Running head: EFFECTS OF CROSSLINGUISTIC OVERLAP

Investigating the Portuguese-English Bilingual Mental Lexicon:

Crosslinguistic Orthographic and Phonological Overlap

in Cognates and False Friends

Leonardo Alves-Soares

Thesis submitted to the

University of Ottawa

In partial fulfillment of the requirements

for the Doctorate in Philosophy degree in Linguistics

Department of Linguistics

Faculty of Arts

University of Ottawa

ã Leonardo Alves-Soares, Ottawa, Canada, 2o20

 ii

EFFECTS OF CROSSLINGUISTIC OVERLAP

Abstract

This dissertation investigates how cognates are organized in the bilingual mental

lexicon and examines whether orthography in one language, via phonological

representations, influences the processing of cognates and false friends in the other

language. In light of the framework of two well-known models of bilingual visual

word recognition, the Bilingual Interactive Activation (BIA) and the Bilingual

Interactive Activation Plus (BIA+), the premise is that there is activation from

orthography to phonology across a bilingual’s two languages and that this activation

is modulated by the degree of orthographic and phonological code overlap.

Two objective metrics were used to assess crosslinguistic similarity of

Portuguese-English cognates and false friends that were selected for a cross-language

lexical decision task with masked priming. Dynamic time warping (DTW), an

algorithm that was originally conceived to compare different speech patterns in

automatic speech recognition and to measure acoustic similarity between two time-

dependent sequences, was used to compute crosslinguistic phonological similarity.

The Normalized Levenshtein Distance (NLD), an algorithm that calculates the

minimum number of single-character insertions, deletions or substitutions required

to change one word into another and normalizes the result by their lengths, was used

to compute crosslinguistic orthographic similarity. Portuguese-English bilinguals

who acquired their second language after reaching puberty, and English functional
 iii

EFFECTS OF CROSSLINGUISTIC OVERLAP

monolinguals who grew up speaking primarily English were recruited to participate

in the experimental task.

Based on collected reaction time and accuracy data, mixed-effects models

analyses are used to estimate the individual effects of crosslinguistic orthographic,

phonological and semantic similarity and the role each of them, along with English

proficiency, word frequency and length play in the organization of the Portuguese-

English bilingual mental lexicon.

Keywords: bilingual mental lexicon, Portuguese, English, Portuguese-English

bilinguals, English functional monolinguals, cognates, false friends, crosslinguistic

overlap, crosslinguistic similarity, semantic, lexical decision task, LDT, masked

priming, spreading activation, RHM, BIA, BIA+, Multilink, orthographic

representations, phonological representations, dynamic time warping, DTW,

Normalized Levenshtein Distance, NLD, mixed-effects models, lmer, glmer,

reaction time, accuracy

 iv

EFFECTS OF CROSSLINGUISTIC OVERLAP

Acknowledgements

If there is one thing 2020 will be remembered for is that it has been a very tough year

for everyone due to COVID-19. These are strange times. Besides social isolation,

physical distancing and having to wear masks in order to protect ourselves and

others, there is much uncertainty about the future all around. Finishing a dissertation

in 2020 makes acknowledging those that stuck by me and helped me through this

journey even more important.

First and foremost, I would like to express my sincerest and perpetual

gratitude to my PhD supervisor, Juana Muñoz-Liceras, who took a genuine interest

in my research and never gave up on me even at the times when I felt there was no

point moving ahead. Juana always encouraged me to keep my head up, shoulders

back and to move ahead at full throttle gracefully, never looking back. Your wit,

passion and tenacity, even when the politics of the world of academia get in the way,

are an inspiration to me. I am extremely proud of having had the opportunity to work

closely with you and consider it a privilege being among one of your doctoral

students.

I would like to thank professor Marc Brunelle for opening my mind to the

possibilities with mixed-effects models and helping me see a much brighter and more

exciting world beyond ANOVAs. I will be forever grateful for all the time you took to

get me started with R and mixed-effects models. It was not an easy road for me to
 v

EFFECTS OF CROSSLINGUISTIC OVERLAP

navigate in the beginning, particularly when all I knew since my bachelor’s in

psychology was SPSS; however, you guided me in the right direction.

I also would like to thank all my committee members, Anahi Alba de la Fuente,

Elena Valenzuela, Nikolay Slavkov and Ricardo Augusto de Souza, not only for taking

the time to carefully read my work and provide me with invaluable feedback, but also

for having made my dissertation defense feel more like a conversation among

academics. It was certainly one of the hallmarks of this entire experience.

I owe my friend and fellow academic, Estela García-Alcaraz, an enormous

thank you for all her assistance with reviewing my data analysis chapter, and, in

particular, for going over my mixed-effects models with a fine-tooth comb. They say

there is no better way to learn than to teach; I can say I have learned a great deal by

imparting my knowledge of running mixed-effects models in R to you. Your

challenging questions pushed me to delve deeper into my analyses and to literally go

the extra mile.

I would like to recognize a few specific Faculty of Arts members who ensured

my tenure at the University of Ottawa as a doctoral student was as rewarding as

possible. They are Vice-Dean Frans De Bruyn, Rachel Fontaine-Azzi, Christine

Bertrand, Suzanne Dalrymple and Roxanne Lacelle. In addition, I would like to

express my appreciation to my colleagues in the English Intensive Program at the

Official Languages and Bilingualism Institute at the University of Ottawa, who were

willing to serve as functional monolingual participants in my experimental task.

 vi

EFFECTS OF CROSSLINGUISTIC OVERLAP

I would like to thank my family in Brazil for supporting me to pursue my

dreams in North America. And, finally, Pumba, thank you for always being there for

me unconditionally along the way, through thick and thin. You have always been my

biggest cheerleader, helping me keep my eyes on the prize and not lose track of a

much brighter future. I would have never been able to accomplish all of this without

you. And, to Caio Caiana, my little girl, thank you for being my furball, jumping on

my desk when I was writing this dissertation, keeping me calm, and for showering

me with lots of meows and purrs.

 vii

EFFECTS OF CROSSLINGUISTIC OVERLAP

Table of Contents

1. Introduction ................................................................................................................. 1

2. Theoretical Background and State of the Art ........................................................... 7

2.1. Spreading Activation as the Foundation of the Mental Lexicon .............................................. 7

2.2. The Bilingual Mental Lexicon ................................................................................................................. 9

2.2.1. The Revised Hierarchical Model (RHM) .................................................................................................. 10

2.2.2. The Bilingual Interactive Activation Model (BIA)................................................................................ 14

2.2.3. The Bilingual Interactive Activation Model Plus (BIA+) ................................................................... 18

2.2.4. Multilink ................................................................................................................................................................. 20

2.3. Testing the Organization of Bilingual Mental Lexicon: The Lexical Decision Task and

The Masked-Priming Paradigm ................................................................................................................... 23

2.4. Cognates and false friends .................................................................................................................... 31

2.5. The Current Study..................................................................................................................................... 35

2.5.1. Experimental Predictions and Hypotheses ............................................................................................ 37

3. The Study: Methodology ........................................................................................... 41

3.1. Participants.................................................................................................................................................. 41

3.1.1. Recruitment .......................................................................................................................................................... 41

3.1.2. Composition ......................................................................................................................................................... 43

3.1.3. Language Background Profile and Language Proficiency Assessment ...................................... 44

3.2. Materials/Stimuli ...................................................................................................................................... 46

3.2.1. Assembling Word Lists A and B ................................................................................................................... 47

3.3. Obtaining Word Frequency Count ..................................................................................................... 53

 viii

EFFECTS OF CROSSLINGUISTIC OVERLAP

3.4. Obtaining Crosslinguistic Orthographic Similarity Estimates ............................................... 54

3.5. Obtaining Crosslinguistic Phonological Similarity Estimates ................................................ 58

3.6. Experimental Procedure ........................................................................................................................ 68

4. The Study: Results ..................................................................................................... 71

4.1. Preliminary Analyses .............................................................................................................................. 72

4.1.1. Dependent Variables ........................................................................................................................................ 72

4.1.2. Independent Variables .................................................................................................................................... 76

4.1.3. Interim Summary of the Preliminary Analyses .................................................................................... 95

4.2. Mixed-Effects Models .............................................................................................................................. 99

4.2.1. Model 1 - Liner Mixed-Effects Model fitted with LMER ................................................................ 103

4.2.2. Model 2 - Liner Mixed-Effects Model fitted with GLMER ............................................................... 107

5. The Study: Discussion ............................................................................................... 111

5.1. Findings .......................................................................................................................................................113

6. Conclusion and Suggestions for Further Research................................................. 131

7. References ................................................................................................................ 139

8. Appendix .................................................................................................................. 159

 ix

EFFECTS OF CROSSLINGUISTIC OVERLAP

List of Tables

Table 1. LDT Trial Utilizing the Masked-Priming Paradigm ......................................... 25

Table 2. Experimental Predictions (Hypotheses) ......................................................... 39

Table 3. Participant Population Breakdown ................................................................. 46

Table 4. Portuguese primes – English targets (examples) ........................................... 49

Table 5. Word Lists Breakdown ..................................................................................... 52

Table 6. Acoustic similarity computations performed in PCT .................................... 67

Table 7. RT (Descriptive Statistics) ............................................................................... 73

Table 8. RAW_ACCURACY (Count) .............................................................................. 75

Table 9. CORRECT_PERCENTAGE (Descriptive Statistics) ........................................ 76

Table 10. ERS and SEMAC Independent samples t-test (Descriptive Statistics)..........77

Table 11. Pearson product-moment correlations between RT and ERS and SEMAC ... 78

Table 12. Point-biserial correlations between ERS and SEMAC with RAW_ACCURACY

............................................................................................................................... 79

Table 13. RT by SEMANTIC: Descriptive Statistics .......................................................82

Table 14. RAW_ACCURACY by SEMANTIC and LANGUAGE......................................83

Table 15. NLD: Descriptive Statistics ............................................................................ 85

Table 16. Original DTW crosslinguistic phonological similarity data vs. consolidated

data ........................................................................................................................ 88

Table 17. PHONETIC: Descriptive Statistics ................................................................ 89

Table 18. English_Freq and Portuguese_Freq by SEMANTIC (Descriptive Statistics) 91

 x

EFFECTS OF CROSSLINGUISTIC OVERLAP

Table 19. lmer Model Results ....................................................................................... 106

Table 20. glmer Model Results ...................................................................................... 110

List of Figures

Figure 1. (1984)’s Word Association Model as depicted in Kroll and Stewart (1994, p.

150). The arrows represent lexical links. ................................................................. 11

Figure 2. Potter et al. (1984)’s Concept Mediation Model as depicted in Kroll and

Stewart (1994, p. 150). The arrows represent lexical links..................................... 12

Figure 3. The Revised Hierarchical Model as depicted in Kroll and Stewart (1994, p.

158). ......................................................................................................................... 13

Figure 4. The Interactive Activation Model (IAM) as depicted in McClelland and

Rumelhart (1981, p. 378). ........................................................................................ 15

Figure 5. The Bilingual Interactive Activation Model (BIA) depicted in in Dijkstra &

van Heuven (2002, p. 177). Excitatory connections are represented by

arrowheads. Inhibitory connections are represented by filled circles. ................. 17

Figure 6. The Bilingual Interactive Activation Model Plus as depicted in Dijkstra &

van Heuven (2002, p. 182). Activation flow is represented by arrows. Inhibitory

connections are not depicted in the illustration. .................................................. 19

Figure 7. The network architecture of Multilink as depicted in Dijkstra et al. (2018, p.

662). EN = English, NL = Dutch. Arrows represent activation flow through the

 xi

EFFECTS OF CROSSLINGUISTIC OVERLAP

various network layers of lexical representation, from the input word (at the

bottom) to output (at the top)............................................................................... 23

Figure 8: Dynamic time warping illustration of the English-Portuguese cognate pair

grafite-graphite and the false friend pair smoking ‘tuxedo’-smoking. Areas that

are darker represent points where the accoustic distance between the waveforms

is less (greater acoustic similarity). ...................................................................... 61

Figure 9. PCT screen capture illustrating the parameters users can set for DTW

algorithm to be taken into account during the acoustic analysis ....................... 64

Figure 10. Acoustic similarity results table. In the Result column, higher values

represent greater acoustic distances (less acoustic similarity) between the

waveforms. ............................................................................................................. 65

Figure 11. Kernel density plots illustrating the distribution of RT (overall and by

LANGUAGE)........................................................................................................... 73

Figure 12. Mean differences in crosslinguistic orthographic overlap across semantic

categories showing a higher mean NLD for COG and FF pairs. ......................... 85

Figure 13. Kernel density plots of the DTW crosslinguistic phonological similarity

data before (left) and after (right) consolidation................................................. 88

Figure 14. Mean differences in DTW crosslinguistic phonological overlap across

semantic categories showing a lower mean for COG and FF pairs. Lower values

represent a greater the degree of crosslinguistic phonological overlap between a

prime and its target. .............................................................................................. 89

 xii

EFFECTS OF CROSSLINGUISTIC OVERLAP

Figure 15: Differences between functional monolinguals and bilinguals with regards

to whether or not they have the knowledge of another European language

besides English and Portuguese. ........................................................................... 95

List of Examples

Example 1. Weber’s Grapheme Similarity (GS) Index ................................................... 55

Example 2. Van Ordern’s Orthographic Similarity (OS) Algorithm ........................... 56

Example 3. Normalized Levenshtein Distance (NLD) .................................................. 57

List of Appendices

Appendix A. Research Ethics Approval Forms ............................................................ 159

Appendix B. Participant Recruitment Form ................................................................ 165

Appendix C: Informed Consent Forms ........................................................................ 166

Appendix D: Language Background Questionnaire ....................................................172

Appendix E: Cloze Proficiency Test ............................................................................. 178

Appendix F: Experimental Word Lists ......................................................................... 180

Appendix G: RStudio lmer model raw output ............................................................212

Appendix H: RStudio glmer model raw output......................................................... 223

 1

EFFECTS OF CROSSLINGUISTIC OVERLAP

1. Introduction

For a Portuguese-English bilingual – or any other bilingual for that matter – riding

the subway in New York City can be a truly remarkable experience. Among all the

hustle and bustle, all the bright, flashing lights and all the announcements in

English, the feeling of reading or hearing a Portuguese word in an English-speaking

country can lead to moments of confusion and ambiguity. For example, seeing the

English word push printed on access doors to stations, emergency exits or on the

windows in the trains can automatically trigger the Portuguese word puxe, which

sounds similar to the English word but has a completely different meaning in the

language. Puxe in Portuguese means to exert force on someone (or something) so as

to cause movement toward oneself (or itself), which in English is conveyed by the

word pull. In other words, the pronunciation similarities between puxe and push are

merely superficial. Similarly, the English word platform can automatically trigger the

Portuguese word plataforma, except that these two words have the same meaning

across both languages. Words like puxe and push, which share some phonemes,

graphemes and yet have completely different meanings crosslinguistically, are often

defined in the literature as false friends, or sometimes false cognates. Words like

platform and plataforma are defined as cognates. They share many graphemes,

phonemes and meaning crosslinguistically. Despite the fact that English is a

Germanic language and Portuguese is a Romance language typologically, both

 2

EFFECTS OF CROSSLINGUISTIC OVERLAP

languages have a considerable number of words whose etymological origin is Latin.

As such, both languages also have considerable number of cognates, e.g., computador

and computer, piano and piano, televisão and television, robô and robot, zero and

zero, etc. English and Portuguese also possess a considerable number of false friends,

e.g., batom ‘lipstick’ and baton, chute ‘kick’ and chute, sapo ‘frog’ and sap, tampão

‘drain cover’ and tampon, etc.

The extant psycholinguistic literature on the processing of cognates and false

friends has helped elucidate the organization of the bilingual mental lexicon,

particularly the mechanisms that trigger that sense of ambiguity that bilinguals can

often experience when they read or hear a word in one of their languages that look

and sound like a word in the other language. For example, in translation recognition

and lexical decision tasks with masked priming, in which accuracy and reaction time

are the dependent variables, highly fluent bilinguals show a significantly greater

priming effect for cognates than for non-cognates in both L1-to-L2 and L2-to-L1

directions (Dunabeitia et al., 2010). In other words, orthographic, semantic and

phonological similarity crosslinguistically seem to play a role in the organization of

the mental lexicon. Thus, it is possible that crosslinguistically cognates may share a

single mental representation. There is evidence suggesting that bilinguals are able to

name cognates more accurately and faster than both non-cognates and false friends

(Sunderman & Schwartz, 2008). Additionally, Jared and Szucs (2002) observed that

French-English bilinguals were slower to name interlingual homographs (false

 3

EFFECTS OF CROSSLINGUISTIC OVERLAP

friends with a great deal of crosslinguistic orthographic overlap with distinct

phonological codes) than matched English controls in an English naming task.

Similarly, Schwartz, Kroll and Diaz (2007) showed highly proficient English-Spanish

bilinguals displayed greater naming latencies (or RTs) for cognates with high

crosslinguistic orthographic overlap when their corresponding phonological codes

were different crosslinguistically. This effect was even observed irrespective of the

bilinguals’ dominant language. Finally, there is evidence showing there is cross-

language feed-forward activation from orthography to phonology, like during a visual

word recognition task; cognates with a high degree of crosslinguistic orthographic

overlap that sound rather different crosslinguistically are processed more slowly than

cognates whose degree of orthographic and phonological code overlap is equal

(Schwartz & Kroll, 2006).

Despite all of the evidence supporting the special status of cognates, many

outstanding issues still remain. For example, most studies do not make a clear

distinction among the various kinds of cognates, viz., true vs. false friends, semi-false

vs. false friends, interlingual homophones vs. homographs, etc. Additionally, the

effects of cross-language activation during the processing of cognates with varying

degrees of phonological similarities (variations of how differently they sound

crosslinguistically) has not been yet established for many language pairs, in particular

English and Portuguese.

In light of this, the primary goal of this dissertation is to investigate how

 4

EFFECTS OF CROSSLINGUISTIC OVERLAP

cognates are organized in the Portuguese-English bilingual mental lexicon and to

understand whether orthography in Portuguese, via phonological representations,

influences Portuguese-English bilinguals’1 processing of cognates and false friends in

their second language. The premise is that there is activation from orthography to

phonology across a bilingual’s two languages and that this activation is modulated by

the degree of crosslinguistic orthographic and phonological code overlap. As such,

this dissertation will further advance the understanding of the interplay between

orthography and phonology in the organization of the bilingual mental lexicon.

Two groups of participants were recruited for a cross-language lexical decision

task with masked priming, Portuguese-English bilinguals and English functional

monolinguals, who served as controls. The stimuli selected for the experiment were

comprised of Portuguese-English cognates and false friends. Their degree of

crosslinguistic orthographic and phonological code overlap was determined through

objective metrics such as the Normalized Levenshtein Distance and dynamic time

warping. Two mixed-effects models were created to account participants’ reaction

times and accuracy.

This dissertation is organized as follows.

Chapter 2 offers a broad overview of early literature on the organization of the

1
While in the extant literature there are multiple definitions of bilingualism, this study is primarily
concerned with late sequential bilingualism, as defined by Montrul (n.d., p. 17). Specifically,
Portuguese speakers who acquired English as their second language (L2) during or after puberty.
 5

EFFECTS OF CROSSLINGUISTIC OVERLAP

mental lexicon, from its origins on spreading activation and connectionist models, to

current literature on the organization of the bilingual mental lexicon. A synopsis of

the structural elements of five well-known models of bilingual word recognition, viz.,

the Revised Hierarchical Model (RHM), the Interactive Activation Model (IAM), the

Bilingual Interactive Activation Model (BIA) and its derivatives, the Bilingual

Interactive Activation Model Plus (BIA+) and Multilink, is also provided.

Chapter 3 explains the methodology employed in the study, from the process

of participant recruitment and English proficiency testing to how all the stimuli were

selected for the cross-language lexical decision task using objective metrics of

crosslinguistic code overlap as well as how the experimental word lists were

assembled.

Chapter 4 reports the results of the cross-language lexical decision task. It is

divided into two sections, Preliminary Analyses and Mixed-Effects Models, with the

former serving as groundwork for the latter. In Preliminary Analyses, each of the

variables of interest are analyzed using conventional statistical analyses tools, such

as t-tests, ANOVAs and correlations. In Mixed-Effects Models, a linear mixed-effect

model (lmer), with reaction time as the dependent variable, and a generalized linear

mixed-effect model (glmer), with accuracy as the dependent variable, are presented.

Both mixed-effects analyses allowed for precise quantification of the individual fixed

effects of crosslinguistic similarity, as well as their interactions with one another,

which was one of the primary goals of this study.

 6

EFFECTS OF CROSSLINGUISTIC OVERLAP

Chapter 5 discusses the key findings from the experimental components of

this study and formalizes the contribution to the existing research in light of the

framework of two important models of the bilingual mental lexicon, the BIA and the

BIA+.

Chapter 6 summarizes the study and offers the main conclusions as well as

suggestions for further research.

 7

EFFECTS OF CROSSLINGUISTIC OVERLAP

2. Theoretical Background and State of the Art

2.1. Spreading Activation as the Foundation of the Mental Lexicon

The mental lexicon is theorized to be a mental dictionary that stores information

pertaining to meaning, spelling and pronunciation of all the words a speaker has

acquired throughout their life. One of the greatest challenges when proposing a

model of the mental lexicon concerns the representation of concepts and form,

specifically, how the conceptual representation of a word in memory (semantics) can

activate graphemes (orthographic representations) and phonemes (phonological

representations) and vice-versa. Collins and Quillian (1969), based upon spreading

activation, proposed a model of lexicosemantic encoding and decoding in which they

envisaged the mental dictionary as a network of interconnected nodes. In their view,

the network nodes represent acquired linguistic knowledge and the pathways, or the

connections/links between the nodes, represent their semantic associations. When a

node in the network becomes activated, it automatically spreads (or percolates) its

activation throughout its pathways iteratively (in a chain-like reaction) to other

nodes with which it is semantically associated. For example, when the word plant is

presented to an English speaker, the network nodes linked to the conceptual

representation of the word in memory spread (or percolate) their activation to nodes

linked to the conceptual representation of the word flower, which, in turn, spread

their activation to nodes linked to the conceptual representation of the English word

rose. Words in the mental lexicon are also thought to have a resting-level (baseline
 8

EFFECTS OF CROSSLINGUISTIC OVERLAP

level) activation; words that are repeatedly activated together are posited to

collectively retain a portion of their activation; thus, they are posited to be at a higher

resting-level activation, which facilitates the speed with which activation spreads

between them and to other associated words (Anderson, 1983a, 1983b, 2015; Samani

& Sharifian, 1997).

Over the years the underpinnings of the mental lexicon were further

advanced. For example, a hierarchical organization consisting of three layers of

linguistic representation was proposed: a semantic or conceptual layer, which sits

atop the hierarchy, an intermediate lexical layer, and a sublexical layer, which sits at

the bottom of the hierarchy. In this framework, during the course of visual word

recognition, activation is thought to spread iteratively, i.e., in a feed-forward fashion,

within each level of linguistic representation as well as backwards from higher to

lower level representations. In addition, the existence of a lexical selection system,

which is responsible for ensuring that words in the mental lexicon that match a

speaker’s communicative intention are properly retrieved from semantic memory

and produced is also introduced (Caramazza, 1997; Costa, La Heij, & Navarrete, 2006;

Costa & Santesteban, 2004; Dell, 1986; Levelt, 1993, 2001; Levelt, Roelofs, & Meyer,

1999; Motley & Camden, 1985). Levelt (1993, 2001) argued that without a lexical

selection system, spreading activation would create an insurmountable obstacle for

communication, because a profusion of lexical items (or words) would be

simultaneously active and compete with one another for production, which would
 9

EFFECTS OF CROSSLINGUISTIC OVERLAP

make human communication a rather slow and error-ridden process. Levelt states

that speakers err no more than once or twice every 1,000 words, in spite of the fact

that they amass a mental lexicon of 50,000-100,000 words throughout their lifetime.

Putting it simply, because human communication is both effective and fluid, a lexical

selection system must be responsible for ensuring a speaker’s intention is conveyed

in a message. Levelt theorizes that the lexical selection system takes into account the

amount of activation lexical and sublexical representations receive from semantic

representations in order to ultimately select a lexical item with the highest level of

activation, which in normal communication should match the speaker’s

communicative intention. Once selected, this lexical item is sent to subsequent

linguistic processing or articulation.

2.2. The Bilingual Mental Lexicon

Highly proficient sequential bilinguals, without much apparent delay or effort, are

able to switch between their L1 and L2 as well as map a specific word they read or hear

to one of their two languages. This remarkable ability has been the subject of much

investigation. Based on the enormous body of research accumulated over the last four

decades, two theoretical issues have consistently remained under the microscope: (a)

whether bilinguals’ two languages are stored together in a single, integrated mental

lexicon, expressed in the literature as the shared-lexicon hypothesis, and (b) whether

the words from bilinguals’ two languages are accessed simultaneously (in parallel)
 10

EFFECTS OF CROSSLINGUISTIC OVERLAP

during word recognition, expressed in the literature as the language non-selective

hypothesis. As a result, a number of models of bilingual word recognition with

different core components, some leapfrogging earlier models, have been postulated

over the years and tackle the shared lexicon hypothesis and the language non-

selective hypothesis in diverse ways. The following is a synopsis of five well-known

models of bilingual word recognition, viz., the Revised Hierarchical Model (RHM),

the Bilingual Interactive Activation Model (BIA) and its precursor the Interactive

Activation Model (IAM), the Bilingual Interactive Activation Model Plus (BIA+), and

Multilink.

2.2.1. The Revised Hierarchical Model (RHM)

Kroll and Stewart (1994) – later revisited in Kroll and Tockowicz (2001) – proposed a

consolidation of two hierarchical models of language interconnection conceived

originally by Potter, So, von Eckhardt, and Feldman (1984) to test whether L2

proficiency could account for translation performance differences in picture naming

and translation-into-L2 tasks. The first model, viz., Word Association Model,

purported that words in the bilingual mental lexicon are stored in two separate, yet

interconnected, lexical memory systems (one for each language), and that concepts

are stored in an abstract semantic system. As Figure 1 below shows, in the Word

Association Model L2 words are associated with L1 words (and vice-versa). However,
 11

EFFECTS OF CROSSLINGUISTIC OVERLAP

only through L1 mediation can L2 words gain access to the abstract semantic system

(concepts).

images
[_, ~-~ L~2

concepts

Figure 1. (1984)’s Word Association Model as depicted in Kroll and Stewart (1994, p. 150). The
arrows represent lexical links.

The second model, viz., Concept Mediation Model, also proposed that words

in the bilingual mental lexicon are stored in two separate lexical memory systems

(one for each language) and that concepts are stored in an abstract semantic system.

However, as Figure 2 shows, unlike in the Word Association Model, in the Concept

Mediation Model the two lexical stores are depicted as completely independent from

each other. Further, L1 and L2 words have direct access to concepts in the common

abstract semantic system (concepts). However, in order for words in either lexical
 12

EFFECTS OF CROSSLINGUISTIC OVERLAP

store to gain access to words in the other lexical store, this access would need to be

mediated via concepts, hence the name Concept Mediation Model.

LI

concepts

Figure 2. Potter et al. (1984)’s Concept Mediation Model as depicted in Kroll and Stewart
(1994, p. 150). The arrows represent lexical links.

In a series of studies that put both models to the test, Porter et al. only found

support for the Concept Mediation Model. They concluded that the semantic store

must always mediate the connections between both lexical stores irrespective of

bilinguals’ L2 proficiency, which was challenged subsequently in Kroll and Curley

(1988) as well as Chen and Leung (1989).

Similar to both the Concept Mediation Model and the Word Association

Model, the Revised Hierarchical Model (RHM) also envisaged the bilingual mental

lexicon following a hierarchical organization. At the lexical level, words from

 13

EFFECTS OF CROSSLINGUISTIC OVERLAP

bilinguals’ two languages are stored in their own respective lexical memory system,

and at the semantic level, concepts are stored in an abstract memory system.

However, unlike its two predecessor models, the RHM posits that as bilinguals

become more proficient in their L2, or as bilinguals’ L2 dominance in respect to L1

increases, L2 words can gain direct access to the abstract conceptual store instead of

via L1 mediation.

lexical
links

Ll L2

conceptual , ' conceptual

links
, , links

concepts

Figure 3. The Revised Hierarchical Model as depicted in Kroll and Stewart (1994, p. 158).

The RHM’s organization can elegantly accommodate translation performance

differences that can arise as a result of L2 proficiency (illustrated by the dashed

arrows in Figure 3 above). Specifically, for unbalanced bilinguals, those who are less

proficient in their L2, translating from L1 to L2 occurs through concept mediation,

 14

EFFECTS OF CROSSLINGUISTIC OVERLAP

i.e., L1 words gain access to L2 words via the semantic store (concepts), thus,

translating in this direction is slower. This conceptual mediation mirrors unbalanced

bilinguals’ reliance on their L1 in order to access the conceptual representations in

the semantic store. Translating from L2 to L1, however, is not mediated via the

semantic store, i.e., L2 words have direct access to L1 words, thus, translating in this

direction is faster (illustrated by the solid arrows in Figure 3 above). This is so because

the connections the L2 lexical store develops to the L1 lexical store are much stronger

than the other way around, which occur through conceptual mediation.

2.2.2. The Bilingual Interactive Activation Model (BIA)

It is impossible to examine the BIA (Dijkstra & van Heuven, 1998) without first

introducing the Interactive Activation Model (IAM), its precursor. The IAM is a

connectionist model devised by McClelland and Rumelhart (1981; 1982) that

postulates that bottom-up and top-down processes occur in parallel during the

course of visual word recognition (or perception). Structurally, the IAM is a

hierarchical model. It specifies various levels of representation (abstraction), viz.,

visual-acoustic feature level, a letter-phoneme level, a word level, and higher levels

of processing that exert top-down input to the word level (see Figure 3). Low level

visual and acoustic feature-based representations spread their activation to letter and

phoneme-based representations at one level higher, which, in turn, spread their

activation to word-based representations one level higher. As its name suggests, the
 15

EFFECTS OF CROSSLINGUISTIC OVERLAP

IAM posits that visual word recognition is a highly interactive process; thus, each

adjacent level of representation can interact with each other, either exciting or

inhibiting activation spread to other adjacent levels. However, activation spread

within the word level is only inhibitory, since one word is posited to receive excitatory

activation from lower levels of representation.

HIGHER LEVEL INPUT

11l

nr
VISUAL INPUT
rn
ACOUSTIC INPUT

Figure 4. The Interactive Activation Model (IAM) as depicted in McClelland and Rumelhart
(1981, p. 378).
 16

EFFECTS OF CROSSLINGUISTIC OVERLAP

The BIA is in its essence an offshoot of the IAM. Thus, its architecture is largely

the same, with the exception of the added language level containing two language

nodes to represent bilinguals’ two languages, since the BIA is a model of bilingual

visual word recognition. Activation at each level of representation in the BIA can

further excite (feed-forward) or inhibit (feedback inhibition) activation at each of the

adjacent levels (see Figure 5 ). In this way, when a string of letters is presented to the

model, low-level feature-representations are activated, which subsequently activate

letter-based representations with features present in the original input string and

also inhibit other letter and phoneme-based representations lacking those features.

The letter-based representations in turn activate words in both languages that

contain the letters in the same position of the input string and also inhibit words that

do not. Thus, regarding the language non-selective hypothesis, activation in the BIA

is imagined to be non-language specific. Language nodes then exert top-down

inhibitory feedback on the activated lexical items from both languages, inhibiting

the ones from the other language to ensure successful identification of the input

string.
 17

EFFECTS OF CROSSLINGUISTIC OVERLAP

Language

Dutch
Word
words
''

Letter

Feature

Visua l input

Figure 5. The Bilingual Interactive Activation Model (BIA) depicted in in Dijkstra & van
Heuven (2002, p. 177). Excitatory connections are represented by arrowheads. Inhibitory
connections are represented by filled circles.
 18

EFFECTS OF CROSSLINGUISTIC OVERLAP

2.2.3. The Bilingual Interactive Activation Model Plus (BIA+)

In order to address the lack of specification for phonological and semantic

representations in the BIA, the Bilingual Interactive Activation Model Plus (BIA+)

was conceived (Dijkstra & van Heuven, 2002). The BIA+, similarly to the BIA, is a

computational model. It envisages bilingual word recognition being comprised of

two primary systems, a Word Identification System and a Task/Decision System.

Orthographic, phonological and semantic representations from bilinguals’ both

languages are integrated in the Word Identification System, much like in a network

of interconnected nodes where sublexical and lexical linguistic representations are

stored.

According to the BIA+ the bilingual mental lexicon is integrated and lexical

access is language non-selective. As depicted in Figure 6, when a written word is

presented to the BIA+, a number of word competitors from both languages are

activated simultaneously based on their similarity to the input word. Additionally,

the BIA+ postulates that bilingual visual word recognition is affected by

crosslinguistic orthographic overlap as well as crosslinguistic phonological and

semantic overlap. In this fashion, when a word is presented to the BIA+, its lexical

and sublexical orthographic representations activate similar lexical and sublexical

orthographic and phonological representations, which then activate semantic

representations. Thus, the greater the degree of crosslinguistic orthographic,

phonological and semantic overlap between a word and stored mental

 19

EFFECTS OF CROSSLINGUISTIC OVERLAP

representations (in both languages), the greater the number of similar word

competitors from both languages that are activated simultaneously.

Task schema

· Specific processing steps for task in hand

· Receivescontinuous input from the
identificationsystem
· Decision criteria determine when a
response is made based on relevant
codes

Identification system

Languagenodes Semantics

Lexical Orthography
- Lexical Phonology

Sublexical Orthography Sublexical Phonology

Figure 6. The Bilingual Interactive Activation Model Plus as depicted in Dijkstra & van Heuven
(2002, p. 182). Activation flow is represented by arrows. Inhibitory connections are not
depicted in the illustration.

The BIA+ also posits that other factors, such as word frequency, recency of use

(how recently a word was last used by the bilingual) and L2 proficiency modulate
 20

EFFECTS OF CROSSLINGUISTIC OVERLAP

activation spread crosslinguistically, because they can raise or lower the resting-level

activation of stored representations. In this fashion, compared to words that are not

used often, frequently used words are thought to have a higher resting-level of

activation. Consequently, not only less input activation is required to activate stored

representations linked to a frequently used word but also activation is believed to

spread much more quickly between frequently used words. Likewise, a delay in

activation spread is posited to occur among L2 representations compared to L1

representations, which is more pronounced for unbalanced bilinguals. Because this

group of speakers is not equally proficient and fluent in their L2, their L2

representations are posited to have a lower resting-level of activation, thus activation

spreads to linked L1 mental representations much m0re quickly than it does to L2

representations.

2.2.4. Multilink

Multilink (Dijkstra et al., 2018) is a state-of-the-art revision to earlier models of the

bilingual mental lexicon and, as such, incorporates several features of its

predecessors, in particular the RHM, the BIA and the BIA+. Multilink is a

computational cognitive model based on the localist-connectionist paradigm and

written in an object-oriented approach in Java. It allows precise simulation of

cognitive processes using a series of computations (algorithm) that enable language

researchers to generate new hypotheses for empirical testing and better understand
 21

EFFECTS OF CROSSLINGUISTIC OVERLAP

how the mechanisms underlying bilingual (and monolingual) word recognition and

production interact and play out both qualitatively and quantitatively with an

unparalleled level of precision never thought possible before. Moreover, Multilink

allows language researchers to advance more robust theories regarding the

organization of the bilingual mental lexicon, particularly with respect to how

cognates might be stored. Its task/decision system can replicate several experimental

tasks, such as word processing during lexical decision, orthographic and semantic

priming, word naming, and word translation (in both forward and backward

directions), which are often utilized in psycholinguistic studies. Furthermore,

Multilink’s lexicon parameter can be fine-tuned to account for L2-proficiency levels,

frequency of use, word length and crosslinguistic overlap. In fact, Multilink is such a

new model that its lexicon at the moment only comprises English-Dutch words.

Multilink’s lexicon, however, is integrated across both languages. Over time, as the

model becomes more mainstream with language researchers, its architects hope

Multilink’s lexicon will expand to include several more language pairs.

One of Multilink’s core assumptions is that the mental lexicon is integrated

across both languages of a bilingual. This is in line with the findings of many

psycholinguistic studies suggesting there is little scientific evidence to conceive the

existence of two separate lexicons, one for each language (Marc Brysbaert & Duyck,

2010; Van Heuven, Dijkstra, & Grainger, 1998). Multilink depicts the mental lexicon

as layered lexical network interconnected by links that vary in strength and in which
 22

EFFECTS OF CROSSLINGUISTIC OVERLAP

activation flow is largely bidirectional (see Figure 7 below). Nevertheless, despite

having a layered structure, unlike the BIA+, a sublexical level is not specified in

Multilink, thus, activation flow occurs lexically. Multilink posits each stored word as

having a resting level activation (RLA), which is a function of the frequency of use of

the word itself (measured in occurrences per million) and the highest and lowest

frequencies of all other words stored in the lexicon. In this framework, when a written

word, irrespective of its length, is presented to Multilink as input, it simultaneously

activates a number of stored lexical-orthographic representations (word candidates)

from both languages proportionate to the strength of the orthographic similarity

between the input word and the stored representations as well as their RLA. Multilink

utilizes the Levenshtein distance (Levenshtein, 1966) normalized for word length as

a metric of crosslinguistic orthographic similarity. Word recognition in Multilink

takes place in a number of time steps (or cycles). Thus, the degree of semantic and

phonological crosslinguistic similarity between the input word and each word

candidate can further strengthen or attenuate activation flow among word

candidates in subsequent time cycles. However, it is important to point out that in

Multilink crosslinguistic phonological similarity can only strengthen activation flow

through semantics. The task/decision system lastly checks for language membership

of the input word and word candidates as well as their degree of orthographic,

phonological and semantic activation required for output in the experimental task at

hand.
 23

EFFECTS OF CROSSLINGUISTIC OVERLAP

LAYERS Standard Architecture of Multilink

Output I I

Phonology

Semantics

Language

, X

Orthography

G
l Input
I

Figure 7. The network architecture of Multilink as depicted in Dijkstra et al. (2018, p. 662). EN
= English, NL = Dutch. Arrows represent activation flow through the various network layers
of lexical representation, from the input word (at the bottom) to output (at the top).

2.3. Testing the Organization of Bilingual Mental Lexicon: The Lexical

Decision Task and The Masked-Priming Paradigm

The masked priming paradigm is one of the most well-known techniques employed

by language researchers to test their hypotheses regarding the organization of the

mental lexicon. It consists of sequentially presenting participants in the center of a

computer screen a forward mask, a prime, sometimes a backward mask, and a target.
 24

EFFECTS OF CROSSLINGUISTIC OVERLAP

This paradigm allows an experimenter to manipulate the properties of the prime in

order to determine its effect on participants’ response latencies (or their response

accuracy) to the target, thus illuminating the processing of linguistic code (Tzur &

Frost, 2007). Priming occurs when the prime facilitates participants’ response to the

target in comparison to some neutral/control baseline (Forster, Mohan, & Hector,

2003, Chapter 1). The most common explanation for the facilitation effect induced by

priming is that it is an unconscious (non-strategic) automatic process due to

spreading-activation. Specifically, the prime activates a linguistic representation, and

this activation spreads associatively to other linguistic representations, facilitating

their identification (Hutchison, 2007). In order to measure the effect of priming, the

masked-priming paradigm is combined with a lexical decision task (LDT) where

participants are instructed to respond to a word (target), usually by pressing a key on

a button box or keyboard, to indicate whether or not the target is a real word in a

particular language, e.g. is this an English word? (see Table 1 below outlining the

procedure).
 25

EFFECTS OF CROSSLINGUISTIC OVERLAP

Table 1. LDT Trial Utilizing the Masked-Priming Paradigm

Presentation Sequence Computer Screen Display Duration

#####
forward mask
500ms
hash marks covering the
same visuospatial field of the
prime

2
prime fruta 30-70ms
SOA
in lowercase letters

backward mask
(optional)
##### 500ms

3
target FRUIT 500ms

in uppercase letters

4 3,000ms before
participant’s response the next stimulus
usually pressing a key on a
is presented
button box (yes/no)

As Table 1 outlines, a number of procedural parameters needs to be carefully

considered prior the design stage and during the administration of an LDT. A

particular important one is the stimulus onset asynchrony (SOA), which is the
 26

EFFECTS OF CROSSLINGUISTIC OVERLAP

amount of time that will elapse between the presentation of the prime and

subsequent target. A shorter SOA invokes early word recognition processes and

ensures that participants are completely unaware of the prime, thus mitigating the

likelihood that any cognitive processes (linguistically related and potentially non-

linguistically as well) are triggered (Forster & Davis, 1984; Forster et al., 2003;

Grainger & Holcomb, 2009; Rastle, Davis, Marslen-Wilson, & Tyler, 2000; Sabourin,

Brien, & Burkholder, 2014). Long SOAs make the prime completely visible to

participants, i.e., they are able to completely read the prime through, and, since they

are conscious of it, postperceptual cognitive processes, not all connected to the

language system, are automatically induced. (Forster et al., 2003, Chapter 1).

Specifically, long SOAs allow participants to consciously (and strategically) compute

probable targets (conscious expectancy generation). As a result, long SOAs yield

rather limiting priming effects, because other cognitive operations interfere with the

time course of the processing of a linguistic code (Colombo, 1986; Martin & Jensen,

1988; Tzur & Frost, 2007).

As can be seen in Table 1, a trial in an LDT is comprised of four major events

that occur in the following sequence: (1) a forward mask consisting of a sequence of

hash marks/tags (#####) covering the same visuospatial field as the prime is

presented for 500ms before the prime; (2) the prime in lowercase letters is presented

briefly (SOA 30-70 ms); (3) the target, usually in uppercase letters, is presented; and

(4) participants have a preset amount of time to record their response (usually by
 27

EFFECTS OF CROSSLINGUISTIC OVERLAP

the press of a button on a button box) before the next trial is displayed. It is

important to highlight the case change between the prime and target, for it ensures

that the two stimuli are visually distinct. Additionally, in order to avoid participants

perceiving the target as a continuation of the prime due to their form overlap, some

studies have opted to add a backward mask between the presentation of the prime

and the subsequent presentation of the target (Forster et al., 2003, Chapter 1).

Distinct types of priming effects are reported in the literature when the

masked-priming technique is used in experimental tasks and have helped

researchers to postulate and advance models of visual word recognition and lexical

organization. “These effects are assessed relative to a baseline condition, in which

the prime differs from the target in all baseline conditions” (Forster et al., 2003, para.

6). Among the effects observed, the translation priming effect is of particular interest

because the prime in one language is posited to facilitate the processing of a

translation equivalent target in the other language, e.g., the Portuguese prime cão

would facilitate the processing of its English translation equivalent dog. This effect

is attributed to the mental representations of both words being connected in

memory either at a conceptual or a lexical level. In addition, translation priming

asymmetry has been reported across several different language pairs, and is

significantly greater in the forward priming direction, viz., from L1-to-L2, than in

the backward direction, viz., L2-to-L1 (Duñabeitia, Dimitropoulou, Uribe-

Etxebarria, Laka, & Carreiras, 2010; Gollan, Forster, & Frost, 1997; Grainger & Frenck-
 28

EFFECTS OF CROSSLINGUISTIC OVERLAP

Mestre, 2002; Jiang, 1999). In the forward-priming direction, L1 translation primes

consistently yield faster reaction times to L2 targets, with an average magnitude

effect of 39ms. In contrast, in the backward-priming direction, the results are not as

consistent, with an average magnitude of the effect remarkably less robust at around

6ms (Duñabeitia, Dimitropoulou, et al., 2010). This priming asymmetry is thought

to be the result of uneven proficiency between the L1 and the L2, i.e., bilinguals who

are not equally proficient in both of their languages (Duñabeitia, Dimitropoulou, et

al., 2010; Gollan et al., 1997; Grainger & Frenck-Mestre, 2002; Jiang, 1999).

Nevertheless, more recently Wen and van Heuven (2017) conducted a meta-analysis

of 24 studies that have utilized the masked-priming technique to quantitively assess

the effect sizes of translation priming effects in both directions. The results revealed

significant priming effects in both priming directions, with the effect being

significantly larger in the L1-L2 than the L2-L1 direction.

The semantic priming effect is attributed to the automatic activation of a

shared semantic store, i.e., the prime in L1 and the target in L2, or vice versa, share

a single conceptual (semantic) representation. In this way, when an L1 prime is

presented to a bilingual participant, it automatically activates a target in L2, or vice-

versa (Dimitropoulou, Duñabeitia, & Carreiras, 2011a; Van Hell & Dijkstra, 2002). For

example, for a Portuguese-English bilingual a Portuguese prime like café ‘coffee’ is

expected to activate English targets like milk or sugar, since coffee is usually served

with milk and sugar in many Portuguese speaking countries.

 29

EFFECTS OF CROSSLINGUISTIC OVERLAP

The orthographic priming effect (also described as form priming) is a type of

priming effect reported by studies that have utilized the masked-priming paradigm.

The effect is reported when the processing of a target, e.g., faster reaction times, is

facilitated by the presentation of a preceding orthographically similar prime. In this

condition, when the orthographic overlap between the prime and target is high, i.e.,

when they share many letters, there is minimal competition between them, thus

perception of the target is facilitated.

The phonological priming effect is another priming effect that has helped

enhance several models of bilingual visual word recognition and is often reported in

the literature. This effect is of particular interest because it is often present for

bilingual participants but absent for monolingual participants. A few studies that

have utilized the masked-priming technique to investigate whether it is possible for

a homophonic prime in one language to facilitate the recognition of a target in the

other language have suggested that during the initial stages of visual word

recognition, in addition to its orthographic representation in both languages, a

word’s phonological representation (in both languages as well) is also automatically

activated (Marc Brysbaert, Van Dyck, & Van De Poel, 1999; Dimitropoulou,

Duñabeitia, & Carreiras, 2011b; Perfetti & Bell, 1991; Pexman, Lupker, & Jared, 2001;

Van Wijnendaele & Brysbaert, 2002). “Visual word recognition is phonologically

mediated to a large extent, not only in L1 but also in L2” (Marc Brysbaert, 2003, p.

186). Therefore, recognition of an L2 target, for example, can also be facilitated if a

 30

EFFECTS OF CROSSLINGUISTIC OVERLAP

homophonic2 prime in L1 is presented to bilingual participants.

While employing the masked priming technique has been proven useful at

establishing a richer understanding of bilingual lexical organization, estimating the

role of crosslinguistic phonological similarity in isolation requires a much more

refined approach from language researchers. It is methodologically challenging to

disentangle orthographic similarity from phonological similarity due to the fact that

in most languages letters (orthography) map to sounds (phonemes). More often

than not, words that are orthographically similar crosslinguistically (have large

crosslinguistic orthographic code overlap/similar spelling) are also judged to be

phonologically similar crosslinguistically (large phonological code overlap/similar

pronunciation). Simply put, when a word in one language sounds similar to a word

in another language, their spelling is also similar. Nevertheless, newer models of

bilingual visual word recognition, viz., BIA+ and Multilink, have incorporated

crosslinguistic phonological similarity as crucial component separate from

crosslinguistic orthographic similarity.

2
The word “homophonic” is used here with the meaning of sounding similar or having a similar
pronunciation.
 31

EFFECTS OF CROSSLINGUISTIC OVERLAP

2.4. Cognates and false friends

Many well-known models of the bilingual mental lexicon theorize that

bilinguals’ two languages share a single semantic store and that words are comprised

of abstract linguistic representations. If the premise of a shared semantic store is

accurate, due to spreading activation, during visual word recognition, presenting an

L1 prime to a bilingual should automatically activate an L2-semantically-related

target (or vice-versa). In addition, crosslinguistic form overlap (phonological and

orthographic similarity) should also play a role. With this in mind, the masked-

priming paradigm has been extensively used to elucidate the organization of the

bilingual mental lexicon. To that end, two classes of words, viz., cognates and false

friends, have been under close inspection by language researchers that, for the past

few decades, have persistently tried to further their understanding of how L1 and L2

words are stored and organized in the bilingual mental lexicon. Cognates are words

whose meaning is largely the same crosslinguistically, i.e., have near complete

semantic overlap. Likewise, their orthographic overlap (form similarity) is also high,

sometimes complete. However, their phonological overlap (pronunciation

similarity), which sometimes can be high, is subject to more variation

crosslinguistically (da Luz Fontes & Schwartz, 2010; Dijkstra, Grainger, & Van

Heuven, 1999; Dijkstra & van Heuven, 2002; Schepens, Dijkstra, Grootjen, & van

Heuven, 2013). For example, the English word kit [kɪt] and the Portuguese word kit

[ˈki.tʃi] have complete crosslinguistic semantic and orthographic overlap.

 32

EFFECTS OF CROSSLINGUISTIC OVERLAP

Specifically, their meaning is largely the same and their orthographic form is identical

across both languages. However, there is variation in how speakers of Portuguese and

English pronounce both words. Like cognates, false friends (often characterized as

interlingual homographs3) have a high degree of orthographic overlap and varying

degrees of phonological overlap crosslinguistically. However, unlike cognates, false

friends denote different concepts across languages, i.e., their semantic overlap is

rather low or null (da Luz Fontes & Schwartz, 2010; Dijkstra et al., 1999; Dijkstra &

van Heuven, 2002; Schepens et al., 2013). For example, while the English word rim

[ɹĩm] and the Portuguese word rim [χĩŋ] ‘kidney’ have complete orthographic overlap

and partial phonological overlap, their meanings are completely different across both

languages.

In the research literature cognates appear to have a clear processing

advantage. They generally yield significantly more robust and more consistent

priming effects (de Groot & Nas, 1991; Gollan et al., 1997; Kim & Davis, 2003,

Sánchez-Casas, Davis, & García-Albea, 1992). Cognates and false friends, either as

primes or targets in LDTs, have often been compared against controls and

pseudowords (Dimitropoulou et al., 2011b; Ferrand & Grainger, 1993; Grainger,

Kiyonaga, & Holcomb, 2006; Grainger, Spinelli, Farioli, Diependaele, & Ferrand,

3
Interlingual homographs are sometimes referred to in the literature as interlingual
pseudohomophones (Dijkstra et al., 1999).
 33

EFFECTS OF CROSSLINGUISTIC OVERLAP

2003; Perfetti & Tan, 1998; Pollatsek, Perea, & Carreiras, 2005; Ziegler, Ferrand,

Jacobs, Rey, & Grainger, 2000). Pseudowords are words that have been manipulated

either via computer software or manually. They resemble real words in a specific

language in terms of both their form and pronunciation.

When orthographic overlap is the primary focus of research, facilitatory

effects, viz., faster reaction times, have been reported for cognates with partial as

well as 100% orthographic overlap, e.g., the Dutch-English cognate pair tomaat-

tomato, in comparison to control words (Cristoffanini, Kirsner, & Milech, 1986;

Dijkstra et al., 1999; Lemhöfer & Dijkstra, 2004; Van Hell & Dijkstra, 2002). When

cross-linguistic phonological overlap is investigated, such as when cognates are

pronounced similarly across two languages, evidence from LDTs shows faster

reaction times as well, suggesting that phonological representations might be

activated during visual word recognition, and, consequently, might play a role in the

organization of the bilingual mental lexicon. This facilitatory effect is observed even

with languages pairs that do not share the same orthographic script, such as Greek

and Spanish, Japanese and English, as well as Greek and French (Dimitropoulou et

al., 2011a; Nakayama, Sears, Hino, & Lupker, 2012; Voga & Grainger, 2007).

Costa, Caramazza and Sebastián-Gallés (2000) theorize that complete

semantic overlap between a prime in one language and a target in the other could

potentially serve as proof that activation spreads backward from semantic to

orthographic (and potentially to phonological) representations, accounting for

 34

EFFECTS OF CROSSLINGUISTIC OVERLAP

some of the facilitatory effects encountered in the literature. Nevertheless, Voga and

Grainger (2007) present a different theory. Specifically, they argue that the cognate

advantage over noncognates may simply be the result of strong morphological,

orthographic and phonological relationships that exist between primes and targets

across some language pairs. It is possible that words that share a common

etymological root also share a common morphemic representation in memory, thus,

are organized differently in the bilingual mental lexicon.

In sum, research on cognates has thus far provided, to a certain extent,

valuable evidence for language non-selectivity, i.e., bilinguals’ two languages appear

to be simultaneously activated during visual word recognition. This parallel

activation points to a continuous state of co-activity in the bilingual mental lexicon,

in which not only semantic representations but also orthographic and phonological

representations are also activated as a result of spreading activation when a word

from one of bilinguals’ two languages is presented (de Groot, 1992; J. F. Kroll &

Stewart, 1994; Perea, Duñabeitia, & Carreiras, 2008; Soares & Grosjean, 1984; van Hell

& de Groot, 1998). Nevertheless, teasing apart the individual contributions of

crosslinguistic semantic, orthographic and phonological similarity has remained a

challenge due to the vastly different methodologies employed by language

researchers.
 35

EFFECTS OF CROSSLINGUISTIC OVERLAP

2.5. The Current Study

In light of the framework provided primarily by the BIA and BIA+, and to some extent

Multilink, the primary goal of this study is to quantify the individual effects of

crosslinguistic orthographic, phonological and semantic similarity on the

organization of the bilingual mental lexicon. In addition, this study seeks to

understand whether language proficiency plays a significant role in this organization.

To accomplish that, Portuguese-English cognates and false friends were selected as

experimental stimuli for a cross-language LDT with masked priming, for which

Portuguese-English bilinguals and English functional monolinguals were selected t0

participate.

To date no study has concerned primarily with Portuguese-English cognates

and false friends in a cross-language LDT, despite the great deal of crosslinguistic

similarity between the two languages. As Brazil, the most populous Portuguese

speaking country, has become one of the world’s most powerful economies,

consistently ranking among the top 10, and experienced dramatic economic growth,

the country’s participation in the world’s economic stage has also become more

prominent (“The World’s Top 10 Largest Economies,” n.d.). As a result, the number of

native speakers of Portuguese receiving training to become Portuguese-English

bilinguals has also increased. Thus, a more scientific understanding of how words

from two languages are stored and interact is needed and it will certainly assist in the

development of better teaching materials and linguistic outcomes.

 36

EFFECTS OF CROSSLINGUISTIC OVERLAP

Despite the fact that the likelihood that two words across two different

languages are cognates is highly correlated with how similarly they are pronounced

(Kondrak & Sherif, 2006), crosslinguistic phonological similarity has not received

much attention from language researchers; thus, its individual effect on bilingual

mental organization has not been clearly established or understood (Dimitropoulou

et al., 2011a; Lemhöfer & Dijkstra, 2004; Nakayama et al., 2012; Van Hell & Dijkstra,

2002; Voga & Grainger, 2007).

Because there currently is no standard crosslinguistic phonological similarity

metric readily available for language research, researchers have opted to reduce

crosslinguistic phonological similarity to segmental similarity and explain it in terms

of distinctive features and other formal primitives of phonological theory (Chomsky

& Halle, 1992; G. Clements & Hume, 1995; G. N. Clements, 1985; inter alia). In fact,

influential studies on the bilingual mental lexicon have, for the most part, either

relied on this traditional framework for assessing crosslinguistic phonological

similarity or on native and non-native speakers’ personal intuitions for selecting

cognates as stimuli for their experiments (Comesaña et al., 2012; Friesen & Jared, 2012;

Pallier, Colomé, & Sebastián-Gallés, 2001; Pexman, Lupker, & Jared, 2001; Schwartz &

Kroll, 2006; Schwartz, Kroll, & Diaz, 2007; inter alia).

Unfortunately, employing subjective measures of crosslinguistic phonological

similarity pose numerous challenges and limit the results of these studies, in

particular, because the mechanisms involved in speech perception and production

 37

EFFECTS OF CROSSLINGUISTIC OVERLAP

are far more multidimensional than what is postulated by phonological theory and

personal judgments are not only often biased but also based on the personal

experiences of the speakers of the languages being studied (Bradlow, Clopper,

Smiljanic, & Walter, 2010). These pitfalls, however, can be successfully overcome if an

objective metric is used instead. To that end, dynamic time warping (DTW), an

algorithm that was originally conceived for automatic speech recognition to measure

acoustic similarity between two waveforms, was leveraged to create an objective

crosslinguistic phonological similarity metric. This approach is largely an offshoot of

the technique developed by Mielke (2012).

To better understand the individual effect of crosslinguistic orthographic

similarity as well as its interplay with crosslinguistic phonological similarity on

bilingual lexical organization, it is essential that an objective metric of crosslinguistic

orthographic similarity also be used. To that end, in light of the work of Schepens,

Dijkstra and Grootjen (2013), the Normalized Levenshtein Distance (NLD) was

chosen.

2.5.1. Experimental Predictions and Hypotheses

Taking into account spreading activation as well as the framework of two well-

known models of the bilingual mental lexicon discussed in 2.2., specifically the BIA

and BIA+, this study hypothesizes that activation in the bilingual mental lexicon

spreads in a feed-forward fashion from the semantic store, which is the site where

bilinguals’ two languages are posited to be integrated, to the orthographic

 38

EFFECTS OF CROSSLINGUISTIC OVERLAP

representations from both bilinguals’ two languages during visual word recognition.

Phonological representations are activated via orthographic representations, not

through the semantic store. In this framework, cognates will have a clear processing

advantage over false friends. Specifically, Portuguese-English bilinguals will

recognize an English target significantly faster when it is matched to a Portuguese

cognate prime compared to when it is matched to a Portuguese false friend prime.

The shared semantics coupled with orthographic and phonological code overlap will

accelerate the recognition of English cognate targets. When an English target is

matched to a Portuguese false friend prime, the orthographic overlap and

phonological overlap will still facilitate the processing of the English target, but the

advantage will be less pronounced than for Portuguese cognate primes due to no

semantic overlap. Bilingual participants should recognize English targets matched to

a semantically unrelated Portuguese prime with low or null orthographic and

phonological overlap with the English target (control pairs) slower because the

orthographic and phonological code overlap is minimum or absent. Finally, bilingual

participants will recognize pseudoword targets matched to a Portuguese prime

(distractor pairs) the slowest due to the absence of facilitation provided by semantic,

orthographic and phonological overlap. These predictions are detailed in Table 2

below.
 39

EFFECTS OF CROSSLINGUISTIC OVERLAP

Table 2. Experimental Predictions (Hypotheses)

Crosslinguistic Overlap Bilinguals’

with English Target Rec0gnition of
Prime Category English Targets
Semantic Orthographic Phonological (Reaction
Time)

Portuguese partial or partial or

complete fastest
cognate complete complete

Portuguese partial or partial or

absent second fastest
false friend complete complete

Portuguese word
matched with
minimal or minimal or
unrelated English absent slower
none none
target
(control pairs)
Portuguese word
matched with
absent none none slowest
pseudoword
(distractor pairs)

In the extant literature, ANOVAs and t-tests have been the primary statistical

analysis (means-based parametric) tools used to analyze bilingual data collected in

LDTs, and the reliance on these tools has continued to this day. Because one of the

primary goals of this study is to quantify the individual effects of crosslinguistic

similarity and because many of the variables of interest are continuous predictors, or

a combination of categorical and continuous predictors, this study will instead utilize

linear mixed-effects models, a much more robust statistical analysis tool, to analyze

the experimental data collected. Besides more precisely estimating the individual
 40

EFFECTS OF CROSSLINGUISTIC OVERLAP

effects of crosslinguistic similarity (fixed factors) on both dependent variables

(reaction time and accuracy), mixed-effects models will enable participants and

prime-target pairs (items) to be treated as random effects, which will allow the

individual effects of crosslinguistic similarity to be further assessed by each

individual participant and each individual prime-target pair, which in the end will

yield a more realistic overall representation of the processes involved in visual word

recognition in the bilingual mental lexicon.

The following chapter provides a thorough account of the study’s

methodology, from the creation of the stimuli lists for the LDT using objective

metrics of crosslinguistic orthographic and phonological similarity to recruiting both

Portuguese-English bilinguals and functional English monolingual participants as

well as assessing their level of English proficiency. It also describes the experimental

procedure used in the cross-language LDT.

 41

EFFECTS OF CROSSLINGUISTIC OVERLAP

3. The Study: Methodology

3.1. Participants

3.1.1. Recruitment

Participants for the experiment were recruited primarily via a recruitment email,4

which provided potential participants with a brief description of the experiment,

qualification requirements, a step-by-step account of what their participation would

entail, and instructed them to contact the researcher via email to express their

interest in serving as volunteers. In order to recruit Portuguese-L1 English-L2 late

sequential bilinguals, the researcher sent out the recruitment email (see Appendix B)

to the International Office at the University of Ottawa, and requested that they

forward the message to any current international graduate or undergraduate student

whose native language is Portuguese. Since a large proportion of students with this

language profile conducts their studies at the University in English, they were

presumed to possess a level of English that allows them to function in an academic

environment, which qualified them to participate in the experiment.

To serve as functional monolingual controls, the researcher sent out the

recruitment email to the teaching staff of the English Intensive Program (EIP), a non-

credit English as a second language program in the Official Languages and

4
The text included in recruitment email was approved by the University of Ottawa Office of
Ethics and Research Integrity (file number 09-12-13). The approval notice is included in
Appendix A.
 42

EFFECTS OF CROSSLINGUISTIC OVERLAP

Bilingualism Institute at the University of Ottawa. The EIP is designed for

international students and francophone Canadians who do not meet the English

proficiency requirements of the University so that they might attend classes either at

the graduate or undergraduate level. Teachers in the EIP are highly trained

professionals who must at a minimum have a Masters in linguistics, second language

education, or related disciplines in order to be a member of the teaching staff, and

many of them are native speakers of English.

It is important to note that finding potential monolingual participants in the

National Capital Region in Canada is a rather dauting endeavor, as residents of the

area are exposed to both official languages, English and French, on a daily basis.

Street signs, advertisements, pamphlets, bus schedules, routes, etc. are omnipresent

in both languages. To further complicate matters, Canada has a rather ethnically

diverse and multicultural population, so many native speakers of English are also

heritage speakers of another language, i.e., they can converse with their parents and

family members in another language with varying degrees of fluency. Thus, it is

virtually impossible to find bona-fide native speakers of English who do not know,

speak, and are exposed to another language to serve as monolingual controls. To

minimize the effects of exposure and having some knowledge of a second language,

potential English monolingual participants were only allowed in the experiment if

they disclosed during a prescreening interview that they could not comfortably
 43

EFFECTS OF CROSSLINGUISTIC OVERLAP

function in an environment that would require them to use their second language in

the same manner as they would English.

Many participants during the prescreening interview indicated that they

knew other potential participants that met the language requirements of the

experiment and who would probably be interested in serving as volunteers.

Therefore, a few participants in both experimental groups actually became aware of

the experiment via word-of-mouth.

3.1.2. Composition

In total 56 participants, specifically, 29 L1-Portuguese L2-English late

sequential bilinguals as well as 27 L1-English functional monolinguals, participated

in the experiment. All of them were 18 years of age or older at the time of their

participation. All bilingual participants were born and grew up in country which

Portuguese is the official language and indicated they had learned English as a second

language in private language schools in their home country during or after reaching

puberty. A considerable number of bilingual participants from Brazil were enrolled

at the University of Ottawa under the Science Without Boarders program, a large-

scale nationwide scholarship program primarily funded by the Brazilian federal

government. The program seeks to strengthen and expand the initiatives of science

and technology, innovation and competitiveness through international mobility of

undergraduate and graduate students and researchers. Many of the students enrolled
 44

EFFECTS OF CROSSLINGUISTIC OVERLAP

under this program at the University of Ottawa obtained scores in the Test of English

as a Foreign Language (TOEFL) or the International English Language Testing

System (IELTS) that satisfied the English proficiency requirements of the University.

Students who did not take or obtained a low score either of these tests had to

complete a few sessions in the EIP in order to satisfy the English proficiency

requirements of the University before they could start taking classes at either the

undergraduate or graduate level.

Most participants from Brazil were from the State of São Paulo. All

monolingual participants were born and grew up in a country in which English is one

of the official languages.

3.1.3. Language Background Profile and Language Proficiency

Assessment

In order to establish participants' language background profile, all

participants completed an abridged version the Language Background

Questionnaire (see Appendix D), which was originally developed for the

psycholinguistic experiments conducted in the Brain and Language Laboratory in the

Department of Linguistics at the University of Ottawa and was previously used in

Alves-Soares (2013). The questionnaire was converted to Google Forms so that

participants could fill it out online on a computer and to make it convenient for them

to complete it in a timelier fashion. To assess participants' overall level of proficiency

 45

EFFECTS OF CROSSLINGUISTIC OVERLAP

in English, consistent with the methodology employed in Alves-Soares (2013), all

participants were also required to take an integrative language test, specifically

Brown’s (1980) cloze test (see Appendix E). This kind of test was selected for the

purposes of this experiment because it is easily administered and provides an overall

picture of communicative competency and overall linguistic ability. In addition, cloze

tests require that test takers make highly complex series of grammatical and lexical

decisions, specifically having to deduce the deleted words based upon all available

contextual clues, which are posited to reflect test takers’ acquired language skills (Litz

& Smith, 2006).

Brown’s (1980) cloze test consists of a short passage of approximately 400

words in length and with 50 missing words or blanks entitled Man and His Progress

(see Appendix E). Participants were instructed to first read the passage as a whole to

get the general meaning and then fill in the blanks with the words they judged

appropriate for the context. They were given 30 minutes complete the test.

Participants’ answers were scored using both the semantically acceptable

scoring procedure (SEMAC) and the exact replacement scoring method (ERS). In the

SEMAC procedure, the test taker’s answers are compared with the original deleted

words. If the supplied words preserve the original intended meaning, even if they are

not direct synonyms of the deleted words, they are deemed acceptable and marked

as correct answers. In the ERS method only supplied words that are exactly identical

to the deleted words are marked as correct answers. It is important to note that
 46

EFFECTS OF CROSSLINGUISTIC OVERLAP

SEMAC and ERS scores correlate very highly with one another (Litz & Smith, 2006).

Table 3 below provides further details about the participant population.

Table 3. Participant Population Breakdown

Age Cloze Proficiency Score

Native Country
N Mean (with Mean (with Std. Dev.)
Language of Origin
Range) ERS SEMAC

Brazil 26 33 (23-56) 49.67 (11.91) 75.83 (16.92)

Portugal 2 36 (33-39) 43.00 (1.00) 78.00 (6.00)

Portuguese
The
1 41 68.00 78.00
Azores

Canada 23 38.65(25-66) 62.23 (9.43) 92.46 (5.33)

English United
4 54.75 (46-64) 70.50 (2.19) 94.50 (2.59)
States

3.2. Materials/Stimuli

The stimuli list for the experiment was created in six stages: (1) compiling an

experimental items list comprised of Portuguese primes - English targets; (2)

matching each English target with an unrelated Portuguese prime to serve as

controls; (3) compiling a list of pseudowords and matching each of them with a

semantic and orthographic unrelated Portuguese prime to serve as distractors; (4)

compiling a list of practice pairs; (5) obtaining a frequency-per-million-words count

 47

EFFECTS OF CROSSLINGUISTIC OVERLAP

for each Portuguese and English word; (6) obtaining an orthographic and an acoustic

similarity overlap value for each prime-target pair.

3.2.1. Assembling Word Lists A and B

3.2.1.1. Experimental Items List

This study is primarily interested in two classes of word with respect to their

form and meaning overlap across languages: cognates and false friends. As indicated

in 2.4, cognates are words whose meaning remains conceptually stable

crosslinguistically, i.e., have near complete semantic overlap. In addition, their

spelling and sound similarities (orthographic and phonological overlap, respectively)

are also high, sometimes complete (da Luz Fontes & Schwartz, 2010; Dijkstra et al.,

1999; Dijkstra & van Heuven, 2002; Schepens et al., 2013). For example, the

Portuguese word prisão [pri.ˈzãũ] forms a cognate pair with the English word prison

[ˈprɪ.zə̃n], since their semantic, orthographic and phonological overlap remains

stable across both languages. False friends, in contrast, are words that very much

resemble cognates in terms of form similarity, i.e., they also possess a high degree of

orthographic and phonological overlap, but they denote different concepts across

languages (da Luz Fontes & Schwartz, 2010; Dijkstra et al., 1999; Dijkstra & van

Heuven, 2002; Schepens et al., 2013). For example, the Portuguese noun sapo [ˈsa.pʊ]

‘frog’ shares many of its letters and phonemes with the English noun sap [sæp] ‘fluid

that circulates in the vascular system of a plant.’ However, their form similarities are
 48

EFFECTS OF CROSSLINGUISTIC OVERLAP

only superficial because the two words ultimately represent different concepts in

both languages.

For the current study, an experimental items list comprised of Portuguese–

English cognates and false friends was assembled using the official word list utilized

in Alves-Soares (2013) as its foundation. The experimental stimuli on that list were

comprised of Portuguese-English cognates and false friends. Specifically, Portuguese

words served as primes; English words served as targets. As expected, the selected

cognates and false friends on that earlier list had various degrees of crosslinguistic

orthographic and phonological overlap.

To enhance the earlier list utilized in Alves-Soares (2013), an additional

number of Portuguese-English cognates and false friends with various degrees of

crosslinguistic orthographic and phonological overlap were chosen in consultation

with native speakers of Portuguese who speak English as a second language

(Portuguese-English sequential bilinguals). The pairs from both lists were then

consolidated into a master Excel spreadsheet and classified as either cognates or false

friends, which became the experimental items list for the current study. Table 4 below

provides a sample of the experimental items list. The full list is provided in Appendix

F.
 49

EFFECTS OF CROSSLINGUISTIC OVERLAP

Table 4. Portuguese primes – English targets (examples)

Classification Portuguese primes English targets

alarme alarm
bagagem baggage
coma coma
computador computer
dama dame
face face
gelatina gelatin
Cognates irado irate
memória memory
normal normal
paciência patience
papel paper
robô robot
veia vein
zebra zebra
batom ‘lipstick’ baton
braço ‘arm’ brass
carta ‘letter’ cart
chato ‘boring/annoying’ chat
êxito ‘success/achievement’ exit
fábrica ‘factory’ fabric
gripe ‘cold/flu’ grip
False Friends hospício ‘psychiatric hospital’ hospice
limpo ‘clean’ limp
mal ‘bad/disorder’ mall
máscara ‘mask’ mascara
rim ‘kidney’ rim
sapo ‘frog’ sap
sorte ‘luck’ sort
tampão ‘lid’ tampon
 50

EFFECTS OF CROSSLINGUISTIC OVERLAP

3.2.1.2. Control Items List

To serve as experimental controls, in consultation with native speakers of

Portuguese who speak English as a second language (Portuguese-English sequential

bilinguals), each of the Portuguese words in the experimental items list was matched

with a semantically unrelated English word that did not orthographically resemble

the Portuguese word, i.e., had a different spelling. For example, the Portuguese word

rim ‘kidney’ in the false friend pair rim-rim was matched with the semantically and

orthographically unrelated English word sun, yielding the control pair rim-sun. The

full list of control pairs is provided in Appendix F.

3.2.1.3. Distractor Items List

Wuggy (Keuleers & Brysbaert, 2010) a multilingual pseudoword generator, was

used to generate pseudowords that resembled genuine English words as closely as

possible. The pseudowords were then matched with Portuguese words that were not

in the experimental and control items lists. The total number of Portuguese prime –

pseudoword target pairs matched the total combined number of experimental and

control pairs together. The full list of distractor pairs is provided in Appendix F.

3.2.1.4. Practice Items List

To familiarize participants with the experimental task and before each trial

began, practice items were used. To that end, 36 practice pairs were created. Thirty-
 51

EFFECTS OF CROSSLINGUISTIC OVERLAP

six words of Portuguese that had not been previously used as experimental, control

or distractor items served as primes. They were matched with 18 unrelated words of

English that had not been previously included in the preceding lists. The remaining

18 Portuguese primes were matched with pseudowords that were generated by

Wuggy. The full list of practice pairs is provided in Appendix F.

3.2.1.5. Official Word Lists A and B

Once the creation of the experimental, control, distractor and practice items

lists was complete, two official word lists (A and B) were created for the experimental

task. One half of the experimental and control pairs as well as all of distractor and

practice pairs formed list A. Conversely, the second half of the experimental and

control pairs as well as all of the distractor and practice pairs formed list B. It is

important to note that while all distractor and practice pairs in both lists were exactly

the same, the experimental and control pairs in list A were different in list B. As Table

5 shows, each official word list contained a total of 624 Portuguese prime - English

target pairs, specifically, 147 experimental, 147 control, 294 distractor and 36 practice

pairs.
 52

EFFECTS OF CROSSLINGUISTIC OVERLAP

Table 5. Word Lists Breakdown

Word List Prime-Target Pairs N

Cognates 83
Experimental
False friends 64
A
Control 147

Distractor 294

Practice 36

Total 624

Cognates 83
Experimental
False friends 64
B
Control 147

Distractor 294

Practice 36

Total 624

All the experimental and control pairs from both word lists as well as all the

distractor and practice pairs were consolidated into a master Excel spreadsheet. This

was done to further facilitate data entry, which will be explained in the following

subsections. See Appendix F for the complete word lists.

 53

EFFECTS OF CROSSLINGUISTIC OVERLAP

3.3. Obtaining Word Frequency Count

A word frequency count was obtained for each Portuguese-English cognate and false

friend in the official word lists. This was done because word frequency has major

implications for this study. Specifically, the frequency in which speakers of a language

are exposed to words, particularly in reading, may shape the organization of their

mental lexicon. Words that are more frequently encountered are more salient to

speakers of that language, which probably increases their baseline level of activation.

In addition, since a vast number of words in Portuguese and English are of Latin

roots, it is possible the effects of word frequency interact with the degree of

orthographic (and possibly phonological) overlap.

A word frequency count is obtained by measuring how often a particular word

occurs per million words in collected corpora (Schepens et al., 2013). To that end, to

obtain the frequency of occurrence of each word in the preliminary word list O corpus

do Português5 (Davies, 2017b), a large online corpus of Portuguese words, was used to

obtain word frequency count for the Portuguese words, while Corpus of

Contemporary American English (COCA6) (Davies, 2017a), was used to obtain word

5
O Corpus do Português is a one-billion-word corpus of Brazilian and European Portuguese
words. It was created by Mark Davies (Brigham Young University) and funded by the US
National Endowment for the Humanities (Davies, 2017b).
6
Corpus of Contemporary American English (COCA) is the largest freely available and balanced
corpus of American English. It contains over 560 million words from spoken, fiction, popular
magazines, newspapers, and academic texts (Davies, 2017a).
 54

EFFECTS OF CROSSLINGUISTIC OVERLAP

frequency count for the English words. The word frequency count for each of the

Portuguese-English (prime-target) word pairs was then entered into in the master

spreadsheet in two separate columns, one for Portuguese and one for English. It is

important to note that no word frequency count was obtained for the pseudowords

(distractor items) in this experiment, since the selected pseudowords are not actually

real, bona fide English words.

3.4. Obtaining Crosslinguistic Orthographic Similarity Estimates

Crosslinguistic orthographic similarity is concerned with how much spelling

similarity, how many letters are shared between word pairs from different languages.

Thus, orthographically similar words, also referred to as homographs, have a high

degree of orthographic overlap (Schepens et al., 2013). The degree of crosslinguistic

orthographic similarity is often calculated using the orthographic similarity metric

(OS), an algorithm that was adapted from the graphic similarity index (GS), which

was developed by Weber (1970). The original GS index, which was devised as an

attempt to explain the degree to which reading substitution errors approximated

participants’ correct responses in terms of letters, includes seven sub-indices to

which Weber assigned distinctive weights based upon her own intuitions about the

importance of cues purported to be used during word identification. Weber’s GS

index is expressed by the following algorithm and comprises the following indices, as

Example 1 below shows.

 55

EFFECTS OF CROSSLINGUISTIC OVERLAP

Example 1. Weber’s Grapheme Similarity (GS) Index

[50) + 30, + 10-)

!" = 10 & 0 + 51 + 274 + 186)
/

A: average number of letters between two words.

B: 1 if the first letter of the two words is the same; otherwise, B = 0

C: number of single letters shared by the two words.

E: 1 if the last letter of the two words is the same; otherwise, E = 0

F: number of pairs of adjacent letters in the same order shared by the two words.

T: ratio of number of letters in the shorter word to the number of letters in the
longer one.
V: number of pairs of adjacent letters in reverse order shared the two words.

The GS index between two words ranges from 0 [zero] (no common letters) to

1,400 (identical letters). However, the application of the formula by itself can yield

results that are difficult to interpret because the formula is sensitive to word length.

Specifically, not all word pairs comprised of identical words will have the same GS

index (Guasch, Boada, Ferré, & Sánchez-Casas, 2013). For example, the GS index of a

word like trim with itself is 975, whereas the GS index of a word like grime with itself

is 1,000. In order to address Weber’s GS index, van Orden (1987) developed an

 56

EFFECTS OF CROSSLINGUISTIC OVERLAP

improved orthographic similarity algorithm that is based on Weber’s GS index

algorithm. Van Ordern’s algorithm (OS) yields similarity values that vary between 0

(no shared letters – minimum orthographic similarity) and 1 (identical letters –

maximum orthographic similarity) and are not sensitive to word length(Guasch et

al., 2013). Van Orden’s algorithm is expressed below, where P is the first word of a pair

and R the second word, as in Example 2 below.

Example 2. Van Ordern’s Orthographic Similarity (OS) Algorithm

!"(9, ;)
7" =
!"(;, ;)

Van Orden’s algorithm, however, lacks an important property of similarity: the

commutative property. Specifically, when comparing the words in a word pair,

reversing the order in which they are input in the algorithm yields completely

different orthographic similarity indices. With this drawback in mind, more recent

studies have started to utilize the Levenshtein distance (Levenshtein, 1966) as a

metric of orthographic similarity. Simply put, the Levenshtein distance between two

words is the minimum number of single-character insertions, deletions or

substitutions required to change one word into the other. Still, the metric is highly

sensitive to word length. In order to neutralize those effects, Schepens et al. (2013)

devised the normalized Levenshtein distance (NLD), which normalizes the metric
 57

EFFECTS OF CROSSLINGUISTIC OVERLAP

and yields similarity scores that always range between 0 [zero] (when word pairs are

orthographically dissimilar) and 1 (when word pairs are orthographically identical)

irrespective of word length (Guasch et al., 2013). The NLD is calculated according to

the algorithm shown in Example 3 below.

Example 3. Normalized Levenshtein Distance (NLD)

=@A@BCℎE@FB HFCEIBJ@ (CEKFBL 1, CEKFBL 2)

<=> = 1 −
MINFOPO CEKFBL Q@BLEℎ (CEKFBL 1, CEKFBL 2)

Since one of the major goals of the current study is to further understand how

orthographic similarity between Portuguese-English cognates and false friends can

mirror the organization of the bilingual mental lexicon, needless to say, choosing the

most reliable as well as the most up-to-date metric of orthographic similarity was

paramount. For this reason, the NLD seemed the most apposite orthographic

similarity metric to use. To that end, NIM (Guasch et al., 2013), a web-based software

developed by psycholinguists in the Research Center for Behavior Assessment at

Universitat Rovira iVirgili in Tarragona, Spain, was utilized to calculate the NLD

between the word pairs in the preliminary word list. One of the greatest features of

NIM is the fact that it calculates both Levenshtein distance and NLD as well as

Weber’s GS and van Orden’s OS indices. In addition, researchers can copy and paste

lists of word pairs directly into NIM’s interface, and the software will output a table
 58

EFFECTS OF CROSSLINGUISTIC OVERLAP

with all the orthographic similarity indices, which can also be downloaded in

Microsoft Excel format. Once the orthographic similarity values were obtained for all

the word pairs in both word lists, they were entered into the master Excel spreadsheet

according to their associated word pairs.

3.5. Obtaining Crosslinguistic Phonological Similarity Estimates

Phonological similarity (or distance), despite lacking a formally agreed upon

definition in the literature, is often reduced to segmental similarity and explained in

terms of distinctive features (Chomsky & Halle, 1992; Clements & Hume, 1995; G. N.

Clements, 1985, inter alia) and other formal primitives of phonological theory.

Assessing crosslinguistic phonological similarity via this traditional framework poses

numerous challenges, in particular, because native and non-native speakers’

mechanisms involved in speech perception and production are more

multidimensional, taking into account not only segmental features but also prosodic

and phonotactic properties (Bradlow et al., 2010; Mielke, 2012).

Because there currently is no standard phonological similarity metric readily

available, many influential studies on the bilingual mental lexicon have often relied

on subjective measures of phonological similarity, i.e., authors’ and consultants’

personal judgments as both native and non-native speakers, for crosslinguistic

stimuli selection (cf., Comesaña et al., 2012; Friesen & Jared, 2012; Schwartz & Kroll,

2006; Schwartz, Kroll, & Diaz, 2007; inter alia). The present study avoided this
 59

EFFECTS OF CROSSLINGUISTIC OVERLAP

shortcoming by computing an objective measure of crosslinguistic phonological

similarity using a technique previously employed by Mielke (2012), viz., dynamic time

warping (DTW), which was originally conceived to compare different speech patterns

in automatic speech recognition and measures acoustic similarity between two time-

dependent sequences, such as sound waveforms. In addition, because the likelihood

that two words across different languages are cognates is highly correlated with their

phonological similarity (Kondrak & Sherif, 2006), it is of utmost importance to this

study to employ an objective metric such as DTW to measure phonological similarity.

DTW in its essence is a time sequence alignment algorithm that, for the

purposes of the current study, iteratively aligns two wave files in a distance matrix by

warping them at their beginning as well as at their end until an optimal match

between the two wave files is found (Shinde & Pawar, 2014). To that end, the optimal

path between two acoustically similar wave files is a diagonal line. The more the

optimal path deviates from a diagonal line, the less acoustically similar two wave files

are. Figure 8 below illustrates the dynamically time warped the waveforms of the

Portuguese-English cognate pair grafite-graphite and the false friend pair smoking

‘tuxedo’-smoking recorded by native speakers of the respective languages. For

comparison purposes, the dynamically time warped waveforms of the Portuguese

word grafite with itself and the English word smoking with itself are also provided.

Although the optimal path between the cognate pair and the false friend pair is not

a straight diagonal line, it deviates less from a diagonal line in the false friend pair
 60

EFFECTS OF CROSSLINGUISTIC OVERLAP

smoking-smoking, since the pronunciation of smoking in Portuguese is more

acoustically similar to the English pronunciation of smoking. The optimal path for

the cognate pair grafite-graphite starts rather jagged and completely deviates from

a diagonal to a horizontal line after the fricative [f], indicating that the

pronunciation of the Portuguese word grafite is less acoustically similar to the

pronunciation the English word smoking. When Portuguese grafite and English

smoking are compared with themselves, the optimal path is a straight diagonal line.
 61

EFFECTS OF CROSSLINGUISTIC OVERLAP

Portuguese-English Cognates Portuguese-English False Friends

DTW of the Portuguese word grafite DTW of the Portuguese word smoking
with the English word grafite with the English word smoking
grafite [ˈɡræ.faɪt]

smoking
0.732404209

grafite [ɡra.ˈfi.tʃi] smoking [ĩʃ.ˈmõũ.kĩŋ.ɡɪ] ‘tuxedo’

DTW of the Portuguese word grafite DTW of the English word smoking with
with itself itself
grafite [ɡra.ˈfi.tʃi]

smoking

grafite [ɡra.ˈfi.tʃi] smoking

Figure 8: Dynamic time warping illustration of the English-Portuguese cognate pair

grafite-graphite and the false friend pair smoking ‘tuxedo’-smoking. Areas that are
darker represent points where the accoustic distance between the waveforms is less
(greater acoustic similarity).
 62

EFFECTS OF CROSSLINGUISTIC OVERLAP

The DTW algorithm is usually implemented via software packages that

quantify the degree to which waveforms of linguistic objects (sounds or words) are

similar to each other. For the purposes of this study, Phonological Corpus Tools (PCT)

(Hall, Allen, Fry, Mackie, & McAuliffe, 2016) was chosen for its easy-to-use interface,

in particular, for its ability to perform all the acoustic-similarity-related

computations unaided once a tab-delimited file that lists all the pairwise

comparisons of individual waveforms to be compared is chosen. In addition, PCT also

allows users to customize the acoustic analysis according to specific parameters

related to the acoustic properties of a sound. The first of these customization

parameters is frequency limits (or range), where researchers can set the minimum

and the maximum frequency range (in Hz) for the acoustic analysis. According to

Johnson (2011), human sensitivity to frequencies above 10,000 Hz is rather limited.

Thus, “frequency components above 10,000 Hz are not likely to be useful for speech

communication even if the speaker has perfect hearing” (p. 23). In light of this fact,

this study set the limits of the frequency range to 80Hz minimum (to filter out low

background noise) and 10,000 Hz maximum. The second customization parameter

is frequency resolution, where researchers can set the number of filters to be used to

divide up the frequency range they previously specified. The default setting of 26

filters was chosen for this study. The third parameter is the number of Mel-frequency

cepstral coefficients (MFCCs), where researchers can set the number MFCCs they

desire, which are a type of representation of a sound that is based on the human ear
 63

EFFECTS OF CROSSLINGUISTIC OVERLAP

response to sounds and are extensively used in automatic speech and speaker

recognition systems. The default setting of 12 coefficients was chosen, which is

consistent with the methodology in Mielke (2012). The fourth customization

parameter is whether PCT should output the acoustic similarity results in a scale that

ranges between 0 [zero] and 1, or use the default inverse similarity scale, which was

chosen for this study. Finally, researchers can request that PCT generates a results

table in the form of a tab-delimited text file, which can be saved on to the computer

and later easily imported into Excel. Figure 9 below is a screen capture of the several

parameters users can set in PCT. Figure 10 illustrates the results table PCT generates

once it completes all the acoustic similarity computations in the pairwise comparison

list.
 64

EFFECTS OF CROSSLINGUISTIC OVERLAP

Acoustic Similarity

Compariso n type
Represena tion
Q ..'\nalyze sIrq e d irecto ry
MFCC
o .recto ry :
Q /\mpl nude envelopes

c··,oose {lirec to --y.. Distance algorithm

Dynamic nme warp ing

Q Compare t\VO d recw nes
F rs1 d ir ectory :
Q Cross-c orrelat ion
Frequency limi ts
c··,oose oirec to--y..
rvhru1rn.•rn treq1..er·
cy (Hz) : JL
s_o______ __,
Secorw directory :
Max im u m freq uency ('-lz) : "1
1-"
Q-"Q-"
Q-"Q____ __,

c··,oose oirec to--y.. Freauency reso lut ion

Use I st o~ ~u oath comparisons Nu mbe r ol f ters : Ll

2_6_______ ~
~
go;Lst 1i Ph 1- E1r 1 lisi 1.t xtl I C ··,oo se ' 1le.. Nu mbe r ol coefl 1c en ts (l,F CC only) : Ll1_2_______

0 Ou tp u t as s11111I
a- ty \0 10 ·1)
/\cou st c sIir11ar ty bere t ts from M ult 1proces s,nq .
Me li voce ssIr g canoe enable<I In P-eferences .

Ca cu a te aCOL:SHC S,r'";)l
larn y Ca cu ate aco1.. st1c s:--11larny
(start ne·N resu ts tao le )
Cance /\bou t ac ou stic sir-1ilarny ..
(add to cu rr ent resu ts ta ole )

Figure 9. PCT screen capture illustrating the parameters users can set for DTW algorithm to be
taken into account during the acoustic analysis
 65

EFFECTS OF CROSSLINGUISTIC OVERLAP

l>•,m1Y1ty

AVG AVG MFCC --I-D_Y"_"_ 11c 11me warping 80 .0 -~00 57148 I 26 12

10000 .0 61494 26
abrigo- 1.wav ad{)!)t 1ve .w_"_-+-M_ Fc_c_ --f-D~Y"_"_ 11c ti me warp1ns1 80 .0 10000 .0 61417 26 12
- -f-- -
abrup:o-1 .wav abruotwav MFCC Dy<1am1c 11mewa rping 80 .0 10000 .0 43 ,562 26

_::aca ,e-1.w;,_•, --+-' '_"m_._.,.,av--t-'- " -"----+D_y_o,_ m1c ti me warp1ns1 80 0 10000 .0 63 48 1 26 12

acen:o-1 .wav MFCC Dy<1o1m
1c 11mewarping 80 .0 48 264 26
.icuce-1 wav .ilcohcl.wav t.lFCC Dyna mic ti me warping 80 .0 10000 .0 66 .389 26 12
adec:o-1 .wav adeptw.iv MFCC Dy<1o1
m 1c t ime warping 80 .0 45 142 26
ades ivo- 1:,~;,v .illgrmert .wav MFCC Dyna m ic ti me warping 80 0 10000 .0 60 519 26 12
adJL-
ntc-1.w av ad1wic1.w a_, --+-M_ Fc_c_ ~ 11me ~ ~ 10000 .0 42 859 26 12
- -f-- f--- - f---
ado11,10
- 1.wav MFCC Dyna m ic t ime warping 80 0 10000 .0 46 44 1 26 12
ilger::i. , wav --+M_F_c_c---+--Dy_<1o1 m 1c ti me wii rp1n9 ~ 48 .6 24 26 12
agun.i- 1.w.i v MFCC Dynam ic t m1e warping 80 0 61 454 26 12
aipo- 1.wa,, -+-M_F_cc
s,qnawre .·,,_,.,_. ___ -+o_y_o,_m1c 11me~ ~ 10000 .0 76 .781 26
alagamen10- 1.1~av MFCC Dynam ic t ime warping 80 o 10000 0
~.i trao-1 .wa·t bag,1aqt>_w_"_-+M_F_cc ___ -+D_y_o,_m1c ti me ~ ~ -~0 .0 66 .609 __ ~ 12
alcool- 1.w.iv iilcohol wav MFCC Dyna m ic ti me wa rping 80 0 10000 0 52.64 2 26
alecrim -1 wav bard .wav MFCC Dy'lam 1c li me warping 80 .0 10000 .0 59 693 26 12
-alho-1 ,',lidV --+--- --f---
M FCC
f--- --
Dyna m ic ti me warping 80 o
-f--
10000 0 67 .209 26
-
alvenan a-1 ,_·,a_,-+-'-"-"_•,a·t --+-M_F_CC ___ -+D_y_s,_m1c li me ~~ 10000 .0 73 .371 26 12
ameaca-1 wav sm ,::, nq .,~·av MFCC Dy'la m 1c 11me warping 80 .0 10000 .0
~qo- 1.,•:av batte'',' .wav__ +-M_
F_CC
___ -+D_y_o,_m1c ti me warping 80 .0 10000 .0 67 .481 26 12
ar.ic- l wav MFCC Dy'la m 1c t ime wa rping 80 0 10000 .0

~ l wa·t --1-'"_ , -_b_w.i·, __ +-'-" -"----+D_Y_"'_ m 1c ti me warping 80 .0 59 523 26 12

arg u - 1 wav brave wav MFCC Dy<1am1c time wa rping 80 .0
MFCC Dy'la m 1c ti me warp1ns1 80 .0 10000 .0 52 .6 11 26 12
ap l1c.ic ao- l .wav appl 1cat 1onw.iv MFCC Dy<1am1c 11mewarping 80 0
apco-1 w.iv capaciyw; H t.lFCC Dyna m ic t ime warping 80 0 10000 .0 61 988 26 12
arn ca- 1.·wav__ l-ca-rg_o_w.iv-~c ___ -+-o_y,_a,_11c 11me~ ~ 10000 .0
--1-'-"-'-5 ~ 12
ass,ste -1.wav <'>-Sst wav MFCC Dyna mic ti me w;i rp1ng 80 .0 10000 .0 53 071 26 12
..:'.::_denc1a
- 1.'.vav aL-d ence ~_1av_-+M_ F_cc
___ -+o_,_,._ m1c 11me ~ ~ -~00 55. 489 -~ 12
aves:ruz -1.wav MFCC Dyna m ic t m1e warping 80 0 10000 0 75 978 26
baba -1 •Nil'/ cheque wil_, --+-M_ Fc_c_ ~ m 1c 11me~ ~ 10000 .0 12
- -f-- --f--- -
baderr,a-1.',~av chocolate .wav MFCC Dynam ic t ime warping 80 0 10000 0 12
ba11co-l wav__ .,·_,·av--+-M_F_cc
1-'•_·,_~ ___ -+D_y_o,_m1c li me ~~ 49. 713 26 12
-~
band o- 1 wav MFCC Dynam ic t ime warping 80 o 10000 0
~e-1 .wav__ l--- 10000 .0 62 .667 26 12
~-~

barata- 1.v,av comedy wav 10000 0 60 564 26

Reopen tunct 1on d1alo,1

Figure 10. Acoustic similarity results table. In the Result column, higher values represent greater
acoustic distances (less acoustic similarity) between the waveforms.

To obtain the waveforms for each Portuguese and English word in the official

word lists, two male native speakers of each language of interest were recorded. To

accomplish this, one native speaker of American English (midwestern accent, native

of Ohio) and one native speaker of Canadian English (southeastern Ontario accent,

native of Cornwall) were recorded reading aloud each English word. Conversely, two

native speakers of Brazilian Portuguese, one native of São Paulo and one native of Rio

de Janeiro, who were born and raised in the metro areas of both cities, were recorded
 66

EFFECTS OF CROSSLINGUISTIC OVERLAP

reading aloud each Portuguese word. For the recording session, two spreadsheets

were created (one for each language) and then converted to individual slides in a PDF

file. Each slide had a white background with one word in a legibly large font in its

center. Recordings took place in the Language Acquisition Research Lab at the

University of Ottawa. Readers sat at a desk in front of a computer screen and were

accompanied by the researcher. They were instructed to read aloud each of the words

that appeared in the center of the computer screen as naturally as possible. The

researcher controlled the pace in which each word was displayed. Each reader was

recorded twice, with 15-minute break in between recordings to control for

pronunciation discrepancies. Audacity® (“Audacity®: Free Audio Editor and

Recorder,” 2019) was selected for the recordings. Audio was captured using a

condenser microphone connected to a preamplifier and saved into a waveform file

(.wav) at 44,100 Hz. Audacity® was also used to create a label track that contained the

orthographic representation of each word in the recording. Each labeled word was

then exported as a separate waveform file, with its orthographic representation for its

name, and saved according to reader’s name and recording session. Both official word

lists (A + B) were used to create tab-delimited text files that listed all the pairwise

comparisons in the list according to their specific file location on the computer,

which were then used in PCT to perform the required acoustic similarity

computations.
 67

EFFECTS OF CROSSLINGUISTIC OVERLAP

In order for the phonological similarity metric to be as accurate as possible, it

was necessary to account for readers’ dialectal differences in pronunciation as well as

any pronunciation discrepancies that might have occurred between the first and

second recording (before and after the break). To that end, each recording from each

reader of Portuguese was compared with each recording from each reader of English.

In the end, eight sets acoustic similarity comparisons (see Table 6 below) were

performed using PCT. Once all the DTW acoustic similarity values were obtained,

they were entered into an Excel spreadsheet so that the mean crosslinguistic

phonological similarity overlap could be calculated for the data analysis.

Table 6. Acoustic similarity computations performed in PCT

Reader’s Recording Acoustic similarity

Native Language and r1 – before break calculations performed in
Place of Origin r2 – after break PCT
P1 – Portuguese • P1r1-E1r1
r1 r2
Rio de Janeiro • P1r1-E1r2
P2 – Portuguese • P1r2-E1r1
r1 r2
São Paulo
• P1r2-E1r2
E1 – English
r1 r2 • P2r1-E2R1
Ohio
• P2r1-E2r2
E2 – English • P2r2-E2r1
r1 r2
Ontario
• P2r2-E2r2
 68

EFFECTS OF CROSSLINGUISTIC OVERLAP

3.6. Experimental Procedure

The experimental task consisted of a lexical decision task with masked priming,

which was solely conducted in Presentation® (“Presentation®: Precise, Powerful

Stimulus Delivery,” 2017), a stimulus delivery and experiment control software widely

used in psycholinguistic experiments. Data collection occurred over a period of ten

months in the Language Acquisition Research Laboratory at the University of

Ottawa. Participants sat at a large computer desk in front of a 23-inch LCD monitor.

Before participants were allowed to initiate the experimental task, they engaged in a

prescreening interview with the researcher, in which they were asked a few questions

regarding their personal background and level of education, as well as how they had

learned about the experiment and whether they had any particular questions or

concerns. Participants who were native speakers of Portuguese were interviewed in

Portuguese. Those who were speakers of English speakers were interviewed in

English. The purpose of the prescreening interview was to ensure participants met

the language requirements of the study and to give them a chance to get themselves

situated in the laboratory. Following the prescreening interview, participants were

provided with the consent form (see Appendix C) and were given a few minutes to

read and sign it. Participants were then assigned a pseudocode that consisted of a

sequence of random letters and numbers so that their data would remain anonymous

and were instructed to use their assigned pseudocode thereafter. Subsequently,

participants were asked to complete the online Language Background Questionnaire

 69

EFFECTS OF CROSSLINGUISTIC OVERLAP

(introduced in 3.1., Google Forms, see Appendix D) on the computer as well as the

Language Proficiency Test (cloze), which was administered on paper. Most

participants completed these steps in approximately 20-30 minutes.

Afterwards, the experimenter started Presentation®, and the software

randomly assigned participants to one of the two word lists as well as instructed them

to enter their assigned pseudocode in a text box that appeared on the computer

screen. Presentation® would then prompt participants to press the blue button on a

Cedrus RB-730 Response Pad (button box) to indicate they were ready commence the

experimental task. The buttons on the button box had four distinct colors (blue,

green, red and white), and the ones participants would be pressing during the task

(blue, green and red) had a label in a legible font indicating their specific function in

the experimental task. Once participants pressed the blue button, they were

presented with the task instructions and instructed to press the blue button again to

begin the (20) practice trials. The purpose of the practice trials was to ensure

participants had understood the task instructions and become familiar with the

procedure. Once participants completed the practice trials, the researcher checked

whether they needed further clarification about the procedure. When participants

were cleared to continue, they were instructed to press the blue button one more time

to commence the experimental trials.

Each experimental trial consisted of a sequence of five events. First, a forward

mask consisting of ten hash marks (##########) was presented in the center of the
 70

EFFECTS OF CROSSLINGUISTIC OVERLAP

computer screen for 500ms. Second, the forward mask was followed by the

Portuguese prime in lowercase letters for 60ms. Third, the English target was

presented in uppercase letters for 500ms. Fourth, once the English target was

displayed, participants had 3,000ms (3s) to decide by the press of a button (green =

yes, red = no) whether or not the target was a real word of English. Fifth, as soon as

participants pressed the button to indicate their response, the next prime-target pair

was presented automatically. If participants did not press any button during the

allotted 3,000ms, the next prime-target pair was presented.

The experimental task was comprised of a total of four blocks of 150 trials, of

which four practice pairs were always displayed at the beginning of each block. A

pause was added at the end of each block, and participants were encouraged to get

up, stretch, rehydrate or use the restroom during this break. On average, participants

completed the task within 45 minutes, including break time. Their reaction time and

accuracy were recorded for the data analysis.

The following chapter presents the results of the study in detail.

 71

EFFECTS OF CROSSLINGUISTIC OVERLAP

4. The Study: Results

In order to facilitate the presentation of the results of the study, this chapter was

divided into two sections, Preliminary Analyses and Mixed-Effects Models. In the

Preliminary Analyses section, both dependent variables, reaction time and accuracy,

and each of the independent variables that comprised the master dataset were

independently analyzed, and whenever plotting the variable enhanced the results

being discussed, plots were provided. The results provided in this section serve as

groundwork for the ensuing section, since only more conventional statistical analysis

tools, such as correlations, t-tests and ANOVAs, were used. In the Mixed-Effects

Models section, one linear mixed-effects analysis of the relationship between reaction

time and the independent variables (lmer model), as well as one logistic regression

analysis between accuracy and the independent variables (glmer model) are

presented. Both mixed-effects analyses allowed for precise quantification of the

individual effects of crosslinguistic similarity, which was one of the primary goals of

this study. Both SPSS (“IBM SPSS Statistics 25,” 2018) and RStudio (RStudio Team,

2018), were used extensively to carry out the analyses in this section. In addition, both

SPSS and ggplot2 (Wickham, 2016), a data visualization package developed for the

statistical programing language R, were used to facilitate the presentation of the

results.

Before proceeding with the data analysis, it is important to underscore the fact

that prime-target pairs that were originally coded as Practice did not enter the
 72

EFFECTS OF CROSSLINGUISTIC OVERLAP

analyses. As discussed in chapter 3, 20 practice pairs were displayed to participants at

the beginning of the experiment in order to give participants an opportunity to

practice the experimental task and ask the researcher any questions they might have

had about the procedure. Four practice pairs were also displayed at the beginning of

a new block after each of the planned pauses/breaks.

4.1. Preliminary Analyses7

4.1.1. Dependent Variables

4.1.1.1. RT (Reaction Time/Response Latency)

As mentioned in Chapter 3, Presentation (“Presentation®: Precise, Powerful

Stimulus Delivery,” 2017), a stimulus delivery and experiment control software, was

used to control the presentation of the stimuli and record participants’ reaction time

during the study. To that end, Presentation records reaction time in tenths of

milliseconds. However, for the sake of consistency with psycholinguistic studies,

which normally report reaction time in milliseconds, the variable RT was converted

from tens of milliseconds to milliseconds. The descriptive statistics for RT, including

a breakdown by language group (hereinafter referred to as the independent variable

7
A brief description of the variable in the master dataset is provided hereinafter between

parentheses.
 73

EFFECTS OF CROSSLINGUISTIC OVERLAP

LANGUAGE) are provided in Table 7 below. Kernel density plots (Figure 11) are also

provided to facilitate visualization of the distribution of RT.

Table 7. RT (Descriptive Statistics)

N Mean Std. Dev. Minimum Maximum

Overall 33,516 769.01 316.04 291.20 3,000
Bilinguals 16,464 808.60 348.29 291.20 3,000
Functional
17,052 730.79 276.08 305.60 3,000
Monolinguals

0.0020

0.002

0.0015
Density
Density

0.001
0.0010

0.0005

0.000

0 250 500 750 1000 1250 1500 1750 2000 2250 2500 2750 3000
0.0000 Mean Reaction Time

Language Groups
□ Bilinguals □ Functional Monolinguals
0 250 500 750 1000 1250 1500 1750 2000 2250 2500 2750 3000
Mean Reaction Time

Figure 11. Kernel density plots illustrating the distribution of RT (overall and by LANGUAGE)

To further examine the differences in mean reaction time between bilinguals

and functional monolinguals, an independent samples t-test was performed on RT

 74

EFFECTS OF CROSSLINGUISTIC OVERLAP

with LANGUAGE as the grouping factor. The t-test showed that functional

monolinguals were on average significantly faster (M = 730.79, SD = 276.08) than

bilinguals (M = 808.60, SD = 348.29), t(33,514) = – 22.705, p < .001. Whether or not

this significant difference between bilinguals and functional monolinguals played a

major role in mean reaction time when all other variables are jointly taken into

account will be explored in the Mixed-Effects Models section.

4.1.1.2. RAW_ACCURACY (Response Accuracy)

Presentation was also used to record participant’s accuracy. As described in

Chapter 3, participants had 3,000ms in each trial (prime-target pair) to decide by the

press of a button whether or not the target word presented was a genuine English

word. Accuracy was thus recorded as a hit (correct answer), incorrect, or miss (when

the allocated 3,000ms was exceeded) for each trial, making it a categorical variable.

As Table 8 shows, irrespective of language group, the number of correct answers in

the entire experiment, was higher than the number of incorrect answers. However,

when the data is broken down by language group, bilinguals had a higher number of

correct answers than functional monolinguals.

 75

EFFECTS OF CROSSLINGUISTIC OVERLAP

Table 8. RAW_ACCURACY (Count)

Hit (correct) Incorrect Miss

Overall 30,956 2,472 88

Bilinguals 16,297 713 42

Functional
14,659 1,759 46
Monolinguals

A point-biserial correlation between RT and RAW_ACCURACY was run to

better understand the relationship between the two dependent variables. Since this

type of correlation is run between a continuous and a dichotomous variable, only

correct and incorrect answers were considered. The correlation yielded a weak yet

significant negative correlation between RT and RAW_ACCURACY, r(33,428) = –

.186, p < .001, suggesting that on average as participants’ accuracy increased, their

reaction time decreased.

In order to determine the percentage of hits (correct answers) per participant

and to weed out participants with an unusually low percentage of correct answers, a

new variable (CORRECT_PERCENTAGE) was created, which in essence transformed

RAW_ACCURACY from a categorical to a continuous variable. To accomplish this,

the total number of hits (correct answers) per participant was divided by the total

number of trials in the experiment. This result was then multiplied by 100%. Because

the percentage of correct answers for one of the bilingual participants was
 76

EFFECTS OF CROSSLINGUISTIC OVERLAP

significantly lower than bilingual participants’ mean, that specific participant had to

be removed from the analysis.

Table 9. CORRECT_PERCENTAGE (Descriptive Statistics)

Language Group N Mean Std. Dev. Min. Max.

Combined 33,428 92.17 6.59 64.10 99.20

Bilinguals 16,418 88.72 7.43 64.10 98.60

Functional Monolinguals 17,010 95.50 3.08 84.80 99.20

An independent samples t-test was performed on CORRECT_PERCENTAGE

with LANGUAGE as a grouping factor. The t-test showed that functional

monolinguals’ percentage of correct answers (M = 95.50, SD = 3.08) was significantly

higher than bilinguals’ (M = 88.72, SD = 7.43), t(33,426) = 109.54, p < .001. Although

this result is only preliminary, it shows that functional monolinguals were on average

more accurate at deciding whether the presented word was a genuine word of English.

4.1.2. Independent Variables

4.1.2.1. SEMAC and ERS (English Proficiency)

As discussed in Chapter 3, English proficiency was assessed via a cloze

proficiency test. Participants’ answers on the test were scored using both the
 77

EFFECTS OF CROSSLINGUISTIC OVERLAP

semantically acceptable scoring procedure (SEMAC) and the exact replacement

scoring method (ERS), producing two language proficiency variables, SEMAC and

ERS, which, as indicated by previous research, are positively correlated, r(33,516) =

.839, p < .001. An independent samples t-test was performed on both ERS and SEMAC

with LANGUAGE as the grouping factor. With respect to SEMAC, the t-test revealed

that functional monolinguals were significantly more proficient in English than

bilinguals, t(33,514) = 119.97, p < .001. Similarly, with respect to ERS, functional

monolinguals were also shown to be more proficient in English than bilinguals,

t(33,514) = 116.88, p < .001. These results show that irrespective of scoring method

functional monolingual participants were on average significantly more proficient in

English than bilinguals, which was expected since this group of participants was

comprised of native speakers of the language. Please refer to Table 10 below for the

mean differences between both variables.

Table 10. ERS and SEMAC Independent samples t-test (Descriptive Statistics)

Std. Mean
Language Group N Mean
Dev Diff.
Bilinguals 16,464 50.07 11.75
ERS 13.51*
Functional Monolinguals 17,052 63.59 9.31
Bilinguals 16,464 76.86 16.28
SEMAC 15.69*
Functional Monolinguals 17,052 92.55 5.09
*p < .001
 78

EFFECTS OF CROSSLINGUISTIC OVERLAP

In order to determine the relationship between both ERS and SEMAC with RT,

irrespective of language group, a Pearson product-moment correlation was run. A

weak yet significant negative correlation was obtained for both ERS, r(33,516) = – .10,

p < .001, and SEMAC, r(33,516) = – .13, p < .001. These results indicate that on average

as mean English proficiency increased, mean reaction time decreased irrespective of

language group.

The relationship between English proficiency and RT presents itself

differently when the data is broken down by LANGUAGE. For bilinguals, a weak

negative correlation is maintained with both ERS and SEMAC (both ps <.001). For

functional monolinguals, a weak positive correlation was found with ERS (p <. 001).

However, no significant correlation was found with SEMAC. See Table 11 below for

the results of the tests.

Table 11. Pearson product-moment correlations between RT and ERS and SEMAC

Language Group Scoring Method r

ERS r(16,464) = – .10*

Bilinguals
SEMAC r(16,464) = – .11*

ERS r(17,052) = .04*

Functional Monolinguals
SEMAC r(17,052) = 0

* p < .001
 79

EFFECTS OF CROSSLINGUISTIC OVERLAP

A point-biserial correlation was run in order to evaluate the relationship

between both ERS and SEMAC with RAW_ACCURACY. A significant positive

correlation was found with both variables (both ps < .001). When the English

proficiency data was split by LANGUAGE, significant positive correlations were also

found, indicating that as mean English proficiency increased, so did mean accuracy.

Table 12 below shows the results of the point-biserial correlations.

Table 12. Point-biserial correlations between ERS and SEMAC with RAW_ACCURACY

Language Group Scoring Method r

ERS r(16,418) = .11*

Bilinguals
SEMAC r(16,418) = .12*

ERS r(17,010) = .05*

Functional Monolinguals
SEMAC r(17,010) = .05*

Since the correlations between SEMAC scores and RT and RAW_ACCRURACY

were stronger, SEMAC scores were selected to enter the mixed-model analyses as an

indicator of participants’ English proficiency level (PROFICIENCY).

 80

EFFECTS OF CROSSLINGUISTIC OVERLAP

4.1.2.2. SEMANTIC (Crosslinguistic Semantic Similarity/Overlap)

Crosslinguistic semantic similarity was coded in the master dataset as a

categorical variable consisting of four categories according to the degree of

crosslinguistic semantic overlap between a prime and its target: cognates (COG), false

friends (FF), controls (CTRL), and distractors (DISTR). Consistent with the results

discussed in Alves-Soares, (2013) and other previous linguistic experiments that have

utilized masked priming in their methodology, participants’ reaction time was highly

influenced by the degree of semantic relationship between prime and target. Mean

reaction time was lowest for targets that were matched to a prime with which they

share a high degree of crosslinguistic semantic overlap (cognate items), and highest

for pseudoword targets (DISTR items), i.e., when a random word of Portuguese was

matched with a pseudoword target. This pattern is also present when the data is split

by language groups, suggesting that a priming effect of crosslinguistic semantic

similarity was present for both language groups. Splitting the data by LANGUAGE

shows that functional monolingual participants were on average faster to decide

whether or not the target was a word of English than bilinguals. Refer to Table 13 for

the descriptive statistics.

A two-way ANOVA was then performed with RT as the dependent variable and

LANGUAGE and SEMANTIC and as factors to further investigate these initial

observations. The ANOVA revealed a significant main effect of semantic similarity,

F(3, 33,508) = 875.88, p < .001, a significant main effect of language group, F(1, 33,508)
 81

EFFECTS OF CROSSLINGUISTIC OVERLAP

= 333.88, p < .001, as well as a significant interaction between semantic similarity and

language group, F(3, 33,508) = 10.06, p < .001. Pairwise comparisons using the

Bonferroni adjustment indicated that functional monolinguals were on average 73ms

faster than bilinguals (p < .001). Regarding reaction time differences due to semantic

similarity, in comparison to controls, participants from both language groups were

on average 45ms faster to respond to cognates and 194ms slower to respond to

distractors (both ps < .001). Participants were also on average 29ms faster to respond

to cognates than false friends (p < .001). No significant mean differences in reaction

time were observed between control and false friends (p = .079).

 82

EFFECTS OF CROSSLINGUISTIC OVERLAP

Table 13. RT by SEMANTIC: Descriptive Statistics

Language Semantic
Group Similarity N Mean* Std. Dev.* Min.* Max.*
COG 5,489 657.99 261.42 291.20 3,000
FF 2,890 687.34 285.97 326.00 3,000
Combined
CTRL 8,379 703.50 280.18 339.70 3,000
DIST 16,758 852.22 331.32 313.4 3,000
COG 2,700 680.35 290.99 291.20 3,000
FF 1,416 733.33 332.78 341.00 3,000
Bilinguals
CTRL 4,116 737.77 308.32 345.80 3,000
DIST 8,232 899.03 362.28 332.90 3,000
COG 2,789 636.35 227.17 305.60 3,000
Functional FF 1,474 643.17 223.68 326.00 3,000
Monolinguals CTRL 4,263 670.42 245.58 339.70 3,000
DIST 8,526 807.02 291.37 313.40 3,000
* in milliseconds

As SEMANTIC, LANGUAGE, and ACCURACY_RAW are all categorical

variables, in order to examine the effect of semantic similarity on accuracy by

language group, a breakdown of total number of hits and incorrect answers as by

language groups and semantic similarity was obtained (see Table 14). The percentage

of correct and incorrect answers was also entered into the breakdown so that any

trends could be more easily spotted.

 83

EFFECTS OF CROSSLINGUISTIC OVERLAP

Table 14. RAW_ACCURACY by SEMANTIC and LANGUAGE

Raw
Language Group Percentage (%)
Count
incorrect 79 2.9
COG hit 2,615 97.1
Total 2,694 100.0
incorrect 107 7.6
FF hit 1,305 92.4
Total 1,412 100.0
Bilinguals
incorrect 256 6.2
CTRL hit 3,849 93.8
Total 4,105 100.0
incorrect 1,317 16.0
DISTR hit 6,890 84.0
Total 8,207 100.0
incorrect 45 1.6
COG hit 2,739 98.4
Total 2,784 100.0
incorrect 32 2.2
FF hit 1,441 97.8
Functional Total 1,473 100.0
Monolinguals incorrect 111 2.6
CTRL hit 4,138 97.4
Total 4,249 100.0
incorrect 525 6.2
DISTR hit 7,979 93.8
Total 8,504 100.0

A particular trend clearly indicated by Table 14 is the fact that functional

monolinguals’ percentage of correct answers is higher than bilinguals’ across all

 84

EFFECTS OF CROSSLINGUISTIC OVERLAP

semantic categories. In addition, distractor pairs yielded the lowest percentage of

correct answers for both language groups while cognate pairs yielded the highest.

Together these two trends indicate a possible main effect of LANGUAGE and

SEMANTIC on RAW_ACCURACY.

4.1.2.3. NLD (Crosslinguistic Orthographic Similarity/Overlap)

As discussed in Chapter 3, to measure crosslinguistic orthographic similarity

(overlap) between primes and targets that were selected for the experiment, the

Normalized Levenshtein Distance (NLD) (Schepens et al., 2013) was selected. The

NLD is the most up-to-date and reliable metric for the purposes of this study. It yields

similarity values that range between 0 [zero] (when prime-target pairs are

orthographically dissimilar) and 1 (when prime-target pairs are orthographically

identical), irrespective of word length (Guasch et al., 2013). Table 15 below provides

the descriptive statistics of the variable NLD by SEMANTIC. The degree of

crosslinguistic orthographic overlap between prime-target pairs classified as COG

and FF is significantly higher than in pairs classified as CTRL and DISTR. The mean

differences can be more clearly seen in both Table 15 and Figure 12 below.
 85

EFFECTS OF CROSSLINGUISTIC OVERLAP

Table 15. NLD: Descriptive Statistics

Std.
Semantic Similarity N Mean Min. Max.
Dev

COG 5,489 .71 .12 0 1

FF 2,890 .70 .20 0 1

CTRL 8,379 .12 .12 0 1

DISTR 16,758 .11 .12 0 1

1.00
Normalized Levenshtein Similarity

.80
-L-
*
- *
*
.60 -

0
.40

0
.20 '
0
0
8 '---

.00

COG FF CTRL DISTR

Semantic Similarity

Figure 12. Mean differences in crosslinguistic orthographic overlap across semantic categories
showing a higher mean NLD for COG and FF pairs.
 86

EFFECTS OF CROSSLINGUISTIC OVERLAP

4.1.2.4. PHONETIC (Crosslinguistic Phonological Similarity/Overlap)

As discussed in Chapter 3, the present study computed an objective

crosslinguistic phonological similarity metric using a technique previously employed

by Mielke (2012), viz., dynamic time warping (DTW). To that end, recordings of two

native speakers of Portuguese and of two native speakers of English reading aloud the

selected experimental words in their native language were made and used to compute

a mean crosslinguistic phonological similarity value between prime-target pairs.

Lower crosslinguistic phonological similarity values represent a greater degree of

crosslinguistic phonological overlap between a prime and its target.

One hurdle that became apparent during the preliminary stages of the data

analysis pertained to distractor pairs, because no crosslinguistic phonological

similarity value for them could have been computed via the DTW algorithm. As

prefaced in Chapter 3, distractor pairs in the experiment consisted of a genuine word

of Portuguese as a prime that was matched with a pseudoword as a target. Even

though the pseudowords selected for the experiment were generated by a

pseudoword generator application that had been programed to generate

pseudowords that resembled words of English as closely as possible, some of the

consonant clusters as well as consonant-vowel combinations were not easily

pronounceable by a native speaker of English due to the phonotactics of the language.

As a result, recording a native speaker of English reading the list of pseudowords

aloud as naturally as possible was a major challenge. While programing a speech

 87

EFFECTS OF CROSSLINGUISTIC OVERLAP

synthesizer and requesting Siri to speak the pseudowords to the make the recordings

were two options explored, the end result sounded quite artificial and robotic,

lacking the connected and coarticulatory features present in human speech.

Moreover, leaving the cells pertaining to distractor pairs blank in the crosslinguistic

phonological similarity data would certainly lead to convergence complications in the

later stages of the data analysis, particularly for fitting linear mixed-effects models.

Thus, it was paramount that the distractor pairs cells be populated with data.

The most elegant solution reached was to populate the cells with data that

followed the same distribution of the DTW crosslinguistic phonological similarity

data. To achieve this, first the crosslinguistic phonological similarity data was plotted

to ensure it followed a Laplace-Gaussian (normal) distribution and its descriptive

statistics were obtained. With this information available, using the rnorm function

in R, random numbers that followed a normal distribution with the exact same means

and standard deviation of the original crosslinguistic phonological similarity data

were generated. Afterward, both the original data and the randomly generated data

were consolidated in a new variable, PHONETIC. Finally, a Pearson product-moment

correlation was carried out and descriptive statistics were obtained to ensure no

disparities had arisen from the procedure. Table 16 and Figure 13 below show the

descriptive statistics and the kernel density plots of the crosslinguistic phonological

similarity data before and after the procedure. To more easily illustrate the mean
 88

EFFECTS OF CROSSLINGUISTIC OVERLAP

differences across the four semantic categories, descriptive statistics and box plots of

the data were also obtained for the variable PHONETIC (see Table 17 and Figure 14).

Table 16. Original DTW crosslinguistic phonological similarity data vs. consolidated data

Original Data Consolidated Data

N Mean SD Min. Max. N Mean SD Min. Max.

16,758 57.07 5.56 45.50 85.88 33,516 57.07 5.56 35.42 85.87

r(33,516) = 1, p < .001

0.08

0.06
0.06
Density
Density

0.04
0.04

0.02
0.02

0.00 0.00

25 35 45 55 65 75 85 25 35 45 55 65 75 85
Mean Crosslinguistic Phonetic Overlap Mean Crosslinguistic Phonetic Overlap
Original Data Consolidated Data

Figure 13. Kernel density plots of the DTW crosslinguistic phonological similarity data before
(left) and after (right) consolidation.
 89

EFFECTS OF CROSSLINGUISTIC OVERLAP

Table 17. PHONETIC: Descriptive Statistics

Std.
Semantic Similarity N Mean Min. Max.
Dev

COG 5,489 54.32 4.35 45.60 71.82

FF 2,890 53.74 3.44 45.51 67.98

CTRL 8,379 60.03 5.26 48.62 85.87

DISTR 16,758 57.06 5.55 35.42 80.74

90.00

80.00
DTW Phonetic Similarity

70.00

60.00

50.00

40.00

30.00

COG FF CTRL DISTR

Semantic Similarity

Figure 14. Mean differences in DTW crosslinguistic phonological overlap across semantic
categories showing a lower mean for COG and FF pairs. Lower values represent a greater the
degree of crosslinguistic phonological overlap between a prime and its target.

Page 1
 90

EFFECTS OF CROSSLINGUISTIC OVERLAP

With the transformations on the DTW crosslinguistic phonological similarity data

complete, the relationship between the PHONETIC with both the two dependent

variables could be evaluated. A Pearson product-moment correlation between

PHONETIC and RT yielded a weak yet significant positive correlation between the

two variables, r(33,516) = .02, p = .002, suggesting that on average participants reacted

more slowly to more phonologically dissimilar pairs. A point biserial correlation

between PHONETIC and RAW_ACCURACY did not yield a statistically significant

result, r(33,428) = –.006, p = .31.

4.1.2.5. English_Freq and Portuguese_Freq (Word Frequency)

As discussed in chapter 3, a word frequency count, viz., how often a particular

word occurs per million words in collected corpora was obtained for each Portuguese

and English word selected for the present study. To that end, O corpus do Português

(Davies, 2017b), a large online corpus of Portuguese words, was used to obtain word

frequency count for the Portuguese words, while Corpus of Contemporary American

English (COCA) (Davies, 2017a), was used to obtain word frequency count for the

English words.

Table 18 below shows the descriptive statistics for both English_Freq and

Portuguese_Freq by SEMANTIC. It is important to underscore the fact that COCA

yielded a word frequency count of zero for all of the pseudoword targets in distractor

pairs selected for the experiment, since these targets are not words that exist in the
 91

EFFECTS OF CROSSLINGUISTIC OVERLAP

English language. Because of this, distractor pairs have zero mean English frequency

in Table 18 below.

Table 18. English_Freq and Portuguese_Freq by SEMANTIC (Descriptive Statistics)

SEMANTIC N Mean Std. Dev. Min. Max.

English_Freq 5,489 31.77 76.76 0.14 735.39

COG
Portuguese_Freq 5,489 36.58 84.74 0.03 688.23
English_Freq 2,890 36.15 94.06 0.08 774.33
FF
Portuguese_Freq 2,890 36.28 83.68 0.00 562.51
English_Freq 8,379 33.77 84.83 0.08 774.33
CTRL
Portuguese_Freq 8,379 18.50 50.58 0.00 443.81
English_Freq 16,758 0.00 0.00 0.00 0.00
DISTR
Portuguese_Freq 16,758 19.48 66.77 0.00 688.23

A Pearson product-moment correlation was run to examine the relationship

between English_Freq and Portuguese_Freq as well as between both word frequency

variables and RT. The results of the correlation between English_Freq and

Portuguese_Freq indicated positive relationship between the two variables, r(33,516)

= .15, p < .001. The correlation with RT indicated a weak yet significant negative

relationship with both English_Freq, r(33,516) = – .11, p < .001, and Portuguese_Freq,

r(33,516) = – .04, p < .001. A point biserial correlation between both word frequency

variables and RAW_ACCURACY indicated a weak yet significant positive correlation

 92

EFFECTS OF CROSSLINGUISTIC OVERLAP

for both English_Freq, r(33,428) = .06, p < .001, and Portuguese_Freq, r(33,428) = .02,

p < .001.

4.1.2.6. Time

Time as an independent variable indicates the length of time (measured in

tenths of milliseconds) that elapsed from the presentation of the very first prime-

target pair until the presentation of each of the subsequent pair until the very end of

the experiment. Similar to RT, Time was recorded in tenths of milliseconds, but it was

converted to milliseconds for the data analysis.

A Pearson product-moment correlation was run in order to examine the

relationship between Time and RT. A weak yet significant negative correlation was

found between the two variables, specifically, r(33,516) = – .08, p < .001. The

correlation was also run with LANGUAGE as a split factor in order to verify whether

the relationship manifested differently for both language groups. The weak negative

correlation between Time and RT was maintained for both language groups,

specifically, r(16,464) = – .06, p < .001 for bilinguals, and r(17,052) = – .14, p < .001 for

functional monolinguals. While these results suggest that, irrespective of language

group membership, participants were getting faster as the experiment progressed,

the effect of Time on RT was stronger for functional monolingual participants.

With respect to accuracy, a point biserial correlation was run between Time

and RAW_ACCURACY. The result of this correlation was not significant, r(33,428) =

.005, p = 401. However, running it with LANGUAGE as a split factor yielded a

 93

EFFECTS OF CROSSLINGUISTIC OVERLAP

significant result only for functional monolinguals, specifically, r(17,010) = .03, p <

.001, suggesting, despite being weak, that functional monolinguals were improving

(getting more accurate) as the experiment progressed.8

4.1.2.7. PRIME and TARGET

PRIME, as an independent variable, comprises each of the Portuguese primes

in the experiment. Likewise, TARGET, comprises each of the English targets and

pseudowords included in the experiment. Refer to Appendix F for the complete list

of primes, targets, and pseudowords. Treating both PRIME and TARGET as

independent variables allowed them to be entered into both mixed-effects models as

random intercepts.

4.1.2.8. PAIR

Each of the prime-target pairs in the experiment were assigned a unique

number code, which was then encoded by the variable PAIR. This allowed PAIR to be

entered into the mixed-effects analyses as a random intercept. Refer to Appendix F

for the unique number code assigned to each of the prime-target pairs.

8
The point biserial correlation between Time and RAW_ACCURACY was not significant for
bilinguals, r(16,418) = .010, p = .179.
 94

EFFECTS OF CROSSLINGUISTIC OVERLAP

4.1.2.9. PARTICIPANT

In order to account for the great deal of individual variation resulting from

individual idiosyncrasies that cannot be easily (if not impossibly) controlled in an

experiment like this, a subject variable must be included in a mixed-effects model

analysis as a random intercept. To that end, as each participant in the experiment was

assigned a pseudocode, the independent variable PARTICIPANT comprises the

pseudocode assigned to each participant.

4.1.2.10. OTHER_LANGUAGES

All participants in the experiment had to indicate in the Language

Background Questionnaire whether they had been exposed to (i.e., had received

formal instruction) of another language. The variable OTHER_LANGUAGES

encodes the participants’ responses to this particular question as either YES or NO.

As Figure 15 illustrates, the majority of the bilingual participants in the experiment

(89.3%) answered NO to the question. This is in stark contrast with functional

monolingual participants, whose majority (62.1%) answered YES to this question.

This reflects the fact that most functional monolingual participants in the

experiment were from the National Capital Region in Canada, an area where English

speakers are exposed to French on a daily basis and often study the language in school

as their second official language. Because of the great degree of orthographic,

semantic and phonological overlap between French and Portuguese, both Romance
 95

EFFECTS OF CROSSLINGUISTIC OVERLAP

Languages, it was important to control for the participants’ knowledge of another

language to further understand its role in response latency. Thus, the variable

OTHER_LANGUAGES was entered in both mixed-effects models as a control

variable, which will be discussed in the next section.

100.0%
Language Groups
Functional Monolinguals
Bilinguals

80.0%

60.0%
Percent

89.3%
40.0%

62.1%

20.0% 37.9%

10.7%

.0%
NO YES
OTHER_LANGUAGES

Figure 15: Differences between functional monolinguals and bilinguals with regards to
whether or not they have the knowledge of another European language besides English and
Portuguese.

4.1.3. Interim Summary of the Preliminary Analyses

Although the Preliminary Analysis section, as its name suggests, serves

primarily as groundwork for the mixed-effects models that are

Page 1 presented in the

ensuing section, an interim summary of the most import findings is provided below.
 96

EFFECTS OF CROSSLINGUISTIC OVERLAP

• With respect to overall RT, functional monolinguals were on average

significantly faster (M = 730.79, SD = 276.08) than bilinguals (M = 808.60, SD

= 348.29), t(33,514) = – 22.705, p < .001.

• With respect to overall accuracy, functional monolinguals’ percentage of

correct answers (M = 95.50, SD = 3.08) was significantly higher than

bilinguals’ (M = 88.72, SD = 7.43), t(33,426) = 109.54, p < .001.

• A weak yet significant negative correlation between RT and

RAW_ACCURACY, r(33,428) = – .186, p < .001, showed that on average as

participants’ accuracy increased, their reaction time decreased.

• With respect to English proficiency, based on SEMAC scoring, functional

monolinguals were significantly more proficient in English than bilinguals,

t(33,514) = 119.97, p < .001. Similarly, based on ERS scoring, functional

monolinguals were more proficient in English than bilinguals, t(33,514) =

116.88, p < .001.

• An ANOVA with RT as the dependent variable and LANGUAGE and

SEMANTIC as factors yielded a significant main effect of semantic similarity,

F(3, 33,508) = 875.88, p < .001, a significant main effect of language group, F(1,

33,508) = 333.88, p < .001, as well as a significant interaction between semantic

similarity and language group, F(3, 33,508) = 10.06, p < .001. Functional

monolinguals were on average 73ms faster than bilinguals (p < .001).

Participants from both language groups were on average 45ms faster to

 97

EFFECTS OF CROSSLINGUISTIC OVERLAP

respond to cognates and 194ms slower to respond to distractors in relation to

controls (both ps < .001). Participants were also on average 29ms faster to

respond to cognates than false friends (p < .001).

• Functional monolinguals’ percentage of correct answers was higher than

bilinguals’ across all semantic categories. Distractor pairs yielded the lowest

percentage of correct answers for both language groups while cognate pairs

yielded the highest. These two findings indicate a possible main effect of

LANGUAGE and SEMANTIC on RAW_ACCURACY.

• The degree of crosslinguistic orthographic overlap (NLD) between prime-

target pairs classified as cognates and false friends is significantly higher than

in pairs classified as controls and distractors, suggesting a possible main

effect of NLD.

• A weak yet significant positive correlation between PHONETIC and RT was

found, r(33,516) = .02, p = .002; on average participants reacted more slowly

to more phonologically dissimilar pairs crosslinguistically. The correlation

between PHONETIC and RAW_ACCURACY was not statistically significant

(p = .31).

• The correlation between English_Freq and Portuguese_Freq was significant,

r(33,516) = .15, p < .001. The correlation between RT and English_Freq was

significant, r(33,516) = – .11, p < .001, and between RT and Portuguese_Freq,

r(33,516) = – .04, p < .001. The correlation between both types of word
 98

EFFECTS OF CROSSLINGUISTIC OVERLAP

frequency and RAW_ACCURACY was significant for both English_Freq,

r(33,428) = .06, p < .001, and Portuguese_Freq, r(33,428) = .02, p < .001.

• A significant correlation was found between RT and Time, r(33,516) = – .08, p

< .001. The correlation was maintained for both language groups, specifically,

r(16,464) = – .06, p < .001 for bilinguals, and r(17,052) = – .14, p < .001 for

functional monolinguals. Despite the low rs, these correlations show that

irrespective of their language group, participants’ RT improved as the

experiment progressed. With respect to RAW_ACCURACY, the correlation

between this dependent variable with Time was not significant (p = 401).

Using LANGUAGE as a split factor, the correlation between

RAW_ACCURACY and Time was only significant for functional

monolinguals, r(17,010) = .03, p < .001. Again, despite the weak r, this

correlation shows that functional monolinguals were more accurate the

experiment progressed.

• 89.3% of the bilingual participants indicated spoke no other European

language besides English and Portuguese. In contrast, 62.1% of the functional

monolingual participants indicated they spoke another European language

besides English.
 99

EFFECTS OF CROSSLINGUISTIC OVERLAP

4.2. Mixed-Effects Models

The two dependent variables, viz., RT (response latency) and RAW_ACCURACY

(response accuracy), were modeled separately as a function of all the independent

variables introduced in the Preliminary Analyses section above. RT, as a continuous

variable, was modeled using a linear mixed-effects model fitted with lmer.

RAW_ACCURACY, as a dichotomous variable, was modeled using a logistic

regression mixed-effects model with glmer. Both lmer and glmer are part of the

lme4 package (Bates, Mächler, Bolker, & Walker, 2015) in R (R Core Team, 2019). Both

mixed-effects models were fit in RStudio running remotely on a virtual Linux virtual

instance using Amazon Elastic Compute Cloud (Amazon EC2)9. The results of each

model are presented independently.

9
Amazon Elastic Compute Cloud (Amazon EC2) is a subscription-based web service that is part

of Amazon Web Services (AWS), a collection of remote computing services that together

make up a cloud computing platform offered over the Internet by Amazon (“Amazon Elastic

Compute Cloud (Amazon EC2),” 2018). Amazon EC2 provides secure, resizable compute

capacity in the cloud, and offers a large number of instances, i.e., virtual computer

configurations (Linux or Windows) that can be further customized to suit subscribers’

computational and data processing needs. For the purposes of this study, Amazon EC2 was

extremely useful during data analysis due to its computation power and processing speed,

which significantly expedited model fitting in R Studio.

 100

EFFECTS OF CROSSLINGUISTIC OVERLAP

The random effects of PARTICIPANT, PAIR, PRIME, and TARGET were

specified in both models. Their random intercepts were always included. Random

slopes were included whenever possible; they were excluded if they correlated too

highly (r ≥ .90) with their random intercept or other random slopes. SEMANTIC,

NLD, PHONETIC, English_Freq, Portuguese_Freq, and Time were entered as

potential PARTICIPANT slopes.

A maximal mixed-effects structure was always sought for both models (Barr,

Lev, Scheepers, & Tily, 2013). Iterations of the maximal model were successively

simplified until the best fit was achieved. During the model simplification process,

the most non-significant effect (highest p-value), starting with any of the specified

interactions, was always removed first. This new iteration of the model was then

refitted. If this iteration of the model successfully converged, subsequently, the anova

function in R was called to compare the simplified (new iteration) model against its

originator (previous iteration) model. The iteration that yielded the lowest Akaike

Information Criterion (AIC) value (Akaike, 1998) continued to be further simplified

in consecutive steps. The best model fit was deemed to have been reached when any

further simplification to the model structure prevented it from successfully

converging, produced a singularity fit, yielded a higher AIC value or was statistically

the same its previous iteration (Chi-Square results provided in the anova).

In order to significantly minimize the occurrence of convergence errors when

fitting both models, a few data cleaning techniques had to be applied. First, all
 101

EFFECTS OF CROSSLINGUISTIC OVERLAP

continuous independent variables, viz., NLD, PHONETIC, English_Freq,

Portuguese_Freq, and Time, were grand mean centered in order to reduce

multicollinearity (Linck & Cunnings, 2015) and then standardized. Second,

nloptwrap2, an optimizer function suggested by Bolker (2014) that calls for the

optimx package (Nash, 2014) in R, was used to fit every iteration of both models.

Finally, an inverse transformation (-1000/RT) was applied to the reaction time data

in order to reduce its positive skew. Reaction time data, more often than not,

particularly in language priming studies, usually turns out to be positively skewed.

However, one of the major assumptions of several tests of statistical significance is

that the dependent variable follows a normal (Gaussian) distribution. A positively

skewed continuous dependent variable, as it is often the case with reaction time,

violates that assumption. One of the suggested ways to mitigate this is to use invRT,

which is applying an inverse transformation to the reaction time data (-1000/RT).

Refer to Brysbaert & Stevens (2018) for further details on this technique.

Four random effects were always specified in both models: PARTICIPANT,

PRIME, TARGET, and PAIR. Their random intercepts were always included. Random

slopes with PARTICIPANT were included whenever possible. They were excluded if

they correlated too highly (r ³ .90) with their random intercept or other random

slopes.

The following fixed effects, as well as their interactions with one another, were

specified in the lmer and glmer maximal models: LANGUAGE, SEMANTIC, NLD,
 102

EFFECTS OF CROSSLINGUISTIC OVERLAP

PHONETIC. Several control variables were also included, specifically, LIST,

English_Freq, Portuguese_Freq, target_length, prime_length, AGE, SEMAC (English

proficiency), Time, and OTHER_LANGUAGES (whether or not participants had

been exposed to or studied another European language besides English or

Portuguese). As presented below, a number of fixed effects and control variables did

not end up making to the best fit of both models, because they simply did not

significantly improve model fit.

The coefficients of effect size (R2 marginal and R2 conditional) were computed

for both models using the r.squaredGLM function included in the MuMIn package

(Bartoń, 2019). R2 marginal estimates the amount of variance accounted for by all the

fixed effects included in the model (i.e., fixed effects, control variables, and

interactions). R2 conditional estimates the amount of variance accounted for by all

the random effects and fixed effects included in the model (P. C. D. Johnson, 2014;

Nakagawa & Schielzeth, 2013). Additionally, the variance inflation factor (VIF) and

the VIF tolerances were computed for both models in order to assess multicollinearity

between predictors (Field, Miles, & Field, 2012, p. 293). Both models yielded VIF

values well below 10 and VIF tolerance values above .2. Based on these values, it can

be safely concluded that there is no multicollinearity within the data (VIF values for

both models are provided in the Appendix).

 103

EFFECTS OF CROSSLINGUISTIC OVERLAP

4.2.1 Mixed-Effects Models Results

4.2.1. Model
LMER.BEST.FIT 1 - Liner Mixed-Effects
<- lmer(invRT Model+fitted
~ LANGUAGE*SEMANTIC with LMER
ZNLD_gmc*ZPHONETIC_gmc +

ZEnglish_Freq_gmc + Zlength_TARGET_raw_gmc + ZTime_ms_gmc + OTHER_LANGUAGES +

(1|PAIR) + (1|PRIME) + (1|TARGET) + (SEMANTIC|PARTICIPANT) +

(0+ZEnglish_Freq_gmc|PARTICIPANT) + (0+ZTime_ms_gmc|PARTICIPANT) +

(0+Zlength_TARGET_raw_gmc| PARTICIPANT), data = Dissertation_Dataset,

REML = FALSE, control=lmerControl(optimizer="nloptwrap2"))

4.2.1.1. Primary Fixed Effects

There was a significant main effect of LANGUAGE (p < .01). In comparison to

functional monolingual participants (reference category), bilingual participants were

on average 77ms slower. A significant main effect of SEMANTIC (semantic similarity)

was also found for cognates (p < .001), false friends (p < .01) and distractors (p < .001).

Specifically, irrespective of language group, in comparison to control pairs (reference

category), cognate pairs yielded on average 22ms faster reaction times, false friend

pairs yielded 14ms faster reaction times, and distractor pairs yielded 97ms slower

reaction times. These findings are in-line with what was initially predicted;

crosslinguistic semantic similarity or overlap on average had a significant effect on

participants’ RT.
 104

EFFECTS OF CROSSLINGUISTIC OVERLAP

4.2.1.2. Secondary Fixed Effects.

There was a significant main effect of English_Freq (p < .001). Specifically,

participants’ reaction time decreased on average six milliseconds as target frequency

count increased. A significant main effect of target_length was also found (p < .001).

Longer targets, those comprised of longer letter strings, slowed participants’ reaction

times 11ms on average. In addition, a main effect of Time was encountered (p < .01).

Later trials yielded significantly faster reaction times than earlier ones, i.e.,

participants got on average 19ms faster as they progressed through the experiment.

Finally, there was a trend towards a main effect of OTHER_LANGUAGES (p = .07).

Specifically, participants with the knowledge of another language besides English

and Portuguese were on average 45ms slower than participants who only spoke

English or both English and Portuguese.

4.2.1.3. Interactions.

The LANGUAGE x SEMANTIC interaction was significant for cognate pairs (p

< .01). Specifically, bilingual participants were on average 14ms faster on cognate pairs

in comparison to functional monolinguals. In addition, the NLD (crosslinguistic

orthographic overlap) x PHONETIC (crosslinguistic phonological overlap)

interaction was also significant (p < .01). Specifically, prime-target pairs with a high

degree of crosslinguistic orthographic overlap but with a low degree of crosslinguistic

phonological overlap slowed down participants three milliseconds on average.

 105

EFFECTS OF CROSSLINGUISTIC OVERLAP

LMER model effect size.

The model’s marginal effect size estimate was 16.14%. The primary and

secondary fixed effects plus their interactions did not account for much of the

observed variability in response latency. Nevertheless, the model’s conditional effect

size estimate was 61.43%. Adding the variance accounted for by PARTICIPANT, PAIR,

PRIME and TARGET random effects explained over half of the observed variability.

Table 19 contains all of the lmer model results.10

10
The raw RStudio output of the lmer model is provided in Appendix G.

Appendix G
 106

EFFECTS OF CROSSLINGUISTIC OVERLAP

Table 19. lmer Model Results

Fixed effects:
invRT RT ms
Effect in ms Std. Error df t value F Pr(>|t|) ,---- d
Estimate Estimate
(Intercept) -1.713 584 0.064 70.91 -26.818 719.205 < 2E-16 ***
LANGUAGEBilinguals 0.199 660 77 0.076 64.02 2.601 6.765 0.011542 * 0.44
SEMANTICCOG -0.068 562 -22 0.020 287.80 -3.451 11.909 0.000643 *** -0.15
SEMANTICDISTR 0.244 681 97 0.027 79.54 8.997 80.946 9.31E-14 *** 0.54
SEMANTICFF -0.041 570 -14 0.020 442.00 -2.044 4.178 0.041527 * -0.09
ZNLD_gmc -0.008 581 -3 0.006 1987.00 -1.295 1.677 0.195404 -0.02
ZPHONETIC_gmc 0.003 585 1 0.003 5485.00 1.267 1.605 0.205051 0.01
ZEnglish_Freq_gmc -0.019 577 -6 0.005 461.70 -3.527 12.440 0.000462 *** -0.04
Zlength_TARGET_raw_gmc 0.030 594 11 0.006 366.70 5.121 26.225 4.92E-07 *** 0.07
ZTime_ms_gmc -0.057 565 -19 0.017 59.40 -3.389 11.485 0.00125 ** -0.13
OTHER_LANGUAGESYES 0.123 629 45 0.067 56.91 1.826 3.334 0.07305 . 0.27
LANGUAGEBilinguals:SEMANTICCOG -0.042 570 -14 0.018 55.20 -2.32 5.382 0.023631 * -0.09
LANGUAGEBilinguals:SEMANTICDISTR -0.005 582 -2 0.035 56.56 -0.128 0.016 0.898777 -0.01
LANGUAGEBilinguals:SEMANTICFF 0.018 590 6 0.015 53.07 1.157 1.339 0.252305 0.04
ZNLD_gmc:ZPHONETIC_gmc 0.008 586 3 0.003 1671.00 2.299 5.285 0.021632 * 0.02

Random effects:

Groups Name Variance Std. Dev. Corr.

PRIME (Intercept) 0.0043107 0.06566
PAIR (Intercept) 0.0022357 0.04728
TARGET (Intercept) 0.0092168 0.096
PARTICIPANT Zlength_TARGET_raw_gmc 0.0004832 0.02198
PARTICIPANT.1 ZTime_ms_gmc 0.0152537 0.12351
PARTICIPANT.2 ZEnglish_Freq_gmc 0.0001005 0.01003
PARTICIPANT.3 (Intercept) 0.0655086 0.25595
SEMANTICCOG 0.0031251 0.0559 0.46
SEMANTICDISTR 0.0170481 0.13057 -0.36 0.19
SEMANTICFF 0.001071 0.03273 0.54 0.63 0.22
Residual 0.0828356 0.28781
 107

EFFECTS OF CROSSLINGUISTIC OVERLAP

4.2.2. Model 2 - Liner Mixed-Effects Model fitted with GLMER

GLMER.BEST.FIT <- glmer(RAW_ACCURACY ~ LANGUAGE*SEMANTIC + ZEnglish_Freq_gmc

+ ZTime_ms_gmc + Zlength_TARGET_raw_gmc + ZSEMAC_gmc + (1| PARTICIPANT) +

(1|PRIME) + (1|TARGET) + (1|PAIR), data = Dissertation_Dataset_LMER,

family=binomial, control=glmerControl(optimizer="nloptwrap2"))

4.2.2.1. Primary Fixed Effects

There was a significant main effect of LANGUAGE (p = .038). Functional

monolinguals were overall more accurate than bilinguals. A main effect of

SEMANTIC was also found (p < .001). Specifically, irrespective of their language

group membership, participants were less accurate overall on distractor pairs.

4.2.2.2. Secondary Fixed Effects.

There was a significant main effect of English_Freq (p < .001). Specifically,

irrespective of their language group membership, participants were more accurate in

judging whether or not a target was a genuine English word when they were shown

targets that occur more frequently in the language. A significant main effect of Time

was also found (p < .001). Specifically, irrespective of their language group

membership, participants’ accuracy increased as they progressed through the task,

i.e., they were more accurate at judging a target as a genuine English word when it

occurred in later trials in the experiment. A main effect of target length

 108

EFFECTS OF CROSSLINGUISTIC OVERLAP

(length_TARGET) was also found (p < .001). Specifically, irrespective of language

group membership, participants’ accuracy increased as target length increased, i.e.,

they were more accurate in longer targets. Finally, a significant main effect of English

proficiency (SEMAC) was found (p < .001). Specifically, irrespective of language group

membership, participants who were on average more proficient in English, i.e., who

had scored higher in the cloze proficiency test, were also more accurate in the

experimental task.

4.2.2.3. Interactions.

The LANGUAGE x SEMANTIC interaction was significant (p = .028).

Specifically, bilingual participants were more accurate than functional monolinguals

controls in trials where the target were matched with a Portuguese prime cognate.

4.2.2.4. GLMER model effect size.

The model’s marginal effect size estimate, which takes into account the

response accuracy variability associated with all of the primary and secondary fixed

effects plus their interactions, was 19.39%. The model’s conditional effect size

estimate was 56.25%. Adding the variance accounted for by PARTICIPANT, PAIR,

PRIME and TARGET random effects explained over half of the observed variability.
 109

EFFECTS OF CROSSLINGUISTIC OVERLAP

Table 20 shows all of the glmer model results11.

11
The raw RStudio output of the glmer model is provided in Appendix H.
 110

EFFECTS OF CROSSLINGUISTIC OVERLAP

Table 20. glmer Model Results

Estimates
Fixed effects: Log Odds (Logit) Odds Ratio ...__
Std. Error z value Pr(>|z|)
Log Odds (Logit) CI Lower Limit CI Upper Limit Odds Ratio CI Lower Limit CI Upper Limit
(Intercept) 4.877 -5.337 -4.418 131.248 0.005 0.012 0.234 20.817 < 2e-16 ***
LANGUAGEBilinguals -0.613 0.033 1.194 0.542 1.033 3.299 0.296 -2.073 0.038353 *
SEMANTICCOG 0.284 -0.740 0.172 1.329 0.477 1.187 0.233 1.222 0.221837
SEMANTICDISTR -1.372 0.997 1.748 0.253 2.710 5.744 0.192 -7.166 7.94E-13 ***
SEMANTICFF 0.185 -0.713 0.343 1.203 0.490 1.410 0.269 0.686 0.492957
ZEnglish_Freq_gmc 0.264 -0.431 -0.097 1.303 0.650 0.907 0.085 3.099 0.001942 **
ZTime_ms_gmc 0.174 -0.241 -0.108 1.191 0.786 0.898 0.034 5.136 2.80E-07 ***
Zlength_TARGET_raw_gmc 0.220 -0.363 -0.077 1.246 0.695 0.926 0.073 3.017 0.002549 **
ZSEMAC_gmc 0.496 -0.753 -0.238 1.642 0.471 0.788 0.131 3.773 0.000162 ***
LANGUAGEBilinguals:SEMANTICCOG 0.540 -1.022 -0.058 1.716 0.360 0.944 0.246 2.196 0.028062 *
LANGUAGEBilinguals:SEMANTICDISTR -0.136 -0.161 0.433 0.873 0.851 1.543 0.152 -0.900 0.368654
LANGUAGEBilinguals:SEMANTICFF -0.348 -0.170 0.867 0.706 0.843 2.381 0.265 -1.317 0.18815

Random effects:
Groups Name Variance Std.Dev.
PRIME (Intercept) 0.2766 0.5259
PAIR (Intercept) 0.3275 0.5723
TARGET (Intercept) 1.5102 1.2289
PARTICIPANT (Intercept) 0.6580 0.8112
 111

EFFECTS OF CROSSLINGUISTIC OVERLAP

5. The Study: Discussion

Several well-known models of the mental lexicon leverage spreading activation to

represent language schematically as a multilayered network of interconnected nodes,

which is analogous to how cognitive psychologists posit human memory is organized

(Anderson, 1983a, 1983b; Collins & Loftus, 1975; Collins & Quillian, 1969; Samani &

Sharifian, 1997). In this framework, linguistic knowledge is depicted as network

nodes; the associations between and among the countless types of linguistic

knowledge are depicted as the connection links. Thus, when a network node is

activated, it automatically spreads its activation, in iterative fashion, to other

linked/associated nodes. Network nodes that are repeatedly activated together are

represented closer to one another, with shorter connection links, in order to indicate

their stronger linguistic relationships.

A hypothesis proposed by several well-known models of the bilingual mental

lexicon, viz., RHM (J. F. Kroll & Stewart, 1994), BIA (Dijkstra & van Heuven, 1998;

Van Heuven et al., 1998), BIA+ (Dijkstra & van Heuven, 2002) and Multilink (Dijkstra

et al., 2018), is that bilingual language processing is non-selective. Specifically, during

bilingual visual word recognition, lexical candidates from both L1 and L2 are

simultaneously activated, which is a direct consequence of the bilingual mental

lexicon being integrated, i.e., lexical representations from both languages are stored

together.

The current study was conceived upon the principles set forth by spreading
 112

EFFECTS OF CROSSLINGUISTIC OVERLAP

activation and concerned with Portuguese-English cognates and false friends with

respect to their crosslinguistic form (orthographic and phonological) and meaning

(semantic) overlap. From a theoretical perspective, its chief goal was to examine how

lexical representations are organized in the Portuguese-English bilingual mental

lexicon as well as to elucidate the lexical selection process. In addition, this study

sought to quantify the specific effects that crosslinguistic form and meaning overlap,

and any of their interactions, as well as word frequency, length, second language

proficiency as control variables exert in this organization and during lexical selection.

To that end, a lexical decision task with masked priming in the forward L1-to-L2

priming direction (Portuguese-English) was used. The experimental stimuli were

comprised of Portuguese-English cognates and false friends. Portuguese primes used

as experimental stimuli were also matched with English targets without form or

meaning overlap crosslinguistically and served as experimental controls. Distractor

items were comprised of Portuguese primes that had not been used as experimental

or control stimuli and matched with pseudowords (nonwords) as targets. L1-

Portuguese-L2-English bilinguals served as the experimental group and functional

English monolinguals served as controls. Two mixed-effects analyses were carried out

to analyze the data, and participants’ response latency and accuracy served as

dependent variables. The findings provide empirical support for many of the well-

known models of the bilingual mental lexicon discussed in Chapter 2, in particular

to the BIA+. Thus, the following discussion was structured primarily in regard to the
 113

EFFECTS OF CROSSLINGUISTIC OVERLAP

BIA+’s framework.

5.1. Findings

A main effect of LANGUAGE was observed for response latencies (p < .05). When

other variables are not considered, bilingual participants’ reaction times were

significantly slower (77ms on average) than functional monolingual participants’,

indicating that lexical decision happens in a slower fashion overall for bilinguals. The

main effect of LANGUAGE was also observed for response accuracy (p < .o5).

Bilingual participants’ accuracy, i.e., deciding whether or not a target was an English

word, was lower than functional monolingual participants’. These two results point

to a language switching cost; when bilinguals are primed with an L1 word and the

lexical decision is carried out in their L2, their reaction times are significantly slower

relative to functional monolingual controls. Further, besides the cost of switching

from their L1 to their L2, bilinguals’ lexical decisions were less accurate compared to

functional monolingual controls’. This decrease in lexical decision accuracy can be

accounted for by Portuguese primes spreading their activation to other (non-target)

English competitors. This reasoning is in line with the findings reported by von

Studnitz & Green (1997), which also employed a masked priming lexical decision

task.

A crosslinguistic priming effect was observed on participants’ response

latencies, but not on response accuracy. Both cognates (SEMANTICCOG) and false
 114

EFFECTS OF CROSSLINGUISTIC OVERLAP

friends (SEMANTICFF) yielded on average faster reaction times in comparison to

controls (both ps < .001). Cognates, however, had a significant 8ms speed-advantage

over false friends. It follows then that crosslinguistic semantic overlap between a L1

prime and an L2 target exerts a facilitatory effect (22ms on average) on participants

lexical decision times, potentially a result of the semantic representations exerting

strong feedback activation on orthographic representations. It follows then that the

greater the degree of semantic overlap between an L1 prime and L2 target, as it was

the case with cognates, the faster participants’ reaction times were on average. To

further support this assertion, two other findings regarding crosslinguistic semantic

overlap need also to be taken into consideration. Participants’ reaction times in false

friend trials (SEMANTICFF), i.e., word pairs in which the crosslinguistic form

(orthographic and phonological) overlap between the Portuguese prime and English

target was high but semantic overlap was low (or non-existent), despite being on

average significantly faster than control trials (p < .05), were still less robust relative

to cognate trials. Further, participants’ reaction times on distractor trials

(SEMANTICDISTR), i.e., pairs in which a bona fine Portuguese prime was matched

to a pseudoword (nonword) target (no crosslinguistic form and meaning overlap

whatsoever), were the slowest of all (97ms on average) in comparison to control trials

(p < .001). Soares and Grosjean (1984), report similar findings with nonwords, with

Portuguese-English bilinguals showing significantly slower reaction times to

nonwords in comparison to genuine words from either language.

 115

EFFECTS OF CROSSLINGUISTIC OVERLAP

The aforementioned findings are, however, puzzling; bilinguals’ and

functional monolinguals’ reaction times were significantly faster in cognate and false

friend trials. This begs the question how functional monolingual participants could

have been subject to crosslinguistic semantic priming, at least with respect to

cognates, even when they expressed not having any knowledge of Portuguese during

the intake assessment interview. Cognate and false friend trials have a common

characteristic: a high degree of crosslinguistic form (orthographic and phonological)

overlap compared to unrelated controls. It is thus possible that upon presentation of

the masked Portuguese prime, its lexical and sublexical orthographic (and

phonological) representations may have been sufficiently able to activate a number

of lexical and sublexical orthographic and phonological representations in both the

target and non-target language, which, in turn, facilitated participants’ lexical

decisions, even in the absence of shared semantic representations to exert feedback

activation. Further, since the effect was also significant in false friend trials

irrespective of language group, it is safe to assume that crosslinguistic orthographic

priming was in fact the culprit.

The interaction between language group and crosslinguistic semantic overlap

(LANGUAGEBilinguals:SEMANTICCOG) was significant, i.e., a facilitatory

crosslinguistic priming effect (often reported in the literature as cognate facilitation

effect) was observed on bilinguals’ response latencies and response accuracy (both ps

< .05). Bilinguals were on average 14ms faster to respond to an English target that was
 116

EFFECTS OF CROSSLINGUISTIC OVERLAP

semantically, orthographically, and phonologically related to a matched Portuguese

prime (cognate) than functional monolingual controls. In addition, bilingual

participants were more accurate at deciding whether a target as a bona fide English

word when it was matched to a semantically, orthographically and phonologically

related Portuguese prime relative to an unrelated Portuguese c0ntrol prime, which

corroborates the results reported in Alves-Soares (2013). Because bilingual

participants were not only significantly faster to react but also significantly more

accurate to respond to English targets in cognate trials compared with functional

monolingual participants, it is safe to conclude that both semantic and

crosslinguistic form overlap facilitate bilinguals’ lexical decisions.

These findings align well with the results of several noteworthy studies on the

bilingual mental lexicon (Altarriba & Basnight-Brown, 2007; Caramazza & Brones,

1979; Cristoffanini et al., 1986; Dijkstra et al., 1999; Dijkstra, Miwa, Brummelhuis,

Sappelli, & Baayen, 2010; Duñabeitia, Perea, & Carreiras, 2010; Duyck, Van Assche,

Drieghe, & Hartsuiker, 2007; Forster et al., 2003; Gerard, Linda, & Scarborough, 1989;

Guasch et al., 2013; Kim & Davis, 2003; Lemhöfer & Dijkstra, 2004; Nakayama et al.,

2012; Peeters, Dijkstra, & Grainger, 2013; Sanchez-Casas, Davis, & Garcia-Albea, 1992;

Schwartz et al., 2007). Taking into account the framework proposed by the BIA+ and

its updated version named the Multilink (both discussed in Chapter 2), they also have

major implications for the inner workings as well as the organization of the

Portuguese-English bilingual mental lexicon. First, these findings show that

 117

EFFECTS OF CROSSLINGUISTIC OVERLAP

crosslinguistic priming effects are automatic and not under direct control of the

bilingual. Second, a certain level of crosslinguistic integration exists in the bilingual

mental lexicon. Specifically, presentation of a prime, even when it is masked, in one

language automatically activates a number of candidates in the target language

proportional to the degree of crosslinguistic orthographic overlap between the prime

and target. Third, when the degree of crosslinguistic form and semantic overlap

between an L1 prime and an L2 target is high, i.e., when they are cognates, their

shared semantic representations exert strong feedback activation on their

orthographic representations; thus, the cognate facilitation effect is a direct result of

cognates receiving increased activation from their shared semantics, which

accelerate bilinguals’ responses for this group of words. In addition, in the absence of

crosslinguistic semantic overlap between an L1 prime and an L2 target, facilitatory

crosslinguistic priming effects can still occur due to crosslinguistic form overlap.

Finally, these findings support for the language non-selective access hypothesis, i.e.,

lexical candidates from bilinguals’ both languages are simultaneously activated upon

presentation of word in either language during the first stages of visual word

recognition. With respect to the current study, presentation of a masked Portuguese

prime facilitated bilinguals’ recognition of a cognate English target.

Despite the fact that crosslinguistic orthographic (ZNLD_gmc) and

phonological overlap (ZPHONETIC_gmc) individually yielded no significant effect

on participants’ response times and accuracy, their interaction

 118

EFFECTS OF CROSSLINGUISTIC OVERLAP

(ZNLD_gmc:ZPHONETIC_gmc) was significant (p < .05) on participants’ response

latencies. This interaction, however, did not enter the second mixed-effects model

(logistic regression glmer model), as it did not contribute significantly to

participants’ response accuracy.

In order to better understand how crosslinguistic form overlap affected

participants’ response latencies, some facts regarding stimuli selection need to be

recollected (see chapter 3, Methodology for further detail on stimuli selection). The

Normalized Levenshtein Distance (Schepens et al., 2013) was selected as an objective

metric of crosslinguistic orthographic overlap. It yields values that range between 0,

when orthographic overlap between a prime and its matched target is nonexistent,

and 1, when the prime and target are orthographically identical. Dynamic time

warping was selected as an objective metric of crosslinguistic phonological overlap,

and Phonological Corpus Tools (Hall et al., 2016) was the software package used to

implement the DTW algorithm. These two objective metrics allowed this study to

more elegantly intuit how these two variables interact in the Portuguese-English

bilingual mental lexicon. In addition, examining this interaction in a mixed-effects

analysis with this language pair had not been yet attempted.

The significant interaction between crosslinguistic orthographic and

phonological overlap suggests that a certain level of integration exists between both

the L1 and the L2 and that orthographic representations spread their activation to

associated phonological representations. Specifically, during the course of visual

 119

EFFECTS OF CROSSLINGUISTIC OVERLAP

word recognition, upon presentation of an L1 prime, orthographic representations of

the prime become activated and spread their activation in a feed-forward fashion to

orthographic representations from both languages based on their degree of

orthographic overlap with the prime. Likewise, phonological representations from

both languages are also activated subsequently based on their degree of phonological

overlap with the prime. In addition, this significant interaction between

crosslinguistic orthographic and phonological overlap may indicate the BIA+’s

“temporal delay assumption” (Dijkstra & van Heuven, 2002, p. 183) deserves merit.

The model proposes that there are time course differences in the contribution of the

two codes in lexical decision, with lexical and sublexical orthographic code being

activated first and L2 lexical and sublexical phonological code being activated slightly

later. As such, crosslinguistically phonological overlap appears to modulate

activation spread, particularly in the forward L1-to-L2 direction. It follows then the

significant 3ms inhibitory effect observed with this interaction may be a consequence

of the temporal delay. Alternatively, it is possible that this 3ms delay is a result of the

inconsistencies in mapping Portuguese orthographic code to English phonological

code (and vice-versa). To that end, a certain grapheme (or its combinations) in

Portuguese may be consistently pronounced the same way in the language, whereas

in English, the pronunciation of the grapheme is much less consistent. These

differences in crosslinguistic phonological overlap relative to orthographic overlap

may induce lateral inhibition, since the non-overlapping phonological code would
 120

EFFECTS OF CROSSLINGUISTIC OVERLAP

also activate lexical representations that are less phonologically similar to both prime

and target, which would reduce the magnitude/robustness of the facilitatory

crosslinguistic form overlap effect. Although both these explanations can be

accommodated within the BIA+ framework, it is important to point out that it is

rather difficult (perhaps nearly impossible) to precisely disentangle crosslinguistic

phonological effects from orthographic effects in languages such as Portuguese and

English (as well as French, Spanish and Italian), which despite sharing the same

roman script and having a great number of words with 100% orthographic overlap,

almost never have words with 100% phonological overlap.

One of the control variables that was of particular interest in this study was

word frequency. While Portuguese (prime/L1) word frequency

(ZPortuguese_Freq_gmc) yielded no significant main effect on both response

latencies and accuracy, probably a consequence of the experiment having been

conducted in the forward L1-to-L2 priming direction (Portuguese-English), English

(target/L2) word frequency (ZEnglish_Freq_gmc) did have an effect on both of

dependent variables. Participants, irrespective of language group membership, were

on average 6ms faster (p < .001) to select targets that occur more frequently in English,

i.e., have a higher frequency count per million words in the language. In addition,

irrespective of their language group membership, i.e., whether they were classified

bilingual or functional monolingual, participants were more accurate at recognizing

high frequency count targets as genuine words of the English language, compared to
 121

EFFECTS OF CROSSLINGUISTIC OVERLAP

low frequency targets (p < .01).

These two findings provide further insight of the inner workings of the

Portuguese-English bilingual mental lexicon. To that end, L2 word frequency

facilitates lexical decision in terms of speed and accuracy, but this facilitatory effect

is not substantial (only a 6ms facilitation in response times, d= .03). Both groups of

participants were significantly faster and more accurate on average to respond to

English targets that occur more frequently in the language. This finding is in line with

the BIA+ proposal. The model purports to say that word frequency affects how

activation spreads in the Word Identification System, which, in turn, affects how the

Task Schema selects a specific target. Specifically, frequency of use modulates how

activation spreads from the input string to linked representations in the Word

Identification System by raising (or even lowering) the resting-level activation of

stored representations. In this fashion, representations linked to frequently used

words have sufficiently higher resting-level activation and, consequently, spread their

activation to other linked representations much faster, which, during the course of

visual word recognition, facilitates lexical decision. Likewise, frequently used L2

targets have a higher baseline (resting) level of activation, i.e., they are more salient,

thus, less activation is required from an L1 prime to effect activation of L2 target

candidates and the spread of activation between prime and target occurs at a

significantly faster rate.

The BIA+ accounts for how representations from one language can have
 122

EFFECTS OF CROSSLINGUISTIC OVERLAP

sufficiently higher resting-level activation relative to representations from the other

language. In respect to the current study, because English was the target language of

the lexical decision task and was also the primary language of communication for all

of the participants, it would be safe to assume that English representations had ample

time to establish themselves in the mental lexical, i.e., increase their resting-level

activation, and exert strong feedback activation on other competitor English

representations, which facilitated the Task Schema’s verification (checking) whether

a specific target belonged to English, even when competing Portuguese

representations might have been available for the bilingual participants. In this

framework, as bilingual and functional monolingual participants were using English

as their primary language of communication, study and work on a daily basis, and the

lexical decision task was carried out in English, it is possible that “recency of use of

the target language” not only accelerated lexical selection in English but also exerted

top-down inhibition that made lexical selection in the language significantly more

accurate. Had the experimental task been conducted in the backwards priming

direction (English prime – Portuguese target), it is conceivable that L1 word

frequency effects on both response latency and accuracy would have been detected,

but only for bilingual participants. In order to more precisely quantify the effects of

recency of use, it would be important for future studies to compare bilinguals living

in their L2 with bilinguals living their L1 in future studies.

Since the effect of L2 word frequency was shown not to be substantial, instead
 123

EFFECTS OF CROSSLINGUISTIC OVERLAP

of adding word frequency as an isolated control variable in a mixed-effects model, as

it was done in the current study, future studies should assemble an experimental

stimuli list in which the frequency of the readings of cognates and false friends is

carefully controlled for relative to controls and pseudowords. Additionally, it would

be important verify whether switching the direction of the lexical decision task, viz.,

Portuguese prime-English target vs. English prime-Portuguese-target vs. Portuguese

prime-Portuguese target vs. English-prime vs. English target, yields substantive

differences.

Target length (Zlength_TARGET_raw_gmc) is another control variable that

had a significant effect on participants’ response latency. Participants, irrespective of

language group membership, processed longer targets, i.e., were comprised of more

letters, on average 11ms slower (p < .001). Besides potentially having been influenced

by the priming direction of the experimental task, this finding suggests a temporal

delay/cost (inhibitory effect) associated with the processing of longer words,

potentially due to greater memory (cognitive) load requirements. In this way, upon

the presentation of an L1 prime, a number of lexical orthographic representations are

simultaneously activated dependent on their degree of orthographic of overlap with

prime during the first states of visual word recognition. As prime-target pairs in the

experimental task were matched for length, and both languages of the experiment

share the same alphabet, it is possible that the number of lexical competitors (or

neighbors) activated by the L1 prime is a function of the total number of letters that
 124

EFFECTS OF CROSSLINGUISTIC OVERLAP

comprises the prime. Specifically, longer Portuguese primes activated a greater

number of lexical competitors in English (the target language) due to an increased

number of acceptable letter permutations crosslinguistically, slowing down lexical

selection by 11ms on average.

Target word length also had a significant effect on response accuracy,

irrespective of language group membership. Participants were more accurate at

recognizing a target as a genuine English word when they were longer in length (p <

.01). This shows that despite the cost associated with the simultaneous activation of

English competitors, which slowed participants down 11ms on average, lexical

selection still occurred in a more accurate fashion. This finding provides support for

the existence a task schema, an idea first purported in the BIA+. According to Dijkstra

& van Heuven (2002), the Task Schema, which is in constant communication with the

Word Identification System, is responsible for ensuring proper “safe execution” of the

task at hand by verifying the membership of possible lexical candidates within a

specific language. Because the effect of target word length affected both bilinguals

and functional monolinguals, it would be safe to assume a task schema had to

properly verify the membership of all the active lexical competitors in the target

language, in this case English, in order to accurately narrow down a specific target.

As word length affected the number of possible competitors simultaneously active, a

greater number of lexical competitors allowed the task schema to more accurately
 125

EFFECTS OF CROSSLINGUISTIC OVERLAP

select an English target, thereby yielding the significant effect of response accuracy

observed.

The variable ZTime_ms_gmc, which encoded the amount of time

participants, irrespective of language group membership, spent on the experimental

task, yielded a significant effect on both response latency (p < .01) as well as response

accuracy (p < .001). Bilinguals’ and functional monolinguals’ lexical decisions were

on average faster (19ms) and more accurate on later trials compared to earlier ones.

Two scenarios may explain these findings. (a) A facilitatory effect of practice:

as participants got more familiar with the lexical decision task, their reaction times

and accuracy in later trials significantly improved, probably also due to the rather

large total number of trials that comprised the experiment. In fact, in the extant

literature the effect of practice has been observed in crosslinguistic lexical decision

tasks with respect to response latency and accuracy when data was analyzed using

mixed effects models (Keuleers, Diependaele, & Brysbaert, 2010). Nevertheless,

despite the highly significant (p < .001) effect of practice, in practical terms it was

described as rather small, (40ms difference in response latency and 2% difference in

response accuracy) and deemed to be a result of the large number of trials. In the

current study, the effect of practice may have been partly an upshot of the large

number of trials as well. In addition, the observed effect was also rather small (19ms

difference in response latency by the end of the task). (b) A facilitatory recency effect:

the linguistic representations of earlier prime-target pairs may have retained a

 126

EFFECTS OF CROSSLINGUISTIC OVERLAP

proportion of their activation and facilitated the activation of linguistic

representations of prime-target pairs in later trials, thereby expediting the lexical

selection process. As Portuguese primes in the experimental condition were matched

with English targets that were unrelated in both form and meaning, it is reasonable

to assume that despite pseudorandomization, Portuguese primes from earlier trials

could have retained a proportion of their activation and facilitated lexical selection

of targets in later trials. In addition, it is plausible that this recency effect was further

accentuated (additive effect) by the fact that all participants were using English as

their primary language of communication and that the experimental task carried out

in the Portuguese-English direction, with English being the target language. In other

words, living in an English-speaking environment may have given English targets an

edge, allowing participants to process them faster and more accurately.

English proficiency had no impact on response latency, and, in fact, it did not

enter the lmer model. Nevertheless, English proficiency (ZSEMAC_gmc) yielded a

main effect that was highly significant for response accuracy (p < .001), irrespective

of language group membership, and thus, entered the glmer model. Specifically,

participants who obtained a higher score on the cloze proficiency test were on average

significantly more accurate at judging a target as a genuine word of the English

language.

Several studies have reported robust translation priming effects (asymmetry)

in the forward L1-L2 priming direction using masked priming technique in

 127

EFFECTS OF CROSSLINGUISTIC OVERLAP

unbalanced bilinguals (Altarriba & Basnight-Brown, 2007; Dimitropoulou,

Duñabeitia, & Carreiras, 2011c; Finkbeiner, Forster, Nicol, & Nakamura, 2004; Jiang,

1999; Nakayama, Ida, & Lupker, 2016; Nakayama, Sears, Hino, & Lupker, 2013; Wang,

2013; Wen & van Heuven, 2017; Xia & Andrews, 2015). The robust translation

asymmetries are suggested to occur due to an imbalance between bilinguals’ L1 and

L2. Specifically, L1 lexical representations have a higher resting-level activation

compared to L2 lexical representations, thus activation spreads more quickly and

easily in the forward L1-to-L2 masked priming direction than the other way around.

In addition, it has been reported that in order to obtain similar effects in the

backwards L2-to-L1 direction, unbalanced bilinguals require that the L2 prime be

presented for longer, greater than the standard Stimuli Onset Asynchrony (SOA) of

50ms (Wen & van Heuven, 2017).

Nevertheless, it has also been reported in the literature that the translation

priming effect is less robust or completely disappears for balanced bilinguals

(Duñabeitia, Dimitropoulou, et al., 2010; Duñabeitia, Perea, et al., 2010; Perea et al.,

2008). Additionally, the BIA+ model (Dijkstra & van Heuven, 2002), proposes that as

L2 proficiency increases, the resting-level activation of L2 lexical representation also

increases. As a result, translation priming asymmetry effects for balanced bilinguals

as well as highly proficient bilinguals would not be significant or, if significant, their

effect size (d) would be small (Wen & van Heuven, 2017).

The aforementioned findings indicate that the bilinguals who participated in

 128

EFFECTS OF CROSSLINGUISTIC OVERLAP

the current study may have been more proficient in English than the cloze proficiency

test was able to detect. In addition, it is likely that using their L2 (English) as their

primary language of communication in Canada on a daily basis affected the balance

between the L1 and the L2, placing English in a more equal position to their L1

(Portuguese). These two factors explain why English proficiency had no effect on

response latency for both bilinguals and functional monolinguals. Regarding the

significant effect of English proficiency on response accuracy, from a theoretical

standpoint, this finding is compatible with the BIA+ framework. Specifically, daily

exposure to their L2 not only helped bilinguals to become more proficient in English

but also allowed them to develop more stable lexical representations in the language

(higher resting-level activation), in turn, improving lexical decision accuracy. In

addition, it is likely that more stable L2 representations exert stronger top-down

inhibition suppressing the activation of competitor lexical representations in the

non-target language leading to better overall accuracy (Dijkstra & van Heuven, 2002,

p. 189).

Finally, the control variable OTHER_LANGUAGES yielded a main effect that

approached statistical significance (p = .08) with respect to response latencies, thus

deserves further examination. The variable encoded whether or not participants

spoke another European language besides English and Portuguese, which for the

purposes of this study was important to account for. Many models of the bilingual

mental lexicon (in particular the BIA and its updated version BIA+) propose that the
 129

EFFECTS OF CROSSLINGUISTIC OVERLAP

bilingual lexicon is integrated, that lexical access occurs in a non-selective fashion,

and that during the first stages of visual word recognition a number of orthographic

candidates from bilinguals’ both languages become activated depending on their

orthographic similarity with the input string (Van Heuven et al., 1998). This

(priming) effect is hypothesized to be proportional to the degree of crosslinguistic

form overlap (Cristoffanini et al., 1986). Therefore, participants that speak languages

with a high degree of crosslinguistic form overlap, such as French and Portuguese,

may manifest slower reaction times on a crosslinguistic lexical decision task. To that

end, this (almost significant) main effect of OTHER_LANGUAGES is potentially a

result of the high degree of crosslinguistic orthographic overlap between Portuguese

and French, both Romance Languages, and between both of these languages and

English. In the current study, all of the selected primes were in Portuguese. Upon

presentation of the Portuguese prime, Portuguese, English (and possibly French)

candidates were simultaneously activated for the bilingual participants. The

Portuguese prime activated English and potentially French candidates for functional

monolingual participants. Since the experiment took place at the University of

Ottawa, an institution where students are exposed to both English and French

regularly, in Canada’s National Capital Region, an area where residents are also

exposed to both English and French regularly, it is likely that lexical decision was

slowed down by their knowledge of French. Despite not being the target language,

French is historically related to Portuguese; consequently, both languages share a

 130

EFFECTS OF CROSSLINGUISTIC OVERLAP

great degree of orthographic code. Thus, having knowledge of French, or having been

exposed to it regularly, may have exerted an inhibitory effect on participants’ reaction

times. Furthermore, it is also possible that a few functional monolingual participants

may have learned other Romance Languages, such as Spanish and Italian, whose

words are be more similar orthographically to Portuguese than French. Ultimately,

this result indicates that more research is required in order to cognize how

crosslinguistic orthographic overlap between a prime and a non-target, yet

historically similar, third language can affect bilinguals’ lexical decision.

 131

EFFECTS OF CROSSLINGUISTIC OVERLAP

6. Conclusion and Suggestions for Further Research

In sum, the aforementioned findings provide strong empirical support for the BIA+.

First, the significant condition effects encountered suggest some level of integration

in the bilingual mental lexicon. Specifically, presentation of a Portuguese prime

facilitated participants’ reaction times to an English target when crosslinguistic form

(orthographic and phonological) overlap between them was high (cognates and false

friends). Because the effect was significant irrespective of language group and

crosslinguistic form overlap in cognates and false friends is high, it follows that

crosslinguistic form overlap was likely the responsible factor. Second, crosslinguistic

semantic overlap plays a leading role in the organization of the bilingual mental

lexicon. When the Portuguese prime and the English target shared the same

conceptual representation in permanent memory (cognates), bilingual participants’

reaction times were significantly faster than when the Portuguese prime and the

English target did not share the same conceptual representation (false friends). This

effect suggests that cognates may have a special representation in the bilingual

mental lexicon. Specifically, besides sharing a single semantic representation

crosslinguistically, cognates may share a single morphemic representation comprised

of common lexical and sublexical orthographic plus phonemic representations in

both bilingual’s languages. Besides providing strong evidence in favor integration,

the findings of this study also support the BIA+’s depiction of the bilingual mental

lexicon as being hierarchically organized into two cognitive systems, a Word

 132

EFFECTS OF CROSSLINGUISTIC OVERLAP

Identification System, which sits atop the hierarchy, and a Task Decision/Schema

System. The aforementioned crosslinguistic priming effects observed with bilingual

participants can be accounted for as being the result of these two complex, yet highly

interactive, cognitive processes. To that end, during the course of visual word

recognition, lexical and sublexical orthographic representations that comprised a

masked Portuguese prime automatically activated a number of similar (linked)

lexical and sublexical orthographic representations simultaneously in both

Portuguese and English, which, subsequently activated a number of similar lexical

and sublexical phonological representations in both languages as well. As English

was bilingual participants’ primary language of communication, English word

frequency, recency of use as well as proficiency level in the language had increased

the resting-level activation of words in the language relative to Portuguese words,

which allowed activation to spread to other similar English representations much

faster. Semantic representations further facilitated or inhibited activation spread to

similar lexical and sublexical orthographic and phonological representations in both

languages, which competed against one another. Recognition of the target ultimately

occurred when the Task Decision/Schema System, after continuous weighing of

activation spread in the Word Identification System and taking into account the

lexical decision task’s instructions, determined that the target had met recognition

threshold.

Despite all of the evidence gathered in support of the BIA+, the findings above
 133

EFFECTS OF CROSSLINGUISTIC OVERLAP

warrant future study in other to address the limitations encountered. First, the pool

of bilingual participants was not uniform. A sizeable number of bilingual participants

was comprised of international undergraduate students from Brazil who had come

to the University of Ottawa to participate in a one-year undergraduate exchange

program in their respective fields of study. Only a few of the bilingual participants

reported to be pursuing, or already had, post-graduate level of education. In addition,

while all bilingual participants indicated during the intake assessment interview that

English had become their primary language since arriving in Canada, most of them

had learned the language after starting high school in Brazil, which is usually the

norm in the country. Moreover, many of the bilingual participants admitted to having

studied French; some even mentioned they had attended L’Alliance Française in

Brazil for many years while they were also learning English. Adding to this, many

reported they were living in Gatineau, Québec, and had francophone roommates,

i.e., were exposed to French on a daily basis. Although one of the original goals of this

study was to quantify the effect of language dominance, age of English acquisition

and manner of acquisition on both dependent variables, aforementioned

circumstances made it impossible for this to be carried out.

Similarly, despite the fact that our functional monolingual participants were

native speakers of English and did not speak or know Portuguese, living in Ottawa,

an officially bilingual city, and working at University of Ottawa, the largest bilingual

university in Canada, meant that this group of participants was still exposed to
 134

EFFECTS OF CROSSLINGUISTIC OVERLAP

French on a daily basis. Some of the functional monolingual participants also

reported having studied Spanish and Italian. Since Portuguese, Spanish, Italian and

French are Romance Languages and share a great deal of morphological and

phonological similarities, it is possible these very facts skewed the results by making

functional monolingual participants more similar to bilingual participants.

Therefore, it would be critical for future studies to be stricter in selecting participants

for the control group, i.e., separating native speakers of English who speak only

English from native speakers of English who speak other romance languages, and not

conflate both into one single group.

Second, few studies have shed light on the dynamics of the Task Decision

Schema and the Word Identification System, in particular with respect to how the

types pseudowords (or nonwords) can sway how activation threshold is crossed for

recognition of a word to take place during a lexical decision task (Grainger & Jacobs,

1996; Lima & Huntsman, 1997; Lupker, Brown, & Colombo, 1997; Taylor & Lupker,

2001). To that end, after decision parameters are set during practice trials for an

English-only lexical decision task, the Task Decision Schema can be adapted or fine-

tuned over the course of the experiment, which can significantly skew target

recognition. Consequently, participants’ reaction times to pseudoword targets may

be facilitated, or participants may start automatically selecting a NO response for

pseudoword targets, if, for instance, all of the pseudowords are orthographically or

even phonologically unpronounceable in the target language. Despite the fact the
 135

EFFECTS OF CROSSLINGUISTIC OVERLAP

results of this study demonstrated that both group of participants were slowest on

average to respond to pseudoword targets (relative to unrelated English controls), in

order to minimize potential stimuli composition effects, it is vital for future studies

to continue to exercise restraint and select only pseudowords that are

orthographically and phonologically legal in English for distractor pairs. In addition,

it is vital that pseudowords match Portuguese primes in length and that their

orthographic and phonological overlap is low, i.e., they must not be interlingual

homographs or homophones.

Finally, although using objective metrics to assess the effects of crosslinguistic

overlap on the organization of the Portuguese-English bilingual mental lexicon

demonstrated to be a steppingstone that yielded meaningful results, future studies

should also factor in subjective metrics in mixed-effects model analyses. For instance,

it would be productive to establish whether there are significant differences in model

effect size between objective and subjective metrics, or even whether it would be

possible to create a composite metric, combining subjective and objective

crosslinguistic overlap metrics together.

In conclusion, this dissertation found empirical support for the BIA+ model,

in particular with respect to the language non-selective hypothesis, which suggest

that linguistic representations from bilinguals’ both languages were activated during

the course of visual word recognition. Additionally, even though crosslinguistic

orthographic and phonological overlap alone did not facilitate bilinguals’ recognition
 136

EFFECTS OF CROSSLINGUISTIC OVERLAP

of English targets, their interaction did. Specifically, bilingual participants, compared

with functional monolingual participants, were on average significantly faster to

respond and more accurate to identify English targets that were matched with

orthographically and phonologically similar Portuguese primes. Additionally, this

effect was further facilitated when the Portuguese prime and the English target

shared the same semantic representation, i.e., when they were cognates. These

findings suggest that cognates may share a single morphemic representation for in

the Portuguese-English bilingual mental lexicon. Moreover, it was found that both

bilingual and functional monolingual participants were sensitive to frequency of use

and word length in English only.

Taking into account the framework proposed by the BIA+, crosslinguistic

overlap facilitates how bilinguals (relative to functional monolingual controls)

process cognates. The significant crosslinguistic semantic priming effect observed

with this group of participants and cognates showed that crosslinguistic overlap

indeed facilitates bilinguals’ lexical decisions. Specifically, because bilingual

participants were not only significantly faster to react but also significantly more

accurate to respond to cognates compared with unrelated controls than functional

monolingual participants did, it is safe to conclude that lexical decisions are

facilitated by both semantic and crosslinguistic orthographic overlap. In addition, it

appears that semantic similarity in fact plays a major role in how activation spreads

crosslinguistically, since in its absence, as it was the case with false friends, no
 137

EFFECTS OF CROSSLINGUISTIC OVERLAP

crosslinguistic priming effect was observed. The BIA+ is able to accommodate the

significant effects of target word length on both response times and accuracy, as well

as the effect of time spent in the experiment.

This dissertation contributes to the advancement of the linguistic and

psycholinguistic fields in three major ways. First, despite the great deal of

crosslinguistic similarity or semantic, orthographic and phonological code overlap

between Portuguese and English, little research has been directed at expounding how

Portuguese-English bilinguals process cognates and false friends in the two

languages. In fact, the role that both crosslinguistic similarity and English proficiency

play in the organization of the Portuguese-English bilingual mental lexicon has been

largely overlooked. To that end, this dissertation serves as a steppingstone in

addressing these issues in a more quantitative fashion and also provides undeniable

evidence in support of the BIA and BIA+. Second, from a methodological standpoint,

this dissertation contributes to a core issue in bilingualism by demonstrating that

crosslinguistic similarity can be effectively measured using objective metrics of

similarity such as dynamic time warping (DTW) and the normalized Levenshtein

distance (NLD); thus, the hurdles and limitations many second language researchers

encounter for relying on subjective metrics of crosslinguistic similarity can now be

successfully overcome. Third, although ANOVAs and t-tests have been the primary

statistical analysis (means-based parametric) tools utilized in many second language

studies to analyze bilingual data collected in LDTs, liner mixed-effects models were
 138

EFFECTS OF CROSSLINGUISTIC OVERLAP

chosen instead for this dissertation. Besides being more robust and more precise at

estimating the individual effects of crosslinguistic similarity on both dependent

variables (reaction time and accuracy), mixed-effects models yielded a more realistic

representation of the processes involved in visual word recognition. Finally, at the

applied level, the fields of translation and language pedagogy could undoubtedly

benefit from the results presented in this dissertation by incorporating cognates in

teacher training and in the preparation of pedagogical materials.

 139

EFFECTS OF CROSSLINGUISTIC OVERLAP

7. References

Akaike, H. (1998). Information Theory and an Extension of the Maximum Likelihood

Principle. In E. Parzen, K. Tanabe, & G. Kitagawa (Eds.), Selected Papers of

Hirotugu Akaike (pp. 199–213). New York, NY: Springer New York.

https://doi.org/10.1007/978-1-4612-1694-0_15

Altarriba, J., & Basnight-Brown, D. M. (2007). Methodological considerations in

performing semantic- and ..., 39(1), 1–18.

Alves-Soares, L. (2013). Investigating the English-Portuguese Bilingual Mental Lexicon:

Effects of Orthography and Phonology. University of Ottawa, Ottawa, ON.

Amazon Elastic Compute Cloud (Amazon EC2). (2018). Amazon Web Services, Inc.

Retrieved from https://aws.amazon.com/ec2/

Anderson, J. R. (1983a). A Spreading Activation Theory of Memory. Journal of Verbal

Learning and Behavior, 22, 261–265. https://doi.org/10.1016/B978-1-4832-1446-

7.50016-9

Anderson, J. R. (1983b). The Architecture of Cognition. Cambridge, MA, USA:

Harvard University Press.

Anderson, J. R. (2015). Cognitive psychology and its implications (8th ed.). New York,

NY: Worth Publishers.

Audacity®: Free Audio Editor and Recorder. (2019). Audacity Team. Retrieved from

https://audacityteam.org/

Barr, D. J., Lev, R., Scheepers, C., & Tily, H. J. (2013). Keep it maximal Appendix.
 140

EFFECTS OF CROSSLINGUISTIC OVERLAP

Journal of Memory and Language, 68(3), 1–5.

https://doi.org/10.1016/j.jml.2012.11.001.Random

Bartoń, K. (2019). MuMIn: Multi-Model Inference. https://doi.org/10.1111/j.1420-

9101.2010.02210.x

Bates, D., Mächler, M., Bolker, B., & Walker, S. (2015). Fitting Linear Mixed-Effects

Models Using {lme4}. Journal of Statistical Software, 67(1), 1–48.

https://doi.org/10.18637/jss.v067.i01

Bolker, B. (2014). Convergence error for development version of lme4. Retrieved

October 30, 2017, from

https://stackoverflow.com/questions/21344555/convergence-error-for-

development-version-of-lme4

Bradlow, A., Clopper, C., Smiljanic, R., & Walter, M. A. (2010). A perceptual phonetic

similarity space for languages: Evidence from five native language listener

groups. Speech Communication, 52(11–12), 930–942.

https://doi.org/10.1016/j.specom.2010.06.003

Brown, J. D. (1980). Relative Merits of Four Methods for Scoring Cloze Tests. The

Modern Language Journal, 64(3), 311–317. https://doi.org/10.1111/j.1540-

4781.1980.tb05198.x

Brysbaert, M., & Stevens, M. (2018). Power Analysis and Effect Size in Mixed Effects

Models: A Tutorial. Journal of Cognition, 1(1)(9), 1–20.

https://doi.org/10.5334/joc.10
 141

EFFECTS OF CROSSLINGUISTIC OVERLAP

Brysbaert, Marc. (2003). Bilingual visual word recognition: Evidence from masked

phonological priming. In Masked Priming: The State of the Art (pp. 178–188).

https://doi.org/10.4324/9780203502846

Brysbaert, Marc, & Duyck, W. (2010). Is it time to leave behind the Revised

Hierarchical Model of bilingual language processing after fifteen years of

service? Bilingualism, 13(3), 359–371. https://doi.org/10.1017/S1366728909990344

Brysbaert, Marc, Van Dyck, G., & Van De Poel, M. (1999). Visual word recognition in

bilinguals: Evidence from masked phonological priming. Journal of Experimental

Psychology: Human Perception and Performance. https://doi.org/10.1037/0096-

1523.25.1.137

Caramazza, A. (1997). How Many Levels of Processing Are There in Lexical Access?

Cognitive Neuropsychology, 14(1), 177–208.

https://doi.org/10.1080/026432997381664

Caramazza, A., & Brones, I. (1979). Lexical access in bilinguals. Bulletin of the

Psychonomic Society, 13(4), 212–214. https://doi.org/10.3758/BF03335062

Chen, H. C., & Leung, Y. S. (1989). Patterns of Lexical Processing in a Nonnative

Language. Journal of Experimental Psychology: Learning, Memory, and

Cognition. https://doi.org/10.1037/0278-7393.15.2.316

Chomsky, N., & Halle, M. (1992). The Sound Pattern of English. PsycCRITIQUES.

Cambridge, MA.: MIT Press.

Clements, G., & Hume, E. V. (1995). The internal organization of speech sounds. In J.
 142

EFFECTS OF CROSSLINGUISTIC OVERLAP

Goldsmith (Ed.), The handbook of phonological theory (pp. 245–306).

Cambridge, MA.

Clements, G. N. (1985). The geometry of phonological features. Phonology, 2(1985),

225. https://doi.org/10.1017/S0952675700000440

Collins, A. M., & Loftus, E. F. (1975). A Spreading-Activation Theory of Semantic

Processing. Psychological Review, 82(6), 407–428. https://doi.org/10.1037/0033-

295X.82.6.407

Collins, A. M., & Quillian, M. R. (1969). Retrieval time from semantic memory.

Journal of Verbal Learning and Behavior, 8, 240–247.

https://doi.org/10.1016/S0022-5371(69)80069-1

Colombo, L. (1986). Activation and Inhibition With Orthographically Similar Words.

Journal of Experimental Psychology: Human Perception and Performance.

https://doi.org/10.1037/0096-1523.12.2.226

Comesaña, M., Sánchez-Casas, R., Soares, A. P., Pinheiro, A. P., Rauber, A., Frade, S.,

& Fraga, I. (2012). The interplay of phonology and orthography in visual cognate

word recognition: An ERP study. Neuroscience Letters, 529(1), 75–79.

https://doi.org/10.1016/j.neulet.2012.09.010

Costa, A., Caramazza, A., & Sebastian-Galles, N. (2000). The Cognate Facilitation

Effect: Implications for Models of Lexical Access. Journal of Experimental

Psychology: Learning Memory and Cognition. https://doi.org/10.1037/0278-

7393.26.5.1283
 143

EFFECTS OF CROSSLINGUISTIC OVERLAP

Costa, A., La Heij, W., & Navarrete, E. (2006). The dynamics of bilingual lexical

access. Bilingualism, 9(2), 137–151. https://doi.org/10.1017/S1366728906002495

Costa, A., & Santesteban, M. (2004). Lexical access in bilingual speech production:

Evidence from language switching in highly proficient bilinguals and L2

learners. Journal of Memory and Language, 50, 491–511.

https://doi.org/10.1016/j.jml.2004.02.002

Cristoffanini, P., Kirsner, K., & Milech, D. (1986). Bilingual lexical representation: The

status of spanish-english cognates. The Quarterly Journal of Experimental

Psychology Section A, 38(3), 367–393.

https://doi.org/10.1080/14640748608401604

da Luz Fontes, A. B. A., & Schwartz, A. I. (2010). On a different plane: Cross-language

effects on the conceptual representations of within-language homonyms.

Language and Cognitive Processes, 25(4), 508–532.

https://doi.org/10.1080/01690960903285797

Davies, M. (2017a). Corpus of Contemporary American English. Retrieved from

https://corpus.byu.edu/coca/

Davies, M. (2017b). O corpus do Português. Retrieved from

http://www.corpusdoportugues.org/x.asp

de Groot, A. M. B. (1992). Determinants of Word Translation. Journal of Experimental

Psychology: Learning, Memory, and Cognition. https://doi.org/10.1037/0278-

7393.18.5.1001
 144

EFFECTS OF CROSSLINGUISTIC OVERLAP

Dell, G. S. (1986). A Spreading-Activation Theory of Retrieval in Sentence

Production. Psychological Review, 93(3), 283–321. https://doi.org/10.1037/0033-

295X.93.3.283

Dijkstra, T., Grainger, J., & Van Heuven, W. J. B. (1999). Recognition of Cognates and

Interlingual Homographs: The Neglected Role of Phonology. Journal of Memory

and Language, 41(4), 496–518. https://doi.org/10.1006/jmla.1999.2654

Dijkstra, T., Miwa, K., Brummelhuis, B., Sappelli, M., & Baayen, H. (2010). How cross-

language similarity and task demands affect cognate recognition. Journal of

Memory and Language, 62(3), 284–301. https://doi.org/10.1016/j.jml.2009.12.003

Dijkstra, T., & van Heuven, W. J. B. (1998). The BIA model and bilingual word

recognition. Localist Connect. Approaches to Hum. Cogn., 189–225.

Dijkstra, T., & van Heuven, W. J. B. (2002). The architecture of the bilingual word

recognition system: From identification to decision. Bilingualism: Language and

Cognition, 5(3), 175–197. https://doi.org/10.1017/s1366728902003012

Dijkstra, T., Wah, BUYTENHUIJS, F., VAN HALEM, N., AL-JIBOURI, Z., DE KORTE,

M., & REKKÉ, S. (2018). Multilink: a computational model for bilingual word

recognition and word translation. Bilingualism: Language and Cognition, 22(4),

657–679. https://doi.org/10.1017/S1366728918000287

Dimitropoulou, M., Duñabeitia, J. A., & Carreiras, M. (2011a). Masked translation

priming effects with low proficient bilinguals. Memory and Cognition, 39(2),

260–275. https://doi.org/10.3758/s13421-010-0004-9
 145

EFFECTS OF CROSSLINGUISTIC OVERLAP

Dimitropoulou, M., Duñabeitia, J. A., & Carreiras, M. (2011b). Phonology by itself:

Masked phonological priming effects with and without orthographic overlap.

Journal of Cognitive Psychology, 23(2), 185–203.

https://doi.org/10.1080/20445911.2011.477811

Dimitropoulou, M., Duñabeitia, J. A., & Carreiras, M. (2011c). Two words, one

meaning: Evidence of automatic co-activation of translation equivalents.

Frontiers in Psychology, 2(AUG), 1–20. https://doi.org/10.3389/fpsyg.2011.00188

Duñabeitia, J. A., Dimitropoulou, M., Uribe-Etxebarria, O., Laka, I., & Carreiras, M.

(2010). Electrophysiological correlates of the masked translation priming effect

with highly proficient simultaneous bilinguals. Brain Research, 1359, 142–154.

https://doi.org/10.1016/j.brainres.2010.08.066

Duñabeitia, J. A., Perea, M., & Carreiras, M. (2010). Masked translation priming

effects with highly proficient simultaneous bilinguals. Experimental Psychology,

57(2), 98–107. https://doi.org/10.1027/1618-3169/a000013

Dunabeitia, J. A., Perea, M., Carreiras, M., Duñabeitia, J. A., Perea, M., & Carreiras, M.

(2010). Masked translation priming effects with highly proficient simultaneous

bilinguals. Experimental Psychology, 57(2), 98–107. https://doi.org/10.1027/1618-

3169/a000013

Duyck, W., Van Assche, E., Drieghe, D., & Hartsuiker, R. J. (2007). Visual Word

Recognition by Bilinguals in a Sentence Context: Evidence for Nonselective

Lexical Access. Journal of Experimental Psychology: Learning Memory and

 146

EFFECTS OF CROSSLINGUISTIC OVERLAP

Cognition, 33(4), 663–679. https://doi.org/10.1037/0278-7393.33.4.663

Ferrand, L., & Grainger, J. (1993). The time course of orthographic and phonological

code activation in the early phases of visual word recognition. Bulletin of the

Psychonomic Society. https://doi.org/10.3758/BF03334157

Field, A., Miles, J., & Field, Z. (2012). Discovering Statistics Using R. London, England:

SAGE Publications Ltd.

Finkbeiner, M., Forster, K., Nicol, J., & Nakamura, K. (2004). The role of polysemy in

masked semantic and translation priming. Journal of Memory and Language,

51(1), 1–22. https://doi.org/10.1016/j.jml.2004.01.004

Forster, K. I., & Davis, C. (1984). Repetition priming and frequency attenuation in

lexical access. Journal of Experimental Psychology: Learning, Memory, and

Cognition. https://doi.org/10.1037/0278-7393.10.4.680

Forster, K. I., Mohan, K., & Hector, J. (2003). The mechanics of masked priming. In

Masked Priming: The State of the Art (pp. 2–20). Psychology Press.

https://doi.org/10.4324/9780203502846

Friesen, D. C., & Jared, D. (2012). Cross-language phonological activation of meaning:

Evidence from category verification. Bilingualism, 15(1), 145–156.

https://doi.org/10.1017/S1366728910000489

Gerard, Linda, & Scarborough, D. (1989). Language-Specific Lexical Access of

Homographs by Bilinguals - Tags: HOMONYMS SEMANTICS. Journal of

Experimental Psychology. Learning; Memory & Cognition, 15(2), 305–315.

 147

EFFECTS OF CROSSLINGUISTIC OVERLAP

Retrieved from http://connection.ebscohost.com/c/articles/14387219/language-

specific-lexical-access-homographs-by-bilinguals

Gollan, T. H., Forster, K. I., & Frost, R. (1997). crucial for Translation Priming With

Different Scripts : Masked Priming With providing and Noncognates in

Bilinguals the effect of. Journal of Experimental Psychology: Learning, Memory,

and Cognition, 23(5), 1122–1139. Retrieved from

http://content.ebscohost.com/ContentServer.asp?T=P&P=AN&K=1997-06939-

004&S=L&D=pdh&EbscoContent=dGJyMMvl7ESeqa44v%2BbwOLCmr1Cep7NS

sKu4TLCWxWXS&ContentCustomer=dGJyMPGot1CzrrZLuePfgeyx44Dt6fIA

Grainger, J., & Frenck-Mestre, C. (2002). Masked Priming by Translation Equivalents

in Proficient Bilinguals. Language and Cognitive Processes, 13(6), 601–623.

https://doi.org/10.1080/016909698386393

Grainger, J., & Holcomb, P. J. (2009). Watching the word go by: On the time-course of

component processes in visual word recognition. Linguistics and Language

Compass. https://doi.org/10.1111/j.1749-818X.2008.00121.x

Grainger, J., & Jacobs, A. M. (1996). Orthographic Processing in Visual Word

Recognition: A Multiple Read-Out Model Word recognition is an integral

component of language pro- cessing in general. It raises a fundamental question

about the. Psychological Review Carr & Pollatsek, 103(3), 518–565.

Grainger, J., Kiyonaga, K., & Holcomb, P. J. (2006). The time course of orthographic

and phonological code activation. Psychological Science.

 148

EFFECTS OF CROSSLINGUISTIC OVERLAP

https://doi.org/10.1111/j.1467-9280.2006.01821.x

Grainger, J., Spinelli, E., Farioli, F., Diependaele, K., & Ferrand, L. (2003). Masked

Repetition and Phonological Priming within and across Modalities. Journal of

Experimental Psychology: Learning Memory and Cognition, 29(6), 1256–1269.

https://doi.org/10.1037/0278-7393.29.6.1256

Guasch, M., Boada, R., Ferré, P., & Sánchez-Casas, R. (2013). NIM: A Web-based Swiss

army knife to select stimuli for psycholinguistic studies. Behavior Research

Methods, 45(3), 765–771. https://doi.org/10.3758/s13428-012-0296-8

Hall, K. C., Allen, B., Fry, M., Mackie, S., & McAuliffe, M. (2016). Phonological

CorpusTools: A free, open-source tool for phonological analysis. 14th Conference

for Laboratory Phonology. Retrieved from

http://corpustools.readthedocs.io/en/latest/index.html

Hutchison, K. A. (2007). Attentional Control and the Relatedness Proportion Effect

in Semantic Priming. Journal of Experimental Psychology: Learning Memory and

Cognition. https://doi.org/10.1037/0278-7393.33.4.645

IBM SPSS Statistics 25. (2018). Armonk, NY: IBM Corp.

Jared, D., & Szucs, C. (2002). Phonological activation in bilinguals: Evidence from

interlingual homograph naming. Bilingualism: Language and Cognition, 5(03).

https://doi.org/10.1017/S1366728902003024

Jiang, N. (1999). Testing processing explanations for the asymmetry in masked cross-

language priming. Bilingualism: Language and Cognition, 2(1), 59–75.

 149

EFFECTS OF CROSSLINGUISTIC OVERLAP

https://doi.org/10.1017/s1366728999000152

Johnson, K. (2011). Acoustic and Auditory Phonetics (3rd ed.). Malden, MA: Wiley-

Blackwell.

Johnson, P. C. D. (2014). Extension of Nakagawa & Schielzeth’s R2GLMMto random

slopes models. Methods in Ecology and Evolution, 5(9), 944–946.

https://doi.org/10.1111/2041-210X.12225

Keuleers, E., & Brysbaert, M. (2010). Wuggy: A multilingual pseudoword generator.

Behavior Research Methods, 42(3), 627–633.

https://doi.org/10.3758/BRM.42.3.627

Keuleers, E., Diependaele, K., & Brysbaert, M. (2010). Practice effects in large-scale

visual word recognition studies: A lexical decision study on 14,000 dutch mono-

and disyllabic words and nonwords. Frontiers in Psychology, 1(NOV), 1–15.

https://doi.org/10.3389/fpsyg.2010.00174

Kim, J., & Davis, C. (2003). Task effects in masked cross-script translation and

phonological priming. Journal of Memory and Language, 49(4), 484–499.

https://doi.org/10.1016/S0749-596X(03)00093-7

Kondrak, G., & Sherif, T. (2006). Evaluation of several phonetic similarity algorithms

on the task of cognate identification. Proceedings of the Workshop on Linguistic

Distances, 43–50. https://doi.org/10.3115/1641976.1641983

Kroll, J., & Curley, J. (1988). Lexical memory in novice bilinguals: The role of concepts

in retrieving second language words. In Practical Aspects of Memory.

 150

EFFECTS OF CROSSLINGUISTIC OVERLAP

Kroll, J. F., & Stewart, E. (1994). Category Interference in Translation and Picture

Naming: Evidence for Asymmetric Connections Between Bilingual Memory

Representations. Journal of Memory and Language, 33(2), 149–174.

https://doi.org/10.1006/jmla.1994.1008

Kroll, J. F., & Tokowicz, N. (2001). The development of conceptual representation for

words in a second language. One Mind, Two Lang. Biling. Lang. Process., 49–71.

Lemhöfer, K., & Dijkstra, T. (2004). Recognizing cognates and interlingual

homographs: Effects of code similarity in language-specific and generalized

lexical decision. Memory and Cognition, 32(4), 533–550.

https://doi.org/10.3758/BF03195845

Levelt, W. J. M. (1993). Speaking: From Intention to Articulation (ACL-MIT Series in

Natural Language Processing). Cambridge, MA: MIT Press.

Levelt, W. J. M. (2001). Spoken word production: a theory of lexical access. In

Proceedings of the National Academy of Sciences (Vol. 98 (23), pp. 13464–13471).

Nijmegen, The Netherlands. https://doi.org/10.1073/pnas.231459498

Levelt, W. J. M., Roelofs, A., & Meyer, A. S. (1999). A theory of lexical access in speech

production. Behavioral and Brain Sciences, 22(1), 1–75.

https://doi.org/10.1017/S0140525X99001776

Levenshtein, V. I. (1966). Binary codes capable of correcting deletions, insertions, and

reversals. Soviet Physics Doklady. https://doi.org/citeulike-article-id:311174

Lima, S. D., & Huntsman, L. A. (1997). Sequential dependencies in the lexical

 151

EFFECTS OF CROSSLINGUISTIC OVERLAP

decision task. Psychological Research, 60(4), 264–269.

https://doi.org/10.1007/BF00419412

Linck, J. A., & Cunnings, I. (2015). The Utility and Application of Mixed-Effects

Models in Second Language Research. Language Learning, 65(S1), 185–207.

https://doi.org/10.1111/lang.12117

Litz, D. R., & Smith, A. K. (2006). Semantically Acceptable Scoring Procedures (

SEMAC ) Versus Exact Replacement Scoring Methods ( ERS ) For ’ Cloze ’ Tests :

A Case Study. Asian EFL Journal Quaterly, 8(1), 1–28. Retrieved from

https://www.asian-efl-journal.com/1133/quarterly-journal/semantically-

acceptable-scoring-procedures-semac-versus-exact-replacement-scoring-

methods-ers-for-cloze-tests-a-case-study/#squelch-taas-tab-content-0-0

Lupker, S. J., Brown, P., & Colombo, L. (1997). Strategic Control in a Naming Task.

Journal of Experimental Psychology: Learning, Memory, and Cognition, 23(3),

570–590.

Martin, R. C., & Jensen, C. R. (1988). Phonological priming in the lexical decision

task: A failure to replicate. Memory & Cognition.

https://doi.org/10.3758/BF03197052

McClelland, J. L., & Rumelhart, D. E. (1981). An interactive activation model of

context effects in letter perception: I. An account of basic findings. Psychological

Review. https://doi.org/10.1037/0033-295X.88.5.375

Mielke, J. (2012). A phonetically based metric of sound similarity. Lingua, 122(2), 145–
 152

EFFECTS OF CROSSLINGUISTIC OVERLAP

163. https://doi.org/10.1016/j.lingua.2011.04.006

Montrul, S. A. (n.d.). Incomplete Acquisition in Bilingualism: Re-Examining the Age

Factor. Philadelphia, PA, USA: John Benjamins Publishing Co.

Motley, M. T., & Camden, C. T. (1985). Nonlinguistic influences on lexical selection:

Evidence from double entendres. Communications Monographs, 52(2), 124–135.

https://doi.org/10.1080/03637758509376100

Nakagawa, S., & Schielzeth, H. (2013). A general and simple method for obtaining R2

from generalized linear mixed-effects models. Methods in Ecology and Evolution,

4(2), 133–142. https://doi.org/10.1111/j.2041-210x.2012.00261.x

Nakayama, M., Ida, K., & Lupker, S. J. (2016). Cross-script L2-L1 noncognate

translation priming in lexical decision depends on L2 proficiency: Evidence from

Japanese-English bilinguals. Bilingualism, 19(5), 1001–1022.

https://doi.org/10.1017/S1366728915000462

Nakayama, M., Sears, C. R., Hino, Y., & Lupker, S. J. (2012). Cross-script phonological

priming for Japanese-English bilinguals: Evidence for integrated phonological

representations. Language and Cognitive Processes, 27(10), 1563–1583.

https://doi.org/10.1080/01690965.2011.606669

Nakayama, M., Sears, C. R., Hino, Y., & Lupker, S. J. (2013). Masked translation

priming with Japanese-English bilinguals: Interactions between cognate status,

target frequency and L2 proficiency. European Journal of Cognitive Psychology,

25(8), 949–981. https://doi.org/10.1080/20445911.2013.839560

 153

EFFECTS OF CROSSLINGUISTIC OVERLAP

Nash, J. C. (2014). On Best Practice Optimization Methods in {R}. Journal of

Statistical Software, 60(2), 1–14. Retrieved from

http://www.jstatsoft.org/v60/i02/

Pallier, C., Colomé, A., & Sebastián-Gallés, N. (2001). The Influence of Native-

Language Phonology on Lexical Access: Exemplar-Based Versus Abstract Lexical

Entries. Psychological Science, 12(6), 445–449. https://doi.org/10.1111/1467-

9280.00383

Peeters, D., Dijkstra, T., & Grainger, J. (2013). The representation and processing of

identical cognates by late bilinguals: RT and ERP effects. Journal of Memory and

Language, 68(4), 315–332. https://doi.org/10.1016/j.jml.2012.12.003

Perea, M., Duñabeitia, J. A., & Carreiras, M. (2008). Masked associative/semantic

priming effects across languages with highly proficient bilinguals. Journal of

Memory and Language, 58(4), 916–930. https://doi.org/10.1016/j.jml.2008.01.003

Perfetti, C. A., & Bell, L. (1991). Phonemic activation during the first 40 ms of word

identification: Evidence from backward masking and priming. Journal of

Memory and Language. https://doi.org/10.1016/0749-596X(91)90017-E

Perfetti, C. A., & Tan, L. H. (1998). The time course of graphic, phonological, and

semantic activation in Chinese character identification. Journal of Experimental

Psychology: Learning Memory and Cognition, 24(1), 101–118.

https://doi.org/10.1037/0278-7393.24.1.101

Pexman, P. M., Lupker, S. J., & Jared, D. (2001). Homophone Effects in Lexical
 154

EFFECTS OF CROSSLINGUISTIC OVERLAP

Decision. Journal of Experimental Psychology: Learning Memory and Cognition,

27(1), 139–156. https://doi.org/10.1037//0278-7393.27.1.139

Pollatsek, A., Perea, M., & Carreiras, M. (2005). Does conal prime CANAL more than

cinal? Masked phonological priming effects in Spanish with the lexical decision

task. Memory and Cognition, 33(3), 557–565. https://doi.org/10.3758/BF03193071

Potter, M. C., So, K. F., Eckardt, B. Von, & Feldman, L. B. (1984). Lexical and

conceptual representation in beginning and proficient bilinguals. Journal of

Verbal Learning and Verbal Behavior. https://doi.org/10.1016/S0022-

5371(84)90489-4

Presentation®: Precise, Powerful Stimulus Delivery. (2017). Berkeley, CA:

Neurobehavioral Systems, Inc. Retrieved from https://www.neurobs.com

R Core Team. (2019). R: A Language and Environment for Statistical Computing.

Vienna, Austria. https://doi.org/10.1007/978-3-540-74686-7

Rastle, K., Davis, M. H., Marslen-Wilson, W. D., & Tyler, L. K. (2000). Morphological

and semantic effects in visual word recognition: A time-course study. Language

and Cognitive Processes. https://doi.org/10.1080/01690960050119689

RStudio Team. (2018). RStudio: Integrated Development Environment for R. Boston,

MA. Retrieved from http://www.rstudio.com/

Rumelhart, D. E., & McClelland, J. L. (1982). An interactive activation model of

context effects in letter perception: II. The contextual enhancement effect and

some tests and extensions of the model. Psychological Review.

 155

EFFECTS OF CROSSLINGUISTIC OVERLAP

https://doi.org/10.1037/0033-295X.89.1.60

Sabourin, L., Brien, C., & Burkholder, M. (2014). The effect of age of L2 acquisition on

the organization of the bilingual lexicon: Evidence from masked priming.

Bilingualism: Language and Cognition, 17(3), 542–555.

https://doi.org/10.1017/S1366728913000643

Samani, R., & Sharifian, F. (1997). Cross-language hierarchical spreading of

activation. In F. Sharifian (Ed.), Conference on Language, Cognition, and

Interpretation (pp. 11–23). Isfahan: IAU Press. Retrieved from

http://cogprints.org/1663/00/act2.htm

Sanchez-Casas, R. M., Davis, C. W., & Garcia-Albea, J. E. (1992). Bilingual Lexical

Processing: Exploring the Cognate/ Non-cognate Distinction. European Journal

of Cognitive Psychology, 4(4), 293–310.

https://doi.org/10.1080/09541449208406189

Schepens, J., Dijkstra, T., Grootjen, F., & van Heuven, W. J. B. (2013). Cross-Language

Distributions of High Frequency and Phonetically Similar Cognates. PLoS ONE,

8(5). https://doi.org/10.1371/journal.pone.0063006

Schwartz, A. I., & Kroll, J. F. (2006). Bilingual lexical activation in sentence context.

Journal of Memory and Language, 55(2), 197–212.

https://doi.org/10.1016/j.jml.2006.03.004

Schwartz, A. I., Kroll, J. F., & Diaz, M. (2007). Reading words in Spanish and English:

Mapping orthography to phonology in two languages. Language and Cognitive

 156

EFFECTS OF CROSSLINGUISTIC OVERLAP

Processes, 22(1), 106–129. https://doi.org/10.1080/01690960500463920

Shinde, R. B., & Pawar, V. P. (2014). Dynamic time Warping using MATLAB & PRAAT.

International Journal of Scientific & Engineering Research, 5(5), 37–42.

Soares, C., & Grosjean, F. (1984). Bilinguals in a monolingual and a bilingual speech

mode: The effect on lexical access. Memory & Cognition, 12(4), 380–386.

https://doi.org/10.3758/BF03198298

Sunderman, G., & Schwartz, A. I. (2008). Using Cognates to Investigate Cross-

Language Competition in Second Language Processing. TESOL Quarterly, 42(3),

527–536. https://doi.org/10.1002/j.1545-7249.2008.tb00145.x

Taylor, T. E., & Lupker, S. J. (2001). Sequential Effects in Naming: A Time-Criterion

Account. Journal of Experimental Psychology: Learning Memory and Cognition,

27(I), 117–138.

The World’s Top 10 Largest Economies. (n.d.). Retrieved October 1, 2019, from

https://www.focus-economics.com/blog/the-largest-economies-in-the-world

Tzur, B., & Frost, R. (2007). SOA does not Reveal the Absolute Time Course of

Cognitive Processing in Fast Priming Experiments. J Mem Lang (Vol. 56).

van Hell, J. G., & de Groot, A. M. B. (1998). Conceptual representation in bilingual

memory: Effects of concreteness and cognate status in word association.

Bilingualism: Language and Cognition, 1(3), 193–211.

https://doi.org/10.1017/s1366728998000352

Van Hell, J. G., & Dijkstra, T. (2002). Foreign language knowledge can influence
 157

EFFECTS OF CROSSLINGUISTIC OVERLAP

native language performance in exclusively native contexts. Psychonomic

Bulletin and Review. https://doi.org/10.3758/BF03196335

Van Heuven, W. J. B., Dijkstra, T., & Grainger, J. (1998). Orthographic Neighborhood

Effects in Bilingual Word Recognition. Journal of Memory and Language, 39(3),

458–483. https://doi.org/10.1006/jmla.1998.2584

van Orden, G. C. (1987). A ROWS is a ROSE: Spelling, sound, and reading. Memory &

Cognition, 15(3), 181–198. https://doi.org/10.3758/BF03197716

Van Wijnendaele, I., & Brysbaert, M. (2002). Visual word recognition in bilinguals:

Phonological priming from the second to the first language. Journal of

Experimental Psychology: Human Perception and Performance.

https://doi.org/10.1037/0096-1523.28.3.616

Voga, M., & Grainger, J. (2007). Cognate status and cross-script translation priming.

Memory & Cognition, 35(5), 938–952.

von Studnitz, R. E., & Green, D. W. (1997). Lexical decision and language switching.

International Journal of Bilingualism, 1(1), 3–24.

https://doi.org/10.1177/136700699700100102

Wang, X. (2013). Language dominance in translation priming: Evidence from

balanced and unbalanced Chinese-English bilinguals. Quarterly Journal of

Experimental Psychology, 66(4), 727–743.

https://doi.org/10.1080/17470218.2012.716072

Weber, R. (1970). A Linguistic Analysis of First-Grade Reading Errors. Reading

 158

EFFECTS OF CROSSLINGUISTIC OVERLAP

Research Quarterly, 5(3), 427–451. Retrieved from

http://www.jstor.org/stable/747079

Wen, Y., & van Heuven, W. J. B. (2017). Non-cognate translation priming in masked

priming lexical decision experiments: A meta-analysis. Psychonomic Bulletin and

Review, 24(3), 879–886. https://doi.org/10.3758/s13423-016-1151-1

Wickham, H. (2016). ggplot2: Elegant Graphics for Data Analysis. Springer-Verlag

New York. Retrieved from http://ggplot2.org

Xia, V., & Andrews, S. (2015). Masked translation priming asymmetry in Chinese-

English bilinguals: Making sense of the Sense Model. Quarterly Journal of

Experimental Psychology, 68(2), 294–325.

https://doi.org/10.1080/17470218.2014.944195

Ziegler, J. C., Ferrand, L., Jacobs, A. M., Rey, A., & Grainger, J. (2000). Visual and

phonological codes in letter and word recognition: Evidence from incremental

priming. Quarterly Journal of Experimental Psychology Section A: Human

Experimental Psychology. https://doi.org/10.1080/713755906

 159

EFFECTS OF CROSSLINGUISTIC OVERLAP

8. Appendix

Appendix A. Research Ethics Approval Forms

FileNun-1>•:09-12-13
~~,I;.
~
...
~~
••
~;:.
,I
Dale <,nniddiY»')'):02/I0/JJl5

Universite d 'Ottawa
Bureau d'ethique et d'integrite de la recherche
University of Ottawa
Off,ce of ResearchEthics and Integrity

Ethics Approval Notice

Social Sciences and Humanities REB

Principal Investigator/ Supervisor/ Co-investigator(s) / Student(s)

Fa-st Name Last Name

Juana MWlOz-Liceras Arts/ Modem Languages and Literatures Supetvisor

Marc Brunelle Arts /Linguistics Co-Supetv:isor

Leonardo Alves-Soares Arts/ Linguistics Studem Researcher

FileNumber: 09-12-13

Type of Project: PhD Thesis

Title: Investigatingthe Effects Cross-Language Activation during Cross-Modal Language Processing

Renewal Date (mm/dd/yyyy) Expiry Date (mm/dd/yyyy) Approval Type

01,22/2015 01,21/2016 Ia
(la: Approval,lb: Approval fur initial stage only)

Special Conditions/ Comments:

NIA

550,rue Cumberland,piece 154 550CumberlandStreet, mom 154

Ottawa (Ontario)Kl N 6N5 Canada Ottawa,Ontario KIN 6N5 Canada
(613) 562-5387• Telec./Fax(613) 562-5338
www.recherche.uottawa.ca/deontologie/ www.research.uottawa.ca/ethicsl
 160

EFFECTS OF CROSSLINGUISTIC OVERLAP

Filo Nun.nr: 09--12-13

i .-·,,.,.,.,
;,;in.,,.
... t~ .,,_
~~
Date (mm dd/yyyy) : 02/10/JJI 5

Universite d'Ottawa
Bureau d'ethique et d'integrite de Ia recherthe
University of Ottawa
Offa e R.esearthEthics
of Integrity
and

This is to confirm that the University of Ottawa Research Ethics Board identified above, which operates in
accordance with the Tri-Council Policy Statement (2010) and other applicable laws and regul aliens in Ontario,
has examined and approved the ethics application for the ab ave named research project . Ethics approval is
valid for the period indicated above and subject to the conditions listed in the section entitled "Special
Conditions/ Comments ".

During the course of the project, the protocol may not be modified without prior written approval from the
REB except when necessary to remove participants from immediate endangerment or when the modifica!ion(s)
pertain to only aclministra!ive or logistical components of the project (e.g., change of telephone number) .
Investigators must also promptly alert the REB of any changes which increase the risk to participant(s). any
changes which considerably affect the conduct of the project, all unanticipated and harmful events that occur,
and new information that may negatively affect the conduct of the project and safety of the participant(s)
Mo difica!ions to the project, including consent and recruitment documenta!ion, should be submitted to the
Ethics Office for approval using the "Modi fication to research project" form available at:
http :// re search .uo ttawa er/ethics/ submi s si on s-and-revi ews.

Please submit an annual rep art to the Ethics Office four weeks before the ab ave-referenced expiry date to
request a renewal of this ethics approval . To close the file, a final report must be submitted . The se documents
can be found at: http://re search .uottawa ca/ethics/submissions-and-reviews .

If you have any questions, please do not hesitate to contact the Ethics Office at extension 5387 or by e-mail at:
ethics@uOttawa .ca .

Si~ature:

Melanie Rioux
Ethics Coordinator
For Catherine Paquet, Director of the Office of Research Ethics and Integrity

2
550,rue Cwnberland,piece 154 550 CwnberlandStreet, JOom154
Ottawa(Ontario)KIN 6N5 Canada Ottawa,Ontario KIN 6N5 Canade
(613) 562-5387 • Telec./Fax (613) 562-5338
wwwu:cb1urbe
nnttawca/cleontnloeie/
WYCVrereamb
unttawca/ethics!
 161

EFFECTS OF CROSSLINGUISTIC OVERLAP

FileNuni>..-:09-12-13

"'
~1$,,,
~ ;i,.
Dale 4)nnidil/Y»')'): 04124/JJl 7

®~
Universite d 'Ottawa
Bureau d'ethique et d'integrite de la recherche
University of Ottawa
Off,ce of Research Ethics and Integrity

Ethics Approval Notice

Social Sciences and Humanities REB

Principal Investigator/ Supervisor/ Co-investigator(s) / Student(s)

Fn-stName Last Name Affiliation

Juana Munoz-Liceras Arts I Modem Languages and Literatures Supervisor

Marc Brunelle Arts /Linguistics Co-Supetv:isor

Leonardo Alves-Soares Arts I Linguistics Studem Researcher

File Number: 09-12-13

Type of Project: PhD Thesis

Title: Investigatingthe Effects Cross-Language Ac1iva1ion during Cross-Modal Language Processing

Renew al Date (mm/dd/yyyy) Expiry Date (mm/dd/yyyy) Approval Type

01,22/2016 01,21/2017 Renewal

Special Conditions/ Comments:

NIA

550, rue Cumberland, piece 154 550 Cumberland Street, room 154
Ottawa (Ontario) Kl N 6N5 Canada Ottawa, Ontario KIN 6N5 Canada
(613) 562-5387 • Telec./Fax (613) 562-5338
www .recherche.uottawa.caideontologie/www.reseaxch.uottawa.cafethicsl
 162

EFFECTS OF CROSSLINGUISTIC OVERLAP

FileNuni>..-:09-12-13

"'
~1$,,,
~ ;i,.
Dale 4)nnidil/Y»')'): 04124/JJl 7

®~
Universite d 'Ottawa
Bureau d'ethique et d'integrite de la recherche
University of Ottawa
Off,ce of Research Ethics and Integrity

Ethics Approval Notice

Social Sciences and Humanities REB

Principal Investigator/ Supervisor/ Co-investigator(s) / Student(s)

Fn-stName Last Name Affiliation

Juana Munoz-Liceras Arts I Modem Languages and Literatures Supervisor

Marc Brunelle Arts /Linguistics Co-Supetv:isor

Leonardo Alves-Soares Arts I Linguistics Studem Researcher

File Number: 09-12-13

Type of Project: PhD Thesis

Title: Investigatingthe Effects Cross-Language Ac1iva1ion during Cross-Modal Language Processing

Renew al Date (mm/dd/yyyy) Expiry Date (mm/dd/yyyy) Approval Type

01,22/2017 01,21/2018 Renewal

Special Conditions/ Comments:

NIA

550, rue Cumberland, piece 154 550 Cumberland Street, room 154
Ottawa (Ontario) Kl N 6N5 Canada Ottawa, Ontario KIN 6N5 Canada
(613) 562-5387 • Telec./Fax (613) 562-5338
www .recherche.uottawa.caideontologie/www.reseaxch.uottawa.cafethicsl
 163

EFFECTS OF CROSSLINGUISTIC OVERLAP

Filo Nun.nr: 09--12-13

i .-·,,.,.,.,
;,;in.,,.
... t~ .,,_
~~
Date (mm dd/yyyy) : 04f.l4/JJI 7

Universite d'Ottawa
Bureau d'ethique et d'integrite de Ia recherthe
University of Ottawa
Offa e of R.esearthEthics Integrity
and

This is to confirm that the University of Ottawa Research Ethics Board identi lied above, which operates in
accordance with the Tri-Council Policy Statement (2010) and other applicable laws and regul aliens in Ontario,
has examined and approved the ethics application for the ab ave named research project . Ethics approval is
valid for the period indicated above and subject to the conditions listed in the section entitled "Special
Conditions/ Comments ".

During the course of the project, the protocol may not be modified without prior written approval from the
REB except when necessary to remove participants from immediate endangerment or when the modifica!ion(s)
pertain to only adrninistra!ive or logistical components of the project (e.g., change of telephone number) .
Investigators must also promptly alert the REB of any changes which increase the risk to participant(s). any
changes which considerably affect the conduct of the project, all unanticipated and harmful events that occur,
and new information that may negatively affect the conduct of the project and safety of the participant(s)
Mo difica!ions to the project, including consent and recruitment documenta!ion, should be submitted to the
Ethics Office for approval using the "Modi fication to research project" form available at:
https.//research.uottawa.ca /ethics/farms .

Please submit an annual rep art to the Ethics Office four weeks before the ab ave-referenced expiry date to
request a renewal of this ethics approval . To close the file, a final report must be submitted. These documents
can be found at: https://research.uottawa.ca/ethics/forms .

If you have any questions, please do not hesitate to contact the Ethics Office at extension 5387 or by e-mail at:
ethics@uOttawa .ca.

Signature:

Melanie Rioux
Ethics Coordinator
For Catherine Paquet, Director of the Office of Research Ethics and Integrity

2
550,rue Gwnberland,piece 154 550 GwnberlandStreet, room 154
Ottawa(Ontario)KIN 6N5 Canada Ottawa,OntarioKIN 6N5Ganade
(613) 562-5387• Telec.iFax (613) 562-5338
www.recherthe.uottaIW.caideon1ologie/www.researth .uotta"" ·ca/e!hies/
 164

EFFECTS OF CROSSLINGUISTIC OVERLAP

Filo Nun.nr: 09--12-13

i .-·,,.,.,.,
;,;in.,,.
... t~ .,,_
~~
Date (mm dd/yyyy) : 04f.l4/JJI 7

Universite d'Ottawa
Bureau d'ethique et d'integrite de Ia recherthe
University of Ottawa
Offa e of R.esearthEthics Integrity
and

This is to confirm that the University of Ottawa Research Ethics Board identi lied above, which operates in
accordance with the Tri-Council Policy Statement (2010) and other applicable laws and regul aliens in Ontario,
has examined and approved the ethics application for the ab ave named research project . Ethics approval is
valid for the period indicated above and subject to the conditions listed in the section entitled "Special
Conditions/ Comments ".

During the course of the project, the protocol may not be modified without prior written approval from the
REB except when necessary to remove participants from immediate endangerment or when the modifica!ion(s)
pertain to only adrninistra!ive or logistical components of the project (e.g., change of telephone number) .
Investigators must also promptly alert the REB of any changes which increase the risk to participant(s). any
changes which considerably affect the conduct of the project, all unanticipated and harmful events that occur,
and new information that may negatively affect the conduct of the project and safety of the participant(s)
Mo difica!ions to the project, including consent and recruitment documenta!ion, should be submitted to the
Ethics Office for approval using the "Modi fication to research project" form available at:
https.//research.uottawa.ca /ethics/farms .

Please submit an annual rep art to the Ethics Office four weeks before the ab ave-referenced expiry date to
request a renewal of this ethics approval . To close the file, a final report must be submitted. These documents
can be found at: https://research.uottawa.ca/ethics/forms .

If you have any questions, please do not hesitate to contact the Ethics Office at extension 5387 or by e-mail at:
ethics@uOttawa .ca.

Signature:

Melanie Rioux
Ethics Coordinator
For Catherine Paquet, Director of the Office of Research Ethics and Integrity

2
550,rue Gwnberland,piece 154 550 GwnberlandStreet, room 154
Ottawa(Ontario)KIN 6N5 Canada Ottawa,OntarioKIN 6N5Ganade
(613) 562-5387• Telec.iFax (613) 562-5338
www.recherthe.uottaIW.caideon1ologie/www.researth .uotta"" ·ca/e!hies/
 165

EFFECTS OF CROSSLINGUISTIC OVERLAP

Appendix B. Participant Recruitment Form

INVITATION TO PARTICIPATE IN A LANGUAGE STUDY

Target Participants

► Portuguese · English bilinguals (all proficiency leve ls)

► Eng li sh · Portuguese bilinguals (all proficiency leve ls)
► English monolingual s

Description

You are being invited to participate in a Cross -Modal Language Processing study , which is
being conducted by L eonardo Alves -Soares, a Ph .D. student in the Department of Linguistics
at the Un iversity of Ottawa and sup ervised by professor Juana Mu.ii.oz·Liceras. The purpose of
the study is to furth er und erstand cross -language activation during the processing of words
with varying degrees of orthographic and phonetic overlap across the two languag es of a
bilingual. Th e study will take a approximate ly 90 minut es of your time , and it will take place
eith er in the Sound Patterns Laboratory , Simard Hall , room 333 (3 rd floor) , or in the Second
La nguag e Acquisition Research Laboratory , Arts Hall , room 253 , which are locat ed at the
Unive rsity of Ottawa main campus.

Your participation in the study is compl ete ly voluntary .

You will not r ece ive compensation for your participation in the study . However , if you wish , we
can provid e you with a cert ificat e of participation , which you can include with your CV.

Contact

If you are interested in participating in the study , or if you h ave any questions b efore deciding
whether or not to participate , pleas e contact the exp erim enter , Leonardo Alves -Soares
 166

EFFECTS OF CROSSLINGUISTIC OVERLAP

Appendix C: Informed Consent Forms

INFORMED CONSENT FORM

iJ
uOttawa
Exper imenter: Leonardo Alves-Soares

Contact Information : Department of Linguistics

University of Ottawa
Hamelin Hall , 70 Laurier Ave. E. , Room 415
Ottawa , ON KlN 6N5

Universit d'Onaw a
Facu l!desans
Unguisrique
Supervisors: Professor Juana Muiioz·Liceras
University of Ottaw a Department of Modern Languages and
F u~yolArt>
Department of Linguistics
Ungulst,cs University of Ottawa
Arts Hall, 70 Laurier Ave. East , room 217
Ottawa , ON K l N 6N5

Professor Marc Brunelle

Department of Linguistics
University of Ottawa
Arts Hall , 70 Laurier Avenue East , room 429

Invitation to Parti cipat e

You are invited to participate in a study titled Investigating the
Portuguese-English Bilingual Mental Lexicon : Effects of
Orthography and Phonology

Purpose of the Study

The purpose of the study is to investigate cross -langu age
activation during the processing of words across the two
langua ges of a bilingual .

Parti cipation
The following is the sequence of events that will take place during
the stud y.

(1) You will fill out a langu age background questionnaire.

U 613-56] ·5'86
W 61l-562 ·S141 (2) You will take an English proficiency test (cloze).
70LMlrierE.
Ottawa ON K1N6N5 Cll'lbda
(3) You will be asked to sit at a comp ut er workstation for the
www..u0tu,w.,.ca
experimental task. Your accuracy and reaction time will be
recorded for analysis.
 167

EFFECTS OF CROSSLINGUISTIC OVERLAP

[i] Risks
This study does not involve risk or deception.
uOttawa Confidentiality and Anonymity
Your name and contact information will remain strictly
confidential. Your data will be identified by code only. All data
linked to you will match the code previously assigned to you. The
data will not contain your name or any other biographical
Universit~ d'Ottawa information.
Faculte de> arts
Linguisrique Conservation of Data
University of Ottaw a
The data will be kept forever. All (digital ) data collected for the
facultyof Art~ study will be stored in the investigator's main computer, which is
UnguistJCS password ·protected.

The original consent forms as well as t heir copies will be stored

in secured filing cabinet in the Sound Patterns Laboratory (SMD
333). No one, besides the experimenter himself and /or his
supervisors , will have access to them.

Further Use of Data

All data collected in the study could be reanalyzed by other
resear chers to verify the validity of the results or compare them
with the results of other studies. In such a case, your data will be
transmitted anonymously, and all your personal information will
be kept confidential. Please indicate your prefer enc e by circling
one of the tw o options below.

You agree that your data be used in further analyses

You disagree that your data be used in further analyses.

Voluntary Participation
You are under no obligation to participate in the study. If you
choose to participate , you may withdraw at any t ime . If you
choose to withdraw, all data gathered up to the time of
withdrawal will be discarded. No negative consequences will
result from either
your refusal to participate or withdraw.

'U 6 B ·S61·5i86
Q 613---561·5 14 1

t
101.am<!r
Ottiwa ON KlN 6N S Clr'lada
WW\v .u()Uow11.u

2
 168

EFFECTS OF CROSSLINGUISTIC OVERLAP

[i] Acceptance
By signing below, you acknowledge that you understand the
terms, parameters and conditions of this research study
uOttawa conducted by Leonardo Alves-Soares, Ph.D. student (the
"Experimenter") in the Department of Linguistics at the
University of Ottawa. This study is under supervision of
Professor Juana Muiioz· Liceras and Professor Mar c Brunelle (the
((Supervisors").

Universit~ d'Ottawa If you have any questions about the study , you may contact
Faculte de> arts
either the Experimenter or the Supervisor using the contact
Linguisrique information provided in the beginning of this form.
University of Ottaw a Alternatively, you may also contact the Protocol Officer for
facultyof Art~ Ethics in Research, University of Ottawa, Tabaret Hall, 550
UnguistJCS Cumberland Street, Room 154, Ottawa, Ontario, KlN 6N 5, 613·
562·5387, or at ethics@uottawa. ca.

There are two copies of this consent form, one of which is yours
to keep.

Participant

Print Name Date

Signature

Experimenter

'U 6 B ·S61·5i86
Q 613---561·5 14 1

t
101.am<!r
Ottiwa ON KlN 6N S Clr'lada Print Name Signature Date
WW\v .u()Uow11.u

3
 169

EFFECTS OF CROSSLINGUISTIC OVERLAP

FORMULARIO DE CONSENTIMENTO

i] lnvestigador Principal: Leonardo Alves-S oares

(estudante de doutorado)
uOttawa Departarnento de Linguistica
Universidade de Ottawa

Contato: Departarnento de Linguistica

Universidade de Ottawa
Universite d'Ottawa Pavilhao Artes , Avenida Laurier Leste n° 70, 4° andar
f acu edes arts
linguisrique
Ottawa , ON KlN 6N5
Universityof Ottawa
fa Ul)'ofArb
Ungulst1<1

Professores Supervisores:
Professora Juana Muiioz·Liceras
Departmento de Linguas Modernas e
Departmento de Linguistica
Universidade de Ottawa
Pavilhao Artes
Avenida Laurier Leste n° 70, sala 2 17
Ottawa , ON KlN 6N5

Professor Marc Brunelle

Departmento de Lingui st ica
Univ ersidade de Ottawa
Pavilhao Artes
Avenida Laurier Leste n° 70, sala 429
Ottawa , ON KlN 6N 5

Convite para Participa,;;iio

Voce esta sendo convidado para participar no estudo intitulado
Inv estigating the Effects Cross-Language A ctivation during Cross ·
Modal Languag e Processing

Proposito do Estudo
0 prop6sito do estudo einv esti gar a ativru;ao simultan ea das duas
lin gu as de um bilingue durante processarnento de palavras
'D 61l-S6Hl86
W 61l-S6H14 1 cognatas .
70 l.awierE.
Ottawa ON Kl N 6NS Canada Parti cipa ,;;iio
www.u0ttow•.ca Durante o estudo, a seguinte cadeia de eventos ira acontecer.

(1) Voce preenchera urn questionario corn questoes sabre o seu background
linguistico.
 170

EFFECTS OF CROSSLINGUISTIC OVERLAP

2) Participante bilingues fariio dais testes de proficiencia , um na

[i] lingua nativa e o outro na segunda lingua.

(3) Assim que voce tenninar o teste de proficiencia, voce ira sentar-se em
u Ottawa ftente a uma workstation e fara uma tarefa sabre decisoes lexicas. A sua
acuidade e tempos de respostas serao medidos.

(4;) Ap6s o experimento, participantes bilingues seriio dados uma

lista de palavras em sua segunda lingua e eles teriio que traduzir
essas palavras para a sua lingua nativa.
Universit~d'Onawa,
Facultede, arts
Riscos
Linguisciqu
e
Este estudo niio envolve riscos au decep<;iio.
Universityof Ottawa
Facultycl Art~ Confidencialidade e Anonimato
Unguisucs 0 seu name e informa~6es pessoais permaneceriio estritamente
confidenciais. Seus dados seriio somente identificaveis par um
c6digo primeiramente associado a voce, e todos dados coletados
durante este estudo serao correlacionados com esse c6digo. Os
dados coletados nao conterao o seu nome ou sequer outra inform~ao
bibliografica sua.

Conserva()iio de Dados
Os dados coletados durante este estudo jamais serao destruidos.
Todos dados (digitais) coletados seriio arquivados no computador
pessoal do pesquisador principal, que e protegido par senha, e uma
c6pia sera armazenada num pen drive criptografado, que ser a
guardado no escrit6rio dos superv1sores no campus da
universidade.

Os formularios de consentimento e suas c6pias seriio arquivados

num arquivo no laborat6rio. Ninguem, a nao ser o pesquisador
principal e seus professores supervisores, teriio acesso a eles.

Uso dos Dados Posteriormente

Todos as dados coletados neste estudo poderao ser reanalisados
par outros pesquisadores a fim de que possam verificar a val i dade
dos resultados deste estudo au compara·los com resultados de
outros estudos. Em ta! caso, seus dados serao transmitidos
anonimamente, e todas suas inform~6es pessoais permanecerao
confidenciais. Par favor, indique sua preferencia circulando uma
das duas op<;6es abaixo.

Voce concorda que seus dados poderiio ser utilizados em analises

futuras.
'U 6 U·S61·5i&6
W 613---561·5141
Voce discorda que seus dados poderao ser utilizados em analises
70lawfert
Ottiwa ON K1N6NS Canada futuras.
WMv.uOl:Ulw•.ca

2
 171

EFFECTS OF CROSSLINGUISTIC OVERLAP

Participaf,iio Voluntaria

[i] Voce nao estii sob nenhwna obriga<;ao sequer de participar neste
estudo. Se voce decidir em participar, voce pode rescindir sua
participa<;ao em qualquer momenta. Se voce decidir rescindir sua
uOttawa participa9ao, todos os seus dados coletados serao descartados.
Nenhuma consequencia negativa sequer ocorrerii coma resultado
de voce recusar·se de parti cipar neste estudo ou rescindir sua
partici pa9ao

Universit~d'Onawa, Aceita
Facultede, arts Assinando abaixo, voce concorda que voce entende os termos,
Linguisciqu
e param.etros e condi96es deste estudo conduzido par Leonardo
Alves-Soares, estudante de doutorado (o "Investigador Principal")
University of Ottawa
Facultycl Art~ no Departamento de Linguistica na Universidade de Ottawa,
Unguisucs estudo supervisionado pelos pro fessores Juana Munoz·Li ceras e
Marc Brunelle (os "Pro fessores Supervisores").

Se voce tiver qualquer pergunta sabre este estudo, favor contatar o

investigador principal ou os professores supervisores us ando as
informa 96es de cantata providenciadas no come<;o deste
documento. Alternativamente, voce pode contatar o Oficial de
Protocolo de Etica em Pesquisas pessoalmente no pavilhao
Tabaret, rua Cumberland n ° 550, sala 154, Ottawa, Ontario, KI N
6N5, 613·562·5387, ou pelo email ethics @uottwa.ca .

Haduas c6pias deste formuliirio de consentimento, uma delas e

sua.

Participante

Nome Assinatura Data

Investigador Principal

Nome Assinatura Data

'U 6 U·S61·5i&6
W 613---561·5141
70lawfert
Ottiwa ON K1N6NS Canada
WMv.uOl:Ulw•.ca

3
 172

EFFECTS OF CROSSLINGUISTIC OVERLAP

Appendix D: Language Background Questionnaire

f- LanguageBackgro • *
• • I 61

Section 1 of 6

LanguageBackgroundQuestionnaire
Form description

After section 1 Continue to next section

Section 2 of 6 ,..
V

ParticipantInformation
Description (optional)

Participant Code *

This code will be provided to you by the researcher.

Short answer text

After section 2 Continue to next section

Section 3 of 6 ,..
V
 173

EFFECTS OF CROSSLINGUISTIC OVERLAP

PersonalInformation
Description (optional)

Month and year of blrth (MM-YYYY) *

Please enter tl1e month and the year you were born MM-YYYY

Sl1ort answer text

Place of blrth (clty - country) *

Please enter tl1e city and the country in which you were born.

Sl1ort answer text

Gender *

Please select your gender

1. Female

2. Male

3. other

After section 3 Continue to next section ...

Sect10n 4 of 6
 174

EFFECTS OF CROSSLINGUISTIC OVERLAP

Native LanguageInformation
Description (optional)

What ls your natlve or first language? *

0 Englisl1

Q Portuguese

Q otl1er...

Were you exposed to another language from blrth? *

0 YES

0 NO

After section 4 Continue to next section ...

y
Section 5 of 6 A

Other LanguageInformation
Description (optional)

What other language(s) were you exposed to from blrth? *

Sl1ort answer text

 175

EFFECTS OF CROSSLINGUISTIC OVERLAP

After section 5 Continue to next section ...

y
Section 6 of 6 A

Second Language Information

Description (optional)

What is your second language? *

Q Englisl1

Q Portuguese

Q I do not speak a second language

0 Otl1er...

At what age did you start learning your second language? *

Sl1ort answer text

 176

EFFECTS OF CROSSLINGUISTIC OVERLAP

Where dfd you learn your second language? *

Q elementary school

Q l1igl1school

Q college/university

Q Janguage scl10ol

Q family (parents orrelatives}

Q otl1er...

How often do you use your second language now? *

Q everyday

Q veryoften

Q often

Q sometimes

Q rarely

Q never

On the scale below, what ls your overall level of proflcfency fn your second *
language?

2 3 4 5 6 7 8 9 10

low proficiency Q Q Q Q O Q Q Q Q Q high proficiency

 177

EFFECTS OF CROSSLINGUISTIC OVERLAP

On the scale below, please rate your level of wrltlng proflclency ln your *
second language?

2 3 4 5 6 7 8 9 10

low proficiency Q Q Q Q Q Q Q Q Q Q high proficiency

On the scale below, please rate your level of readlng proflclency ln your *
second language?

2 3 4 5 6 7 8 9 10

low proficiency Q Q Q Q Q Q Q Q Q Q high proficiency

On the scale below, please rate your level of oral proflclency ln your second *
language?

2 3 4 5 6 7 8 9 10

low proficiency Q Q Q Q O Q Q Q Q Q high proficiency

 178

EFFECTS OF CROSSLINGUISTIC OVERLAP

Appendix E: Cloze Proficiency Test

Participant Code: _________ _

Cloze Test (Brown, 1980)

DIRECTIONS

1. Read the passage quickl y to get the general meaning.

2. Write only one word in eac h blank next to the it e m number.

3. Contractions such a s don't and John's bi cy cl e are one word.

4. Check your answers.

5. Spelling will not count against you as long as the words in the blanks are legible.

EXAMPLE
The boy walked up the street. He stepped on a piece of ice. He fell (1) down, but
h e didn't hurt himself.

MAN AND HIS PROGRESS

Man is th e only living creature that can mak e and us e tools. He is th e most

teachable of living b eings, earni ng the name of Homo sapiens. (l) ________ eve r

restless brain has used the 2)________ and the wisdom of his ancestors

(3)______ improve his way of life. Since (4)_______ is ab le to walk and

run (5)_______ his feet, his hands h ave always (6)_______ free to carry

and to use (7)________ . Man's h a nds have served him well (8) ______ _

his life on earth . His development, (9)_______ can be divided into three major

(10) _______ , is marked by seve r al different ways (11) _______ life .

Up to 10 ,000 years ago, (12) _______ human beings lived by huntin g and

(13) _______ . Th ey also pick e d be rri e s a nd fruits, (14 ) _______ dug for

various edib le roots. Most (15) _______ , th e m en wer e th e hunt er s, a nd

(16) ________ wom e n act e d as food ga th erers. Since (17) ________ wom e n

wer e busy with th e c hildr en, (18) ________ m e n handl e d th e tools. In a

p.1
 179

EFFECTS OF CROSSLINGUISTIC OVERLAP

( 19) _______ hand , a dead branch became a (20) _______ to knock

do wn fruit or to (21) _______ for tasty roots. Sometimes, an animal

(22) _______ s erv ed as a club, and a (23)_______ pi ece of ston e , fitting

comfortabl y into (24 ) _______ hand, could b e us ed to br eak (25) _____ or

to throw a t a n a nimal. (26 ) _______ ston e wa s chipped aga inst a noth er until

(27) _______ had a sharp edg e. Th e primitiv e (28) _______ who first

thought of putti ng a (29 ) _______ stone at the end of a (30) ______ _

mad e a brilliant discov ery : h e (31) join ed two thin gs to mak e a

(32) _______ us eful tool, the spear. Flint, found (33 ) _______ man y

rocks, became a common cutting (34) _______ in the Paleolithic period of man's

(35) _______ . Since no wood or bone tools (36 ) _______ survived , we

know of this man (37 )_______ his stone implements, with whi ch he

(38 ) _______ kill animals, cut up the meat, (39 ) _______ scra pe the

skins, as w ell as (40) ______ p ictur e s on the wal ls of th e (4 1) where

h e lived during the wint er.

(42) _______ t h e warmer s ea sons, man wandered on (43 ) ______ _

st epp es of Europ e with ou t a fixed (44) _______ , always for ag ing for food . Perha p s

the (45 )_______ carried nuts an d berries in shells (46 ) _____ skins or even

in light , woven (47 ) _______ . Wh erever they campe d, the primitive peop le

(48 ) _______ fir es by strik in g flin t for sparks (49 ) _______ using dri ed

seeds, moss , and rotten (50 ) _______ for tinder . With fir es that h e kin d led

hims elf, man could keep wild an imals away and co uld cook thos e that h e k illed, as well

as provide warmth and light for hims e lf.

p.2
 180

EFFECTS OF CROSSLINGUISTIC OVERLAP

Appendix F: Experimental Word Lists

C: cognate; F: false friend; D: distractor/pseudoword; P: practice; CTRL: control

Word List 1

prime TARGET Exp Status Item Nbr

abate PRESS CTRL 2
abrigo ADOPTIVE CTRL 3
acaraje ALARM CTRL 4
a<;ude ALCOHOL CTRL 6
adesivo ALIGNMENT CTRL 7
agulha ANGLE CTRL 8
aipo SIGNATURE CTRL 9
alagamento ASTUTE CTRL 11

alcatrao BAGGAGE CTRL 12

alecrim BAND CTRL 13
alho RIM CTRL 14
alvenaria BASE CTRL 16
amea<;a SMOKING CTRL 18
amigo BATTERY CTRL 19
anao BEIGE CTRL 20
anca BOMB CTRL 21
angu BRAVE CTRL 23
apoio CAPACITY CTRL 26
arnica CARGO CTRL 27
avestruz CAR CTRL 28
baba CHEQUE CTRL 30
baderna CHOCOLATE CTRL 31
bando COMA CTRL 34
baque COMATOSE CTRL 35
barata COMEDY CTRL 36
batente COMPUTER CTRL 38
ber<;o CONFIDENT CTRL 40
bilhete COMFORT CTRL 41
boi CONVENIENT CTRL 43
bolo CREAM CTRL 45
cama PULL CTRL 51
caminhao DETERGENT CTRL 53
capuz DIFFERENT CTRL 54
carne DISTRIBUTION CTRL 55
carrapato ELEVATOR CTRL 57
cartomante ENTRANCE CTRL 58
cebola EXCELLENT CTRL 60
cilio FATALITY CTRL 63
cisco FEVER CTRL 64
ciume FECES CTRL 65

Page 1 of 16
 181

EFFECTS OF CROSSLINGUISTIC OVERLAP

Word List 1
prime TARGET Exp Status Item Nbr
cochilo FLEXIBLE CTRL 66
congelador SYMPATHETIC CTRL 68
corneta FRUIT CTRL 71
couve GEL CTRL 73
criarn;:a GENIAL CTRL 75
cumulo HOUR CTRL 78
curral INGENUITY CTRL 79
dan o SORT CTRL 81
da rdo TAMP O N CTRL 82
debit INTERPR ETATION CTRL 83
delito LICENSE CTRL 86
demao LANGUAGE CTR L 87
demasia LINE CTRL 88
dente MANNE Q UIN CTRL 89
destrrn;:o ULTIM ATELY CTRL 92
eg ua RANGER CTRL 94
elo MINERAL CTRL 96
fantoche MOUSE CTRL 102
far da HASTE CTRL 103
fenda NAT UR E CTRL 105
frei SUM CTRL 110
freio TAX CTRL lll

furao TENANT CTRL 114

ga mba OCE AN CTRL 116
golpe OTHER CTRL 119
granada PARTICULAR CTRL 124
grava c;a o PIANO CTR L 126
ida VENT CTRL 132
isca PROPER CTRL 134
jabuti PSYCHOLOGY CTRL 135
jaca PSYCHOSIS CTRL 136
jarara ca PUMA CTRL 138
joa ninha RAT CTRL 142
la birinto REAL CTRL 145
laca REFEREN CE CTR L 146
ladra REMEDY CTRL 148
laguna REPAIR CTRL 149
laje REQU IREM ENT CTRL 150
lib elula ROSE CTRL 156
lima o RUDE CTRL 157

Page 2 ofi6
 182

EFFECTS OF CROSSLINGUISTIC OVERLAP

Word List 1
prime TARGET Exp Status Item Nbr
limo RUMOR CTRL 158
linho SAMBA CTRL 159
liquidificador SENTIMENT CTRL 160
lixo SIGNIFICANCE CTRL 162
loc;ao SIMPLE CTRL 164
Iona TEMPERATURE CTRL 167
louco TOMATO CTRL 170
manchete TRANSFORMATION CTRL 172
manicure TREPIDATION CTRL 174
manobra TRIUMPH CTRL 175
marimbondo ULTIMATUM CTRL 177
meado VEGETABLE CTRL 179
meia VEIN CTRL 180
melancia VEHICLE CTRL 181
metade RECIPIENT CTRL 183
m e tido VINEGAR CTRL 184
multa VENG EANCE CTRL 186
musculac;ao VOWEL CTRL 188
nada VOTE CTRL 189
nau ZEBRA CTRL 190
obra ARM CTRL 197
olaria ASSIST CTRL 198
ombro ACTUAL CTRL 199
orn;a AUDIENCE CTRL 200
paiol BANK CTRL 203
eaeo BEEF CTRL 205
paquiderme BOND CTR L 206
pipa COLLAR CTRL 217
prata COMPROMISE CTRL 221
preguic;a CONCOURSE CTRL 223
prenda CONCEAL CTRL 224
propina COSTUME CTRL 226
punho CUSP CTRL 229
eure DATA CTRL 230
quadra DESSERT CTR L 231
raio ENJOY CTRL 237
ralo EXPERT CTRL 238
ramo EXQUISITE CTRL 239
rapaz OR DINARY CTRL 292
rena EXCITING CTRL 242

Page 3 ofi6
 183

EFFECTS OF CROSSLINGUISTIC OVERLAP

Word List 1
prime TARGET Exp Status Item Nbr
renda EXIT CTRL 2 43
reu FACILITY CTRL 246
roda GRIP CTRL 249
ruga HOME CTRL 251
saguao HOSPICE CTRL 252
saia IDIOM CTRL 2 53
saldo INSCRIPTION CTRL 256
sarda INGENIOUS CTRL 254
sinal INSTANCE CTRL 257
sobra JAR CTRL 258
sobrado JOURNAL CTRL 259
solteiro LACE CTRL 260
solrn;:o LAMP CTRL 261
sortudo RECOURSE CTRL 263
sujeira LASER CTRL 264
toldo LENS CTRL 267
trilho LOCATION CTRL 269
tristeza LUNCH CTRL 270
unha MAGAZINE CTRL 271
urubu MASCARA CTRL 273
velho MALL CTRL 275
velocimetro MASTER CTRL 276
verruga NOTICE CTRL 278
veu OUTDOOR CTRL 281
visom PAVEMENT CTRL 287
vovo POST CTRL 289
voz PREJUDICE CTRL 291
aba HEVA D
abestalhado CHECKLEBURST D 2
a<;ucar HARETS D 3
adorno NULIPS D 4
agriao QUELTER D 5
agu a CAHED D 6
alavanca BETTULO D 7
alcatra VALLUME D 227
alergia OFFEN D 8
alfazema HOOPLES D 9
aluguel SUNTELO D 10

amaciante FRONG D 11

amendoim BELISM D 12

Page 4 ofi6
 184

EFFECTS OF CROSSLINGUISTIC OVERLAP

Word List 1
prime TARGET Exp Status Item Nbr
andaime SEGRETS D 13
arco-iris OUTBRAW D 14
asno OLEALD D 266
ata KIFF D 15
aterrisagem PAMA D 16
azeitona CUNWREMY D 17
babosa WINOOZE D 275
bafo VEAST D 18
bagulho REOCCUS D 19
bandeja PUNOUN D 20
banguela MAIATROSE D 21
banheira INWORM D 22
baranga MUNPY D 23
barco OVENCY D 24
barro ESUYP D 284
barulho RERARD D 25
baunilha RERATE D 26
berimbau BESLOOR D 2 53
bina TOOGIT D 282
biscate CHOR ENN D 270
bocejo VOQUEV D 267
bofe MICRONY D 27
bofetada TRONSHAT D 28
borra ch eiro MOGININ D 29
brega BETIPE D 30
bru xa HOLS D 31
brn;o voov D 226
cac;amba CHEAHOOF D 242
cachac;:a NUCK E D 32
cafetao SWOPOM D 241
cafofo PLOCK D 33
cafun e HUNM D 34
calc;a da CHEIR D 35
camada MUEBA D 36
cambio MERIP D 37
caneca UN ELIND D 235
canet a IADLOP D 38
can gote MAIATED D 39
capim BELYCH D 40
capo CHEESTNAIR D 41

Page 5 ofi6
 185

EFFECTS OF CROSSLINGUISTIC OVERLAP

Word List 1
prime TARGET Exp Status Item Nbr
caraca JIOFRAX D 246
carango TINEREW D 276
careca MODGONE D 237
carioca WERRADITH D 271
caroc,:o RODRUP D 278
carrasco KARPLE D 42
carreta TELCI D 43
carroc,:a MORATH D 279
casa PIFFURE D 44
cascavel INTIENT D 45
caxumba DRIRATHIEL D 245
cenoura BEORARNI D 46
cera KREEPEE D 47
chacara UNDERDOUG D 248
chacrete TOXIMBLE D 264
chafariz POLLNEST D 48
chao SOTTA D 49
chicoria ODRANOEL D 50
chinelo OPERTOT D 51
chuchu SHABI D 52
cidadao ENFLEP D 53
cinto REVENTIAL D 54
coisa TEBS D 55
col eira JOHA CKLE D 236
colete ADNANREF D 56
comicio JOVAPHILE D 225
comid a PROFINROR D 57
coqueluche KULATOR D 58
correio HARP ENITAGE D 59
couro FOMMER D 60
coxinha PLOOSNAR D 230
cozinha CHOOR D 61
dende SUMNIKA D 6z
desemerego FUTUKE D 63
deus WOTNOT D 64
dragao TESBAR D 65
erva ZESHO D 66
esfre gao DARNET D 67
esgoto LEPSET D 68
estrum e CHACAKA D 247

Page 6 ofi6
 186

EFFECTS OF CROSSLINGUISTIC OVERLAP

Word List 1
prime TARGET Exp Status Item Nbr
faro! RESAIX D 281
fechadura DINNY D 69
feijoada PROTHORACLED D 70
ferro }UNSET D 71
fiel SERAOS D 72
foca WENSH D 73
fogo BIEDE D 74
fogueira MBSON D 75
framboesa RELONE D 76
frigideira KECHIN D 77
fronha CHURRUSTUL D 78
frouxo WALNERT D 79
furacao TRINETTE D So
furo TWM D 294
gaita NIETWE D 277
garfo PLAMPLE D 81
gentalha BROXNER D 82
geringorn;a MONTAQUEST D 83
girico GROJET D 84
goela MIRESA D 251
goma NEKMIT D 283
graveto OBBISN D 85
gravura NOURINE D 86
gula BUENO D 293
hene JOURC H D 87
hieoteca JOENTELINITY D 88
h6spede FEA D 89
im6vel CAKESKI D 90
impressora DOCTURE D 91
inhame HEMICAL D 92
invalidez NEXTING D 93
jacare PROUSE D 94
janela PUBIEN D 95
jeca TROSP D 96
jejum NONVICT D 97
joia CHIER D 98
jumento ADEMPT D 99
lagar ta WITHNEN D 100
lagar tixa MINSE D 101
lap ela PORTHOLOBATE D 102

Page 7ofi6
 187

EFFECTS OF CROSSLINGUISTIC OVERLAP

Word List 1
prime TARGET Exp Status Item Nbr
laranja FROMMIN D 103
lareira ANUNAT D 104
leite MEGTA D 105
lerdo IBBLY D 106
lesma QUONBI D 107
liminar KROWT D 108
lula RACHLER D 109
lustre CUTION D 110
luz RELAND D 111
macaco CLORE D 112
ma<;arico LAWMER D 113
macete NUSHTRA D 114
ma<;o PLINGT D 115
macumba PEEVISITION D 116
madeixa RERATE D 117
mala PARIDIK D 118
malha,;ao JAYL D 119
mamadeira EDADIVITAN D 120
mangueira TRIME D 121
manicomio YONNAL D 122
manifestante LOGITUDE D 123
maremoto YENBATUR D 124
marimba SAFOME D 239
matra ca CH EILIT H D 252
me leca AFFER D 125
merenda GLARETRAM D 262
mes DURETS D 126
mesa GLANGING D 127
meta LACTER D 128
mexerica BUMMOXA D 240
milho JUESMA D 129
minh oca WANTEMER D 130
mo<;oila SASAROO D 2 43
mocreia NETTE D 131
moeda SUBAVA D 132
moela TITUI D 133
mofo ZOIT D 134
moita RUCTULE D 135
molamb ento MUFTA D 139
m olh o NEVESY D 136

Page 8 ofi6
 188

EFFECTS OF CROSSLINGUISTIC OVERLAP

Word List 1
prime TARGET Exp Status Item Nbr
monge RACA D 137
moqueca CHERSHOEE D 255
morango CLICEZO D 138
mudo ECKLE D 261
muleta TUPACASE D 254
mulher APINGO D 140
mundo NYNDRA D 141
muquifo CUTLUS D 142
mutreta CATACELLA D 143
nabo RUITA D 144
namoro BETTARAN D 145
nanico ANIRAM D 146
nariz FUDE D 147
nenem COMED D 148
ninho ZIBLA D 149
6culos SIEDCASE D 150
olheira SANNY D 151
olho DRAKET D 152
ouvinte RUXTA D 153
earceiro FLIPKNIPS D 154
easta EVENUL E D 155
eata TUPIB D 288
pauta MULTIVE D 156
paviu CHILLAID D 263
pei xe FUEVE D 157
een a WASMAC D 158
perici a VIOTTIS D 244
periguete AFFERARON D 268
peruca BIASDO D 257
picanha PELOO ZOID D 228
pichai;:ao OHCAMARG D 159
pn eu HAPPER D 160
po chete NAPERONE D 286
eocilg a LALONTENT D 161
pomar NECCA D 162
praia HYDRING D 163
presunto PADA D 164
erimavera INHOFT D 165
queda LAUDEITNESS D 166
quib e BUMOLA D 232

Page 9 ofi6
 189

EFFECTS OF CROSSLINGUISTIC OVERLAP

Word List 1
prime TARGET Exp Status Item Nbr
quinze PHICOLY D 167
rabisco GROLE D 168
rabo SCHNOL D 169
rachadura SPIPSCOPS D 170
ram;o KERBAL D 171
raeosa POMODO D 172
reda<;ao GULLSTONT D 173
reduto KUCALIR D 174
regua SADOKIN D 175
relogio ACAVO D 176
remo LANKERTORT D 177
revoga<;ao KROSHIST D 178
rissole PRUVIA D 231
rodovia KATASTRY D 179
roleta BOCILILE D 249
rolha OVERSAL D 180
rotulo PORIC D 181
sabor SETENFI D 182
sacana VASAGLE D 292
saci OSHUN D 183
sacole TESHEKUR D 184
saleicao DIGISOL D 234
salsinha HIABBYA D 185
sap eca OLIELLE D 238
sarro BISTUP D 287
sebo JEREN D 273
sede YEAVE D 186
serelepe LOOPLAB D 229
seta ROINAD D 233
sinuca SEILIU D 256
sol ENENTINESS D 187
surdo YOFFA D 260
sutia COSOMOQ D 188
tabefe SPUSIOUS D 291
tamandua MARESOM D 189
tapete FROWING D 190
tarado QIAMETH D 285
tatu ERAOW D 259
teclado MO VINY D 191
terr em oto AVILI D 192

Page 10 ofi6
 190

EFFECTS OF CROSSLINGUISTIC OVERLAP

Word List 1
prime TARGET Exp Status Item Nbr
tes ta OPONT D 193
tigela ARIGAIT D 194
tormenta IMPLOCE D 195
torresmo FLIO NDESO D 250
toucinho RANTOSTN ER D 196
trai c;:ao KONKERS D 197
tranco ACAER D 272
trema ALQUE D 198
tribufu HENDASSA D 274
trinco REILTAS D 289
troco SATEIL D 290
troc;:o YABATA D 199
tromba MURER D 200
trono QUINK D 201
trote YIMELLO D 280
tr uta FRUGTER D 202
tubarao CO NT ERAB D 203
tutu UMPRE D 204
vagao O NAMA D 269
varal ELOCI N D 205
var eio COPIC D 206
vazamento ZUZUUT D 207
vazao SOAXCAS E D 208
vela PUSOON D 209
ven cimento CINT RERY D 210
ver ao YULANAR D 211
vereda DORITY D 212
ve rniz TAMTOP D 213
vesgo GORMET D 214
vista TELVIT D 215
vitrin e HIPPLED D 216
vit ro BETIFT D 217
vitro la NOVALY D 258
viveiro ZEANDROT D 218
xav eco NOCOBOT D 265
xilindr6 GAULIN D 219
xinxim PHREST D 220
xod6 ATINTE D 221
zabumba SHINESH D 222
zagu eiro BRUXET D 223

Page 11ofi6
 191

EFFECTS OF CROSSLINGUISTIC OVERLAP

Word List 1
prime TARGET Exp Status Item Nbr
zarolh o NENURIA D 224
acento ACCENT F
adepto ADEPT F 2
adjunto ADJUNCT F 3
agenda AGENDA F 4
ae licac;:ao APPLICATION F 5
assist e ASSIST F 7
au dienc ia AUDIENCE F 9
banco BANK F 12
bife BEEF F 14
brando BRAND F 17
carta CART F 19
cartao CARTON F 20
ch oe e CHOP F 23
chute CHUTE F 24
co lar CO LLAR F 26
co nce ito CO NCEIT F 31
cova COVE F 36
cus e CUSP F 38
data DATA F 39
deserto DESSERT F 40
diversao DIVERSION F 42
dra ga DRAG F 43
ed itor EDITOR F 44
esperto EXPERT F 46
estr angeiro ESTRANGED F 48
estr anho STRANGER F 49
exito EXIT F 51
fabrica FABRIC F 53
facilidade FACILITY F 54
gradu ac;:ao GRADUATI O N F 55
gratuito GRATUITY F 56
inscric;:ao INSCRIPTION F 65
lanche LUN CH F 77
legend a LEGEND F 72
locac;:ao LOCATION F 76
maior MAYOR F 79
ma scara MASCARA F 80
mau MALL F 82
noticia NOTI CE F 85

Page 12 ofi6
 192

EFFECTS OF CROSSLINGUISTIC OVERLAP

Word List 1
prime TARGET Exp Status Item Nbr
novela NOVEL F 86
oficio OFFICE F 87
ordinario ORDINARY F 127
outdoor OUTDOOR F 88
pao PAN F 89
earente PARENT F 90
eastel PASTEL F 91
pavimento PAVEMENT F 93
pote POT F 96
erejuizo PREJUDICE F 97
preservativo PRESERVATIVE F 98
pressa PRESS F 99
pretende PRETEND F 100
recurso RECOURSE F 106
resumo RESUME F 107
simpatico SYMPATHETIC F 112
smoking SMOKING F 113
sorte SORT F 115
sueorte SUPPORT F 117
tameao TAMPON F 118
taxa TAX F 119
tenente TENANT F 120
turno TURN F 123
vento VENT F 125
vila VILE F 126
abajur BANDOVA p
acre DRIM p 2
apito WHISTL E p 3
aranha SHENDON p 4
aumento RAISE p 5
balde WOMT p 6
batata DONCE p 7
bebida BEVERAGE p 8
broca YOENK p 9
caixote KAHOE p 10
campo DIELB p 11

carretel REEL p 12
cogumelo MUSHROOM p 13
crern;a BELIEF p 14
descarado BLATANT p 15

Page 13 ofi6
 193

EFFECTS OF CROSSLINGUISTIC OVERLAP

Word List 1
prime TARGET Exp Status Item Nbr
desejo URGE p 16
en uia EEL p 17
ferrugem RUST p 18
gema YOLK p 19
irritado LASATY p 20
isoeor STYROFOAM p 21
lingui<;a SAUSAGE p 22
luva GLOVE p 23
magricela LANKY p 24
manivela CRANK p 25
marinheiro SITUYA p 26
mobilia FURNITURE p 27
6bito PRINKT p 28
oficina HENTIOM p 29
recife REEF p 30
s6tao ATTIC p 31
tesoura SCISSORS p 32
tesoura SCISSORS p 32
touro BULL p 33
valsa WALTZ p 34
vagueiro COWBOY p 35
abrueto ABRUPT C
adotivo ADOPTIVE C 2
alcool ALCOHOL C 4
angulo ANGLE C 6
base BASE C 13
bateria BATTERY C 14
botao BUTTON C 17
capacidade CAPACITY C 21
caso CASE C 24
cheque CHEQUE C 25
coluna COLUM N C 28
comatoso COMAT OSE C 30
comed ia COM EDY C 31
complicado COMPLICATED C 32
computador COMPUTER C 33
condicional CONDITIONAL C 34
confidente CONFIDENT C 35
conforto CO MFORT C 36
cortina CURTAIN C 39
Page 14 ofi6
 194

EFFECTS OF CROSSLINGUISTIC OVERLAP

Word List 1
prime TARGET Exp Status Item Nbr
criticismo CRITICISM C 41
cubo CUBE C 42
dama DAME C 43
detergente DETERGENT C 47
diferente DIFFERENT C 48
distribuic;:ao DISTRIBUTION C 49
elastico ELASTIC C 50
elevado r ELEVATOR C 51
entrada ENTRANCE C 52
estueido STUPID C 53
excelente EXCELLENT C 54
fantasia FANTASY C 56
febre FEVER C 58
fezes FECES C 59
flexivel FLEXIBLE C 60
frustrado FRUSTRATED C 63
fruta FRUIT C 64
gel GEL C 66
genial GENIAL C 68
girafa GIRAFFE C 69
maneguim MANNEQUIN C 79
maguina MACHINE C 80
mineral MINERAL C 84
m osaico MOSAIC C 86
mosquito MOSQUITO C 87
mostarda MUSTARD C 88
museu MUSEUM C 91
natureza NATURE C 92
ocasiao OCCASION C 96
outro OTHER C 100
paciencia PATIENCE C 101
pagam ento PAYMENT C 102
pape l PAPER C 103
eargue PARK C 104
piano PIANO C 107
planta PLANT C 108
prisao PRISON C 111
eroficiencia PROFICIEN CY C 112
projetor PROJECTOR C 113
pr 6pri o PROPER C 114

Page 15 ofi6
 195

EFFECTS OF CROSSLINGUISTIC OVERLAP

Word List 1
prime TARGET Exp Status Item Nbr
rato RAT C 120
referencia REFERENCE C 123
reparo REPAIR C 126
resposta RESPONSE C 128
restaurante RESTAURANT C 129
ritmo RHYTHM C 131
rosa ROSE C 133
shampoo SHAMPOO C 138
sofa SOFA C 142
sucesso SUCCESS C 143
testamento TESTAMENT C 145
texto TEXT C 146
transformac;:ao TRANSFORMATION C 149
trepidac;:ao TREPIDATION C 151
ultimato ULTIMATUM C 154
vagabundo VAGABOND C 155
vegetal VEGETABLE C 156
veia VEIN C 157
verme VERMIN C 159
vinagre VINEGAR C 160
vinganc;:a VENG EANC E C 161
vocac;:ao VOCATION C 162
zebra ZEBRA C 165
zero ZERO C 166

Page 16 ofi6
 196

EFFECTS OF CROSSLINGUISTIC OVERLAP

Word List 2

prime TARGET Exp Status ItemNbr

abalo ABRUPT CTRL
aceno SERVICE CTRL 5
ala ASSISTANCE CTRL 10

almoxarifado BAR CTRL 15

alvo SAP CTRL 17
an el BUTTON CTRL 22
animal COFFEE CTRL 24
anzol CAMEL CTRL 25
azulejo CASE CTRL 29
baldea<;ao SURGERY CTRL 32
baleia COLUMN CTRL 33
batedeira COMPLICATED CTRL 37
beco CONDITIONAL CTRL 39
birra CONSTIPATION CTRL 42
boina CURTAIN CTRL 44
breu CRITICISM CTRL 46
brinco CUBE CTRL 47
bujao DAME CTRL 48
cabelo DEMAND CTRL 49
cal<;a DISGUST CTRL 50
cambalhota DISGRACE CTRL 52
cara ELASTIC CTRL 56
caspa STUPID CTRL 59
cha FACE CTRL 6i
ciente FANTASY CTRL 62
coice FORMA T CTRL 67
cor FREEZER CTRL 69
corja FRUSTRATED CTRL 70
corno FUNDAMENTAL CTRL 72
cra cha GELATIN CTRL 74
crism a GIRAFFE CTRL 76
cueca GRAPHITE CTRL 77
dado SOAP CTRL So
declive IRATE CTRL 84
dedo LIBERTY CTRL 85
derrota MACHINE CTRL 90
despa cho MEDAL CTRL 91
dor MEMORY CTRL 93
eixo METAL CTRL 95

Pag e 1 of 16
 197

EFFECTS OF CROSSLINGUISTIC OVERLAP

Word List 2

prime TARGET Exp Statu s ItemNbr

embalagem MONSTER CTRL 97
escada MOSAIC CTRL 98
escama MOSQUITO CTRL 99
eseantalho MUSTARD CTRL 100
falcatrua MOTORCYCLE CTRL 101
fatia MUSEUM CTRL 104
feriado NEGOTIATE CTRL 106
ferias NORMAL CTRL 107
folha TURN CTRL 108
forno NUMBER CTRL 109
fuligem OCCASION CTRL 112
fundo TOQUE CTRL 113
gaiola TERRACE CTRL 115
garrafa OIL CTRL 117
gaveta OPTION CTRL 118
gosma PATIENCE CTRL 120
gota PAYMENT CTRL 121
goteira PAPER CTRL 122
grampo PARK CTRL 123
granel PERSONAL CTRL 125
greve PLANT CTRL 127
grife PRACTICE CTRL 128
grilo PRECARIOUS CTRL 129
grinalda PRISON CTRL 130
grito PROFICIENCY CTRL 131
infancia PROJECTOR CTRL 133
janta VILE CTRL 137
jarro QUANTITY CTRL 139
jaula RECEIPT CTRL 140
jeit o RADIO CTRL 141
joelho REASON CTRL 143
jovem PRETEND CTRL 144
!ado RELATIONSHIP CTRL 147
la.pis RESPONSE CTRL 151
laudo RESTAURANT CTRL 152
lavabo RHYME CTRL 153
lavadeira RHYTHM CTRL 154
lavradio ROBOT CTRL 155
lirio SHAMPOO CTRL 161

Page 2 of 16
 198

EFFECTS OF CROSSLINGUISTIC OVERLAP

Word List 2

prime TARGET Exp Status ItemNbr

lobo SIMILAR CTRL 163
lodo SOFA CTRL 165
lombriga SUCCESS CTRL 166
lontra TESTAMENT CTRL 168
lote TEXT CTRL 169
malandro TRANSLATION CTRL 171
mane TRANSIT CTRL 173
margarida TROPHY CTRL 176
martelo VAGABOND CTRL 178
merluza VERMIN CTRL 182
mosca PUSH CTRL 185
muro VOCATION CTRL 187
navalha ZERO CTRL 191
navio ACCENT CTRL 192
neblina ADEPT CTRL 193
negocio ADJUNCT CTRL 194
neve AGENDA CTRL 195
nexo APPLICA T ION CTRL 196
onda HAZARD CTRL 201
padeiro BALCONY CTRL 202
palito BATON CTRL 204
paquita BRASS CTRL 207
parapeito BRAND CTRL 2 08
pato COMPASS CTRL 209
pedra CART CTRL 210
12egada CARTON CTRL 211
12eixaria CHAPEL CTRL 212
eerdiz CHAT CTRL 213
pesquisa CHOP CTRL 2 14
piada CHUT E CTRL 2 15
pilantr a CLIQUE CTRL 2 16
pirulito COLLEGE CTRL 2 18
pomada COMPETITION CTRL 219
ponte COMPREHENSIBLE CTRL 220
pregao CONCEIT CTRL 222
privada CUP CTRL 225
prosti bul o COVE CTRL 2 27
pulga CURT CTRL 22 8
quente DISCUSSION CTRL 23 2

Page 3 of 16
 199

EFFECTS OF CROSSLINGUISTIC OVERLAP

Word List 2

prime TARGET Exp Status ItemNbr

quiabo DIVERSION CTRL 233
guitute DRAG CTRL 234
rabanete SUPPORT CTRL 235
racha EDITOR CTRL 236
raguete ESTRANGED CTRL 240
rastro STRANGER CTRL 241
requinte FAN CTRL 244
reto FABRIC CTRL 245
risco GRADUATION CTRL 247
riso GRATUITY CTRL 248
ronco DUEL CTRL 293
rosca GROCERY CTRL 250
sarna INJURY CTRL 255
soneca RESUME CTRL 262
taiher LEGEND CTRL 265
tela LECTURE CTRL 266
tren6 LIMP CTRL 268
urtiga MAYOR CTRL 272
vapor MASS CTRL 274
verde MOISTURE CTRL 277
vespa NOVEL CTRL 279
vestido OFFICE CTRL 280
viagem PAN CTRL 282
vibora PARENT CTRL 283
videira RETIRED CTRL 284
vidro PASTEL CTRL 285
virilha PATRON CTRL 286
voraz POLICY CTRL 288
vovo POT CTRL 290
zonzo PRESERVATIVE CTRL 292
aba HEVA D
abestalh ado CHECKLEBURST D 2
ai;:ucar HARETS D 3
adorno NULIPS D 4
agriao QUELTER D 5
agua CAHED D 6
alavanca BETTULO D 7
alcatra VALLUME D 227
alergia OFFEN D 8

Page 4 of 16
 200

EFFECTS OF CROSSLINGUISTIC OVERLAP

Word List 2

prime TARGET Exp Status ItemNbr

alfazema HOOPLES D 9
alu[luel SUNTELO D 10

amaciante FRONG D 11

amendoim BELISM D 12
andaime SEGRETS D 13
arco-iris OUTBRAW D 14
asno OLEALD D 266
ata KIFF D 15
aterrisagem PAMA D 16
azeitona CUNWREMY D 17
babosa WINOOZE D 275
bafo VEAST D 18
bagulho REOCCUS D 19
band eja PUNOUN D 20
banguel a MALATROSE D 21
banheira INWORM D 22
baran!la MUNPY D 23
barco OVENCY D 24
barro ESUYP D 284
barulho RERARD D 25
baunilha RERATE D 26
berimbau BESLOOR D 253
bina TOOGIT D 282
biscate CHORENN D 270
bocejo VOQUEV D 267
bofe MICRONY D 27
bofetada TRONSHAT D 28
borracheiro MOGININ D 29
bre ga BETIPE D 30
bruxa HOLS D 31
buc;o voov D 226
cac;amba CHEAHOOF D 242
cachac;a NUCKE D 32
cafetao SWOPOM D 241
cafofo PLOCK D 33
cafune HUNM D 34
calc;ada CHEIR D 35
cam ada MUEBA D 36
cambio MERIP D 37

Pag e 5 of 16
 201

EFFECTS OF CROSSLINGUISTIC OVERLAP

Word List 2

prime TARGET Exp Status ItemNbr

caneca UNELIND D 235
caneta LADLOP D 38
can!lote MALATED D 39
caeim BELYCH D 40
caeo CHEESTNAIR D 41
caraca JIOFRAX D 246
carango TINEREW D 276
careca MODGONE D 237
carioca WERRADITH D 271
caroc;o RODRUP D 278
carrasco KARPLE D 42
carreta TELCI D 43
carroc;a MORATH D 279
casa PIFFURE D 44
cascave l INTIENT D 45
caxumba DRIRATHIEL D 2 45
cenoura BEORARNI D 46
cera KREEPEE D 47
chacara UNDERDOUG D 248
chacrete TOXIMBLE D 264
chafari z POLLNEST D 48
cha o SOTTA D 49
chicor ia ODRANOEL D 50
chinelo OPERTOT D 51
chuchu SHABI D 52
cidadao ENFLEP D 53
cinto REVENTIAL D 54
coisa TEBS D 55
coleira JOHACKLE D 236
colete ADNANREF D 56
comic io JOVAP HILE D 225
comida PROFINROR D 57
coqueluche KULATOR D 58
correio HARPENITAGE D 59
couro FOMMER D 60
coxinha PLOOSNAR D 230
cozin ha CHO OR D 61
dende SUMNIKA D 62
desemprego FUTUKE D 63

Page 6 of 16
 202

EFFECTS OF CROSSLINGUISTIC OVERLAP

Word List 2

prime TARGET Exp Status ItemNbr

deus WOTNOT D 64
dra[laO TESBAR D 65
erva ZESHO D 66
esfregao DARNET D 67
esgoto LEPSET D 68
estrum e CHACAKA D 2 47
faro! RESAIX D 281
fechadura DINNY D 69
feijoada PROTHORACLED D 70
ferro ]UNSET D 71
tie I SERAOS D 72
foca WENSH D 73
fogo BIEDE D 74
fogueira MBSON D 75
framboesa RELONE D 76
frigideira KECHIN D 77
fronha CHURRUSTUL D 78
frouxo WALNERT D 79
furacao TRINETIE D So
furo TWM D 294
gaita NIETWE D 277
garfo PLAMPLE D 81
gentalha BROXNER D 82
geringorn;a MONTAQUEST D 83
girico GROJET D 84
[lOela MIRESA D 251
goma NEKMIT D 283
graveto OBBISN D 85
gravura NOURINE D 86
gula BUENO D 293
hene JOUR CH D 87
hipoteca JOENTELINITY D 88
h6spede FEA D 89
im6vel CAKESKI D 90
impressora DOCTURE D 91
inhame HEMICAL D 92
invalid ez NEXTIN G D 93
jacar e PROUSE D 94
janela PUBIEN D 95

Page 7 of 16
 203

EFFECTS OF CROSSLINGUISTIC OVERLAP

Word List 2

prime TARGET Exp Status ItemNbr

jeca TROSP D 96

jeium NONVIC T D 97
joia CHIER D 98
jumento ADEMPT D 99
lagarta WITHNEN D 100

lagartixa MINSE D 101

lapela PORTHOLOBATE D 102

laranja FROMMIN D 103

lareira ANUNAT D 104
leite MEGTA D 105
lerdo IBBLY D 106

lesma QUONBI D 107

liminar KROWT D 108

lula RACHLER D 109

lustre CUTION D 110

luz RELAND D 111

macaco CLORE D 112

mac;:arico LAWMER D 113

macete NUSHTRA D 114
mac;:o PLINGT D 115
macumba PEEVISITION D 116

mad eixa RERATE D 117

mala PARIDIK D 118

malhac;:ao JAYL D 119

mamadeira EDADIVITAN D 120

man1:1ueira TRIME D 121

manicomio YONNAL D 122

manifestante LOGITUDE D 123

maremoto YENBATUR D 12 4
marimba SAFOME D 2 39

matraca CHEILITH D 252

meleca AFFER D 125
merenda GLARETRAM D 262

mes DURETS D 126

mesa GLANGING D 127

meta LACTER D 128

m exerica BUMMOXA D 2 40
milho JUESMA D 12 9
minhoca WANTEMER D 13 0

Page 8 of 16
 204

EFFECTS OF CROSSLINGUISTIC OVERLAP

Word List 2

prime TARGET Exp Statu s ItemNbr

m0<;oila SASAROO D 243
mocreia NETTE D 131
moeda SUBAVA D 132
moela TITUI D 133
mofo ZOIT D 134
moita RUCTULE D 135
molambento MUFTA D 139
mo lho NEVESY D 136
monge RACA D 137
moqueca CH ERSHO EE D 2 55
morango CLICEZO D 138
mudo ECKLE D 261
muleta TUPACASE D 2 54
mulh er APINGO D 140
mundo NYNDRA D 141
m ugui fo CUTLUS D 142
mutreta CATACELLA D 143
nabo RUITA D 144
namoro BETTARAN D 145
nanico ANIRAM D 146
nari z FUD E D 147
nenem COMED D 148
ninho ZIBLA D 149
6c ulo s SIEDCASE D 150
olheira SANNY D 151
olho DRAKET D 152
ouvinte RUXTA D 153
earc eiro FLIPKNIPS D 154
pasta EVENULE D 155
pat a TUPIB D 288
pauta MULTIVE D 156
paviu CH ILLAID D 263
pei xe FUEVE D 157
pena WASMAC D 158
pe ric ia VIOTTIS D 244
periguet e AFFERARON D 268
peruca BIASDO D 257
picanha PELOOZOID D 228
pichac;ao OHCAMA RG D 159

Page 9 of 16
 205

EFFECTS OF CROSSLINGUISTIC OVERLAP

Word List 2

prime TARGET Exp Status ItemNbr

pneu HAPPER D 160
eochete NAPERONE D 286
eocil~a lALONTENT D 161
eomar NECCA D 162
praia HYDRING D 163
presunto PADA D 164
primavera INHOFT D 165
queda lAUDEITNESS D 166
quibe BUMOlA D 232
quinze PHICOLY D 167
rabisco GROLE D 168
rabo SCHNOL D 169
rachadura SPIPSCOPS D 170
ran<;o KERBAL D 171
raposa POMODO D 172
reda<;ao GULLSTONT D 173
reduto KUCALIR D 174
regua SADOKIN D 175
rel6gio ACAVO D 176
remo lANKERTORT D 177
revoga<;ao KROSHIST D 178
rissole PRUVIA D 231
rodovia KATASTRY D 179
roleta BOCILILE D 249
rolha OVERSAL D 180
r6tulo PORIC D 181
sabor SETENFI D 182
sacana VASAGLE D 292
saci OSHUN D 183
saco le TESHEKUR D 184
salpic ao DIGISOL D 234
salsinha HIABBYA D 185
sapeca OLIELLE D 238
sarro BISTUP D 287
sebo JEREN D 273
sede YEAVE D 186
serele pe LOO PlAB D 229
seta ROINAD D 233
sinuc a SEILIU D 256

Page 10 of 16
 206

EFFECTS OF CROSSLINGUISTIC OVERLAP

Word List 2

prime TARGET Exp Status ItemNbr

sol ENENTINESS D 187
surdo YOFFA D 260
sutia COSOMOQ D 188
tabefe SPUSIOUS D 291
tamandua MARESOM D 189
tapete FROWING D 190
tarado QIAMETH D 285
tatu ERAOW D 259
teclado MOVINY D 191
terremoto AVILI D 192
testa OPONT D 193
tigela ARIGAIT D 194
tormenta IMPLOCE D 195
torresmo FLIONDESO D 250
toucinho RANTOSTNER D 196
trai<;ao KONKERS D 197
tranco ACAER D 272
trema ALQUE D 198
tribufu HENDASSA D 274
trinco REILTAS D 289
troco SATEIL D 290
tro <;o YABATA D 199
tromb a MURER D 200
trono QUINK D 201
trote YIMELLO D 280
truta FRUGTER D 202
tubarao CONTERAB D 203
tutu UMPR E D 204
vagao ONAMA D 269
varal ELOCIN D 205
varej o COPIC D 206
vazamento ZUZUUT D 207
vazao SOAXCASE D 208
vela PUSOON D 209
vencimento CINTRERY D 210
verao YULANAR D 211
ver eda DORITY D 212
verniz TAMTOP D 213
ves go GORMET D 214

Page 11of 16
 207

EFFECTS OF CROSSLINGUISTIC OVERLAP

Word List 2

prime TARGET Exp Statu s ItemNbr

vista TELVIT D 215
vitrine HIPPL ED D 216
vitro BETIFT D 217
vitrola NOVALY D 258
viveiro ZEANDROT D 218
xaveco NOCOBOT D 265
xilind r6 GAULIN D 219
xinxim PHREST D 220
xod6 ATINTE D 221
zabumba SHIN ESH D 222
zagueiro BRUXET D 223
zarolho N ENURIA D 224
arma ARM F 6
atua l ACTUAL F 8
azar HAZARD F 10

balcao BALCONY F 11

batom BATON F 13
bonde BOND F 15
brac;:o BRASS F 16
bussola COMPASS F 18
chape u CHAP EL F 21
ch ato CHAT F 22
cliq ue CLIQUE F 25
co legio COLLEGE F 27
competic;:ao COMPETITION F 28
comereensivel COMPREHENSIBLE F 29
comeromisso COMPROMISE F 30
conc ur so CONCOU RSE F 32
co ns elho CON CEAL F 33
copo CUP F 34
costume COSTUME F 35
curto CURT F 37
discus sao DISCUSSION F 41
duelo DUEL F 128
enjoo ENJOY F 45
esq uis ito EXQUISITE F 47
exc itant e EXCITING F 50
fa FAN F 52
grip e GRIP F 57

Page 12 of 16
 208

EFFECTS OF CROSSLINGUISTIC OVERLAP

Word List 2

prime TARGET Exp Status ItemNbr

grosseria GROCERY F 58
haste HASTE F 59
homem HOME F 60
hoseicio HOSPICE F 61
idioma IDIOM F 62
ingenuo INGENIOUS F 63
inju ria INJURY F 64
instancia INSTANCE F 66
jarra JAR F 67
jornal JOURNAL F 68
lac;o LACE F 69
lampada LAMP F 70
lazer LASER F 71
leit ura LECTURE F 73
lern;o LENS F 74
limo LIMP F 75
ma!lazine MAGAZINE F 78
massa MASS F 81
mestre MASTER F 83
mistura MOISTURE F 84
patrao PATRON F 92
poli cia POLICY F 94
poste POST F 95
pulo PULL F 101
puxe PUSH F 102
ran!l er RANGER F 103
receita RECEIPT F 104
recipiente RECIPIENT F 105
retirado RETIRED F 108
rim RIM F 109
sapo SAP F 110
servic;o SERVICE F 111
sopa SOAP F 114
sumo SUM F 116
terrac;o TERRACE F 121
toque TOQUE F 122
ultimamente ULTIMATELY F 124
abajur BANDOVA p
acre DRIM p 2

Page 13 of 16
 209

EFFECTS OF CROSSLINGUISTIC OVERLAP

Word List 2

prime TARGET Exp Status ItemNbr

apito WHISTLE p 3
aranha SHENDON p 4
aumento RAISE p 5
balde WOMT p 6
batata DONCE p 7
bebida BEVERAGE p 8
broca YOENK p 9
caixote KAHOE p 10

campo DIELB p 11

carretel REEL p 12
cogumelo MUSHROOM p 13
crem,:a BELIEF p 14
descarado BLATANT p 15
des ejo URGE p 16
engui a EEL p 17
ferrugem RUST p 18
ema YOLK p 19
irritado LASATY p 20
isopor STYROFOAM p 21
lingui<,:a SAUSAGE p 22
luva GLOVE p 23
magri cela LANKY p 24
m anivel a CRANK p 25
marinheiro SITUYA p 26
mobilia FURNITURE p 27
6bito PRINKT p 28
oficina HENTIOM p 29
recife REEF p 30
s6tao ATTIC p 31
tesoura SCISSORS p 32
tesour a SCISSORS p 32
touro BULL p 33
valsa WALTZ p 34
vaqueiro COWBO Y p 35
alarme ALARM C 3
alinhamento ALIGNMENT C 5
assinatura SIGNATURE C 7
assistenci a ASSISTANCE C 8
astuto ASTUTE C 9

Page 14 of 16
 210

EFFECTS OF CROSSLINGUISTIC OVERLAP

Word List 2

prime TARGET Exp Status ItemNbr

bagagem BAGGAGE C 10

banda BAND C 11

bar BAR C 12
bee BEIGE C 15
bomba BOMB C 16
bravo BRAVE C 18
cafe COFFEE C 19
camelo CAMEL C 20
cargo CARGO C 22
carro CAR C 23
chocolate CHOCOLATE C 26
cirurgia SURGERY C 27
coma COMA C 29
constipai;:ao CONSTIPATION C 37
convenien te CONVENIENT C 38
creme CREAM C 40
demanda DEMAND C 44
desgosto DISGUST C 45
desgrai;:a DISGRACE C 46
face FACE C 55
fatalidade FATALITY C 57
formato FORMAT C 61
freezer FREEZER C 62
fundamental FUNDAMENTAL C 65
gelatina GELATIN C 67
!:lrafite GRAPHITE C 70
hora HOUR C 71
in!:lenuidade INGENUITY C 72
interpretai;:ao INTERPRETATION C 73
irado IRATE C 74
liberd ade LIBERTY C 75
liceni;:a LICENSE C 76
lingua LANGUAGE C 77
linha LINE C 78
medalha MEDAL C 81
memoria MEMORY C 82
m eta l METAL C 83
monstro MONSTER C 85
motocicleta MOTOR CYCLE C 89

Page 15 of 16
 211

EFFECTS OF CROSSLINGUISTIC OVERLAP

Word List 2

prime TARGET Exp Status ItemNbr

mouse MOUSE C 90
negociar NEGOTIATE C 93
normal NORMAL C 94
numero NUMBER C 95
oceano OCEAN C 97
oleo OIL C 98
opi;:ao OPTION C 99
particular PARTICULAR C 105
pessoal PERSONAL C 106
pratica PRACTICE C 109
precario PRECARIOUS C 110
psicologia PSYCHOLOGY C 115
psicose PSYCHOSIS C 116
puma PUMA C 117
qu antidade QUANTITY C 118
radio RADIO C 119
razao REASON C 121
real REAL C 122
relacionamento RELATIONSHIP C 124
remedio REMEDY C 125
requ erimento REQUIREM ENT C 127
rima RHYME C 130
rob6 ROBOT C 132
rude RUDE C 134
rumor RUMOR C 135
samba SAMBA C 136
sentimento SENTIMENT C 137
significancia SIGNIFICANCE C 139
simila r SIMILAR C 140
simples SIMPLE C 141
temperatur a TEMPERATURE C 144
tom ate TOMATO C 147
tradui;:ao TRANSLATION C 148
tran sito TRANSIT C 150
triunfo TRIUMPH C 152
trofeu TROPHY C 153
veiculo VEHICLE C 158
vogal VOWEL C 163
voto VOTE C 164

Page 16 of 16
 212

EFFECTS OF CROSSLINGUISTIC OVERLAP

Appendix G: RStudio lmer model raw output

LMER Model

install.packages(c("lmerTest", "optimx", "nloptr", "MuMIn", "r2glmm", "car"))

## Installing packages into '/home/rstudio-user/R/x86_64-pc-linux-gnu-library/3.5'

## (as 'lib' is unspecified)

#Loading required packages

library(foreign)
library(lmerTest)

## Loading required package: lme4

## Loading required package: Matrix

##
## Attaching package: 'lmerTest'

## The following object is masked from 'package:lme4':

##
## lmer

## The following object is masked from 'package:stats':

##
## step

library(optimx)
library(nloptr)
 213

EFFECTS OF CROSSLINGUISTIC OVERLAP

library(MuMIn)
library(r2glmm)
library(car)

## Loading required package: carData

#Data optimizer
defaultControl <- list(algorithm="NLOPT_LN_BOBYQA",xtol_rel=1e-6,maxeval=1e5)

nloptwrap2 <- function(fn,par,lower,upper,control=list(),...) {

for (n in names(defaultControl))
if (is.null(control[[n]])) control[[n]] <- defaultControl[[n]]
res <- nloptr(x0=par,eval_f=fn,lb=lower,ub=upper,opts=control,...)
with(res,list(par=solution,
fval=objective,
feval=iterations,
conv=if (status>0) 0 else status,
message=message))
}

#Reading the data file

setwd("/cloud/project/LMER Model")
Dissertation_Dataset <- read.csv("/cloud/project/Master Datasets/Dissertation_Master_Dataset.csv", sep = ",
", dec = ".", encoding = "UTF-8")

#Configuring the data variables

Dissertation_Dataset$RT <- as.numeric(Dissertation_Dataset$RT)
Dissertation_Dataset$invRT <- as.numeric(Dissertation_Dataset$invRT)
Dissertation_Dataset$PAIR <- as.factor(Dissertation_Dataset$PAIR)
Dissertation_Dataset$LIST <- as.factor(Dissertation_Dataset$LIST)
 214

EFFECTS OF CROSSLINGUISTIC OVERLAP

Dissertation_Dataset$ZTime <- as.numeric(Dissertation_Dataset$ZTime)

Dissertation_Dataset$Time <- as.numeric(Dissertation_Dataset$Time)
Dissertation_Dataset$AGE <- as.numeric(Dissertation_Dataset$AGE)
Dissertation_Dataset$GENDER <- as.factor(Dissertation_Dataset$GENDER)
Dissertation_Dataset$SEMANTIC <- as.factor(Dissertation_Dataset$SEMANTIC)
Dissertation_Dataset$SEMANTIC_OLD <- as.factor(Dissertation_Dataset$SEMANTIC_OLD)
Dissertation_Dataset$SEMANTIC_ORIGINAL <- as.factor(Dissertation_Dataset$SEMANTIC_ORIGINAL)
Dissertation_Dataset$ZPHONETIC <- as.numeric(Dissertation_Dataset$ZPHONETIC)
Dissertation_Dataset$PHONETIC_RAW <- as.numeric(Dissertation_Dataset$PHONETIC_RAW)
Dissertation_Dataset$PARTICIPANT <- as.factor(Dissertation_Dataset$PARTICIPANT)
Dissertation_Dataset$TARGET <- as.factor(Dissertation_Dataset$TARGET)
Dissertation_Dataset$PRIME <- as.factor(Dissertation_Dataset$PRIME)
Dissertation_Dataset$ZEnglish_Freq <- as.numeric(Dissertation_Dataset$ZEnglish_Freq)
Dissertation_Dataset$English_Freq <- as.numeric(Dissertation_Dataset$English_Freq)
Dissertation_Dataset$ZPortuguese_Freq <- as.numeric(Dissertation_Dataset$ZPortuguese_Freq)
Dissertation_Dataset$Portuguese_Freq <- as.numeric(Dissertation_Dataset$Portuguese_Freq)
Dissertation_Dataset$logPortuguese_Freq <- as.numeric(Dissertation_Dataset$logPortuguese_Freq)
Dissertation_Dataset$logEnglish_Freq <- as.numeric(Dissertation_Dataset$logEnglish_Freq)
Dissertation_Dataset$ZPROFICIENCY <- as.numeric(Dissertation_Dataset$ZPROFICIENCY)
Dissertation_Dataset$SEMAC <- as.numeric(Dissertation_Dataset$SEMAC)
Dissertation_Dataset$ERS <- as.numeric(Dissertation_Dataset$ERS)
Dissertation_Dataset$GS <- as.numeric(Dissertation_Dataset$GS)
Dissertation_Dataset$OS <- as.numeric(Dissertation_Dataset$OS)
Dissertation_Dataset$Lev <- as.numeric(Dissertation_Dataset$Lev)
Dissertation_Dataset$NLD <- as.numeric(Dissertation_Dataset$NLD)
Dissertation_Dataset$ZGS <- as.numeric(Dissertation_Dataset$ZGS)
Dissertation_Dataset$ZOS <- as.numeric(Dissertation_Dataset$ZOS)
Dissertation_Dataset$ZLev <- as.numeric(Dissertation_Dataset$ZLev)
Dissertation_Dataset$ZNLD <- as.numeric(Dissertation_Dataset$ZNLD)
 215

EFFECTS OF CROSSLINGUISTIC OVERLAP

Dissertation_Dataset$length_prime_raw <- as.numeric(Dissertation_Dataset$length_prime_raw)

Dissertation_Dataset$length_TARGET_raw <- as.numeric(Dissertation_Dataset$length_TARGET_raw)
Dissertation_Dataset$Zlength_prime_raw <- as.numeric(Dissertation_Dataset$Zlength_prime_raw)
Dissertation_Dataset$Zlength_TARGET_raw <- as.numeric(Dissertation_Dataset$Zlength_TARGET_raw)
Dissertation_Dataset$English_Freq_gmc <- as.numeric(Dissertation_Dataset$English_Freq_gmc)
Dissertation_Dataset$Portuguese_Freq_gmc <- as.numeric(Dissertation_Dataset$Portuguese_Freq_gmc)
Dissertation_Dataset$SEMAC_gmc <- as.numeric(Dissertation_Dataset$SEMAC_gmc)
Dissertation_Dataset$English_Freq_log_gmc <- as.numeric(Dissertation_Dataset$English_Freq_log_gmc)
Dissertation_Dataset$Portuguese_Freq_log_gmc <- as.numeric(Dissertation_Dataset$Portuguese_Freq_log_gmc)
Dissertation_Dataset$length_TARGET_raw_log_gmc <- as.numeric(Dissertation_Dataset$length_TARGET_raw_log_gmc
)
Dissertation_Dataset$Englisg_Freq_gmc <- as.numeric(Dissertation_Dataset$English_Freq_gmc)
Dissertation_Dataset$NLD_gmc <- as.numeric(Dissertation_Dataset$NLD_gmc)
Dissertation_Dataset$NLD_log <- as.numeric(Dissertation_Dataset$NLD_log)
Dissertation_Dataset$NLD_log_gmc <- as.numeric(Dissertation_Dataset$NLD_log_gmc)
Dissertation_Dataset$PHONETIC_gmc <- as.numeric(Dissertation_Dataset$PHONETIC_gmc)
Dissertation_Dataset$PHONETIC_RAW_gmc <- as.numeric(Dissertation_Dataset$PHONETIC_RAW_gmc)
Dissertation_Dataset$PHONETIC_gmc <- as.numeric(Dissertation_Dataset$PHONETIC_gmc)
Dissertation_Dataset$PHONETIC_log <- as.numeric(Dissertation_Dataset$PHONETIC_log)
Dissertation_Dataset$PHONETIC_RAW_log_gmc <- as.numeric(Dissertation_Dataset$PHONETIC_RAW_log_gmc)
Dissertation_Dataset$AGE_gmc <- as.numeric(Dissertation_Dataset$AGE_gmc)
Dissertation_Dataset$Time_ms_gmc <- as.numeric(Dissertation_Dataset$Time_ms_gmc)
Dissertation_Dataset$ERS_gmc <- as.numeric(Dissertation_Dataset$ERS_gmc)
 216

EFFECTS OF CROSSLINGUISTIC OVERLAP

#Releveling Pertinent Variables

Dissertation_Dataset$SEMANTIC<-relevel(Dissertation_Dataset$SEMANTIC,"CTRL")
Dissertation_Dataset$LANGUAGE<-relevel(Dissertation_Dataset$LANGUAGE,"Functional Monolinguals")

# Running best fit model

LMER_best_fit_model <- lmer(invRT ~ LANGUAGE*SEMANTIC + ZNLD_gmc*ZPHONETIC_gmc + ZEnglish_Freq_gmc + Zlengt
h_TARGET_raw_gmc + ZTime_ms_gmc + OTHER_LANGUAGES + (1|PAIR) + (1|PRIME) + (1|TARGET) + (SEMANTIC|PARTICIPA
NT) + (0+ZEnglish_Freq_gmc|PARTICIPANT) + (0+ZTime_ms_gmc|PARTICIPANT) + (0+Zlength_TARGET_raw_gmc|PARTICIP
ANT), data = Dissertation_Dataset, REML = FALSE, control=lmerControl(optimizer="nloptwrap2"))

#Obtaining model summary

summary(LMER_best_fit_model)

## Linear mixed model fit by maximum likelihood . t-tests use Satterthwaite's

## method [lmerModLmerTest]
## Formula:
## invRT ~ LANGUAGE * SEMANTIC + ZNLD_gmc * ZPHONETIC_gmc + ZEnglish_Freq_gmc +
## Zlength_TARGET_raw_gmc + ZTime_ms_gmc + OTHER_LANGUAGES +
## (1 | PAIR) + (1 | PRIME) + (1 | TARGET) + (SEMANTIC | PARTICIPANT) +
## (0 + ZEnglish_Freq_gmc | PARTICIPANT) + (0 + ZTime_ms_gmc |
## PARTICIPANT) + (0 + Zlength_TARGET_raw_gmc | PARTICIPANT)
## Data: Dissertation_Dataset
## Control: lmerControl(optimizer = "nloptwrap2")
##
## AIC BIC logLik deviance df.resid
## 14364.2 14633.7 -7150.1 14300.2 33484
##
 217

EFFECTS OF CROSSLINGUISTIC OVERLAP

## Scaled residuals:
## Min 1Q Median 3Q Max
## -7.4402 -0.6202 -0.0386 0.5881 6.1591
##
## Random effects:
## Groups Name Variance Std.Dev. Corr
## PRIME (Intercept) 0.0043107 0.06566
## PAIR (Intercept) 0.0022357 0.04728
## TARGET (Intercept) 0.0092168 0.09600
## PARTICIPANT Zlength_TARGET_raw_gmc 0.0004832 0.02198
## PARTICIPANT.1 ZTime_ms_gmc 0.0152537 0.12351
## PARTICIPANT.2 ZEnglish_Freq_gmc 0.0001005 0.01003
## PARTICIPANT.3 (Intercept) 0.0655086 0.25595
## SEMANTICCOG 0.0031251 0.05590 0.46
## SEMANTICDISTR 0.0170481 0.13057 -0.36 0.19
## SEMANTICFF 0.0010710 0.03273 0.54 0.63 0.22
## Residual 0.0828356 0.28781
## Number of obs: 33516, groups:
## PRIME, 901; PAIR, 852; TARGET, 591; PARTICIPANT, 57
##
## Fixed effects:
## Estimate Std. Error df t value
## (Intercept) -1.713e+00 6.388e-02 7.091e+01 -26.818
## LANGUAGEBilinguals 1.986e-01 7.636e-02 6.402e+01 2.601
## SEMANTICCOG -6.777e-02 1.964e-02 2.878e+02 -3.451
## SEMANTICDISTR 2.437e-01 2.709e-02 7.954e+01 8.997
## SEMANTICFF -4.078e-02 1.995e-02 4.420e+02 -2.044
## ZNLD_gmc -8.221e-03 6.348e-03 1.987e+03 -1.295
## ZPHONETIC_gmc 3.245e-03 2.560e-03 5.485e+03 1.267
 218

EFFECTS OF CROSSLINGUISTIC OVERLAP

## ZEnglish_Freq_gmc -1.888e-02 5.354e-03 4.617e+02 -3.527

## Zlength_TARGET_raw_gmc 3.036e-02 5.928e-03 3.667e+02 5.121
## ZTime_ms_gmc -5.692e-02 1.679e-02 5.940e+01 -3.389
## OTHER_LANGUAGESYES 1.226e-01 6.715e-02 5.691e+01 1.826
## LANGUAGEBilinguals:SEMANTICCOG -4.164e-02 1.789e-02 5.520e+01 -2.328
## LANGUAGEBilinguals:SEMANTICDISTR -4.531e-03 3.546e-02 5.656e+01 -0.128
## LANGUAGEBilinguals:SEMANTICFF 1.761e-02 1.522e-02 5.307e+01 1.157
## ZNLD_gmc:ZPHONETIC_gmc 7.533e-03 3.277e-03 1.671e+03 2.299
## Pr(>|t|)
## (Intercept) < 2e-16 ***
## LANGUAGEBilinguals 0.011542 *
## SEMANTICCOG 0.000643 ***
## SEMANTICDISTR 9.31e-14 ***
## SEMANTICFF 0.041527 *
## ZNLD_gmc 0.195404
## ZPHONETIC_gmc 0.205051
## ZEnglish_Freq_gmc 0.000462 ***
## Zlength_TARGET_raw_gmc 4.92e-07 ***
## ZTime_ms_gmc 0.001250 **
## OTHER_LANGUAGESYES 0.073059 .
## LANGUAGEBilinguals:SEMANTICCOG 0.023631 *
## LANGUAGEBilinguals:SEMANTICDISTR 0.898777
## LANGUAGEBilinguals:SEMANTICFF 0.252305
## ZNLD_gmc:ZPHONETIC_gmc 0.021632 *
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1

##
## Correlation matrix not shown by default, as p = 15 > 12.
 219

EFFECTS OF CROSSLINGUISTIC OVERLAP

## Use print(x, correlation=TRUE) or

## vcov(x) if you need it

#Obtaining confidence intervals

lower <- coef(summary(LMER_best_fit_model))[,1] + qnorm(.025)*coef(summary(LMER_best_fit_model))[,2]
upper <- coef(summary(LMER_best_fit_model))[,1] + qnorm(.975)*coef(summary(LMER_best_fit_model))[,2]
cbind(coef(summary(LMER_best_fit_model)), lower, upper)

## Estimate Std. Error df

## (Intercept) -1.713136770 0.063879946 70.91278
## LANGUAGEBilinguals 0.198577234 0.076356564 64.02065
## SEMANTICCOG -0.067771841 0.019639952 287.84904
## SEMANTICDISTR 0.243740424 0.027090214 79.54454
## SEMANTICFF -0.040780256 0.019949363 441.95256
## ZNLD_gmc -0.008221424 0.006347647 1986.76272
## ZPHONETIC_gmc 0.003245229 0.002560453 5485.08733
## ZEnglish_Freq_gmc -0.018883343 0.005354073 461.67348
## Zlength_TARGET_raw_gmc 0.030355769 0.005927736 366.73161
## ZTime_ms_gmc -0.056920741 0.016794282 59.40272
## OTHER_LANGUAGESYES 0.122629148 0.067147924 56.90869
## LANGUAGEBilinguals:SEMANTICCOG -0.041641526 0.017890928 55.20199
## LANGUAGEBilinguals:SEMANTICDISTR -0.004531108 0.035460415 56.55559
## LANGUAGEBilinguals:SEMANTICFF 0.017614627 0.015219558 53.06934
## ZNLD_gmc:ZPHONETIC_gmc 0.007533500 0.003276959 1670.90311
## t value Pr(>|t|) lower
## (Intercept) -26.8180685 7.466639e-39 -1.838339164
## LANGUAGEBilinguals 2.6006570 1.154209e-02 0.048921118
## SEMANTICCOG -3.4507132 6.428742e-04 -0.106265439
## SEMANTICDISTR 8.9973607 9.310208e-14 0.190644581
 220

EFFECTS OF CROSSLINGUISTIC OVERLAP

## SEMANTICFF -2.0441884 4.152708e-02 -0.079880288

## ZNLD_gmc -1.2951924 1.954043e-01 -0.020662583
## ZPHONETIC_gmc 1.2674434 2.050506e-01 -0.001773166
## ZEnglish_Freq_gmc -3.5269117 4.624682e-04 -0.029377133
## Zlength_TARGET_raw_gmc 5.1209716 4.920232e-07 0.018737619
## ZTime_ms_gmc -3.3892929 1.249909e-03 -0.089836929
## OTHER_LANGUAGESYES 1.8262538 7.305927e-02 -0.008978365
## LANGUAGEBilinguals:SEMANTICCOG -2.3275219 2.363102e-02 -0.076707101
## LANGUAGEBilinguals:SEMANTICDISTR -0.1277793 8.987769e-01 -0.074032243
## LANGUAGEBilinguals:SEMANTICFF 1.1573678 2.523049e-01 -0.012215159
## ZNLD_gmc:ZPHONETIC_gmc 2.2989299 2.163180e-02 0.001110778
## upper
## (Intercept) -1.587934377
## LANGUAGEBilinguals 0.348233350
## SEMANTICCOG -0.029278243
## SEMANTICDISTR 0.296836268
## SEMANTICFF -0.001680224
## ZNLD_gmc 0.004219735
## ZPHONETIC_gmc 0.008263624
## ZEnglish_Freq_gmc -0.008389553
## Zlength_TARGET_raw_gmc 0.041973918
## ZTime_ms_gmc -0.024004553
## OTHER_LANGUAGESYES 0.254236661
## LANGUAGEBilinguals:SEMANTICCOG -0.006575952
## LANGUAGEBilinguals:SEMANTICDISTR 0.064970027
## LANGUAGEBilinguals:SEMANTICFF 0.047444413
## ZNLD_gmc:ZPHONETIC_gmc 0.013956222
 221

EFFECTS OF CROSSLINGUISTIC OVERLAP

#Obtaining R2 values
r.squaredGLMM(LMER_best_fit_model)

## Warning: 'r.squaredGLMM' now calculates a revised statistic. See the help page.

## R2m R2c
## [1,] 0.1614303 0.614307

#Obtaining VIF values

vif(LMER_best_fit_model)

## GVIF Df GVIF^(1/(2*Df))
## LANGUAGE 1.940343 1 1.392962
## SEMANTIC 7.556333 3 1.400829
## ZNLD_gmc 2.780622 1 1.667520
## ZPHONETIC_gmc 1.540855 1 1.241312
## ZEnglish_Freq_gmc 1.050145 1 1.024766
## Zlength_TARGET_raw_gmc 1.034999 1 1.017349
## ZTime_ms_gmc 1.000053 1 1.000027
## OTHER_LANGUAGES 1.395945 1 1.181501
## LANGUAGE:SEMANTIC 4.502283 3 1.285007
## ZNLD_gmc:ZPHONETIC_gmc 1.519332 1 1.232612
 222

EFFECTS OF CROSSLINGUISTIC OVERLAP

1/vif(LMER_best_fit_model)

## GVIF Df GVIF^(1/(2*Df))
## LANGUAGE 0.5153727 1.0000000 0.7178946
## SEMANTIC 0.1323393 0.3333333 0.7138629
## ZNLD_gmc 0.3596318 1.0000000 0.5996931
## ZPHONETIC_gmc 0.6489905 1.0000000 0.8055995
## ZEnglish_Freq_gmc 0.9522495 1.0000000 0.9758327
## Zlength_TARGET_raw_gmc 0.9661841 1.0000000 0.9829466
## ZTime_ms_gmc 0.9999467 1.0000000 0.9999734
## OTHER_LANGUAGES 0.7163604 1.0000000 0.8463807
## LANGUAGE:SEMANTIC 0.2221095 0.3333333 0.7782059
## ZNLD_gmc:ZPHONETIC_gmc 0.6581842 1.0000000 0.8112855
 223

EFFECTS OF CROSSLINGUISTIC OVERLAP

Appendix H: RStudio glmer model raw output

GLMER Model

install.packages(c("lmerTest", "optimx", "nloptr", "MuMIn", "r2glmm", "car"))

## Installing packages into '/home/rstudio-user/R/x86_64-pc-linux-gnu-library/3.6'

## (as 'lib' is unspecified)

#Loading required packages

library(foreign)
library(lmerTest)

## Loading required package: lme4

## Loading required package: Matrix

##
## Attaching package: 'lmerTest'

## The following object is masked from 'package:lme4':

##
## lmer

## The following object is masked from 'package:stats':

##
## step
 224

EFFECTS OF CROSSLINGUISTIC OVERLAP

library(optimx)
library(nloptr)
library(MuMIn)
library(r2glmm)
library(car)

## Loading required package: carData

## Registered S3 methods overwritten by 'car':

## method from
## influence.merMod lme4
## cooks.distance.influence.merMod lme4
## dfbeta.influence.merMod lme4
## dfbetas.influence.merMod lme4

#Data optimizer
defaultControl <- list(algorithm="NLOPT_LN_BOBYQA",xtol_rel=1e-6,maxeval=1e5)

nloptwrap2 <- function(fn,par,lower,upper,control=list(),...) {

for (n in names(defaultControl))
if (is.null(control[[n]])) control[[n]] <- defaultControl[[n]]
res <- nloptr(x0=par,eval_f=fn,lb=lower,ub=upper,opts=control,...)
with(res,list(par=solution,
fval=objective,
feval=iterations,
conv=if (status>0) 0 else status,
message=message))
}

#Reading the data file

setwd("/cloud/project/GLMER Model")
 225

EFFECTS OF CROSSLINGUISTIC OVERLAP

Dissertation_Dataset <- read.csv("/cloud/project/Master Datasets/Dissertation_Master_Dataset_GLMER.csv", se

p = ",", dec = ".", encoding = "UTF-8")

#Configuring the data variables

Dissertation_Dataset$RT <- as.numeric(Dissertation_Dataset$RT)
Dissertation_Dataset$invRT <- as.numeric(Dissertation_Dataset$invRT)
Dissertation_Dataset$PAIR <- as.factor(Dissertation_Dataset$PAIR)
Dissertation_Dataset$LIST <- as.factor(Dissertation_Dataset$LIST)
Dissertation_Dataset$ZTime <- as.numeric(Dissertation_Dataset$ZTime)
Dissertation_Dataset$Time <- as.numeric(Dissertation_Dataset$Time)
Dissertation_Dataset$AGE <- as.numeric(Dissertation_Dataset$AGE)
Dissertation_Dataset$GENDER <- as.factor(Dissertation_Dataset$GENDER)
Dissertation_Dataset$SEMANTIC <- as.factor(Dissertation_Dataset$SEMANTIC)
Dissertation_Dataset$SEMANTIC_OLD <- as.factor(Dissertation_Dataset$SEMANTIC_OLD)
Dissertation_Dataset$SEMANTIC_ORIGINAL <- as.factor(Dissertation_Dataset$SEMANTIC_ORIGINAL)
Dissertation_Dataset$ZPHONETIC <- as.numeric(Dissertation_Dataset$ZPHONETIC)
Dissertation_Dataset$PHONETIC_RAW <- as.numeric(Dissertation_Dataset$PHONETIC_RAW)
Dissertation_Dataset$PARTICIPANT <- as.factor(Dissertation_Dataset$PARTICIPANT)
Dissertation_Dataset$TARGET <- as.factor(Dissertation_Dataset$TARGET)
Dissertation_Dataset$PRIME <- as.factor(Dissertation_Dataset$PRIME)
Dissertation_Dataset$ZEnglish_Freq <- as.numeric(Dissertation_Dataset$ZEnglish_Freq)
Dissertation_Dataset$English_Freq <- as.numeric(Dissertation_Dataset$English_Freq)
Dissertation_Dataset$ZPortuguese_Freq <- as.numeric(Dissertation_Dataset$ZPortuguese_Freq)
Dissertation_Dataset$Portuguese_Freq <- as.numeric(Dissertation_Dataset$Portuguese_Freq)
Dissertation_Dataset$logPortuguese_Freq <- as.numeric(Dissertation_Dataset$logPortuguese_Freq)
Dissertation_Dataset$logEnglish_Freq <- as.numeric(Dissertation_Dataset$logEnglish_Freq)
Dissertation_Dataset$ZPROFICIENCY <- as.numeric(Dissertation_Dataset$ZPROFICIENCY)
Dissertation_Dataset$SEMAC <- as.numeric(Dissertation_Dataset$SEMAC)
Dissertation_Dataset$ERS <- as.numeric(Dissertation_Dataset$ERS)
Dissertation_Dataset$GS <- as.numeric(Dissertation_Dataset$GS)
 226

EFFECTS OF CROSSLINGUISTIC OVERLAP

Dissertation_Dataset$OS <- as.numeric(Dissertation_Dataset$OS)

Dissertation_Dataset$Lev <- as.numeric(Dissertation_Dataset$Lev)
Dissertation_Dataset$NLD <- as.numeric(Dissertation_Dataset$NLD)
Dissertation_Dataset$ZGS <- as.numeric(Dissertation_Dataset$ZGS)
Dissertation_Dataset$ZOS <- as.numeric(Dissertation_Dataset$ZOS)
Dissertation_Dataset$ZLev <- as.numeric(Dissertation_Dataset$ZLev)
Dissertation_Dataset$ZNLD <- as.numeric(Dissertation_Dataset$ZNLD)
Dissertation_Dataset$length_prime_raw <- as.numeric(Dissertation_Dataset$length_prime_raw)
Dissertation_Dataset$length_TARGET_raw <- as.numeric(Dissertation_Dataset$length_TARGET_raw)
Dissertation_Dataset$Zlength_prime_raw <- as.numeric(Dissertation_Dataset$Zlength_prime_raw)
Dissertation_Dataset$Zlength_TARGET_raw <- as.numeric(Dissertation_Dataset$Zlength_TARGET_raw)
Dissertation_Dataset$English_Freq_gmc <- as.numeric(Dissertation_Dataset$English_Freq_gmc)
Dissertation_Dataset$Portuguese_Freq_gmc <- as.numeric(Dissertation_Dataset$Portuguese_Freq_gmc)
Dissertation_Dataset$SEMAC_gmc <- as.numeric(Dissertation_Dataset$SEMAC_gmc)
Dissertation_Dataset$English_Freq_log_gmc <- as.numeric(Dissertation_Dataset$English_Freq_log_gmc)
Dissertation_Dataset$Portuguese_Freq_log_gmc <- as.numeric(Dissertation_Dataset$Portuguese_Freq_log_gmc)
Dissertation_Dataset$length_TARGET_raw_log_gmc <- as.numeric(Dissertation_Dataset$length_TARGET_raw_log_gmc
)
Dissertation_Dataset$Englisg_Freq_gmc <- as.numeric(Dissertation_Dataset$English_Freq_gmc)
Dissertation_Dataset$NLD_gmc <- as.numeric(Dissertation_Dataset$NLD_gmc)
Dissertation_Dataset$NLD_log <- as.numeric(Dissertation_Dataset$NLD_log)
Dissertation_Dataset$NLD_log_gmc <- as.numeric(Dissertation_Dataset$NLD_log_gmc)
Dissertation_Dataset$PHONETIC_gmc <- as.numeric(Dissertation_Dataset$PHONETIC_gmc)
Dissertation_Dataset$PHONETIC_RAW_gmc <- as.numeric(Dissertation_Dataset$PHONETIC_RAW_gmc)
Dissertation_Dataset$PHONETIC_gmc <- as.numeric(Dissertation_Dataset$PHONETIC_gmc)
Dissertation_Dataset$PHONETIC_log <- as.numeric(Dissertation_Dataset$PHONETIC_log)
Dissertation_Dataset$PHONETIC_RAW_log_gmc <- as.numeric(Dissertation_Dataset$PHONETIC_RAW_log_gmc)
Dissertation_Dataset$AGE_gmc <- as.numeric(Dissertation_Dataset$AGE_gmc)
Dissertation_Dataset$Time_ms_gmc <- as.numeric(Dissertation_Dataset$Time_ms_gmc)
Dissertation_Dataset$ERS_gmc <- as.numeric(Dissertation_Dataset$ERS_gmc)
 227

EFFECTS OF CROSSLINGUISTIC OVERLAP

#Releveling Pertinent Variables

Dissertation_Dataset$SEMANTIC<-relevel(Dissertation_Dataset$SEMANTIC,"CTRL")
Dissertation_Dataset$LANGUAGE<-relevel(Dissertation_Dataset$LANGUAGE,"Functional Monolinguals")
Dissertation_Dataset$RAW_ACCURACY<-relevel(Dissertation_Dataset$RAW_ACCURACY,"incorrect")

#Running best fit model

GLMER_best_fit_model <- glmer(RAW_ACCURACY ~ LANGUAGE*SEMANTIC + ZEnglish_Freq_gmc + ZTime_ms_gmc + Zlength
_TARGET_raw_gmc + ZSEMAC_gmc + (1| PARTICIPANT) + (1|PRIME) + (1|TARGET) + (1|PAIR), data = Dissertation_Da
taset, family=binomial, control=glmerControl(optimizer="nloptwrap2"))

#Obtaining model summary

summary(GLMER_best_fit_model)

## Generalized linear mixed model fit by maximum likelihood (Laplace

## Approximation) [glmerMod]
## Family: binomial ( logit )
## Formula:
## RAW_ACCURACY ~ LANGUAGE * SEMANTIC + ZEnglish_Freq_gmc + ZTime_ms_gmc +
## Zlength_TARGET_raw_gmc + ZSEMAC_gmc + (1 | PARTICIPANT) +
## (1 | PRIME) + (1 | TARGET) + (1 | PAIR)
## Data: Dissertation_Dataset
## Control: glmerControl(optimizer = "nloptwrap2")
##
## AIC BIC logLik deviance df.resid
## 12955.9 13090.6 -6462.0 12923.9 33412
##
## Scaled residuals:
## Min 1Q Median 3Q Max
## -19.6469 0.0663 0.1250 0.2319 2.8842
##
 228

EFFECTS OF CROSSLINGUISTIC OVERLAP

## Random effects:
## Groups Name Variance Std.Dev.
## PRIME (Intercept) 0.2766 0.5259
## PAIR (Intercept) 0.3275 0.5723
## TARGET (Intercept) 1.5102 1.2289
## PARTICIPANT (Intercept) 0.6580 0.8112
## Number of obs: 33428, groups:
## PRIME, 901; PAIR, 852; TARGET, 591; PARTICIPANT, 57
##
## Fixed effects:
## Estimate Std. Error z value Pr(>|z|)
## (Intercept) 4.87710 0.23429 20.817 < 2e-16 ***
## LANGUAGEBilinguals -0.61333 0.29588 -2.073 0.038183 *
## SEMANTICCOG 0.28415 0.23245 1.222 0.221563
## SEMANTICDISTR -1.37249 0.19154 -7.166 7.74e-13 ***
## SEMANTICFF 0.18464 0.26908 0.686 0.492601
## ZEnglish_Freq_gmc 0.26429 0.08528 3.099 0.001941 **
## ZTime_ms_gmc 0.17440 0.03395 5.136 2.80e-07 ***
## Zlength_TARGET_raw_gmc 0.22030 0.07301 3.017 0.002550 **
## ZSEMAC_gmc 0.49578 0.13141 3.773 0.000161 ***
## LANGUAGEBilinguals:SEMANTICCOG 0.53983 0.24560 2.198 0.027950 *
## LANGUAGEBilinguals:SEMANTICDISTR -0.13628 0.15150 -0.900 0.368368
## LANGUAGEBilinguals:SEMANTICFF -0.34842 0.26453 -1.317 0.187791
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
##
 229

EFFECTS OF CROSSLINGUISTIC OVERLAP

Correlation of Fixed Effects:

## (Intr) LANGUAGEBl SEMANTICC SEMANTICD SEMANTICF ZEn_F_
## LANGUAGEBln -0.685
## SEMANTICCOG -0.321 0.195
## SEMANTICDIS -0.580 0.247 0.394
## SEMANTICFF -0.271 0.169 0.220 0.330
## ZEnglsh_Fr_ -0.039 -0.004 0.006 0.174 -0.015
## ZTim_ms_gmc 0.026 -0.007 -0.010 -0.005 0.005 -0.011
## Zl_TARGET__ 0.025 0.003 -0.032 -0.011 0.050 0.152
## ZSEMAC_gmc -0.293 0.492 -0.001 -0.003 0.000 0.001
## LANGUAGEB:SEMANTICC 0.241 -0.259 -0.698 -0.293 -0.204 0.004
## LANGUAGEB:SEMANTICD 0.396 -0.430 -0.384 -0.572 -0.332 0.009
## LANGUAGEB:SEMANTICF 0.219 -0.242 -0.217 -0.268 -0.734 0.006
## ZTm_m_ Z_TARG ZSEMAC LANGUAGEB:SEMANTICC
## LANGUAGEBln
## SEMANTICCOG
## SEMANTICDIS
## SEMANTICFF
## ZEnglsh_Fr_
## ZTim_ms_gmc
## Zl_TARGET__ 0.002
## ZSEMAC_gmc 0.025 -0.002
## LANGUAGEB:SEMANTICC 0.005 0.002 0.002
## LANGUAGEB:SEMANTICD -0.002 0.008 -0.002 0.511
## LANGUAGEB:SEMANTICF -0.001 0.003 0.000 0.291
## LANGUAGEB:SEMANTICD
## LANGUAGEBln
## SEMANTICCOG
 230

EFFECTS OF CROSSLINGUISTIC OVERLAP

## SEMANTICDIS
## SEMANTICFF
## ZEnglsh_Fr_
## ZTim_ms_gmc
## Zl_TARGET__
## ZSEMAC_gmc
## LANGUAGEB:SEMANTICC
## LANGUAGEB:SEMANTICD
## LANGUAGEB:SEMANTICF 0.474

#Obtaining confidence Intervals

lower <- coef(summary(GLMER_best_fit_model))[,1] + qnorm(.025)*coef(summary(GLMER_best_fit_model))[,2]
upper <- coef(summary(GLMER_best_fit_model))[,1] + qnorm(.975)*coef(summary(GLMER_best_fit_model))[,2]
cbind(coef(summary(GLMER_best_fit_model)), lower, upper)

## Estimate Std. Error z value Pr(>|z|)

## (Intercept) 4.8770982 0.23428968 20.8165304 3.065976e-96
## LANGUAGEBilinguals -0.6133302 0.29588170 -2.0728900 3.818252e-02
## SEMANTICCOG 0.2841483 0.23245440 1.2223829 2.215629e-01
## SEMANTICDISTR -1.3724860 0.19153675 -7.1656537 7.741623e-13
## SEMANTICFF 0.1846397 0.26908458 0.6861772 4.926014e-01
## ZEnglish_Freq_gmc 0.2642935 0.08527900 3.0991633 1.940680e-03
## ZTime_ms_gmc 0.1744004 0.03395427 5.1363335 2.801504e-07
## Zlength_TARGET_raw_gmc 0.2203011 0.07301016 3.0174025 2.549510e-03
## ZSEMAC_gmc 0.4957788 0.13141050 3.7727487 1.614589e-04
## LANGUAGEBilinguals:SEMANTICCOG 0.5398262 0.24560000 2.1979895 2.794985e-02
## LANGUAGEBilinguals:SEMANTICDISTR -0.1362771 0.15149746 -0.8995339 3.683684e-01
## LANGUAGEBilinguals:SEMANTICFF -0.3484214 0.26452814 -1.3171431 1.877907e-01
## lower upper
 231

EFFECTS OF CROSSLINGUISTIC OVERLAP

## (Intercept) 4.41789884 5.33629750

## LANGUAGEBilinguals -1.19324768 -0.03341275
## SEMANTICCOG -0.17145397 0.73975055
## SEMANTICDISTR -1.74789119 -0.99708091
## SEMANTICFF -0.34275640 0.71203579
## ZEnglish_Freq_gmc 0.09714978 0.43143730
## ZTime_ms_gmc 0.10785130 0.24094959
## Zlength_TARGET_raw_gmc 0.07720376 0.36339835
## ZSEMAC_gmc 0.23821894 0.75333862
## LANGUAGEBilinguals:SEMANTICCOG 0.05845907 1.02119339
## LANGUAGEBilinguals:SEMANTICDISTR -0.43320667 0.16065247
## LANGUAGEBilinguals:SEMANTICFF -0.86688703 0.17004422

#Obtaining R2
r.squaredGLMM(GLMER_best_fit_model)

## Warning: 'r.squaredGLMM' now calculates a revised statistic. See the help page.

## Warning: The null model is correct only if all variables used by the original
## model remain unchanged.

## R2m R2c
## theoretical 0.1938626 0.5625173
## delta 0.0744979 0.2161653
 232

EFFECTS OF CROSSLINGUISTIC OVERLAP

#Obtaining VIF values

vif(GLMER_best_fit_model)

## GVIF Df GVIF^(1/(2*Df))
## LANGUAGE 1.756417 1 1.325299
## SEMANTIC 5.929369 3 1.345348
## ZEnglish_Freq_gmc 1.087451 1 1.042809
## ZTime_ms_gmc 1.001749 1 1.000874
## Zlength_TARGET_raw_gmc 1.036538 1 1.018105
## ZSEMAC_gmc 1.425754 1 1.194050
## LANGUAGE:SEMANTIC 6.720202 3 1.373716

1/vif(GLMER_best_fit_model)

## GVIF Df GVIF^(1/(2*Df))
## LANGUAGE 0.5693409 1.0000000 0.7545468
## SEMANTIC 0.1686520 0.3333333 0.7433019
## ZEnglish_Freq_gmc 0.9195813 1.0000000 0.9589480
## ZTime_ms_gmc 0.9982537 1.0000000 0.9991265
## Zlength_TARGET_raw_gmc 0.9647503 1.0000000 0.9822170
## ZSEMAC_gmc 0.7013831 1.0000000 0.8374862
## LANGUAGE:SEMANTIC 0.1488050 0.3333333 0.7279523