principal (fem) of being very *benevolent (masc)*in his
treatment of others.
18a. El doctor aseguro a la paciente estar muy dispuesto a luchar hasta el final.
18a. ‘The doctor (masc) assured the patient (fem) to be very willing (masc) to fight till
the end’.
The doctor (masc) assured the patient (fem) (that) he was very willing to fight till the
end.
*18b. El doctor aseguro a la paciente estar muy dispuesta a luchar hasta el final
*18b. ‘The doctor (masc) assured the patient (fem) to be very *willing (fem)* to fight
till the end’.
M. Betancort et al. / Cognition 100 (2006) 217–282258
The doctor (masc) assured the patient (fem) (that) he was very *willing (fem)* to fight
till the end.
18c. El doctor recomendo a la paciente estar muy dispuesta a luchar hasta el final
18c. ‘The doctor (masc) recommended the patient (fem) to be very willing (fem) to
fight till the end’.
The doctor (masc) recommended to the patient (fem) that she be very willing to fight till
the end.
*18d. El doctor recomendo a la paciente estar muy dispuesto a luchar hasta el final
*18d. ‘The doctor (masc) recommended the patient (fem) to be very *willing (masc)*
to fight till the end’.
The doctor (masc) recommended to the patient (fem) that she be very *willing (masc)*
to fight till the end.
19a. El director confeso a la empleada ser muy estricto con el horario.
19a. The director (masc) confessed to the employee (fem) to being very strict (masc)
with the schedule.
*19b. El director confeso a la empleada ser muy estricta con el horario
*19b. The director (masc) confessed to the employee (fem) to being very *strict (fem)*
with the schedule.
19c. El director mando a la empleada ser muy estricta con el horario
19c. The director (masc) ordered the employee (fem) to be very strict (fem) with the
schedule.
*19d. El director mando a la empleada ser muy estricto con el horario
*19d. The director (masc) ordered the employee (fem) to be very *strict (masc)* with the
schedule.
20a. El presidente reconocio a la directora ser mucho mas ambicioso que el resto.
20a. The president (masc) acknowledged to the director (fem) being much more
ambitious (masc) than the rest.
*20b. El presidente reconocio a la directora ser mucho mas ambiciosa que el resto.
*20b. The president (masc) acknowledged to the director (fem) being much more
*ambitious (fem)* than the rest.
20c. El presidente encargo a la directora ser mucho mas ambiciosa que el resto.
20c. ‘The president (masc) asked/ordered the director (fem) to be much more ambitious
(fem) than the rest’.
The president (masc) entrusted the director (fem) with being much more ambitious (fem)
than the rest.
*20d. El presidente encargo a la directora ser mucho mas ambicioso que el resto.
*20d. ‘The president (masc) asked/ordered the director (fem) to be much more
*ambitious (masc)* than the rest’.
M. Betancort et al. / Cognition 100 (2006) 217–282 259
The president (masc) entrusted the director (fem) with being much more *ambitious
(masc)* than the rest.
21a. El abogado revelo a la diputada estar muy disgustado con el presidente.
21a. The lawyer (masc) revealed to the representative (fem) being very displeased
(masc) with the president.
*21b. El abogado revelo a la diputada estar muy disgustada con el presidente
*21b. The lawyer (masc) revealed to the representative (fem) being very *displeased
(fem)* with the president.
21c. El abogado reprocho a la diputada estar muy disgustada con el presidente
21c. The lawyer (masc) reproached the representative (fem) for being very displeased
(fem) with the president.
*21d. El abogado reprocho a la diputada estar muy disgustado con el presidente
*21d. The lawyer (masc) reproached the representative (fem) for being very *displeased
(masc)* with the president.
22a. El senador admitio ante la periodista ser extremadamente cauto tratanto el tema.
22a. The senator (masc) admitted to the journalist (fem) (to) being extremely cautious
(masc) dealing with the subject.
*22b. El senador admitio ante la periodista ser extremadamente cauta tratanto el tema
*22b. The senator (masc) admitted to the journalist (fem) (to) being extremely *cautious
(fem)* dealing with the subject.
22c. El senador aconsejo a la periodista ser extremadamente cauta tratanto el tema
22c. The senator (masc) advised the journalist (fem) to be extremely cautious (fem)
dealing with the subject.
*22d. El senador aconsejo a la periodista ser extremadamente cauto tratanto el tema
*22d. The senator (masc) advised the journalist (fem) to be extremely *cautious (masc)*
dealing with the subject.
23a. El diputado confirmo a la ministra estar muy involucrado en el fraude electoral.
23a. The representative (masc) confirmed to the government minister (fem) being very
involved (masc) in the election fraud.
*23b. El diputado confirmo a la ministra estar muy involucrada en el fraude electoral
*23b. The representative (masc) confirmed to the government minister (fem) being very
*involved (fem)* in the election fraud.
23c. El diputado obligo a la ministra a estar muy involucrada en el fraude electoral
23c. The representative (masc) obligated the government minister (fem) to be very
involved (fem) in the election fraud.
*23d. El diputado obligo a la ministra a estar muy involucrado en el fraude electoral
M. Betancort et al. / Cognition 100 (2006) 217–282260
*23d. The representative (masc) obligated the government minister (fem) to be very
*involved (masc)* in the election fraud.
24a. El policia confeso a la periodista ser muy discreto en declaraciones publicas.
24a. The policeman confessed to the journalist (fem) to being very discreet (masc) in
(making) public statements.
*24b. El policia confeso a la periodista ser muy discreta en declaraciones publicas
*24b. The policeman confessed to the journalist (fem) to being very *discreet (fem)* in
(making) public statements.
24c. El policia forzo a la periodista a ser muy discreta en declaraciones publicas
24c. The policeman forced the journalist (fem) to be very discreet (fem) in (making)
public statements.
*24d. El policia forzo a la periodista a ser muy discreto en declaraciones publicas
*24d. The policeman forced the journalist (fem) to be very *discreet (masc)* in (making)
public statements.
25a. La joven acepto ante su amigo estar muy obsesionada con el peso.
25a. The young girl admitted to her friend (masc) (to) being very obsessed (fem) with
her weight.
*25b. La joven acepto ante su amigo estar muy obsesionado con el peso
*25b. The young girl admitted to her friend (masc) (to) being very *obsessed (masc)*
with her weight.
25c. La joven recrimino a su amigo estar muy obsesionado con el peso
25c. The young girl recriminated her friend (masc) for being very obsessed (masc) with
his weight.
*25d. La joven recrimino a su amigo estar muy obsesionada con el peso
*25d. The young girl recriminated her friend (masc) for being very *obsessed (fem)*
with his weight.
26a. La profesora prometio al alumno ser mas respetuosa con la clase.
26a. The professor (fem) promised the student (masc) to be more respectful (fem) to the
class.
*26b. La profesora prometio al alumno ser mas respetuoso con la clase
*26b. The professor (fem) promised the student (masc) to be more *respectful (masc)* to
the class.
26c. La profesora sugirio al alumno ser mas respetuoso con la clase
26c. ‘The professor (fem) suggested the student (masc) be more respectful to the
class’
The professor (fem) suggested to the student (masc) that he be more respectful to the
class.
M. Betancort et al. / Cognition 100 (2006) 217–282 261
*26d. La profesora sugirio al alumno ser mas respetuosa con la clase
*26d. ‘The professor (fem) suggested the student (masc) be more *respectful (fem)* to
the class’.
The professor (fem) suggested to the student (masc) that she be more *respectful (fem)*
to the class.
class.
27a. La detenida nego ante el policıa estar muy involucrada en dicho asunto.
27a. The woman under arrest denied to the policeman being very involved (fem) in
said/the matter.
*27b. La detenida nego ante el policıa estar muy involucrado en dicho asunto
*27b. The woman under arrest denied to the policeman being very *involved (masc)* in
said/the matter.
27c. La detenida acuso al policıa de estar muy involucrado en dicho asunto
27c. The woman under arrest accused the policeman of being very involved (masc) in
said/the matter.
*27d. La detenida acuso al policıa de estar muy involucrada en dicho asunto
*27d. The woman under arrest accused the policeman of being very *involved (fem)* in
said/the matter.
28a. La maestra prometio al director ser muy objetiva con ese alumno.
28a. The teacher (fem) promised the principal (masc) to be very objective (fem) with
that student.
*28b. La maestra prometio al director ser muy objetivo con ese alumno
*28b. The teacher (fem) promised the principal (masc) to be very *objective (masc)* with
that student.
28c. La maestra urgio al director ser muy objetivo con ese alumno
28c. The teacher (fem) urged the principal (masc) to be very objective (masc) with that
student.
*28d. La maestra urgio al director ser muy objetiva con ese alumno
*28d. The teacher (fem) urged the principal (masc) to be very *objective (fem)* with that
student.
29a. La abogada alego ante el juez ser muy solidaria con el defendido.
29a. The lawyer (fem) claimed to the judge (masc) to be very supportive (fem) of the
defendant.
*29b. La abogada alego ante el juez ser muy solidario con el defendido
*29b. The lawyer (fem) claimed to the judge (masc) to be very *supportive (masc)* of the
defendant.
29c. La abogada animo al juez a ser muy solidario con el defendido
M. Betancort et al. / Cognition 100 (2006) 217–282262
29c. The lawyer (fem) encouraged the judge (masc) to be very supportive (masc) of the
defendant.
*29d. La abogada animo al juez a ser muy solidaria con el defendido
*29d. The lawyer (fem) encouraged the judge (masc) to be very *supportive (fem)* of the
defendant.
30a. La secretaria alego ante el director estar muy descontenta con el sueldo.
30a. ‘The secretary (fem) complained to the director (masc) that she was very
dissatisfied (fem) with her salary’.
The secretary argued with the director over her dissatisfaction with her salary.
*30b. La secretaria alego ante el director estar muy descontento con el sueldo
*30b. ‘The secretary (fem) complained to the director (masc) that she was very
*dissatisfied (masc)* with her salary’.
The secretary argued with the director over *his* dissatisfaction with her salary.
30c. La secretaria reprocho al director estar muy descontento con el sueldo
30c. The secretary (fem) reproached the director (masc) for being very dissatisfied
(masc) with his salary.
*30d. La secretaria reprocho al director estar muy descontenta con el sueldo
*30d. The secretary (fem) reproached the director (masc) for being very *dissatisfied
(fem)* with his salary.
31a. La actriz desmintio ante el director estar nerviosa durante la actuacion.
31a. The actress denied to the director (masc) being nervous (fem) during the
performance.
*31b. La actriz desmintio ante el director estar nervioso durante la actuacion
*31b. The actress denied to the director (masc) being *nervous (masc)* during the
performance.
31c. La actriz recrimino al director estar nervioso durante la actuacion
31c. The actress recriminated the director (masc) for being nervous (masc) during the
performance.
*31d. La actriz recrimino al director estar nerviosa durante la actuacion
*31d. The actress recriminated the director (masc) for being *nervous (fem)* during the
performance.
32a. La diputada declaro al general estar muy involucrada en la caıda del presidente.
32a. The representative (fem) declared to the general (masc) being very involved (fem)
in the president’s downfall.
*32b. La diputada declaro al general estar muy involucrado en la caıda del presidente
*32b. The representative (fem) declared to the general (masc) being very *involved
(masc)* in the president’s downfall.
M. Betancort et al. / Cognition 100 (2006) 217–282 263
32c. La diputada obligo al general a estar muy involucrado en la caıda del presidente
32c. The representative (fem) obligated the general (masc) to be very involved (masc) in
the president’s downfall.
*32d. La diputada obligo al general a estar muy involucrada en la caıda del presidente
*32d. The representative (fem) obligated the general (masc) to be very *involved (fem)*
in the president’s downfall.
Appendix B. Sentences used in Experiment 2
We have tried as much as possible to translate our ‘por clauses’ using equivalent non-
finite clauses in English (with gerunds instead of infinitives). Whenever that has resulted in
an awkward translation we have added a finite version with ‘because’.
1a. Juan se enfado con Marıa para irse del baile y parecer estupido ante todos.
1a. Fran got angry with Marıa (in order) to go to the dance and look stupid (masc) in
front of everyone.
*1b. Juan se enfado con Marıa para irse del baile y parecer estupida ante todos
*1b. Fran got angry with Marıa (in order) to go to the dance and look *stupid (fem)* in
front of everyone.
1c. Juan se enfado con Marıa por irse del baile y parecer estupida ante todos.
1c. Fran got angry with Marıa for going to the dance and looking stupid (fem) in front
of everyone.
*1d. Juan se enfado con Marıa por irse del baile y parecer estupido ante todos.
*1d. Fran got angry with Marıa for going to the dance and looking *stupid (masc)* in
front of everyone.
2a. Marıa abofeteo a Jorge para montar una escena y ser cınica con los demas
2a. Marıa slapped Jorge (in order) to make a scene and be cynical (fem) with the
others.
*2b. Marıa abofeteo a Jorge para montar una escena y ser cınico con los demas
*2b. Marıa slapped Jorge (in order) to make a scene and be *cynical (masc)* with the
others.
2c. Marıa abofeteo a Jorge por montar una escena y ser cınico con los demas.
2c. Marıa slapped Jorge for making a scene and being cynical (masc) with the others.
*2d. Marıa abofeteo a Jorge por montar una escena y ser cınica con los demas.
*2d. Marıa slapped Jorge for making a scene and being *cynical (fem)* with the others.
3a. Esteban critico a Marta para presumir ante todos y proclamarse dueno del club.
3a. Esteban criticized Marıa (in order) to show off to others and proclaim himself
owner of the club.
M. Betancort et al. / Cognition 100 (2006) 217–282264
*3b. Esteban critico a Marta para presumir ante todos y proclamarse duena del club
*3b. Esteban criticized Marıa (in order) to show off to others and proclaim *herself*
owner of the club.
3c. Esteban critico a Marta por presumir ante todos y proclamarse duena del club.
3c. Esteban criticized Marıa for showing off to others and proclaiming herself owner
of the club.
*3d. Esteban critico a Marta por presumir ante todos y proclamarse dueno del club.
*3d. Esteban criticized Marıa for showing off to others and proclaiming *himself*
owner of the club.
4a. Julia dejo a Luıs para olvidarse de todo y marcharse de monja a un monasterio
4a. Julia left Luis (in order) to forget about/leave behind everything and to go off and
become a nun in a monastery.
*4b. Julia dejo a Luıs para olvidarse de todo y marcharse de monje a un monasterio
*4b. Julia left Luıs (in order) to forget about/leave behind everything and to go off to
become a *monk* in a monastery.
4c. Julia dejo a Luıs por olvidarse de todo y marcharse de monje a un monasterio.
4c. Julia left Luıs for forgetting about/leaving behind everything and going off to
become a monk in a monastery.
4c. Julia left Luıs because he forgot about/ left behind everything and went off to
become a *monk* in a monastery.
*4d. Julia dejo a Luıs por olvidarse de todo y marcharse de monja a un monasterio
*4d. ‘Julia left Luıs for forgetting about/leaving behind everything and going off to
become a *nun* in a monastery’.
*4d. Julia left Luıs because he/she forgot about/ left behind everything and went off to
become a *nun* in a monastery.
5a. Pedro acuso a Lucıa para hacer trampa y nombrarse heredero del dinero familiar
5a. Pedro accused Lucıa (in order) to cheat and name himself heir to the family
money.
*5b. Pedro acuso a Lucıa para hacer trampa y nombrarse heredera del dinero familiar
*5b. Pedro accused Lucıa (in order) to cheat and name herself *heiress* to the family
money.
5c. Pedro acuso a Lucıa por hacer trampa y nombrarse heredera del dinero familiar.
5c. Pedro accused Lucıa for cheating and naming herself heiress to the family money’.
5c. Pedro accused Lucıa of cheating and naming herself heiress to the family money.
*5d. Pedro acuso a Lucıa por hacer trampa y nombrarse heredero del dinero familiar.
*5d. Pedro accused Lucıa for cheating and naming himself * heir* to the family
money.
*5d. Pedro accused Lucıa of cheating and naming himself *heir* to the family
money.
6a. Luisa peleo con Antonio para huir sutilmente y ser mala con los hijos
6b. Luisa fought/argued with Antonio (in order) to flee subtly and be bad (fem) with the
children.
*6b. Luisa peleo con Antonio para huir sutilmente y ser malo con los hijos
*6b. Luisa fought/argued with Antonio (in order) to flee subtly and be *bad (masc)*
with the children.
6c. Luisa peleo con Antonio por huir sutilmente y ser malo con los hijos.
6c. Luisa fought/argued with Antonio for fleeing subtly and being bad (masc) with the
children’.
6c. Luisa fought with Antonio because he fled subtly and was bad (masc) with the
children.
*6d. Luisa peleo con Antonio por huir sutilmente y ser mala con los hijos
*6d. ‘Luisa fought/argued with Antonio because he fled subtly and was *bad (fem)*
with the children’.
*6d. Luisa fought/argued with Antonio because he/she fled subtly and was *bad (fem)*
with the children.
7a. Antonio se encaro con Marıa para enfadarse con todos y marcharse solo del
cine
7a. Antonio faced/stood up to Marıa (in order) to get angry with everyone and leave
the cinema alone (masc).
*7b. Antonio se encaro con Marıa para enfadarse con todos y marcharse sola del cine
*7b. Antonio faced/stood up to Marıa (in order) to get angry with everyone and leave
the cinema *alone (fem)*.
7c. Antonio se encaro con Marıa por enfadarse con todos y marcharse sola del cine.
7c. Antonio faced/stood up to Marıa for getting angry with everyone and leaving the
cinema alone (fem).
*7d. Antonio se encaro con Marıa por enfadarse con todos y marcharse solo del cine
*7d. Antonio faced/stood up to Marıa for getting angry with everyone and leaving the
cinema *alone (masc)*.
8a. Juan se peleo con Cecilia para huir del trabajo y hacerse voluntario en Peru
8a. Juan fought/argued with Cecilia (in order) to run away from work and become a
volunteer (masc) in Peru.
*8b. Juan se peleo con Cecilia para huir del trabajo y hacerse voluntaria en Peru
*8b. Juan fought/argued with Cecilia (in order) to run away from work and become a
*volunteer (fem)* in Peru.
8c. Juan se peleo con Cecilia por huir del trabajo y hacerse voluntaria en Peru.
8c. Juan fought/argued with Cecilia for running away from work and becoming a
volunteer (fem) in Peru.
*8d. Juan se peleo con Cecilia por huir del trabajo y hacerse voluntario en Peru.
M. Betancort et al. / Cognition 100 (2006) 217–282 265
M. Betancort et al. / Cognition 100 (2006) 217–282266
*8d. Juan fought/argued with Cecilia for running away from work and becoming a
*volunteer (masc)* in Peru.
9a. El medico hablo con la paciente para tener datos y estar informado del caso.
9a. The doctor (masc) spoke with the patient (fem) (in order) to have/obtain
information and keep himself informed about the case.
*9b. El medico hablo con la paciente para tener datos y estar informada del caso
*9b. The doctor (masc) spoke with the patient (fem) (in order) to have/obtain
information and keep *herself* informed about the case.
9c. El medico hablo con la paciente por tener datos y estar informada del caso.
9c. The doctor (masc) spoke with the patient (fem) for having information and being
informed (fem) about the case.
9c. The doctor (masc) spoke with the patient (fem) because she had information and
was informed (fem) about the case.
*9d. El medico hablo con la paciente por tener datos y estar informado del caso.
*9d. The doctor (masc) spoke with the patient (fem) for having information and being
informed (masc) about the case.
*9d. The doctor (masc) spoke with the patient (fem) because he/she had information and
was *informed (masc)* about the case.
10a. Lucıa felicito a Pedro para ganar la votacion y salir elegida en el comite
10a. Lucıa congratulated Pedro (in order) to win the vote and be elected (fem) to the
committee.
*10b. Lucıa felicito a Pedro para ganar la votacion y salir elegido en el comite
*10b. Lucıa congratulated Pedro (in order) to win the vote and be *elected (masc)* to
the committee.
10c. Lucıa felicito a Pedro por ganar la votacion y salir elegido en el comite.
10c. Lucıa congratulated Pedro for winning the vote and being elected (masc) to the
committee.
*10d. Lucıa felicito a Pedro por ganar la votacion y salir elegida en el comite
*10d. Lucıa congratulated Pedro for winning the vote and being *elected (fem)* to the
committee.
11a. Montse critico a Julian para sacar provecho y hacerse rica con la empresa
11a. Montse criticized Julian (in order) to derive benefit/benefit (from the situation) and
become rich (fem) in the company.
*11b. Montse critico a Julian para sacar provecho y hacerse rico con la empresa
*11b. Montse criticized Julian (in order) to derive benefit/benefit (from the situation) and
become *rich (masc)* in the company.
11c. Montse critico a Julian por sacar provecho y hacerse rico con la empresa.
M. Betancort et al. / Cognition 100 (2006) 217–282 267
11c. Montse criticized Julian for deriving benefit/ benefiting (from the situation) and
becoming rich (masc) in the company.
*11d. Montse critico a Julian por sacar provecho y hacerse rica con la empresa
*11d. Montse criticized Julian for deriving benefit/ benefiting (from the situation) and
becoming *rich (fem)* in the company.
12a. Juan apoyo a Laura para despedir al contable y hacerse dueno de la situacion
12a. Juan supported/backed Laura (in order) to fire the accountant and take
control/command (masc) of the situation.
*12b. Juan apoyo a Laura para despedir al contable y hacerse duena de la situacion
*12b. Juan supported/backed Laura (in order) to fire the accountant and *take
control/command (masc)* of the situation.
12c. Juan apoyo a Laura por despedir al contable y hacerse duena de la situacion.
12c. Juan supported/backed Laura for firing the accountant and her taking
control/command of the situation.
*12d. Juan apoyo a Laura por despedir al contable y hacerse dueno de la situacion.
*12d. Juan supported/backed Laura for firing the accountant and *his* taking
control/command of the situation.
13a. Nacho felicito a Sonia para ganarse al director y convertirse en jefe de ventas
13a. Nacho congratulated Sonia (in order) to win over the director and become head
(masc) of sales.
*13b. Nacho felicito a Sonia para ganarse al director y convertirse en jefa de ventas
*13b. Nacho congratulated Sonia (in order) to win over the director and become *head
(fem)* of sales.
13c. Nacho felicito a Sonia por ganarse al director y convertirse en jefa de ventas.
13c. Nacho congratulated Sonia for winning over the director and becoming head (fem)
of sales.
*13d. Nacho felicito a Sonia por ganarse al director y convertirse en jefe de ventas.
*13d. Nacho congratulated Sonia for winning over the director and becoming *head
(masc)* of sales.
14a. La directora despidio al contable para robar dinero y hacerse rica rapidamente
14a. The director (fem) fired the accountant (masc) (in order) to steal money and
become rich (fem) quickly.
*14b. La directora despidio al contable para robar dinero y hacerse rico rapidamente
*14b. The director (fem) fired the accountant (masc) (in order) to steal money and
become *rich (masc)* quickly.
14c. La directora despidio al contable por robar dinero y hacerse rico rapidamente.
M. Betancort et al. / Cognition 100 (2006) 217–282268
14c. The director (fem) fired the accountant (masc) for stealing money and becoming
rich (masc) quickly.
*14d. La directora despidio al contable por robar dinero y hacerse rica rapidamente
*14d. The director (fem) fired the accountant (masc) for stealing money and becoming
*rich (fem)* quickly.
15a. Violeta contrato a Pedro para ganar el proyecto y llegar a jefa del centro.
15a. Violeta hired Pedro (in order) to win the project and become the boss (fem) of the
center.
*15b. Violeta contrato a Pedro para ganar el proyecto y llegar a jefe del centro.
*15b. Violeta hired Pedro (in order) to win the project and become the *boss (masc)* of
the center.
15c. Violeta contrato a Pedro por ganar el proyecto y llegar a jefe del centro.
15c. VioletahiredPedro forwinning theproject andbecoming theboss (masc)of thecenter.
15c. Violeta hired Pedro because he won the project and became the boss of the center.
*15d. Violeta contrato a Pedro por ganar el proyecto y llegar a jefa del centro.
*15d. Violeta hired Pedro for winning the project and becoming the *boss (fem)* of the
center.
*15d. Violeta hiredPedrobecause he/shewon theproject andbecame the *boss (fem)* of the
center.
16a. Juan se divorcio de Lucıa para emigrar al Nepal y empezar solo una nueva vida
16a. JuandivorcedLucıa (in order) to emigrate toNepal and begin a new life alone (masc).
*16b. Juan se divorcio de Lucıa para emigrar al Nepal y empezar sola una nueva vida.
*16b. Juan divorced Lucıa (in order) to emigrate to Nepal and begin a new life *alone
(fem)*.
16c. Juan se divorcio de Lucıa por emigrar al Nepal y empezar sola una nueva vida.
16c. ‘Juan divorced Lucıa for emigrating to Nepal and beginning a new life alone (fem).
16c. JuandivorcedLucıa because she emigrated toNepal and begananew life alone (fem).
*16d. Juan se divorcio de Lucıa por emigrar al Nepal y empezar solo una nueva vida.
*16d. JuandivorcedLucıa for emigrating toNepal and beginning a new life *alone (masc)*.
*16d. Juan divorced Lucıa because he/she emigrated to Nepal and began a new life *alone
(masc)*.
17a. Marıa se enfrento a Juan para escurrir el bulto y hacerse la sueca ante el juez.
17a. Marıa confronted Juan (in order) to dodge the issue and to play dumb (fem) in
front of the judge.
*17b. Marıa se enfrento a Juan para escurrir el bulto y hacerse el sueco ante el juez.
*17b. Marıa confronted Juan (in order) to dodge the issue and to play *dumb (masc)* in
front of the judge.
17c. Marıa se enfrento a Juan por escurrir el bulto y hacerse el sueco ante el juez.
17c. Marıa confronted Juan for dodging the issue and playing dumb (masc) in front of
M. Betancort et al. / Cognition 100 (2006) 217–282 269
the judge.
*17d. Marıa se enfrento a Juan por escurrir el bulto y hacerse la sueca ante el juez.
*17d. Marıa confronted Juan for dodging the issue and playing *dumb (fem)* in front of
the judge.
18a. Juan rino con Lisa para liar la situacion y quedar disculpado ante el decanato
18a. Juan quarreled with Lisa (in order) to complicate the situation and get
excused/pardoned (masc) by the dean’s office.
*18b. Juan rino con Lisa para liar la situacion y quedar disculpada ante el decanato
*18b. Juan quarreled with Lisa (in order) to complicate the situation and get
*excused/pardoned (fem)* by the dean’s office.
18c. Juan rino con Lisa por liar la situacion y quedar disculpada ante el decanato.
18c. Juan quarreled with Lisa for complicating the situation and getting excused/
pardoned (fem) by the dean’s office.
*18d. Juan rino con Lisa por liar la situacion y quedar disculpado ante el decanato
*18d. Juan quarreled with Lisa for complicating the situation and getting *excused/
pardoned (masc)* by the dean’s office.
19a. Teresa alabo a Martın para captar votos y resultar ganadora en el concurso
19a. Teresa praised Martin (in order) to capture votes and come out the winner (fem) of
the contest.
*19b. Teresa alabo a Martın para captar votos y resultar ganador en el concurso
*19b. Teresa praised Martin (in order) to capture votes and come out the *winner (masc)
* of the contest.
19c. Teresa alabo a Martın por captar votos y resultar ganador en el concurso.
19c. Teresa praised Martin for capturing votes and coming out the winner (masc) of the
contest.
*19d. Teresa alabo a Martın por captar votos y resultar ganadora en el concurso
*19d. Teresa praised Martin for capturing votes and coming out the *winner (fem)* of
the contest.
20a. Maite chantajeo a Juan para falsear datos y declararse vencedora de la votacion
20a. Maite blackmailed Juan (in order) to falsify information and declare herself
winner of the vote/election.
*20b. Maite chantajeo a Juan para falsear datos y declararse vencedor de la votacion
*20b. Maite blackmailed Juan (in order) to falsify information and declare *himself*
winner of the vote/election.
20c. Maite chantajeo a Juan por falsear datos y declararse vencedor de la votacion.
20c. Maite blackmailed Juan for falsifying information and declaring himself winner
(masc) of the vote/election.
M. Betancort et al. / Cognition 100 (2006) 217–282270
*20d. Maite chantajeo a Juan por falsear datos y declararse vencedora de la votacion
*20d. Maite blackmailed Juan for falsifying information and declaring *herself* winner
of the vote/election.
21a. La maestra grito al alumno para llamar la atencion y darselas de lista en clase.
21a. The teacher (fem) screamed at the student (masc) (in order) to attract attention and
act smart (fem) in class.
*21b. La maestra grito al alumno para llamar la atencion y darselas de listo en clase
*21b. The teacher (fem) screamed at the student (masc) (in order) to attract attention and
act *smart (masc)* in class.
21c. La maestra grito al alumno por llamar la atencion y darselas de listo en clase.
21c. The teacher (fem) screamed at the student (masc) for attracting attention and
acting smart (masc) in class.
*21d. La maestra grito al alumno por llamar la atencion y darselas de lista en clase.
*21d. The teacher (fem) screamed at the student (masc) for attracting attention and
acting *smart (fem)* in class.
22a. Julia se encaro con Andres para cambiar el plan y sentirse unica responsable.
22a. Julia faced/stood up to Andres (in order) to change the plan and think herself the
sole person in charge.
*22b. Julia se encaro con Andres para cambiar el plan y sentirse unico responsable
*22b. Julia faced/stood up to Andres (in order) to change the plan and think*himself* the
sole (*masc) person in charge.
22c. Julia se encaro con Andres por cambiar el plan y sentirse unico responsable.
22c. Julia faced/stood up to Andres for changing the plan and thinking himself the sole
(masc) person in charge.
*22d. Julia se encaro con Andres por cambiar el plan y sentirse unica responsable.
*22d. Julia faced/stood up to Andres for changing the plan and thinking *herself* the
sole *(fem) person in charge.
23a. La abogada critico al juez para retrasar el caso y darselas de dura en la sala.
23a. The lawyer (fem) criticized the judge (masc) (in order) to delay the case and act
harsh (fem) in the courtroom.
*23b. La abogada critico al juez para retrasar el caso y darselas de duro en la sala.
*23b. The lawyer (fem) criticized the judge (masc) (in order) to delay the case and act
*harsh (masc)* in the courtroom.
23c. La abogada critico al juez por retrasar el caso y darselas de duro en la sala.
23c. The lawyer (fem) criticized the judge (masc) for delaying the case and acting harsh
(masc) in the courtroom.
*23d. La abogada critico al juez por retrasar el caso y darselas de dura en la sala.
M. Betancort et al. / Cognition 100 (2006) 217–282 271
*23d. The lawyer (fem) criticized the judge (masc) for delaying the case and acting
*harsh (fem)* in the courtroom.
24a. La doctora estudio al enfermo para ofrecer datos y estar dispuesta a todo.
24a. The doctor (fem) studied the patient (masc) (in order) to offer information and be
willing (fem) (to do anything).
*24b. La doctora estudio al enfermo para ofrecer datos y estar dispuesto a todo.
*24b. The doctor (fem) studied the patient (masc) (in order) to offer information and be
*willing (masc)* (to do anything).
24c. La doctora estudio al enfermo por ofrecer datos y estar dispuesto a todo.
24c. The doctor (fem) studied the patient (masc) for offering information and being
willing (masc) (to do anything).
24c. The doctor (fem) studied the patient (masc) because he offered information and
was willing (masc) (to do anything).
*24d. La doctora estudio al enfermo por ofrecer datos y estar dispuesta a todo.
*24d. ‘The doctor (fem) studied the patient (masc) for offering information and being
*willing (fem)* (to do anything).
*24d. The doctor (fem) studied the patient (masc) because he/she offered information and
was *willing (fem)* (to do anything).
25a. El agente alabo a la acusada para evitar rumores y ser honesto ante todos
25a. The agent (masc) praised the woman under arrest (in order) to avoid rumors and
be honest with everyone.
*25b. El agente alabo a la acusada para evitar rumores y ser honesta ante todos
*25b. The agent (masc) praised the woman under arrest (in order) to avoid rumors and
be *honest (fem)* with everyone.
25c. El agente alabo a la acusada por evitar rumores y ser honesta ante todos.
25c. The agent (masc) praised the woman under arrest for avoiding rumors and being
honest (fem) with everyone.
*25d. El agente alabo a la acusada por evitar rumores y ser honesto ante todos.
*25d. The agent (masc) praised the woman under arrest for avoiding rumors and being
*honest (masc)* with everyone.
26a. Juan consulto a Ana para tener un criterio y estar informado de la situacion.
26a. Juan consulted Ana (in order) to have good judgment/discretion and be informed
(masc) about the situation.
*26b. Juan consulto a Ana para tener un criterio y estar informada de la situacion.
*26b. Juan consulted Ana (in order) to have good judgment/discretion and be*informed
(fem)* about the situation.
26c. Juan consulto a Ana por tener un criterio y estar informada de la situacion.
M. Betancort et al. / Cognition 100 (2006) 217–282272
26c. Juan consulted Ana for having good judgment/discretion and being informed (fem)
about the situation’.
26c. Juan consulted Ana because she had good judgment/discretion and was informed
(fem) about the situation.
*26d. Juan consulto a Ana por tener un criterio y estar informado de la situacion
*26d. Juan consulted Ana for having good judgment/discretion and being *informed
(masc)* about the situation.
*26d. Juan consulted Ana because he/she had good judgment/discretion and was
*informed (masc)* about the situation.
27a. Silvia se enfado con Andres para eludir la realidad y estar pasiva siempre.
27a. Silvia got angry with Andres (in order) to always evade reality and always be
passive (fem).
*27b. Silvia se enfado con Andres para eludir la realidad y estar pasivo siempre.
*27b. Silvia got angry with Andres (in order) to always evade reality and always be
*passive (masc)*.
27c. Silvia se enfado con Andres por eludir la realidad y estar pasivo siempre.
27c. Silvia got angry with Andres for always evading reality and always being passive
(masc).
*27d. Silvia se enfado con Andres por eludir la realidad y estar pasiva siempre.
*27d. Silvia got angry with Andres for always evading reality and always being *passive
(fem)*.
28a. Borja felicito a Silvia para evitar los rumores y estar preparado en la reunion.
28a. Borja congratulated Silvia (in order) to avoid rumors and be prepared (masc) at
the meeting.
*28b. Borja felicito a Silvia para evitar los rumores y estar preparada en la reunion.
*28b. Borja congratulated Silvia (in order) to avoid rumors and be *prepared (fem)* at
the meeting.
28c. Borja felicito a Silvia por evitar los rumores y estar preparada en la reunion.
28c. Borja congratulated Silvia for avoiding rumors and being prepared (fem) at the
meeting.
*28d. Borja felicito a Silvia por evitar los rumores y estar preparado en la reunion.
*28d. Borja congratulated Silvia for avoiding rumors and being *prepared (masc)* at
the meeting.
29a. Ivan defendio a Nuria para dejarse involucrar e ir voluntario a Kosovo.
29a. Ivan defended Nuria (in order) to get involved and become a volunteer (masc) in
Kosovo.
*29b. Ivan defendio a Nuria para dejarse involucrar e ir voluntaria a Kosovo.
M. Betancort et al. / Cognition 100 (2006) 217–282 273
*29b. Ivan defended Nuria (in order) to get involved and become a *volunteer (fem)* in
Kosovo.
29c. Ivan defendio a Nuria por dejarse involucrar e ir voluntaria a Kosovo.
29c. Ivan defended Nuria for getting involved and becoming a volunteer (fem) in Kosovo.
*29d. Ivan defendio a Nuria por dejarse involucrar e ir voluntario a Kosovo.
*29d. Ivan defended Nuria for getting involved and becoming a *volunteer (masc)* in
Kosovo.
30a. Eva denuncio a Marcos para revelar datos y sentirse protegida por el director.
30a. Eva condemned Marcos (in order) to reveal information and be under (fem) the
director’s protection.
*30b. Eva denuncio a Marcos para revelar datos y sentirse protegido por el director.
*30b. Eva condemned Marcos (in order) to reveal information and be *under (masc)*
the director’s protection (protected (*masc) by the director).
30c. Eva denuncio a Marcos por revelar datos y sentirse protegido por el director.
30c. Eva condemned Marcos for revealing information and (his) being under the
director’s protection.
*30d. Eva denuncio a Marcos por revelar datos y sentirse protegida por el director.
*30d. Eva condemned Marcos for revealing information and *(her)* being under the
director’s protection (protected (*fem) by the director).
31a. La doctora acuso al cirujano para eludir el caso y fingirse cansada ese dıa.
31a. The doctor (fem) accused the surgeon (masc) (in order) to evade the case and fake
being fatigued (fem) that day.
*31b. La doctora acuso al cirujano para eludir el caso y fingirse cansado ese dıa.
*31b. The doctor (fem) accused the surgeon(masc) (in order) to evade the case and fake
being *fatigued (masc)* that day.
31c. La doctora acuso al cirujano por eludir el caso y fingirse cansado ese dıa.
31c. The doctor (fem) accused the surgeon (masc) of evading the case and faking being
fatigued (masc) that day.
*31d. La doctora acuso al cirujano por eludir el caso y fingirse cansada ese dıa.
*31d. The doctor (fem) accused the surgeon (masc) of evading the case and faking being
*fatigued (fem)* that day.
32a. Antonio discutio con Julia para armar una escena y ser gracioso en la fiesta.
32a. Antonio argued with Julia (in order) to cause a scene and be funny (masc) at the
party.
*32b. Antonio discutio con Julia para armar una escena y ser graciosa en la fiesta
*32b. Antonio argued with Julia (in order) to cause a scene and be *funny (fem)* at the
party.
M. Betancort et al. / Cognition 100 (2006) 217–282274
32c. Antonio discutio con Julia por armar una escena y ser graciosa en la fiesta.
32c. Antonio argued with Julia for causing a scene and being funny (fem) at the party.
*32d. Antonio discutio con Julia por armar una escena y ser gracioso en la fiesta.
*32d. Antonio argued with Julia for causing a scene and being *funny (masc)* at the
party.
33a. Ines contrato al asesor para ser interesante y estar bien documentada del tema.
33a. Ines hired the consultant (masc) (in order) to be interesting and well versed (fem)
in the subject.
*33b. Ines contrato al asesor para ser interesante y estar bien documentado del tema.
*33b. Ines hired the consultant (masc) (in order) to be interesting and *well versed
(masc)* in the subject.
33c. Ines contrato al asesor por ser interesante y estar bien documentado del tema.
33c. Ines hired the consultant (masc) for being interesting and well versed (masc) in the
subject.
33c. Ines hired the consultant (masc) because he was interesting and well versed in the
subject.
*33d. Ines contrato al asesor por ser interesante y estar bien documentada del tema.
*33d. Ines hired the consultant (masc) for being interesting and *well versed (fem)*in the
subject.
*33d. Ines hired the consultant (masc) because he was interesting and *well versed (fem)
* in the subject.
34a. Angeles defendio a Martın para mantener una postura y parecer ıntegra en todo.
34a. Angeles defended Martın (in order) to maintain a stance and seem upright (fem) in
everything.
*34b. Angeles defendio a Martın para mantener una postura y parecer ıntegro en todo.
*34b. Angeles defended Martın (in order) to maintain a stance and seem *upright (masc)
* in everything.
34c. Angeles defendio a Martın por mantener una postura y parecer ıntegro en todo.
34c. Angeles defended Martın for maintaining a stance and seeming upright (masc) in
everything.
*34d. Angeles defendio a Martın por mantener una postura y parecer ıntegra en todo.
*34d. Angeles defended Martın for maintaining a stance and seeming *upright (fem)* in
everything.
35a. Monica insulto a Jose para buscar una disculpa y largarse sola del hotel
35a. Monica insulted Jose (in order) to look for an excuse and leave the hotel alone
(fem).
*35b. Monica insulto a Jose para buscar una disculpa y largarse solo del hotel
M. Betancort et al. / Cognition 100 (2006) 217–282 275
*35b. Monica insulted Jose (in order) to look for an excuse and leave the hotel *alone
(masc)*.
35c. Monica insulto a Jose por buscar una disculpa y largarse solo del hotel.
35c. Monica insulted Jose for looking for an excuse and leaving the hotel alone (masc).
*35d. Monica insulto a Jose por buscar una disculpa y largarse sola del hotel
*35d. Monica insulted Jose for looking for an excuse and leaving the hotel *alone (fem)*.
36a. Ines discutio con Juan para olvidar el trabajo y hacerse la loca en la oficina.
36a. Ines argued with Juan (in order) to blow off work and act crazy (fem) at the office.
*36b. Ines discutio con Juan para olvidar el trabajo y hacerse el loco en la oficina
*36b. Ines argued with Juan (in order) to blow off work and act *crazy (masc)* at the
office.
36c. Ines discutio con Juan por olvidar el trabajo y hacerse el loco en la oficina.
36c. Ines argued with Juan for blowing off work and acting crazy (masc) at the office.
*36d. Ines discutio con Juan por olvidar el trabajo y hacerse la loca en la oficina.
*36d. Ines argued with Juan for blowing off work and acting*crazy (fem)* at the office.
37a. Juan dejo a Sandra para cambiar de vida y hacerse reportero de repente.
37a. Juan left Sandra (in order) to turn over a new leaf and suddenly become a reporter
(masc).
*37b. Juan dejo a Sandra para cambiar de vida y hacerse reportera de repente.
*37b. Juan left Sandra (in order) to turn over a new leaf and suddenly become a
*reporter (fem)*.
37c. Juan dejo a Sandra por cambiar de vida y hacerse reportera de repente.
37c. Juan left Sandra for turning over a new leaf and suddenly becoming a reporter
(fem).
37c. Juan left Sandra because she turned over a new leaf and suddenly became a
reporter.
*37d. Juan dejo a Sandra por cambiar de vida y hacerse reportero de repente.
*37d. Juan left Sandra for turning over a new leaf and suddenly becoming a *reporter
(masc)*.
*37d. Juan left Sandra because he/she turned over a new leaf and suddenly became a
*reporter (masc)*.
38a. Yolanda se caso con Jorge para tener dinero y ser heredera de una fortuna.
38a. Yolanda married Jorge (in order) to have money and be an heiress to a fortune.
*38b. Yolanda se caso con Jorge para tener dinero y ser heredero de una fortuna.
*38b. Yolanda married Jorge (in order) to have money and be an *heir* to a fortune.
38c. Yolanda se caso con Jorge por tener dinero y ser heredero de una fortuna.
38c. Yolanda married Jorge for having money and being an heir to a fortune.
M. Betancort et al. / Cognition 100 (2006) 217–282276
38c. Yolanda married Jorge because he had money and was an heir to a fortune.
*38d. Yolanda se caso con Jorge por tener dinero y ser heredera de una fortuna.
*38d. Yolanda married Jorge for having money and being an *heiress’ to a fortune.
*38d. Yolanda married Jorge because he/she had money and was an *heiress* to a
fortune.
39a. Marisa critico a Julio para darse fama y sentirse salvadora de la patria.
39a. Marisa criticized Julio (in order) to become famous and thinking herself the
country’s savior.
*39b. Marisa critico a Julio para darse fama y sentirse salvador de la patria.
*39b. Marisa criticized Julio (in order) to become famous and think *himself* the
country’s savior.
39c. Marisa critico a Julio por darse fama y sentirse salvador de la patria.
39c. Marisa criticized Julio for becoming famous and thinking himself the country’s
savior.
*39d. Marisa critico a Julio por darse fama y sentirse salvadora de la patria.
*39d. Marisa criticized Julio for becoming famous and thinking *herself* the country’s
savior.
40a. Juan denuncio a Teresa para encubrir la trama y convertirse en jefe del negocio.
40a. Juan condemned Teresa (in order) to cover up the conspiracy and become the boss
(masc) of the business.
*40b. Juan denuncio a Teresa para encubrir la trama y convertirse en jefa del negocio.
*40b. Juan condemned Teresa (in order) to cover up the conspiracy and become the
*boss (fem)* of the business.
40c. Juan denuncio a Teresa por encubrir la trama y convertirse en jefa del negocio.
40c. Juan condemned Teresa for covering up the conspiracy and becoming the boss
(fem) of the business.
*40d. Juan denuncio a Teresa por encubrir la trama y convertirse en jefe del negocio.
*40d. Juan condemned Teresa for covering up the conspiracy and becoming the boss
(masc)* of the business.
41a. Marıa se burlo de Juan para llamar la atencion y presumir de seductora en clase.
41a. Marıa made fun of Juan (in order) to attract attention in class and boast about
being a seductress.
*41b. Marıa se burlo de Juan para llamar la atencion y presumir de seductor en clase.
*41b. Marıa made fun of Juan (in order) to attract attention in class and boast about
being a *seducer*.
41c. Marıa se burlo de Juan por llamar la atencion y presumir de seductor en clase.
M. Betancort et al. / Cognition 100 (2006) 217–282 277
41c. Marıa made fun of Juan for attracting attention in class and boasting about being a
seducer.
*41d. Marıa se burlo de Juan por llamar la atencion y presumir de seductora en clase.
*41d. Marıa made fun of Juan for attracting attention in class and boasting about being a
*seductress*.
42a. Juan se enfado con Sara para falsear el informe y fingirse atacado por el grupo.
42a. Juan got angry with Sara (in order) to falsify the report and fake being attacked
(masc) by the group.
*42b. Juan se enfado con Sara para falsear el informe y fingirse atacada por el grupo.
*42b. Juan got angry with Sara (in order) to falsify the report and fake being *attacked
(fem)* by the group.
42c. Juan se enfado con Sara por falsear el informe y fingirse atacada por el grupo.
42c. Juan got angry with Sara for falsifying the report and faking being attacked (fem)
by the group.
*42d. Juan se enfado con Sara por falsear el informe y fingirse atacado por el grupo.
*42d. Juan got angry with Sara for falsifying the report and faking being *attacked
(masc)* by the group.
43a. Juan elogio a Susana para solucionar el problema y quedar como amigo ante todos.
43a. Juan praised Susana (in order) to solve the problem and remain everyone’s friend
(masc).
*43b. Juan elogio a Susana para solucionar el problema y quedar como amiga ante todos.
*43b. Juan praised Susana (in order) to solve the problem and remain everyone’s *friend
(fem)*.
43c. Juan elogio a Susana por solucionar el problema y quedar como amiga ante todos.
43c. Juan praised Susana for solving the problem and remaining everyone’s friend
(fem).
*43d. Juan elogio a Susana por solucionar el problema y quedar como amigo ante todos.
*43d. Juan praised Susana for solving the problem and remaining everyone’s *friend
(masc)*.
44a. Isabel se burlo de Jose para evitar el ridıculo y parecer tonta ante el publico.
44a. Isabel made fun of Jose (in order) to avoid making a fool of herself and to look silly
(fem) in public.
*44b. Isabel se burlo de Jose para evitar el ridıculo y parecer tonto ante el publico.
*44b. Isabel made fun of Jose (in order) to avoid making a fool of herself and to look
*silly (masc)* in public.
44c. Isabel se burlo de Jose por evitar el ridıculo y parecer tonto ante el publico.
M. Betancort et al. / Cognition 100 (2006) 217–282278
44c. Isabel made fun of Jose for avoiding making a fool of himself and looking silly
(masc) in public.
*44d. Isabel se burlo de Jose por evitar el ridıculo y parecer tonta ante el publico.
*44d. Isabel made fun of Jose for avoiding making a fool of himself and looking *silly
(fem)* in public.
45a. Juan acuso a Marıa para correr el rumor y convertirse en dueno de la situacion.
45a. Juan accused Marıa (in order) to spread the rumor and take control/command
(his) of the situation.
*45b. Juan acuso a Marıa para correr el rumor y convertirse en duena de la situacion.
*45b. Juan accused Marıa (in order) to spread the rumor and take *control/command
(fem)* of the situation.
45c. Juan acuso a Marıa por correr el rumor y convertirse en duena de la situacion.
45c. Juan accused Marıa of spreading the rumor and (her) taking control/command of
the situation.
*45d. Juan acuso a Marıa por correr el rumor y convertirse en dueno de la situacion.
*45d. Juan accused Marıa of spreading the rumor and *(his)* taking control/command
of the situation.
46a. Pedro se molesto con Lucıa para faltar al protocolo y ser grosero en la cena.
46a. Pedro got annoyed with Lucıa (in order) to break protocol and be rude (masc) at
dinner.
*46b. Pedro se molesto con Lucıa para faltar al protocolo y ser grosera en la cena
*46b. Pedro got annoyed with Lucıa (in order) to break protocol and be *rude (fem)* at
dinner.
46c. Pedro se molesto con Lucıa por faltar al protocolo y ser grosera en la cena.
46c. Pedro got annoyed with Lucıa for breaking protocol and being rude (fem) at
dinner.
*46d. Pedro se molesto con Lucıa por faltar al protocolo y ser grosero en la cena.
*46d. Pedro got annoyed with Lucıa for breaking protocol and being *rude (masc)* at
dinner.
47a. Belen abandono a Luıs para vivir una aventura y hacerse bohemia por un tiempo.
47a. Belen abandoned Luıs (in order) to live out an adventure and become a bohemian
(fem) for a while.
*47b. Belen abandono a Luıs para vivir una aventura y hacerse bohemio por un tiempo.
*47b. Belen abandoned Luıs (in order) to live out an adventure and become a *bohemian
(masc)* for a while.
47c. Belen abandono a Luıs por vivir una aventura y hacerse bohemio por un tiempo.
M. Betancort et al. / Cognition 100 (2006) 217–282 279
47c. Belen abandoned Luıs for living out an adventure and becoming a bohemian
(masc) for a while.
47c. Belen abandoned Luıs because he lived out an adventure and became a bohemian
for a while.
*47d. Belen abandono a Luıs por vivir una aventura y hacerse bohemia por un tiempo.
*47d. Belen abandoned Luıs for living out an adventure and becoming a *bohemian
(fem)* for a while.
*47d. Belen abandoned Luıs because she/he lived out an adventure and became a
*bohemian (fem)* for a while.
48a. Esteban apoyo a Rosa para defender la causa y hacerse voluntario de Caritas.
48a. Esteban supported/backed Rosa (in order) to defend the cause and become a
Caritas volunteer (masc).
*48b. Esteban apoyo a Rosa para defender la causa y hacerse voluntaria de Caritas.
*48b. Esteban supported/backed Rosa (in order) to defend the cause and become a
Caritas *volunteer (fem)*.
48c. Esteban apoyo a Rosa por defender la causa y hacerse voluntaria de Caritas.
48c. Esteban supported/backed Rosa for defending the cause and becoming a Caritas
volunteer (fem).
*48d. Esteban apoyo a Rosa por defender la causa y hacerse voluntario de Caritas.
*48d. Esteban supported/backed Rosa for defending the cause and becoming a Caritas
*volunteer (masc)*.
References
Betancort, M., Meseguer, E., & Carreiras, M. (2004). The empty category PRO: Processing what can’t be seen. In
M. Carreiras, & C. Clifton (Eds.), The on-line study of sentence comprehension: Eyetracking, ERP, and
beyond. Brighton: Psychology Press.
Bever, T. G., & McElree, B. (1988). Empty categories access their antecedent during comprehension. Linguistic
Inquiry, 19, 35–43.
Bever, T. G., & Sanz, M. (1997). Empty categories access their antecedents during comprehension of Spanish
unaccusatives. Linguistic Inquiry, 28, 69–91.
Boland, J., & Tanenhaus, M. K. (1991). The role of lexical representation in sentence processing. In G. B.
Simpson (Ed.), Understanding word and sentence. Amsterdam/Elsevier: North Holland.
Boland, J. E., Tanenhaus, M. K., & Garnsey, S. M. (1990). Evidence for the immediate use of verb control
information in sentence processing. Journal of Memory and Language, 29, 413–432.
Bresnan, J. (1982). The theory of complementation in English syntax Doctoral dissertation. Cambridge, MA:
MIT.
Brysbaert, M., & Mitchell, D. C. (1996). Modifier attachment in sentence parsing: Evidence from Dutch.
Quarterly Journal of Experimental Psychology, Section A—Human Experimental Psychology, 49(3), 664–
695.
Cattell, R. (1984). Composite predicates in English. New York, NY: Academic Press.
M. Betancort et al. / Cognition 100 (2006) 217–282280
Chierchia, G. (1988). Structured meanings, thematic roles, and control. In G. Chierchia, B. Partee, & R. Turner
(Eds.), Properties, types, and meaning (Vol. 2) (pp. 131–136). Dordrecht: Kluwer.
Chomsky, N. (1981). Lectures on government and binding. Dordrecht: Foris.
Chomsky, N. (1986). Knowledge of language: Its nature, origin, and use. New York, NY: Praeger.
Clifton, C., & Frazier, L. (1986). The use of syntactic information in filling gaps. Journal of Psycholinguistic
research, 15, 209–224.
Clifton, C., & Frazier, L. (1989). Comprehending sentence with long distance dependencies. In G. Carlson, & M.
Tanenhaus (Eds.), Linguistics structure in language processing (pp. 273–317). Dordrech, The Netherlands:
Klewer, 273–317.
Cloitre, M., & Bever, T. G. (1988). Linguistic anaphors, levels of representation, and discourse. Language and
Cognitive Processes, 3, 293–322.
Coulson, S., King, J., & Kutas, M. (1998). Expect the unexpected: Event-related brain response to
morphosyntactic violations. Language and Cognitive Processes, 13, 21–58.
Cowart, W., & Cairns, H. S. (1987). Evidence for an anaphoric mechanism within syntactic processing: Some
reference relations defy semantic and pragmatic constraints. Memory & Cognition, 15, 318–331.
Crain, S., & Fodor, J. D. (1985). How can grammars help parsers?. In D. R. Dowty, L. Karttunen, & A. M. Zwicky
(Eds.), Natural language parsing. New York, NY: Cambridge University Press.
Culicover, P. W., & Jackendoff, R. (2001). Control is not movement. Linguistic Inquiry, 32/3, 493–511.
Demestre, J., Meltzer, S., Garcıa Albea, J. E., & Vigil, A. (1999). Identifying the null subject: Evidence from
event-related brain potential. Journal of Psycholinguistic Research, 28, 293–312.
De Vincenci, M. (1998). Syntactically-based parsing strategies: evidence from typologically different languages.
In D. Hillert, Syntax and semantics (Vol. 31) (pp. 337–344). New York, NY: Academic Press.
Duffy, S. A., Morris, R. K., & Rayner, K. (1988). Lexical ambiguity and fixation times in reading. Journal of
Memory and Language, 27, 429–446.
Featherston, S., Gross, M., Munte, T. F., & Clahsen, H. (2000). Brain potentials in the processing of complex
sentences: An ERP study of control and raising constructions. Journal of Psycholinguistic Research, 29, 141–
154.
Fernandez Lagunilla, M., & Anula Rebollo, A. (1995). Sintaxis y cognicion Introduccion al conocimiento, el
procesamiento y los deficits sintacticos. Madrid: Sıntesis.
Ferreira, F., & Clifton, C. (1986). The independence of syntactic processing. Journal of Memory and Language,
25, 348–368.
Ferreira, F., & Henderson, J. (1990). The use of verb information in syntactic parsing: Evidence from eye-
movements and word-by-word self-paced reading. Journal of Experimental Psychology: Learning, Memory,
and Cognition, 16, 555–568.
Fiebach, C. J., Schlesewsky, M., & Friederici, A. (2001). Syntactic working memory and the establishment of
filler-gap dependencies: Insights of ERPs and fMRI. Journal of Psycholinguistics Research , 30321–30338.
Fiebach, C. J., Schlesewsky, M., & Friederici, A. (2002). Separating syntactic memory costs and syntactic
integration costs during parsing: The processing of German WH-questions. Journal of Memory and
Language, 47, 250–272.
Fodor, J. A. (1983). Modularity of mind. Cambridge, MA: MIT Press.
Fodor, J. D. (1978). Parsing strategies and constraints in transformations. Linguistic Inquiry, 9, 427–474.
Fodor, J. D. (1988). On modularity in syntactic processing. Journal of Psycholinguistic Research, 17, 125–163.
Fodor, J. D. (1989). Empty categories in sentence processing. In G. Altman (Ed.), Special issue on language and
congnitive processes. Parsing and interpretation (Vol. 4(3/4)), SI155–SI209.
Ford, M., & Dalrymple, M. (1988). A note on some psychological evidence and alternative grammars. Cognition,
29, 63–71.
Frazier, L. (1987). Sentence processing: A tutorial review. In M. Coltheart (Ed.), Attention and performance (pp.
559–586). Hillsdale, NJ: Lawrence Erlbaum Associates.
Frazier, L. (1993). Processing Dutch sentence structures. Journal of Psycholinguistic Research, 22, 85–108.
Frazier, L. (1998). Getting there (slowly). Journal of Psycholinguistic Research, 27, 123–126.
Frazier, L., & Clifton, C. (1989). Sucessive cyclicity in the grammar and parser. Language and Cognitive
Processes, 4, 93–126.
Frazier, L., & Clifton, C. (1996). Construal. Cambridge, MA: MIT Press.
M. Betancort et al. / Cognition 100 (2006) 217–282 281
Frazier, L., Clifton, C., & Randall, J. (1983). Filling gaps: decision principles and structure in sentence
comprehension. Cognition, 13, 187–222.
Frazier, L., & Flores D’Arcais, G. (1989). Filler-driven parsing: A study of gap filling in Dutch. Journal of
Memory and Language, 28, 331–344.
Friederici, A. D., Pfeifer, E., & Hahne, A. (1993). Event-related brain potentials during natural speech processing:
Effects of semantic, morphological, and syntactic violations. Cognitive Brain Research, 1, 183–192.
Garcıa-Albea, J. E., & Meltzer, S. (1996). Processing empty categories in Spanish. In M. Carreiras, J. E. Garcıa-
Albea, & N. Sebastian-Galles (Eds.), Language processing in Spanish. Nahwah: Lawrence Erlbaum.
Garnsey, S. M., Pearlmutter, N. J., Myers, E., & Lotocky, M. A. (1997). The contributions of verb bias and
plausibility to the comprehension of temporarily ambiguous sentences. Journal of Memory and Language, 37,
58–93.
Garnsey, S. M., Tanenhaus, M. K., & Chapman, R. M. (1989). Evoked potential and the study of comprehension.
Journal of Psycholinguistic Research, 18, 51–60.
Haegemann, L. (1991). Introduction to government and binding theory. Blackwell: Oxford.
Haegeman, L. (1994). Introduction to government and binding theory. Oxford: Blackwell.
Haggort, P., Brown, C., & Groothusen, J. (1993). The syntactic positive shift as an ERP measure of syntactic
processing. Language and Cognitive Processes, 8, 439–483.
Hare, M., McRae, K., & Elman, J. (2004). Admitting that admitting verb sense into corpus analysis makes sense.
Language and Cognitive Processes, 19(2), 181–224.
Hornstein, N. (1991). A primer on PRO (and other empty categories). Journal of Psycholinguistic Research, 20,
187–195.
Hornstein, N. (1999). Movement and control. Linguistic Inquiry, 30, 69–96.
Huddleston, R., & Pullum, G. K. (2002). The Cambridge grammar of the English language. Cambridge, MA:
Cambridge University Press.
Jackendoff, R. (1972). Semantic interpretation in generative grammar. Cambridge, MA: MIT Press.
Jackendoff, R. (1974). A deep structure projection rule. Linguistic Inquiry, 5, 481–506.
Konieczny, L., Hemforth, B., Scheepers, C., & Strube, G. (1997). The role of lexical heads in parsing: Evidence
from German. Language and Cognitive Processes, 12, 307–348.
Liversedge, S. P. (1994). Referential context, relative clauses, and syntactic parsing. Unpublished doctoral
dissertation. University of Dundee.
MacDonald, M. C. (1989). Priming effects from gaps to antecedents. Language and Cognitive Processes, 4, 35–
36.
MacDonald, M. C., Pearlmutter, N., & Seidenberg, M. S. (1994). The lexical nature of syntactic ambiguity
resolution. Psychological Review, 101, 676–703.
Manzini, R. (1983). On control and control theory. Linguistic Inquiry, 14, 421–446.
Marcos-Marın, F., Satorre, F. J., & Viejo, L. (1998). Gramatica espanola. Madrid: Sıntesis.
Mauner, G., & Koenig, J. P. (2000). Lexical encoding of event participant information. Journal of Memory and
Language, 43, 110–134.
Manuer, G., Melinger, A., Koenig, J. P., & Bienvenue, B. (2002). When is schematic participant information
encoded? Evidence from eye-monitoring. Journal of Memory and Language, 47, 386–406.
Mauner, G., Tanenhaus, M., & Carlson, G. (1995). Implicit arguments in sentence processing. Journal of Memory
and Language, 34, 357–382.
McConkie, G. W., & Zola, D. (1987). Visual attention during eye fixations while reading. In M. Coltheart,
Attention and performance (Vol. 12) (pp. 385–401). London: Erlbaum, 385–401.
McElree, B., & Bever, T. (1989). The psychological reality of linguistically defined gaps. Journal of
Psycholinguistic Research, 18, 21–35.
Nicol, J. (1988). Co-references in parsing. Unpublished doctoral dissertation. MIT.
Nicol, J., & Swinney, D. (1989). The role of structure in coreference assignment during sentence comprehension.
Journal of Psycholinguistic Research, 18, 5–19.
Osterhout, L. (1997). On the brain responses to syntactic anomalies: Manipulations of word position and word
class reveal individual differences. Brain and Language, 59, 494–522.
Osterhout, L., & Swinney, D. A. (1993). On the temporal course of gap-filling during comprehension of verbal
passives. Journal of Psycholinguistics Research, 22, 273–286.
M. Betancort et al. / Cognition 100 (2006) 217–282282
Rayner, K., Well, A. D., & Pollatsek, A. (1982). The availability of useful information to the right of fixation in
reading. Perception & Psychophysics, 31, 537–550.
Rosenbaum, P. (1967). The grammar of English predicate complement constructions. Cambridge, MA: MIT
Press.
Sag, I., & Fodor, J. (1995). Extraction without traces. Proceedings of the 13th west coast conference on formal
linguistics (pp. 365–384).
Sag, I., & Pollard, C. (1991). An integrated theory of complement control. Language, 67, 63–113.
Schmauder, A. R. (1991). Argument structure frames: A lexical complexity metric? Journal of Experimental
Psychology: Learning: Memory, and Cognition, 17, 49–65.
Sebastian, N., Martı, M. A., Carreiras, M., & Cuetos, F. (2000). LEXESP: Una base de datos informatizada del
espanol. Barcelona: Servicio de Publicaciones de la Universitat de Barcelona.
Shapiro, L., & Nagel, N. (1993). Preference for a verb’s complements and their use in sentence processing.
Journal of Memory and Language, 32, 96–114.
Stowe, L. (1984). Models of gap location in the human language processor. Bloomington, IN: Indiana University
Linguistics Club.
Swinney, D. A., Ford, U., Frauenfelder, U., & Bresnan, J. (1988). On the temporal course of gap-filling and
antecedent-assignment during sentence comprehension. In B. Grosz, R. Kaplan, M. Machen, & I. Sag (Eds.),
Language structure and processing. Standford, CA: CSLI.
Tanenhaus, M. K., Boland, J., Garsney, S., & Carlson, G. (1989). Lexical structure in parsing long-distance
dependencies. Journal of Psycholinguistic Research, 18, 37–50.
Tanenhaus, M. K., & Trueswell, J. C. (1995). Sentence comprehension. In J. Miller, & P. Eimas, Handbook of
perception and cognition: Speech, language, and communication (2nd ed.) (Vol. 11) (pp. 217–262). San
Diego, CA: Academic Press, 217–262.
Traxler, M., Pickering, M., & Clifton, C. (1998). Adjunt attachment is not a form of lexical ambiguity resolution.
Journal of Memory and Language, 39, 558–592.
Trueswell, J. C., Tanenhaus, M. K., & Garnsey, S. M. (1994). Semantic influences on parsing: Use of thematic
role information in syntactic disambiguation. Journal of Memory and Language, 33, 285–318.
Underwood, N. R., & McConkie, G. W. (1985). Perceptual span for letter distinction during reading. Reading
Research Quarterly, 20, 153–162.