You are on page 1of 28

Second Language Research

http://slr.sagepub.com/

Cross-linguistic variation and the acquisition of pronominal reference in L2


Spanish
Jill Jegerski, Bill VanPatten and Gregory D Keating
Second Language Research 2011 27: 481 originally published online 6 June 2011
DOI: 10.1177/0267658311406033

The online version of this article can be found at:


http://slr.sagepub.com/content/27/4/481

Published by:

http://www.sagepublications.com

Additional services and information for Second Language Research can be found at:

Email Alerts: http://slr.sagepub.com/cgi/alerts

Subscriptions: http://slr.sagepub.com/subscriptions

Reprints: http://www.sagepub.com/journalsReprints.nav

Permissions: http://www.sagepub.com/journalsPermissions.nav

Citations: http://slr.sagepub.com/content/27/4/481.refs.html

>> Version of Record - Sep 19, 2011


OnlineFirst Version of Record - Jun 6, 2011

What is This?

Downloaded from slr.sagepub.com at ARIZONA STATE UNIV on August 18, 2012


second
language
research

Second Language Research

Cross-linguistic variation and


27(4) 481–507
© The Author(s) 2011
Reprints and permission: sagepub.
the acquisition of pronominal co.uk/journalsPermissions.nav
DOI: 10.1177/0267658311406033
reference in L2 Spanish slr.sagepub.com

Jill Jegerski
College of Staten Island, City University of New York, USA

Bill VanPatten
Michigan State University, USA

Gregory D Keating
San Diego State University, USA

Abstract
The current investigation tested two predictions regarding second language (L2) processing at
the syntax–discourse interface: (1) that L2 performance on measures of interface phenomena can
differ from that of native speakers; and (2) that cross-linguistic influence can be a source of such
divergence. Specifically, we examined the offline interpretation of ambiguous subject pronouns
with intrasentential antecedents in Spanish and English, including discourse–syntactic constraints
that are active in pro-drop Spanish and principles of discourse structure that affect pronominal
reference in English. Two participant groups of English-speaking learners of L2 Spanish – an
intermediate group and an advanced group – failed to show categorically native-like differentiation
between null and overt pronouns in Spanish. Both groups, however, did show marginal effects for
Discourse Structure (coordination or subordination of clauses), an effect that was also present
in their native English. Furthermore, there was a significant interaction with the advanced group
between Pronoun and Discourse Structure, so this group seemed to employ to a certain degree
a hybrid strategy. This outcome suggests that pre-existing referential strategy persists even at an
advanced level of L2 proficiency and may be a primary barrier to native-like performance, even
after target-like L2 principles are acquired and begin to apply.

Keywords
pronominal reference, null subjects, pronoun interpretation, sentence comprehension, first language
influence, non-native Spanish

Corresponding author:
Jill Jegerski, College of Staten Island-CUNY, 2800 Victory Boulevard, Staten Island, NY 10314, USA
Email: jill.jegerski@csi.cuny.edu

Downloaded from slr.sagepub.com at ARIZONA STATE UNIV on August 18, 2012


482 Second Language Research 27(4)

I Introduction
The Interface Hypothesis (Sorace, 2000, 2004; Sorace and Filiaci, 2006), in its simplest
form, proposes that grammatical properties that are wholly syntactic are in principle
acquirable in non-native languages, whereas those properties of the grammar that exist at
the interface between the syntax and other areas or modules of linguistic knowledge,
such as semantics or discourse, do not reach a stable state of native-likeness.1 In their
elaboration of the Interface Hypothesis, Sorace and Filiaci (2006) point out that there are
principally two possible explanations for the difficulty that interfaces present during
acquisition; either there are underlying deficiencies in the mental representation of lan-
guage, or the limitation is actually one of processing resources required to simultane-
ously integrate and apply information from different domains of linguistic knowledge. In
the latter case, cognitive capacity such as working memory would be more likely to be
insufficient for second language (L2) tasks, as language operations such as retrieval and
computation could be less efficient than with a native language. Although distinct meas-
urement of these two sources of non-target performance can be quite difficult, much
recent discussion has focused on the importance of processing limitations in second lan-
guage acquisition (SLA) (e.g. Birdsong, 2006; Clahsen and Felser, 2006; White, 2007).
A third and often-neglected potential source of non-nativeness at linguistic interfaces
is influence from the existing language system. The Shallow Structure Hypothesis (Clah-
sen and Felser, 2006) proposes that inter-linguistic influence cannot occur during L2
processing because of learners’ inability to represent abstract and hierarchical syntactic
detail. Sorace (2006), however, questions whether the complete exclusion of first lan-
guage (L1) influence from L2 processing may be premature and outlines how insuffi-
cient processing resources could in some cases foster rather than preclude L1 influence,
because zero strategy is not always the most efficient. In other words, both innate limita-
tions and existing native language principles could each affect interface behaviors. Fur-
thermore, the two cannot always be empirically distinguished. For example, Sorace and
Filiaci (2006) studied the processing of null and overt pronouns by near-native speakers
of Italian whose L1 was English. In this experiment the near-natives showed a clear pro-
cessing strategy, meaning they distinguished between null and overt pronouns, but the
strategy was not entirely native-like. The authors concluded that what these non-native
participants were lacking was the processing resources to consistently apply their native-
like syntactic and discourse knowledge of pro-drop. However, it is not clear whether the
indeterminacy that occurred with the interpretation of overt pronouns in this study was
additionally or even entirely due to L1 influence from non-pro-drop English, as the
investigation only tested the interpretation of pronominal reference in the participants’
L2. Indeed, Sorace and Filiaci explicitly leave open the possibility of some degree of
cross-linguistic influence from English, acknowledging that their experiment was not
specifically designed to explore this factor.
Thus, there is a growing interest in the acquisition of linguistic knowledge that exists
at the interfaces between syntax and other linguistic domains, which is frequently attrib-
uted to processing limitations, or a limited capability to integrate multiple sources of
knowledge during comprehension. Influence from the pre-existing language system as
an alternative or additional source of divergence has often been overlooked or

Downloaded from slr.sagepub.com at ARIZONA STATE UNIV on August 18, 2012


Jegerski et al. 483

discounted. Such influence has been proposed on theoretical grounds (Sorace, 2006), but
so far there is no clear empirical evidence to corroborate this claim. Previous research on
interface phenomena has largely focused on testing L2 behavior without direct compari-
son to L1 behavior, which can make it impossible to tease out cross-linguistic influence.
The present study of English-speaking learners of Spanish explored the interpretation of
null and overt subjects in Spanish as well as the interpretation of overt pronouns in the
participants’ native English, in order to more clearly establish interpretation strategy in
the L1 and thus to identify potential cross-linguistic influence. Based on the outcomes of
a series of four experiments, we will propose that a primary obstacle to learning can be
the ongoing inclusion of pre-existing strategy during non-native interpretation, although
some degree of L2 learning also appears to occur independently of cross-linguistic influ-
ence in this case.

II  Anaphora resolution in English


Anaphora resolution, or the mental association of real-world entities with referential
linguistic expressions such as pronouns, is a crucial component of language comprehen-
sion. Interestingly, the language user often encounters ambiguous pronouns and is forced
to choose between two or more referents in order to arrive at the accurate interpretation.
Additionally, the factors that can and do affect the resolution of such ambiguity can vary
from language to language. Thus, the interpretation of pronouns presents ideal ground
for exploring and describing the nature of comprehension in a non-native language
because it is possible to tease apart L1 strategy from L2 strategy when these two differ.
This is the case with the interpretation of ambiguous pronouns in English and Spanish.
Much of the theory and research on anaphora resolution in English has been carried
out from a discourse perspective, which fundamentally assumes that the comprehension
of written or spoken language involves some type of cognitive representation of dis-
course. Such a mental model is analogous to syntax in that where syntax is a representa-
tion of a series of lexical items and the relations and dependencies between them, dis-
course is a broader model of a series of events and entities and the relations between
these. Discourse-based theories of pronominal reference in particular seek to determine
what specific features in a discourse cue the link between a pronoun and its correspond-
ing antecedent.
One such discourse-based account was synthesized by Hudson-D’Zmura and Tanen-
haus (1998), who developed and applied existing work in centering theory (Gordon et al.,
1993) in order to explain how the processor might resolve referential ambiguity in dis-
courses like the one presented in example (1).

(1) John saw Bill walking down the street.


He said hello to him.

In this discourse, the pronoun he is ambiguous, but instinctively refers to John rather
than to Bill. According to Hudson-D’Zmura and Tanenhaus, pronouns are used to con-
tinue a previously established discourse center (essentially equivalent to what is also
commonly referred to as the discourse Topic), while nouns such as proper names are

Downloaded from slr.sagepub.com at ARIZONA STATE UNIV on August 18, 2012


484 Second Language Research 27(4)

used to shift the center. The center usually corresponds to the syntactic subject, so in the
example given in (1) the pronoun he refers to John because it is the established discourse
center, the subject of the previous sentence.
These predictions for antecedent assignment were tested empirically by Hudson-
D’Zmura and Tanenhaus (1998) in a series of four online experiments, the last of which
is most relevant to the present study. In Experiment 4 of the investigation, native English
speakers judged the coherence of multi-sentence discourses with ambiguous pronouns.
In this computerized experimental task, participants first read two sentences that estab-
lished a basic discourse context, as in example (2), then read a target sentence appearing
in one of the four conditions illustrated in (2a)–(2d).

(2) Jack helped Dick wash the car. context sentences


He washed the windows as Dick waxed the car.
   a.   He soaped a pane. pronoun/subject
   b.   He buffed the hood. pronoun/object
   c.   Jack soaped a pane. noun/subject
   d.   Dick buffed the hood. noun/object

Lastly, after reading both the context and the target, participants were prompted with a
question mark on the screen and asked to judge whether the final sentence was a logical
continuation of the first two. The target sentence variants used either a pronoun or a proper
noun to refer to either the subject or the object of the first sentence of the discourse context.
The target condition in which a pronoun referred to a subject was judged as more coherent
(78% vs. 48%) and was read slightly faster on average (2,459 ms vs. 2,519 ms) than when
a pronoun referred to an object. Thus, it appears that in this type of short discourse in Eng-
lish, anaphora are most often linked to subject antecedents. The authors interpret this obser-
vation as empirical support for discourse center theory, concluding that ‘local discourse
structure plays an important role in guiding initial pronoun assignment’ (Hudson-D’Zmura
and Tanenhaus, 1998: 222). Thus, their study served to corroborate empirically the
fundamental assumption that English subject pronouns most often refer to the most dis-
course-prominent antecedent. In addition to this basic phenomenon, additional discourse
constraints may further affect the referential function of subject pronouns in English. Of
interest for the present study, which focuses on intrasentential anaphora in complex sen-
tences, is to what degree the discourse relation between two clauses or sentences affects
pronominal reference. In other words, it remains to be seen whether the reference assign-
ment strategy observed by Hudson-D’Zmura and Tanenhaus applies equally to pronouns
appearing in discourse contexts different from those tested in their original investigation.
Segmented Discourse Representation Theory (SDRT; Asher, 1993; Asher and Vieu,
2005) is a hierarchical model of discourse structure – or the relations between a series of
sentential or clausal constituents – that proposes that the hierarchy in discourse arises from
binary relations of coordination and subordination (note that these purportedly occur inde-
pendently of the syntactic relations of the same names). SDRT defines these two types of
connection in terms of the discourse function of a constituent with respect to its predeces-
sors, with Narration and Elaboration being two prototypical functions. In Narration two or
more constituents combine in series to provide an account of connected events, whereas

Downloaded from slr.sagepub.com at ARIZONA STATE UNIV on August 18, 2012


Jegerski et al. 485

(3a)

Elaboration
Elaboration

(3b) (3e)
Narration

Elaboration Elaboration

(3c) (3d)
Narration

Figure 1  Graphic representation of the SDRT discourse structure of the sentences in (3)
Source: adapted from Asher and Vieu (2005)

Elaboration occurs when a constituent provides additional or supporting detail regarding


the ideas in a previous constituent. The authors use the examples in (3) to illustrate.

(3) a.   John had a great evening last night.


b.   He had a great meal. elaboration of (a)
c.   He ate salmon. elaboration of (b)
d.   He devoured lots of cheese. narration of (c); elaboration of (b)
e.   He then won a dancing competition. narration of (b); elaboration of (a)
f. # It was a beautiful pink. incoherent continuation

If a constituent is dependent on its predecessor – as is the case with (3b), which is an


Elaboration of (3a) – the second item takes a lower position in the hierarchy, entering
into a subordinate relation with the first. Additional discourse functions that are classi-
fied as subordination include Instance, Topic, Explanation, Precondition, and Commen-
tary. On the other hand, if the second sentence or clause is a Narration of the first, as is
the case of (3d) with respect to (3c), the two are at the same level in the hierarchy and
joined through a relation of coordination. In addition to Narration, other discourse func-
tions that lead to coordination include Parallel, Background, Result, Continuation, and
Contrast. Thus, in the SDRT model each constituent of a discourse connects to one or
more previous constituents, to which it is either parallel or inferior. A graphic representa-
tion of the hierarchical discourse presented in (3) is given in Figure 1. According to
Asher and Vieu, this crucial distinction between coordination and subordination affects
the way new information can be incorporated into a discourse and thereby influences
processes like reference assignment.
Asher (1993) provides a more in-depth analysis of the interaction between anaphora
resolution and discourse structure, emphasizing that these are two independent phenomena

Downloaded from slr.sagepub.com at ARIZONA STATE UNIV on August 18, 2012


486 Second Language Research 27(4)

and that the latter restricts the former. In SDRT the assignment of referents is restricted by
the structure of the discourse in which referential expressions and their potential anteced-
ents occur. Asher dubs this constraint ‘accessibility’ and defines it as follows:

The potential referents of an anaphor are found in the discourse constituents to its left and in
those to which it is inferior.

Thus, (3f) would not be a coherent continuation of the discourse presented in (3a)–(3e)
because salmon from (3c) is no longer available as a referent for it. Although SDRT has
not been developed to address the specific case of intrasentential anaphora with ambigu-
ity of the antecedent that is the target of the present empirical study, it stands to reason
that the subordination/coordination distinction that forms the foundation of discourse
structure in this model may play a role in such interpretation. For this reason, the distinc-
tion was a target of the present study.

III  Anaphora resolution in pro-drop languages


Pro-drop languages such as Spanish are characterized by the potential of finite verbs to
appear with null subjects, or phonologically empty pronouns that are also referred to as
pro. As with explicit pronouns, the processing of pro entails the assignment of an ante-
cedent. An example of a null subject is given in (4).

(4) Raúl vio a Alejandro cuando (pro) bajaba de un taxi.


‘Raúl saw Alejandro when (he) was getting out of a taxi.’

Accounts of anaphora resolution in Spanish and other pro-drop languages initially


attempted to employ the discourse-based theories that had previously been articulated
for English (e.g. Lujan, 1985, 1986). However, while such models may provide
elegant explanations of pronominal reference in languages with only one type of
pronoun, they do not take into consideration the occurrence of both null and overt
pronoun variants that is made possible by the distinct syntax of pro-drop (Toribio,
2000; Zagona, 2002).
The inadequacy of English-oriented theories was noted by Carminati (2002), who
proposed and tested empirically the Position of Antecedent (PA) hypothesis. This prin-
ciple, which describes the referential tendencies of null and overt pronouns in intrasen-
tential contexts in Italian, proposes the following: in cases of intrasentential anaphora,
the null pronoun prefers an antecedent that is in the SpecIP or canonical subject position,
while the overt pronoun prefers an antecedent that is not in the SpecIP position (i.e.
object or other position). To illustrate, consider the pronouns in the second clauses of two
of Carminati’s sentences in (5).

(5) a. Marta scriveva frequentemente a Piera quando (pro) era negli Stati Uniti.
‘Marta frequently wrote to Piera when (she) was in the United States.’
b. Marta scriveva frequentemente a Piera quando lei era negli Stati Uniti.
‘Marta frequently wrote to Piera when she was in the United States.’

Downloaded from slr.sagepub.com at ARIZONA STATE UNIV on August 18, 2012


Jegerski et al. 487

These examples are ambiguous because the null pronoun in (5a), and the overt pronoun,
lei, in (5b) could grammatically refer to either of the two NP antecedents in the first clause.
However, the PA hypothesis states that the null pronoun in (5a) tends to refer to the NP in the
subject position, Marta, while the overt pronoun in (5b) prefers the antecedent in the object
position, Piera. In a series of 14 psycholinguistic experiments, Carminati used both offline
questionnaires and online self-paced reading tasks to assess how native speakers of Italian
parsed these pronouns with intrasentential referents. For instance, the participants of the first
experiment were presented with ambiguous stimuli like those in (5), each of which was
followed by the question Who was in the United States? The responses differed according to
the type of pronoun that was present in the stimulus. For the null pronoun, participants chose
the subject antecedent, Marta, 80.72% of the time, while for overt pronouns they chose the
subject antecedent only 16.67% of the time. This result is consistent with the PA hypothesis,
but could equally be interpreted in terms of discourse-based theories because the subject in
such canonical constructions holds both the syntactic position of SpecIP and the discourse
role of prominent element or Topic. For this reason, Carminati also conducted additional
tests that specifically tested the purported referential link to the SpecIP position indepen-
dently of discourse prominence. This was accomplished with experimental stimuli that con-
tained non-canonical constructions like dative subjects in psychological verb constructions
and expletive subjects, where the SpecIP position is not occupied by the discourse Topic.
Overall, the PA hypothesis was corroborated by these experiments as well. Thus, it does
appear that the PA tendency is motivated at least in part by syntactic structure, meaning
occupation of the SpecIP position. Crucially, the outcome of Carminati’s investigation high-
lights the differential usage of null and overt pronouns and the importance of accounting for
this distinction in models of pronominal reference in null subject languages.
Alonso-Ovalle et al. (2002) investigated whether Carminati’s (2002) PA hypothesis
for Italian also applied to intersentential reference in Spanish. These researchers used a
series of five questionnaire-based experiments to assess the preferences of Spanish
monolinguals in selecting antecedents for null and overt pronouns. In the first experi-
ment, participants read two-sentence discourses with referential ambiguity, in a version
with either a null pronoun or an overt pronoun, as illustrated in (6).

(6) a. Juan pegó a Pedro. (pro) Está enfadado.


‘Juan hit Pedro. (he) Is angry.’
b. Juan pegó a Pedro. Él está enfadado.
‘Juan hit Pedro. He is angry.’

In both of these examples, the pronouns pro and él are ambiguous because they could
refer back to either Juan or Pedro. Participants indicated their choice of antecedent for the
pronoun by answering the question ¿Quién está enfadado? (‘Who is angry?’). Answers
varied according to the type of pronoun, with the subject antecedent being chosen 78.2%
of the time with pro and only 50.2% of the time with an overt pronoun. The remaining
four experiments investigated whether Spanish monolinguals prefer the null pronoun to
an overt one when referring to a subject, whether the PA hypothesis applies when the
antecedent is a quantifier and, finally, whether the PA hypothesis is affected by the posi-
tion of the null or overt pronoun in relation to the verb (i.e. pre-verbal or post-verbal). The

Downloaded from slr.sagepub.com at ARIZONA STATE UNIV on August 18, 2012


488 Second Language Research 27(4)

results of these experiments were quantitatively different from those obtained for Italian
in that Carminati (2002) had found that overt pronouns have a clear preference for a non-
SpecIP antecedent, whereas Alonso-Ovalle et al. found that overt pronouns have a roughly
neutral preference. Two important differences between these two studies were the lan-
guage tested, Spanish or Italian, and whether the pronominal reference in question
occurred within or between sentences. As the explicitness of a pronoun appeared to affect
its referential behavior in both studies, however, Alonso-Ovalle et al. concluded that the
PA hypothesis applies to Spanish as well as in contexts of intersentential reference. This
could be considered a broader interpretation of the PA hypothesis, which allows for only
the null pronoun to have a strong antecedent preference. The referential functions of null
and overt pronouns are not in complementary distribution, as suggested in the original PA
hypothesis, as there is a certain degree of overlap between the two, but there is still a
qualitative distinction in their antecedent tendencies.
It is also important to note that the Alonso-Ovalle et al. investigation did not replicate the
Carminati experiments that had teased apart the syntactic position of SpecIP from the dis-
course role of Topic, so the observations obtained for Spanish could perhaps be alternatively
accounted for in terms of classic discourse principles, at least if one assumes that Italian and
Spanish are fundamentally different in this regard. Still, the referential tendencies of Spanish
null and overt pronouns were shown to differ empirically, so syntactic principles must have
been at work on some level. In other words, it is possible that the PA for Spanish does not
operate on the basis of syntactic position, as Carminati (2002) has claimed it does for Italian,
but at the same time it is not possible that PA-like interpretation behavior is achieved in
complete isolation from the syntax. Therefore, for the purposes of the present study, we
assume that PA behavior – or differential interpretation of null vs. overt pronouns – is on
some level syntactic. At the same time, we assume an interface between the syntax of pro-
drop and discourse principles, but make no specific assumptions about the nature of this
interface, as previous research has been quite limited, especially with Spanish.

IV  The present study


To summarize previous theory and research on anaphora resolution in English and in null
subject languages then, there is evidence that pronoun interpretation in languages like Eng-
lish is determined by the local center of discourse. There is also reason to think that addi-
tional factors in discourse structure, such as the fundamental coordination/subordination
distinction proposed in SDRT (Asher, 1993; Asher and Vieu, 2005), may further influence
the selection of pronoun antecedents, but this prediction has yet to be tested empirically. In
null subject languages like Spanish, on the other hand, it appears that the processing of
pronouns is influenced by the syntactic distinction between null and overt pronouns.
Given the observation that there is a certain degree of cross-linguistic variation in the
interpretation of pronouns, some fundamental SLA questions follow:

xx How does the language learner handle anaphora resolution in a non-native


language?
xx Can native-like L2 strategy be acquired over time?
xx Where behavior is divergent from the native target, is there any similarity to L1
behavior?

Downloaded from slr.sagepub.com at ARIZONA STATE UNIV on August 18, 2012


Jegerski et al. 489

The present study addressed these questions in a series of four offline experiments meas-
uring the assignment of antecedents to ambiguous pronouns in intrasentential contexts.
In Experiments 1 and 2 we tested native speakers of English and then Spanish to see if
referential ambiguity is indeed affected by discourse structure in English and by the PA
hypothesis (and not by discourse structure) in Spanish. Experiment 3 examined the strat-
egy of low intermediate L2 learners for evidence of target-like L2 behavior in Spanish
and of L1 transfer from English. Finally, in Experiment 4 we tested advanced learners to
see whether they had acquired native-like L2 strategy in Spanish and whether they still
employed any principles from their native English.

V  Experiment 1
The purpose of the first experiment was to explore whether discourse structure (i.e. coor-
dination vs. subordination) influences the resolution of referential ambiguity in intrasen-
tential contexts in English.

1 Method
a Participants:  Forty-three undergraduate students, graduate students, and profession-
als recruited at two large public universities in Texas and California participated in this
experiment. Nineteen participants were male and 24 were female, and their mean age
was 26.3 years (range 19–61). All were native speakers of American English.

b Materials:  The materials consisted of 20 experimental stimuli with ambiguous


anaphora appearing in either subordinate or coordinate discourse, as defined in the SDRT
theory described above (Asher 1993; Asher and Vieu, 2005). These were complex sen-
tences comprised of two clauses each; 10 pairs of clauses were linked by the word while,
which evokes parallel events or a coordinate discourse relationship (not to be confused
with the syntactic relation of coordination, which exists independently of discourse
structure), while the other 10 pairs were linked with the words after or until, so that the
discourse relation was one of subordination (again, not to be confused with syntactic
subordination). In all cases the first clause presented two human nouns of the same gen-
der, a subject and an object, while the second clause contained a pronoun that was ambig-
uous in that it could refer back to either of the two nouns from the previous clause. A
sample coordination stimulus with while is given in (7) and a subordination stimulus
with after is provided in (8).

(7) Jeffrey saw Ricky while he was hunting for coordination


coins in the fountain.
(8) Anita talked to her sister after she had the baby. subordination

If the model of discourse structure proposed in SDRT accounts for anaphoric reference
in native English, it was expected that participants would be more likely to interpret the
two clauses in (7) as having the same Topic, whereas in (8) a change of Topic would be
relatively more plausible. In other words, the ambiguous pronoun he in (7) would be more
likely to refer back to the subject antecedent Jeffrey than would the she in (8) be likely to

Downloaded from slr.sagepub.com at ARIZONA STATE UNIV on August 18, 2012


490 Second Language Research 27(4)

refer to Anita. Each stimulus was followed by a binary choice interpretation question
probing the participant’s choice of antecedent for the ambiguous pronoun, as illustrated in
(9). The 20 target stimuli (10 items in each of two conditions) appeared with 20 distract-
ers, for a total of 40 trials. The complete set of English stimuli is listed in Appendix 1.

(9) Who had a baby? interpretation


a. Anita b. Her sister

c Procedure:  The sentence interpretation task was administered via computer using
SuperLab 4.0 stimulus presentation software (Cedrus Corporation, 1992). Target items
appeared in a different randomized order for each participant and were preceded by 10
practice items. Each stimulus appeared in isolation in the center of the computer screen;
the corresponding question and answer choices appeared separately on the following
screen. Participants advanced through the stimuli and answered the interpretation ques-
tions using the seven keys on a Cedrus RB-730 response pad. The task was not timed but
participants were asked to read quickly and to give their first and most instinctive
response to the binary choice questions.

2 Results
An alpha level of < .05 was interpreted as significant and an alpha level of < .10 was
interpreted as marginally significant for this and all further analyses. When the Discourse
Structure involved the subordination of clauses, participants chose as an antecedent the
noun in the SpecIP position 53.26% of the time, whereas with coordinate clauses they
chose SpecIP antecedents 64.19% of the time.
The first set of statistical analyses examined whether the different stimuli types were
associated with different interpretation behavior. Upon comparison of scores with a
paired samples t-test by participants, a significant difference emerged, t1(42) = 3.026, p
= .004. However, in the t-test by items, which was an independent samples test because
the items were not matched across conditions, the difference was not significant, t2(18)
= 1.259, p > .10. This was probably because of a lack of experimental power with only
10 items in each condition. Thus, participants were more likely to select a SpecIP or
subject antecedent for an ambiguous pronoun when it appeared in the second clause of a
coordinate rather than a subordinate discourse.
A second set of statistical tests was conducted on the advice of an anonymous reviewer,
who had suggested that the analyses should specifically assess the likelihood that the anteced-
ent choices for each set of stimuli would arise by chance, given that the experimental task was
all binary choice. To this end, one-sample binomial tests were performed on the total raw
numbers of SpecIP and non-SpecIP choices for each of the two stimulus types. These tests
revealed that the distribution of responses for the subordination stimuli was marginally sig-
nificant, p = .091, and the distribution of responses for the coordination stimuli was signifi-
cant, p < .001. Thus, the observed trends were not likely due to chance. In other words, there
was probably a reliable interpretation preference with both types of stimuli. It should be
noted, however, that the reliability of this preference was most certain with the coordination
stimuli, as the distribution with the subordination stimuli was only marginally significant.

Downloaded from slr.sagepub.com at ARIZONA STATE UNIV on August 18, 2012


Jegerski et al. 491

3 Discussion
In Experiment 1 the difference between coordinate and subordinate discourse structure
influenced intrasentential reference assignment for ambiguous pronouns in English. This
outcome broadly corroborates the model presented in SDRT (Asher, 1993; Asher and Vieu,
2005), where discourse structure is comprised of binary relations of either coordination or
subordination, and this fundamental distinction plays a critical role in the resolution of
anaphora and of referential expressions in general. Specifically, discourse structure is often
correlated with change or continuation of Topic. Coordinate structure reflects discourse
functions such as Narration, where each constituent of the discourse continues to relate a
series of events pertaining to the same Topic. Because the topic of the first clause in exam-
ple (7) is Jeffrey, the topic of the second clause in the narration is often assumed to be the
same. Subordinate structure, on the other hand, is often found with functions like Elabora-
tion, which can involve a shift of Topic. Thus, the first clause in (8) is about Anita, but the
second clause could be an elaboration or aside that refers to a different Topic, her sister.

VI  Experiment 2
The second experiment tested whether discourse structure and the syntax of null and
overt subjects – specifically the division of labor outlined in Carminati’s (2002) PA
hypothesis for Italian – influence the resolution of referential ambiguity in intrasentential
contexts in the Spanish of native speakers.

1 Method
a Participants:  The 26 participants in this experiment were international undergradu-
ates, graduate students, and professionals recruited at two large public universities in
Texas and California. There were 9 males and 17 females, with a mean age of 26.7 years
(range 20–53). All were native speakers of Spanish from Spain and Latin America. None
had acquired English or spent time in the USA until after the age of 15. A few of the
Spaniards had been raised speaking another language of Spain in addition to Castilian, but
these were also null subject languages, such as Catalan and Basque. There were no speak-
ers of Caribbean Spanish included in this group, since that particular dialect may differ
from other varieties of Spanish with regard to the null subject parameter (Toribio, 2000).

b Materials:  The Spanish materials were similar to the English materials in that they were
also complex sentences in which two clauses were joined by either coordination, using the
word mientras ‘while’, or subordination created by después de ‘after’ or antes de ‘before’.
The second of each pair of clauses again contained a pronoun that was ambiguous in terms
of its referent, and a subsequent interpretation question was given to determine which of the
two potential antecedents each reader had selected. In addition, the number of target stimuli
in Spanish was doubled from 20 to 40 because the pro-drop nature of the language intro-
duced a second variable, explicitness of the pronoun. In other words, a given English
sentence would have two equivalent sentences in Spanish, one with a null pronoun and one
with an overt pronoun. The 20 coordination stimuli with mientras were the same ones used

Downloaded from slr.sagepub.com at ARIZONA STATE UNIV on August 18, 2012


492 Second Language Research 27(4)

80
Percentage of SpecIP antecedents
70

60

50 Null pronoun

40 Overt pronoun

30

20
Subordinate clause Coordinate clause

Figure 2  Antecedent choice of native Spanish group according to discourse structure and
pronoun

in a related study (Keating et al., 2011), and the 20 subordination stimuli with después de and
antes de were approximate Spanish translations of those used by Carminati (2002) for Italian,
with adaptations made to eliminate culture-specific references. Thus, this was a 2 (coordinate,
subordinate) × 2 (null, overt) design with 10 stimuli in each of the four conditions, for a total
of 40 experimental stimuli. The 40 experimental items appeared with 40 distractors, for a
total of 80 trials. All 40 experimental Spanish stimuli are given in Appendix 2.

c Procedure:  The procedure for Experiment 2 was the same used for Experiment 1,
although the duration was slightly longer because the total number of stimuli read was
greater.

2 Results
Mean scores reflected 61.23% SpecIP antecedents chosen in subordinate discourse and
63.98% SpecIP antecedents chosen in coordinate discourse; 72.83% SpecIP antecedents
were selected for null pronouns and 52.39% SpecIP antecedents were selected for overt
pronouns. These descriptive statistics are also depicted in Figure 2.
As with Experiment 1, the first set of statistical analyses for Experiment 2 tested
whether antecedent preference differed according to the type of stimulus. For analysis by
participant, the data were submitted to a 2 × 2 repeated measures ANOVA with Discourse
Structure (Subordinate, Coordinate) and Pronoun (Null, Overt) as within-participants fac-
tors. For the analysis by items both Discourse Structure and Pronoun were between-items
factors, to take into account the fact that the stimuli were not matched across conditions.
These analyses revealed a main effect for Pronoun, F1(1, 25) = 10.577, p = .003, F2(1, 36)
= 27.999, p < .001, no main effect for Discourse Structure, F1(1, 25) = .784, p > .10, F2(1,
36) = .552, p > .10, and no significant interaction between the two factors, F1(1, 25) =
.304, p > .10, F2(1, 36) = .062, p > .10. This indicates that whether a Spanish pronoun was
null or overt affected how it was linked to a referent, but that appearing in a coordinate or
subordinate discourse did not affect such intrasentential anaphora resolution.

Downloaded from slr.sagepub.com at ARIZONA STATE UNIV on August 18, 2012


Jegerski et al. 493

The second set of statistical analyses determined whether the observed antecedent
preferences reflected reliable strategies or were likely to have arisen simply by chance,
meaning randomly chosen responses to the binary choice experimental items. One-sam-
ple binomial tests were conducted on the total raw numbers of SpecIP and non-SpecIP
choices for each of the four stimulus types in the 2 × 2 design. The analyses for the sub-
ordination stimuli revealed that the distribution of responses to a stimulus with an ambig-
uous pronoun that was null was significant, p < .001, while the distribution of responses
to a stimulus with an ambiguous pronoun that was overt was not, p > .10. The same was
true of the responses to the coordination stimuli, which varied significantly from chance
when the ambiguous pronoun was null, p < .001, but not when the ambiguous pronoun
was overt, p > .10. Thus, there was a reliable interpretation strategy with null pronouns,
while the interpretation of overt pronouns was more likely due to chance. Furthermore,
these trends held across both types of discourse structure.

3 Discussion
The native Spanish speakers in this experiment were more likely to choose a SpecIP or
subject antecedent for a null subject pronoun than for an overt subject pronoun. Viewed
from the perspective of the secondary set of analyses, there was a reliable interpretation
strategy observed with null pronouns, while interpretation of overt pronouns appeared to
be indifferent or neutral, guided by chance. This result is partially consistent with the PA
hypothesis for Italian applying to Spanish in that it appears that null and overt pronouns
are not equal with regard to their referential functions. However, in the original study with
Italian, Carminati (2002) found that pro was most often linked to a referent in SpecIP and
that the overt pronoun was usually assigned an antecedent in a non-SpecIP position, so the
two tended towards a more complementary distribution. In the present study, only the first
outcome was replicated, as the overt pronoun did not show any preference for a particular
antecedent position, but rather a neutral strategy (around 50%). A similar pattern was
observed by Alonso-Ovalle et al. (2002) for two-sentence discourses in Spanish, but in
that case it was not clear whether the different outcome could be attributed to Spanish or
to the pronominal reference being intersentential. Based on the results of the present
experiment, in which the only difference from the original Carminati study was the lan-
guage tested, it would appear that it is Spanish that may yield a weaker antecedent prefer-
ence for overt pronouns, rather than the intersentential contexts tested by Alonso-Ovalle
et al. However, as overt pronouns have also been shown to be less compliant with the PA
hypothesis among other populations of Italian speakers as well (Sorace and Filiaci, 2006),
it may be that this is a general trend that was not accurately captured in the bulk of the
original Carminati experiments (2002). Still, despite the unpredictability of overt pro-
nouns, our results clearly reflect differential interpretation of null vs. overt pronouns.
Thus, this outcome corroborates a largely PA-like processing strategy in Spanish, whereby
anaphora resolution is affected by the explicitness of the pronoun in question.
Interestingly, the type of discourse structure manipulated in this experiment (i.e.
subordinated vs. coordinated clauses) did not influence the interpretation of subject pro-
nouns in Spanish. This outcome may be surprising, considering that Experiment 1
showed that discourse structure is a significant factor in English and the same structures

Downloaded from slr.sagepub.com at ARIZONA STATE UNIV on August 18, 2012


494 Second Language Research 27(4)

occur in Spanish. It appears that the pro-drop nature of Spanish leads the reader to focus
on the character of the subject pronoun itself, perhaps drawing attention away from the
larger discourse structure during anaphora interpretation. It may be the case that, given
the importance of syntactic principles (i.e. the null/overt distinction) to reference assign-
ment for subject pronouns in Spanish, the discourse structure cues that are significant in
English are relatively peripheral.
Another possible explanation is detailed in the Syntax–Discourse Model (Burkhardt,
2005), which is articulated on the basis of several different referential phenomena in English
and Dutch, including reflexives, and tested empirically with online sentence processing and
ERP (event-related potentials) methods. Burkhardt proposes that the language processor
looks first to the mental representation of syntax in resolving dependencies, such as linking
an anaphor with its antecedent, because a syntactic resolution is most efficient. When the
syntax does not serve to resolve a reference, the processor then undertakes the more time-
consuming task of looking to the representation of discourse for a solution. Such an explana-
tion, while originally articulated to account for data from online processing experiments,
could account for the current observations regarding the offline interpretation of referential
ambiguity in English and Spanish. In English, there is nothing in the non-pro-drop syntax to
aid in the choice between two possible antecedents (except in the case of distinguishing phi
features such as gender or number, which were controlled for in the present study), so the
processor must turn immediately to discourse principles, which then receive primary impor-
tance. In Spanish, however, the syntax licenses both null and overt subjects and this syntac-
tic distinction appears to facilitate pronominal reference, so the initial stages of selection
of antecedents for ambiguous pronouns would be syntactic. According to the Syntax–
Discourse Model, once the syntactic representation has been utilized to make a choice, there
is no need for the processor to move towards the discourse representation. This is not to say
that only syntactic principles are involved in antecedent selection in Spanish, as basic dis-
course principles are likely at play as well, but perhaps the broader representation of dis-
course structure is simply not prioritized in the referential process.
It is important to note, however, that such a syntax–discourse explanation for Spanish
may well only apply under the specific circumstances of the present study, that is, where the
examination of discourse structure was quite narrow, the explicitness of pronouns were also
manipulated at the same time, and, perhaps most importantly, the stimuli were only of sen-
tence length. Different manipulations of discourse structure could affect pronoun interpreta-
tion in Spanish or even this same coordination/subordination manipulation of discourse
structure could prove significant in other contexts, such as in the absence of the pronoun
manipulation or with stimuli of multiple sentences. For these reasons, we emphasize that the
present experiment has tested but one of an innumerable set of discourse factors that could
facilitate pronoun comprehension, and there is good reason to continue exploring the poten-
tially important role of discourse relations in pronominal reference in Spanish.

VII  Experiment 3
The third experiment tested the behavior of intermediate level L2 learners in interpreting
ambiguous pronouns in intrasentential contexts in Spanish to see whether cross-linguistic
influence from English may affect interpretation by early-stage learners. Crucially, this

Downloaded from slr.sagepub.com at ARIZONA STATE UNIV on August 18, 2012


Jegerski et al. 495

was a reasonable comparison because the participants who were tested in Spanish in the
present experiment were the same ones tested in L1 English in Experiment 1. This exper-
iment also tested whether the target L2 strategy had been acquired.

1 Method
a Participants:  Twenty-two undergraduate students at a large public university in
Texas participated in this experiment. All were native speakers of English who had stud-
ied Spanish as adults and were enrolled in a third semester university Spanish course at
the time of testing. This group included 7 males and 15 females and the mean age was
20.7 years (range 19–23). This entire participant group also participated in Experiment 1
in L1 English.

b Materials:  The 40 stimuli for this experiment were the same used for Experiments 2
(reported above) and 4 (reported below) with native and advanced L2 speakers of
Spanish. Sentences contained either null or overt ambiguous pronouns that appeared in
either subordinate or coordinate discourse structures. All of the Spanish stimuli are listed
in Appendix 2.

c Procedure:  Experiment 3 employed the same sentence interpretation task that was
used for Experiments 2 and 4. In order to counterbalance any potential cross-language
priming, half of these participants completed the Spanish experiment first and half com-
pleted the English experiment first. In addition, independent samples t-tests were con-
ducted to compare each set of scores for the two different presentation orders (i.e.
Spanish–English, English–Spanish). No significant differences were found between the
two groups of participants: all ts < 1.2 and all ps > .240. Thus, the fact that these partici-
pants read similar stimuli in Spanish and English did not appear to affect their
performance.

2 Results
The descriptive statistics are displayed in Figure 3. Overall, the participants in this experi-
ment chose 52.95% SpecIP antecedents in subordinate discourse and 62.96% SpecIP ante-
cedents in coordinate discourse. Additionally, 60.68% SpecIP antecedents were selected
for null pronouns and 55.23% SpecIP antecedents were selected for overt pronouns.
Again, the first set of statistical analyses tested whether different stimuli types were
associated with different antecedent preferences. In the analysis by participants, the
interpretation data were submitted to a 2 × 2 repeated measures ANOVA with Discourse
Structure (Subordinate, Coordinate) and Pronoun (Null, Overt) as within-participants
factors, while the analysis by items included Discourse Structure and Pronoun as
between-items factors to take into account the fact that the stimuli were not matched
across experimental conditions. The effect for Discourse Structure was marginally sig-
nificant by participants, F1(1, 21) = 3.540, p = .074, and significant by items, F2(1, 36) =
8.669, p = .006. There was no significant main effect for Pronoun, F1(1, 21) = 2.866, p >
.10, F2(1, 36) = 2.587, p > .10, and no interaction between the two factors, F1(1, 21) =

Downloaded from slr.sagepub.com at ARIZONA STATE UNIV on August 18, 2012


496 Second Language Research 27(4)

80
Percentage of SpecIP antecedents
70

60

50 Null pronoun

40 Overt pronoun

30

20
Subordinate clause Coordinate clause

Figure 3  Antecedent choice of intermediate Spanish group according to discourse structure


and pronoun

.025, p > .10, F2(1, 36) = .018, p > .10. Thus, anaphora resolution among this participant
group was not affected by the type of pronoun, but was likely affected by the type of
discourse structure.
A second set of statistical analyses examined whether the observed antecedent prefer-
ences for each type of stimulus were likely to have arisen by chance. One-sample bino-
mial tests were conducted on the total raw numbers of SpecIP and non-SpecIP choices
for each of the four stimulus types in the 2 × 2 design. The tests for the subordination
stimuli revealed that the distribution of antecedent choices was not significant when the
ambiguous pronoun was null, nor when the pronoun was overt, ps > .10. With the coor-
dination stimuli, on the other hand, the distribution varied significantly from chance with
both null pronouns, p < .001, and overt pronouns, p = .004. Hence, overall there was a
reliable interpretation strategy with coordinate discourse, while the interpretation of pro-
nouns in subordinate discourse was more likely due to chance. Furthermore, these trends
held across both types of pronoun.

3 Discussion
The low intermediate level learners of Spanish in this experiment did not exhibit any
PA-like effects in their interpretation of null and overt pronouns, but they did show evi-
dence of English-like effects for coordinate vs. subordinate discourse structure. Further-
more, these participants employed a reliable interpretation strategy with the coordination
stimuli but not with the subordination stimuli, and in both cases this was regardless of
pronoun type. The first outcome is probably of little surprise, given the relatively low
level of proficiency of the participants, who were enrolled in a third semester university
language course. The syntactic distinction between null and overt pronouns did not affect
the interpretation behavior of this group, so it is possible that either they have not acquired
the syntax of pro-drop in their L2 systems, or perhaps that they do have the syntax but it
is somehow inaccessible in this context and therefore cannot be applied. Another possi-
ble explanation is that the syntax of the null subject is there, but the PA principles of

Downloaded from slr.sagepub.com at ARIZONA STATE UNIV on August 18, 2012


Jegerski et al. 497

reference assignment that rely on that syntax are what has yet to be acquired. While fur-
ther research would be needed to tease apart these potential causes with any degree of
certainty, at this point it is clear that PA-like behavior does not appear early on in L2
Spanish acquisition.
The second outcome of Experiment 3 – that the main effect of Discourse Structure
was significant by items and marginally significant by participants (p = .074) – is perhaps
more interesting. The group showed a trend towards applying a discourse-based interpre-
tation strategy in their L2 that is similar to that observed in their native English. In terms
of deviation from chance, the strategy was the same as that observed in L1 English,
whereby reliable strategy was employed only in cases of coordination. Such behavior
could therefore be the product of cross-linguistic influence from the native language.
Furthermore, it seems that the absence (or inactivity) of pro-drop syntax and PA princi-
ples allowed principles of discourse structure to take prominence.

VIII  Experiment 4
The purpose of the final experiment was to explore whether adult L2 learners of Spanish
at a more advanced level still show evidence of L1 transfer of the discourse-based inter-
pretation strategy from English and whether they have acquired a native-like pronoun
interpretation strategy in the L2. As with Experiment 3, the possible effect of transfer
could be considered, as the participants who were tested in L2 Spanish in the present
experiment were also tested in English in Experiment 1.

1 Method
a Participants:  The 21 participants in this experiment were native speakers of English
who had completed at least a university minor in Spanish or the equivalent through
immersion abroad. There were 12 males and 9 females in this group and the mean age
was 32.1 years (range 20–61). Most were graduate students and professors at two large
public universities in Texas and California, while a few were professionals from the
broader university communities. All participants in this group completed the LEAP-Q
Language Experience and Proficiency Questionnaire (Marian et al., 2007), a language
background survey for bilinguals. Some key factors from the questionnaire, including
age of acquisition and self-rated proficiency in Spanish for this group, are summarized
in Table 1.

Table 1  Spanish background information for advanced L2 participants in experiment 4

M SD Minimum Maximum

Age 32.1 13.4 21 61


Age of first exposure 15.0  3.5  8 24
Immersion abroad (years)  1.6  1.5  0  5.5
Self-rated proficiency (1–10):
  Speaking  7.9  1.2  5 10
  Understanding  8.0  1.6  5 10
  Reading  8.0  1.4  5 10

Downloaded from slr.sagepub.com at ARIZONA STATE UNIV on August 18, 2012


498 Second Language Research 27(4)

80
Percentage of SpecIP antecedents
70

60

50 Null pronoun

40 Overt pronoun

30

20
Subordinate clause Coordinate clause

Figure 4  Antecedent choice of advanced Spanish group according to discourse structure and
pronoun

b Materials:  The 40 stimuli for this experiment were the same items used for
Experiment 2 with native speakers of Spanish and for Experiment 3 with the intermedi-
ate learners of Spanish. All were complex sentences in which two clauses were joined by
either coordination or subordination, and the second clause of each sentence contained a
null or overt pronoun that was ambiguous with regard to its antecedent. The complete set
of Spanish stimuli is provided in Appendix 2.

c Procedure:  The procedure for Experiment 4 was the same used for Experiments 2
and 3.

2 Results
Mean scores for this experiment were 53.81% SpecIP antecedents in subordinate dis-
course and 63.33% SpecIP antecedents in coordinate discourse; 60.95% SpecIP anteced-
ents were selected for null pronouns and 56.19% SpecIP antecedents were selected for
overt pronouns. The descriptive statistics by condition are provided in Figure 4.
As with the first three experiments in this study, the first set of statistical analyses for
Experiment 4 determined whether the different stimuli types were associated with differ-
ent interpretation behavior. The data by participants were submitted to a 2 × 2 repeated
measures ANOVA with Discourse Structure (Subordinate, Coordinate) and Pronoun
(Null, Overt) as within-participants factors. The data by items were submitted to a 2 × 2
ANOVA with both Discourse Structure and Pronoun as between-items factors, in order
to take into account the fact that the stimuli were not matched across conditions. These
analyses revealed a marginally significant effect for Discourse Structure, F1(1, 20) =
4.098, p = .056, F2(1, 36) = 3.070, p = .088, and no main effect for Pronoun, F1(1, 20) =
1.438, p > .10, F2(1, 36) = 1.332, p > .10. The interaction between the two factors was
significant by participants, F1(1, 20) = 5.974, p = .024, but not significant by items, F2(1,
36) = 2.409, p > .10.2 Post-hoc t-tests were conducted to probe the interaction; as with
the ANOVAs, these were paired by participants and independent by items. A comparison

Downloaded from slr.sagepub.com at ARIZONA STATE UNIV on August 18, 2012


Jegerski et al. 499

of null and overt pronouns in subordinate discourse revealed no significant difference,


t1(20) = .616, p > .10, t2(18) = .236, p > .10, while the same comparison for coordinate
discourse showed significantly more SpecIP antecedents chosen for null pronouns than
for overt pronouns, t1(20) = 2.280, p = .034, t2(18) = 2.530, p = .021. Thus, the advanced
L2 participants in this group showed a marginal effect for discourse and also differenti-
ated between null and overt pronouns in coordinate discourse, but not in subordinate
discourse.
A second set of statistical tests was conducted to assess the likelihood that the ante-
cedent choices for each type of stimulus would arise by chance, as opposed to through
the application of a specific interpretation strategy. One-sample binomial tests were con-
ducted on the total raw numbers of SpecIP and non-SpecIP choices for each of the four
stimulus types in the 2 × 2 design. These analyses revealed that the distribution of ante-
cedent choices for the subordination stimuli was not significant when the ambiguous
pronoun was null, nor when the pronoun was overt, ps > .10. With the coordination
stimuli, the distribution was significant with the null pronoun, p < .001, but not with the
overt pronoun, p > .10. Thus, there was a reliable interpretation strategy only with coor-
dination stimuli with null pronouns, while the distributions of responses with the other
three stimulus types likely arose by chance.

3 Discussion
The advanced L2 speakers of Spanish who participated in this experiment showed a
borderline main effect for Discourse Structure. This tendency to rely on the discourse
relationships of coordination and subordination between two clauses for the purpose of
pronoun interpretation is an English-like strategy that potentially reflects influence from
the pre-existing language representation. It is interesting to note that the nearly signifi-
cant effect observed among intermediate learners in Experiment 3 is maintained at an
advanced level of proficiency in the current experiment. This result suggests that cross-
linguistic influence can affect L2 processing for quite some time.
On the other hand, there was no main effect for Pronoun or general PA-like effect. It
appears that these participants have not completely acquired a native-like Spanish strat-
egy for assigning intrasentential referents to null and overt pronouns. However, the inter-
action between Pronoun and Discourse Structure was due to an effect for Pronoun, or a
native-like PA effect, that occurred in the stimuli with coordination, but not in those with
subordination. In terms of deviation from chance, only those stimuli with both a null
pronoun and coordination were interpreted with a reliable strategy, as opposed to indif-
ferent behavior guided by chance. In other words, these participants seemed to be
employing to a certain degree both L1-like and L2-like strategy. The preference for a
SpecIP antecedent was only realized when both the L1-like discourse-based strategy and
the L2-like PA strategy coincided among null pronouns in coordinate discourse. When
only one of the two strategies necessitated the selection of a SpecIP antecedent (i.e. with
null pronouns in subordinate discourse or with overt pronouns in coordinate discourse),
the two seemed to be at odds or even to cancel each other out, resulting in essentially
random behavior. This apparent dual activation of discourse-based and PA principles
perhaps reflects a transition from an earlier period of wholly discourse-based strategy to

Downloaded from slr.sagepub.com at ARIZONA STATE UNIV on August 18, 2012


500 Second Language Research 27(4)

native-like PA behavior that occurs only at a more advanced level, towards ultimate
attainment. Alternatively, this hybrid processing performance may indicate that PA-like
behavior is in principle acquirable by L2 learners and that what is lacking is not the
capacity to learn L2 processing strategy, but rather the ability to completely disregard or
deactivate the previously established strategy that is active in the native language.

IX  General discussion and conclusions


In Experiments 1 and 2 the discourse structural distinction of coordination vs. subordina-
tion affected anaphora resolution in English, whereas pronoun interpretation behavior in
Spanish followed instead a strategy that was broadly similar to that predicted by the Posi-
tion of Antecedent (PA) hypothesis (Carminati, 2002). The intermediate L2 Spanish par-
ticipants in Experiment 3 showed no native-like PA effects, but instead used discourse
structure as a cue in their interpretation of null and overt pronouns. In Experiment 4 the
advanced L2 Spanish participants also demonstrated this discourse-based strategy, which
influenced their interpretation of ambiguous pronouns, while showing no entirely native-
like distinction between Spanish null and overt pronouns in general. However, these
participants did more narrowly apply a PA-like strategy in only those stimuli with coor-
dinate discourse structure.
The results of Experiment 1 for native English represent novel empirical corroboration
for the critical importance of coordination and subordination in discourse structure. The
coordination/subordination distinction, which is fundamental in Segmented Discourse
Representation Theory (SDRT; Asher, 1993; Asher and Vieu, 2005), proved to be a sig-
nificant cue to readers in the interpretation of ambiguous pronouns with intrasentential
referents in English. Furthermore, Experiment 2 demonstrated that this same discourse
cue was not significant in native Spanish, which suggests that there is some degree of
cross-linguistic variation in the importance of coordination/subordination as a cue, at least
as it pertains to pronoun interpretation in the specific type of isolated sentences employed
in the present study. This variation between languages may correspond to the activation of
pro-drop syntax, as Spanish is a null subject language and English is not. It appears that
behaviors that rely at least in part on the syntactic-based principles of pro-drop, such as
PA-like behavior (Carminati, 2002), may hierarchically obviate some of the broader dis-
course principles, such as the specific instance of discourse structure manipulated in the
current investigation. Further research is needed to test the proposition that null subject
variation is paralleled by cross-linguistic variation in the importance of discourse struc-
ture variables such as coordination and subordination to pronoun interpretation, and to
explore more generally the complex interaction between syntax and discourse principles
during the linking of pronouns with intrasentential antecedents.
The outcomes of Experiments 3 and 4 answer to the broad SLA issue of how the
learner handles anaphora resolution in a non-native language. The first research question
guiding this study was: Can native-like L2 strategy be acquired over time? The results
presented herein do not address this question directly, but do suggest that it can in prin-
ciple be acquired. On the one hand, both the intermediate and the advanced learner par-
ticipant groups failed to employ a wholly native-like PA strategy while interpreting
Spanish sentences, as is evident in Figure 5. Parenthetically, it is worth noting that this

Downloaded from slr.sagepub.com at ARIZONA STATE UNIV on August 18, 2012


Jegerski et al. 501

Percentage of SpecIP antecedents 80

70

60
Null pronoun

Overt pronoun
50

40
Intermediate Advanced Spanish Nave Spanish
Spanish

Figure 5  Antecedent choice according to pronoun in Spanish

non-native behavior is not likely due to an underlying grammatical deficit, since previ-
ous research on the acquisition of the null subject parameter has shown that the relevant
grammar is learned at a level lower than that of our advanced group (Rothman and Iver-
son, 2007a, 2007b; Rothman, 2008, 2009). What is more, as Figure 4 shows, the advanced
group did follow the PA hypothesis in one specific context, in coordinated discourse,
which suggests that some degree of learning of L2 interpretation principles takes place
with linguistic development. As the advanced group of participants was not of the high-
est proficiency (i.e. at ultimate attainment), it is plausible that further learning of target-
like PA strategy may take place and be manifest at a more advanced level. Such was the
case in Sorace and Filiaci’s (2006) study of anaphora resolution among near-native
speakers of Italian. In any case, our experimental results suggest that this particular inter-
face phenomenon does indeed present a challenge in acquisition, as we did not find evi-
dence of entirely native-like behavior at the syntax–discourse interface, not even among
the advanced participant group. This outcome is consistent with the developmental
sequence predicted by the Interface Hypothesis (Sorace, 2004; Sorace and Filiaci, 2006)
and even by those who have presented contradictory evidence (e.g. Rothman, 2009), but
should not be taken as direct support for the purportedly insurmountable nature of resid-
ual optionality at linguistic interfaces predicted by the Interface Hypothesis, given that
the participants in the present study were probably not at ultimate attainment. This is
especially true in the specific case of the acquisition of pronominal reference, as child L1
acquirers of Spanish are known to take more than 10 years to develop adult-like prefer-
ences (Shin and Cairns, 2009), while the participants in the present study had an average
of less than two years immersion in Spanish. Finally, as with previous research, the PA
data alone do not indicate whether cross-linguistic influence may be a source of non-
nativeness in this case.
The English and discourse structure data from the present study are critical in answer-
ing our second research question: Where behavior is divergent from the native target, is

Downloaded from slr.sagepub.com at ARIZONA STATE UNIV on August 18, 2012


502 Second Language Research 27(4)

Percentage of SpecIP antecedents 80

70

60
Subordinate

Coordinate
50

40
Nave English Intermediate Advanced Nave
Spanish Spanish Spanish

Figure 6  Antecedent choice according to discourse structure in English and Spanish

there any similarity to L1 behavior? Both the intermediate and advanced L2 learners of
Spanish in Experiments 3 and 4 showed differential interpretation behavior according to
Discourse Structure, which is an effect that these same participants showed in their
native English in Experiment 1, and that native Spanish speakers did not exhibit in
Experiment 2. This outcome, illustrated in Figure 6, suggests that the interpretation strat-
egy present in the existing linguistic system may begin to influence the developing L2 at
some level of proficiency prior to the intermediate level of those tested, and that this
L1-like strategy remains active even at the advanced level. Furthermore, the Discourse
Structure by Pronoun interaction displayed by the advanced L2 Spanish participants
indicates that some degree of L1-like strategy persists even after target-like L2 strategy
begins to appear. There is in a sense a doubling-up of the L1 and L2 strategy, as reliable
interpretation behavior was only evident where both preferences overlapped, as can be
seen in Table 2. Thus, it seems that the English-speaking L2 learner of Spanish initially
relies solely on discourse-based principles, and subsequently begins to acquire and apply
PA-like principles in combination with the pre-existing strategy. In other words, cross-
linguistic influence occurs and may even be a primary cause of non-native behavior, up
through the advanced level.
Of course, there is an alternative interpretation of the results of Experiments 3 and 4
with the non-native groups. In addition to being a strategy present in native English, the
effect of discourse coordination and subordination on subject pronoun interpretation
could prove to be a universal tendency in language comprehension, meaning a default
behavior that would arise in any context in which it was not trumped by a more powerful
language-specific strategy. Under such an interpretation, the discourse-based strategy
would have been present but invisible among the native Spanish participants in the pres-
ent study, where PA-like principles would have taken precedence. Additionally, the
apparent cross-linguistic influence among the L2 Spanish participants would not have
stemmed directly from their being native speakers of English and could thus be observed

Downloaded from slr.sagepub.com at ARIZONA STATE UNIV on August 18, 2012


Jegerski et al. 503

Table 2  Binomial tests for all experiments and all stimulus types

Experiment Subordination Coordination

Null Overt Null Overt

1. English L1 – p = .091 – p < .001


2. Spanish L1 p < .001 p = .664 p < .001 p = .352
3. Spanish L2 intermediate p = .121 p = .946 p < .001 p = .004
4. Spanish L2 advanced p = .946 p = .544 p < .001 p = .381
Note: Effects significant at α = .05 appear in boldface.

among any population of L2 learners of Spanish, even those whose native language is
also pro-drop. PA strategy of the type observed in Spanish, on the other hand, cannot be
universally active in null subject languages, as the opposite pattern of behavior has been
documented for Arabic (Ioup et al., 1994). While the distinction of universal language
tendencies from native language influence is certainly not trivial and should be the goal
of future research, in either case, the challenge for the L2 acquirer of Spanish pronominal
reference principles remains essentially similar: to learn, on the one hand, a new strategy
and to deactivate, on the other, an underlying strategy that predates the L2 principles.
There are some limitations to the present study that should be acknowledged. Like
most previous research in the area of pronominal reference in Spanish and in Italian, the
current investigation employed an offline measure of pronoun interpretation. Thus, we
were unable to make any temporal distinctions between different phases of processing,
which may prove to be of interest in the acquisition of anaphoric reference. In addition,
there were age differences between the participants for the four experiments in the cur-
rent investigation. The inclusion of professors and other professionals in the advanced
L2 Spanish group and both native speaker groups increased the ages to a mean of 26.3
(range 19–61) for Experiment 1, a mean of 26.7 (range 20–53) for Experiment 2, and a
mean of 32.1 (range 20–61) for Experiment 4. Only the intermediate L2 group in Experi-
ment 3 was comprised entirely of undergraduate students with a traditional age mean of
20.7 (range 19–23). However, this could equally be viewed as a strength of the design,
as it means the conclusions may be generalizable beyond the typically studied university
undergraduate population. Finally, our specific manipulation of subordination and coor-
dination was but one narrow instance of the numerous and complex set of discourse
structure variables that could influence pronominal reference in English or Spanish, so
our conclusions regarding the role of discourse structure are not in any way definitive.
On the contrary, the observation that one type of discourse structure affected pronominal
reference in English and in L2 Spanish is a motive for continued research in this area.
In conclusion, this study provides empirical evidence that linguistic phenomena that
occur at the interface between syntax and discourse, such as the interpretation of ambigu-
ous intrasentential anaphora in Spanish, present a particular challenge to the L2 learner.
Most importantly, we have shown that cross-linguistic influence can be a factor when L2
interpretation performance diverges from the native-like target, even at an advanced
level of L2 proficiency and even after some degree of native-like behavior has emerged.
Thus, the acquisition of native-like L2 interpretation strategy may be as much about

Downloaded from slr.sagepub.com at ARIZONA STATE UNIV on August 18, 2012


504 Second Language Research 27(4)

abandoning pre-existing language behavior as it is about learning to apply new princi-


ples. Further research in this area will determine whether cross-linguistic influence might
also be important in other types of sentence comprehension, in addition to the linking of
pronouns to intrasentential antecedents.

Acknowledgements
This study was funded in part by a research grant awarded to Gregory D. Keating by the San Diego
State University Research Foundation, as well as research monies allocated to Bill VanPatten from
the Dean of Arts and Sciences at Texas Tech University. An earlier version of this article was pre-
sented at the Conference on Second Language Processing and Parsing in May 2009 in Lubbock,
TX, USA. We are grateful to that audience and to three anonymous Second Language Research
reviewers for their many insightful comments on our work.

Notes
1 It should be noted that the Interface Hypothesis has been revised on the basis of more recent
contradictory evidence. Sorace and Serratrice (2009), for instance, suggest that it is the inter-
faces between grammar modules and grammar-external representations in memory that are
most likely to present challenges during early and late bilingual acquisition, while those areas
of language that occur at the interfaces between grammar-internal components are more read-
ily acquired. The target of the present study, as a syntax–discourse phenomenon, would
remain a challenge under the revised version of the hypothesis.
2 One anonymous reviewer noted that ANOVAs conducted on proportional data can produce
questionable results. Specifically, Jaeger (2008) outlines how in some cases an interaction
could be an artifact of the statistical model itself and argues that a mixed logit model would
yield more accurate results than an ANOVA. However, while this is admittedly a potential
limitation of the present study, the results of Experiment 4 suggested by the interaction in the
ANOVA were also corroborated by the post-hoc t-tests, which are uncontrovertibly appropri-
ate for proportional data, so the overall outcome is reliable. What is more, the pattern of reli-
able strategy vs. chance outcome that was revealed by the series of binomial tests further
bolsters the results of the ANOVAs.

References
Alonso-Ovalle L, Fernandez-Solera S, Frazier L, and Clifton C (2002) Null vs. overt pronouns and
the topic–focus articulation in Spanish. Rivista di Linguistica 14: 151–69.
Asher N (1993) Reference to abstract objects in discourse. Dordrecht: Kluwer Academic.
Asher N and Vieu L (2005) Subordinating and coordinating discourse relations. Lingua 115: 591–610.
Birdsong D (2006) Why not fossilization. In: Han Z and Odlin T (eds) Studies of fossilization in
second language acquisition. Clevedon: Multilingual Matters, 173–88.
Burkhardt P (2005) The syntax–discourse interface: Representing and interpreting dependency.
Amsterdam: John Benjamins.
Carminati M (2002) The processing of Italian subject pronouns. Unpublished Doctoral Dissertation,
University of Massachusetts, Amherst, MA.
Cedrus Corporation (1992) SuperLab 4.0.5 [software]. San Pedro, CA: Cedrus Corporation.
Clahsen H and Felser C (2006) Grammatical processing in language learners. Applied
Psycholinguistics 27: 3–42.
Gordon P, Grosz B, and Gilliom L (1993) Pronouns, names, and the centering of attention in dis-
course. Cognitive Science 17: 311–47.

Downloaded from slr.sagepub.com at ARIZONA STATE UNIV on August 18, 2012


Jegerski et al. 505

Hudson-D’Zmura S and Tanenhaus M (1998) Assigning antecedents to ambiguous pronouns: The


role of center of attention as the default assignment. In: Walker M, Joshi A, and Prince E (eds)
Centering theory in discourse. Oxford: Oxford University Press, 199–226.
Ioup G, Boustagui E, El Tigi M, and Moselle M (1994) Reexamining the critical period hypothesis:
A case study in a naturalistic environment. Studies in Second Language Acquisition 16: 73–98.
Jaeger TF (2008) Categorical data analysis: Away from ANOVAs (transformation or not) and
towards logit models. Journal of Memory and Language 59: 434–46.
Keating GD, VanPatten B, and Jegerski J (2011) Who was walking on the beach? Anaphora reso-
lution in Spanish heritage speakers and adult second language learners. Studies in Second
Language Acquisition 33: 193–221.
Lujan M (1985) Binding properties of overt pronouns in null pronominal languages. In: Eilfort
W, Kroeler P, and Peterson K (eds) Proceedings of the Twenty-first Regional Meeting of the
Chicago Linguistic Society. Chicago, IL: University of Chicago, 424–38.
Lujan M (1986) Stress and binding of pronouns. In: Farley A, Farley P, and McCullough K-E
(eds) Papers from the Parasession on Pragmatics and Grammatical Theory at the Chicago
Linguistics Society. Chicago, IL: University of Chicago, 248–62.
Marian V, Blumenfeld HK, and Kaushanskaya M (2007) The Language Experience and Proficiency
Questionnaire (LEAP-Q): Assessing language profiles in bilinguals and multilinguals. Journal
of Speech, Language and Hearing Research 50: 940–67.
Rothman J (2008) How pragmatically odd! Interface delays and pronominal subject distribution in
L2 Spanish. Studies in Hispanic and Lusophone Linguistics 1: 317–39.
Rothman J (2009) Pragmatic deficits with syntactic consequences?: L2 pronominal subjects and
the syntax–pragmatics interface. Journal of Pragmatics 41: 951–73.
Rothman J and Iverson M (2007a) On L2 clustering and resetting the null subject parameter in L2
Spanish: Implications and observations. Hispania 90: 329–42.
Rothman J and Iverson M (2007b) The syntax of pronominal subjects in L2 Spanish: Comparing
two L2 populations with different exposure. Revista Española de Lingüística Aplicada 20:
185–214.
Shin NL and Cairns HS (2009) Subject pronouns in child Spanish and Continuity of Reference.
In: Collentine J, García M, Lafford B, and Marín FM (eds) Selected Proceedings of the 11th
Hispanic Linguistics Symposium. Somerville, MA: Cascadilla Proceedings Project, 155–64.
Sorace A (2000) Syntactic optionality in non-native grammars. Second Language Research 16:
93–102.
Sorace A (2004) Native language attrition and developmental instability at the syntax–discourse
interface: Data, interpretations and methods. Bilingualism: Language and Cognition 7: 143–45.
Sorace A (2006) Possible manifestations of shallow processing in advanced second language
speakers. Applied Psycholinguistics 27: 88–91.
Sorace A and Filiaci F (2006) Anaphora resolution in near-native speakers of Italian. Second
Language Research 22: 339–68.
Sorace A and Serratrice L (2009) Internal and external interfaces in bilingual language develop-
ment: Beyond structural overlap. International Journal of Bilingualism 13: 195–210.
Toribio AJ (2000) Setting parametric limits on dialectal variation in Spanish. Lingua 110: 315–41.
White L (2007) Grammatical theory, interfaces and L2 knowledge. Plenary paper presented at the
Second Language Research Forum, Urbana, IL, USA.
Zagona K (2002) The syntax of Spanish. Cambridge: Cambridge University Press.

Downloaded from slr.sagepub.com at ARIZONA STATE UNIV on August 18, 2012


506 Second Language Research 27(4)

Appendix 1  English stimuli for experiment 1


Coordinate discourse
 1. Janice saw Marilyn while she was playing Frisbee with her dog.
 2. Mary ran into Helen while she was shopping for a new blouse at Macy’s.
 3. Wanda saw her girl friend while she was drinking coffee at the local diner.
 4. The maid saw her daughter while she was driving by the house on Grand
Street.
 5. The woman ran into her sister while she was running errands downtown.
 6. David ran into Harry while he was strolling with his dog in the park.
 7. Jeffrey saw Ricky while he was hunting for coins in the fountain.
 8. Phillip saw his father while he was washing the car in the driveway.
 9. Robert saw Sam while he was walking on the beach in Santa Monica.
10. The man saw Mr. Smith while he was running in the park yesterday afternoon.

Subordinate discourse
11. Anita talked to her sister after she had the baby.
12. Gloria didn’t speak to her sister until she had won a million dollars in Las
Vegas.
13. Janet didn’t speak to Mary until she had become an attorney in Chicago.
14. Julia never saw Martha again after she left the clinic in Minnesota.
15. The waitress didn’t see her girl friend after she got divorced last December.
16. John called his brother after he bought a new S-class Mercedes.
17. Ralph didn’t see Peter after he left for Hollywood to become famous.
18. Stan called his brother after he won the national tennis championship.
19. The man didn’t see his dad until he’d returned from a vacation to Cancun.
20. Tom did not talk to his buddy until he returned from his lunch break.

Appendix 2  Spanish stimuli for experiments 2, 3, and 4


Coordinate discourse, null pronoun
 1. Susana dio con Roberto mientras corría en el parque esta mañana.
 2. Amanda vio al chico mientras cenaba en el restaurante anoche.
 3. La señorita vio al maestro mientras tomaba un café con leche en Starbucks.
 4. Luisa vio al vecino mientras compraba una entrada al cine.
 5. La profesora vio al estudiante mientras caminaba en la calle Grand.
 6. El niño vio a la niña mientras jugaba en el jardín de la escuela ayer.
 7. José vio a Leticia mientras caminaba en la playa de Malibú.
 8. El señor dio con su suegra mientras paseaba con un amigo enfrente de la casa.
 9. Ricardo vio a Evita mientras jugaba con su perro en el parque.
10. El señor vio a la abogada mientras esperaba un taxi enfrente del edificio.

Downloaded from slr.sagepub.com at ARIZONA STATE UNIV on August 18, 2012


Jegerski et al. 507

Coordinate discourse, overt pronoun


11. Alicia se encontró con Elena mientras ella corría en el parque esta mañana.
12. Rebeca vio a su hermana mientras ella estudiaba en la biblioteca anoche.
13. Marta vio a la vecina mientras ella trabajaba en el jardín.
14. La actriz dio con su hermana mientras ella tomaba un cóctel con un amigo.
15. La señora vio a su buena amiga mientras ella hacía compras en el centro.
16. El niño vio a su hermano mientras él jugaba futbol con los vecinos.*
17. Antonio vio a Rafael mientras él caminaba en el centro ayer.
18. El hombre dio con su empleado mientras él paseaba por la calle.
19. Carlos vio a David mientras él compraba una corbata en Bloomingdale’s.
20. El abogado vio al doctor mientras él esperaba a su mujer en el lobby.

Subordinate discourse, null pronoun


21. Marta le escribía frecuentemente a Lorena cuando estaba en los Estados Unidos.
22. Maribela se hizo amiga de Sandra después de que se divorció.
23. Carolina llamaba a Paloma cada semana cuando vivía en Madrid.
24. La chica vio a su amiga cuando estaba en la tienda.
25. Clara sale con Mariana cuando no está trabajando.
26. Daniel ya no ve a Miguel desde que se casó.
27. Pedro no habla con Víctor desde que dio su resignación.
28. Marco se reconcilió con su hermano después de que regresó del ejército.
29. Humberto pegó a José cuando estaba jugando.
30. Alejandro visitó al profesor después de que regresó de Barcelona.

Subordinate discourse, overt pronoun


31. Gloria le enviaba cartas a su mamá cuando ella estaba en Buenos Aires.
32. Cristina veía a Lina todos los días cuando ella vivía en Cancún.
33. Susana llamó a Paola cuando ella estaba en la oficina.
34. María hace las compras con Lucía cuando ella está libre.
35. Natalia conoció a Isabela después de que ella se mudó a Chicago.
36. Roberto insultó a Hugo cuando él estaba borracho.
37. Diego le escribió a Eduardo después de que él se mudó a San Francisco.
38. Ramón ya no charla con Paco desde que él perdió su trabajo.
39. Tomás no pelea con Antonio desde cuando él tuvo un accidente.
40. Francisco se encontró con Carlos después de que él se casó.

Note: * It is worth noting that the Spanish verb jugar is associated with dialectal variation
as to whether or not it must be followed by the preposition a, while the noun futbol varies
with regard to which of its two syllables receives tonic stress. The variants found in our
experimental materials are those common to most Latin American dialects.

Downloaded from slr.sagepub.com at ARIZONA STATE UNIV on August 18, 2012

You might also like