ABSTRACT 
 
 
 

 
Title of Dissertation: HOW BILINGUALS' COMPREHENSION OF 

CODE-SWITCHES INFLUENCES 
ATTENTION AND MEMORY 

  
 Lauren Kathleen Salig, 

Doctor of Philosophy, 2024  
  
Dissertation directed by: Associate Professor Jared Novick,  

Department of Hearing and Speech Sciences  
 

Associate Professor L. Robert Slevc, 
Department of Psychology 

 
 

Bilinguals sometimes code-switch between their shared languages. While 

psycholinguistics research has focused on the challenges of comprehending code-switches 

compared to single-language utterances, bilinguals seem unhindered by code-switching in 

communication, suggesting benefits that offset the costs. I hypothesize that bilinguals orient their 

attention to speech content after hearing a code-switch because they draw a pragmatic inference 

about its meaning. This hypothesis is based on the pragmatic meaningfulness of code-switches, 

which speakers may use to emphasize information, signal their identity, or ease production 

difficulties, inter alia. By considering how code-switches may benefit listeners, this research 

attempts to better align our psycholinguistic understanding of code-switch processing with actual 

bilingual language use, while also inspiring future work to investigate how diverse language 

contexts may facilitate learning in educational settings. 



  

In this dissertation, I share the results of three pre-registered experiments with Spanish-

English bilinguals that evaluate how hearing a code-switch affects attention and memory. 

Experiment 1a shows that code-switches increase bilinguals’ self-reported attention to speech 

content and improve memory for that information, compared to single-language equivalents. 

Experiment 1b demonstrates that this effect requires bilingual experience, as English-speaking 

monolinguals did not demonstrate increased attention upon hearing a code-switch.  

Experiment 2 attempts to replicate these results and establish the time course of the 

attentional effect using an EEG measure previously associated with attentional engagement (alpha 

power). However, I conclude that alpha power was not a valid measure of attention to speech 

content in this experiment.  

In Experiment 3, bilinguals again showed better memory for information heard in a code-

switched context, with a larger benefit for those with more code-switching experience and when 

listeners believed the code-switches were natural (as opposed to inserted randomly, removing the 

element of speaker choice). This suggests that the memory benefit comes from drawing a 

pragmatic inference, which likely requires prior code-switching experience and a belief in code-

switches’ communicative purpose. 

These experiments establish that bilingual listeners derive attentional and memory benefits 

from ecologically valid code-switches—challenging a simplistic interpretation of the traditional 

finding of “costs.” Further, these findings motivate future applied work assessing if/how code-

switches might benefit learning in educational contexts. 

 

 

 
 



  

 
 
 
 
 
 
 
 

HOW BILINGUALS' COMPREHENSION OF CODE-SWITCHES INFLUENCES 
ATTENTION AND MEMORY  

 
 
 

by 
 

 
 
 
 
 
 

Dissertation submitted to the Faculty of the Graduate School of the  
University of Maryland, College Park, in partial fulfillment 

of the requirements for the degree of 
Doctor of Philosophy 

2024 
 
 
 
 
 
 
 
 
 
 
Advisory Committee: 
Associate Professor Jared Novick, Co-chair 
Associate Professor L. Robert Slevc, Co-chair 
Associate Professor Jorge Valdés Kroff, Special Member 
Professor Samira Anderson 
Associate Professor Ellen Lau, Dean’s Representative 

 
 
 

Lauren Kathleen Salig 



  

 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

© Copyright by 
Lauren Kathleen Salig 

2024 
 
 
 
 
 
 
 
 
 
 
 
 
 



 

 

ii 
 

Foreword 
 

The research included in this dissertation was highly collaborative, with much 

intellectual input from L. Robert Slevc, Jared Novick, and Jorge Valdés Kroff. In 

particular, note that Chapter 2 of this dissertation is a manuscript that was co-authored 

by Lauren Salig with Jorge Valdés Kroff, L. Robert Slevc, and Jared Novick. Lauren 

Salig was the first author of all drafts of the manuscript and made substantial 

contributions to the manuscript. Very minor edits have been made to the co-authored 

manuscript to fit this dissertation. A letter certifying Lauren’s substantial contribution 

to any co-authored work and the committee’s approval of its inclusion is on the 

following page. 

 

 

 

 

 

 

 

 

 

 

 

 



 

 

iii 
 

 

  



 

 

iv 
 

Acknowledgements 

I am extremely grateful to the many people who supported me academically 

and emotionally as I worked towards this dissertation. I have been so lucky to be 

surrounded by incredible academic communities and the best family and friends. Thank 

you for not only “getting me through” these past five years but for making them 

genuinely fun and insightful.  

 

First, I want to acknowledge my amazing advisors. Jared, thank you for always 

being so excited by our research and believing so steadfastly in its value. When I lost 

sight of what I was trying to do or got bogged down by details, your enthusiasm 

reminded me why I was here and why I love doing research. Thank you, also, for being 

such a caring advisor and making sure I put my wellbeing above everything else. This 

sounds simple, but it’s not; it was obvious that this was your first priority and never an 

afterthought, and that type of support helped me immensely. And, of course, thank you 

for prompting me to provide more “connective tissue” in my writing. I joke about it, 

but it has made me a better science communicator. If I could offer you one suggestion 

for improvement, it would be to upgrade your cornhole set. 

Bob, it is so fun to do research with you. Thank you for being up for anything 

and for being the one to keep asking questions that push my thinking further along. 

Sometimes you did make me question the entire basis of my research agenda, and even 

though that could be scary, it made me think more deeply and become a better scientist. 

Your willingness to joke around, to be wrong about things, and to ask questions about 

what you don’t understand made me feel safe and reminded me that part of the fun of 



 

 

v 
 

science is being wrong and learning from that. Thank you also for being willing to 

hunker down and problem-solve with me. When I was banging my head against a wall 

trying to learn new technical skills (especially in R), it was amazing to have you sit 

down and actively work with me to figure out a solution. It was really nice to be able 

to come to a meeting with any sort of problem—from a theoretical question to a coding 

error—and know that I could ask for help.  

Jorge, I will forever be grateful to you. You didn’t need to act as my advisor 

and mentor over the past five years, but you did, going above and beyond being just a 

collaborator. I have learned so much from you about how to approach research in a way 

that challenges the status quo and that steps outside of the siloed view of an individual 

field. When you un-muted in a Zoom call, I could be sure that I was about to hear an 

insightful perspective. Thank you, also, for being such a wonderful collaborator and 

co-writer; you give the most supportive feedback. I always felt that we were on the 

same team, working together to make the research as good as it could be. 

To all of my advisors, thank you for always filling the first ten minutes or so of 

our meetings with small talk about penguins, references from before my time, and (my 

personal favorite) commentary about the small talk itself. We wasted a lot of valuable 

time making our meetings a comfortable, fun space; thank you for that. 

 

Second, I want to thank all of the people who have provided academic support 

to me over the years. Thank you to Shevaun Lewis, Caitlin Eaves, Colin Phillips, and 

everyone else who worked to create an incredible community at the Language Science 



 

 

vi 
 

Center, where I found my academic home at UMD. Thank you to my NACS and HESP 

cohorts for your camaraderie.  

A big thank you to all of my lab mates—Rachel Thompson, Kelly Marshall, 

Mine Muezzinoglu, Zoe Ovans, Zach Maher, Tal Ness, Yi Wei, and Amy Gionfriddo—

for being there when I had questions about study design, data analysis, funding, and 

life in general. Your support and help have meant a lot to me. Thank you to Yi Wei and 

Tal Ness for so kindly going out of your way to help me learn how to analyze EEG data 

and to Mattson Ogg for teaching me how to cap an EEG participant so that actual signal 

can be recorded rather than a whole lot of nothing.  

Thank you to Al Kim and the members of his lab in Colorado for your 

thoughtful insight on my research at all stages and especially for feedback on my EEG 

research.  

Thank you, Eleonora Rossi, Edith Kaan, and Keng-Yu Lin for sitting down and 

teaching me how to actually design, program, run, and analyze an EEG study. You took 

something that felt overwhelming and opaque and made it approachable. And an 

additional thank you to Eleonora and Edith for your collaboration in my work.  

Thank you to Pam Komarek (and recently to Claire Morse as well) for all that 

you do to keep the NACS program running smoothly. Pam, when you are involved, 

things get done, and that made a huge difference throughout my time in the program.  

Thank you to my committee members: Ellen and Samira. I so appreciate the 

thoughtful questions you have asked to encourage me to improve my research as well 

as your encouraging enthusiasm. 



 

 

vii 
 

And thank you to Kim Epting, who introduced me to the idea of studying 

language scientifically and who was simply the best undergraduate mentor I could have 

ever had. You set a high bar for who I want to work with and be mentored by, and this 

standard has served me well. 

 

 Third, I want to thank everyone who has provided practical support as I did this 

research. Tracy Riggins, thank you for allowing me to use your EEG equipment and 

for tolerating my research assistants and I meandering through your lab meetings as we 

tested participants. Your generosity opened up new research opportunities for me.  

 I need to acknowledge that none of this research would have been possible 

without the support of so many research assistants who I loved working with. To Alex 

Medrano, Chiara Sforza, Ciaran Stone, Dominic Marcinelli, Ethan Steuernagle, Haley 

Tosh, Harini Muthu, Jeannine Lederman, Jeymi Menendez, Julieta Magud, Mia Lulli, 

Michael Egan, Priyal Dhingra, Roxanna Bakhtiari, Sofia Villagomez, Tessa 

Schauermann, and others: Thank you. You are such talented, intelligent individuals. I 

appreciate the insight and support you provided on many research projects (not just the 

ones presented in this dissertation). This research truly would not have been possible 

without your contributions. 

 I also need to acknowledge the many funding sources that made my dissertation 

and my time as a Ph.D. student possible. My research and/or myself has been supported 

by funding from: the National Science Foundation (NSF #1449815), the William Orr 

Dingwall Foundation’s Foundations of Language fellowship, a P.E.O. scholar award, 

the UMD Flagship Fellowship, and by both the Psychology and Hearing and Speech 



 

 

viii 
 

Sciences Departments at UMD as well as by the Neuroscience and Cognitive Science 

program. 

  

 Fourth, I want to thank my graduate student friends and communities that 

supported me through the emotional highs and lows of doing a Ph.D. program. Thank 

you to my fellow language science outreachers—Craig Thorburn, Erika Exton, 

Kathleen Oppenheimer, and Alex Krauska. After the pandemic began shortly into my 

second semester as a Ph.D. student, being part of our outreach group was one of the 

first things that made me feel like I belonged at UMD. It meant a lot to be working 

together with you all towards something that felt meaningful. Thank you for welcoming 

me into the group and for playing around with green screens together while we were 

locked in our homes. 

 I also need to thank my graduate student therapy group. To our moderator, 

Allison Asarch, and to everyone who attended the group from summer 2021 through 

spring 2024, thank you. From our interactions, I learned so much about how to form 

meaningful connections with other people, how to recognize patterns in my 

thoughts/behaviors, and how to be my authentic self. The space that we created together 

each semester meant a lot to me. 

 Thank you, also, to my writing groups. To NACS writing crew (hey Rachel 

Thompson, Gloria Kim, and Deena Shariq), thank you for creating a space for us to 

write together over the past few months; the later parts of my dissertation research and 

writing got done in part during our time together, and you made it a lot more fun. And 

to my writing accountability group, y’all are the best. Since I joined in spring 2020, our 



 

 

ix 
 

group became the place where I felt I most belonged at UMD and where I could ask 

for academic and personal support. To Anna Tinnemore, thank you for being an 

incredible listener and a voice of reason; I strive to be like you in many ways. To Kelly 

Marshall, thank you for tolerating my unsolicited advice about being co-advised and 

for being so optimistic. To Mine Muezzinoglu, thank you for being such an open person 

who is always up for a much-needed chat. To Carly Rosvold and Kristin Hoch, thank 

you for being our newest additions; you fit in perfectly, and it makes me happy to think 

about writing group continuing on long after I graduate. To Rachel Thompson, thank 

you for always pasting adorable little images into our writing spreadsheet—and for so 

much else, which I’ll have to thank you for below. To all of writing group, thank you 

for listening to me complain endlessly and then offering unwavering support. You are 

all kings of the hill in my book, and I will miss our semesterly cookie parties. 

 

 Finally, I want to try my best to thank the family and friends that have supported 

me so much as I worked on this dissertation, although these are the people for whom I 

don’t think I could ever fully express how much their support has meant.  

 Rachel, we started this thing together, and I will always be grateful to have gone 

through grad school with you. In our first few months especially, you were the person 

who made me feel like it was normal to not have it all figured out, even when it sure 

felt like everyone else had it figured out. Thank you for sharing my love for tea, 

homemade cards, and DND. Being able to talk about and bond over non-academic 

interests made life a lot more fun. And being able to talk about the tough things made 

the low points more bearable. I will miss sharing an office with you, drinking copious 



 

 

x 
 

amounts of tea together, and going for our walks, but I look forward to many more 

memes shared back and forth. 

 Katie, I am so glad that you started talking to me on the bus back to Silver 

Spring one day. You are such a kind and welcoming person—I immediately felt at 

home with you as if we’d been friends for years. Thank you for being such a genuine 

listener and for caring so deeply. I have immensely enjoyed talking about bilingualism 

and life with you. And thank you for introducing me to so many fun games that helped 

me relax as I worked on this dissertation (if one can “relax” while playing Overcooked). 

Charles and Amber, despite living so obnoxiously far away, you have made the 

past few years (and many years before that) so much fun. You both have a zest for life 

that, when I’m down, reminds me that there are tons of things to get excited about—

like plants and hats and dressing up for our DND sessions that fall near Halloween. 

Thank you to both of you for always keeping our weekly DND sessions and our annual 

weekend on your calendar; they have often been the highlight of my week and of my 

year. I love you both. 

Sarah, I feel so lucky to have you as my best friend. Thank you for being a 

constant source of support and joy throughout my life, especially over the past few 

years. My trips to Philly and your trips to D.C. gave me something to look forward to 

while I worked on this dissertation. There’s just nothing like sitting on the couch, 

stuffing my face with snacks, and chatting with you over some either really bad or 

really good TV. Being able to talk about silly little things and also big life things with 

you means a lot to me. 



 

 

xi 
 

 To my family—Mom, Dad, Matt, Colin, Alex, Jen, Joe, Kari, and Alyssa—I 

appreciate you and all of your unwavering support and love so much. Thank you for 

making me feel loved and accepted. All of the games, the laughs, and the meals shared 

with you over the past few years are what makes life great. Mom, thank you for being 

a great mother. And thank you for being the one to encourage my love of writing and 

to teach me the importance of editing. Dad, thank you for being a great father. I like to 

think that your analytical brain wore off on me and makes me a better researcher. Both 

of you have always been so supportive of the things I want to do in life, and I don’t 

think I would have made it to where I am today without that support. 

And finally, Brian. Thank you for everything—for moving to Maryland without 

a second thought, for making sure I ate and slept, for dragging me into D.C. to explore 

new things, for going on parking lot laps with me, for vegging out to Dimension 20 

when I needed that, and for planning adventures for us to balance out the couch potato-

y-ness. There are simply too many things to thank you for, but I am so grateful for all 

of them. I could not have done it without you. Since I started grad school, you’ve gone 

from being my partner to my fiancé to my husband. You’ve earned two degrees while 

I worked on this one. I’m proud of both of us. What a journey we’ve been on over the 

past five years. I love you, and I can’t wait to see what’s next. 



 

 

xii 
 

Table of Contents 
 
Foreword ....................................................................................................................... ii 
Acknowledgements ...................................................................................................... iv 
Table of Contents ........................................................................................................ xii 
List of Tables ............................................................................................................. xiv 
List of Figures ............................................................................................................. xv 
Chapter 1: Introduction ................................................................................................. 1 

1.1 Switch Costs ........................................................................................................ 1 
1.1.1 Exploring Nuance in Switch Costs ............................................................... 2 

1.2 Switch Benefits ................................................................................................... 4 
1.2.1 Code-switches and Attention........................................................................ 5 
1.2.2 Overall Approach......................................................................................... 6 

1.3 My Positionality .................................................................................................. 8 
Chapter 2: Establishing a Relationship between Hearing Code-switches and Increased 
Attention and Memory ................................................................................................ 10 

2.1 Introduction ....................................................................................................... 10 
2.1.1 Code-switches and Attention: Three Accounts .......................................... 11 
2.1.2 The Current Experiments: Preliminaries and Predictions ........................ 17 

2.2 Experiment 1a Method ...................................................................................... 18 
2.2.1 Participants ................................................................................................ 18 
2.2.2 Materials and Design ................................................................................. 20 

2.3 Experiment 1a Results ...................................................................................... 26 
2.3.1 Effects of Code-switching on Bilinguals’ Attention ................................... 26 
2.3.2 Effects of Code-switching and Attention on Bilinguals’ Memory .............. 28 
2.3.3 Participant Enjoyment ............................................................................... 32 

2.4 Experiment 1a Discussion................................................................................. 32 
2.5 Experiment 1b Method ..................................................................................... 33 

2.5.1 Participants ................................................................................................ 34 
2.6 Experiment 1b Results ...................................................................................... 35 

2.6.1 Effects of Code-switching on Monolinguals’ Attention ............................. 36 
2.6.2 Effects of Code-switching and Attention on Monolinguals’ Memory ........ 37 

2.7 Experiment 1b Discussion ................................................................................ 39 
2.8 General Discussion ........................................................................................... 41 

2.8.1 Exploring the Source of Bilinguals’ Increased Attention: The Role of 
Stimulus Salience versus Bilingual Experience .................................................. 42 
2.8.2 Are our Attention and Memory Effects Simply Due to Processing a Less-
dominant Language? .......................................................................................... 46 
2.8.3 Comparing the Costs and Benefits of Code-switching .............................. 48 
2.8.4 Broader Impacts and Closing Remarks ..................................................... 49 

Chapter 3: Specifying the Time Course of Code-switches’ Effect on Attention ........ 52 
3.1 Introduction ....................................................................................................... 52 

3.1.1 Why Time Course Matters ......................................................................... 53 
3.1.2 EEG to Address Time Course .................................................................... 55 
3.1.3 Prior Time Frequency Analyses of Code-switch Comprehension ............. 58 
3.1.4 The Current Experiment ............................................................................ 59 



 

 

xiii 
 

3.2 Method .............................................................................................................. 59 
3.2.1 Participants ................................................................................................ 60 
3.2.2 Materials and Design ................................................................................. 62 
3.2.3 Procedure ................................................................................................... 65 
3.2.4 Analysis ...................................................................................................... 67 

3.3 Results ............................................................................................................... 70 
3.3.1 Behavioral Analysis: Code-switches and Memory .................................... 71 
3.3.2 Time Frequency Analysis: Alpha Power in Critical Region ...................... 73 
3.3.3 Time Frequency Analysis: Alpha Power in Pre-critical Region ............... 75 
3.3.4 Time Frequency Analysis: Link between Attention and Memory .............. 76 
3.3.5 ERP Analysis of Critical Word .................................................................. 80 

3.4 Discussion ......................................................................................................... 81 
3.4.1 Does Alpha Track In-the-Moment Attention to External Information? ..... 83 
3.4.2 Why Was There No LPC Effect of Code-Switches? ................................... 90 
3.4.3 Conclusion ................................................................................................. 95 

Chapter 4: Testing the Role of Listeners’ Inferences in Code-switches’ Effect on 
Memory ..................................................................................................................... 101 

4.1 Introduction ..................................................................................................... 101 
4.1.1 The Inference Possibility for Attention and Memory Effects of Code-
switches ............................................................................................................. 102 

4.2 Method ............................................................................................................ 110 
4.2.1 Participants .............................................................................................. 110 
4.2.2 Materials .................................................................................................. 112 
4.2.3 Design and Procedure ............................................................................. 115 
4.2.4 Coding ...................................................................................................... 118 

4.3 Results ............................................................................................................. 119 
4.3.1 Effect of the Manipulations on Recall ...................................................... 119 
4.3.2 Belief Group Manipulation Check ........................................................... 124 
4.3.3 Vignette Type: Narrative versus Conversational ..................................... 127 
4.3.4 Single-language Recall ............................................................................ 129 

4.4 Discussion ....................................................................................................... 130 
4.4.1 Code-switches Boost Memory for Information ........................................ 131 
4.4.2 Bilinguals’ Pragmatic Inferences Contribute to the Effect of Code-switches 
on Attention/Memory ........................................................................................ 132 
4.4.3 Connecting Back to Linguistic Focus ...................................................... 140 
4.4.4 Conclusion ............................................................................................... 142 

Chapter 5: General Discussion.................................................................................. 143 
5.1 Summary of Key Findings and Their Interpretation ....................................... 143 

5.1.1 What Happens When Bilinguals Hear a Code-switch? ........................... 145 
5.2 Considering Costs and Benefits of Code-switches in Comprehension ........... 146 
5.3 Broader Impacts .............................................................................................. 150 
5.4 Additional Future Directions to Address Limitations ..................................... 153 
5.5 Conclusion ...................................................................................................... 157 

Bibliography ............................................................................................................. 158 
 



 

 

xiv 
 

List of Tables 
 

2.1 Bilingual participants’ characteristics (N=92)  ......................................................20 
2.2 Pre-Registered Regression Model Predicting Bilinguals’ Attention from Language 
Context.  .......................................................................................................................27 
2.3 Exploratory Regression Model Predicting Bilinguals’ Comprehension Question 
Accuracy from Language Context Alone  ...................................................................29 
2.4 Pre-Registered Regression Model Predicting Bilinguals’ Comprehension Question 
Accuracy from Language Context and Attention to Information ................................31 
2.5 Regression Model Predicting Monolinguals’ Attention from Language Context .36 
2.6 Logistic Regression Model Predicting Monolinguals’ Comprehension Question 
Accuracy from Language Context and Attention to Information ................................38 
3.1 Participants’ characteristics (N=30) .......................................................................61 
3.2 Example critical code-switches and single-language equivalents. ........................64 
3.3 Regression Model Predicting LPC Magnitude by Condition. ...............................81 
4.1 Participants’ characteristics (N=72) .....................................................................111 
4.2 Pre-Registered Logistic Regression Model Predicting Recall Accuracy (N=72).
....................................................................................................................................121 
 
 

 



 

 

xv 
 

List of Figures 
 

2.1 Study Schematic with Example Stimuli. ...............................................................25 
2.2 Bilinguals’ Attention Responses by Language Context  .......................................27 
2.3 Bilinguals’ Accuracy of Remembering Information. ............................................29 
2.4 Bilinguals’ Memory Accuracy of and Attention to Relevant Information. ...........31 
3.1 Memory Accuracy Based on Sentence Language Context of Information. ..........72 
3.2 Topography Plot of Alpha Differences between Code-switches and Single-
language Equivalents. ..................................................................................................74 
3.3 Time Frequency Representation of Power Differences between Code-switches 
and Single-language Equivalents. ................................................................................75 
3.4 Time Frequency Representation of Power Differences between Correct Epochs and 
Wrong Epochs. .............................................................................................................77 
3.5 Time Frequency Representation of Power Differences between Correct Epochs and 
No Idea Epochs. ...........................................................................................................79 
3.6 Significant Positive Cluster for Alpha Difference Between Correct and No Idea 
Epochs. .........................................................................................................................79 
3.7 Grand Average ERP Difference Wave for Critical Words ....................................81 
4.1 Recall Accuracy by Language Context................................................................120 
4.2 Correlation between Local Code-switch Effect on Recall and BCSP Scores  ....122 
4.3 Recall Accuracy by Language Context and Belief Group ...................................124 
4.4 Participants’ Endorsed Reasons for Speakers’ Code-switching in Experiment, by 
Belief Group...............................................................................................................126 
4.5 Recall Accuracy by Language Context and Vignette Type) ...............................129 
4.6 Single-language Narrative Vignette Recall Accuracy by Vignette Language. ....130 

  



 

 

1 
 

Chapter 1: Introduction 

While language processing research tends to focus on monolinguals, most 

people worldwide speak multiple languages and deal with different processing 

demands than monolinguals (Kroll et al, 2012). For example, in bilingual communities, 

people “code-switch” between shared languages to varying degrees, skillfully weaving 

both languages into their conversations in a way that follows grammatical principles. 

Code-switches can even occur intrasententially (within phrases; Beatty-Martínez & 

Dussias, 2017; Guzzardo Tamargo et al., 2016; Poplack, 1980)—as in example (1) with 

English and Spanish—and this type of code-switching is the focus of this dissertation. 

(1) Here’s a bolígrafo para que puedas firmar. [Here’s a pen so that you can sign.] 

1.1 Switch Costs 

Although such intrasentential code-switches are common, much research shows 

that bilinguals experience processing costs during the cued production of switches (e.g, 

Gollan & Ferreira, 2008) and—as I focus on in this dissertation—during 

comprehension when they encounter code-switches. For example, bilinguals generally 

take longer to comprehend code-switches than single-language equivalents (e.g., 

Altarriba et al., 1996; Bultena et al., 2015). Additionally, in EEG and pupillometry 

studies, bilinguals demonstrate effects that typically indicate increased processing or 

integration difficulty in response to code-switches (e.g., Beatty-Martínez et al., 2021; 

Litcofsky & Van Hell, 2017; Valdés Kroff et al., 2020). This increased difficulty or 

processing time when comprehending code-switches is often called a “switch cost.” 

 



 

 

2 
 

Given that language comprehension is an incremental, predictive process (e.g., 

Allopenna et al., 1998), it may not be surprising that code-switches seem to add 

difficulty for the comprehender. For instance, in example (1), as a listener hears “Here’s 

a…,” they may begin to make predictions about the unfolding speech—expecting an 

English noun or, given context, perhaps the specific word “pen” to follow. Instead, 

“bolígrafo” arrives in the unfolding speech. In code-switching contexts, there may be 

greater ambiguity/variability in language, it may become more difficult to make 

accurate predictions during comprehension, and there may be increased difficulty 

associated with switching from comprehending one language to comprehending 

another. This narrative of code-switches as costly and difficult in comprehension is 

prevalent in the relevant psycholinguistic literature and also aligns with the general 

educational policy in the United States to use a single-language (English) in the 

classroom, recommending that bilinguals keep their languages separated (see e.g., 

Mallikarjun et al., 2017). 

1.1.1 Exploring Nuance in Switch Costs 

Despite the well-established costs associated with processing a code-switch, 

many bilinguals engage in this behavior frequently without apparent impedances to 

communication (e.g., Poplack, 1980). How does communication remain successful 

despite the purported costs of code-switches? One possibility is that code-switches in 

natural conversations are not so costly to process, despite evidence for switch costs in 

tightly controlled laboratory studies. Psycholinguistic studies often ask participants to 

listen to or read isolated sentences with little to no context about who is producing the 

sentences and in what situation they are being produced. This contrasts with the 



 

 

3 
 

contexts in which code-switches are generally used, where bilinguals generally know 

the topic of conversation and have background information about the producer.  

Indeed, there is psycholinguistic evidence demonstrating that more ecologically 

valid code-switches are easier to process. For example, bilinguals have reduced or no 

switch costs for the types of code-switches that are more frequent in bilingual 

production, compared to the types of code-switches that are rarely used in conversation 

(Beatty-Martínez & Dussias, 2017; Dussias et al., 2014; Guzzardo Tamargo et al., 

2016; Salig et al., 2024; Valdés Kroff et al., 2018). Further, switch costs seem to be 

reduced when processing code-switches in the presence of another bilingual, compared 

to in the contextually-odd presence of a monolingual (Kaan et al., 2020). Thus, 

comprehending code-switches may not be particularly costly when the context is 

ecologically valid.  

Even if/when code-switches are costly, the timescale of those costs may be 

important to consider. Switch costs have generally been measured on the scale of 

milliseconds (e.g., taking ~100ms longer to read code-switched content compared to 

single-language content; Salig et al., 2024). But taking slightly longer to process 

information is not always a detrimental behavior. For example, some work (considering 

comprehension of a single language alone) has shown that when readers spend more 

time on information, they tend to remember it better (e.g., Birch & Rayner, 1997; 

Hartley et al., 1994), suggesting that “costs” in the moment may be balanced by later 

benefits. Thus, in this dissertation, I consider how purported switch costs may be offset 

by benefits of processing code-switches, resulting in successful communication despite 

costs. In doing so, I specifically consider how naturalistic code-switches may offer 



 

 

4 
 

benefits to comprehenders on a longer time scale that spans from seconds to minutes—

moving beyond isolated-sentence studies in the lab that focus on millisecond-level 

switch costs to better understand how bilinguals may capitalize on ecologically valid 

code-switches to aid in comprehension. 

1.2 Switch Benefits 

While the idea that code-switches may have benefits is novel within 

psycholinguistic/cognitive studies of comprehension, benefits of code-switching have 

long been acknowledged within the sociolinguistics and educational fields. Code-

switches are noted as being meaningful speech acts that can: signal an individual’s 

identity, build rapport, highlight important information, be used for interjection, ease 

production demands for the speaker, and play a role in storytelling, among many other 

uses (Beatty Martínez et al., 2020; Bhatt & Bolonyai, 2011; Cashman, 2005; De La 

Cruz, 2022; Gumperz, 1982; MacSwan, 2019).  

I ask if bilingual listeners can capitalize on the social/pragmatic meaningfulness 

of code-switches to gain benefits to comprehension. Given that code-switches can 

convey nuanced pragmatic meanings (signaling identity, interjecting, highlighting 

information, etc.), bilingual listeners may be more attentive when they encounter code-

switches, knowing that code-switches can carry important pragmatic information. 

Moreover, code-switches tend to co-occur with low-frequency or surprising content 

(Calvillo et al., 2020; Myslín & Levy, 2015), which may again warrant additional 

attention if bilingual listeners have learned this pattern.  

Indeed, one recent study suggests that bilinguals can learn this pattern: 

Bilingual listeners began predicting upcoming low-frequency referents after hearing a 



 

 

5 
 

code-switch (Tomić & Valdés Kroff, 2022), contrasting with the typical tendency to 

predict high-frequency referents in a single-language context (Dahan et al., 2001). This 

finding demonstrates that bilinguals seem able to learn associations between code-

switches and upcoming information and can apply that knowledge during in-the-

moment processing. While this past study focuses on how bilinguals successfully 

exploit their knowledge of code-switching in predictive processing, I focus on how 

code-switches might affect bilinguals’ attention and memory. 

If bilinguals are able to draw on their code-switching experience effectively 

during comprehension, then bilingual listeners have a number of sensible reasons to 

increase their attention to linguistic information when they hear a code-switch (e.g., 

anticipating upcoming surprising or pragmatically important information). Further, 

paying increased attention to information around a code-switch may result in more 

thorough processing of that information, making it more likely to be encoded and 

remembered. Thus, I investigate if hearing code-switches may help bilinguals 

beneficially orient their attention during comprehension to allow them to better gather, 

encode, and remember information. 

1.2.1 Code-switches and Attention 

In this dissertation, I hypothesize that upon hearing a code-switch, bilinguals 

draw an inference about the pragmatic information that a code-switch carries and that 

this inference orients attention to speech content to better encode information near the 

code-switch. That is, I predict that through their language experience, bilinguals have 

learned the many pragmatic reasons why speakers code-switch (e.g., to emphasize 

information; Gumperz et al., 1982) and that the inference they draw upon hearing a 



 

 

6 
 

naturalistic code-switch orients their attention to what is being said—allowing them to 

optimally orient their attention in the seconds after a code-switch in a way that benefits 

their later memory for information on the order of minutes or hours. 

Of course, it is also possible that code-switches could have a detrimental or 

simply no effect on bilinguals’ attention to or memory for speech content. In this 

dissertation, I address multiple possibilities for how and why code-switches may affect 

attention and memory by presenting three experimental approaches to assessing the 

link (if any) between hearing code-switches and bilingual listeners’ attention and 

memory. In Chapter 2, I outline different accounts for how hearing a code-switch may 

affect a bilinguals’ attention and memory as well as provide evidence that bilinguals 

capitalize on their language experience upon hearing a code-switch to increase their 

attention to and memory for key information in a way that monolinguals are unable to 

do. In Chapter 3, I present the findings of an electroencephalography (EEG) study in 

which bilinguals listened to naturalistic code-switches and single-language equivalents, 

attempting to use alpha oscillatory power to operationalize attention. In Chapter 4, I 

present further evidence that code-switched information is encoded and recalled better 

than single-language information and additionally show that this effect is largest when 

bilinguals believe code-switches are naturally produced and when bilinguals have more 

code-switching experience—which I take as support for my hypothesis that pragmatic 

inferences cause the attentional orientation effect that code-switches have. 

1.2.2 Overall Approach 

Throughout this dissertation, I report the results of a number of experiments, all 

of which take a similar overall approach to addressing how bilinguals’ linguistic and 



 

 

7 
 

other cognitive processes interact with each other. First, I aim to understand how 

bilinguals differentially orient their attention and encode information from moment-to-

moment based on the current language context. That is, I study how bilinguals engage 

their cognitive processes differently depending on the current state of their 

environment, rather than studying how bilinguals might differ from monolinguals as a 

whole on any given trait (for more on this distinction between states and traits, see 

Salig et al., 2021). Second, I start from the reality that bilinguals do code-switch and 

attempt to align our theoretical understanding of how bilinguals process code-switches 

with that reality by looking for benefits of processing code-switches. Thus, I strive to 

make the code-switches in my studies as ecologically valid as possible to better 

understand how code-switches are processed in real-world interactions, while still 

maintaining the controls needed to draw meaningful psycholinguistic conclusions. 

Along with this, I also look for effects of code-switches on a longer time scale than the 

typical millisecond-level time scale on which switch costs are found in an attempt to 

provide a fuller picture of how code-switches may engage other cognitive processes in 

beneficial ways as comprehension unfolds. Through this work, I hope to better align 

our sociolinguistic and cognitive understandings of how code-switches are 

comprehended while also working to better understand how code-switching contexts 

may affect learning, with the hope that future research will continue to flesh out how 

to best support students of many linguistic backgrounds. 

 



 

 

8 
 

1.3 My Positionality 

Having laid out the overall academic approach that I have intentionally taken 

in this research, it is also important to recognize my personal identity, which has no 

doubt also affected my approach to this research. I am a white, cisgender, able-bodied 

woman who was raised in the northeast of the United States in an English-speaking 

household. I have learned English as my first language from birth and began learning 

Spanish as a second language in school at the age of 12. I formally studied Spanish in 

school for ten years, including a semester in Santiago, Chile, and then spent 9 months 

living in Neuquén, Argentina. Through these experiences, I became fluent, although 

not native-sounding, in Spanish and consider myself to be a second-language learner 

of Spanish and an English-dominant English-Spanish bilingual. Although I consider 

myself to be bilingual, I currently use English almost exclusively and rarely code-

switch in my daily social interactions.  

Given my language background, it is important to note that I am the type of 

bilingual (white, native-English speaking, school-based learner of a Romance 

language) that often is highly-valued in our society but is not representative of the 

bilingual experience in the United States as a whole or of the bilingual participants in 

my research. Although I am intentionally inclusive in how I define who is a Spanish-

English bilingual, many of the Spanish-English bilinguals who participate in my 

research are heritage speakers of Spanish who learned Spanish as a home language and 

English as a school/societal language, which is common in the United States and is not 

an experience that I personally have. In my dissertation, I strive to uncover benefits of 

code-switching (a common bilingual language behavior) and shift away from deficit 



 

 

9 
 

framing of code-switches as costly/difficult, aiming to conduct cognitive research that 

may one day help support bilingual students in their education/learning. Despite this 

intention and the immensely valuable collaboration/feedback of bilingual colleagues 

and mentors, my personal identity affects my research. I encourage consideration and 

criticism of how my white, English-dominant identity may limit this research and of 

how my future research could implement changes to better align my research actions 

with my research goals. 

 
 

 

 



 

 

10 
 

Chapter 2: Establishing a Relationship between Hearing Code-
switches and Increased Attention and Memory 

2.1 Introduction 

During conversations, bilinguals may naturally switch between their languages, 

a behavior known as code-switching. This phenomenon involves alternations from one 

language to another, requiring cross-linguistic integration across multiple levels, such 

as phonological, morphosyntactic, and semantic (Poplack, 1980). While code-

switching is a natural use of a bilingual speaker’s full linguistic repertoire (Beatty-

Martínez et al., 2020; Otheguy et al., 2015), it can introduce processing challenges for 

bilingual comprehenders. For instance, bilinguals take longer to read code-switched 

sentences compared to single-language equivalents (Altarriba et al., 1996; Bultena et 

al., 2015), and encountering a code-switch during real-time comprehension generates 

electrophysiological brain activity that is associated with processing difficulty (e.g., 

Litcofsky & Van Hell, 2017). These costs have been attributed to various time-

consuming cognitive operations, including resolution of the conflict that arises when 

integrating input from two languages, suppression of the prior language following a 

switch, and/or promotion of the switched-into language that was previously inhibited 

(Adler et al., 2020; Litcofsky & Van Hell, 2017). However, other research demonstrates 

no difficulties associated with understanding code-switches (Adamou & Shen, 2017; 

Gosselin & Sabourin, 2021; Johns et al., 2019), or with processing naturalistic, 

common types of switches (Salig et al., 2024). Notably, bilinguals code-switch during 

spontaneous conversations under specific pragmatic contexts (e.g., in the presence of 



 

 

11 
 

another familiar bilingual) with no obvious detriment to communication (Poplack, 

1980).  

Indeed, the reality is that bilinguals frequently engage in code-switching, and 

communication remains successful. While code-switches can sometimes lead to 

processing costs during comprehension—typically measured on the millisecond scale 

in context-less and pragmatically under-supported lab studies—emerging findings 

from psycholinguistics research suggest that code-switches may also offer more 

macroscopic benefits to the bilingual listener that enable successful communication 

(e.g., Valdés Kroff & Dussias, 2023). Based on this understanding, we test the 

hypothesis that code-switches enhance bilingual listeners’ attention to the speech signal 

during story listening, potentially affecting their later memory for information 

contained around the code-switch. In what follows, we explore various accounts that 

describe how such a relationship may form and the extent to which it does. 

2.1.1 Code-switches and Attention: Three Accounts 

Attention refers to the multicomponent cognitive system that allows individuals 

to orient themselves to sensory input, sustain focus throughout a task, and gather 

information over time (Petersen & Posner, 2012; Posner & Petersen, 1990). In this 

section, we sketch various ways in which code-switches might influence bilinguals’ 

attention and, consequently, their memory for information presented in code-switched 

input. We will focus on two main explanations for how code-switches could heighten 

attention to the speech signal, but to distinct features of the input with fundamentally 

different effects on memory.  



 

 

12 
 

Orienting Account: Code-switches direct attention to the content of 

language input. One possibility is that, when hearing a code-switch, listeners may 

orient their attention to speech content, enhancing their ability to collect, encode, and 

later remember information, compared to content that has not been code-switched.  

Such an effect could be driven by listeners’ prior experience with code-

switching. Knowledge about the conditions under which a speaker is likely to code-

switch may enable a listener to infer the kind of information a code-switch carries, thus 

modulating attention to accommodate that inference. For example, code-switches often 

occur before words that might be unexpected or surprising from the comprehender’s 

perspective (Calvillo et al., 2020; Myslín & Levy, 2015). A visual-world eye-tracking 

study demonstrated that bilingual listeners are sensitive to such probabilistic 

occurrences: Although their eye-gaze patterns showed anticipation of high-frequency 

referents after hearing single-language input (as is typical; see Dahan et al., 2001), they 

anticipated low-frequency referents after hearing a code-switch (Tomić & Valdés 

Kroff, 2022). Thus, bilinguals with sufficient code-switching experience may infer the 

communicative intent of a code-switch regarding its sentential or pragmatic content 

(e.g., a switch may signal upcoming mention of a rarer object) and increase their 

attention to that content accordingly.  

While there are several reasons why a speaker may code-switch (e.g., retrieval 

difficulty, marking identity, rhetorical devices; see e.g. Gumperz, 1982), our focus here 

is not on production choices. The essence of our proposal is that bilingual listeners are 

sensitive to the various reasons why a code-switch may occur. Thus, they learn to draw 

appropriate inferences when a speaker code-switches. We speculate that this inference 



 

 

13 
 

process, which would require bilingual experience, may increase listeners’ attention to 

the content of linguistic input when they hear a code-switch.  

Note, however, that the orienting account does not necessarily require an 

inference by the comprehender: A beneficial effect of code-switches on listeners’ 

attention and memory may be elicited by the signal change itself, rather than by 

listeners’ knowledge. Because a code-switch involves a shift between phonological 

systems—relatively low-level acoustic properties of the signal—such a shift might also 

orient listeners’ attention to speech content. Specifically, the language change may be 

salient, thereby directing listeners’ attention to the message in the input through a 

bottom-up process (see Diachek & Brown-Schmidt, 2022 and Fraundorf & Watson, 

2011 for a similar account applied to disfluency processing).  

Whether an orienting effect of code-switches reflects listeners’ bilingual 

knowledge or stems from the signal change, it should generate increased attention to 

the speech signal (specifically to the information it carries) and a corresponding boost 

in encoding the message contained in the signal. Such an increase in attention to the 

linguistic input should, we hypothesize, lead to better memory for content surrounding 

a code-switch compared to non-switched input. 

Diverting Account: Code-switches increase attention to auditory input but 

divert attention from speech content. Instead of orienting listeners’ attention to 

content surrounding a switch, code-switches might capture listeners’ attention and 

draw it away from information conveyed in the input.  

Myslín and Levy (2015) suggest that code-switches serve as a marked or 

distinct type of speech that captures attention because of phonological changes. In 



 

 

14 
 

bilingual conversations, most utterances are not code-switched (Beatty-Martínez et al., 

2020; Fricke & Kootstra, 2016; Piccinini & Arvaniti, 2015), so a switch between 

languages may be salient. Earlier, we suggested that the saliency of the phonological 

change may orient attention to the linguistic message in a bottom-up way that helps the 

listener collect and later remember the information conveyed in the switch. However, 

an alternative possibility is that the phonological change may seize attention, 

disengaging it from sentential or pragmatic content. This may be akin to an ‘acoustic 

oddball effect,’ where unexpected sounds distract from a central task, impairing 

detection of a subsequent auditory target (e.g., Dalton & Lavie, 2004).  

Under this account, although attention to the auditory signal is increased at a 

code-switch, it is captured by a distracting, salient change in sound patterns that does 

not facilitate greater information collection. If this is the case, then bilinguals’ memory 

for information around the switch should not be better and in fact may be worse than 

for single-language material. This is because attention is not specifically directed to 

input of communicative value, but rather captured by a signal change, which disrupts 

information collection and negatively impacts memory. Such an account could align 

with the traditional observation of switch costs, wherein processing a code-switch 

presents difficulty during real-time comprehension. 

Comparing the two attention-boosting accounts. Under both accounts 

discussed thus far, we propose that code-switches increase attention to the speech 

signal, albeit to different aspects of the signal. Moreover, the attentional increase should 

occur immediately, when the switch happens and sustain for at least a short while (in 

line with descriptions of how certain aspects of attention function; Petersen & Posner, 



 

 

15 
 

2012). The key difference between the accounts is that the orienting account predicts a 

memory benefit for linguistic information near code-switches, compared to single-

language content. This is because, under the orienting account, the increase in attention 

is directed at message content, enhancing the encoding of information around the 

switch. In contrast, the diverting account predicts no memory benefit (or even worse 

memory) for information around the switch, compared to non-switched material, 

because the switch is a salient phonological change that temporarily captures attention, 

leading the listener to miss crucial information.  

Null Account: Code-switches do not modulate attention. It is possible that 

code-switches may not affect listeners’ attention or memory. In natural bilingual 

interactions where speakers use their full linguistic repertoire, code-switches might not 

engage attention any differently than single-language content. The phonological 

change in code-switches may appear typical rather than salient, and listeners may not 

infer any specific meaning from a code-switch. However, this scenario seems unlikely 

given the evidence that bilingual listeners detect code-switches early during language 

processing (Kuipers & Thierry, 2010) and can use them as a predictive cue to 

comprehension (Tomić & Valdés Kroff, 2022). Sociolinguistic studies also suggest that 

code-switching serves various communicative functions beyond lexical accessibility 

(Gumperz, 1982). Thus, bilinguals appear to be sensitive to both the phonological 

change that occurs during a code-switch and the pragmatic functions of code-switching. 

This suggests that such sensitivities could lead to a modulation of attention when 

encountering a code-switch. Indeed, prior work has posited a relationship between 

attention and code-switching (Beatty-Martínez et al., 2021; Green, 2019; Salig et al., 



 

 

16 
 

2021; see also Bialystok & Craik, 2022; Nijmeijer et al., 2022; Timmer et al., 2021), 

although none have specifically examined how code-switches affect attention during 

comprehension. 

Considering a Continuum of Bilingualism. We have outlined three accounts 

of how code-switches might impact attention. Of course, it is plausible that code-

switches have different attentional effects on different bilinguals. Habitual code-

switchers may infer pragmatic meanings from code-switches that focus their attention 

on speech content, while non-switchers might find their attention diverted by the 

phonological change that they are not accustomed to processing. Additionally, speakers 

use code-switching to convey a range of communicative functions, which may lead 

listeners to engage attentional resources differently depending on the environmental 

and pragmatic conditions present. The above accounts therefore are not necessarily 

mutually exclusive, as code-switching experience could influence bilinguals’ ability to 

infer meaning from code-switches or how they process code-switches (Beatty-Martínez 

& Dussias, 2017; Gosselin & Sabourin, 2021; Valdés Kroff et al., 2018). Consequently, 

we hypothesize that code-switching experience may interact with the attentional effects 

of hearing code-switches.  

In the experiments reported below, we did not vary the pragmatic function for 

the code-switches that listeners heard, choosing to present naturalistic code-switches 

embedded in long stories. This research serves as an initial step toward differentiating 

among the proposed accounts and assessing whether code-switching experience 

influences the impact of code-switching on attention and memory. These insights could 



 

 

17 
 

lead to a deeper understanding of the varied outcomes observed thus far regarding 

switch costs during online processing. 

2.1.2 The Current Experiments: Preliminaries and Predictions 

We conducted two experiments to examine the effects of code-switched input 

on the attention and memory of Spanish-English bilinguals (Experiment 1a) and 

English-speaking monolinguals (Experiment 1b) during naturalistic story listening. 

Participants heard both single-language and code-switched stories while periodically 

indicating their attention levels. Afterward, they answered comprehension questions to 

assess memory retention. We hypothesized that bilinguals would report higher 

attention levels during code-switched content compared to single-language content, 

and that better-attended information would be more accurately remembered (e.g., 

Boudewyn & Carter, 2018). Additionally, we predicted that code-switches would 

influence memory for information and that bilinguals’ code-switching experience 

would impact the attention and/or memory effects observed. 

To preview our results, bilinguals reliably reported higher attention levels when 

exposed to code-switches compared to single-language input. They also demonstrated 

significantly better memory for material near code-switches than for material in single-

language contexts due to their increased attention. Interestingly however, bilingual 

code-switching experience did not influence these effects. 

Experiment 1b was designed to address whether these effects could be solely 

explained by the saliency of the switch between languages. By examining monolingual 

participants’ responses to code-switched material, we aimed to differentiate between 

stimulus-driven and bilingual-experience-driven effects. This allowed us to distinguish 



 

 

18 
 

between two accounts of our findings: the bottom-up orienting account, which suggests 

that attentional effects are mainly due to the novelty or incongruity of the language 

switch, and the higher-level orienting account, which posits that bilinguals’ increased 

attention and memory for information around a switch result from their exposure and 

sensitivity to code-switching contexts. 

We will show that, in contrast to bilinguals, monolinguals did not exhibit an 

attentional increase in response to code-switches; instead, they reported higher 

attention levels in single-language contexts. This finding challenges a purely bottom-

up saliency account, which would predict similar attention effects in both groups. Thus, 

our results highlight that the attention and memory effects of code-switches in 

bilinguals are influenced by their linguistic experience, including their sensitivity to the 

context and communicative intent conveyed by a code-switch, underscoring the role of 

experience-derived knowledge. 

2.2 Experiment 1a Method 

The method and analyses were pre-registered (https://osf.io/fuahq/). 

Experimental tasks were approved by the University of Maryland’s Institutional 

Review Board. Participants provided informed consent and received $12/hour. 

2.2.1 Participants  

We recruited Spanish-English bilinguals based in the U.S. through Prolific 

(https://www.prolific.com/). Participants first completed a qualifying survey, which 

included the Bilingual Code-switching Profile (BCSP; Olson, 2024), a validated tool 

for assessing self-reported code-switching background and use. The survey also 

https://osf.io/fuahq/


 

 

19 
 

included questions about their language history and LexTALE vocabulary assessments 

in both English and Spanish (Izura et al., 2014; Lemhöfer & Broersma, 2012). Of the 

182 bilinguals who completed the survey, 132 qualified and were invited to participate 

in the main study. To qualify, participants had to indicate that Spanish and/or English 

was their first and most-preferred language, and they had to score 55% or higher on 

both LexTALE assessments.  

Ninety-six out of the 132 accepted our invitation to complete the main study, 

which occurred in a separate experimental session. We excluded data from four 

participants who did not meet pre-registered criteria: Three gave the same response to 

all attention probes, and one failed all engagement checks (see below). The remaining 

92 bilinguals (39 women, 36 men, 1 of another gender, and 16 who did not report 

gender) appeared mostly English dominant, although it is important to note that the 

LexTALE relies on the evaluation of written vocabulary and may underestimate 

Spanish proficiency for our U.S.-based participants, who likely had more verbal than 

written exposure to Spanish. Importantly, our bilingual sample represented varied first 

language backgrounds: 43 had English as their first language, 25 had Spanish, and 24 

had acquired both Spanish and English from birth. They also had moderate, but varied, 

code-switching experience (see Table 2.1). Six participants reported that they did not 

code-switch, but those who did started relatively early in life (M=10.41 years, 

SD=6.12). On average, the bilinguals had spent approximately 14.07 years in a 

region/community where switching languages is common, and they seemed to code-

switch more in family settings than in other settings (e.g., work or school). 

 



 

 

20 
 

Table 2.1. Bilingual participants’ characteristics (N=92) 
 

  Mean (SD) 

Age (years) 33.36 (10.29) 

AoA English (years) 1.51 (2.80) 

AoA Spanish (years) 5.24 (8.23) 

English LexTALE score (out of 100) 91.93 (8.81) 

Spanish LexTALE score (out of 100) 69.66 (11.50) 

Language Exposure Entropy (from 0 to 1.58) 0.81 (0.30) 

BCSP score (out of 100) 48.58 (12.45) 
Note. AoA = Age of Acquisition. Language Exposure Entropy was calculated based on 
participants’ self-reports of what percentage of their daily time on average they are 
exposed to English, Spanish, or other languages (higher indicates more balanced 
exposure; Gullifer & Titone, 2020). BCSP = Bilingual Code-switching Profile, which 
measures code-switching experience (higher indicates more code-switching 
experience/engagement). 
 

2.2.2 Materials and Design 

The study materials, including links to the audio files on GitHub, are available 

in the OSF repository (https://osf.io/fuahq/). Participants completed the experiment 

remotely using PCIbex (Zehr & Schwarz, 2018). Each participant listened to two 

Sherlock Holmes stories, adapted from Boudewyn and Carter (2018): one story entirely 

in English and the other mostly in English but with code-switches into Spanish. The 

stories were approximately 35 minutes long (to reduce study length, audio files were 

sped to 1.07x the original rate, which still sounded natural). The stories were recorded 

by an early Colombian-Spanish/American-English bilingual woman who frequently 

code-switches in her daily life. She was raised in South Florida, a highly bilingual 

https://osf.io/fuahq/


 

 

21 
 

geographic area of the U.S. with many Spanish dialects in contact. She exhibited the 

characteristics of early Spanish-English bilingual speakers of the region, including 

dialect features prevalent in Miami English (Carter & Lynch, 2015; Carter et al., 2020) 

and in Spanish influenced by contact with multiple dialects and American English. 

 Attention Probes. During the stories, an attention probe appeared 

approximately every 1-2 minutes, following a method from the mind-wandering 

literature (e.g., Mooneyham & Schooler, 2013; Smallwood et al., 2008). Each story had 

24 or 25 probes. When a probe appeared, participants had ten seconds to report their 

attention level by selecting a point along a vertical line with a red-face icon at the 

bottom to indicate off-task, a yellow-face icon at the midpoint to indicate split-

attention, and a green-face icon at the top to indicate on-task attention (see Figure 2.1). 

Participants’ responses were recorded as a value on a continuous 100-point (0-99) 

scale. The scale was explained before the study; participants were encouraged to 

respond honestly and were assured that their responses would not affect compensation. 

Trials without responses to a probe were re-coded as that person’s lowest-used attention 

response value within that story. It is important to note that while this approach 

measures attention subjectively, it is not arbitrary: Responses to similar attention 

probes reliably correlate with objective measures of attention, including neural 

oscillations in the alpha band from scalp-recorded electroencephalography (Boudewyn 

& Carter, 2018). Moreover, responses to such attention probes predict expected 

behavioral outcomes; for example, when participants report lower attention, their error 

detection and comprehension question performance is less accurate than when they 

report higher attention (see Mooneyham & Schooler, 2013 for a review). 



 

 

22 
 

Code-switching Manipulation. We created the code-switched versions of the 

stories by inserting English-to-Spanish code-switches shortly before approximately 

half of the attention probes (within three sentences of the probe) in the single-language 

(English-only) versions. In the code-switched versions, one story had critical code-

switches shortly before 11 of the 24 attention probes, and the other had critical code-

switches shortly before 15 of the 25 attention probes. This approach allowed us to 

measure attention near a recent code-switch, assessing the immediate impact of hearing 

a code-switch on attention in real time (see Figure 2.1). The remaining attention probes 

in the code-switched stories did not closely follow a code-switch, enabling us to 

measure attention within a code-switched story, but after stretches of single-language 

content—namely, far from a code-switch. Such probes allowed us to evaluate whether 

hearing code-switches in a global story-listening context modulated attention 

independently of specific code-switching events. Code-switches were naturally spoken 

and recorded without splicing. Audio files were edited to remove lengthy silences and 

errors. Story version (code-switched versus single-language) and the order of the two 

stories were counterbalanced across participants, resulting in four possible lists to 

which participants could be assigned: Story1CodeSwitch-Story2SingleLanguage 

(n=25), Story1SingleLanguage-Story2CodeSwitch (n=22), Story2CodeSwitch-

Story1SingleLanguage (n=22), and Story2SingleLanguage-Story1CodeSwitch (n=23).  

Critical code-switches were intrasentential, occurring within sentences at noun 

or verb phrases, after which the content remained in Spanish until after the next 

attention probe. In a separate norming study, 49 Spanish-English bilinguals on Prolific 

rated these critical code-switches as more naturalistic (M=6.53 on a 1-9 scale, 



 

 

23 
 

SD=0.78) than intentionally ill-formed code-switches (M=4.77, SD=0.90, p<0.01). 

Experimental switches that were rated poorly in norming were revised.  

To prevent participants from reliably predicting an upcoming attention probe 

based solely on the occurrence of a code-switch, we strategically inserted 5-6 filler 

code-switches into the code-switched stories. With 11-15 critical code-switches and 5-

6 filler code-switches, attention probes followed approximately two-thirds of the code-

switches. This intentional inconsistency likely prevented participants from learning an 

association between code-switches and probes. The filler code-switches were placed at 

least five sentences away from the next attention probe, occurred at varied syntactic 

sites, and included at least one sentence in Spanish before switching back to English, 

equivalent to the amount of Spanish in the critical code-switches.  

Overall, approximately 5% of the words in the code-switched stories were in 

Spanish, a proportion within the range of what occurs naturally in bilingual speech (see 

Fricke & Kootstra, 2016). Moreover, the code-switched segments occurred at 

morphosyntactic sites considered typical and naturalistic in the Spanish-English code-

switching literature (Poplack, 1980), which our norming data corroborated. Care was 

taken to limit Spanish dialect differences across lexical items and morphosyntactic 

features, with Spanish-speaking research assistants and co-authors independently 

verifying the intelligibility of the stories. In fact, in Experiment 1a, participants rated 

the naturalness of the code-switched stories at a 5.33 out of 7 (SD=1.48), and bilinguals 

with more code-switching experience tended to rate the naturalness of the code-

switched stories higher (r=0.24, p<0.001). This indicates that the code-switched stories 



 

 

24 
 

did not seem unnatural to the listeners, particularly to those with significant code-

switching experience. 

Post-story Memory Assessment. After listening to each story, participants 

answered multiple-choice comprehension questions (modified from Boudewyn & 

Carter, 2018) that tested memory for story material preceding attention probes (Figure 

2.1). One story had 30 comprehension questions, and the other had 44, following the 

original paradigm from Boudewyn & Carter (2018). After the code-switched story, 

questions tested material that appeared either near or far from a code-switch. Note that 

this design feature was necessary to test whether memory for story content was directly 

linked to a local code-switching event, or more broadly to a general code-switching 

story context. However, it also created an asymmetry in our comprehension questions: 

Questions that tested memory for information near code-switched material mostly 

probed content that was presented in Spanish, whereas questions that tested memory 

for information far from code-switched material probed content that was presented in 

English. We return to this issue in the General Discussion. To assess bilinguals’ 

attention while they heard the information interrogated by the question, we used their 

responses from the nearest attention probe after the relevant story material. This helped 

test whether attention influenced later memory for content.1 

 
1 Because our study involved remote participation over the internet, we first conducted a validation study 
with English-speaking monolinguals who listened to English-only versions of the stories (pre-registered 
at https://osf.io/g92at/). We found that greater self-reported attention to information predicted higher 
accuracy when tested on that information later (p<0.01), replicating the link between attention and 
memory observed in prior in-person studies (Boudewyn & Carter, 2018; Smallwood et al., 2008). 
Additionally, we found that story identity (i.e., Story 1 vs. Story 2) did not predict attention probe 
responses (p=0.97) or comprehension question accuracy (p=0.44), indicating that both stories were 
matched in interest and difficulty when presented in English. Data are available on OSF. Note that this 
validation study is distinct from Experiment 1b, reported below, in which English-speaking 
monolinguals heard both English-only and code-switched stories, like the bilinguals in Experiment 1a. 

https://osf.io/g92at/


 

 

25 
 

Engagement Checks. We included four engagement checks, one early in the 

study and three intermixed with comprehension questions, which instructed 

participants to select a specific answer to evaluate general alertness during the 

experiment. For example, one engagement check question instructed: “For this 

question, please select the letter B (Watson) as your answer. This is to check that you 

are engaged in the study task.” 

 

 

Figure 2.1. Study Schematic with Example Stimuli. The illustration represents two of 
the three conditions: An attention probe after a recent code-switch in the code-switched 
version of a story (top panel), and an attention probe in the single-language version of 
the same story (bottom panel). Each participant heard only one version of each story, 
code-switched or single-language. There was also a third condition (not illustrated): 
Attention probes in the code-switched story but after single-language content (far from 
a code-switch). After the stories were over, comprehension questions assessed memory 
for information that had appeared near a code-switch, far from a code-switch, or in a 
single-language story. 



 

 

26 
 

2.3 Experiment 1a Results 

The materials, analysis scripts, and data from the 97% of participants who 

consented to data sharing are available on OSF (https://osf.io/fuahq/). Analyses were 

pre-registered unless indicated as exploratory. Linear mixed effects models were 

conducted in R (R Core Team, 2023, version 4.0.2) using the lmerTest package 

(Kuznetsova et al., 2017), which approximates p-values using the Satterthwaite degrees 

of freedom method. Random effects structures were pre-registered and included 

random slopes and intercepts by participant and by item but were simplified as needed 

to reach model convergence; final model specifications, including random effects, are 

in the Notes section of the model tables below.2 

2.3.1 Effects of Code-switching on Bilinguals’ Attention 

As can be seen in Figure 2.2, bilinguals reported greater attention after a recent 

code-switch (M=78.58, SD=16.11) compared to when probed far from a code-switch 

(M=73.85, SD=18.26), and compared to when probed in a single-language story 

(M=73.57, SD=19.07). This pattern aligns with our prediction, and a multiple 

regression model (Table 2.2) confirmed this observation: Within code-switched stories, 

bilinguals reported significantly greater attention after recent code-switches than when 

probed far from code-switches (p<0.01, ηp
2=0.10). Crucially, there was no attention 

difference in code-switched stories overall compared to single-language stories 

(p=0.06). Together, these results indicate that code-switches did not increase attention 

 
2 After the study, we asked participants if they had previously read either story and if they had an 
attention disorder or took medication for one. The results did not change when we removed 3 participants 
who reported story familiarity or 15 participants who reported an attention disorder or medication. The 
results also did not change when we removed 10 participants who had below chance accuracy (<25%) 
on the comprehension questions. 

https://osf.io/fuahq/


 

 

27 
 

throughout a code-switched story; the attentional increase was localized to story 

sections that were near a code-switch, not sections that were far from a code-switch 

(i.e., in a stretch of single-language input). 

We also hypothesized that code-switching experience (measured by BCSP 

score) would interact with attentional modulations driven by code-switches. However, 

this pattern did not emerge (ps>0.11).  

 

Figure 2.2. Bilinguals’ Attention Responses by Language Context. Mean self-reported 
attention, split by language context. Bars represent standard error. Dots indicate 
individual participants’ average probe response in each language context. 
 
Table 2.2. Pre-Registered Regression Model Predicting Bilinguals’ Attention from 
Language Context. 
 
Factor β SE t p 
Intercept 0.43 1.76 0.25 0.81 
Language Context: Single-language story vs. 
Code-switched story 

2.71 1.40 1.94 0.06 



 

 

28 
 

Language Context: Far vs. Near a code-switch 3.26 1.00 3.26 <0.01 
BCSP Score -0.12 0.14 -0.87 0.39 
Language Context: Single-language vs. Code-
switched story * BCSP Score 

-0.18 0.11 -1.59 0.12 

Language Context: Far vs. Near code-switch * 
BCSP Score 

-0.03 0.07 -0.39 0.70 

Note. Model specification: CenteredProbeResponse ~ LanguageContext * BCSPScore 
+ (LanguageContext | Participant) + (1 | Probe). Language Context was orthogonally 
coded. 
 

2.3.2 Effects of Code-switching and Attention on Bilinguals’ Memory 

Bilinguals largely performed above chance (25%) on comprehension questions. 

As hypothesized, Figure 2.3 shows that bilinguals better remembered information that 

was presented near a code-switch (M=64.7%, SD=24.6%) compared to information far 

from a code-switch (M=58.4%, SD=24.8%), and compared to information in a single-

language story (M=59.2%, SD=24.6%).  

To confirm this, we conducted an exploratory logistic regression predicting the 

log(odds) of comprehension question accuracy solely from the language context in 

which information was presented (Table 2.3). The aim of this exploratory model was 

to determine if language context alone explained significant variance in memory for 

information, before considering how additional variables such as attention to 

information may further mediate this relationship. The model confirmed that bilinguals 

remembered information better when it appeared near a code-switch, compared to far 

from a code-switch, within code-switched stories (p=0.04). Crucially, there was no 

difference in memory for information in code-switched versus single-language stories 

overall (p=0.10). These results suggest that memory for information did not generally 

increase in code-switched contexts. Instead, the benefit was again localized: Memory 

increased for information that appeared in story sections near a code-switch rather than 



 

 

29 
 

in single-language sections far from a code-switch. Alongside the attention findings 

presented above, the data suggest that a code-switching event itself is linked to 

increased memory for information. 

 

Figure 2.3. Bilinguals’ Accuracy of Remembering Information. The pre-registered 
logistic regression model’s estimated probability of accurately answering a 
comprehension question, split by language context. Bars represent standard error. Dots 
indicate individual participants’ average estimated accuracy in each language context. 
 
Table 2.3. Exploratory Regression Model Predicting Bilinguals’ Comprehension 
Question Accuracy from Language Context Alone. 
 
Factor β SE z p 
Intercept 0.55 0.17 3.21 <0.01 
Info Context: Single-language story vs. Code-
switched story 

0.12 0.07 1.66 0.10 

Info Context: Far vs. Near a code-switch 0.29 0.14 2.10 0.04 
Note. Model specification: QuestionAccuracy ~ InfoContext + (1 | Participant) + (1 | 
Question). Info Context was orthogonally coded. 
 

Logically, we also expected bilinguals to generally remember information 

better when they reported paying more attention to it, a factor not evaluated by our 

exploratory model above. Indeed, Figure 2.4 shows that bilinguals remembered 



 

 

30 
 

information more accurately when they reported paying more attention to it, regardless 

of language context. A pre-registered logistic regression confirmed that bilinguals’ 

attention to information significantly predicted memory accuracy (p<0.01; Table 2.4). 

With attention and language context in the model, language context no longer 

explained variation in question accuracy. These findings replicate the link between 

attention during story listening and memory for content (Boudewyn & Carter, 2018) 

and suggest that the improved memory for information near code-switches was driven 

by the increased attention to that information. The general pattern also strongly 

suggests that listeners’ subjective reports of attention level are not random or 

indiscriminate, as they correlate with objective measures of comprehension in a 

sensible way (higher attention leads to better memory; lower attention leads to worse 

memory; see also Boudewyn & Carter, 2018; Mooneyham & Schooler, 2013).  

This pre-registered model also included a factor called Distance of Information 

from Attention Probe. This variable indicated the number of intervening sentences 

between the critical information (required to answer the comprehension question) and 

the next attention probe (which we use in this model as a measure of attention to that 

information). Including this factor as a covariate in the model allowed us to ensure that 

any effect of language context on memory for information was not solely due to that 

information being closer to or farther from an attention probe.  

Additionally, we attempted to include code-switching experience (BCSP 

scores) in this model to determine if code-switching experience interacted with the 

memory effects; however, the BCSP score variable was removed when the model failed 

to converge after employing other methods of simplification.  



 

 

31 
 

 

Figure 2.4. Bilinguals’ Memory Accuracy of and Attention to Relevant Information. 
Relationship between individuals’ average estimated question accuracy and their 
average self-reported attention to relevant story information interrogated by the 
question, split by language context. The pre-registered logistic regression model’s 
estimated probability of accurately answering comprehension questions was used to 
calculate individual participants’ average estimated accuracy in each language context. 
Gray lines indicate the 95% confidence interval.  
 
Table 2.4. Pre-Registered Regression Model Predicting Bilinguals’ Comprehension 
Question Accuracy from Language Context and Attention to Information 
 
Factor β SE z p 
Intercept 0.54 0.16 3.32 <0.01 
Centered Attention Probe Response 0.02 0.00 8.02 <0.01 
Info Context: Single-language story vs. Code-
switched story 

0.04 0.07 0.61 0.54 

Info Context: Far vs. Near a code-switch 0.24 0.14 1.73 0.08 
Distance of Information from the Attention 
Probe 

-0.04 0.02 -1.88 0.06 

Info Context: Single-language vs. Code-
switched story * Centered Probe Response 

0.00 0.00 0.44 0.66 

Info Context: Far vs. Near code-switch * 
Centered Probe Response 

0.00 0.00 0.13 0.90 

Note. Model specification: QuestionAccuracy ~ CenteredProbeResponse * 
InfoContext + InfoDistanceFromProbe + (1 | Participant) + (1 | Question). Info Context 
was orthogonally coded.  



 

 

32 
 

 

2.3.3 Participant Enjoyment 

Although we do not discuss it further, here we report the results of our pre-

registered analyses on bilinguals’ enjoyment of the task. Bilinguals rated the enjoyment 

of their listening experience for both stories on a 1-7 scale, and on average they rated 

the experience a 4.28 out of 7 (SD=1.80). A multiple regression model showed that 

participants rated their enjoyment of stories higher when their average attention to that 

story was greater (p<0.01, ηp
2=0.21). Whether a story was code-switched or not did not 

predict bilinguals’ self-reported enjoyment of that story (p=0.88). 

2.4 Experiment 1a Discussion 

Bilinguals paid more attention to and better remembered code-switched 

information, a finding that is clearly at odds with the null account sketched earlier. The 

data are also inconsistent with the diverting account, as memory was better for 

information near code-switches, which we would not expect if code-switches directed 

attention away from linguistic messages. Overall, the results support an orienting 

account, under which code-switches quickly elicit increases in bilinguals’ attention to 

speech content.  

This orienting effect could be influenced by a higher-level process in which 

listeners draw an inference based on their bilingual experience and/or by a bottom-up 

process, where the salient change in sound patterns at a code-switch automatically 

grabs attention, regardless of linguistic knowledge. These possibilities are further 

discussed in the General Discussion. Experiment 1b assesses the possibility that a 

change in language alone could create a bottom-up attentional effect in the absence of 



 

 

33 
 

any meaningful bilingual language experience. If a bottom-up process driven by the 

novelty of the language change is what increased attention at a code-switch, then even 

monolinguals should demonstrate increased attention after hearing a code-switch. 

However, monolinguals would not be expected to demonstrate a memory benefit for 

code-switched information since they would not understand the content provided in 

one of the languages.  

In Experiment 1b, we aimed to isolate the bottom-up, stimulus-driven process 

within the orienting account by testing whether the attentional increase after a switch 

is solely driven the saliency of the language change. To examine this possibility, 

English-speaking monolinguals completed the same story listening task as bilinguals 

in Experiment 1a. They listened to one story entirely in English and another story that 

included code-switching, periodically reporting their attention throughout the task. 

2.5 Experiment 1b Method 

The method of Experiment 1b was identical to that of Experiment 1a with a few 

exceptions. First, although all participants completed the study using the same online 

platform (PCIbex), Experiment 1b participants were recruited from the University of 

Maryland’s study pool and received course credit for their participation, as opposed to 

Experiment 1a participants who were recruited through Prolific and received payment. 

Second, we specifically recruited English-speaking monolinguals who confirmed that 

English was their first and only language, without significant experience in other 

languages. Thus, participants did not need to complete a qualifying survey to assess 

language proficiency. Third, we added language history questions at the end of the 

study to ascertain if participants had any notable experience with other languages.  



 

 

34 
 

As in Experiment 1a, participants could be assigned to one of four lists: 

Story1CodeSwitch-Story2SingleLanguage (n=19), Story1SingleLanguage-

Story2CodeSwitch (n=24), Story2CodeSwitch-Story1SingleLanguage (n=20), and 

Story2SingleLanguage-Story1CodeSwitch (n=23). In the instructions to the 

experiment, participants were encouraged to pay attention to the stories, even if they 

encountered parts in a language they did not understand. 

Although Experiment 1b was not pre-registered, we followed the same data 

collection and analysis procedures as in the pre-registered Experiment 1a, with only 

minor changes as described below. 

 

2.5.1 Participants 

A total of 110 monolinguals participated in Experiment 1b. The same exclusion 

criteria pre-registered for Experiment 1a were applied, resulting in the exclusion of 24 

participants for failing to respond to 25 or more of the 49 attention probes, failing 2 or 

more of the 4 engagement checks, and/or giving the same response to all attention 

probes. Eighty-six participants were included in the final analyses (49 women, 31 men, 

and 6 who did not report gender). The included participants had a mean age of 19.14 

years (SD=1.18).  

Although all participants identified as being functionally monolingual, they 

reported varying experiences with other languages, including Spanish. Fifty-one 

participants reported having learned or attempted to learn Spanish at some point: 21 

reported one year or less of Spanish experience, 19 reported two or three years, 7 

reported four of five years, and 8 reported five or more years. Most Spanish experience 



 

 

35 
 

was in the form of formal classroom learning that occurred between elementary and 

high school. Overall, responses indicated minimal (and not current) experience learning 

Spanish in a classroom environment. All 86 participants reported that they had never 

read either of the stories used in the study. 

2.6 Experiment 1b Results 

Analysis scripts and data from the 76% of participants who consented to data 

sharing are available on OSF (https://osf.io/fuahq/).3 

In Experiment 1a, we used orthogonal coding in our regression models to 

compare performance in the English-only story versus the code-switched story as a 

whole, and to compare performance in the code-switched sections versus single-

language sections of the code-switched story. However, in Experiment 1b, it was more 

appropriate to use new orthogonal coding to enable more meaningful comparisons. 

Orthogonal coding in Experiment 1b allowed for comparisons between performance in 

single-language English sections (in either the code-switched or English-only story) 

versus in code-switched sections, and between performance in English sections of a 

code-switched story versus English sections of a single-language story. We made this 

change in consideration of the results of Experiment 1a, where attention increased 

shortly after a code-switch but not throughout an entire code-switched story. 

Additionally, the participants in Experiment 1b were monolinguals, which influenced 

the choice of coding approach. The new orthogonal coding allowed us to isolate the 

 
3 The pattern of Experiment 1b results did not change after removing 16 participants who reported an 
attention disorder or medication. Similarly, the patterns did not change when we excluded 21 participants 
who scored below chance (<25%) on the comprehension questions. 

https://osf.io/fuahq/


 

 

36 
 

effect of code-switches and then assess whether the presence of code-switches in a 

story affected monolinguals’ processing of information presented in English (their 

known language). The models with the new orthogonal coding are reported here but 

Experiment 1b models using the original orthogonal coding (identical to Experiment 

1a’s coding) are also available on OSF.  

 

2.6.1 Effects of Code-switching on Monolinguals’ Attention 

Monolinguals reported higher attention in an English-only story (M=59.25, 

SD=22.88) and when probed in a single-language stretch of a code-switched story (far 

from a code-switch, M=57.19, SD=23.79) compared to when probed near a code-

switch (M=56.18, SD=23.07), which was confirmed by our regression model (Table 

2.5; p<0.05, ηp
2=0.04). This finding suggests that monolinguals did not experience the 

same attentional boost from code-switches as bilinguals, indicating that the language 

change alone is unlikely to fully account for the increased attention observed in 

bilinguals in Experiment 1a. Additionally, there was no significant difference in 

attention to single-language content that occurred in a code-switched story (far from a 

code-switch) compared to in a single-language story (p=0.66). This suggests that the 

general presence of code-switches did not disrupt monolinguals’ attention to 

information in their known language (English). 

 

Table 2.5. Regression Model Predicting Monolinguals’ Attention from Language 
Context 
 
Factor β SE t p 
Intercept -0.40 2.28 -0.17 0.86 



 

 

37 
 

Language Context: Near a code-switch vs. 
Single-language context 

3.48 1.73 2.01 <0.05 

Language Context: Single-language section in 
Code-switched story vs. in Single-language 
story 

0.99 2.25 0.44 0.66 

Note. Model specification: CenteredProbeResponse ~ LanguageContext + 
(LanguageContext | Participant) + (1 | Probe). Language Context was orthogonally 
coded. 
 

2.6.2 Effects of Code-switching and Attention on Monolinguals’ Memory  

Monolinguals largely performed above chance (25%) on the comprehension 

questions. Our logistic regression analysis indicated that monolinguals had 

significantly better memory for English-only information (from either the code-

switched or single-language story) compared to code-switched information (Table 2.6; 

p<0.01). This difference appeared to be primarily driven by monolinguals’ stronger 

memory for information in the English-only story (M=47.49%, SD=24.69%), as 

numerically, their memory was comparable for information from single-language 

material within the code-switched story (far from a code-switch; M=42.31%, 

SD=27.44%) and material near a code-switch (M=42.91%, SD=26.58%).  

However, there was no statistically significant difference in memory for 

English information that had appeared in a single-language story versus in a code-

switched story (far from any code-switch; p=0.10), despite the numerical trend for 

better memory in the single-language story. This finding is consistent with the attention 

results, further indicating that the general presence of code-switches in a story did not 

impair monolinguals’ ability to process and retain information in the language that they 

understood. This is notable because, in this experiment, information that was presented 

in Spanish was not reiterated in English. This setup created the potential for 



 

 

38 
 

monolinguals to miss important information (because it was in Spanish) and to become 

increasingly lost as the story progressed. This aspect of the study design could account 

for the non-significant trend towards poorer memory for English information in the 

code-switched story compared to the single-language story. 

 

Table 2.6. Logistic Regression Model Predicting Monolinguals’ Comprehension 
Question Accuracy from Language Context and Attention to Information. 
 
Factor β SE z p 
Intercept -0.41 0.16 -2.57 0.01 
Centered Attention Probe Response 0.02 0.00 7.76 <0.01 
Info Context: Near a code-switch vs. Single-
language context 

0.45 0.13 3.54 <0.01 

Info Context: English information in Code-
switched story vs. in Single-language story 

0.13 0.08 1.66 0.10 

Distance of Information from the Attention 
Probe 

-0.04 0.02 -2.47 0.01 

Info Context: Near code-switch vs. Single-
language context * Centered Probe Response 

0.00 0.00 1.07 0.28 

Info Context: English info in Code-switched story 
vs. in Single-language story * Centered Probe 
Response 

0.00 0.00 -0.57 0.57 

Note. Model specification: QuestionAccuracy ~ CenteredProbeResponse * 
InfoContext + InfoDistanceFromProbe + (CenteredProbeResponse | Participant) + (1 | 
Question). Info Context was orthogonally coded.  
 

Importantly, our model also confirmed that monolinguals demonstrated a link 

between attention and memory: They were more likely to accurately remember 

information that they had paid more attention to (p<0.01). Further, this relationship did 

not interact with language context (ps≥0.28), suggesting that the attention-memory link 

held across all language contexts. 

There was also a significant effect of the distance of the information from the 

attention probe (p=0.01). That is, the critical information needed to answer a 

comprehension question was more likely to be accurately remembered if it had been 



 

 

39 
 

closer to an attention probe. This suggests that the study design may have resulted in 

improved memory for information that more closely preceded attention probes. 

However, we included this variable in the model to account for its contribution to 

variation in memory, and the other discussed effects still emerged with this factor 

included in the model. 

2.7 Experiment 1b Discussion 

Experiment 1b was designed to compare the effects of code-switching on 

attention between bilinguals (who were the focus of Experiment 1a) and monolinguals. 

By including monolingual participants, the study aimed to determine if the attentional 

benefits observed in bilinguals were unique to bilingual language processing or if they 

were attributable to a more general stimulus-driven phenomenon that could also be 

observed in monolinguals. The bottom-up, stimulus-driven hypothesis suggests that the 

salient change in sound patterns at a code-switch may automatically orient attention, 

regardless of linguistic knowledge, contributing to increased attention in both 

bilinguals and monolinguals. This possibility was explored by assessing if 

monolinguals also demonstrated increased attention after hearing a code-switch, 

despite not having the same