GRAMMATICAL GENDER AGREEMENT IN L2 SPANISH:
THE ROLE OF SYNTACTIC CONTEXT
By
Le Anne L. Spino-Seijas

A DISSERTATION
Submitted to
Michigan State University
in partial fulfillment of the requirements
for the degree of
Second Language Studies—Doctor of Philosophy
2017

	
  

ABSTRACT
GRAMMATICAL GENDER AGREEMENT IN L2 SPANISH:
THE ROLE OF SYNTACTIC CONTEXT
By
Le Anne L. Spino-Seijas
A pervasive question in second language (L2) research is whether L2 learners can acquire
parameterized functional features that are not instantiated in their first language (L1). While
some researchers have argued for a representational deficit (e.g., Clahsen & Muysken, 1989;
Hawkins & Chan, 1997), claiming that L2 learners’ competence is fundamentally deficient,
others have argued that learners can indeed acquire features that are not instantiated in their L1
(e.g., Prévost & White, 2000; Schwartz & Sprouse, 1996), and ascribe any optionality to
communication pressure or other external factors. In this dissertation, grammatical gender
agreement was used as a test case to determine if L2 Spanish learners (L1 = English) can indeed
acquire a parametrized feature not present in their L1.
Many researchers that investigate grammatical gender agreement do not have a principled
reason for investigating a particular type of agreement (e.g., determiner-noun, noun-adjective). It
may be the case, though, that not all types of agreement are equally difficult for L2 learners. In
studies that investigate whether L2 learners have a representational deficit for grammatical
gender agreement, it is therefore impossible to conclude whether learners truly have a
representational deficit, or whether they are performing poorly because of the type of agreement
under investigation. Therefore, this dissertation tests grammatical gender agreement in three
different syntactic contexts that are commonly used in this type of research: determiner-noun
(DET-N), noun-adjective (N-ADJ) and null nominal (N-DROP). These syntactic contexts were

	
  

hypothesized to differ in difficulty for L2 learners, with DET-N being the easiest and N-DROP
the most difficult.
Native Spanish speakers and L2 learners read a series of sentences embedded with
violations of these three different types of grammatical gender agreement while their eyemovements were recorded with an eye-tracker. Participants’ sensitivity was measured both via
reading times and self-reports on a post-reading questionnaire. Linear mixed-effects models
indicated that native Spanish speakers were sensitive to all three types of grammatical gender
agreement, as evidenced by longer reading times on ungrammatical relative to grammatical areas
of interest, but L2 Spanish learners were sensitive only to DET-N agreement, and not N-ADJ and
N-DROP agreement. The self-reports paralleled these findings, with L2 learners reporting a
higher instance of seeing DET-N agreement violations than N-ADJ and N-DROP violations in
the experimental stimuli. These results indicate that the L2 learners likely do not have a
representational deficit for grammatical gender agreement, and that the type of grammatical
gender agreement under investigation matters, as the syntactic context of the agreement may
affect performance. Results are discussed in terms of the types of knowledge L2 learners use
during online processing in studies that detect sensitivity to grammatical violations.

	
  

Copyright by
LE ANNE L. SPINO-SEIJAS
2017

	
  

In loving memory of Philip P. Spino
best of fathers and best of men

	
  

v

ACKNOWLEDGEMENTS

I am indebted to so many people for their assistance and guidance as I navigated my way
through this degree. The following is my humble attempt to express my gratitude for those who
have helped me along the way.
First, I’d like to thank Bill VanPatten for being such a limitless source of knowledge. To
say that his outlook on language research and instruction has been instrumental in my career is
an understatement. Among other things, I am thankful for all the time he has spent going over
my research with me, for his patience in showing me how to most effectively create class
activities, and for the many, many times he has made me laugh. I am also greatly indebted to
Aline Godfroid for long and productive conversations about research, academia, and life in
general. My interactions with Aline have undoubtedly made me a better researcher, a better
writer, and a more thoughtful person. I have so many fond memories of our time together in
Michigan, and I am privileged to count her as a friend. I also extend my thanks to Patti Spinner
and Paula Winke for agreeing to being part of this project. They have both provided me with
friendly conversation, thoughtful feedback, and words of wisdom over the years.
In my estimation, the Second Language Studies program at Michigan State University
has professors that are not only incredibly talented, but that also care deeply about their students
and go out of their way to help them succeed. I am thankful to not only my committee members,
but also to many other professors who have taken time from their own research to plan
interesting classes, help students, and answer their queries.

	
  

vi

I am also thankful to the native Spanish speakers and Spanish learners that participated in
this study, and would like to acknowledge financial support from a Language Learning
Dissertation Grant and an internal grant from Michigan State University.
So many of my Michigan friends are also very deserving of my gratitude. A special thank
you to Megan Smith for being my café work buddy, sounding board, and official baker; to
Daniel Trego for always lending me a helping hand and being the other half of our “dream
team”; to Luca Giupponi for being my photographer and craft beer connoisseur; to Angelika
Kraemer for always keeping me entertained; and to Walter Hopkins for making every day
brighter with his infectious laugh. My time in Michigan wouldn’t have been the same without
you all.
I am grateful for my colleagues at Princeton University as well. I am particularly thankful
to my compinche, Anais Holgado Lage, for giving me a place to stay and keeping me company,
especially during long commutes. I would also like to thank the rest of the Aprendo team:
Catalina Méndez Vallejo, Adriana Merino, and Sylvia Zetterstrand.
Finally, I’d like to extend a heartfelt thank you to the people I love most – my family.
They have all rallied around me during difficult times, and for that I am incredibly grateful.
Thanks especially to my mother and late father for their patience with me and unwavering
support. My brother and I hit the lottery as far as parents go, and we know it. Thanks also to
Kaitlin Mignella for her longstanding, tried and true friendship. She’s not technically family, but
she might as well be.
Finally, I would like to thank my top-notch husband, Jorge Méndez Seijas. Jorge has
worked diligently with me over the years to make sure that I remain optimistic, independent, and
well-traveled. I am very thankful for his role in shaping me into the person I am today.

	
  

vii

TABLE OF CONTENTS

LIST OF TABLES

xi

LIST OF FIGURES

xiii

CHAPTER 1: INTRODUCTION AND BACKGROUND
The Nature of L2 Competence
Representational vs. Non-Representational Deficit Accounts
Grammatical Gender: A Test Case
The Present Dissertation
Overview of the Dissertation
Definition of Terms

1
1
1
2
3
4
5

CHAPTER 2: MOTIVATION OF THE CURRENT STUDY
Grammatical Gender Assignment and Agreement
Methodological Tools to Assess Gender Agreement
Gender Agreement in Different Syntactic Contexts
Acquisition of Agreement in L1 Spanish
Acquisition of Agreement in L2 Spanish
Summary of L1 and L2 research
Motivation of the Present Dissertation
Research Questions and Predictions

6
6
10
15
17
18
22
22
24

CHAPTER 3: METHOD
Research Design
Participants
L2 Spanish Learners
Native Spanish Speakers
Materials
Eye-tracking Materials
Section of target nouns: A pilot vocabulary test.
Experimental sentences.
Apparatus.
Additional Experimental Materials
Background questionnaire.
Reading questionnaire .
Vocabulary posttest.
Proficiency test.
Procedures
Analysis
Areas of Interest
Fixation Time Measures
Data Cleaning

26
26
26
27
29
30
30
30
36
41
42
42
42
44
44
45
45
45
46
47

	
  

viii

Statistical Analyses

48

CHAPTER 4: RESULTS
Statistical Analyses: Native Speakers’ and L2 Learners’ Reading Times on
Grammatical and Ungrammatical Sentences
Determiner-Noun Agreement
Descriptive statistics.
First fixation duration.
First-pass time.
Go-past time.
Total time.
Summary of results.
Noun-Adjective Agreement
Descriptive statistics.
First fixation duration.
First-pass time.
Go-past time.
Total time.
Summary of results.
Null Nominal Agreement
Descriptive statistics.
First fixation duration.
First-pass time .
Go-past time.
Total time.
Summary of results.
General Summary of Statistical Results
Additional Statistical Analyses: L2 Learners’ Reading Times on Grammatical
and Ungrammatical Sentences by Gender
Descriptive Statistics
First Fixation Duration
First-Pass Time
Go-Past Time
Total Time
General Summary of Statistical Results
Sensitivity to Violations of Gender Agreement on an Individual Basis
Participants’ Sensitivity as Evidenced by Self-Reports
L2 Learners’ Sensitivity on an Individual Basis
Comparison of L2 Learners’ Reported Sensitivity and Individual
Reading Times
Summary of Results

55

CHAPTER 5: GENERAL DISCUSSION
Summary and Discussion of the Findings
Implications of the Findings
Limitations

89
89
96
97

	
  

ix

55
55
55
55
56
57
58
59
59
59
60
61
62
63
64
64
64
65
66
67
68
69
69
71
73
75
76
77
78
79
79
80
83
85
87

Future Research

98

APPENDICES
Appendix A: Pilot Vocabulary Test
Appendix B: Experimental Stimuli
Appendix C: L2 Learner Background Questionnaire
Appendix D: Native Spanish Speaker Background Questionnaire
Appendix E: Vocabulary Posttest
Appendix F: Additional Gender Analyses

100
101
104
108
113
117
118

REFERENCES

123

	
  

x

LIST OF TABLES

Table 2.1

Comparison of Gender Assignment and Agreement in L2 Spanish

7

Table 3.1

Pilot Vocabulary Test Results

33

Table 3.2


Frequency Ranks for Target Words

35

Table 4.1

Mean (Standard Deviation) Fixation Times for DET-N Condition

55

Table 4.2

Model Results for First Fixation Duration in DET-N Condition

56

Table 4.3

Model Results for First-Pass Time in DET-N Condition

57

Table 4.4

Model Results for Go-Past Time in DET-N Condition

58

Table 4.5

Model Results for Total Time in DET-N Condition

59

Table 4.6

Mean (Standard Deviation) Fixation Times for N-ADJ Condition

60

Table 4.7

Model Results for First Fixation Duration in N-ADJ Condition

61

Table 4.8

Model Results for First-Pass Time in N-ADJ Condition

62

Table 4.9

Model Results for Go-Past Time in N-ADJ Condition

63

Table 4.10

Model Results for Total Time in N-ADJ Condition

64

Table 4.11

Mean (Standard Deviation) Fixation Times for N-DROP Condition

65

Table 4.12

Model Results for First Fixation Duration in N-DROP Condition

66

Table 4.13

Model Results for First-Pass Time in N-DROP Condition

67

Table 4.14

Model Results for Go-Past Time in N-DROP Condition

68

Table 4.15

Model Results for Total Time in N-DROP Condition

69

Table 4.16

Summary of Statistical Analyses

70

Table 4.17

Mean (Standard Deviation) Fixation Times for L2 Learners
by Gender of Target Noun

74

	
  

xi

Table 4.18

Gender Model Results for L2 Learners’ First Fixation Duration
in DET-N Condition

76

Gender Model Results for L2 Learners’ First-Pass Time
in DET-N Condition

77

Gender Model Results for L2 Learners’ Go-Past Time
in DET-N Condition

78

Gender Model Results for L2 Learners’ Total Time
in DET-N Condition

79

Participants’ Reported Sensitivity to Violations in Sentence
Processing Task

82

Table 4.23

Individual Change in Reading Time for L2 Spanish Learners

84

Table 4.24

Comparison of Reported Sensitivity and Individual Change in
Reading Time for L2 Spanish Learners

86

Gender Model Results for L2 Learners’ First Fixation Duration
in N-ADJ Condition

117

Gender Model Results for L2 Learners’ First-Pass Time
in N-ADJ Condition

118

Gender Model Results for L2 Learners’ Go-Past Time
in N-ADJ Condition

118

Gender Model Results for L2 Learners’ Total Time
in N-ADJ Condition

119

Gender Model Results for L2 Learners’ First Fixation Duration
in N-DROP Condition

119

Gender Model Results for L2 Learners’ First-Pass Time
in N-DROP Condition

120

Gender Model Results for L2 Learners’ Go-Past Time
in N-DROP Condition

120

Gender Model Results for L2 Learners’ Total Time
in N-DROP Condition

121

Table 4.19
Table 4.20
Table 4.21
Table 4.22

Table 6.1
Table 6.2
Table 6.3
Table 6.4
Table 6.5
Table 6.6
Table 6.7
Table 6.8

	
  

xii

LIST OF FIGURES

Figure 3.1

Distribution of experimental sentences across conditions

39

Figure 3.2

Histogram depicting total time data for the DET-N condition

50

Figure 3.3 


Residual plot from fictitious data that does not violate the assumption of
absence of heteroscedasticity

51

Figure 3.4 


Residual plot depicting total time data for the DET-N condition

51

Figure 3.5 


Histogram depicting the log-transformed total time data for the
DET-N condition

52

Residual plot depicting the log-transformed total time data for the
DET-N condition

52

Participants’ percent change in reading time between grammatical and
ungrammatical sentences

71

Figure 3.6 

Figure 4.1

	
  

xiii

CHAPTER 1: INTRODUCTION AND BACKGROUND
The Nature of L2 Competence
A perennial question in second language acquisition (SLA) is whether non-native
speakers of a language can acquire native-like competence in a second language (L2). The crux
of the debate centers around whether Universal Grammar (UG), which is assumed to guide first
language (L1) acquisition, is still accessible for late L2 learners. While all normally developing
children do ultimately converge on a target-like grammar for their L1, adults learning a second
language often create a steady state grammar that diverges in certain areas from that of adult L1
speakers of the target language.
Representational vs. Non-Representational Deficit Accounts
To account for these differing outcomes, some researchers have argued that L2 learners’
competence is characterized by a representational deficit. Over the years, there have been
multiple proposals in the literature that argue for a representational deficit account, such as the
No Parameter Setting Hypothesis (Clahsen & Muysken, 1989), the Failed Functional Features
Hypothesis (Hawkins & Chan, 1997), and the Interpretability Hypothesis (Tsimpli &
Dimitrakopoulou, 2007). Although the details of these proposals differ, they all claim that L2
learners cannot acquire parameterized functional features that are not instantiated in their L1.
For this reason, the L2 learners’ competence is deficient, and is therefore fundamentally different
from that of native speakers.
This representational deficit view contrasts with other accounts that claim L2 learners
still have access to UG, and can therefore acquire functional features not instantiated in their L1.
This view includes hypotheses such as the Full Transfer/Full Access Hypothesis (Schwartz &
Sprouse, 1996) and the Missing Surface Inflection Hypothesis (MSIH) (Prévost & White, 2000).

	
  

1

While the details of these hypotheses also differ, the underlying argument is the same: L2
learners’ underlying competence is not impaired. That is, because L2 learners have access to UG,
they are able to reset parameters, and ultimately acquire functional features that are not
instantiated in their L1. To the extent that non-targetlike performance occurs, researchers ascribe
it to different task demands such as communication pressure (Prévost & White, 2000, p. 129) or
processing constraints (Hopp, 2010; Keating, 2010).
Grammatical Gender: A Test Case
Research on grammatical gender, a parameterized functional feature, has especially
informed questions examining L2 learners’ competence. At the heart of the problem is whether
learners whose L1 does not have grammatical gender agreement (e.g., English1) can develop a
native-like gender agreement system in an L2 that does have grammatical gender (e.g., Spanish,
French, Dutch, German). While some researchers studying the acquisition of grammatical gender
agreement have found that these learners’ underlying linguistic systems are not native-like (e.g.,
Franceschina, 2001; Franceschina, 2005; Sabourin, Stowe, & de Haan, 2006), others have found
that they are (e.g., Hopp, 2012; Keating, 2009; White, Valenzuela, Kozlowska-Macgregor, &
Leung, 2004).
Gender agreement occurs between a ‘trigger’ (generally a noun) and multiple targets
(e.g., articles, adjectives, pronouns, demonstratives). It may well be the case that not all these
types of agreement are equally difficult for the L2 language learner. That is, learners may
demonstrate differential levels of accuracy in comprehending and producing2 agreement between

	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  
1
While English does mark natural gender on pronouns, it has no agreement (Corbett, 1991).
2
While the goal of these studies is to investigate the status of L2 learners’ underlying linguistic
system, researchers resort to production measures and/or online or offline comprehension
measures to indirectly access the linguistic system. Therefore, even though these researchers are
	
  

2

the trigger and different categories of targets (e.g., articles vs. adjectives). If this is the case, then
conclusions drawn by researchers supporting the representational deficit account may be due to
at least two possible explanations, either: (a) L2 learners have a representational deficit for
grammatical gender agreement or, (b) L2 learners do not have a representational deficit for
grammatical gender agreement, but perform poorly because of the type of agreement under
investigation.
The Present Dissertation
In this dissertation, I will examine L1 English learners’ acquisition of grammatical
gender agreement in L2 Spanish. I will compare three different types of grammatical gender
agreement: Determiner-Noun (DET-N), Noun-Adjective (N-ADJ) and null nominal agreement
(N-DROP). This study will use an online sentence comprehension task to measure participants’
sensitivity to violations of grammatical gender agreement, which will be measured by recording
participants’ eye-movements with an eye tracker. If sensitivity to the violations of grammatical
gender agreement is found, it will be taken as evidence that learners do not have a
representational deficit for morphology. If no such sensitivity is found, it will be taken as
evidence that either (a) the L2 learners have a representational deficit, or (b) the L2 learners do
have target-like representations, but they could not be detected3. Of particular interest in the
present dissertation is whether the three types of agreement previously stated are equally difficult
	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  
ultimately concerned with linguistic competence, they tend to examine linguistic competence via
performance measures.
3
Unfortunately, it is very difficult to adjudicate between these possibilities. While many
researchers conducting these types of studies do take lack of sensitivity to violations of
grammatical gender agreement as an indication of a representational deficit, there always exists
the possibility that the participants do have target-like representations, but the data collection tool
and experimental design employed were not sensitive enough to detect them. Another possibility
is that learners did not perform like native speakers because of communication or processing
pressures (White et al., 2004). 	
  
	
  

3

for the L2 learners, or whether certain types of agreement are less difficult than others. To further
understand how L2 learners process sentences, I will also explore their sensitivity to violations of
grammatical gender agreement in a novel way: through a reading questionnaire administered
after the eye-tracking experiment. Of particular interest is whether the L2 learners report seeing
agreement violations in the experimental stimuli, and if so, whether their self-reports mirror any
evidenced sensitivity in reading times.
Overview of the Dissertation
The organization of this dissertation is as follows: in Chapter 2, I will provide an
overview of the literature and will explain the motivation for this dissertation. I will explain the
nature of gender assignment and agreement, how different methodologies can be used to
investigate grammatical gender agreement, and why I have chosen to investigate how learners
compute agreement in three different syntactic contexts with the eye-tracking methodology. In
Chapter 3 I will describe the method of the study, including the results of a pilot study that was
critical in constructing the stimuli for the primary data collection. In Chapter 4, I will present the
results of the eye-tracking study in two ways: both by statistical tests performed on both
participant groups and individually on a participant-by-participant basis. I will also triangulate
the L2 learners’ reading times with their responses on a reading questionnaire, designed to
measure if they were aware of the agreement violations in the experimental stimuli. Finally, in
Chapter 5, I will discuss the results and the limitations of the current study, as well as propose
future lines of research.

	
  

4

Definition of Terms
Universal Grammar (UG): A construct within generative theory that argues language is innate,
uniquely human, and different from all other types of cognition. UG consists of principles and
parameters that constrain language acquisition.
Linguistic competence: The mental representation of language that is abstract, implicit and
normally not describable in lay terms.
Gender assignment: An inherent lexical feature that, in Spanish, ascribes nouns to one of two
classes: masculine or feminine.
Grammatical gender agreement: A syntactic process by which controller nouns (called
“triggers”) enter an agreement relationship with targets (e.g., articles, adjectives, etc.).
Functional features: Subcategories of functional categories that express information (e.g.,
gender, tense, finiteness).
Input processing: Term that describes how syntactic and grammatical computations are made
during sentence comprehension.

	
  

5

CHAPTER 2: MOTIVATION OF THE CURRENT STUDY
In this chapter I will describe the motivation of the current study. I will begin by
examining the nature of grammatical gender assignment and agreement and will explore the
syntax of gender agreement. Then, I will explore common methodological tools that researchers
have employed to assess language learners’ ability to compute grammatical gender agreement
and will weigh the relative benefits and drawbacks of each. Next, I will make the case that the
syntactic context in which agreement takes place affects the order in which it is acquired both for
L1 and L2 acquisition. Finally, I will explain the rationale for the current study and present the
research questions.
Grammatical Gender Assignment and Agreement
Gender assignment is the term used to refer to an inherent lexical feature on nouns
(Carroll, 1989). Spanish nouns are divided into two classes: masculine or feminine. By and large,
the endings on nouns are transparent: those that end in /–o/ are masculine while those that end in
/–a/ are feminine. However, gender assignment is not always so straightforward: there are
counterexamples to transparent nouns (e.g., problema, ‘problem,’ m., mano, ‘hand,’ f.) and some
nouns are opaque, meaning that the grammatical gender cannot be deduced from the nominal
ending (e.g., cristal, ‘cristal,’ m., nariz, ‘nose,’ f.). It is generally agreed that gender is an
interpretable φ-feature that is lexically determined on the noun (Chomsky, 1995). Because
masculine is considered default and unmarked, Spanish nouns are marked with the gender
feature [±feminine] (Carroll, 1989; Carstens, 2000).
Gender agreement, on the other hand, is manifested on lexical items in a different
syntactic category—such as determiners and adjectives—that bear some relationship to the noun.

	
  

6

Unlike nouns, these lexical items do not carry intrinsic gender; instead, they are marked for
gender based on an agreement relationship with the noun they modify (e.g., el refresco frío,
‘theMASC drinkMASC coldMASC’; la guitarra negra, ‘theFEM guitarFEM blackFEM,’). While gender
assignment is determined lexically, gender agreement is the result of a syntactic operation
whereby the interpretable features on the triggers and the uninterpretable features on the targets
are matched and checked off in the course of derivation. Table 2.1 summarizes the most salient
differences between grammatical gender assignment and agreement.
Table 2.1
Comparison of Gender Assignment and Agreement in L2 Spanish
Gender Assignment
Gender Agreement
-   Interpretable feature on triggers (i.e.,
-   Uninterpretable feature on targets
nouns)
(e.g., determiners, adjectives)
-   Lexically determined
-   Result of syntactic feature-checking
-   Masculine is the default and is
operation
unmarked (feminine is marked)
In example (1) below, I have depicted a syntactic tree for the DP la guitarra negra. As
previously stated, gender is an intrinsic, interpretable feature of Spanish nouns. In example (1),
the determiner (la) and the adjective (negra) both have uninterpretable gender features. Number
is considered a strong feature in Spanish, which triggers the movement of the noun shown
below4. Gender agreement transpires due to feature-checking and gender is checked as a ‘freerider’ to number agreement in Spanish, because it does not prompt the movement (Carstens,
2000). A more detailed description of this syntactic structure can be found in Franceschina
(2001, 2005) and Carstens (2000)5.
	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  
4
	
  In English, number is a weak feature, accounting for the difference in English adjective
placement (i.e., ‘The black guitar’).	
  
5
	
  In the syntactic structure depicted, gender does not project its own phrase; however, that is not
the stance taken by all researchers. For example, Picallo (1991) proposes a GenP and Bernstein
(1993) proposes a Word Marker Phrase (WMP).	
  
	
  

7

(1)  

In research on L2 Spanish, it is often found that L2 learners exhibit a masculine default,
meaning that masculine inflection occurs in feminine contexts (2a) more often than the reverse
(2b).
(2)  a. *la
the FEM

manzana

delicioso

apple FEM

delicious MASC

cuarto

rosada

room MASC

pink FEM

‘the delicious apple’
b. *el
the MASC
‘the pink room’
Because masculine is considered the unmarked or underspecified gender in Spanish, (2a)
is also referred to as an underspecification error. Gender agreement asymmetry has been widely
documented in Spanish (e.g., Franceschina, 2001; McCarthy, 2007; Montrul, Foote, & Perpiñán,

	
  

8

2008; White et al., 2004), as well as other languages (e.g., a feminine default in German Spinner
& Juffs, 2008).
Representational deficit accounts hypothesize that L2 learners cannot acquire
uninterpretable gender features that are not instantiated in their L1. This means that L2 Spanish
learners (L1 = English) are predicted to be able to assign the correct gender to nouns, but not
acquire the uninterpretable features on determiners and adjectives (Franceschina, 2001, 2005).
For that reason, gender agreement and not gender assignment is of primary interest to researchers
investigating representational deficits.
That said, researchers cannot leave the question of gender assignment completely aside.
To determine if L2 learners can successfully compute gender agreement, researchers must
determine whether the language learner has assigned the correct gender to the noun in the first
place. This represents a challenge for L2 researchers, as the interpretable gender features under
investigation are not directly measurable. Researchers have used the agreement relationship
between a determiner and noun as a reflection of gender assignment in both offline posttests
(e.g., White et al., 2004), and online production measures (e.g., Hopp, 2012), on the assumption
that determiners may be the most immediate reflection of a given lexical item’s gender (Carroll,
1989); however, conflating agreement with assignment can make results difficult to interpret.
What is more, researchers may not be consistent with their terminology within a single study.
For example, Grüter, Lew-Williams and Fernald (2012) took participants’ ability to use the
determiner’s gender as a predictive cue in a looking-while-listening task as their ability to
compute agreement in real time; but in their production task, the researchers used DET-N
agreement as evidence that “the speaker correctly classified the noun with regard to its gender
class” (p. 201). Finally, researchers may also not be consistent in maintaining terminology

	
  

9

employed by the authors they cite. For example, Bruhn de Garavito and White (2002)
investigated whether learners exhibited correct gender agreement in DPs involving a determiner
and noun, yet Hopp (2012) interpreted their results as gender assignment (p. 35). In sum,
Montrul et al. (2008) noted that whether a determiner’s gender is related to gender agreement or
assignment is “very hard to tease apart” (p. 510).
In the present dissertation, the relationship between determiners and nouns will be
considered one of gender agreement and not assignment, as that is how the relationship is
generally conceived of in the literature (Carroll, 1989; Carstens, 2000). Assignment will be
measured in a posttest where learners must assign the correct gender (masculine or feminine) to
the nouns under investigation.
Methodological Tools to Assess Gender Agreement
Because the linguistic system cannot be accessed directly, researchers must rely on
participants’ performance on different tasks to examine whether L2 learners have a
representation for features that are not instantiated in their L1. Researchers can select from
production or comprehension tasks, and the comprehension tasks may be either offline or
online6,7. Minimizing task effects is critical in these types of studies, because they can potentially
obscure the researchers’ measurement of the linguistic system. I will review the potential task
effects of production tasks, offline comprehension tasks and online comprehension tasks in turn.

	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  
6
Online tasks are moment-by-moment measures of what participants are doing (e.g., eye
tracking, SPR), while offline tasks are often paper-and-pencil tasks that do not collect momentby-moment data (e.g., GJTs, cloze tests) (VanPatten & Benati, 2010).	
  
7
	
  Production tasks are generally online by nature. It is difficult to imagine what an offline
production task would look like, and, as Grüter et al. (2012) note, it would likely tap
metalinguistic knowledge, thus making it an inappropriate method for examining the linguistic
system.	
  
	
  

10

When researchers administer production tasks (e.g., Bruhn de Garavito & White, 2002;
McCarthy, 2008; Montrul et al., 2008; Prévost & White, 2000; White et al., 2004), they assume
that participants’ percent accuracy during performance is an indirect, yet relatively accurate,
reflection of the state of their linguistic system. If participants’ performance is highly accurate,
then they are assumed not to have a representational deficit. The problem, however, is that if
non-native speakers do not exhibit 100% accuracy on these tasks, the cause of the optionality is
then up to the researcher’s interpretation. Supporters of a representational account attribute any
optionality to the inability to acquire grammatical gender agreement, however non-deficit
accounts, such as the MSIH, interpret the optionality as being a by-product of communication
pressure (Prévost & White, 2000, p. 129). Put differently, MSIH supporters would argue that
even though learners still exhibit optionality, their underlying representations are target-like, and
the optionality stems from task demands. The problem for researchers, then, is deciding exactly
how much optionality can be explained away by communication pressure, and how much is
indicative of a true representational problem. Since no such consensus exists, researchers could
feasibly arrive at opposing conclusions when examining the same set of production data.
To obviate the communication pressure issue inherent in online production tasks,
researchers have coupled these tasks with offline comprehension tasks such as multiple-choice
tasks (e.g., Bruhn de Garavito, 2003), gender recognition tasks where learners must circle the
agreeing determiner or adjective (Montrul et al., 2008), or a picture identification task (e.g.,
Grüter et al., 2012; McCarthy, 2008; Montrul et al., 2008; White et al., 2004). These offline tasks
sidestep the communication pressure issue because participants are not under time pressure to
respond. Consequently, performance accuracy on these offline measures is arguably more
reflective of the linguistic system. The potential problem with these types of tasks, though, is that

	
  

11

their offline nature may invite explicit reasoning and reliance on metalinguistic knowledge.
Competence, by definition, is implicit in nature, so any task that invites metalinguistic reasoning
introduces an intervening variable.
For this reason, some researchers have turned to online comprehension tasks borrowed
from processing and parsing research to tap the linguistic system. The benefit of these online
comprehension tasks is that moment-by-moment data can be collected on how participants are
processing language, which can then be used as an indirect measure of the linguistic system.
There is a large range of methodologies to select from, including eye-tracking (e.g., Keating,
2009; Lim & Christianson, 2014; Sagarra & Ellis, 2013), self-paced reading (e.g., Jiang, 2004,
2007; Jiang, Novokshanova, Masuda, & Wang, 2011; VanPatten, Keating, & Leeser, 2012), selfpaced listening (e.g., Marinis, 2007a, 2007b), ERPs (e.g., Alemán Bañón, Fiorentino, &
Gabriele, 2014; Barber & Carreiras, 2005; Dowens, Vergara, Barber, & Carreiras, 2010;
Morgan-Short, Sanz, Steinhauer, & Ullman, 2010) and neuroimaging (e.g., Christensen, Kizach,
& Nyvad, 2013).
In many online comprehension studies, researchers attempt to tap participants’
underlying competence for different linguistic phenomena by testing their sensitivity to
grammatical violations (e.g., Kreiner, Garrod, & Sturt, 2013; Lim & Christianson, 2014; Sagarra
& Herschensohn, 2010a). In these studies, researchers present participants with sentences
embedded with grammatical violations of the linguistic phenomena under investigation as well
as matched sentences that are properly formed. These target sentences are then distributed among
distractors and/or fillers that may or may not also contain grammatical violations. After reading
an experimental sentence, participants complete a secondary task, such as responding to a
question about what they read. The researcher then compares the participants’ reading times on

	
  

12

the grammatical and ungrammatical regions of the target sentences, and a relatively longer
reading time on the ungrammatical region is taken as indication that the learner was indeed
sensitive to the violation.
Jiang (2007) notes that one of the online methodologies discussed above, SPR, is
particularly useful for testing competence because during production, L2 learners are more likely
to monitor themselves and therefore use explicit knowledge. Because production is not necessary
during SPR experiments, he notes that SPR “thus eliminates the motivation for applying explicit
knowledge” (p. 12). Jiang does not conclude that SPR is a completely implicit measure;
however, he notes that it elicits “little involvement of explicit knowledge” (p. 12). The same
could be true for other online comprehension tasks (e.g., eye-tracking, SPL)8. Jiang is not alone
in positing that these comprehension measures tap primarily implicit knowledge. In fact,
researchers often pair a comprehension task with a separate grammaticality judgement task
(GJT), on the premise that the former measures more implicit knowledge and the latter more
explicit knowledge (e.g., Coughlin & Tremblay, 2013; Roberts & Liszka, 2013; Sagarra &
Herschensohn, 2010b).
Yet the extent to which researchers avoid tapping explicit knowledge during these
sentence comprehension experiments is still open to debate, and is likely dependent on a host of
methodological decisions. One such decision is the type of secondary task participants engage in
after reading each experimental sentence. Some options for secondary tasks include
comprehension questions (e.g., Keating, 2009; VanPatten et al., 2012), plausibility judgements
(e.g., Wen, Miyao, Takeda, Chu, & Schwartz, 2010), grammaticality judgements (Godfroid et

	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  
8
It should be noted, though, that Jiang (2007) ultimately favors SPR to eye tracking because of
the transient nature of the stimuli presentation in the former.
	
  

13

al., 2015), producing a translation of the experimental sentence (Lim & Christianson, 2014), and
assessing the accuracy of a translation of the experimental sentence (Keating, 2009). It seems,
though, that the type of task may affect how learners process the sentences and what type of
knowledge they employ. For example, Leeser, Brandl and Weissglass (2011) tested L2 Spanish
learners to violations of noun-adjective agreement and subject-verb inversion in wh-questions.
Participants completed two SPR experiments: in one, they answered a yes/no comprehension
question after each experimental sentence, and in the other they made grammaticality
judgements. Leeser et al. (2011) found that task type affected participants’ sensitivity to nounadjective agreement violations: they were sensitive when asked to perform grammaticality
judgements, but not when answering comprehension questions. The participants were not
sensitive to violations of subject-verb inversion in either condition. The grammaticality
judgement Leeser et al. (2011) employed was an untimed judgement, but another potential
mediating factor is whether the GJT is timed or untimed, as the former is often posited to
measure implicit knowledge and the latter explicit knowledge (Ellis, 2005; Godfroid et al.,
2015).
In addition to the type of secondary task employed, another methodological decision that
may affect the extent to which participants tap explicit knowledge is percentage of grammatical
violations that appear in the experimental sentences, both in the targets and distractor sentences,
with more grammatical violations likely increasing the likelihood that participants will tap
explicit knowledge. Researchers vary greatly in the percentages of grammatical violations
present in the sentences. For example, Coughlin and Tremblay (2013) report that half of their
distractor sentences contained violations, yet Sagarra and Herschensohn (2010b) report that only
about 13% of their sentences contain violations.

	
  

14

While it is very difficult to gauge whether participants are aware that the experimental
sentences contain violations and are actively looking for them during these online
comprehension tasks, posttest questionnaires could be used to determine how aware participants
were of the violations during the experiment. In this dissertation, I will report findings from a
reading questionnaire that aims to examine precisely this question.
For an online comprehension measure, I selected eye-tracking methodology to examine
participants’ sensitivity to violations of grammatical gender agreement (for reviews of this
methodology see Dussias, 2010; Holmqvist et al., 2011; Keating, 2014; Roberts, 2012). One of
the benefits of eye tracking is that the stimuli can be presented in their entirety rather than on a
word-by-word or segment-by-segment basis. This allows participants to read more naturally in
eye-tracking experiments than other online comprehension measures, such as self-paced reading
(Dussias, 2010; Roberts, 2012; Witzel, Witzel, & Forster, 2012). Another benefit of the eyetracking methodology is that participants do not need to perform a secondary task while reading9
(e.g., pressing a button) which could potentially affect comprehension (Dussias, 2010). Eye
tracking also provides a plethora of fine-grained data that can reflect early and late cognitive
processing, whereas other online processing measures, such as self-paced reading, provide
reaction time data, which are more unidimensional.
Gender Agreement in Different Syntactic Contexts
In Spanish, grammatical gender marking is very prevalent, and can occur between a
trigger (generally a noun) and multiple different targets (e.g., articles, adjectives, pronouns, and
demonstratives). This preponderance of agreement marking provides researchers with many

	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  
9
A secondary task is usually performed after reading each sentence, though (e.g., answering a
comprehension question).
	
  

15

types of agreement that could potentially be investigated (e.g., DET-N, N-ADJ) either during
oral and written production, or during online and offline receptive tasks (oral or written). While
many researchers do provide a theory-based justification for the types of tasks they employ (e.g.,
Grüter et al., 2012; McCarthy, 2008), a rationale for the type of agreement under investigation is
generally lacking.
For example, McCarthy (2008) investigated intermediate and advanced L2 Spanish
learners’ (L1 = English) ability to correctly produce adjective and direct object clitic agreement
in a production task and comprehend direct object clitic agreement on an interpretation task.
McCarthy found that the learners’ performance was variable on both tasks. Because the
optionality extended into comprehension, McCarthy interprets these results as evidence that the
L2 learners have a representational deficit for gender agreement. However, McCarthy also
noticed that many of the advanced learners’ errors were underspecified, thus showing evidence
of a masculine default10. In her discussion, McCarthy cites White et al. (2004), who found that
L2 learners exhibited optionality for gender agreement, but also perfect accuracy on nounadjective word order (thus showing an unimpaired syntax). McCarthy therefore argues that the
deficit for her participants occurs in the morphology, not in the syntax.
It is unclear why McCarthy would decide to examine agreement on direct object clitics
during comprehension, because they are difficult to begin with for Spanish language learners.
Therefore, performance problems could easily be ascribed to the difficulty of the structure, rather
than agreement problems (cf. VanPatten et al., 2012, p. 112). An intriguing question is whether

	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  
10
Supporters of the MSIH interpret these types of errors as evidence of the MSIH by arguing
that the pattern of the errors is evidence of a functioning system. The errors themselves are
ascribed to communication or processing pressures (Prévost & White, 2000; White et al., 2004).
	
  

16

McCarthy would have obtained the same results, had she tested gender agreement in a less
difficult syntactic context.
To be sure, in studies that examine whether L2 learners can acquire uninterpretable
gender features, the syntactic context question is tangential, since participants only need to show
that they have acquired the features when tested on a single type of agreement. It may well be the
case, though, that there is an interaction between the syntactic context where the agreement
transpires and whether participants can compute the agreement. If this were the case, then studies
that examine more difficult types of agreement may conclude that learners have not acquired the
features, when in reality they have, just in another syntactic context. For this reason, it is
essential to determine in which syntactic contexts agreement is the most and least difficult to
compute. I will first explore the literature examining the acquisition of grammatical gender
agreement in Spanish L1 acquisition in different syntactic contexts, and will then turn to L2
studies.
Acquisition of Agreement in L1 Spanish
L1 Spanish speaking children do exhibit optionality in grammatical gender agreement,
although such optionality is relatively infrequent compared to other morphological errors, such
as verbal morphology (Mariscal, 2009).
Agreement on the definite article seems to be acquired before agreement on the indefinite
article, perhaps owing to the former’s higher frequency in the input (Mariscal, 2009). Also DETN agreement is mastered before N-ADJ agreement (Hernández Pina, 1984). López Ornat (1988)
noted that L1 Spanish-speaking children correctly produce DET-N agreement from about 18 to
24 months. Then, at about 24 months, children exhibit variability in DET-N agreement, leading
López Ornat to attribute the previous stage to the production of unanalyzed chunks. Around 30

	
  

17

months, learners begin to exhibit optionality in the marking of (DET-)N-ADJ agreement, and
these errors continue until about roughly 36 months of age.
L1 Spanish speakers also acquire agreement over longer linear distances. In Spanish,
nouns do not always have to be overtly realized. Null nominals (also known as N-drop or
nominal ellipsis) are DPs in which the noun is dropped or omitted (Bernstein, 1993; Snyder,
1995). This occurs when the referent can be recovered through the gender or number on the
determiner of the null nominal. This is shown in example (3):
(3)  

Gladys quiere la pelota blanca pero Maruja quiere la e roja.
Gladys wants the white ball but Maruja wants the red (one).

In example (3), the noun is overt in the first DP (la pelota blanca), however, in the
second DP it has been omitted (la e roja). The gender and number features on the determiner
allow the noun to be recovered from the context. In English, a similar interpretation is achieved
by using the pronoun ‘one.’
Snyder, Senghas and Inman (2001) investigated the acquisition of N-DROP by two
Spanish-speaking children: María and Koki. María mastered agreement on determiners and
adjectives at around 25 months, and began producing N-DROP around the same time as
attributive adjectives. Koki, on the other hand, mastered agreement on determiners and
adjectives long before producing N-DROP. Koki began producing determiners at 19 months, but
she did not produce N-DROP utterances until 30 months. Liceras, Díaz and Mongeon (2000)
also reported comparable findings.
Acquisition of Agreement in L2 Spanish
In L2 research, syntactic context may also play a role in acquisition. For example, in an
elicited production task, Bruhn de Garavito and White (2002) found that L2 Spanish speakers

	
  

18

(L1 = French) were more accurate marking grammatical gender agreement on definite rather
than indefinite DPs, and that agreement errors were more frequent on feminine than masculine
nouns, thus showing evidence of a masculine default. Participants also exhibited more
optionality on N-ADJ agreement (roughly 29-31% total errors, depending on proficiency group)
than DET-N agreement (only 11-18.5% errors) (for similar results in L2 French see Dewaele &
Véronique, 2001), and that optionality on N-ADJ agreement was more frequent with feminine
than masculine nouns.
In a Spanish production task, White et al. (2004) also found that their participants fared
slightly worse on (DET-)N-ADJ agreement than DET-N agreement, especially at lower levels of
proficiency. Although White et al. examined DET-N and (DET-)N-ADJ agreement during
production, for their receptive task, they elected to examine N-DROP. During this task,
participants read a short dialogue that ended with a null nominal, that is, a DP with only a
determiner and an adjective (no noun). They were then presented with images of three nouns that
differed in gender and/or number and had to indicate to what picture the null nominal referred. If
they had acquired gender and number, then it was hypothesized that they would select the correct
picture. If they had not, then they should select pictures randomly. Participants with an advanced
level of Spanish performed quite well on this measure in terms of gender agreement (over 90%
accuracy), but participants at intermediate and beginner proficiency levels performed worse
(about 80-85% accuracy for intermediate learners and 55-65% for beginners11). It is unclear
whether the intermediate and beginner learners performed worse because they did not understand
the syntactic structure of N-DROP, or because they could not compute the syntactic agreement.
Also, since the researchers used agreement in different syntactic contexts to assess L2 learners’
	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  
11
Chance level was 33% accuracy.
	
  

19

ability to compute grammatical gender agreement, it is impossible to disentangle the difficultly
of the grammatical structure from participants’ ability to compute the agreement.
Grüter et al. (2012) also investigated advanced L2 Spanish (L1 = English) learners’
acquisition of grammatical gender agreement by comparing the learners’ performance on an
offline measure of comprehension, an online measure of comprehension and a production task.
For the offline measure of comprehension, the researchers employed the picture-identification
task created by White et al. (2004). Both native Spanish speakers and Spanish learners performed
at ceiling on this task. To measure online comprehension, the authors employed the lookingwhile listening procedure (Fernald, Zangl, Portillo, & Marchman, 2008). Two pictures were
presented to the participants on a screen and they listened to a sentence naming one of the
images. The pictures showed two nouns that were either of the same or different genders, and the
researchers investigated whether participants could use gender as a cue to select the correct
picture. In their oral production task, Grüter et al. elicited DET-N-ADJ and null nominal
structures (DET-ADJ), but collapsed the two for analysis, making it impossible to determine if
participants were equally accurate on the two types of agreement. Grüter et al. found that in the
production task, native and near-native speakers did not differ in gender agreement, but nearnative speakers were statistically less accurate in gender assignment. This is likely because many
of the targeted nouns were not transparent (for a description of the same study materials see
Montrul et al., 2008). Importantly, assignment for the production task was operationalized as
DET-N agreement, “on the assumption that determiner choice is the most immediate reflection
of a noun’s lexical gender” (p. 201, cf. Carroll, 1989). It should be noted though, that for the
looking-while listening task, determiners were used as a predictive cue indicating ability to
compute gender agreement, not assignment.

	
  

20

Unlike the previous studies, Franceschina (2001) investigated gender agreement in
multiple syntactic contexts in a single task. In this study, oral production data was collected from
Martin, a L2 Spanish learner (L1 = English) who lived in a Spanish immersion context for 24
years. Martin’s accuracy for gender agreement depended heavily on syntactic context:
demonstrative (100%), pronoun (98%), article (94%), adjective (77%). He also showed evidence
of masculine defaults in his production. This non-native-like performance led Franceschina to
conclude that Martin evidenced a syntactic deficit.
In a more complex and comprehensive study, Franceschina (2005) investigated the
performance of near-native L2 Spanish learners from L1s with and without grammatical gender
(henceforth +gen and –gen, respectively). Franceschina collected naturalistic production data as
well as results from six experimental tasks: five that tested gender agreement and one that tested
assignment. In the naturalistic production data, Franceschina found that the native speakers and
L1 +gen participants performed at ceiling for gender agreement on determiners, adjectives and
pronouns, but the L1 –gen participants fared worse, with 93% accuracy on determiners, 90%
accuracy on adjectives and 87% percent accuracy on pronouns. The five experimental tasks that
tested agreement were: (1) a guessing game, (2) a missing word task, (3) a cloze/multiple choice
task, (4) a GJT, and (5) a novel word task. Test 1 investigated nouns, adjectives and pronouns,
test 2 investigated clitic pronouns, test 3 investigated nouns and adjectives, test 4 investigated all
categories marked for gender and test 5 investigated agreement on adjectives and pronouns. In
the battery of tests, the –gen participants consistently performed worse than the +gen
participants, leading Franceschina to conclude that they had a representational deficit for gender
agreement. Tests 4 and 5 were production tasks, tests 1 and 3 were comprehension tests, and test
2 required both production and comprehension; however, it is worth noting that all the

	
  

21

comprehension tests were offline tests. Even though the L2 learners had naturalistic exposure to
the L2, they still may have used their knowledge of grammatical rules to complete the offline
comprehension tasks.
Summary of L1 and L2 Research
In sum, it seems that there are some parallels between the acquisition of grammatical
gender agreement in L1 and L2 Spanish. First, definite articles seem to exhibit less optionality
than indefinite articles for both groups. Second, DET-N agreement seems to be easier to compute
than N-ADJ agreement. It also seems that L1 Spanish learners acquire agreement in overt
constructions before being able to produce N-DROP.
In many of the L2 studies mentioned, syntactic context is a confounding variable. For
example, White et al. (2004) tested DET-N and N-ADJ agreement during production, but NDROP during an offline comprehension task with a picture-identification task. While this task
should be praised for its ingenuity, it is unclear why the researchers decided to examine different
types of agreement in the productive and receptive mode12. Even though Franceschina did
investigate agreement of varying syntactic contexts in a single task, these tasks were either
production tasks (2001, 2005) or offline compression tasks (2005), leaving open the question of
how L2 learners would perform on an online comprehension task.
Motivation of the Present Dissertation
Researchers investigating L2 grammatical gender agreement oftentimes do not seem to
have a principled reason for the types of grammatical gender agreement that they investigate, and

	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  
12
	
  Although N-DROP is like DET-N and N-ADJ agreement in the sense that a target adjective
must agree with a trigger noun, they are different in that the noun is not overtly realized within
the null nominal. Instead, the speaker/listener must rely on discourse factors to recover the
gender and number features of the null noun.	
  
	
  

22

many seem to assume that agreement in all syntactic contexts is the same in difficulty for L2
learners; however, given the results of studies that investigate L1 acquisition, and the work of
Franceschina (2001, 2005), that is likely not the case. In many studies that investigate whether
L2 learners have a representational deficit for grammatical gender agreement, it is impossible,
then, to conclude whether learners truly do have a representational deficit, or whether they are
performing poorly because of the agreement type under investigation. Therefore, to tease apart
the effects of the syntactic context of agreement, various types of agreement must be tested in a
single task.
To this end, I assess three types of agreement to determine if they are equally difficult for
L2 Spanish learners. The three types of agreement I selected are DET-N, N-ADJ and N-DROP
for comparability purposes, as they seem to be the three frequent types of agreement studied in
this line of research (e.g., Grüter et al., 2012; White et al., 2004). Judging by the L1 and L2
research previously reviewed, it seems that in L1 Spanish learners acquire agreement in the
following order:
1.   DET-N agreement
2.   N-ADJ agreement
3.   N-DROP agreement
L2 learners seem to generally be more accurate on DET-N agreement than N-ADJ agreement
(Bruhn de Garavito & White, 2002; Franceschina, 2001). Although task type or the coding of
data area often confounding variables in L2 studies, if L2 learners acquire agreement as L1
learners do (Snyder et al., 2001), then they will likely be more accurate on N-ADJ agreement
than N-DROP.

	
  

23

In this dissertation, participants’ sensitivity to violations of grammatical gender
agreement in each of the three conditions was tested in a reading comprehension task. As they
read, participants’ eye movements were recorded with eye-tracking methodology. Longer
reading times on violations of gender agreement relative to matched grammatical regions was
assumed to reflect a processing cost, which was taken as indirect evidence that the participants
did not have a representational deficit for grammatical gender agreement. Participants were also
given a reading questionnaire after finishing the eye-tracking experiment, to determine if they
were aware of the gender agreement violations. This task was employed to determine what type
of strategies L2 learners are using when processing sentences that contain violations.
Research Questions and Predictions
The following research questions will guide this dissertation:
1.   Are native Spanish speakers and L2 learners sensitive to violations of DET-N, N-ADJ
and N-DROP gender agreement, as evidenced by their reading times during an online
processing task?
2.   Is sensitivity to the grammatical violations contingent on the type of agreement under
investigation?
3.   Do native Spanish speakers and L2 learners report sensitivity to the violations in the
experimental stimuli?
I predict that native Spanish speakers will be sensitive to violations of grammatical
gender agreement in all three conditions. I also predict that the L2 learners will be sensitive to
violations of grammatical gender agreement, but not necessarily in all three conditions. This
could mean that the L2 learners would not have a representational deficit for gender agreement
(since they are indeed sensitive to violations in at least one condition), but that their sensitivity to

	
  

24

the violation is mediated by the agreement condition under consideration. Give the previous
literature in L1 and L2 acquisition of grammatical gender agreement, I predict that L2 learners
will show higher rates of sensitivity to violations of DET-N agreement, followed by N-ADJ
agreement and then N-DROP agreement. In the analyses, I will investigate whether this is true
both for the L2 learner group as a whole, and also for each individual participant. As for research
question 3, I predict that native Spanish speakers will report sensitivity to the violations of
grammatical gender agreement, but the L2 learners will likely only report such sensitivity if they
also evidence sensitivity in the reading time data.

	
  

25

CHAPTER 3: METHOD
This chapter describes the method used to answer the research questions presented in
Chapter 2. Below, I provide details on the research design, participants, materials, procedures,
and analyses employed in the present dissertation.
Research Design
Eye-tracking methodology was used to determine if participants were sensitive to
violations of grammatical gender agreement. Sensitivity to these violations was operationalized
as a relatively longer reading time on ungrammatical regions involving grammatical gender
agreement relative to matched grammatical regions. This increased reading time was assumed to
reflect a processing cost, and was taken as indirect evidence that the participants did not have a
representational deficit for grammatical gender agreement, provided that an increased reading
time was also found in a control group of native Spanish speakers (cf. Sagarra & Herschensohn,
2010a; VanPatten et al., 2012).
In the present dissertation, there were three independent variables and four dependent
variables. The independent variables were group (native Spanish speaker, L2 Spanish learner),
condition (DET-N, N-ADJ, N-DROP), and grammaticality (grammatical, ungrammatical). The
dependent variables were first fixation duration, first-pass time, go past-time, and total time. A
description of the dependent variables is located under analysis.
Participants
Sixty people participated in the current study. Participants were either L2 Spanish
learners or native Spanish speakers. All participants were recruited from the Michigan State
University (MSU) community.

	
  

26

L2 Spanish Learners
The L2 Spanish learners (n = 29) were notified of the opportunity to participate in the
study via three possible routes: (1) an email sent by their undergraduate advisor, (2) an
announcement made in one of their classes, or (3) word of mouth. Only Spanish majors that were
close to completing their degree at MSU were invited to participate in the study13.
In order to be retained for analysis, the L2 Spanish learners had to: (a) be at least 18 years
old, (b) be a native speaker of English, (c) be a Spanish major at MSU, (d) be close to
completing their major (i.e., have no more than three classes left), (e) have studied no other
language with grammatical gender agreement for more than two years, (f) demonstrate they
understood the experimental sentences by scoring at least 80% on the comprehension
questions14, (g) have normal or corrected vision, and (h) consistently pass calibration during the
eye-tracking experiment.
A total of four L2 learners were eliminated from the analyses for having a native
language other than English (n = 1), for studying another language with grammatical gender
(French) for more than two years (n = 1), and for not being close enough to completing their
Spanish major (n = 2). This yielded a total of 25 L2 Spanish learners retained for analysis.
The L2 learners were on average 21.56 years old (SD = 1.23). Nineteen were female and
six were male. Some participants were majoring only in Spanish (n = 7), while others were dual
majors in both Spanish and another subject (n = 18). Some had studied other languages (e.g.,
French, Japanese, Hindi), but no participant had studied any language with grammatical gender

	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  
13
Most L2 learners were seniors, but some were juniors that had taken as many Spanish classes
as seniors.
14
A description of these comprehension questions is provided in the experimental sentences
section below.	
  
	
  

27

agreement for more than two years. Spanish majors at MSU must take a minimum of 12 Spanish
classes (36 credits) that count towards the major. The majors that participated in this study still
had an average of .94 classes left to take (SD = 1.24, range: 0 - 3) before completing the major.
On average, participants began learning Spanish at age 12.32 (SD = 3.11) and reported actively
studying Spanish for an average of 8.84 years (SD = 3.12). On a scale of 0 (none) to 10 (very
good), participants rated their level of proficiency speaking (M = 6.44, SD = 1.78), understanding
spoken language (M = 7.20, SD = 1.78), and reading (M = 7.12, SD = 1.48). Participants
estimated that in a given week they spent 3.92 hours (SD = 5.48) speaking Spanish, 5.36 hours
(SD = 5.81) listening to Spanish, 3.12 hours (SD = 4.81) writing in Spanish and 5.08 hours (SD =
6.21) reading in Spanish.
Many L2 learners had also studied abroad in a Spanish-speaking country (n = 21).
Spanish majors at MSU are highly encouraged to study abroad for at least eight weeks. Spanish
majors who do not study abroad for eight weeks must either (a) complete an internship
experience in a Spanish-speaking environment, (b) complete a service learning experience in a
Spanish-speaking environment, or (c) enroll in an additional class. In the current study, 15
participants had studied abroad, 3 studied abroad and completed an internship, 3 studied abroad
and completed a service learning experience, 1 completed only an internship, 1 completed only a
service learning experience and 2 had not done either of these activities15. Those that studied
abroad studied in Spain, Ecuador or Peru for an average of 13.06 weeks (SD = 4.53, range: 8.0021.70).

	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  
15
These two students presumably took an additional class.
	
  

28

Native Spanish Speakers
The native Spanish speakers (n = 31) were notified of the opportunity to participate in the
study via three possible routes: (1) an email sent to native Spanish speakers on the MSU campus,
(2) flyers posted around campus, or (3) word of mouth.
To be retained for analysis, the native Spanish speakers had to: (a) be at least 18 years
old, (b) be a native speaker of Spanish, (c) be born in a Spanish-speaking country, (d) immigrate
to the United States at or after age 16, (e) identify as Spanish-dominant, (f) demonstrate they
understood the experimental sentences by scoring at least 80% on the comprehension questions,
(g) have normal or corrected vision, and (h) consistently pass calibration during the eye-tracking
experiment.
A total of four native Spanish speakers were eliminated from the analyses for not
reporting Spanish as their dominant language (n = 2), not having normal or corrected vision (n =
1), and scoring below 80% on the comprehension questions during the reading portion of the
experiment (n = 1). This yielded a total of 27 native Spanish speakers that were retained for
analysis.
The native Spanish speakers were on average 25.59 years old (SD = 7.74). Fourteen were
female and thirteen were male. They were born in various Spanish-speaking countries16:
Argentina, Colombia, Costa Rica, the Dominican Republic, Guatemala, Mexico, Panama, Peru,
Puerto Rico, Spain, and Venezuela. On average, they immigrated to the United States at age
22.41 (SD = 6.51) and had spent a total of 3.02 years in the United States (SD = 3.05). All but

	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  
16
Unlike some aspects of Spanish (e.g., subjunctive vs. indicative distribution), grammatical
gender agreement is not variable across dialects.
	
  

29

two had attended at least some college, and nine were pursuing advanced degrees (either
master’s or doctoral degrees).
Materials
Participants completed five tasks: they first read sentences while their eye movements
were recorded with an eye-tracker, and then also completed a reading questionnaire, vocabulary
posttest, background questionnaire, and proficiency test. The creation of the materials for each
of these tasks is described below.
Eye-tracking Materials
The eye-tracking materials consisted of 60 critical experimental sentences that were used
to investigate the research questions stated in Chapter 2. In this section I describe how I selected
the target nouns for inclusion in the experimental sentences, how the experimental sentences
were constructed, and the apparatus used to record participants’ eye movements.
Section of target nouns: A pilot vocabulary test. The target nouns used in the current
study were selected based on the results of a pilot vocabulary test. The pilot vocabulary test was
administered to Spanish 310 students at MSU through Survey Gizmo, an online survey tool.
Spanish 310 is a grammar class, and it is the “gateway” course that all students must take before
taking courses in the major or minor sequence at MSU. I decided to collect the pilot vocabulary
data from these participants because they are generally of a lower proficiency than the Spanish
majors who participated in the eye-tracking experiment. In this way, the Spanish 310 students’
performance on the vocabulary test yielded a conservative estimate of the vocabulary knowledge
of the participants in the primary data collection. Students were made aware of the survey in
their SPA 310 class, and took this survey outside of class, during their free time.

	
  

30

For this pilot vocabulary test, participants were given a list of 32 nouns: 16 masculine and
16 feminine. I judged these words to be nouns that are commonly taught in basic Spanish
language textbooks and/or are common in classroom discourse (e.g., ensayo ‘essay’).
For each of the 32 words, participants were first asked to translate the word into English,
and then rate their knowledge of the word with the following scale:
•   4 − I know this word very well; I translated it correctly and rapidly.
•   3 − I know this word somewhat well; I translated it correctly after some thought.
•   2 − I'm unsure of this word; I'm unsure if my translation is correct
•   1 − I don't think I know this word; I don't think my translation is correct.
•   0 − I definitely don't know this word; My translation is definitely incorrect.
This scale was selected because it is familiar to these participants, as students at MSU are
graded on a scale such as this one. After the participants translated the 32 words and rated their
knowledge, they were asked to identify the gender of each word. The 32 words were always
presented in isolation (i.e., without a sentential context), but participants were told that the words
were nouns. The presentation of the words was randomized for each participant, and the words in
the translation and rating activity were broken up into four blocks of eight words, to reduce
participant fatigue. The pilot vocabulary test is presented in Appendix A.
The design of the pilot vocabulary test ensured that the participants in the primary data
collection were likely to (a) know the gender of the target nouns and, (b) be familiar with their
translations. The former is important because learners’ sensitivity to the agreement violations
hinges on their ability to assign the correct gender to the target noun. The latter is important
because word familiarity affects processing (Clifton, Staub, & Rayner, 2007; Rayner, 1998),
which could potentially introduce a confounding variable into the reading time measures.

	
  

31

A total of 42 Spanish learners completed the pilot vocabulary test. Four participants
reported a native language other than English and were removed. This left a total of 38
participants retained for analyses. The participants had taken an average of 2.21 semesters of
Spanish classes at MSU (SD = 1.07) and had been actively studying Spanish for an average of
6.96 years (SD = 3.71).
I first calculated the average knowledge rating and percentage of correct gender
assignment across participants for each noun. I then coded the translations for each noun into
English as correct or incorrect. Because this coding required a small degree of interpretation, a
native Spanish speaker who is also highly proficient in English coded the translations as well.
Inter-rater agreement was high − 99.18%. All disagreements between raters were discussed and
resolved. The results of the pilot vocabulary test are presented in Table 3.1.
Participants generally performed well on the gender assignment task, assigning the
correct gender on average 99.84% of the time for masculine nouns (SD = 1.01; range: 93.75100.00) and 98.5% of the time for feminine nouns (SD = 3.96; range: 81.25-100.00). Participants
were less accurate translating the words correctly into English. On average, they correctly
translated masculine nouns 89.47% of the time (SD = 10.59; range: 62.50-100.00), and feminine
nouns 86.68% of the time (SD = 11.64; range: 56.25-100.00).

	
  

32

Table 3.1
Pilot Vocabulary Test Results
English
Correct
Knowledge Rating
Correct Gender
Equivalent Translation
Assignment
(%)
Mean
SD
(%)
Masculine
zapato
‘shoe’
100.0
3.95
0.23
97.4
Nouns
sombrero
‘hat’
100.0
3.92
0.40
100.0
almuerzo
‘lunch’
100.0
3.89
0.39
100.0
refresco
‘drink’
100.0
3.61
0.68
100.0
trabajo
‘work’
97.4
3.95
0.23
100.0
mercado
‘market’
94.7
3.71
0.84
100.0
ensayo
‘essay’
92.1
3.76
0.59
100.0
dibujo
‘drawing’
92.1
3.63
0.91
100.0
archivo
‘file’
92.1
2.37
1.26
100.0
regalo
‘gift’
89.5
3.79
0.66
100.0
vestido
‘dress’
89.5
3.76
0.59
100.0
partido
‘game’
86.8
3.68
0.81
100.0
espejo
‘mirror’
84.2
3.16
1.22
100.0
cuchillo
‘knife’
78.9
3.13
1.02
100.0
abrigo
‘coat’
68.4
2.58
1.62
100.0
cuaderno ‘notebook’
65.8
3.42
1.03
100.0
Feminine
comida
‘food’
100.0
3.97
0.16
100.0
Nouns
escuela
‘school’
100.0
3.97
0.16
100.0
ventana
‘window’
100.0
3.95
0.23
100.0
bebida
‘drink’
100.0
3.82
0.61
100.0
iglesia
‘church’
97.4
3.97
0.16
97.4
pregunta ‘question’
97.4
3.97
0.16
97.4
camisa
‘shirt’
97.4
3.79
0.53
100.0
manzana
‘apple’
92.1
3.97
0.16
100.0
cocina
‘kitchen’
92.1
3.95
0.23
100.0
piscina
‘pool’
92.1
3.87
0.41
100.0
mochila ‘backpack’
92.1
3.63
1.03
94.7
pintura
‘painting’
78.9
3.55
1.01
97.4
cerveza
‘beer’
76.3
3.63
0.91
100.0
corbata
‘tie’
63.2
2.45
1.67
92.1
revista ‘magazine’
60.5
3.03
1.17
100.0
maleta
‘suitcase’
47.4
2.45
1.30
97.4
Note. Results are presented in descending order in terms of correct translation percentage and
then mean knowledge rating.

	
  

33

Because the gender assignment scores were all high, the target words were selected based
on the translation accuracy as well as the knowledge rating. The percentage of correct
translations was used as the primary determining factor as to whether a word should be included
as a target word in the primary data collection. The knowledge rating was the secondary
determining factor. The goal of the pilot vocabulary knowledge study was to select the 10
masculine and 10 feminine nouns that were best known to the participants; however, since one of
the words in the top 10 (archivo, ‘file’) had a relatively high translation score (92.1%) but a low
average knowledge score (2.37), only the first eight masculine and feminine words were selected
for inclusion in the main study. The masculine words were zapato ‘shoe’, sombrero ‘hat’,
almuerzo ‘lunch’, refresco ‘drink’, trabajo ‘work’, mercado ‘market’, ensayo ‘essay’, and dibujo
‘drawing’. The feminine words were comida ‘food’, escuela ‘school’, ventana ‘window’, bebida
‘drink’, iglesia ‘church’, pregunta ‘question’, camisa ‘shirt’, and manzana ‘apple’.
Cognates were selected for the remaining 4 words (2 masculine and 2 feminine) to ensure
that participants would be familiar with them. The masculine cognates were momento ‘moment’
and proyecto ‘project’, and the feminine cognates were guitarra ‘guitar’ and cámara ‘camera’.
Because lexical access is affected by cognate status (e.g., Dufour & Kroll, 1995; van Hell & de
Groot, 1998), these cognates were distributed carefully across different lists (see experimental
sentences section below) to wash out the potential influence cognate status may have on reading
times.
In sum, a total of 20 nouns (10 masculine, 10 feminine) were selected for the primary
data collection. All the nouns were three syllable nouns (6-8 letters in length) and had transparent
/-o/ (masculine) or /-a/ (feminine) endings.

	
  

34

I conducted an a posteriori investigation of the frequency of the target nouns using
Davies’ frequency dictionary (2006). This frequency dictionary includes the 5,000 most common
words in the Spanish language, drawn from a 20,000,000-word corpus. Approximately twothirds of this corpus is based on written texts (literary and non-literary) and one-third is based on
spoken Spanish.
A vocabulary test was used to select the target words instead of frequency rank
calculations because the latter may not accurately reflect the majority of the input L2 learners
receive. As a case in point, according to Davies (2006) the verbs bañar ‘to bathe’ and cenar ‘to
eat dinner’ are roughly as frequent (ranked 3224 and 3261, respectively) as the verbs rozar ‘to
touch lightly’ and alentar ‘to encourage’ (ranked 3226 and 3189, respectively). An intermediate
L2 Spanish learner, however, would likely know only the first two. The frequency ranks of the
target nouns in the current study are presented in ascending order in Table 3.2 below.
Table 3.2
Frequency Ranks for Target Words
Masculine Frequency Feminine Frequency
Nouns
Rank
Nouns
Rank
momento
108
pregunta
481
trabajo
145
escuela
532
proyecto
604
comida
873
mercado
609
iglesia
1111
dibujo
1692
cámara
1172
ensayo
1835
ventana
1265
zapato
1932
camisa
2443
sombrero
2899
bebida
2828
almuerzo
3104
manzana
2853
refresco
--guitarra
2885
Most of the target nouns were ranked within the 3,000 most frequent Spanish words. One
word, refresco ‘drink’ was not listed in the frequency dictionary, as its frequency did not fall
within the 5,000 most common words in the Spanish language. This word was retained as a

	
  

35

target word, however, because (a) 100% of the students translated it correctly on the pilot
vocabulary test, and (b) research with similar populations has also found it to be a well-known
word (Keating, 2005).
The overall average frequency rank for the selected target nouns (not including refresco)
was 1545.82 (SD = 1034.96). The average rank for the masculine nouns was 1436.44 (SD =
1129.16) and the average for the feminine nouns was 1644.30 (SD = 993.09).
Experimental sentences. I created 60 NPs, 20 for each condition. For the DET-N
condition, the NP consisted of a definite article and the target noun. For the N-ADJ condition,
the NP consisted of a definite article followed by a noun and then a modifying adjective. In the
N-DROP condition, the NP consisted of a definite article and modifying adjective. In all three
conditions, the target nouns were preceded by definite articles (el and la) rather than indefinite
articles (un and una) for two reasons. First, masculine and feminine definite articles contain the
same number of letters, which makes their reading times more comparable. Second, learners are
more accurate at providing correct agreement on definite than indefinite articles (Bruhn de
Garavito & White, 2002), so definite articles were used to ‘bias learners for the best.’ Each of the
20 target words appeared once in each of the three conditions (i.e., DET-N, N-ADJ, and NDROP) to lessen the likelihood that word knowledge could mediate sensitivity to violations of
grammatical gender agreement across conditions.
In the N-ADJ and null nominal conditions, adjectives with transparent /-o/ and /-a/
endings agreed with the target noun. These adjectives were all 2-3 syllables long, were
comprised of 4-7 letters, and were not cognates. Twenty unique adjectives were used. They are
barato ‘cheap’, bello ‘lovely’, blanco ‘white’, bonito ‘pretty’, bueno ‘good’, caro ‘expensive’,
corto ‘short’, frío ‘cold’, largo ‘long’, limpio ‘clean’, lindo ‘beautiful’, malo ‘bad’, negro

	
  

36

‘black’, nuevo ‘new’, pequeño ‘small’, rojo ‘red’, rosado ‘pink’, sucio ‘dirty’, tonto ‘silly’, and
viejo ‘old’. According to Davies’ (2006) frequency dictionary, all 20 of these adjectives were in
the top 5,000 most frequent Spanish words, with an average frequency rank of 1247.55 (SD =
1143.86; range: 99 - 4661). In the interest of uniformity, the same 20 adjectives were used twice:
once in the N-ADJ condition and again in the N-DROP condition; however, the adjective
modified different nouns in both conditions.
Examples of DET-N (4), N-ADJ (5), and N-DROP (6) experimental sentences are
presented below. A complete list of target sentences for both masculine and feminine nouns is
presented in Appendix B.
(4)

El chico bebe el/*la refresco cuando ve la película en el cine con su familia.
‘The boy drinks theMASC/*theFEM drinkMASC when (he) watches the movie in the
theater with his family.’

(5)

El atleta toma el refresco frío/*fría cuando termina de correr tres millas por la
mañana.
‘The athlete drinks the drinkMASC coldMASC/*coldFEM when (he) finishes running
three miles in the morning.’

(6)

El jefe pide el refresco grande y su empleado pide el pequeño/*la pequeña
cuando van a McDonald’s.
The boss orders theMASC drinkMASC big and his employee orders theMASC
smallMASC/*theFEM smallFEM (one) when (they) go to McDonald’s.

All 60 experimental sentences began with a two-word singular NP (e.g., the child, the
mother), which was the subject of the sentence. Half of these subjects were masculine and half

	
  

37

were feminine. The subjects were followed by a 1 to 3-syllable singular verb in the present tense
and then a direct object.
In the DET-N and N-ADJ conditions, the direct object was an NP with the determiner
and target noun (and adjective in the case of the N-ADJ condition) described above. These NPs
were followed by the words cuando ‘when’ or durante ‘during’ and then 6-10 more words to
complete the sentence.
In the null nominal condition, the direct object was an NP with a definite article, the
target noun and an opaque modifying adjective17 (e.g., el refresco grande in example 5). The
adjective was then followed by a coordinating conjunction (either y ‘and’ or pero ‘but’), another
two-word singular NP (half masculine, half feminine) that served as the subject of the
coordinating clause, and another 1 to 3-syllable verb in the present tense. This verb was followed
by the target NP (i.e., the null nominal, consisting of a determiner, covertly realized target noun
and modifying adjective) which was followed by the words cuando ‘when’ or durante ‘during’
and then 1 to 3 more words to complete the sentence. The linear distance between the target noun
(e.g., el refresco in example 5) and the null nominal (e.g., el pequeño) was always 5-6 words
(between 10-12 syllables) long. All experimental sentences were between 13 and 16 words long.
Each of the 20 experimental sentences contained the same 10 masculine target nouns, and
10 feminine target nouns. There were two versions of each experimental sentence, one
grammatical, and one ungrammatical. Participants only saw one version of each sentence. For
each of the 3 conditions, participants read 10 grammatical sentences and 10 ungrammatical
sentences. This distribution of target nouns is presented in Figure 3.1.

	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  
17
An opaque adjective was used to ensure that the learners must process the gender of the target
noun rather than the subsequent adjective to be able to match the features to the null nominal.
	
  

38

Figure 3.1
Distribution of experimental sentences across conditions
G = grammatical sentences; U = ungrammatical sentences.
Because reading times on the grammatical and ungrammatical sentences were compared
for analyses, when determining which grammatical and ungrammatical sentences each
participant read, I paid special attention to the gender of the target noun, the cognate status of the
target noun, and the number of letters in each target noun and adjective to ensure that the
grammatical and ungrammatical sentences were as comparable as possible.
The experimental sentences were divided across two lists, so that participant A would see
the grammatical versions of sentences 1-5 and the ungrammatical versions of sentences 6-10 and
participant B would see the reverse. Figure 3.1 shows the distribution of sentences for a
hypothetical participant A. The 60 critical items in both lists were interspersed among 88

	
  

39

distractors18. The two presentation lists were randomized twice for a total of four lists, so that
participants read the experimental sentences in one of four orders. Participants were assigned
randomly to one of the four lists.
As previously mentioned, target words were repeated throughout the experimental
stimuli. That is, the 20 target nouns were recycled across all three conditions, and the 20 target
adjectives were also recycled across the N-ADJ and N-DROP conditions. Because repetitions of
these words may increase familiarity and ultimately affect processing (Reichle, Pollatsek, Fisher,
& Rayner, 1998), the lists were also pseudo-randomized so that no two of the same target nouns
or adjectives appeared within 15 experimental sentences of each other.
Each experimental sentence was followed by a comprehension question to ensure that
participants were focused on the meaning of the sentences while reading. For example, the
comprehension question to (4), reprinted below, is the following (7):
(7) a.

El chico bebe el/*la refresco cuando ve la película en el cine con su familia.
‘The boy drinks theMASC/*theFEM drinkMASC when (he) watches the movie in the
theater with his family.’

b. ¿El chico está solo?
A: Sí

B: No

‘Is the boy alone?’
A: Yes

B: No

	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  
18
The distractors tested linguistic phenomena that were not directly related to the present
dissertation (adverb placement, tense morphology, subject-verb agreement, and adjective
placement). Half the distractors the participants read contained ungrammaticalities.
	
  

40

None of the questions tested the participants’ comprehension of the noun or adjective.
Half of the comprehension question required “yes” answers and the other half required “no”
answers.
Previous sentence processing research has not reached a consensus as to how many
comprehension questions participants must respond to correctly to be retained for analysis. For
example, researchers have set cutoffs of 60% (Sagarra & Herschensohn, 2010a, 2010b), 63%
(Jiang, 2004), 75% (Jiang et al., 2011), 80% (Jiang, 2007), and 85% (Lim & Christianson, 2014).
A relatively conservative cutoff point of 80% was selected for the current study.
Apparatus. The data were collected on an EyeLink 1000 eye-tracker with a desktop
tower mount. Participants rested their head in between a chin rest and a forehead rest while
reading. Participants were seated approximately 60 cm from a 20-inch computer screen while
completing the reading portion of the experiment. The experimental sentences were presented on
a single line of text in size 18 Calibri font. Sentences were presented in black upper- and lowercase letters on a white background. The stimuli were divided into 6 blocks (25 sentences in the
first four blocks, 24 sentences in the last two), so that participants could take five breaks during
the experiment, one after each block. A 9-point calibration was performed at the beginning of the
experiment and after each break. Drift correction was performed before each experimental
sentence. Participants progressed through the experimental sentences and responded to
comprehension questions by clicking buttons on a hand-held controller. The experiment was
written with Experiment Builder software (SR Research Ltd.).

	
  

41

Additional Experimental Materials
In addition to the eye-tracking task, participants also completed a background
questionnaire, reading questionnaire, vocabulary posttest and proficiency test. I describe the
creation of these materials below.
Background questionnaire. Both the L2 Spanish learners and the native Spanish
speakers completed a background questionnaire. Both questionnaires were modified versions of
the Language Experience and Proficiency Questionnaire (LEAP-Q), a language background
questionnaire designed specifically for bilinguals (see Marian, Blumenfeld, & Kaushanskaya,
2007). The LEAP-Q is a validated survey used for examining the experience and proficiency of
bilinguals by means of self-report data. I modified the LEAP-Q to include questions specific to
the participants of this study (e.g., the Spanish classes the L2 learners had taken at Michigan
State University) and to reduce its length. The L2 learners’ background questionnaire is located
in Appendix C, and the native Spanish speakers’ background questionnaire is located in
Appendix D.
Reading questionnaire. To determine whether participants were aware that some of the
sentences in the reading experiment contained errors, they completed a short questionnaire
immediately after finishing the reading portion of the experiment. This reading questionnaire
contained five short questions to gauge participants’ awareness of the errors. The questions are
listed below:
1.   Did you notice anything strange about the sentences you read during the eye-tracking
experiment? If so, what?

	
  

42

2.   Where there any grammatical errors in the sentences you read during the eye-tracking
experiment?
Options:
a.   Yes
b.   No
3.   What types of grammatical errors did you notice? Please list all the errors you
remember, and provide examples when possible.
4.   Please check off all the types of errors you noticed in the sentences. If you are unsure
as to what something is, please ask the researcher:
Options:
a.   ADVERBS appeared in the wrong place in the sentence
b.   incorrect tense (present, past, etc.) was used
c.   incorrect gender agreement between articles (e.g., el, la) and nouns
d.   ADJECTIVES appeared in the wrong place in the sentence
e.   incorrect agreement between subjects and verbs
f.   subjunctive was used incorrectly
g.   por and para were used incorrectly
h.   ser and estar were used incorrectly
i.   incorrect gender agreement between nouns and adjectives
j.   incorrect gender agreement between nouns and null nominals (e.g., el rico, la
cómica)
5.   What percentage of the experimental sentences (EXCLUDING comprehension
questions) do you think contained grammatical errors?

	
  

43

The participants completed the questionnaire on Survey Gizmo, an online survey tool. All
participants answered questions 1 and 2; however, if a participant selected “No” for question 2,
indicating that they had not seen any errors while reading, then the survey automatically skipped
questions 3 through 5, which inquired as to the nature of those ungrammaticalities. To ensure
that subsequent questions did not influence responses on previous questions, each question was
presented on a different page online, and participants could not navigate backwards through the
survey to return to previous questions.
Vocabulary posttest. To ensure that participants were familiar with all the target words,
I administered a vocabulary posttest immediately following the reading questionnaire portion of
the experiment. The participants were presented with the 20 target nouns from the experiment.
They were asked to first translate each noun and then mark the gender of the noun by checking
off if it was masculine or feminine. Next, participants were asked to translate two of the null
nominal sentences into English to ensure that they could indeed interpret those sentences. The
vocabulary posttest is presented in Appendix E.
Proficiency test. All participants completed the grammar portion of the Diploma de
Español como Lengua Extranjera (Certificates of Spanish as a Foreign Language, DELE) for
intermediate learners (Instituto Cervantes, 2008). The test consisted of a total of 20 multiple
choice questions in which participants must select from 3 options to fill in a blank in a paragraph.
Participants received 1 point for each correct response for a total of 20 possible points. The L2
Spanish learners scored an average of 10.84 points out of 20, (Mdn = 10.00; SD = 3.24, range =
4-18) while the native Spanish speakers scored an average of 18.30 points (Mdn = 18.00; SD =
1.24, range = 16-20).

	
  

44

Procedures
The data were collected on an individual basis in an eye-tracking lab. Data collection
lasted approximately 1.5 hours, and participants were paid $20 each for their time. Participants
first read and signed the consent form and completed the eye-tracking portion of the experiment.
The eye-tracking portion of the experiment began with 5 practice sentences to familiarize
participants with the procedure, followed by the 148 experimental sentences (60 critical, 88
distractors). Participants were not told that the sentences contained ungrammaticalities, and none
of the practice sentences were ungrammatical. Each experimental sentence was followed by a
comprehension question, to ensure that participants attended to meaning while reading.
Once participants completed the eye-tracking experiment, they completed the reading
questionnaire, vocabulary posttest, the background questionnaire, and finally the proficiency test
on Survey Gizmo, an online survey tool.
Analysis
Areas of Interest
The dependent variables in the current study were reading times on the areas of interest.
The reading times on the grammatical regions were compared to the reading times on the
ungrammatical regions for each of the three syntactic context conditions. Relatively longer
reading times on ungrammatical regions were assumed to reflect a processing cost, which was
taken as evidence that the participant was sensitive to the violation in grammatical gender
agreement.

	
  

45

The area of interest was different for each of the three agreement conditions. In the
DET-N agreement condition, the critical region encompassed both the determiner and noun in
each target sentence, shown in (4) reprinted as (8) below:
(8) El chico bebe el/*la refresco cuando ve la película en el cine con su familia.
Both the determiner and noun combined to form a single area of interest because readers
often skip over short words (Brysbaert, Drieghe, & Vitu, 2005; Brysbaert & Vitu, 1998; FrenckMestre, 2005; Vitu, O'Regan, Inhoff, & Topolski, 1995), making two-letter articles a challenge
to measure directly with eye tracking (cf. Spinner, Gass, & Behney, 2013).
The area of interest for the N-ADJ agreement condition encompassed just the adjective,
and the area of interest for the N-DROP condition was the full null nominal (determiner and
adjective). These areas of interest are shown in examples (9) and (10) below, reprinted from
examples (5) and (6), respectively.
(9) El atleta toma el refresco frío/*fría cuando termina de correr tres millas por la mañana.
(10) El jefe pide el refresco grande y su empleado pide el pequeño/*la pequeña cuando van a
McDonald’s.
To control for spillover effects, the word following the critical regions in each condition
was always either durante ‘during’ or cuando ‘when.’
Fixation Time Measures
I investigated four fixation time measures in the present dissertation:
•   First fixation duration – The duration of the first time a participant fixates on the area
of interest.

	
  

46

•   First-pass time – The sum of all fixations when a participant first fixates on the area
of interest before that participant’s gaze leaves that area. This is also known as gaze
duration when the fixation area is comprised of only one word19.
•   Go-past time (also known as regression path time) – Includes the first fixation and all
subsequent fixations (both in and outside of the area of interest) until exiting the word
to the right (in scripts that are read from left to right).
•   Total time – Sum of all fixations on a single area of interest.
First fixation duration and first-pass time are early measures of processing, and are
thought to reflect word identification processes. Total time is a late processing measure, which
can often indicate processing difficulty (see Pickering, Frisson, McElree, & Traxler, 2004). Gopast time is often considered both an early and late processing measure because it includes both
word integration (an early measure) and the time it takes to overcome any processing difficulties
and move on in the sentence (Clifton et al., 2007).
Data Cleaning
The eye-tracking data were first cleaned manually. Any fixations that were slightly above
or below the target region were moved vertically so that they fell into the target regions. Missing
data due to skipped words or track loss accounted for 4.9% of the total data set. Trials in which
the critical region was not fixated on for at least 80 ms were removed from the data set when the
fixation reports were generated, and trials that were over or under 2 standard deviations away
from each participant’s mean for both conditions (grammatical and ungrammatical) were

	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  
19
The areas of interest for the DET-N and N-DROP conditions are comprised of two words,
while the area of interest for the N-ADJ condition is comprised of only one. In the interest of
consistency, the term first-pass time will be used to refer to all three areas of interest in the
present dissertation.
	
  

47

manually replaced with a value two standard deviations away from that mean (Keating, 2014;
Keating & Jegerski, 2015).
If participants indicated on the vocabulary posttest that they were unfamiliar with the
translation or gender of one of the target nouns, the experimental sentences containing that noun
were excluded on a participant basis from the analyses of all three conditions (DET-N, N-ADJ
and N-DROP). This is a common practice in studies that investigate grammatical gender
agreement (e.g., White et al., 2004), and was done to ensure that the L2 learners’ knowledge of
the target words approximated that of the native Spanish speakers, at least at a declarative level.
A total of eight L2 Spanish learners translated at least one of the target nouns incorrectly, which
resulted in a loss of another 0.89% of the data set. Thirteen L2 Spanish learners marked the
incorrect gender for at least one of the target nouns, which resulted in a loss of 1.43% of the data
set.
All participants correctly translated the null nominal sentences from Spanish into English
on the vocabulary posttest (see Appendix E); however, if they could not, their reading times in
the N-DROP condition would have been removed from the analyses.
Statistical Analyses
These data were analyzed with linear mixed-effects models (Baayen, Davidson, & Bates,
2008). Linear mixed-effects models can be used for data with repeated measures, and for data
that are hierarchical or clustered. They differ from means-based inferential statistics, such as
ANOVAs, in many ways. First, instead of comparing the means of participants per condition,
mixed-effects models consider each of the participants’ observations separately. Second, mixedeffects models are considered to be more robust than ANOVAs because they allow for all the
factors under investigation to be considered simultaneously (Cunnings, 2012; Cunnings &

	
  

48

Finlayson, 2015; Jaeger, 2008; Keating & Jegerski, 2015; Plonsky, 2013), thereby obviating the
need for separate by-participants and by-items analyses (Cunnings, 2012; Cunnings & Finlayson,
2015). Another benefit of linear mixed-effects models is that, unlike ANOVAs, they are
relatively robust against missing data (Quené & van den Bergh, 2004), which is not uncommon
in eye-tracking research because of measurement error. They are also robust against unequal
sample sizes per group (Quené & van den Bergh, 2004).
Linear mixed-effects models measure how well an outcome variable can be predicted by
fixed and random effects. Fixed effects model how the independent variable(s) affect(s) the
outcome variable, while random effects model variance that can be attributed to other factors
inherent in the sampling of the study, such as subject or item variance (Cunnings, 2012).
Whereas fixed effects are hypothesized to have systematic and predictable effects on the
outcome variable, random effects are expected to be idiosyncratic and unpredictable (Cunnings,
2012). In this dissertation, the outcome variable was one of four different reading time measures
(first fixation duration, first-pass time, go-past time and total time). The fixed effects were group
(native Spanish speakers, L2 Spanish learners) and grammaticality (grammatical,
ungrammatical). Subject and item were also included as crossed random effects. For each of the
four fixation time measures, each participant yielded 20 observations per structure. Separate
analyses were conducted for each of the three syntactic context conditions (DET-N, N-ADJ, NDROP) because, for practical reasons, the areas of interest could not be controlled for length and
frequency across those three conditions, thus making direct within-analysis comparisons
impossible.
The data were first checked to make sure they met the assumptions of linear mixedeffects models. One assumption is that the data should be normally distributed. An initial

	
  

49

exploration into the data for this dissertation revealed that the reading times were, in fact, skewed
to the right. I have showed this lack of normality in the data in Figure 3.2. Another assumption of
linear mixed-effects models is the absence of heteroscedasticity20, meaning that the variance in
the residuals should be relatively equal across the range of the predicted values. The residual plot
in Figure 3.3, created with 800 fictitious and random data points, shows residuals that are
homoscedastic, as there is no obvious pattern in the plot. Figure 3.4 shows the residuals for one
of the models run in this dissertation with total reading time in the DET-N condition as the
outcome variable. The plot is cone-shaped, indicating that the larger the predicted means are, the
larger the residuals are.

Figure 3.2
Histogram depicting total time data for the DET-N condition

	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  
20
Some researchers vary in their opinion on how important this assumption is. For example,
Winter (2013) notes that linear mixed-effects models cannot violate the assumption of absence of
heteroscedasticity while Quené and van den Bergh (2004) claim they are still robust in the face
of heteroscedasticity.	
  
	
  

50

Figure 3.3
Residual plot from fictitious data that does not violate the assumption of absence of
heteroscedasticity

Figure 3.4
Residual plot depicting total time data for the DET-N condition
To account for these violations of the assumptions of the model, a log transformation was
performed on all the outcome variables (i.e., the different reading time measures). This log
	
  

51

transformation corrected the problems of lack of normality and absence of homoscedasticity, as
can be seen in Figures 3.5 and 3.6, respectively.

Figure 3.5
Histogram depicting the log-transformed total time data for the DET-N condition

Figure 3.6
Residual plot depicting the log-transformed total time data for the DET-N condition
To summarize, I analyzed the data with linear mixed-effects models with reading times as
the outcome variables, participant group and grammaticality as fixed effects, and subject and
	
  

52

item as random effects. All the outcome variables were log-transformed. Each participant
contributed 20 data points to each of the three conditions (DET-N, N-ADJ, N-DROP), 10 for
grammatical items and 10 for ungrammatical items. I conducted four tests within each condition,
one for each of the four reading time measures specified as outcome variables (first fixation
duration, first-pass time, go-past time and total time). The data were analyzed using models with
and without an interaction between group and grammaticality. Interactions are only reported
when they were significant. All statistics for this dissertation were computed in R (R Core Team,
2015) with the lme(4) package (Bates, Mächler, Bolker, & Walker, 2015). For the interested
reader, Cunnings (2012), Cunnings and Finlayson (2015), Gries (2015) and Winter (2013) all
provide clear explanations on how to compute linear mixed-effects models in R.

	
  

53

CHAPTER 4: RESULTS
This chapter presents the results of this dissertation and is divided into three primary
sections. In the first section, I will present the statistical analyses detailing the native speaker and
L2 learners’ sensitivity to grammatical violations for each of the three conditions (DET-N, NADJ and N-DROP) separately. I will first present the descriptive statistics for each of the four
outcome variables: first fixation duration, first-pass time, go-past time and total time. I will then
present the inferential statistics for each of the four outcome variables.
In the second section, I will present the findings of additional analyses examining only
the L2 learners’ sensitivity to grammatical violations by the gender of the target nouns. The
rationale for and description of these additional analyses will be presented at the beginning of
this section. Then, I will present the descriptive and inferential statistics for each of the four
outcome variables in all three conditions (some of these analyses can also be found in Appendix
H).
In the third section, I will present results examining sensitivity to grammatical violations
on an individual participant basis. In this section, I explore the results of the reading
questionnaire, and also triangulate L2 learners’ reading times in the eye-tracking experiment
with their responses on the reading questionnaire.

	
  

54

Statistical Analyses: Native Speakers’ and L2 Learners’ Reading Times
on Grammatical and Ungrammatical Sentences
Determiner-Noun Agreement
Descriptive statistics. Table 4.1 depicts the mean fixation times for each of the four
fixation measures for the DET-N condition. As is typical in this kind of research, the native
Spanish speakers tended to have shorter reading times than the L2 Spanish learners across
reading measures. Both native Spanish speakers and L2 Spanish learners have longer reading
times for ungrammatical regions relative to grammatical ones for first-pass, go-past and total
time.
Table 4.1
Mean (Standard Deviation) Fixation Times for DET-N Condition

Native Speakers
G
UG
Difference
L2 Learners
G

First Fix.

First-Pass

Go-Past

Total

242.00
(98.91)
244.08
(115.52)
2.08

341.94
(187.12)
411.23
(257.11)
69.29

461.32
(308.41)
591.72
(429.39)
130.40

592.19
(385.18)
754.86
(431.26)
162.67

273.81
481.99
626.46
781.90
(106.33)
(284.37)
(413.40)
(453.82)
UG
254.93
524.89
713.13
938.27
(100.30)
(284.67)
(463.47)
(507.90)
Difference
-18.88
42.90
86.67
156.37
Note. First Fix, First Fixation Duration; First-Pass, First-Pass Time; Go-Past, Go-past Time;
Total, Total Time; G, Grammatical; UG, Ungrammatical. All times are presented in
milliseconds.
First fixation duration. The results of the linear mixed-effects model for first fixation
duration in the DET-N condition are presented in Table 4.2. The native Spanish speakers’
reading times for the grammatical condition were statistically different from zero (t = 179.64, p <
.001). First fixation duration showed an effect of group (t = 2.04, p = .046) with L2 Spanish

	
  

55

learners reading 8%21 more slowly than the native Spanish speakers. There was no main effect of
grammaticality (t = -1.01, p < .311). The final model did not include an interaction term.
Table 4.2
Model Results for First Fixation Duration in DET-N Condition
Description
Predictor
Coefficient
Standard
t
p
(β)
Error
Overall effect of being a
Intercept
5.43
0.03
179.64 < .001
native speaker for the
grammatical condition
Overall main effect of being
Group
0.08
0.04
2.04
.046
an L2 learner
Overall main effect of
Grammaticality
-0.02
0.02
-1.01
.311
reading an ungrammatical
sentence
Random Effects
Subject
0.014
Item
0.000
Residual
10.132
Note. Group and grammaticality are both categorical variables with two levels. Native speaker
reading times on grammatical sentences were taken as the reference category (Intercept).
First-pass time. The results of the linear mixed-effects model for first-pass time in the
DET-N condition are presented in Table 4.3. The native Spanish speakers’ reading times for the
grammatical condition were statically different from zero (t = 125.09, p < .001). First-pass time
showed an effect of group (t = 5.26, p = < .001), with L2 Spanish learners reading 29% more
slowly than the native speakers. There was also an effect of grammaticality (t = 3.56, p < .001),
with participants reading in the ungrammatical condition 11% more slowly than the grammatical
condition. The final model did not include an interaction term.

	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  
21
Because a log transformation was performed on the outcome variable, and the two predictor
variables are binary, the coefficients can be multiplied by 100 to calculate percent change from
the reference category.
	
  

56

Table 4.3
Model Results for First-Pass Time in DET-N Condition
Description
Predictor
Coefficient
Standard
t
p
(β)
Error
Overall effect of being a
Intercept
5.72
0.05
125.09 < .001
native speaker for the
grammatical condition
Overall main effect of being
Group
0.29
0.06
5.26
< .001
an L2 learner
Overall main effect of
Grammaticality
0.11
0.03
3.56
< .001
reading an ungrammatical
sentence
Random Effects
Subject
0.027
Item
0.006
Residual
0.261
Note. Group and grammaticality are both categorical variables with two levels. Native speaker
reading times on grammatical sentences were taken as the reference category (Intercept).
Go-past time. The results of the linear mixed-effects model for go-past time in the DETN condition are presented in Table 4.4. The average native Spanish speaker’s reading times for
the grammatical condition were statistically different from zero (t = 113.82, p < .001). Go-past
time showed an effect of group (t = 4.24, p = < .001), with L2 Spanish learners reading in the
grammatical condition 27% more slowly than the native speakers. There was also an effect of
grammaticality (t = 4.90, p < .001), with participants reading in the ungrammatical condition
17% more slowly than the grammatical condition. The final model did not include an interaction
term.

	
  

57

Table 4.4
Model Results for Go-Past Time in DET-N Condition
Description
Predictor
Coefficient
Standard
t
p
(β)
Error
Overall effect of being a
Intercept
5.98
0.05
113.82 < .001
native speaker for the
grammatical condition
Overall main effect of being
Group
0.27
0.06
4.24 < .001
an L2 learner
Overall main effect of
Grammaticality
0.17
0.04
4.90 < .001
reading an ungrammatical
sentence
Random Effects
Subject
0.038
Item
0.009
Residual
0.294
Note. Group and grammaticality are both categorical variables with two levels. Native speaker
reading times on grammatical sentences were taken as the reference category (Intercept).
Total time. The results of the linear mixed-effects model for total time in the DET-N
condition are presented in Table 4.5. The average native Spanish speaker’s reading times for the
grammatical condition were statistically different from zero (t = 114.63, p < .001). Total time
showed an effect of group (t = 3.83, p = < .001), with L2 Spanish learners reading in the
grammatical condition 27% more slowly than the native speakers. There was also an effect of
grammaticality (t = 6.38, p < .001), with participants reading the ungrammatical sentences 22%
more slowly than the grammatical sentences. The final model did not include an interaction term.

	
  

58

Table 4.5
Model Results for Total Time in DET-N Condition
Description
Predictor
Coefficient
Standard
t
p
(β)
Error
Overall effect of being a
Intercept
6.22
0.05
114.63 < .001
native speaker for the
grammatical condition
Overall main effect of being
Group
0.27
0.07
3.83 < .001
an L2 learner
Overall main effect of
Grammaticality
0.22
0.03
6.38 < .001
reading an ungrammatical
sentence
Random Effects
Subject
0.049
Item
0.005
Residual
0.297
Note. Group and grammaticality are both categorical variables with two levels. Native speaker
reading times on grammatical sentences were taken as the reference category (Intercept).
Summary of results. All four of the models in the DET-N condition showed a main
effect of group, with L2 learners reading more slowly than native Spanish speakers. For firstpass time, go-past time, and total time, there was also a main effect for grammaticality, with
participants reading more slowly in the ungrammatical condition relative to the grammatical one.
There were no interactions between group and grammaticality for any of the four outcome
variables.
Noun-Adjective Agreement
Descriptive statistics. Table 4.6 depicts the mean fixation times for each of the four
fixation measures for the N-ADJ condition. The native Spanish speakers generally tended to
have shorter reading times than L2 Spanish learners. The native Spanish speakers evidenced
increased reading times for ungrammatical regions for all four fixation measures with the largest
increase showing up in total time. The L2 learners evidenced reading times that were roughly
equivalent for first fixation, first-pass and go-past reading times, with a slight increase for the
ungrammatical region in total time.
	
  

59

Table 4.6
Mean (Standard Deviation) Fixation Times for N-ADJ Condition

Native Speakers
G
UG
Difference
L2 Learners
G

First Fix.

First-Pass

Go-Past

Total

275.30
(98.538)
289.67
(121.10)
14.37

300.51
(113.99)
331.13
(144.09)
30.62

371.71
(219.77)
457.37
(316.25)
85.66

425.29
(235.81)
592.25
(393.39)
166.96

298.56
365.64
461.38
531.04
(114.49)
(144.04)
(293.81)
(288.88)
UG
290.81
352.58
464.71
570.40
(106.58)
(183.99)
(328.88)
(358.88)
Difference
-7.75
-13.06
3.33
39.36
Note. First Fix, First Fixation Duration; First-Pass, First-Pass Time; Go-Past, Go-past Time;
Total, Total Time; G, Grammatical; UG, Ungrammatical. All times are presented in
milliseconds.
First fixation duration. The results of the linear mixed-effects model for first fixation
duration in the N-ADJ condition are presented in Table 4.7. The average native Spanish
speaker’s reading times for the grammatical condition were statistically different from zero (t =
195.52, p < .001). However, first fixation duration did not show a main effect of group (t = 1.41,
p = .166), or grammaticality (t = 0.39, p = .694). The final model did not include an interaction
term.

	
  

60

Table 4.7
Model Results for First Fixation Duration in N-ADJ Condition
Description
Predictor
Coefficient
Standard
t
p
(β)
Error
Overall effect of being a
Intercept
5.56
0.03
195.52
< .001
native speaker for the
grammatical condition
Overall main effect of
Group
0.05
0.04
1.41
= .166
being an L2 learner
Overall main effect of
Grammaticality
0.01
0.02
0.39
= .694
reading an ungrammatical
sentence
Random Effects
Subject
0.008
Item
0.002
Residual
0.132
Note. Group and grammaticality are both categorical variables with two levels. Native speaker
reading times on grammatical sentences were taken as the reference category (Intercept).
First-pass time. The results of the linear mixed-effects model for first-pass time in the
N-ADJ condition are presented in Table 4.8. The average native Spanish speaker’s reading times
for the grammatical condition were statistically different from zero (t = 141.79, p < .001). Firstpass time showed a main effect of group (t = 4.34, p < .001) and grammaticality (t = 2.78, p
<.001), and a group*grammaticality interaction (t = -3.15, p < .001). The native Spanish
speakers had reading times that were 9% longer in the ungrammatical condition relative to the
grammatical one, and post hoc analyses confirmed that this difference was statistically
significant (β = -.10, t = -2.78, p = .029). L2 Spanish learners had reading times that were 7%22
shorter on the ungrammatical condition, but this difference was not statistically significant (β =
.06, t = 1.67, p = .343).

	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  
22
This percentage is computed by using the coefficients to first calculate the percentage change
for the L2 learners’ reading times in the grammatical condition relative to the reference category,
(5.62+.20 = 5.82), then to calculate the percentage change for the L2 learners’ reading times in
the ungrammatical condition relative to the reference category (5.62 + .20 +.09 - .16 = 5.75) and
then calculating the difference between the two and multiplying by 100.
	
  

61

Table 4.8
Model Results for First-Pass Time in N-ADJ Condition
Description
Predictor
Coefficient Standard
t
p
(β)
Error
Overall effect of being a Intercept
5.62
0.04
141.79 < .001
native speaker for the
grammatical condition
Overall main effect of
Group
0.20
0.05
4.34 < .001
being an L2 learner
Overall main effect of
Grammaticality
0.09
0.04
2.78 < .001
reading an
ungrammatical sentence
Interaction between
Group*Grammaticality
-0.16
0.05
-3.15 < .001
group and
grammaticality
Random Effects
Subject
0.011
Item
0.011
Residual
0.144
Note. Group and grammaticality are both categorical variables with two levels. Native speaker
reading times on grammatical sentences were taken as the reference category (Intercept).
Go-past time. The results of the linear mixed-effects model for go-past time in the NADJ condition are presented in Table 4.9. The average native Spanish speaker’s reading times
for the grammatical condition were statistically different from zero (t = 117.85, p < .001). Gopast time showed a main effect of group (t = 3.40, p < .001) and grammaticality (t = 3.77, p
<.001), and a group*grammaticality interaction (t = -3.07, p < .001). According to the model, the
native Spanish speakers had reading times that were 17% longer in the ungrammatical condition
relative to the grammatical one, and post hoc analyses confirmed that this difference was
statistically significant (β = -.17, t = -3.77, p = .001). The L2 Spanish learners had reading times
that were 3% shorter on the ungrammatical sentences, but this difference was not statistically
significant (β = .03, t = 0.56, p = .943).

	
  

62

Table 4.9
Model Results for Go-Past Time in N-ADJ Condition
Description
Predictor
Coefficient Standard
t
p
(β)
Error
Overall effect of being a Intercept
5.78
0.05
117.85 < .001
native speaker for the
grammatical condition
Overall main effect of
Group
0.21
0.06
3.40 < .001
being an L2 learner
Overall main effect of
Grammaticality
0.17
0.05
3.77 < .001
reading an
ungrammatical sentence
Interaction between
Group*Grammaticality
-0.20
0.07
-3.07 < .001
group and
grammaticality
Random Effects
Subject
0.023
Item
0.010
Residual
0.243
Note. Group and grammaticality are both categorical variables with two levels. Native speaker
reading times on grammatical sentences were taken as the reference category (Intercept).
Total time. The results of the linear mixed-effects model for total time in the N-ADJ
condition are presented in Table 4.10. The average native Spanish speaker’s reading times for the
grammatical condition were statistically different from zero (t = 95.77, p < .001). Total time
showed a main effect of group (t = 3.40, p = .001) and grammaticality (t = 6.62, p <.001), and a
group*grammaticality interaction (t = -4.11, p < .001). According to the model, the native
Spanish speakers had reading times that were 29% longer in the ungrammatical condition
relative to the grammatical one, and post hoc analyses confirmed that this difference was
statistically significant (β = -.29, t = -6.62, p = < .001). The L2 Spanish learners had reading
times that were 3% longer, and post hoc analyses confirmed that this difference was not
statistically significant (β = -.03, t = -0.70, p = .896).

	
  

63

Table 4.10
Model Results for Total Time in N-ADJ Condition
Description
Predictor
Overall effect of being a
native speaker for the
grammatical condition
Overall main effect of being
an L2 learner
Overall main effect of
reading an ungrammatical
sentence
Interaction between group
and grammaticality

Intercept

Coefficient Standard
t
p
(β)
Error
5.91
0.06
95.77 < .001

Group

0.23

0.07

3.40

= .001

Grammaticality

0.29

0.04

6.62

< .001

Group*Grammaticality

-0.26

0.06

-4.11 < .001

Random Effects
Subject
0.036
Item
0.030
Residual
0.243
Note. Group and grammaticality are both categorical variables with two levels. Native speaker
reading times on grammatical sentences were taken as the reference category (Intercept).
Summary of results. Three of the four models (first-pass time, go-past time and total
time) in the N-ADJ condition showed a main effect of group, with L2 learners reading more
slowly than native Spanish speakers. The same three models also showed a main effect of
grammaticality, and a group*grammaticality interaction. While the native speakers consistently
evidenced longer reading times (between 9% and 29%) in the ungrammatical condition relative
to the grammatical one, the L2 learners showed little difference in reading times between the
grammatical and ungrammatical conditions.
Null Nominal Agreement
Descriptive statistics. Table 4.11 depicts the mean fixation times for each of the four
fixation measures for the N-DROP condition. The native Spanish speakers generally tended to
have shorter reading times than L2 Spanish learners. The native Spanish speakers evidenced
increased reading times for ungrammatical regions for all four fixation measures with the largest

	
  

64

increase showing up in total time, and the smallest increase for first fixation duration. The L2
learners evidenced reading times that were roughly equivalent for all four reading times.
Table 4.11
Mean (Standard Deviation) Fixation Times for N-DROP Condition

Native Speakers
G
UG
Difference
L2 Learners
G

First Fix.

First-Pass

Go-Past

Total

242.84
(76.25)
258.67
(100.76)
15.83

301.56
(145.00)
357.72
(180.19)
56.16

374.02
(238.72)
449.08
(273.58)
75.06

487.91
(308.39)
624.87
(420.26)
136.96

261.69
376.11
449.69
627.37
(93.21)
(189.30)
(264.70)
(423.08)
UG
261.88
381.30
481.96
614.71
(91.99)
(220.57)
(458.01)
(473.02)
Difference
0.19
5.19
32.27
-12.66
Note. First Fix, First Fixation Duration; First-Pass, First-Pass Time; Go-Past, Go-past Time;
Total, Total Time; G, Grammatical; UG, Ungrammatical. All times are presented in
milliseconds.
First fixation duration. The results of the linear mixed-effects model for first fixation
time in the N-DROP condition are presented in Table 4.12. The average native Spanish speaker’s
reading times for the grammatical condition were statistically different from zero (t = 190.62, p <
.001). However, first fixation duration did not show a main effect of group (t = 1.26, p = .215) or
grammaticality (t = 1.26, p = .208). The final model did not include an interaction term.

	
  

65

Table 4.12
Model Results for First Fixation Duration in N-DROP Condition
Description
Predictor
Coefficient
Standard
t
p
(β)
Error
Overall effect of being a
Intercept
5.46
0.03
190.62 < .001
native speaker for the
grammatical condition
Overall main effect of being
Group
0.04
0.04
1.26
= .215
an L2 learner
Overall main effect of
Grammaticality
0.03
0.02
1.26
= .208
reading an ungrammatical
sentence
Random Effects
Subject
0.011
Item
0.003
Residual
0.095
Note. Group and grammaticality are both categorical variables with two levels. Native speaker
reading times on grammatical sentences were taken as the reference category (Intercept).
First-pass time. The results of the linear mixed-effects model for first-pass time in the
N-DROP condition are presented in Table 4.13. The average native Spanish speaker’s reading
times for the grammatical condition were statistically different from zero (t = 118.01, p < .001).
First-pass time showed a main effect of group (t = 3.73, p < .001) and grammaticality (t = 3.86, p
<.001), and a group*grammaticality interaction (t = -2.80, p =.005). The native Spanish speakers
had reading times that were 15% longer in the ungrammatical condition relative to the
grammatical one, and post hoc analyses revealed that this difference was statistically significant
(β = -.29, t = -6.62, p = < .001). The L2 Spanish learners did not evidence any percent change in
reading time, which was confirmed by post hoc analyses (β = .01, t = 0.15, p = 1.000).

	
  

66

Table 4.13
Model Results for First-Pass Time in N-DROP Condition
Description
Predictor
Coefficient Standard
t
p
(β)
Error
Overall effect of being a Intercept
5.61
0.05
118.01 < .001
native speaker for the
grammatical condition
Overall main effect of
Group
0.21
0.06
3.73 < .001
being an L2 learner
Overall main effect of
Grammaticality
0.15
0.04
3.86 < .001
reading an
ungrammatical sentence
Interaction between
Group*Grammaticality
-0.15
0.05
-2.80
= .005
group and
grammaticality
Random Effects
Subject
0.022
Item
0.014
Residual
0.182
Note. Group and grammaticality are both categorical variables with two levels. Native speaker
reading times on grammatical sentences were taken as the reference category (Intercept).
Go-past time. The results of the linear mixed-effects model for go-past time in the NDROP condition are presented in Table 4.14. The average native Spanish speaker’s reading
times for the grammatical condition were statistically different from zero (t = 95.80, p < .001).
Go-past time showed a main effect of group (t = 2.78, p = .007) and grammaticality (t = 3.50, p
<.001), and a group*grammaticality interaction (t = -2.60, p = .009). The native Spanish
speakers had reading times that were 15% longer in the ungrammatical condition relative to the
grammatical one, and post hoc analyses confirmed that this difference was statistically
significant (β = -0.15, t = -3.50, p = .003). The L2 Spanish learners had reading times that were
1% slower, but this difference was not statistically significant (β = 0.01, t = 0.22, p = .996).

	
  

67

Table 4.14
Model Results for Go-Past Time in N-DROP Condition
Description
Predictor
Coefficient Standard
t
p
(β)
Error
Overall effect of being a Intercept
5.79
0.06
95.80 < .001
native speaker for the
grammatical condition
Overall main effect of
Group
0.18
0.07
2.78 = .007
being an L2 learner
Overall main effect of
Grammaticality
0.15
0.04
3.50 < .001
reading an
ungrammatical sentence
Interaction between
Group*Grammaticality
-0.16
0.06
-2.60 = .009
group and
grammaticality
Random Effects
Subject
0.029
Item
0.032
Residual
0.242
Note. Group and grammaticality are both categorical variables with two levels. Native speaker
reading times on grammatical sentences were taken as the reference category (Intercept).
Total time. The results of the linear mixed-effects model for total time in the N-DROP
condition are presented in Table 4.15. The average native Spanish speaker’s reading times for the
grammatical condition was different from zero (t = 72.51, p < .001). Total time showed a main
effect of group (t = 2.66, p = .009) and grammaticality (t = 4.30, p <.001), and a
group*grammaticality interaction (t = -3.64, p < .001). The native Spanish speakers had reading
times that were 19% longer in the ungrammatical condition relative to the grammatical one, and
post hoc analyses confirmed that these differences were statistically significant (β = -0.19, t = 4.31, p < .001). The L2 Spanish learners had reading times that were 4% shorter in the
ungrammatical condition, but this difference was not statistically significant (β = 0.04, t = 0.88, p
= .818).

	
  

68

Table 4.15
Model Results for Total Time in N-DROP Condition
Description
Predictor
Overall effect of being a
native speaker for the
grammatical condition
Overall main effect of
being an L2 learner
Overall main effect of
reading an ungrammatical
sentence
Interaction between group
and grammaticality

Intercept

Coefficient Standard
(β)
Error
6.03
0.08

t

p

72.51

< .001

Group

0.22

0.08

2.66

= .009

Grammaticality

0.19

0.05

4.30

< .001

Group*Grammaticality

-0.23

0.06

-3.64

< .001

Random Effects
Subject
0.061
Item
0.073
Residual
0.257
Note. Group and grammaticality are both categorical variables with two levels. Native speaker
reading times on grammatical sentences were taken as the reference category (Intercept).
Summary of results. Three of the four models (first-pass time, go-past time and total
time) in the N-DROP condition showed a main effect of group, with L2 learners reading more
slowly than native Spanish speakers. The same three models also showed a main effect of
grammaticality, and a group*grammaticality interaction. While the native speakers consistently
evidenced longer reading times (between 15% and 19%) in the ungrammatical condition relative
to the grammatical one, the L2 learners showed little difference in reading times between the
grammatical and ungrammatical conditions.
General Summary of Statistical Results
This section presented the results on native and L2 learners’ reading times for
grammatical and ungrammatical sentences with grammatical gender agreement under three
conditions: DET-N, N-ADJ, and N-DROP. Table 4.16 summarizes the statistical results related
to whether or not both native speakers and L2 learners slowed down in the ungrammatical
condition. To give the reader an indication of the magnitude of the sensitivity to violations across
	
  

69

the different sentential contexts, Figure 4.1 summarizes the participants’ percent change in
reading time, as estimated by the coefficients in the linear mixed-effects models described above.
Positive numbers indicate a longer reading time in the ungrammatical condition.
Table 4.16
Summary of Statistical Analyses
First-Fixation
Slowed down in
ungrammatical
condition?
Native Speakers
DET-N
No
N-ADJ
No
N-DROP
No
L2 Learners
DET-N
No
N-ADJ
No
N-DROP
No

	
  

First-Pass
Slowed down in
ungrammatical
condition?

Go-Past
Slowed down in
ungrammatical
condition?

Total Time
Slowed down in
ungrammatical
condition?

Yes
Yes
Yes

Yes
Yes
Yes

Yes
Yes
Yes

Yes
No
No

Yes
No
No

Yes
No
No

70

Figure 4.1
Participants’ percent change in reading time between grammatical and ungrammatical
sentences
Additional Statistical Analyses: L2 Learners’ Reading Times on Grammatical and
Ungrammatical Sentences by Gender
In the previous section, the L2 learners were sensitive to violations of DET-N agreement,
as evidenced by their statistically slower reading times on first-pass, go-past and total time on
ungrammatical areas of interest. However, because nouns of both genders were grouped for these
analyses, the results cannot determine if the gender of the target nouns plays a role in the L2
learners’ sensitivity to ungrammaticalities. That is, it could be that L2 learners’ sensitivity to
violations is asymmetrical. Previous research has indicated that L2 learners exhibit a masculine

	
  

71

default (e.g., Franceschina, 2001; McCarthy, 2007; Montrul et al., 2008; White et al., 2004),
which would mean that the learners may exhibit greater sensitivity to violations such as (11a)
compared to (11b):
(11)   a. *la
the FEM

proyecto
project MASC

‘the project’
b. *el
the MASC

manzana
apple FEM

‘the apple’
In this section, I aim to determine: (a) if the sensitivity to grammatical violations that L2
learners evidenced in the DET-N condition is asymmetrical and, (b) if the L2 learners’ apparent
lack of sensitivity found in previous statistical analyses can be attributed to the gender of the
target noun acting as an intervening variable.
In the analyses of the previous section, the fixed effects of the mixed-effects models were
group (native Spanish speakers, L2 Spanish learners) and grammaticality (grammatical,
ungrammatical). Subject and item were included as random effects. One benefit of linear mixedeffects models is that many factors can be included as fixed and random effects in a single
model. The number of factors that can be included in a given model, however, is not limitless:
complex models with many fixed and random effects require large sample sizes so the model can
be adequately fit to the data. The gender of the target noun (masculine, feminine) could have
been included as a third fixed effect in the models from the previous section; however, when I
tried to do this the models failed to converge, meaning that the model was too complex for the

	
  

72

sample size. Results from models that fail to converge are not reliable and therefore should not
be reported.
Instead, I opted to run a separate set of analyses on only the L2 learners’ reading times. In
this way, group (native Spanish speakers, L2 learners) could be dropped as a fixed effect and
replaced by gender (masculine, feminine). Therefore, the fixed effects for these additional
analyses were gender (masculine, feminine) and grammaticality (grammatical, ungrammatical),
and subject and item were included again as random effects23. I conducted a log transformation
on all the outcome variables (i.e., the different reading time measures) to correct for problems of
normality and heteroscedasticity. The data were analyzed using models with and without an
interaction between gender and grammaticality. Interactions are only reported when they were
significant.
Descriptive Statistics
Table 4.17 depicts the L2 learners’ mean fixation times for each of the four fixation
measures for all three conditions.

	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  
23
Even though I ran these analyses on a subset of the data, thus lowering the sample size, none
these models failed to converge.	
  
	
  

73

Table 4.17
Mean (Standard Deviation) Fixation Times for L2 Learners by Gender of Target Noun

DET-N
Masculine
G
UG
Difference
Feminine
G
UG
Difference
N-ADJ
Masculine
G
UG
Difference
Feminine
G
UG
Difference
N-DROP
Masculine
G
UG
Difference
Feminine
G

First Fix.

First-Pass

Go-Past

Total

284.92
(125.72)
248.19
(93.10)
-36.73

481.30
(306.58)
554.37
(303.76)
73.07

642.21
(429.17)
730.19
(493.16)
87.98

811.39
(469.77)
934.08
(509.61)
122.69

262.89
(103.87)
261.56
(106.89)
-1.33

482.68
(261.55)
495.41
(262.17)
12.73

610.45
(397.87)
695.93
(432.87)
85.48

752.18
(437.05)
942.49
(508.27)
190.31

294.21
(115.56)
288.79
(107.86)
-5.42

341.76
(136.05)
342.48
(179.56)
0.72

449.74
(316.33)
446.33
(309.34)
-3.41

520.82
(296.78)
570.88
(337.90)
50.06

302.91
(113.74)
292.71
(105.80)
-10.20

389.72
(148.39)
362.00
(189.19)
-27.72

473.23
(269.84)
481.69
(346.40)
8.46

541.52
(281.43)
569.94
(379.16)
28.42

269.22
(105.27)
265.87
(102.69)
-3.35

390.47
(198.26)
376.82
(204.32)
-13.65

472.30
(271.12)
510.27
(568.68)
37.97

625.61
(436.54)
613.66
(487.10)
-11.95

253.91
361.26
426.51
629.18
(78.54)
(179.19)
(257.20)
(410.64)
UG
257.78
385.86
452.92
615.80
(79.79)
(236.76)
(305.51)
(460.10)
Difference
3.87
24.60
26.41
-13.38
Note. First Fix, First Fixation Duration; First-Pass, First-Pass Time; Go-Past, Go-past Time;
Total, Total Time; G, Grammatical; UG, Ungrammatical. All times are presented in
milliseconds.
	
  

74

In the DET-N condition, the L2 learners evidenced substantially longer reading times on
ungrammatical sentences. Numerically, this difference was greater for the masculine nouns
during first-pass, but for the feminine nouns in total time. For the N-ADJ and N-DROP
conditions, the L2 learners did not evidence substantially longer reading times on the
ungrammatical sentences for feminine or masculine noun. These descriptive statistics indicate
that the L2 learners may evidence asymmetric sensitivity in the DET-N condition, but likely not
in the N-ADJ and N-DROP condition. Below, I will explore in detail the statistical analyses
performed on these data in the DET-N condition by reviewing the analyses for each of the four
fixation measures: first fixation, first-pass, go-past and total time. The statistical analyses for the
N-ADJ and N-DROP condition are located in Appendix F.
First Fixation Duration
The results of the linear mixed-effects model for first fixation duration in the DET-N
condition are presented in Table 4.18. The L2 learners’ reading times when reading grammatical
sentences with a masculine target noun were statistically different from zero (t = 159.70, p <
.001). First fixation duration showed no main effect of gender (t = -0.40, p = .691) or
grammaticality (t = -1.78, p = .075), although the latter did approach significance. The final
model did not include an interaction term.

	
  

75

Table 4.18
Gender Model Results for L2 Learners’ First Fixation Duration in DET-N Condition
Description
Predictor
Coefficient
Standard
t
p
(β)
Error
Overall effect of reading a
Intercept
5.54
0.04
159.70
< .001
grammatical sentence with a
masculine target noun
Overall main effect of
Gender
-0.01
0.04
-0.40 = .691
reading a sentence with a
feminine target noun
Overall main effect of
Grammaticality
-0.06
0.04
-1.78 = .075
reading an ungrammatical
sentence
Random Effects
Subject
0.008
Item
0.000
Residual
0.142
Note. Gender and grammaticality are both categorical variables with two levels. L2 learner
reading times on grammatical sentences with a masculine target noun were taken as the reference
category (Intercept).
First-Pass Time
The results of the linear mixed-effects model for first-pass time in the DET-N condition
are presented in Table 4.19. The L2 learners’ reading times when reading grammatical sentences
with a masculine target noun were statistically different from zero (t = 104.53, p < .001). Firstpass time showed no main effect of gender (t = -0.71, p = .484) or grammaticality (t = 1.59, p =
.113). The final model did not include an interaction term.

	
  

76

Table 4.19
Gender Model Results for L2 Learners’ First-Pass Time in DET-N Condition
Description
Predictor
Coefficient
Standard
t
p
(β)
Error
Overall effect of reading a
Intercept
6.06
0.06
104.53 < .001
grammatical sentence with a
masculine target noun
Overall main effect of
Gender
-0.04
0.06
-0.71 = .484
reading a sentence with a
feminine target noun
Overall main effect of
Grammaticality
0.08
0.05
1.59
= .113
reading an ungrammatical
sentence
Random Effects
Subject
0.024
Item
0.007
Residual
0.275
Note. Gender and grammaticality are both categorical variables with two levels. L2 learner
reading times on grammatical sentences with a masculine target noun were taken as the reference
category (Intercept).
Go-Past Time
The results of the linear mixed-effects model for go-past time in the DET-N condition are
presented in Table 4.20. The L2 learners’ reading times when reading grammatical sentences
with a masculine target noun were statistically different from zero (t = 102.26, p < .001). Go-past
time showed no main effect of gender (t = -0.58, p = .566), but did show a main effect of
grammaticality (t = 2.68, p = .008), with L2 learners reading in the ungrammatical condition
13% more slowly than the grammatical condition. The final model did not include an interaction
term.

	
  

77

Table 4.20
Gender Model Results for L2 Learners’ Go-Past Time in DET-N Condition
Description
Predictor
Coefficient
Standard
t
p
(β)
Error
Overall effect of reading a
Intercept
6.29
0.06
102.26 < .001
grammatical sentence with a
masculine target noun
Overall main effect of
Gender
-0.04
0.06
-0.58 = .566
reading a sentence with a
feminine target noun
Overall main effect of
Grammaticality
0.13
0.05
2.68 = .008
reading an ungrammatical
sentence
Random Effects
Subject
0.029
Item
0.009
Residual
0.282
Note. Gender and grammaticality are both categorical variables with two levels. L2 learner
reading times on grammatical sentences with a masculine target noun were taken as the reference
category (Intercept).
Total Time
The results of the linear mixed-effects model for total time in the DET-N condition are
presented in Table 4.21. The L2 learners’ reading times when reading grammatical sentences
with a masculine target noun were statistically different from zero (t = 110.32, p < .001). Total
time showed no main effect of gender (t = -0.83, p = .417), but did show a main effect of
grammaticality (t = 3.94, p < .001), with L2 learners reading in the ungrammatical condition
19% more slowly than the grammatical condition. The final model did not include an interaction
term.

	
  

78

Table 4.21
Gender Model Results for L2 Learners’ Total Time in DET-N Condition
Description
Predictor
Coefficient
Standard
t
p
(β)
Error
Overall effect of reading a
Intercept
6.52
0.06
110.32 < .001
grammatical sentence with a
masculine target noun
Overall main effect of
Gender
-0.04
0.05
-0.83 = .417
reading a sentence with a
feminine target noun
Overall main effect of
Grammaticality
0.19
0.05
3.94 < .001
reading an ungrammatical
sentence
Random Effects
Subject
0.041
Item
0.000
Residual
0.291
Note. Gender and grammaticality are both categorical variables with two levels. L2 learner
reading times on grammatical sentences with a masculine target noun were taken as the reference
category (Intercept).
General Summary of Statistical Results
This section presented the results L2 learners’ reading times for grammatical and
ungrammatical sentences depending on the gender of the target noun (masculine or feminine). In
the DET-N condition, the L2 learners evidenced longer reading times in the ungrammatical
condition relative to the grammatical condition for both go-past time and total time, but not first
fixation duration or first-pass time. There was no statistically significant interaction for any of
the models, meaning the sensitivity to the ungrammaticalities were not mediated by the gender of
the target noun. For the N-ADJ and N-DROP condition (see Appendix F for analyses), no
statistically significant main effect for gender or grammaticality was evidenced in any of these
models, and there were no interactions between gender and grammaticality.
Sensitivity to Violations of Gender Agreement on an Individual Basis
In this section I explore whether the participants evidenced sensitivity to violations of
grammatical gender agreement on an individual basis. I will examine this sensitivity in two
	
  

79

different ways. First, I will explore participants’ sensitivity as evidenced by their responses on
the reading questionnaire administered immediately after the eye-tracking experiment. Then, I
will explore the L2 learners’ sensitivity as evidenced by their individual reading times. Finally, I
will compare the L2 learners’ individual reading times to their responses on the self-reports.
Participants’ Sensitivity as Evidenced by Self-Reports
Immediately after finishing the eye-tracking experiment, participants were asked five
questions to determine whether they were aware that the sentences were embedded with
grammatical violations. These questions can be found in the section titled reading
questionnaire in Chapter 3.
The participants were first asked if they noticed anything strange about the sentences they
read during the eye-tracking experiment and if so, what. Of the 25 L2 learners, 15 reported errors
in the sentences they read. The others either did not report anything strange about the
experimental sentences, or commented solely on the comprehension questions. Of those 15 L2
learners, 10 stated that there were problems with grammatical gender agreement. For example,
one L2 learner noted, “…The grammar was incorrect at times in that the gender of the article
didn't match the gender of the noun” while another stated, “There were some grammatical errors.
For example something like ‘el madre’ instead of ‘la madre.’” Of the 10 L2 learners that reported
seeing violations of grammatical gender agreement, seven reported violations of DET-N
agreement, while the other three did not specify the context of the agreement violations. No L2
learner specifically mentioned violations of N-ADJ or N-DROP agreement.
For question 1, all the native Spanish speakers reported that the sentences were indeed
strange, and 26 of them reported that it was because of ungrammaticalities. Of those 26
participants, 11 specified that there were problems with grammatical gender agreement. Of those

	
  

80

11 participants, five reported violations of DET-N agreement, one reported a violation of N-ADJ
agreement, and five did not report a specific context.
Question 2 asked participants a more pointed question: whether there were any
grammatical errors in the sentences they read during the eye-tracking experiment. All the native
Spanish speakers reported that there were grammatical errors, while only 19 of the 25 L2
learners reported seeing errors. The six participants who stated that they did not notice any
grammatical errors were not asked the next three questions and proceeded directly in the online
survey to the vocabulary posttest.
The other 19 L2 Spanish learners and all the native speakers then navigated to question 3,
where they were asked to name the types of grammatical errors they saw and to provide
examples when possible. Fourteen of the L2 learners reported gender agreement errors. Of those
14, 12 reported DET-N agreement violations while 1 reported N-ADJ agreement violations. One
participant mentioned agreement violations without specifying a context. No L2 learner reported
N-DROP violations24.
Of the 27 native Spanish speakers, 16 reported gender agreement errors. Of those 16, 11
reported DET-N agreement violations and two reported N-ADJ agreement violations. Three
participants mentioned gender agreement violations without specifying one specific context. No
native Spanish speaker reported N-DROP violations.
For question 4, I asked participants to check off from a list all the types of grammatical
violations they noticed in the sentences. Some of the errors did appear in the experimental

	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  
24
This may be because participants were not sensitive to these violations and/or because they did
not have the metalinguistic vocabulary to describe these violations, as N-DROP is not usually
taught explicitly in Spanish language classrooms.
	
  

81

stimuli while others did not. This was to determine if participants were simply marking all
grammatical violations on the list.
Table 4.22 summarizes the participants’ responses for question 4. The three types of
agreement under investigation in the current study are highlighted in gray. The native Spanish
speakers were roughly equally sensitive to all three types of grammatical gender agreement,
while the L2 learners were most sensitive to DET-N agreement (16 participants), followed by NADJ agreement (13 participants), and then N-DROP agreement (11 participants).
Table 4.22
Participants’ Reported Sensitivity to Violations in Sentence Processing Task
Violation
# of native
Grammatical violation
appeared in
speakers
stimuli?
reported
violation
-   ADVERBS appeared in the wrong place in
yes
21/27
the sentence
-   incorrect tense (present, past, etc.) was used
yes
21/27
-   incorrect gender agreement between articles
yes
25/27
(e.g., el, la) and nouns
-   ADJECTIVES appeared in the wrong place in
yes
22/27
the sentence
-   incorrect agreement between subjects and
yes
23/27
verbs
-   subjunctive was used incorrectly
no
11/27
-   por and para were used incorrectly
no
9/27
-   ser and estar were used incorrectly
no
11/27
-   incorrect gender agreement between nouns
yes
25/27
and adjectives
-   incorrect gender agreement between nouns
yes
26/27
and null nominals (e.g., el rico, la cómica)

# of L2
learners
reported
violation
3/19
12/19
16/19
8/19
10/19
2/19
3/19
5/19
13/19
11/19

For the final question, I asked participants what percentage of the experimental sentences
they thought contained grammatical errors. The L2 learners answered that an average of 29% of
sentences (range: 3%-75%) contained grammatical errors and the native Spanish speakers
answered that an average of 71% did (range: 18% to 99%). In reality, half of the sentences

	
  

82

contained violations of some kind. The difference between the saliency of grammatical
violations is striking: native speakers tend to over-estimate the number of errors while L2
learners underestimate them.
L2 Learners’ Sensitivity on an Individual Basis
To calculate individual participants’ sensitivity to the grammatical gender agreement
violations, I subtracted each participant’s mean total reading time in the grammatical condition
from the ungrammatical condition in each of the three syntactic contexts (DET-N, N-ADJ, NDROP). I selected the total time measurement because it was the latest processing measure, and
therefore the one most likely to evidence sensitivity, if there were any. Table 4.23 shows the
change in reading time for the L2 Spanish learners.

	
  

83

Table 4.23
Individual Change in Reading Time for L2 Spanish Learners
NParticipant DET-N N-ADJ
DROP
A
-299
-12
-470
B
-265
113
132
C
-158
-213
-24
D
-84
136
98
E
-62
8
-26
F
-60
78
132
G
-10
56
-86
H
57
-9
-17
I
-46
-2
146
J
-26
174
-338
K
-15
143
-101
L
143
62
-25
M
176
-2
-194
N
262
16
13
O
290
20
-420
P
291
-83
-33
Q
502
106
-120
R
564
120
-217
S
174
-22
155
T
178
-155
323
U
495
12
171
V
537
-63
183
W
188
249
-156
X
236
226
564
Y
815
164
224
Mean
155
45
-3
Note. All times in are milliseconds. Any difference over 140 ms is highlighted in gray.
In Table 4.23, all differences over 140 ms are highlighted in gray to facilitate the
interpretation of the table25. This threshold is, admittedly, arbitrarily defined, just as any
threshold would be. I selected it by examining the data for the DET-N condition and finding a
natural break in the data: participants in the DET-N condition either evidenced at least 140 ms of
	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  
25
A similar procedure was used by Keating (2009) when comparing L2 learners’ sensitivity
across conditions. Keating, however, selected three different thresholds depending on the
condition under investigation (125 ms for agreement in the DP, 117 ms in the VP, and 161 ms in
the subordinate clause). In the interest of uniformity, I have instead selected to use a single
threshold for all conditions.
	
  

84

sensitivity or none at all (range: -299 – 57). I then applied this threshold to the other two
conditions. These data should be interpreted with caution, though, as altering the threshold can
yield different interpretations of the same data.
Using 140 ms as a threshold, 14 out of a total of 25 L2 learners were sensitive to
violations of DET-N agreement, five were sensitive to violations of N-ADJ agreement and seven
were sensitive to N-DROP agreement. The greatest average reading time change was in the
DET-N condition (155 ms), followed by N-ADJ agreement (45 ms). Even though more
participants in the N-DROP condition were sensitive to the violations of agreement than in the
N-ADJ condition when considering the 140 ms threshold, the participants in the N-DROP
condition read the grammatical sentences on average at equal speed (-3 ms).
These data show a great deal of variability at the individual level: some participants did
not show sensitivity to violations of grammatical gender agreement in any of the three conditions
(8 participants), some were sensitive only to DET-N agreement (7), only to N-ADJ agreement
(2), only to N-DROP agreement (1), to both DET-N and N-ADJ agreement (1), to both DET-N
and N-DROP agreement (4), or to all three types of agreement (2).
Comparison of L2 Learners’ Reported Sensitivity and Individual Reading Times
I triangulated the L2 learners’ responses to questions 1, 2, 3, 4, and 5 on the reading
questionnaire with their individual reading times to determine whether their reported sensitivity
was reflected in their reading times. The results are in Table 4.24. Question 1 asked if
participants noticed anything strange about the experimental sentences, question 2 asked if they
saw any grammatical errors, question 3 asked participants to list any grammatical errors they saw
in a free recall fashion, question 4 asked participants to check off the errors they saw from a list,

	
  

85

and question 5 asked participants to report the percentage of sentences they believe contained an
error.
Table 4.24
Comparison of Reported Sensitivity and Individual Change in Reading Time for L2 Spanish
Learners
DET-N
N-ADJ
N-DROP
∆
∆
∆
Reading
Reading
Reading
Q5
Participant
Q1
Q3
Time
Q4
Time
Q4
Time
Q4 (%)
A
/
/
-299
/
-12
/
-470
/
/
B
/
/
-265
/
113
/
132
/
/
C
ambiguous
DET-N
-158
yes
-213
yes
-24
yes 25
D
DET-N
DET-N
-84
yes
136
yes
98
yes 60
E
DET-N
DET-N
-62
yes
8
yes
-26
no
25
F
none
N-ADJ
-60
yes
78
yes
132
yes
/
G
/
/
-10
/
56
/
-86
/
/
H
/
/
57
/
-9
/
-17
/
/
I
none
other
-46
yes
-2
no
146
no
4
J
none
DET-N
-26
yes
174
yes
-338
yes 10
K
/
/
-15
/
143
/
-101
/
/
L
none
other
143
yes
62
no
-25
no
3
M
none
other
176
yes
-2
yes
-194
yes 40
N
ambiguous
DET-N
262
yes
16
no
13
yes 20
O
none
other
290
no
20
no
-420
no
10
P
/
/
291
/
-83
/
-33
/
/
Q
ambiguous
DET-N
502
yes
106
yes
-120
yes 75
R
DET-N
DET-N
564
yes
120
no
-217
no
25
S
none
ambiguous
174
yes
-22
yes
155
no
25
T
DET-N
DET-N
178
yes
-155
yes
323
yes 40
U
DET-N
DET-N
495
no
12
yes
171
yes 35
V
DET-N
DET-N
537
yes
-63
no
183
no
20
W
none
DET-N
188
yes
249
yes
-156
yes 40
X
none
other
236
no
226
yes
564
no
20
Y
DET-N
DET-N
815
yes
164
yes
224
yes 50
Mean
155
45
-3
29
Note. All times in are milliseconds. Any difference over 140 ms is highlighted in gray. A slash
(/) indicates that participants indicated on Q 2 that they did not notice any grammatical violations
in the experimental stimuli. For Q1 and Q 3, “none” indicates that the learners did not report
seeing any agreement errors; “other” indicates that participants did not report seeing gender
agreement errors, but rather a different error found in the distractor sentences; and “ambiguous”
indicates that the participant reported seeing gender agreement errors, but did not specify the
syntactic context.

	
  

86

In general, L2 learners that did not report any grammatical violations (Q2 – those
indicated in the table with a slash) were not sensitive to the violations in their reading times;
however, reported sensitivity to violations of grammatical gender agreement was not always
manifested in reading times.
For question 1, of the 10 participants that reported seeing gender agreement errors, seven
reached the 140 ms threshold of sensitivity in their reading times on at least one of the three
conditions while three did not. Of the seven participants that specifically reported seeing DET-N
errors, five evidenced DET-N sensitivity in their reading times while two did not. For question 3,
of the 12 participants that reported seeing DET-N agreement violations, eight evidenced
sensitivity in their DET-N reading times while four did not. One participant reported seeing NADJ violations for question 3, but did not reach the 140 ms threshold. The participant that
mentioned gender errors but did not specify a context (i.e., ‘ambiguous’) was sensitive to DET-N
and N-DROP agreement, but not N-ADJ agreement. For question 4, sensitivity in reading times
coincided with the L2 learners’ reported sensitivity 57% of the time, and when it did not,
participants were more likely to over-report (72%) than under-report (28%) sensitivity. For
question 5, the percentages that the participants reported must be interpreted with caution, as
participants estimated the percentage of all the violations they saw during the experiment, both in
terms of target items and distractors.
Summary of Results
The results of the reading questionnaire indicate the all the native speakers and most of
the L2 learners reported seeing errors during the experiment. Of the three gender agreement
violations under investigation, DET-N agreement errors were the most salient. In terms of
individual reading times, there was a great deal of variability across participants. Reading times

	
  

87

indicated that the L2 learners were most sensitive to DET-N violations, but performed relatively
similarly on N-ADJ and N-DROP items. Finally, the L2 learners’ reading times coincided with
their reported awareness on the reading questionnaire often, but not always. The results from the
L2 learners’ reading times and reported sensitivity coincided in that the DET-N violations were
the most salient both in terms of reading times and reported sensitivity on the reading
questionnaire.

	
  

88

CHAPTER 5: GENERAL DISCUSSION
Summary and Discussion of the Findings
The first two research questions that guided this dissertation were whether native Spanish
speakers and L2 learners are sensitive to violations of DET-N, N-ADJ and N-DROP gender
agreement during an online processing task and whether any evidenced sensitivity to the
grammatical violations was contingent on the type of agreement under investigation. In the
primary statistical analyses, the native Spanish speakers showed sensitivity to violations of
grammatical gender agreement in all three syntactic contexts. The native speakers’ sensitivity
was evident in their first-pass, go-past and total reading times. L2 learners, on the other hand,
were only sensitive to violations of DET-N agreement, and this sensitivity was evident in their
first-pass, go-past and total reading times. It is unsurprising that sensitivity to violations of
agreement was not evident during first-fixation because, these times generally reflect word
identification processes (Clifton et al., 2007), and some previous research has also not found
sensitivity to violations of grammatical gender agreement in first fixation times (e.g., Keating,
2009). Even though participants did not evidence sensitivity to violations of grammatical gender
agreement during first fixation, they did evidence sensitivity during first-pass time, which, like
first fixation, is considered an early-processing measure (Pickering et al., 2004). The results of
this study seem to suggest that when sensitivity to violations did occur, it was robust and
happened during both early and late stages of processing.
Taken together, these results provide preliminary evidence that the L2 learners do not
have a representational deficit for grammatical gender agreement because, although they are not
sensitive to the violations in all conditions, they are sensitive in at least one, according to the
statistical analyses. Furthermore, an analysis of individual L2 participants’ performance

	
  

89

indicated that at least 50% of the participants showed sensitivity to DET-N violations, suggesting
some form of representation. The findings of this study therefore generally align with nonrepresentational deficit approaches (e.g., Prévost & White, 2000; Schwartz & Sprouse, 1996) and
suggest that learners can indeed acquire functional features that are not instantiated in their L1.
The additional statistical analyses examined whether the L2 learners exhibited evidence
of a masculine default (e.g., McCarthy, 2008; White et al., 2004). While the L2 learners
exhibited sensitivity to agreement violations in the DET-N condition for go-past time and total
time, this sensitivity did not vary depending on the gender of the target noun. For the N-ADJ and
N-DROP conditions, the L2 learners showed no sensitivity to agreement violations, regardless of
the gender of the noun. Therefore, the L2 learners’ reading times did not show any evidence of
default morphology. Some researchers argue that L2 learners’ underspecification errors during
production, although non-nativelike, do not necessarily suggest a representational deficit,
because the pattern of errors may be taken as evidence of a functioning system (e.g., White et al.,
2004). For comprehension studies that examine native speakers’ online processing of agreement,
previous studies yield contradictory findings: while some have found that native speakers
evidence an asymetrical representation for gender (e.g., Alemán Bañón & Rothman, 2016;
Romanova & Gor, 2017), others have not (e.g., Acuña-Fariña, Meseguer, & Carreiras, 2014).
Therefore, the L2 learners’ lack of gender agreement asymmetry during online processing in this
dissertation does not provide clear evidence for or against a non-representational account.
Even though the L2 learners showed some sensitivity to the violations in the DET-N
condition, they did not perform like native speakers in the N-ADJ and N-DROP condition. I
predicted that L2 learners would evidence the greatest sensitivity to violations of DET-N
agreement followed by N-ADJ and then N-DROP. This prediction was only partially borne out

	
  

90

in the data. The statistical results indicated that participants were sensitive to violations of DETN agreement, but were not sensitive to violations of N-ADJ and N-DROP agreement. When I
analyzed L2 learners’ responses individually, the participants showed the greatest sensitivity to
violations in DET-N agreement both in terms of the number of participants deemed sensitive to
the violations (14) and the average amount of sensitivity across participants (155 ms). However,
like the statistical analyses, the individual data could not determine whether the N-ADJ or NDROP condition showed higher rates of sensitivity: while the L2 learners had a greater average
sensitivity for violations of N-ADJ agreement (45 ms), a greater number of participants read the
ungrammatical areas of interest at least 140 ms more slowly in the N-DROP (seven in the NDROP condition compared to five in the N-ADJ condition).
The L2 learners’ sensitivity to DET-N agreement violations, but not N-ADJ violations is
also paralleled in production studies, where learners are more accurate on DET-N agreement
(Bruhn de Garavito & White, 2002; Franceschina, 2001; White et al., 2004). The L2 learners’
lack of sensitivity to N-ADJ agreement could be due to a processing constraint: unlike the DETN condition, the N-ADJ contains an AP which makes it longer, more complex, and therefore
perhaps more difficult to process (Spinner & Juffs, 2008).
A processing constraint could also explain the lack of sensitivity to violations in the NDROP condition. For example, in an eye-tracking study, Keating (2009) tested native Spanish
speakers’ and beginning, intermediate and advanced L2 Spanish learners’ (L1 = English)
sensitivity to grammatical gender violations on postnominal adjectives located in three syntactic
domains: the DP, the VP and a subordinate clause, thereby manipulating both the structural and
linear distance between the nouns and adjectives. Keating found that while the native speakers
were sensitive to the violations in all three conditions, the beginner and intermediate learners

	
  

91

were not26, and the advanced L2 learners showed sensitivity only to violations in the DP. Keating
interprets the learners’ sensitivity in the DP as evidence that they do not have a representational
deficit for gender agreement, and ascribes their lack of sensitivity in the VP and subordinate
clause as shallow processing (Clahsen & Felser, 2006a, 2006b). The deficit, he argues, is not one
of competence, but rather of processing.
Another possible explanation for the L2 learners’ lack of sensitivity in the N-DROP
condition is that the participants had to compute a syntactic dependency to recover the referent of
the null nominal across a linear distance; however, it is unclear whether in the ungrammatical
condition the L2 learners (a) were not sensitive to the agreement violation and were therefore
recovering the correct antecedent and incorrectly linking it to the pro in the null nominal, or (b)
were linking a different antecedent to the pro in the null nominal that does agree in gender.
To illustrate this, consider (12) below. It is possible that the participants either (a) linked
the masculine marked null nominal (elMASC fríoMASC) to the correct antecedent (laFEM bebidaFEM)
because they are not sensitive to the ungrammaticality or, (b) linked the null nominal to an
incorrect antecedent (e.g., su amigoMASC) because they fail to comprehend the sentence. Example
(a) would suggest problems computing agreement, while in example (b) that is not necessarily
the case.

	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  
26
	
  The L2 learners in this dissertation were likely somewhat similar in proficiency to the
intermediate L2 learners in Keating’s study. Because Keating’s intermediate learners were not
sensitive to any violations, he posits that gender is acquired late (p. 525); however, the results of
this dissertation suggest that gender is acquired earlier, but is only evidenced in DET-N
agreement, not N-ADJ agreement.	
  
	
  

92

(12)

El hombre toma la bebida caliente y su amigo toma *el frío cuando van al
restaurante.
The man drinks theFEM drinkFEM hot and his friendMASC drinks *theMASC smallMASC
(one) when (they) go to the restaurant.

During the vocabulary posttest, the participants were asked to translate two Spanish
sentences that contained null nominals into English to ensure that they could comprehend these
types of sentences. While all the L2 learners translated the sentences correctly, the sentences they
were asked to translate contained only correct agreement, which does not resolve what the L2
learners would do when the null nominal contained a violation. It is feasible that the L2 learners
did not recover any referent at all, which would yield similar reading times on both grammatical
and ungrammatical sentences27. Therefore, in this study, the difficulty for L2 learners may not be
agreement, but rather pro indexing. Recovering referents can be quite difficult for both L1 and
L2 language learners. For example, Shin & Cairns (2012) determined that Spanish-speaking
children do not develop completely ‘adult-like’ preferences for overt pronouns in switchreference contexts until age 14. There was no clear developmental trend for children to develop
‘adult-like’ preferences for null pronouns in same-reference contexts for any of the ages tested in
the study (ages 6 to 15). These examples illustrate that for some structures, extensive input is
required to pattern like adult native speakers. This seems to suggest that the L2 learners in this
dissertation have simply not had enough input to be as sensitive to the violations as native
speakers in all three syntactic categories.

	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  
27
This may also explain the advanced learners’ lack of sensitivity to N-ADJ agreement
violations across clauses in Keating (2009), as those learners also had to compute a long-distance
syntactic dependency.
	
  

93

The third research question that guided this dissertation was whether L2 learners reported
seeing violations on a reading questionnaire administered after the eye-tracking experiment. On
question 2 of the reading questionnaire, all native speakers and roughly 75% of the L2 learners
reported seeing errors in the experimental sentences. The self-reports indicated that the most
salient type of errors for the L2 learners was DET-N agreement violations, which also coincided
with their reading times. Therefore, even though participants were not told ahead of time that the
sentences contained ungrammaticalities, and their attention was directed to meaning through the
inclusion of comprehension questions after every experimental sentence, they still developed
explicit awareness of the violations.
For the L2 learners, reported sensitivity (or lack thereof) on the reading questionnaire
(question 4) coincided with reading times only 57% of the time. The triangulation of
participants’ reading times with question 4 could results in four possible outcomes, either
participants evidenced sensitivity to violations: (a) both on reading times and on the self-report,
(b) on reading times but not on the self-report, (c) not on reading times, but on the self-report, or
(d) not on reading times or the self-report. In this dissertation, roughly 23% of the L2 learners’
responses on question 4 fell in category (a), 12% in (b), 31% in (c) and 35% in (d). Because the
goal of studies that examine L2 learners’ competence through sensitivity to grammatical
violations during processing is to measure a linguistic system that is abstract and implicit,
researchers often try to shift participants’ attention away from explicit reasoning by having them
focus on meaning28 (e.g., Jiang, 2004, 2007; Jiang et al., 2011; but see Godfroid et al., 2015).
This means that any scenario without awareness on the self-reports is ideal: either participants

	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  
28
Jiang (2007) and Jiang et al. (2011) discuss integrated and nonintegrated knowledge, which
are akin to implicit and explicit knowledge, respectively.
	
  

94

evidence sensitivity implicitly (b) or not at all (d). When participants report awareness, and are
sensitive in reading times as in (a), the type of knowledge they are using while processing
sentences is difficult to ascertain. Put differently, are the L2 learners sensitive to DET-N
violations because they are using metalinguistic knowledge to detect them, or is their sensitivity
first detected implicitly, and then rise to the level of metalinguistic awareness? For the native
speakers, it is likely that implicit knowledge gives rise to a metalinguistic recognition of the
errors in the sentences; however, whether this directionality is also true for L2 learners is an open
question. One telling finding is that although the L2 learners evidenced sensitivity in the DET-N
condition, they did not in the N-ADJ condition, even though both are explicitly taught in
language classrooms. If metalinguistic knowledge were driving sensitivity, then they should
show sensitivity on reading times in both conditions. Therefore, it seems likely that the L2
learners’ sensitivity on reading times is primarily a reflection of their underlying competence,
rather than metalinguistic knowledge.
That said, to ensure that L2 learners’ explicit reasoning during processing stems from
their underlying system and not the nature of the task they are engaged in, researchers should
employ safeguards when creating experiments. Godfroid and Winke (2015) note that “whether
eye movements signal implicit or explicit processes will depend to a large extent on the
experimental design, target structure, and research questions of the studies involved” (p. 340).
Some specific experimental design questions that L2 researchers should pay attention to are: (a)
the ratio of distractor/filler sentences to experimental sentences, (b) the percentage of
grammatical violations in the distractors, (c) the type of secondary distractor task, and (d) the
online methodology employed (for reviews see Jegerski, 2014; Keating, 2014; Marinis, 2010;
Roberts, 2012).

	
  

95

In this dissertation, the target sentences I tested represented 40% of all experimental
sentences, including distractors/fillers. While similar research has employed roughly the same
proportion of experimental sentences and distractors/fillers (e.g., Keating, 2009; VanPatten et al.,
2012), more research is needed to determine how and if the percentage of distractors affects
sensitivity to violations in experimental sentences. Half of the distractors in this study also
contained ungrammaticalities, again coinciding with similar studies (Coughlin & Tremblay,
2013; VanPatten et al., 2012), but this may have also heightened participants’ awareness to the
violations in the experimental sentences.
Detailed research is also needed as to whether the type of data collection tool researchers
implement affects how learners process sentences. Even though eye tracking may allow
participants to read more naturally than other online comprehension measures, such as self-paced
reading, the issue of reactivity still needs to be further explored (cf. Godfroid & Spino, 2015).
One potential pitfall is that because sentences are presented in their entirety, and participants are
usually not under a time pressure to read them (but see Godfroid et al., 2015), these eye-tracking
studies could invite more explicit reasoning than other types of sentence processing measures,
such as non-cumulative self-paced reading, because the sentences are not continuously projected
on the screen for reanalysis (Jiang, 2004; Marinis, 2010). Therefore, one could argue that eye
tracking is a more natural measure of reading (Dussias, 2010; Roberts, 2012; Witzel et al., 2012),
but other online methodologies may be more appropriate for processing research (for discussion
see Mitchell, 2004).
Implications of the Findings
The results of this dissertation indicate that L2 learners are indeed sensitive to gender
agreement violations during online processing and therefore likely do not have a representational

	
  

96

deficit for gender agreement. The results also indicate that syntactic context can indeed affect
sensitivity to grammatical gender agreement violations during online processing. Therefore,
researchers should select with care the type of agreement under investigation when assessing L2
learners’ competence, as certain types of agreement may be more difficult than others for
learners. Learners should be biased for the best when researchers are testing for the ability to
compute agreement, that is, the easiest type of agreement should be tested. If not, then it is
impossible to determine whether a lack of sensitivity to violations (i.e., during processing tasks)
or optionality (i.e., during production) is caused by a representational deficit or another
intervening variable, such as syntactic context.
The results also suggest that L2 learners may rely on explicit knowledge when processing
sentences that contain grammatical violations, even if their attention is directed to meaning. As a
result, online tests that use this experimental design may be measuring explicit knowledge in
addition to (or perhaps to the exclusion of) underlying competence. This was likely not a
problem in this dissertation, because, even though participants reported high levels of explicit
awareness of the violations, it did not make them sensitive in reading times to both syntactic
contexts that are taught explicitly (DET-N and N-ADJ agreement). However, researchers using
this experimental design to investigate representational deficits should carefully construct their
experimental stimuli to limit this potentially confounding variable.
Limitations
All research is limited to the targets at hand, the population at hand, and the conditions at
hand, and this dissertation is no exception. This dissertation would have benefitted from a larger
sample size as well as a more advanced L2 learner group to tease apart the difficulty of N-ADJ
and N-DROP agreement. It also would have benefitted from an independent test of working

	
  

97

memory, to determine if individual differences in working memory affected processing,
especially in the N-DROP condition (cf. Keating, 2010; Sagarra & Herschensohn, 2010b).
Another limitation involves the target nouns selected for the eye-tracking study. I
selected nouns with canonical –o and –a endings to bias learners with the best (for similar
methodological decisions see Keating, 2009; Sagarra & Herschensohn, 2010b); however, this
yielded near-ceiling performance on gender assignment in the vocabulary posttest, perhaps overestimating their ability to assign gender to the target nouns employed in this dissertation.
Implementing opaque nouns may have improved the quality of the experimental stimuli by
eliminating this confound.
Future Research
This section outlines two future areas of research related to the current study. One future
area of research would be to extend this study to investigate different types of grammatical
gender agreement (e.g., direct object agreement). This could be done with eye tracking, as in the
current study, or with another methodology such as self-paced reading or an oral production task.
Studies such as this one would indicate which types of agreement are more and less difficult for
L2 learners, which, in turn, can be used to reflect upon previous research investigating whether
L2 learners can acquire grammatical gender agreement in their L2 if it is not instantiated in their
L1 (e.g., McCarthy, 2008; Sagarra & Herschensohn, 2010a; White et al., 2004).
Another future potential area of research is methodological in nature. The extent to which
online processing methodologies measure underlying competence instead of explicit knowledge
likely depends on the design of the experiment and the construction of experimental stimuli.
Future research should investigate directly the extent to which both the stimuli design and the
online methodology chosen affect sensitivity to grammatical violations. These studies should

	
  

98

also triangulate eye-movement data with self-reported sensitivity, to determine what type of
knowledge participants are using to process the sentences.

	
  

99

APPENDICES

	
  

100

Appendix A

Pilot Vocabulary Test
1.  
2.  
3.  
4.  
5.  
6.  

Age:
Gender:
First Language(s)/Mother tongue(s):
How many semesters total have you taken Spanish in COLLEGE?
What is/are your major(s)? If you have two, please list both.
Are you a Spanish minor?
Yes/No
7.   At what age did you begin studying Spanish?
8.   How many years did you take Spanish for in HIGH SCHOOL?
9.   For how many years total have you actively studied Spanish?
10.  Have you visited a Spanish-speaking country? If so, how long were you there for and
what was the nature of your visit (e.g., vacation, study abroad, etc.)?
11.  Have you studied any language(s) other than Spanish? If so, explain which language(s)
you studied and for how long.
12.  On the next page you will be asked to translate and rate your knowledge of 32 Spanish
nouns. For each word on the next page please do the following:
A.) Translate the word into English (but do not look up the translation!)
B.) Rate your knowledge of the word on a scale of 4-0.
4: I know this word very well; I translated it correctly and rapidly.
3: I know this word somewhat well; I translated it correctly after some thought.
2: I'm unsure of this word; I'm unsure if my translation is correct.
1: I don't think I know this word; I don't think my translation is correct.
0: I definitely don't know this word; My translation is definitely incorrect.
English
Word Knowledge
Translation
4
3
2
1
0
abrigo
___
()
()
()
()
()
sombrero
___
()
()
()
()
()
cuchillo
___
()
()
()
()
()
trabajo
___
()
()
()
()
()
bebida
___
()
()
()
()
()
pregunta
___
()
()
()
()
()
maleta
___
()
()
()
()
()
comida
___
()
()
()
()
()
Please translate and rate your knowledge.
English
Word Knowledge
Translation
4
3
2
1
	
  

101

0

zapato
almuerzo
ensayo
cuaderno
revista
escuela
mochila
cerveza

___
___
___
___
___
___
___
___

()
()
()
()
()
()
()
()

()
()
()
()
()
()
()
()

()
()
()
()
()
()
()
()

()
()
()
()
()
()
()
()

()
()
()
()
()
()
()
()

Please translate and rate your knowledge.
English
Word Knowledge
Translation
4
3
2
1
regalo
___
()
()
()
()
dibujo
___
()
()
()
()
espejo
___
()
()
()
()
archivo
___
()
()
()
()
cocina
___
()
()
()
()
ventana
___
()
()
()
()
manzana
___
()
()
()
()
pintura
___
()
()
()
()

0
()
()
()
()
()
()
()
()

Please translate and rate your knowledge.
English
Word Knowledge
Translation
refresco
4
3
2
1
partido
___
()
()
()
()
mercado
___
()
()
()
()
vestido
___
()
()
()
()
camisa
___
()
()
()
()
piscina
___
()
()
()
()
corbata
___
()
()
()
()
iglesia
___
()
()
()
()
refresco
___
()
()
()
()

0
()
()
()
()
()
()
()
()

13.  Please decide whether each noun below is masculine or feminine.
Masculine Feminine
abrigo
()
()
sombrero
()
()
cuchillo
()
()
trabajo
()
()
zapato
()
()
almuerzo
()
()
ensayo
()
()
	
  

102

cuaderno
regalo
dibujo
espejo
archivo
refresco
partido
mercado
vestido
bebida
pregunta
maleta
comida
revista
escuela
mochila
cerveza
cocina
ventana
manzana
pintura
camisa
piscina
corbata
iglesia

	
  

()
()
()
()
()
()
()
()
()
()
()
()
()
()
()
()
()
()
()
()
()
()
()
()
()

()
()
()
()
()
()
()
()
()
()
()
()
()
()
()
()
()
()
()
()
()
()
()
()
()

103

Appendix B

Experimental Stimuli
Determiner-Noun Condition
1.   El niño se pone el/*la zapato durante la mañana antes de montarse en el autobús.
2.   El muchacho compra el/*la sombrero cuando viaja a México con su mamá y su papá.
3.   La madre prepara el/*la almuerzo durante la mañana para su hija y su esposo.
4.   El traductor termina el/*la trabajo cuando llega a la oficina a las nueve de la mañana.
5.   La mujer recuerda el/*la momento durante la reunión con su jefe y unos clientes.
6.   El chico bebe el/*la refresco cuando ve la película en el cine con su familia.
7.   La muchacha describe el/*la mercado cuando habla con su madre y su padre por teléfono.
8.   La profesora revisa el/*la ensayo durante sus horas de oficina a las tres de la tarde.
9.   La artista crea el/*la dibujo durante su clase de arte por la tarde con sus estudiantes.
10.  El estudiante termina el/*la proyecto cuando regresa a su casa muy tarde por la noche.
11.  El espectador bebe la/*el bebida durante el partido con sus amigos en el estadio.
12.  El padre visita la/*el escuela durante la noche para la reunión con el maestro.
13.  La muchacha compra la/*el camisa durante el fin de semana en el centro comercial.
14.  La mujer visita la/*el iglesia cuando tiene dinero en el banco para donar.
15.  El hombre usa la/*el guitarra cuando toca música con su banda en el bar.
16.  La cocinera prepara la/*el comida durante la mañana en la cocina con sus asistentes.
17.  La experta contesta la/*el pregunta durante la conferencia al final de su presentación.
18.  El niño rompe la/*el ventana cuando juega con su amigo por la tarde en el jardín.
19.  El abogado come la/*el manzana cuando habla con su cliente a las ocho de la mañana.

	
  

104

20.  La madre necesita la/*el cámara cuando va de vacaciones con su esposo y su bebé.
Noun-Adjective Condition
21.  El hombre busca el zapato rojo/*roja cuando se viste en la mañana para ir a trabajar.
22.  La muchacha lleva el sombrero rosado/*rosada durante el fin de semana en la discoteca.
23.  El estudiante come el almuerzo bueno/*buena cuando llega a la cafetería con sus amigos.
24.  La mujer termina el trabajo corto/*corta durante la noche en la biblioteca de la
universidad.
25.  La niña recuerda el momento bello/*bella cuando habla con su madre en la playa.
26.  El atleta toma el refresco frío/* fría cuando termina de correr tres millas por la mañana.
27.  La cocinera busca el mercado caro/*cara durante el fin de semana en la ciudad.
28.  El profesor lee el ensayo largo/*larga cuando tiene tiempo libre por la noche.
29.  El artista describe el dibujo/*bonita bonito durante la exhibición del Museo Reina Sofía.
30.  La muchacha empieza el proyecto nuevo/*nueva durante la mañana los jueves o viernes.
31.  El hombre compra la bebida barata/*barato cuando sale con sus amigos en la noche.
32.  La madre visita la escuela linda/*lindo durante la tarde para ver a sus hijos.
33.  La chica plancha la camisa limpia/*limpio cuando se despierta por la mañana a las siete.
34.  El niño dibuja la iglesia blanca/*blanco durante la clase en la hoja de papel.
35.  La muchacha vende la guitarra negra/*negro durante el fin de semana a su compañero.
36.  La niña come la comida mala/*malo durante el viaje con su madre y sus amigas.
37.  La esposa hace la pregunta tonta/*tonto cuando habla con su esposo en la noche.
38.  El padre limpia la ventana sucia/*sucio cuando trabaja afuera con su hijo en el jardín.
39.  El hombre compra la manzana pequeña/*pequeño cuando va a la tienda por la tarde.
40.  El muchacho vende la cámara vieja/*viejo durante la noche a alguien en Ebay.

	
  

105

Null Nominal Condition
41.  El chico lleva el zapato marrón y su hermanito lleva el blanco/*la blanca cuando van al
parque.
42.  La madre se pone el sombrero horrible pero su hija se pone el bonito/*la bonita durante el
partido.
43.  La mujer pide el almuerzo simple pero su amiga pide el caro/*la cara cuando salen.
44.  El estudiante escribe el trabajo breve y su compañero escribe el largo/*la larga durante la
clase.
45.  El hombre describe el momento ideal y su tío describe el malo/*la mala durante la
conversación.
46.  El jefe pide el refresco grande y su empleado pide el pequeño/* la pequeña cuando van a
McDonald’s.
47.  La muchacha visita el mercado normal y su abuela visita el lindo/*la linda durante el día.
48.  La estudiante lee el ensayo horrible y su amiga lee el bueno/*la buena cuando hacen la
tarea.
49.  El pintor ve el dibujo reciente y su estudiante ve el viejo/*la vieja cuando estudian.
50.  El profesor describe el proyecto interesante y el chico describe el tonto/*la tonta durante
la reunión.
51.  El hombre toma la bebida caliente y su amigo toma la fría/*el frío cuando van al
restaurante.
52.  La niña prefiere la escuela horrible pero su mamá prefiere la limpia/*el limpio durante las
visitas.
53.  La madre lleva la camisa elegante y su hija lleva la sucia/*el sucio cuando salen.
54.  La mujer visita la iglesia normal y su hermana visita la bella/*el bello durante la semana.
55.  El músico toca la guitarra azul pero su amigo toca la rosada/*el rosado durante el
concierto.
56.  La mujer come la comida especial y su hermana come la barata/*el barato cuando cenan.
57.  El estudiante hace la pregunta interminable y su amigo hace la corta/*el corto durante la
clase.

	
  

106

58.  El arquitecto quiere la ventana tradicional y el cliente quiere la nueva/*el nuevo durante
la cita.
59.  La niña come la manzana verde pero su hermana come la roja/*el rojo cuando desayunan.
60.  El muchacho usa la cámara gris pero el fotógrafo usa la negra/*el negro cuando sacan
fotos.

	
  

107

Appendix C

L2 Learner Background Questionnaire
1.  
1.  
2.  
3.  
4.  

5.  
6.  
7.  
8.  
9.  

	
  

What is your participant number?
Age:
Gender:
What is your native language?
Year in college:
( ) Freshman
( ) Sophomore
( ) Junior
( ) Senior (4th year)
( ) Senior (5th+ year)
( ) Other: _________________________________________________
What is/are your major(s)? If you have two, please list both.
If you are a Spanish major, how many Spanish classes (NOT including the ones you're
enrolled in now) do you still need to take to finish your Spanish Major?
Are you a Spanish minor?
When do you expect to graduate from MSU?
Please select all the courses you have previously taken or are currently taking at MSU:
[ ] SPN 101 Elementary Spanish I
[ ] SPN 102 Elementary Spanish II
[ ] SPN 150 Review of Elementary Spanish
[ ] SPN 201 Second Year Spanish I
[ ] SPN 202 Second Year Spanish II
[ ] SPN 310 Basic Spanish Grammar
[ ] SPN 320 Cultural Readings and Composition
[ ] SPN 330 Phonetics and Pronunciation
[ ] SPN 342 Media and Conversation
[ ] SPN 350 Introduction to Reading Hispanic Literature
[ ] SPN 412 Topics in Hispanic Culture
[ ] SPN 420 Spain and its Literature
[ ] SPN 432 Latin America and its Literature
[ ] SPN 440 The Structure of Spanish
[ ] SPN 452 Topics in Spanish Language I
[ ] SPN 462 Topics in Spanish Literature
[ ] SPN 472 Topics in the Literatures of the Americas
[ ] SPN 482 Topics in Spanish Linguistics
[ ] SPN 490 Independent Study
[ ] SPN 491 Special Topics in Spanish
[ ] SPN 492 Senior Writing Project
[ ] Other(s): _________________________________________________

108

10.  How many SEMESTERS of Spanish have you taken in COLLEGE (including this
semester)?
11.  For how many YEARS did you take Spanish in HIGH SCHOOL?
12.  For many YEARS did you take Spanish in MIDDLE SCHOOL?
13.  At what age did you take your first Spanish class? (Please only consider classes that met
for more than an hour or two a week).
14.  For how many years total have you actively studied Spanish?
15.  How many hours per week do you currently spend...
Speaking Spanish: _____
Reading Spanish: _____
Listening to Spanish: _____
Writing in Spanish: _____
16.  Have you ever studied abroad in a Spanish-speaking country?
17.  Have you ever vacationed in a Spanish speaking country?
If yes, WHERE did
WHEN did you
FOR HOW LONG
you travel?
travel?
did you travel?
Yes No
Response ( ) ( )
_____
_____
_____
18.  Have you ever completed an internship or service learning experience in a Spanishspeaking environment?
[ ] No, neither.
[ ] Yes, an internship.
[ ] Yes, a service learning experience.
[ ] If you selected yes, please briefly describe your experience.:
_________________________________________________
19.  What languages besides Spanish have you studied? (Leave blank if you have not studied
any other language).
Language For how many years did you study this language?
Language 1 _____
_____
Language 2 _____
_____
Language 3 _____
_____
Language 4 _____
_____
20.  Where did you complete your study abroad?
21.  What month and year did you LEAVE to travel abroad?
22.  What month and year did you RETURN FROM traveling abroad?
23.  How long did you study abroad for?
24.  What best describes your living situation?
[ ] I lived with a host family.
[ ] I lived in a dorm with other Americans.
[ ] I lived in a dorm with native-Spanish speakers.
[ ] Other: _________________________________________________
25.  What percentage of the time did you use Spanish (as opposed to English or any other
language) while living abroad (this includes speaking Spanish, watching Spanish TV,
reading in Spanish, conversing in Spanish etc.)?
26.  How did the study abroad experience affect your Spanish abilities?
( ) 7 My Spanish was MUCH BETTER when I returned to the US.
	
  

109

()6
()5
( ) 4 My Spanish was THE SAME when I returned to the US.
()3
()2
( ) 1 My Spanish was MUCH WORSE when I returned to the US.
27.  Compare your Spanish abilities now to when you first arrived home from your study
abroad experience. How would you compare your Spanish abilities now and then?*
( ) 7 My Spanish is MUCH BETTER NOW than when I returned to the US.
()6
()5
( ) 4 My Spanish is THE SAME NOW as when I returned to the US.
()3
()2
( ) 1 My Spanish is MUCH WORSE NOW than when I returned to the US.
28.  Please list all the languages you know in order of dominance. Be sure to include your
native language in the list:
29.  Please list all the languages you know in order of acquisition (your native language first):
30.  Please list the percentage of the time you are currently and on average exposed to each
language.
(Your percentages should add up to 100%)
31.  Do you wear glasses or contacts?
( ) Yes
( ) No
32.  Are you currently having problems reading from books or computer screens that could
negatively impact your performance on the tests you are taking today?
( ) Yes
( ) No
33.  On a scale from zero to ten, please select your level of proficiency in speaking,
understanding, and reading SPANISH from the pull-down menus:
Speaking
Understanding spoken
Reading
language
( ) 0 - none
( ) 0 - none
( ) 0 - none
( ) 1 - very low
( ) 1 - very low
( ) 1 - very low
( ) 2 - low
( ) 2 - low
( ) 2 - low
( ) 3 - fair
( ) 3 - fair
( ) 3 - fair
( ) 4 - slightly less than
( ) 4 - slightly less than
( ) 4 - slightly less than
adequate
adequate
adequate
( ) 5 - adequate
( ) 5 - adequate
( ) 5 - adequate
( ) 6 - slightly more than
( ) 6 - slightly more than
( ) 6 - slightly more than
adequate
adequate
adequate
( ) 7 - good
( ) 7 - good
( ) 7 - good
( ) 8 - very good
( ) 8 - very good
( ) 8 - very good
( ) 9 - excellent
( ) 9 - excellent
( ) 9 - excellent
( ) 10 - perfect
( ) 10 - perfect
( ) 10 - perfect

	
  

110

34.  On a scale from zero to ten, please select how much the following factors contributed to
you learning SPANISH:
Interacting with friends
Interacting with family
Reading
( ) 0 - not a contributor
( ) 0 - not a contributor
( ) 0 - not a contributor
( ) 1 - minimal contributor ( ) 1 - minimal contributor ( ) 1 - minimal contributor
()2
()2
()2
()3
()3
()3
()4
()4
()4
( ) 5 - moderate contributor ( ) 5 - moderate contributor ( ) 5 - moderate contributor
()6
()6
()6
()7
()7
()7
()8
()8
()8
()9
()9
()9
( ) 10 - most important
( ) 10 - most important
( ) 10 - most important
contributor
contributor
contributor
Self instruction
Watching TV
Listening to the
radio/music
( ) 0 - not a contributor
( ) 0 - not a contributor
( ) 0 - not a contributor
( ) 1 - minimal contributor ( ) 1 - minimal contributor ( ) 1 - minimal contributor
()2
()2
()2
()3
()3
()3
()4
()4
()4
( ) 5 - moderate contributor ( ) 5 - moderate contributor ( ) 5 - moderate contributor
()6
()6
()6
()7
()7
()7
()8
()8
()8
()9
()9
()9
( ) 10 - most important
( ) 10 - most important
( ) 10 - most important
contributor
contributor
contributor
35.  Please rate to what extent you are currently exposed to SPANISH in the following
contexts:*
Interacting with friends
Interacting with family
Reading
( ) 0 – never
( ) 0 – never
( ) 0 – never
( ) 1 – almost never
( ) 1 – almost never
( ) 1 – almost never
()2
()2
()2
()3
()3
()3
()4
()4
()4
( ) 5 – half of the time
( ) 5 – half of the time
( ) 5 – half of the time
()6
()6
()6
()7
()7
()7
()8
()8
()8
()9
()9
()9
( ) 10 - always
( ) 10 - always
( ) 10 - always
Self instruction
Watching TV
Listening to the
radio/music
( ) 0 – never
( ) 0 – never
( ) 0 – never

	
  

111

( ) 1 – almost never
( ) 1 – almost never
( ) 1 – almost never
()2
()2
()2
()3
()3
()3
()4
()4
()4
( ) 5 – half of the time
( ) 5 – half of the time
( ) 5 – half of the time
()6
()6
()6
()7
()7
()7
()8
()8
()8
()9
()9
()9
( ) 10 - always
( ) 10 - always
( ) 10 - always
36.  In your perception, how much of a non-native accent do you have in SPANISH?
( ) 0 - none
( ) 1 - almost none
( ) 2 - very light
( ) 3 - light
( ) 4 - some
( ) 5 - moderate
( ) 6 - considerable
( ) 7 - heavy
( ) 8 - very heavy
( ) 9 - extremely heavy
( ) 10 - pervasive
37.  Please rate how frequently others identify you as a non-native speaker based on your
accent in SPANISH:
( ) 0 - never
( ) 1 - almost never
()2
()3
()4
( ) 5 - half of the time
()6
()7
()8
()9
( ) 10 - always

	
  

112

Appendix D

Native Spanish Speaker Background Questionnaire
1.  
2.  
3.  
4.  

What is your participant number?
Age:
Gender:
What is your occupation? If you are a student, please type in your area of study (e.g.,
Spanish literature, mathematics, etc.).
5.   Are you a Spanish TA, instructor or professor?
Yes/No
6.   If you answered yes to the previous question, for how many years have you taught
Spanish?
7.   What languages have you studied and for how long?
Language For how many years did you study this language?
Language 1 _____
_____
Language 2 _____
_____
Language 3 _____
_____
Language 4 _____
_____
8.   Please list all the languages you know (including your native langauge) in order of
dominance:
9.   Please list all the languages you know in order of acquisition (your native language first):
10.  Please list the percentage of the time you are currently and on average exposed to each
language. (Your percentages should add up to 100%)
11.  When choosing to read a text available in all of your languages, in what percentage of
cases would you choose to read it in each of your languages? Assume that the original
was written in another language, which is unknown to you.
12.  When choosing a language to speak with a person who is equally fluent in all your
languages, what percentage of the time would you choose to speak each language? Please
report percent of total time.
13.  Please check your highest education level (or the approximate US equivalent to a degree
obtained in another country):
( ) Less than High School
( ) High School
( ) Professional Training
( ) Some College
( ) College
( ) Some Graduate School
( ) Master's
( ) Ph.D./M.D./J.D.
( ) Other: _________________________________________________
14.  In what country were you born?
15.  At what age did you immigrate to the US?
16.  Do you wear glasses or contacts?

	
  

113

Yes/No
17.  Are you currently having problems reading from books or computer screens that could
negatively impact your performance on the tests you are taking today?
Yes/No
18.  Age when you...
Age
Began acquiring ENGLISH:
___
Become fluent in ENGLISH:
___
Began reading in ENGLISH:
___
Became fluent reading in ENGLISH: ___
19.  Please list the number of years and months you have spent in each language learning
environment:
Years	
   Months	
  
	
  
A country where ENGLISH is spoken:	
  
___	
  
___	
  
A family where ENGLISH is spoken:	
  
___	
  
___	
  
A school and/or working environment where ENGLISH is spoken:	
   ___	
  
___	
  
20.  On a scale from zero to ten, please select your level of proficiency in speaking,
understanding, and reading ENGLISH from the pull-down menus:
Speaking
Understanding spoken
Reading
language
( ) 0 - none
( ) 0 - none
( ) 0 - none
( ) 1 - very low
( ) 1 - very low
( ) 1 - very low
( ) 2 - low
( ) 2 - low
( ) 2 - low
( ) 3 - fair
( ) 3 - fair
( ) 3 - fair
( ) 4 - slightly less than
( ) 4 - slightly less than
( ) 4 - slightly less than
adequate
adequate
adequate
( ) 5 - adequate
( ) 5 - adequate
( ) 5 - adequate
( ) 6 - slightly more than
( ) 6 - slightly more than
( ) 6 - slightly more than
adequate
adequate
adequate
( ) 7 - good
( ) 7 - good
( ) 7 - good
( ) 8 - very good
( ) 8 - very good
( ) 8 - very good
( ) 9 - excellent
( ) 9 - excellent
( ) 9 - excellent
( ) 10 - perfect
( ) 10 - perfect
( ) 10 - perfect
21.  On a scale from zero to ten, please select how much the following factors contributed to
you learning ENGLISH:
Interacting with friends
Interacting with family
Reading
( ) 0 - not a contributor
( ) 0 - not a contributor
( ) 0 - not a contributor
( ) 1 - minimal contributor ( ) 1 - minimal contributor ( ) 1 - minimal contributor
()2
()2
()2
()3
()3
()3
()4
()4
()4
( ) 5 - moderate contributor ( ) 5 - moderate contributor ( ) 5 - moderate contributor
()6
()6
()6
()7
()7
()7
()8
()8
()8
()9
()9
()9
	
  

114

( ) 10 - most important
contributor
Self instruction

( ) 10 - most important
contributor
Listening to the
radio/music
( ) 0 - not a contributor
( ) 0 - not a contributor
( ) 0 - not a contributor
( ) 1 - minimal contributor ( ) 1 - minimal contributor ( ) 1 - minimal contributor
()2
()2
()2
()3
()3
()3
()4
()4
()4
( ) 5 - moderate contributor ( ) 5 - moderate contributor ( ) 5 - moderate contributor
()6
()6
()6
()7
()7
()7
()8
()8
()8
()9
()9
()9
( ) 10 - most important
( ) 10 - most important
( ) 10 - most important
contributor
contributor
contributor
22.  Please rate to what extent you are currently exposed to ENGLISH in the following
contexts:
Interacting with friends
Interacting with family
Reading
( ) 0 – never
( ) 0 – never
( ) 0 – never
( ) 1 – almost never
( ) 1 – almost never
( ) 1 – almost never
()2
()2
()2
()3
()3
()3
()4
()4
()4
( ) 5 – half of the time
( ) 5 – half of the time
( ) 5 – half of the time
()6
()6
()6
()7
()7
()7
()8
()8
()8
()9
()9
()9
( ) 10 - always
( ) 10 - always
( ) 10 - always
Self instruction
Watching TV
Listening to the
radio/music
( ) 0 – never
( ) 0 – never
( ) 0 – never
( ) 1 – almost never
( ) 1 – almost never
( ) 1 – almost never
()2
()2
()2
()3
()3
()3
()4
()4
()4
( ) 5 – half of the time
( ) 5 – half of the time
( ) 5 – half of the time
()6
()6
()6
()7
()7
()7
()8
()8
()8
()9
()9
()9
( ) 10 - always
( ) 10 - always
( ) 10 - always
23.  In your perception, how much of a foreign accent do you have in ENGLISH?*
( ) 0 - none
( ) 1 - almost none

	
  

( ) 10 - most important
contributor
Watching TV

115

( ) 2 - very light
( ) 3 - light
( ) 4 - some
( ) 5 - moderate
( ) 6 - considerable
( ) 7 - heavy
( ) 8 - very heavy
( ) 9 - extremely heavy
( ) 10 - pervasive
24.  Please rate how frequently others identify you as a non-native speaker based on
your accent in ENGLISH:
( ) 0 - never
( ) 1 - almost never
()2
()3
()4
( ) 5 - half of the time
()6
()7
()8
()9
( ) 10 - always

	
  

116

Appendix E

Vocabulary Posttest
1.   What is your participant number?
2.   Please mark whether each word is masculine or feminine, and then translate each word
into English.
If you are unsure about a translation, feel free to guess. If you have no idea what a
particular translation is, write "I don't know".
Gender
English Translation
Masculine Feminine
zapato
()
()
___
sombrero
()
()
___
almuerzo
()
()
___
trabajo
()
()
___
momento
()
()
___
bebida
()
()
___
escuela
()
()
___
camisa
()
()
___
iglesia
()
()
___
guitarra
()
()
___
refresco
()
()
___
mercado
()
()
___
ensayo
()
()
___
dibujo
()
()
___
proyecto
()
()
___
comida
()
()
___
pregunta
()
()
___
ventana
()
()
___
manzana
()
()
___
cámara
()
()
___
3.   Please translate the following two sentences into English to the best of your ability:
El jefe pide el refresco grande y su empleado pide el pequeño cuando van a McDonald’s.
____________________________________________
La niña come la manzana verde pero su hermana come la roja cuando desayunan.
____________________________________________

	
  

117

Appendix F

Additional Gender Analyses
The L2 learners’ model results for the gender analyses in the N-ADJ (Tables 6.1 to 6.4)
and N-DROP (Tables 6.5 to 6.8) conditions are presented below. For all analyses, gender and
grammaticality are both categorical variables with two levels. L2 learner reading times on
grammatical sentences with a masculine target noun were taken as the reference category
(Intercept).
Table 6.1
Gender Model Results for L2 Learners’ First Fixation Duration in N-ADJ Condition
Description
Predictor
Coefficient
Standard
t
(β)
Error
Overall effect of reading a
Intercept
5.62
0.03
179.26
grammatical sentence with a
masculine target noun
Overall main effect of
Gender
0.03
0.04
0.69
reading a sentence with a
feminine target noun
Overall main effect of
Grammaticality
-0.02
0.04
-0.62
reading an ungrammatical
sentence
Random Effects
Subject
0.000
Item
0.001
Residual
0.142

	
  

118

p
< .001

= .505
= .536

Table 6.2
Gender Model Results for L2 Learners’ First-Pass Time in N-ADJ Condition
Description
Predictor
Coefficient Standard
t
(β)
Error
Overall effect of reading a
Intercept
5.77
0.05
115.15
grammatical sentence with a
masculine target noun
Overall main effect of
Gender
0.10
0.06
1.58
reading a sentence with a
feminine target noun
Overall main effect of
Grammaticality
-0.06
0.04
-1.72
reading an ungrammatical
sentence
Random Effects
Subject
0.008
Item
0.012
Residual
0.155

Table 6.3
Gender Model Results for L2 Learners’ Go-Past Time in N-ADJ Condition
Description
Predictor
Coefficient
Standard
(β)
Error
Overall effect of reading a
Intercept
5.95
0.06
grammatical sentence with a
masculine target noun
Overall main effect of
Gender
0.08
0.07
reading a sentence with a
feminine target noun
Overall main effect of
Grammaticality
-0.03
0.05
reading an ungrammatical
sentence
Random Effects
Subject
0.031
Item
0.014
Residual
0.236

	
  

119

p
< .001
= .130
= .087

t

p

92.56

< .001

1.19

= .250

-0.66

= .509

Table 6.4
Gender Model Results for L2 Learners’ Total Time in N-ADJ Condition
Description
Predictor
Coefficient
Standard
(β)
Error
Overall effect of reading a
Intercept
6.13
0.08
grammatical sentence with a
masculine target noun
Overall main effect of
Gender
0.04
0.08
reading a sentence with a
feminine target noun
Overall main effect of
Grammaticality
0.03
0.04
reading an ungrammatical
sentence
Random Effects
Subject
0.052
Item
0.022
Residual
0.228

t

p

80.60

< .001

0.44

= .662

0.66

= .512

Table 6.5
Gender Model Results for L2 Learners’ First Fixation Duration in N-DROP Condition
Description
Predictor
Coefficient
Standard
t
(β)
Error
Overall effect of reading a
Intercept
5.53
0.03
176.48
grammatical sentence with a
masculine target noun
Overall main effect of
Gender
-0.03
0.03
-0.87
reading a sentence with a
feminine target noun
Overall main effect of
Grammaticality
-0.00
0.03
-0.05
reading an ungrammatical
sentence
Random Effects
Subject
0.009
Item
0.000
Residual
0.100

	
  

120

p
< .001
= .383
= .961

Table 6.6
Gender Model Results for L2 Learners’ First-Pass Time in N-DROP Condition
Description
Predictor
Coefficient
Standard
t
(β)
Error
Overall effect of reading a
Intercept
5.84
0.06
94.23
grammatical sentence with a
masculine target noun
Overall main effect of
Gender
-0.03
0.07
-0.48
reading a sentence with a
feminine target noun
Overall main effect of
Grammaticality
-0.01
0.04
-0.23
reading an ungrammatical
sentence
Random Effects
Subject
0.028
Item
0.016
Residual
0.181

Table 6.7
Gender Model Results for L2 Learners’ Go-Past Time in N-DROP Condition
Description
Predictor
Coefficient
Standard
(β)
Error
Overall effect of reading a
Intercept
6.01
0.08
grammatical sentence with a
masculine target noun
Overall main effect of
Gender
-0.07
0.10
reading a sentence with a
feminine target noun
Overall main effect of
Grammaticality
-0.01
0.05
reading an ungrammatical
sentence
Random Effects
Subject
0.037
Item
0.036
Residual
0.248

	
  

121

p
< .001
= .635
= .820

t

p

73.70

< .001

-0.76

= .458

-0.23

= .815

Table 6.8
Gender Model Results for L2 Learners’ Total Time in N-DROP Condition
Description
Predictor
Coefficient
Standard
(β)
Error
Overall effect of reading a
Intercept
6.25
0.11
grammatical sentence with a
masculine target noun
Overall main effect of
Gender
0.01
0.13
reading a sentence with a
feminine target noun
Overall main effect of
Grammaticality
-0.04
0.05
reading an ungrammatical
sentence
Random Effects
Subject
0.087
Item
0.068
Residual
0.256

	
  

122

t

p

57.354

< .001

0.07

= .944

-0.92

= .358

REFERENCES

	
  

123

REFERENCES

Acuña-Fariña, J. C., Meseguer, E., & Carreiras, M. (2014). Gender and number agreement in
comprehension in Spanish. Lingua, 143, 108-128. doi:10.1016/j.lingua.2014.01.013
Alemán Bañón, J., Fiorentino, R., & Gabriele, A. (2014). Morphosyntactic processing in
advanced second language (L2) learners: An event-related potential investigation of the
effects of L1–L2 similarity and structural distance. Second Language Research, 30(3),
275-306. doi:10.1177/0267658313515671
Alemán Bañón, J., & Rothman, J. (2016). The role of morphological markedness in the
processing of number and gender agreement in Spanish: an event-related potential
investigation. Language, Cognition and Neuroscience, 31(10), 1273-1298.
doi:10.1080/23273798.2016.1218032
Baayen, R. H., Davidson, D. J., & Bates, D. M. (2008). Mixed-effects modeling with crossed
random effects for subjects and items. Journal of Memory and Language, 59(4), 390-412.
doi:10.1016/j.jml.2007.12.005
Barber, H., & Carreiras, M. (2005). Grammatical gender and number agreement in Spanish: An
ERP comparison. Journal of Cognitive Neuroscience, 17(1), 137-153.
doi:10.1162/0898929052880101
Bates, D. M., Mächler, M., Bolker, B. M., & Walker, S. (2015). Fitting linear mixed-effects
models using lme4. Journal of Statistical Software, 67(1), 1-48.
doi:10.18637/jss.v067.i01
Bernstein, J. B. (1993). Topics in the syntax of nominal structure across Romance. (Doctoral
Dissertation), City University of New York.
Bruhn de Garavito, J. (2003). The (dis)association between morphology and syntax: The case of
L2 Spanish. In S. Montrul & F. Ordoñez (Eds.), Linguistic theory and language
development in Hispanic languages (pp. 398-417). Somerville, MA: Cascadilla Press.
Bruhn de Garavito, J., & White, L. (2002). The second language acquisition of Spanish DPs: The
status of grammatical features. In A. T. Pérez-Leroux & J. Muñoz Liceras (Eds.), The
acquisition of Spanish morphosyntax: The L1/L2 connection (pp. 153-178). Netherlands:
Kluwer Academic Publishers.

	
  

124

Brysbaert, M., Drieghe, D., & Vitu, F. (2005). Word skipping: Implications for theories of eye
movement control in reading. In G. Underwood (Ed.), Cognitive processes in eye
guidance (pp. 53-77). Oxford: University Press.
Brysbaert, M., & Vitu, F. (1998). Word skipping: Implications for theories of eye movement
control in reading. In G. Underwood (Ed.), Eye guidance in reading and scene perception
(pp. 125-147). Oxford, England: Elsevier.
Carroll, S. (1989). Second language acquisition and the computational paradigm. Language
Learning, 39(4), 535-594. doi:10.1111/j.1467-1770.1989.tb00902.x
Carstens, V. M. (2000). Concord in minimalist theory. Linguistic Inquiry, 31(2), 319-355.
doi:10.1162/002438900554370
Chomsky, N. (1995). The minimalist program. Cambridge, MA: MIT Press.
Christensen, K. R., Kizach, J., & Nyvad, A. M. (2013). The processing of syntactic islands – An
fMRI study. Journal of Neurolinguistics, 26(2), 239-251.
doi:10.1016/j.jneuroling.2012.08.002
Clahsen, H., & Felser, C. (2006a). Continuity and shallow structures in language processing.
Applied Psycholinguistics, 27(1), 107-126. doi:10.1017/S0142716406060206
Clahsen, H., & Felser, C. (2006b). Grammatical processing in language learners. Applied
Psycholinguistics, 27(1), 3-42. doi:10.1017/S0142716406060024
Clahsen, H., & Muysken, P. (1989). The UG paradox in L2 acquisition. Second Language
Research, 5(1), 1-29. doi:10.1177/026765838900500101
Clifton, C., Staub, A., & Rayner, K. (2007). Eye movements in reading words and sentences. In
R. P. G. van Gompel, M. H. Fischer, W. S. Murray, & R. L. Hill (Eds.), Eye movements:
A window on the mind and brain (pp. 341-371). Oxford, UK: Elsevier.
Coughlin, C. E., & Tremblay, A. (2013). Proficiency and working memory based explanations
for nonnative speakers’ sensitivity to agreement in sentence processing. Applied
Psycholinguistics, 34(03), 615-646. doi:10.1017/s0142716411000890

	
  

125

Cunnings, I. (2012). An overview of mixed-effects statistical models for second language
researchers. Second Language Research, 28(3), 369-382.
doi:10.1177/0267658312443651
Cunnings, I., & Finlayson, I. (2015). Mixed effects modeling and longitudinal data analysis. In
L. Plonsky (Ed.), Advancing quantitative methods in second language research (pp. 159181). New York, NY: Routledge.
Davies, M. (2006). A frequency dictionary of Spanish: Core vocabulary for learners. New York:
Routledge.
Dewaele, J., & Véronique, D. (2001). Gender assignment and gender agreement in advanced
French interlanguage: A cross-sectional study. Bilingualism: Language and Cognition,
4(3), 275-297. doi:10.1017/S136672890100044X
Dowens, M. G., Vergara, M., Barber, H., & Carreiras, M. (2010). Morphosyntactic processing in
late second-language learners. Journal of Cognitive Neuroscience, 22(8), 1870-1887.
doi:10.1162/jocn.2009.21304
Dufour, R., & Kroll, J. (1995). Matching words to concepts in two languages: A test of the
concept mediation model of bilingual representation. Memory & Cognition, 23(2), 166180. doi:10.3758/BF03197219
Dussias, P. E. (2010). Uses of eye-tracking data in second language sentence processing
research. Annual Review of Applied Linguistics, 30, 149-166.
doi:10.1017/s026719051000005x
Ellis, R. (2005). Measuring implicit and explicit knowledge of a second language: A
psychometric study. Studies in Second Language Acquisition, 27(2), 141-172.
doi:10.1017/S0272263105050096
Fernald, A., Zangl, R., Portillo, A. L., & Marchman, V. A. (2008). Looking while listening:
using eye movements to monitor spoken language comprehension by infants and young
children. In I. Sekerina, E. M. Fernández, & H. Clahsen (Eds.), Developmental
psycholinguistics: On-line methods in children's language processing. (pp. 97-135).
Amsterdam: John Benjamins.
Franceschina, F. (2001). Morphological or syntactic deficits in near-native speakers? An
assessment of some current proposals. Second Language Research, 17(3), 213-247.
doi:10.1177/026765830101700301
	
  

126

Franceschina, F. (2005). Fossilized second language grammars: The acquisition of grammatical
gender. Amsterdam: John Benjamins.
Frenck-Mestre, C. (2005). Eye-movement recording as a tool for studying syntactic processing
of second language: A review of methodologies and experimental findings. Second
Language Research, 21(2), 175-198. doi:10.1191/0267658305sr257oa
Godfroid, A., Loewen, S., Jung, S., Park, J., Gass, S., & Ellis, R. (2015). Timed and untimed
grammaticality judgments measure distinct types of knowledge: Evidence from eyemovement patterns. Studies in Second Language Acquisition, 37(2), 269-297.
doi:10.1017/S0272263114000850
Godfroid, A., & Spino, L. (2015). Reconceptualizing reactivity of think-alouds and eye-tracking:
Absence of evidence is not evidence of absence. Language Learning, 65(4), 896-928.
doi:10.1111/lang.12136
Godfroid, A., & Winke, P. (2015). Investigating implicit and explicit processing using L2
learners’ eye-movement data. In P. Rebuschat (Ed.), Implicit and explicit learning of
languages (pp. 325-348). Amsterdam: John Benjamins.
Gries, S. T. (2015). The most under-used statistical method in corpus linguistics: Multi-level
(and mixed-effects) models. Corpora, 10(1), 95-125. doi:10.3366/cor.2015.0068
Grüter, T., Lew-Williams, C., & Fernald, A. (2012). Grammatical gender in L2: A production or
a real-time processing problem? Second Language Research, 28(2), 191-215.
doi:10.1177/0267658312437990
Hawkins, R., & Chan, C. Y. (1997). The partial availability of Universal Grammar in second
language acquisition: The 'failed functional features hypothesis'. Second Language
Research, 13(3), 187-226.
Hernández Pina, F. (1984). Teorías psicolingüísticas y su aplicación a la adquisición del español
como lengua materna. Madrid: Siglo XXI de España Editores.
Holmqvist, K., Nyström, M., Andersson, R., Dewhurst, R., Jarodzka, H., & van de Weijer, J.
(2011). Eye-tracking: A comprehensive guide to methods and measures. Oxford: Oxford
University Press.

	
  

127

Hopp, H. (2010). Ultimate attainment in L2 inflection: Performance similarities between nonnative and native speakers. Lingua, 120(4), 901-931. doi:10.1016/j.lingua.2009.06.004
Hopp, H. (2012). Grammatical gender in adult L2 acquisition: Relations between lexical and
syntactic variability. Second Language Research, 29(1), 33-56.
doi:10.1177/0267658312461803
Instituto Cervantes. (2008). Diploma de español. Nivel intermedio. Retrieved from
http://www.dele.org/uploads/test/intermedio/textos_orales.pdf
Jaeger, T. F. (2008). Categorical data analysis: Away from ANOVAs (transformation or not) and
towards logit mixed models. Journal of Memory and Language, 59(4), 434-446.
doi:10.1016/j.jml.2007.11.007
Jegerski, J. (2014). Self-paced reading. In J. Jegerski & B. VanPatten (Eds.), Research Methods
in Second Language Psycholinguistics (pp. 20-49). New York: Routledge.
Jiang, N. (2004). Morphological insensitivity in second language processing. Applied
Psycholinguistics, 25, 603-634. doi:10.1017/S0142716404001298
Jiang, N. (2007). Selective integration of linguistic knowledge in adult second language learning.
Language Learning, 57(1), 1-33. doi:10.1111/j.1467-9922.2007.00397.x
Jiang, N., Novokshanova, E., Masuda, K., & Wang, X. (2011). Morphological congruency and
the acquisition of L2 morphemes. Language Learning, 61(3), 940-967.
doi:10.1111/j.1467-9922.2010.00627.x
Keating, G. D. (2005). Processing gender agreement across phrases in Spanish: Eye movements
during sentence comprehension. (Doctoral Dissertation), The University of Illinois,
Chicago, IL.
Keating, G. D. (2009). Sensitivity to violations of gender in native and nonnative Spanish: An
eye-movement investigation. Language Learning, 59(3), 503-535. doi:10.1111/j.14679922.2009.00516.x
Keating, G. D. (2010). The effects of linear distance and working memory on the processing of
gender agreement in Spanish. In B. VanPatten & J. Jegerski (Eds.), Research in Second
Language Processing and Parsing (pp. 113-134). Amsterdam: John Benjamins.

	
  

128

Keating, G. D. (2014). Eye-tracking with text. In J. Jegerski & B. VanPatten (Eds.), Research
methods in second language psycholinguistics (pp. 69-92). London: Taylor & Francis.
Keating, G. D., & Jegerski, J. (2015). Experimental designs in sentence processing research: A
methodological review and user's guide. Studies in Second Language Acquisition, 37(1),
1-32. doi:10.1017/s0272263114000187
Kreiner, H., Garrod, S., & Sturt, P. (2013). Number agreement in sentence comprehension: The
relationship between grammatical and conceptual factors. Language and Cognitive
Processes, 28(6), 829-874. doi:10.1080/01690965.2012.667567
Leeser, M. J., Brandl, A., & Weissglass, C. (2011). Task effects in second language sentence
processing research. In P. Trofimovich & K. McDonough (Eds.), Applying priming
methods to L2 learning, teaching and research. Philadelphia: John Benjamins.
Liceras, J. M., Díaz, L., & Mongeon, C. (2000). N-drop and determiners in native and non-native
Spanish: More on the role of morphology in the acquisition of syntactic knowledge. In R.
P. Leow & C. Sanz (Eds.), Spanish applied linguistics at the turn of the millenium (pp.
67-96). Somerville, MA: Cascadilla Press.
Lim, J. H., & Christianson, K. (2014). Second language sensitivity to agreement errors: Evidence
from eye movements during comprehension and translation. Applied Psycholinguistics,
36(6), 1-33. doi:10.1017/s0142716414000290
López Ornat, S. (1988). On data sources on the acquisition of Spanish as a first language.
Journal of Child Language, 15, 679-686. doi:10.1017/S0305000900012642
Marian, V., Blumenfeld, H. K., & 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(4), 940-967.
doi:10.1044/1092-4388(2007/067)
Marinis, T. (2007a). On-line processing of passives in L1 and L2 Children. In A. Belikova, L.
Meroni, & M. Umeda (Eds.), Proceedings of the 2nd Conference on Generative
Approaches to Language Acquisition North America (GALANA) (pp. 265-276).
Somerville, MA: Cascadilla Proceedings Project.
Marinis, T. (2007b). On-line processing of sentences involving reflexive and non-reflexive
pronouns in L1 and L2 children. In A. Gavarro & M. J. Freitas (Eds.), Proceedings of
GALA (pp. 348-358). Newcastle, UK: Cambridge Scholars Publishing.
	
  

129

Marinis, T. (2010). Using on-line processing methods in language acquisition research. In E.
Blom & S. Unsworth (Eds.), Experimental methods in language acquisition research (pp.
139-162). Amsterdam: John Benjamins.
Mariscal, S. (2009). Early acquisition of gender agreement in the Spanish noun phrase: Starting
small. Journal of Child Language, 36, 143-171. doi:10.1017/S0305000908008908
McCarthy, C. (2007). Morphological variability in second language Spanish. (Doctoral
Dissertation), McGill University, Montreal.
McCarthy, C. (2008). Morphological variability in the comprehension of agreement: An
argument for representation over computation. Second Language Research, 24(4), 459486. doi:10.1177/0267658308095737
Mitchell, D. C. (2004). On-line methods in language processing: Introduction and historical
review. In M. Carreiras & C. Clifton (Eds.), The on-line study of sentence
comprehension: eyetracking, ERPs and beyond (pp. 15-32). New York: Psychology
Press.
Montrul, S., Foote, R., & Perpiñán, S. (2008). Gender agreement in adult second language
learners and Spanish heritage speakers: The effects of age and context of acquisition.
Language Learning, 58(3), 503-553. doi:10.1111/j.1467-9922.2008.00449.x
Morgan-Short, K., Sanz, C., Steinhauer, K., & Ullman, M. T. (2010). Second language
acquisition of gender agreement in explicit and implicit training conditions: An eventrelated potential study. Language Learning, 60(1), 154-193. doi:10.1111/j.14679922.2009.00554.x
Picallo, M. C. (1991). Nominals and nominalizations in Catalan. Probus, 3(3), 279-316.
doi:10.1515/prbs.1991.3.3.279
Pickering, M. J., Frisson, S., McElree, B., & Traxler, M. J. (2004). Eye movements and semantic
composition. In M. Carreiras & C. Clifton (Eds.), The on-line study of sentence
comprehension: Eye-tracking, ERPs and beyond (pp. 33-50). New York: Psychology
Press.
Plonsky, L. (2013). Study quality in SLA: An assessment of designs, analyses and reporting
practices in quantitative L2 research. Studies in Second Language Acquisition, 35(4),
655-687. doi:10.1017/S0272263113000399

	
  

130

Prévost, P., & White, L. (2000). Missing Surface Inflection or Impairment in second language
acquisition? Evidence from tense and agreement. Second Language Research, 16(2),
103-133. doi:10.1191/026765800677556046
Quené, H., & van den Bergh, H. (2004). On multi-level modeling of data from repeated
measures designs: A tutorial. Speech Communication, 43(1-2), 103-121.
doi:10.1016/j.specom.2004.02.004
Rayner, K. (1998). Eye movements in reading and information processing: 20 years of research.
Psychological Bulletin, 124(3), 372-422. doi:10.1037/0033-2909.124.3.372
Reichle, E. D., Pollatsek, A., Fisher, D. L., & Rayner, K. (1998). Toward a model of eye
movement control in reading. Psychological Review, 105(1), 125-157. doi:10.1037/0033295X.105.1.125
Roberts, L. (2012). Psycholinguistic techniques and resources in second language acquisition
research. Second Language Research, 28(1), 113-127. doi:10.1177/0267658311418416
Roberts, L., & Liszka, S. A. (2013). Processing tense/aspect-agreement violations on-line in the
second language: A self-paced reading study with French and German L2 learners of
English. Second Language Research, 29(4), 413-439. doi:10.1177/0267658313503171
Romanova, N., & Gor, K. (2017). Processing of gender and number agreement in Russian as a
second language. Studies in Second Language Acquisition, 39(1), 97-128.
doi:10.1017/s0272263116000012
Sabourin, L., Stowe, L. A., & de Haan, G. J. (2006). Transfer effects in learning a second
language grammatical gender system. Second Language Research, 22(1), 1-29.
doi:10.1191/0267658306sr259oa
Sagarra, N., & Ellis, N. (2013). From seeing adverbs to seeing verbal morphology. Studies in
Second Language Acquisition, 35(2), 261-290. doi:10.1017/s0272263112000885
Sagarra, N., & Herschensohn, J. (2010a). Proficiency and animacy effects on L2 gender
agreement processes during comprehension. Language Learning, 61(1), 80-116.
doi:10.1111/j.1467-9922.2010.00588.x

	
  

131

Sagarra, N., & Herschensohn, J. (2010b). The role of proficiency and working memory in gender
and number agreement processing in L1 and L2 Spanish. Lingua, 120(8), 2022-2039.
doi:10.1016/j.lingua.2010.02.004
Schwartz, B., & Sprouse, R. (1996). L2 cognitive states and the Full Transfer/Full Access model.
Second Language Research, 12(1), 40-72. doi:10.1177/026765839601200103
Shin, N. L., & Cairns, H. S. (2012). The development of NP selection in school-age children:
Reference and Spanish subject pronouns. Language Acquisition, 19(1), 3-38.
doi:10.1080/10489223.2012.633846
Snyder, W. (1995). Language acquisition and language variation: The role of morphology.
(Doctoral Dissertation), MIT.
Snyder, W., Senghas, A., & Inman, K. (2001). Agreement morphology and the acquisition of
noun-drop in Spanish. Language Acquisition, 9(2), 157-173.
doi:10.1207/S15327817LA0902_02
Spinner, P., Gass, S. M., & Behney, J. (2013). Ecological validity in eye-tracking: An empirical
study. Studies in Second Language Acquisition, 35(2), 389-415.
doi:10.1017/s0272263112000927
Spinner, P., & Juffs, A. (2008). L2 grammatical gender in a complex morphological system: The
case of German. International Review of Applied Linguistics in Language Teaching,
46(4). doi:10.1515/iral.2008.014
Tsimpli, I. M., & Dimitrakopoulou, M. (2007). The Interpretability Hypothesis: Evidence from
wh-interrogatives in second language acqusition. Second Language Research, 23(2), 215242. doi:10.1177/0267658307076546
van Hell, J. G., & de Groot, A. M. (1998). Conceptual representation in bilingual memory:
Effects of concreteness and cognate status in word association. Bilingualism: Language
and Cognition, 1(3), 193-211. doi:10.1017/S1366728998000352
VanPatten, B., & Benati, A. G. (2010). Key Terms in Second Language Acquisition. London:
Continuum.

	
  

132

VanPatten, B., Keating, G. D., & Leeser, M. J. (2012). Missing verbal inflections as a
representational problem: Evidence from self-paced reading. Linguistic Approaches to
Bilingualism, 2(2), 109-140. doi:10.1075/lab.2.2.01pat
Vitu, F., O'Regan, K., Inhoff, A. W., & Topolski, R. (1995). Mindless reading: Eye-movement
characteristics are similar in scanning letter strings and reading texts. Perception and
Psychophysics, 57(3), 352-364. doi:10.3758/BF03213060
Wen, S., Miyao, M., Takeda, A., Chu, W., & Schwartz, B. (2010). Proficiency effects and
distance effects in nonnative processing of English number agreement. In K. Franich, K.
Iserman, & L. Keil (Eds.), Proceedings of the 34th Boston University Conference on
Language Development (pp. 445-456). Somerville, MA: Cascadilla Press.
White, L., Valenzuela, E., Kozlowska-Macgregor, M., & Leung, Y.-K. I. (2004). Gender and
number agreement in nonnative Spanish. Applied Psycholinguistics, 25, 105-133.
doi:10.1017/S0142716404001067
Winter, B. (2013). Linear models and linear mixed effects models in R with linguistic
applications. Retrieved from http://arxiv.org/pdf/1308.5499.pdf
Witzel, N., Witzel, J., & Forster, K. (2012). Comparisons of online reading paradigms: Eye
tracking, moving-window, and maze. Journal of Psycholinguistic Research, 41(2), 105128. doi:10.1007/s10936-011-9179-x

	
  

133