Orthography-induced Transfer in the Production of

Orthography-induced Transfer in the Production of
Novice Adult English-speaking Learners of Spanish
by
Yasaman Rafat
A thesis submitted in conformity with the requirements
for the degree of Doctor in Philosophy
Department of Spanish and Portuguese
University of Toronto
© Copyright by Yasaman Rafat (2011)
Orthography-induced Transfer in the Production of Novice Adult
English-speaking Learners of Spanish
Yasman Rafat
Doctor in Philosophy
Department of Spanish and Portuguese
University of Toronto
2011
Abstract
This study provides a thorough examination of the role of orthography in promoting first
language-based phonological transfer. Specifically, it analyzes the role of auditoryorthographic condition, type of grapheme-to-phoneme correspondence and aspects of
phonological memory on shaping transfer. Although, there has been previous work on the
role of orthography in the acquisition of second language phonology, not much is known
about the factors that shape orthography-induced transfer. In addition, the role of
orthography remains to be formalized in the future models of the acquisition of second
language phonology.
In this experiment, data was elicited via a primary Spanish-based picture-naming task and a
secondary Farsi-based non-word repetition phonological memory task. In the picture-naming
task, participants were divided into four groups and assigned to four conditions, three with
different degrees of exposure to orthography and one auditory condition. The data based on
the productions of 40 novice adult English-speaking learners of Spanish, reveal a robust
effect of orthography on phonological transfer leading to non-target-like productions at the
very beginning stages of second language acquisition. There is also strong evidence that
individual grapheme-to-phoneme correspondences differ in the extent to which they trigger
ii
phonological transfer. In addition, the findings show that while the presence of orthography
at learning or at production induces transfer, the presence of orthography at learning has a
stronger effect. The results also indicate some effect for the different aspects of phonological
memory, namely, primacy and repetition effects. However, there was no correlation between
individual phonological memory and the quantity of transfer.
Based on the findings, I argue that when a shared grapheme corresponds to two different
phonemes in the learners’ first language and the second language, the less salient the
acoustic/phonetic difference between the target language and the first language phonemes,
the higher the probability of first language transfer. I also argue for an effect of first
language grapheme-to-phoneme frequency on transfer, suggesting that when there is
variability in the realization of a particular grapheme in the first language, transfer will be
based on the most frequent first language realization. Moreover, based on the findings in this
study and previous research on the effect of orthography on second language production, I
propose that exposure to orthography may interfere with the establishment of second
language phonological categories.
iii
Acknowledgements
I am grateful to many people who have helped me in the PhD program. My deepest gratitude
for my advisor Professor Laura Colantoni, as I have been extremely fortunate to have her
guide me through this exciting and challenging period in my life. Laura, I have been in awe
of your intelligence, patience, kindness, commitment, advice, suggestions, sensitivity, and
understanding throughout the program. When it seemed like I would not be able to push
forward you always gave me the strength to believe in myself and my work. I cannot thank
you enough for all that you have done for me. You have given me the tools to become an
independent researcher and as I pursue my professional development, I will always know
that I can rely on all that you have given me.
My warmest gratitude to my committee members: Professors Jeffrey Steele and Ana Teresa
Pérez-Leroux.
Jeff, your role has been instrumental in the development of this thesis. I cannot express how
thankful I am for your careful, constructive and caring feedback. Your positive energy,
insights and enthusiasm for my work and encouragement has been invaluable. Your
intellectual support was as forthcoming as your emotional and spiritual encouragement. I
truly feel blessed to have had the opportunity to learn from you. Thank you for being such an
amazing mentor to me.
Ana, I have benefitted immensely from your knowledge and perspective. Whenever I was
caught up in the details of my work, you took the time to help me see the big picture and my
overall purpose. Thank you for sharing your wisdom with me.
My external committee members: Professors Martha Young-Scholten and Alister Cumming.
Thank you for your graceful pronunciations on my thesis. At this early stage of my career to
have had your feedback has been immensely valuable. You have been very generous in
offering your time and consideration.
iv
Many many thanks to other UofT faculty members: Professors María Cristina Cuervo,
Robert A. Davidson, Stephen Rupp and Olesya Falenchuk. Thank you all for all your
support.
My heartfelt appreciation goes out to UofT staff: Ms. Blanca Talesnik, Ms. Rosinda Raposo
and Ms. Heather Kelly. Thank you for always being there for me and your willingness to
help.
I am grateful to my friends for their help, commiserating and cajoling: Natalia Mazzaro,
Irina Marinescu, Anna Limanni, Yadira Alvarez, Orchid Fung, Elizabeth Chavez, Danielle
Thomas, Clelia Rodriguez, Olivia Marasco, Joanne Markle Lamontagne, Violeta Lorenzo,
Gorana Pobric, Maryam Soheil, Julieta Burdeanu, Greg Orencsak, Thea Urbina, and
Cornelia and Alex Farcas. Thank you for the fond memories and most importantly just being
there whenever I needed you.
This thesis would not have been possible without the help and kindness of the participants.
My loving family: my husband, brother, parents and my extended family: my aunt and
cousins and my parents-in-law. I am very lucky to have you in my life and would not have
started or completed this amazing journey if it were not for your faith in me.
I dedicate my thesis to my husband, Bahman who wholeheartedly supported and
encouraged me in the PhD program, including the writing of this thesis.
v
Table of Contents
ABSTRACT..…………………………………………………………………………….….ii
TABLE OF CONTENTS.…………………………………………………………………..iv
LIST OF TABLES..………………………………………………………………………..vii
LISTOF FIGURES...…………………………………………………………....................ix
LIST OF APPENDICES…....………………………………………………………….......x
Chapter 1.
1.1
Introduction………………………………………………………………….1
Objectives of the study………………………………………………….……1
1.2
Research questions and overview of the methodology and findings………2
1.3
Motivation and contributions.………………...……………………………..6
1.4
Thesis structure……………………………………..………………………9
Chapter 2.
Previous research on transfer, orthography and phonological memory in the
acquisition of L2 phonology…………………………………………………..10
2.1
Transfer and orthographic influence in models of the acquisition of L2
phonetics and phonology…………………… ………………………………12
2.2 The role of orthography in shaping L1 perception and production.…………...16
2.3
2.4
The role of orthography in L2 phonological acquisition..…………………..29
2.3.1
Positive effects of orthography on L2 phonological acquisition.....29
2.3.2
Negative effects of orthography on L2 phonological acquisition...34
PM capacity and L2 acquisition…..…………………………………….…42
2.4.1
2.4.2
PM………………………………………………………….……….43
Primacy and recency effects…...……..………………………...……44
vi
2.4.3
Effect of repetition on L2 vocabulary learning…………………….48
2.4.4
PM capacity and individual variation in L2 acquisition……………54
2.5 Chapter Summary……………………………………………………………..62
Chapter 3.
Hypotheses & methodology………………………………………………….65
3.1
Hypotheses…………………………………………………………………....66
3.2
Participants…………………………………………………………………….70
3.3
Task 1: Spanish picture-naming task…………………………………………71
3.4
3.5
3.3.1
Task design…………………………………………………………..72
3.3.2
Stimuli………………………………………………………………..74
Task 2: PM task……………………………………………………………….80
3.4.1
Stimuli…………………………………………………………….…..81
3.4.2
Task design……………………………………..…………………….84
Testing protocol………………………………………………………………84
Chapter 4.
Data analysis and results: world learning and PM tasks…………………..88
4.1
Picture-naming task…………………………………………………………...89
4.1.1
Data analysis……………………………...………………………….89
4.1.2
Results: picture-naming task ..………………………………………90
4.1.2.1
Overall effect of orthography and differences between
orthographic conditions…...………………………………91
4.1.2.2
4.1.2.3
Effect of grapheme-to-phoneme inconsistency………......93
Effect of auditory-orthographic condition on individual
grapheme-to-phoneme correspondences ……………….....98
vii
4.1.2.4
4.1.2.5
Effect of primacy and recency…………………………....102
Effect of primacy and recency on individual grapheme-to
phoneme correspondences……………………………….104
4.1.2.6
Effect of primacy within the word………..……………...109
4.1.2.7
Effect of round/repetition………….…………………….111
4.1.2.8
Effect of round/repetition on individual grapheme-tophoneme correspondence……………………………….113
4.2
Data analysis and results: PM task and individual variation in proportion of
transfer……….………………………………………………………….119
4.2.1
Data Analysis…………………………………………………..120
4.2.2
Results: PM scores and individual variation in orthographyinduced-transfer.……………………………..…………………122
4.3
Summary of results………………………………………………………………….125
Chapter 5.
Discussion and conclusions………………………….…………………………….129
5.1
Effect of auditory-orthographic condition……………………………………………130
5.2
Effect of grapheme-to-phoneme inconsistency between English and Spanish…..…..132
5.3
PM………………...…………………………………………………………..140
5.4
5.3.1
Primacy and recency effects………………………………………..……….141
5.3.2
Effect of round/repetition…………………………………………..………..145
5.3.3
Individual variation in PM and orthography-induced transfer……..………..147
Conclusions and future directions…………………………………………………150
viii
List of Tables
Table 3.1
Modality of presentation of input at learning and production (Auditoryorthographic condition)………………………………………………………...73
Table 3.2
Picture-naming task: target stimuli……………………………………………..76
Table 3.3
Picture-naming task distracters………………………………………...……….78
Table 3.4
Picture-naming task: positional composition of grapheme-to-phoneme
correspondences at learning and production per triplet……………….……….79
Table 3.5
PM non-word repetition task: Farsi stimuli…………………………………….83
Table 4.1
Mean proportion transfer and standard deviations by condition….…………….92
Table 4.2
Mann-Whitney test results for the effects of condition on the mean proportion
transfer.…………………………………………………………………………92
Table 4.3
Mean proportion transfer and standard deviations for Spanish grapheme-tophoneme correspondences by conditions………………………………………94
Table 4.4
Mann-Whitney test results for pair-wise comparisons of Spanish grapheme-tosound correspondences by condition……………………………………………..96
Table 4.5
Mann-Whitney results for the effect of condition for Spanish grapheme-tophoneme correspondences………...............………………………………...…100
Table 4.6
Mean proportion transfer and standard deviations for position within triplet by
condition………………..……………………………………………………103
Table 4.7
Mean proportion transfer and standard deviations for position within triplet by
Spanish grapheme-to-phoneme correspondence by condition…………………105
Table 4.8
Kruskal-Wallis test results for effect of position within triplet on mean proportion
transfer by Spanish grapheme-to-sound correspondence by condition……….108
Table 4.9
Mean proportion scores and standard deviations for <ll>-/j/ by position by
Condition………………………………………………………………………110
Table 4.10 Mann-Whitney test results for <ll>-/j/ by position by condition..…………….110
ix
Table 4.11 Cross-tabulations: Mean proportion transfer scores and standard deviations for
round by condition…………………………………………………………….112
Table 4.12 Pearson chi-square results: rounds by condition………………………………112
Table 4.13 Cross-tabulations: Mean proportion transfer scores and standard deviations for
the effect of round for Spanish grapheme-to-phoneme correspondence by
condition……………………………………………………………………......114
Table 4.14 Kruskal-Wallis test results for the effect of round on Spanish grapheme-tophoneme correspondences by condition………………………………………....116
Table 4.15 Pearson chi-square results: pair-wise comparisons of rounds by Spanish
grapheme-to-phoneme correspondence by condition………………………...117
Table 4.16 Individual PM score and mean proportion transfer in orthographic
conditions………………………………,,,……………………………….....122
Table 5.1
Lexical frequency for silent <h> by position within the word in English……133
Table 5.2
Hierarchy of Spanish grapheme-to-phoneme correspondences in accordance
to their corresponding mean proportion transfer..…………………………136
x
List of Figures
Figure 1 The revised working memory model,, (Baddeley, 2003, p.196)………………..…..43
Figure 2 Typical serial position curves observed for different list lengths and presentation
rates in a free recall task, (Murdock, 1962, p. 483)…………………...………....46
xi
List of Appendices
Appendix A Background questionnaire………………………………………………………156
Appendix B Picture-naming task: assigned meanings………………………………...……..161
Appendix C
Picture-naming: stimuli meanings………………………………………………….162
xii
1
Chapter 1
Introduction
1.1 Objectives of the study
The overall objective of this study is to contribute to our understanding of the role of
orthography in the acquisition of second language (L2) phonology by conducting a systematic
examination of the effect of orthography on first language (L1)-based transfer. The purposes of
this study are two-fold. The first is to determine whether orthography will promote transfer
leading to non-target-like productions in novice adult English-speaking learners of Spanish. The
second purpose of this study is to determine the factors that affect the quantity of orthographyinduced transfer. Specifically, this study aims to examine the effect of three main factors on
promoting orthography-induced transfer. The first is the auditory-orthographic condition at
learning and production. In this study, participants were assigned to 4 different conditions in
which the presence of orthographic (written) and auditory input was manipulated. The factor
auditory-orthographic condition specifically refers to the presence or absence of auditory and
orthographic input at learning and/or at production in a given condition. The second factor
examined in this study is the effect of grapheme-to-phoneme inconsistency between English and
Spanish. Grapheme-to-phoneme inconsistency refers to cases in which a shared grapheme
(letter) corresponds to two different phonemes in the target language (TL) and L1. For example,
whereas the grapheme <ll> corresponds to the phoneme /j/ (e.g., <pallete> [pajete]) in the
variety of Spanish used in this study, it corresponds to /l/ in English (e.g., <pillow> [pɪlo]). The
third factor investigated in relation to its impact to orthography-induced phonological transfer is
Phonological memory (PM), in other words our ability to encode and recall verbal information
(e.g., Baddeley & Hitch, 1974, 2000). With respect to PM-related factors, the research to be
2
presented here specifically seeks to examine: (i) the effect of well-known primacy and recency
effects (e.g., Deese & Kaufman, 1957; Murdock, 1962; Waugh & Norman, 1965; Brown &
McNeill, 1966; Horowitz, White & Atwood, 1968; Craik, 1970; Rundus & Atkinson, 1970;
Rundus, 1971; Foreit, 1976; Gathercole & Baddeley, 1993; Gupta, 2005), in which initial and
final word and list items are best recalled in comparison with list- and word-medial items (ii)
repetition effects (e.g., Hebb, 1961; Atkinson & Shiffrin, 1968; Mathews & Tulving, 1973) and
(iii) the relationship between individual variation in (PM) capacity and orthography-induced
transfer.
In this section, I have provided the objectives of this study. In the next section, I will
outline the research questions and an overview of the study. In doing so, I will further elaborate
on the factors whose effect on orthography-induced transfer I have proposed to study.
1.2 Research questions and overview of the
methodology and findings
The research questions in this study are as follows:
1. Does exposure to orthography at learning and/or production promote L1-based phonological
transfer in production leading to non-target-like productions?
2. Do different types of auditory-orthographic condition affect the quantity of orthographyinduced transfer in different degrees?
3. What is the impact of inconsistency between English and Spanish grapheme-to-phoneme
correspondences on the quantity of transfer leading to non-target-like productions?
3
4. What role, if any, does phonological memory play in shaping orthography-induced transfer in
novice adult English-speaking learners of Spanish? Specifically, (i) given that initial and most
recent items are more readily recalled, do primacy and recency effects lower the proportion of
orthography-induced transfer? (ii) Does repetition of grapheme-to-phoneme correspondences in
the TL lower the quantity of transfer? (iii) Is there a negative correlation between PM capacity
and orthography-induced transfer?
In order to examine the above questions, I recruited 45 but tested 40 novice adult
English-speaking learners of Spanish. The participants had no prior knowledge of Spanish nor
had they been previously exposed to it. The participants were required to perform a primary
Spanish-based picture-naming and a secondary non-word repetition PM task in one session. The
purpose of the picture-naming task was to determine the effect of presence of orthography at
learning and production, inconsistency between the L1 and TL grapheme-to-phoneme
correspondences, primacy and recency as well as repetition on the quantity of orthographyinduced transfer. The purpose of the PM task, on the other hand was to determine whether there
was a correlation between PM capacity and orthography-induced transfer in production.
I tested the effect of presence of orthography at learning and/or production on the
quantity of L1-based transfer in production by assigning the participants to 4 different auditoryorthographic conditions. In all the conditions, at learning, the learners heard the Spanish words
one at a time, together with their assigned images, in groups of three (36 triplets all together),
and were required to learn them. Immediately after learning, the participants were presented
with the three pictures one at a time and were required to name them in Spanish. Whereas the
conditions did not differ in terms of the presence of auditory input at learning and at production,
they varied in terms of the presence of orthography at learning and production. There conditions
4
were as follows: (a) orthography at learning & production; (b) orthography at learning only; (c)
orthography at production only, (c) auditory only.
In order to test the effect of grapheme-to-phoneme inconsistency between English and
Spanish, I included two types of stimuli: (i) words with grapheme-to-phoneme correspondences
that are identical in English and Spanish and (ii) words with grapheme-to-phoneme
correspondences that differ between English and Spanish; they share the same graphemes but
correspond to two different sounds in English and Spanish. An example of a grapheme-tophoneme correspondence that is identical in both English and Spanish is <s>-/s/ word initially
(e.g., <sotera>-[soteɾa] ‘a type of spade in Spanish and <Sam>-[sæm] in English). An example
of a grapheme-phoneme correspondence that is different, as mentioned above is <ll>-/j/ (e.g.,
<pallete>-[pajete]). Whereas <ll> corresponds to /j/ in Spanish, it corresponds to /l/ in English
(e.g., <pillow>-[pɪlo]).
In the picture-naming task, the stimuli also controlled for primacy and recency effects at
the list level and for primacy effects at the word level. Learners were presented with the stimuli
in groups of three and immediately tested on them. Therefore, it was possible to control for
primacy and recency effects by manipulating the position of a given stimuli at learning and at
production. The positions that were controlled for were as follows: (i) first at learning and first
at production, (ii) first at learning and last at production, (ii) middle at learning and middle at
production, (iii) last at learning and first at production and (iii) last at learning and last at
production. At the word level on the other hand, primacy effects were controlled by having two
types of stimuli: (i) words with the target grapheme-to-phoneme correspondence word initially
and (ii) words with the target grapheme-to-phoneme correspondence word medially. For
example, <ll>-/j/ was word-initial in <llanura>-[januɾa] and word medial in <pollero>-[pojeɾo].
5
In addition to controlling for the effect of primacy and recency effects, I also controlled
for the effect of task repetition. In order to examine the effect of task repetition on the quantity
of orthography-induced transfer, I asked the participants to repeat the picture-naming task three
times (three rounds) in the same session. This would show whether learning of TL grapheme-tophoneme correspondences would occur as learners were exposed to the target graphemephoneme correspondences in Spanish.
Finally, in an effort to see whether PM capacity impacts the quantity of orthographyinduced transfer in production, upon the completion of the Spanish picture-naming task, I asked
the learners to complete a PM non-word repetition task. In the latter task, the learners heard
Farsi words and were required to repeat each word immediately after hearing it. The words
varied in the number of syllables (e.g., 3-9 syllables).
The overall results showed that presence of orthography at learning and/or production
promotes phonological transfer in the production of novice adult English-speaking learners of
Spanish. For example, in the presence of a Spanish grapheme-to-phoneme correspondence that
was different from English (e.g., <z>-/s/ in Spanish vs. <z>-/z/ in <zoo>-[zu] in English), the
learners substituted their L1 phoneme for the TL phoneme (e.g., the <zatico>-[satiko]) was
produced as [zatiko]). In addition, when the grapheme-to-phoneme correspondences that
resulted in transfer were collapsed, the following hierarchy was obtained with respect to the
effect of the presence of orthography at learning and/or production, wherein the quantity of
transfer significantly decreased from left to right: orthography at learning & production ~
orthography at learning > orthography at production > auditory only. I now turn to the rationale
and contributions of this study.
6
1.3 Motivation and contributions
The present study promises to make a number of important contributions to our
understanding of the influence of written language on oral production. First, although there is a
growing body of research on the effect of orthography on the acquisition of a L2 phonology
(e.g., Young-Scholten, Akita & Cross, 1999; Young-Scholten, 2002; Erdener & Burnham, 2005;
Steele, 2005; Bassetti, 2007; Escudero, Hayes-Harb & Mitterer, 2008; Escudero & Wanrooij,
2010; Hayes-Harb, Nicol & Barker, 2010), not much is known about the factors that promote
orthography-induced transfer. By studying the effect of exposure to orthographic input at
learning and/or production, this study will provide a more fine-grained picture of the degree to
which different auditory-orthographic modalities affect transfer in production. Analyzing the
effect of auditory-orthographic condition at the time of learning and/or production is also
important because it has pedagogical implications for pronunciation teaching. Most language
instructors and/or researchers would agree that orthography is a major source of input in the
process of L2 acquisition for learners. Not only L2 learners who are literate heavily rely on
written input, but also traditionally, language instructors have heavily used written material in
classroom settings for teaching foreign language pronunciation (Erdener & Burnham, 2005).
The findings in the present work point to a negative effect of orthography on L2 production and
highlight an advantage for an audio-lingual method for pronunciation teaching (e.g., Richards &
Rodgers, 2001). Specifically, it is recommended that in order to enhance learners’
pronunciation, in cases in which grapheme-to-phoneme correspondences differ between English
and Spanish, language instructors expose learners to the auditory input, prior to exposing them
to orthographic input.
7
Second, the research presented in this thesis is important because, even though most
previous L2 studies have involved literate learners, most often-cited models of acquisition of L2
phonology (e.g., Flege, 1995; Brown, 1998, 2000) have not incorporated the role of orthography
in their predictions. That is, despite the fact that most participants in L2 research studies are
literate learners, these models have mainly been concerned with the influence of phonetic (e.g.,
Flege, 1995; Best & Tyler, 2007) and phonological categories (Brown, 1998, 2000) on L2
acquisition. The only exception is Best and Tyler’s (2007) PAM-L2 that briefly mentions that
orthography may bias category assimilation of new L2 sounds. That literacy and orthography
affect L1 categories has been previously argued in L1 literature (e.g., Burnham, Earnshaw &
Clark, 1991; Burnham, 2003; Mazzaro, 2011; Ranbom & Connine, 2011). If orthography can
affect L1-based representations, then it is important to investigate and formalize the effect of
orthography in L2 acquisition. Currently, the body of empirical evidence on the influence of
orthography on L2 phonological acquisition is growing (e.g., Young-Scholten, Akita & Cross,
1999; Young-Scholten, 2000, 2002; Erdener & Burnham, 2005; Steele, 2005; Hayes-Harb,
Nicol & Barker, 2010). The present study will further shed light on our understanding of the role
of orthography in L2 acquisition, specifically on orthography-induced transfer and will highlight
the need for future models to incorporate the role of orthography in their formulation of
hypotheses on L2 phonological acquisition.
Third, this study will contribute to our understanding of the PM effects on orthographyinduced transfer. Previously, phonological memory, the working memory component
responsible for storing verbal/acoustic information (Baddeley and Hitch, 1974, 2003) has been
shown to positively impact vocabulary learning. For example, primacy and recency effects
(Brown & McNeill, 1966), namely a higher probability of recall for list initial and final items,
have been shown to exist in L1 vocabulary learning, and repetition has been shown to enhance
8
L2 vocabulary learning (e.g., Saragai, Nation & Meister, 1978; Horst, Cobb & Meara, 1998;
Waring & Takaki, 2003; Webb, 2007). In addition, a positive correlation has been reported
between individual learners’ PM capacity and L2 vocabulary learning (e.g., Service 1992;
Service & Kohonen, 1995; Cheung, 1996; Dufva & Voeten, 1999; French, 2004). Given that
both the process of vocabulary learning as well as learning new grapheme-to-phoneme
correspondences entail encoding TL sounds in memory, it is plausible that memory effects will
impact orthography-induced transfer. However, the effect of PM on orthography-induced
transfer has not been previously investigated and needs further examination.
In addition to the fact that the factors explored in this study will help us gain a better
understanding of the effect of orthography on phonological transfer, the population investigated
here is generally understudied. In particular, there are very few studies that have focused on the
absolute initial stage of acquisition. This is especially true for studies that have examined the
role of orthography in L2 phonological acquisition (e.g., Erdener & Burnham, 2005). Studying
novice learners will show to what extent we can expect to see cross-linguistic influence at the
very initial stages of L2 acquisition.
Finally, this study will contribute to the empirical body of evidence on orthographyinduced transfer in the acquisition of Spanish, which is very limited. Currently the only studies
that have examined the effect of orthography on L2 acquisition in Spanish are Erdner and
Burnham (2005) and Zampini (1994,1997). Because the learners who participated in the present
study did not have any exposure to Spanish prior to this study, such an empirical study lays the
ground work for future studies that would want to investigate the effect of orthography on
phonological transfer in intermediate and advanced learners of Spanish. In other words, the
9
present study lays the foundation for future research that seek to investigate the effect of
orthography in the development of L2 phonology.
1.4 Thesis structure
This thesis is composed of five chapters. In the present chapter, I have introduced the objective
of the study, provided an overview of the research questions, the methodology and the findings
and pointed to the factors that motivated this study as well as its contributions. Chapter 2
reviews the literature on transfer in light of major models of acquisition of L2 phonology and
the effect of orthography on L1 and L2 acquisition. It also reviews the influence of PM in L2
acquisition, with a focus on vocabulary acquisition. Specifically, I define PM and discuss its
different aspects such as primacy and recency effects as well as the effect of repetition. In
addition, I provide an overview of the studies that have examined the role of individual variation
in PM capacity and L2 acquisition. Chapter 3 presents the hypotheses and outlines the
methodological design of the experimental study. In this chapter, I describe the two tasks,
namely the picture-naming task and the Farsi-based non-word repetition task as well as the data
collection procedure by describing the participants, the procedures, stimuli, and testing protocol.
Chapter 4 presents the data analysis and results for both the picture-naming and non-word
repetition task. I conclude with Chapter 5, where the findings are discussed in light of the
previous literature; contributions, implications and limitations are highlighted; and future studies
are suggested.
10
Chapter 2
Previous research on transfer, orthography and phonological
memory in the acquisition of L2 phonology
In the previous chapter, I provided the objective, research questions and an overview of the
study to be presented in Chapters 3 and 4. Specifically, I stated that this study is concerned with
the overall effect of orthography on phonological transfer as well as the factors that shape
orthography-induced transfer, including grapheme-to-phoneme inconsistency between the L1
and TL, auditory-orthographic condition and the PM-related factors of primacy and recency
effects, repetition and PM capacity. In this chapter, I will review the previous studies that are
related to phonological transfer, effect of orthography in L1 phonology and L2 acquisition as
well as PM in the context of L2 acquisition.
The remainder of this chapter is structured as follows. Because the present study is
concerned with the overall effect of orthography on phonological transfer, in reviewing previous
research, I first provide an overview of the formalization of transfer in models of L2 acquisition
of phonology (Section 2.1). A review of some of the prominent models of L2 phonological
acquisition shows that whereas transfer has been formalized in light of phonetic (e.g., Flege,
1995; Best & Tyler, 2007) and phonological categories (Brown, 1998, 2000), the influence of
orthography has not been fully incorporated in these models. Indeed, the only model that briefly
mentions a potential role for the effect of orthography in the perception of new TL sounds is
Best and Tyler (2007). I then discuss the role of orthography in L1 phonology (Section 2.2)
prior to moving to reviewing the studies that have investigated the effect of orthography in L2
phonological acquisition (Section 2.3). In discussing the role of orthography on the acquisition
of L1, I highlight the importance of the effect of onset of reading (e.g., Burnham et al., 1991;
11
Burnham, 2003) in the formation of L1 categories, the effect of literacy in L1 perception (e.g.,
Morais, Alegria, Cary, & Bertelson, 1979; Bertelson, De Gelder, Tfouni, & Morais, 1989; Reid,
Zhang, Nie & Ding, 1986) and production (e.g., Mazzaro, 2011). This section demonstrates that
much work has focused on the importance of orthographic inconsistency in L1 perception (e.g.,
Seidenberg & Tanenhaus, 1979; Taft & Hambly, 1985; Zeigler & Ferrand, 1998; Halle,
Chereau, & Segui, 2000; Tyler & Burnham, 2006; Ranbom & Connine, 2011). Orthographic
inconsistency refers to the extent that a single grapheme corresponds to one or more phonemes
in the same language (Ranbom & Connine, 2011). For example, whereas the grapheme <b>
only maps to the phoneme /b/ in English, the grapheme <s> can correspond to either /s/ or /z/
depending on the context (eg., initial position - <Sue>-[su] versus medial or final position <rosy> [ɹozi], <rose> [ɹoz]). In addition, orthographic inconsistency refers to the extent to
which a phoneme maps to different graphemes in the same language (Ranbom & Connine,
2011). For example, whereas the grapheme <b> alone represents the phoneme /b/ in English,
both <gh> and <f> represent /f/ (e.g., <cough>, <fish>). In Section 2.3, the studies on the effect
of orthography on L2 phonological acquisition report both positive and negative effects of
orthography in the acquisition of L2 phonology. Because the study to be presented in Chapters 3
and 4 is concerned with the effect of inconsistency between L1 and TL grapheme-to-phoneme
correspondences as well as the effect of auditory-orthographic condition on shaping
orthography-induced transfer, particular attention is paid to previous studies that have
investigated the former (e.g., Young-Scholten, 2000; Erderner & Burnham, 2005) and the latter
factors (e.g., Young-Scholten et al., 1999; Young-Scholten, 2002; Erdener & Burnham, 2005).
Other factors, including auditory-orthographic condition (Young-Scholten, 1999; Erdener &
Burnham, 2005), are also discussed in this section. In Section 2.4, I move to reviewing the
relationship between primacy and recency effects (e.g., Deese & Kaufman, 1957; Murdock,
12
1962; Brown & McNeill, 1966; Gathercole & Baddeley, 1993; Gupta, 2005), repetition (e.g.,
Waring & Takaki, 2003; Webb, 2007) and PM capacity on L2 acquisition. There are currently
no previous studies that have examined the effect of these factors on orthography-induced
transfer, therefore, I focus on studies that have been concerned with these PM-related factors in
the context of vocabulary learning (e.g., Service 1992; Service & Kohonen, 1995; Cheung,
1996; Dufva & Voeten., 1999; French, 2004). I now turn to the discussion of transfer and
orthographic influence in some of the prominent models of L2 acquisition of phonology.
2.1 Transfer and orthographic influence in models of the
acquisition of L2 phonetics and phonology
In this section, I will first review some of the well-known characterizations of transfer (e.g.,
Weinreich, 1953; Odlin, 1989; Kellerman & Sharwood Smith, 1986). This will be followed by
an overview of the ways in which transfer has been formalized in certain often cited models of
the acquisition of L2 phonology; particular attention will be paid to the limited extent to which
orthography has been integrated as a conditioning factor for promoting phonological transfer. In
reviewing these models, it will become apparent that, while a central role is always posited for
L1 influence, such influence is normally formalized via the interaction of L1 phonological or
phonetic categories with the TL input. Moreover, with the exception of Best and Tyler (2007),
such models do not integrate the effect of orthography on learners’ speech learning.
In the general L2 literature, a number of terms have been used to characterize L1
influence during L2 acquisition. One of the most well known of this was proposed by Weinreich
(1953) who used the term interference to refer to “instances of language deviation from the
norms of either language which occur in the speech of bilinguals as a result of their familiarity
13
with more than one language” (p.1). Odlin (1989), on the other hand, makes reference to
transfer as “the influence resulting from similarities and differences between the target language
and any other language that has been previously (and perhaps imperfectly) acquired” (p.27).
Whereas Weinreich’s characterization of L1 influence proposes that transfer has a negative
effect and is a phenomenon leading to non-target-like production, Odlin’s (1989) definition is
more comprehensive, as it includes both the facilitative and hindering effects of L1 influence.
Terms other than transfer have been used to refer to the influence of L1 on L2. One of these is
cross-linguistic influence, which was first proposed by Kellerman and Sharwood Smith (1986).
This term was deemed more appropriate because it was considered to refer to a wide range of
domains in which a person’s knowledge of one language could influence another including the
influence of an L2 on an individual’s L1. In present work, the terms transfer and cross-linguistic
influence will be used interchangeably.
Having briefly reviewed the general conceptualization of transfer, I will now describe
how it has been formalized in some of the most often cited models of the L2 acquisition of
phonology, highlighting the limited degree to which orthography has been integrated as a factor
promoting phonological transfer. The influence of L1 categories on L2 categories has been
formalized mainly in the area of perception via acoustic properties or articulatory gestures (e.g.,
Flege, 1995, inter alia; Best & Tyler., 2007) or, much less often, via phonological features (e.g.,
Brown, 1998, 2000). Flege’s Speech Learning Model (1995, inter alia) formalizes transfer in
terms of the influence of L1 phonetic categories on the perception and subsequent acquisition of
TL sounds. This model proposes that (i) ‘old sounds’, that is sounds that already exist in the L1,
do not need to be acquired and will not be problematic for learners; (ii) ‘similar’ sounds, that is
sounds that are phonetically similar to L1 sounds, will be difficult to acquire if they are
acoustically too similar and thus not differentiated as being different by learners; and (iii) ‘new’
14
sounds that are phonetically different from the L1 sounds will be easily acquired. In sum,
transfer is formalized in terms of the use of L1 phonetic categories for parsing the TL input. The
potential role of orthography on the acquisition of TL sounds is not considered in this model.
The issue of transfer has also been addressed in other phonetic models that seek to
formalize L1 influence on cross-linguistic perception including Best and Tyler’s (2007) PAML2, a revised version of Best’s (1995) Perceptual Assimilation Model (PAM). In this model,
transfer is conceptualized in terms of the use of L1 phonetic categories – in this model,
articulatory gestures – in the perceptual assimilation of TL sounds. PAM-L2 is based on the
premise that sounds are perceived in terms of articulatory gestures (e.g., Browman & Goldstein,
1989, 1990, 1992, 1995). Moreover, it is the patterns of assimilation to L1 categories that
determine the accuracy of the discrimination of TL contrasts and subsequent category
formation. Very good to excellent discrimination is predicted for Two Category assimilation, in
which two TL phones are perceived as acceptable exemplars of two different L1 phones; poor
discrimination is predicted when two TL sounds are perceived as equally good or poor
exemplars of the same L1 phoneme; and Single Category assimilation and intermediate
discrimination is predicted when two TL sounds differ in the extent to which they are good
exemplars in relation to a single L1 phoneme. Of all the formal models of (perceptual) speech
learning, this model is the only one that mentions the effect of orthography on the categorization
of TL sounds, albeit very briefly. Specifically, it refers to the possible biasing effect of L1
orthography on the categorization of new sounds. In the particular case discussed, the authors
propose that the French uvular rhotic may be perceived as the English approximant rhotic
because English and French share the same grapheme <r> to represent these sounds. Although
this model explicitly mentions the potential role of orthography in the categorization of a sound
that does not exist in a learner’s L1, it does not state whether orthographic inconsistency may
15
play a role in the categorization of a TL sound that already exists in the L1. For example, it does
not discuss the effect of orthography when a shared grapheme such as <z> in Spanish and
English, corresponds to [z] in the TL but to [s] in the L1 and [s] is an old sound in the L1. As
will be discussed in Section (2.3.2), Young-Scholten (2000) has shown that it is possible for
orthography to promote phonological transfer by showing that orthography inhibits the German
final devoicing rule (e.g., the realization of /b,d,g/ as [p,t,k] in word final position) for English
learners. That is while /p,t,k/ exist in English, because these phonemes correspond to the
graphemes <b,d,g> respectively, they are realized as [b,d,g] by English-speaking learners,
although they exist in English. In this thesis, I will also focus on examining the effect of
orthography in the presence of grapheme-to-phoneme inconsistencies, where the TL already
exists in the L1.
As mentioned above, the influence of the L1 sound inventory on L2 category formation
has also been formalized in phonological terms. In Brown’s (1998, 2000) model, learners’
perceptual categorization depends on feature geometry (e.g., Clements, 1985). According to
Clement’s theory of feature geometry, phonemes consist of an internal structure composed of a
hierarchy of phonological features that are contained in the phonological component of
Universal Grammar (UG). Brown’s model, inspired by Trubetzkoy (1939/1958), proposes that
TL sound structures whose representation involves phonological features absent from the L1
will be impossible to perceive accurately and thus acquire, whereas phonemes whose
representation involves features already employed in the native language phonological inventory
can be acquired by adult L2 learners. In sum, this model acknowledges the role of L1 influence
in the acquisition of L2 phonological acquisition and formalizes transfer via feature geometry.
However, similarly to Flege (1995), it does not address the role of orthography.
16
The models reviewed in this section are concerned with the phenomenon of L1 influence
in perception either phonetically or phonologically. They all acknowledge a role for L1
influence on L2 category formation with the common theme that similarity and differences can
mitigate cross-linguistic influence. In addition, with the exception of Best and Tyler (2007),
most models are only concerned with auditory input as opposed to multimodal input and fail to
address the potential role of orthography, including the effect of mappings between L1 and TL
grapheme-to-phoneme correspondences in the L2 acquisition of phonology. Moreover, while
Best and Tyler (2007) do not discuss the role of orthography in category formation for TL
sounds that already exist in the L1. Given that the majority of adult learners studied in L2
acquisition research are literate, the learners’ degree of exposure to orthographic input in the
process of L2 acquisition (Erdener & Burnham, 2005; Bassetti, 2009), and the large body of
research documenting the effect of orthography on L1 perception and category formation, it is
arguably very important for models of L2 acquisition to address and integrate the potential role
of both orthographic and auditory input in shaping L1-based transfer. I now turn to previous
empirical work that provides evidence for the effect of orthography on L1 perception as well as
L2 phonological acquisition.
2.2 The role of orthography in shaping L1 perception and
production
One of the topics in native speech perception research that has received considerable attention
has been the impact of exposure to orthography. In describing the effect of the role of
orthography on the acquisition of L1, it will become apparent that some of the previous research
has demonstrated the importance of the effect of onset of reading in the formation of L1
17
categories (e.g., Burnham et al., 1991; Burnham, 2003) and others have shown an effect of
literacy in L1 perception (e.g., Morais et al, 1979; Reid et al., 1986; Bertelson et al., 1989) as
well as in production (e.g., Mazzaro, 2011). This section will also the importance of the effect of
orthographic inconsistency in L1 perception (e.g., Seidenberg & Tanenhaus, 1979; Taft &
Hambly, 1985; Zeigler & Ferrand, 1998; Halle et al., 2000; Tyler & Burnham, 2006; Ranbom &
Connine, 2011). Given that the work presented in Chapters 3 and 4 also focuses on the effect of
orthography and considers the factor grapheme-to-phoneme differences in the acquisition of L2
phonology in novice adult literate learners, as well as the assumption that the factors that shape
L1 phonology may also shape L2 phonology, it is important that I review the above mentioned
studies.
Evidence in support of the effect of literacy on speech perception has been provided by
studies that have investigated the effect of the development of reading experience on categorical
speech perception in children and adults (e.g., Burnham et al., 1991; Burnham, 2003). Burnham
et al. (1991) examined categorical speech perception in infants, in English-speaking children
aged between 2 to 6, and with adults ranging in age between18-29. An identification task to test
the perception of native (voiced/voiceless bilabial stops) and non-native (pre-voiced/voiced
bilabial stops) was conducted. Analysis of category boundary sharpness showed a significant
linear trend for age, where perception of the native contrast became increasingly more
categorical with age, especially between two and six. However, for the non-native contrast,
boundary sharpness improved notably between infancy and two years, went down to chance
level at six years, and then improved slightly by adulthood. The researchers attributed the
intensification of language specific speech perception between two and six years to the onset of
reading instruction. It was speculated that language instruction would encourage children to
process language in terms of separate phonemes, and instruction in phoneme-to-grapheme
18
mappings would promote language processing in terms of separate phonemes of the ambient
language and direct children’s attention away from previously perceived non-native contrasts.
Burnham (2003) also explored the relationship between categorical speech perception of
native and non-native linguistic contrasts, age, and onset of reading. This latter study was more
controlled than that of Burnham et al. (1991): the stimuli involved eight contrasts as opposed to
two; both identification and discrimination tasks were included; the effect of synthetic and
natural stimuli was controlled; a wider age range (e.g., 4, 6 and 8) as well as the year of
schooling (Kindergarten, grades 1, 2, 3 and 4) were tested for in the children and results were
correlated with reading ability as measured by a comprehension, articulation, and reading ability
test. The results once again pointed to an increase in language specific speech perception at the
onset of acquiring reading skills. Whereas in Burnham et al. (1991) the results were consistent
with an increase in language specific speech perception between the ages of two to six, in
Burnham (2003), language specific speech perception increased at the age of six. In other
words, categorical speech perception was significantly worse at the age of six than four or eight
years. Moreover, the results showed that six and eight year old children that were good readers
for their age were also those whose speech perception was significantly more influenced by their
L1 phonetic inventory. Furthermore, the results showed that whereas for each of the three sets of
contrasts, peak levels of language specific speech perception occurred at the end of the
Kindergarten year – the point where reading and phoneme segmentation were rapidly improving
– language specific perception was attenuated at the end of the first year of school. In all, both
these studies provide evidence that children’s learning of their L1 orthographic rules at the onset
of reading acquisition shapes their perception. Given the link between the onset of reading and
category formation in L1, it is plausible that, exposure to orthography at the very initial stages of
phonological acquisition in L2 will also interfere with category formation.
19
Literacy impacts not only native speech perception; it promotes also phonological
awareness. Phonological awareness refers to the ability to identify and manipulate sounds in a
language at the following levels of sound structure: syllable, onset, rhyme and phoneme. Morais
et al. (1979) conducted phoneme deletion and phoneme addition tasks with 30 literate and 30
illiterate Portuguese-speaking participants. In the deletion task, the participants had to delete the
first phoneme from the utterance provided by the experimenter. For example, when the
experimenter provided the word /purso/, the participants were required to delete the initial
phoneme in order to produce [urso]. In the phoneme addition task, the participants had to
introduce an additional phoneme (e.g., /m, p, ʃ/) at the beginning of the utterances provided by
the experimenter. For example, upon hearing the word /osa/ and the consonant /m/, participants
were expected to produce [mosa]. Both real words and non-words were included. The results
revealed that literate adults significantly outperformed illiterate ones on phoneme deletion and
addition tasks with both real words and non-words. For example, literate participants’ mean
percentage correct responses for the non-word phone deletion task (71%) was almost four times
that of the illiterate participants (19%). It was concluded that the ability to deal explicitly with
phonetic units of speech is not acquired spontaneously as a normal outcome of cognitive growth
and/or linguistic experience but is rather a consequence of learning to read. Subsequent studies
have shown that literacy plays a role in phoneme manipulation tasks but not other types of
phonological awareness tasks such as rhyme judgment tasks. For example, Bertelson et al.
(1989) carried out two rhyme judgment tasks as well as initial syllabic vowel deletion and initial
consonant deletion tasks with 9 literate and 16 illiterate adult native speakers of Brazilian
Portuguese. In the main rhyme task, the experimenter provided disyllabic, stress-initial rhyming
pairs (e.g., <cola> and <mola>) and non-rhyming pairs (e.g., <nossa> and <sujo>). For each
trial, subjects first repeated the words and then said whether they rhymed or not. The other
20
rhyme judgment task included words that represented different patterns of phonological overlap
from rhyming assonances: pairs of words with same vowels, different consonants (e.g., <bota>
and <sola>), pairs of words with same stressed beginning (e.g., <faca> and <fado>). As with the
first rhyme judgment task, participants repeated the words and stated whether they rhymed or
not. The latter rhyme-judgment task was administered only for participants that had reached the
criterion of six consecutive correct responses in the main task. In the vowel deletion task,
participants were asked to elide the initial /a/ from a VCV or VCVC pseudo-word uttered by the
experimenter, and produce the resulting string. For example, if the experimenter said <ako>,
participants needed to say <ko>. Similarly, a consonant deletion task involved deleting the
initial /f/ of a CVC pseudo-word. Their results showed that, among the illiterate participants,
while the majority reached the criterion in the first rhyme-judgment task and in the vowel
deletion task, all failed in the consonant deletion task; these differences were significant. On the
other hand, the literate participants all reached the criterion in the main rhyme-judgment task
and all but one and two participants in the vowel deletion and consonant reached the criterion in
the deletion tasks, respectively. The differences between the two groups of participants on each
task were found to be non-significant for the main rhyme-judgment task and vowel deletion
task. However, the group differences were highly significant for the consonant deletion task.
Moreover, the results of the second rhyme-judgment task showed that literate and illiterate
learners performed similarly with the exception that literate participants tended to accept pairs
with same endings as rhymes more frequently than the illiterate participants.
Other research has demonstrated that the relationship between literacy and phonological
awareness is not observed with speakers of all languages. For example, Reid et al. (1986)
compared 12 adults who had learned Hanju Pinyin (alphabetic group) with 18 adults who had
learned Chinese characters (non-alphabetic group). The two groups were similar in education:
21
the mean years of schooling for the alphabetic group was 10 years and 7 for the non-alphabetic
group. In their experiment, the phoneme to be added or deleted was /d/, /s/, or /n/. For example,
in the phoneme addition task, the experimenter would provide the word /an/ as well as the
phoneme /s/ and the participants would have to say /san/. In the same way that the literate
participants in Morais et al. (1979) outperformed their illiterate counterparts, in this later study,
the differences between the proportion of correct trials between the alphabetic and nonalphabetic groups for both the addition and deletion tasks were highly significant. These results
led Reid et al. (1986) to conclude that it is alphabetic literacy in particular as opposed to any
other type of literacy (e.g., logographic) that leads to superior phonological awareness. That is,
they suggested that the phonemic segmentation skills only develop in the process of learning an
alphabetic language where, in order to read and write alphabetically, learners must learn to
segment spoken syllables into phonemes.
The fact that orthography impacts native language speech perception is also evident in
studies that examine the impact of orthographic inconsistency on orthographic activation during
word recognition. One study that examined the effect of orthographic consistency on rhyme
judgments was Seidenberg and Tanenhaus (1979). In this study, rhyming words with identical
coda spellings (e.g., light, bright) were compared to rhyming pairs with different coda spellings
(e.g., key, knee). The effect of orthographic consistency on rhyme judgments was examined in
three similar experiments. In the first experiment, 40 English-speaking undergraduate students
were tested. On each trial, the participants were presented with a word in isolation (the cue),
followed two seconds later by a list of five words at a rate of one second per word. Their task
was to detect the single word that rhymed with the cue. Cues were presented in two modes. In
the auditory mode, the participants heard the cues prior to the auditory list. In the visual mode,
participants read cues aloud from index cards prior to hearing the target list. The same target
22
lists were used in both tasks. The reaction times showed that, whereas the effect of orthographic
type was highly significant, there was no effect of cue mode (i.e., the magnitude of the
orthographic effect was similar for both cue presentation modes: 56ms for auditory presentation
and 48 ms for visual presentation). The first experiment was replicated with a larger stimuli set
in the auditory mode only. As in the first experiment, rhyme monitoring latencies were
significantly faster by 63ms with orthographically similar rhymes than with orthographically
dissimilar rhymes. In a third experiment, participants were presented with a pair of stimuli that
either had orthographically identical or dissimilar rhymes and were asked to say whether they
were the same or different. As in the previous experiments, there was a highly significant effect
of orthography: for the rhymes, reaction times with orthographically similar pairs were 99ms
faster than reaction times with dissimilar pairs. The opposite pattern was obtained for nonrhymes with reaction time to orthographically similar pairs 58ms longer than with dissimilar
pairs. The researchers proposed that their results supported the role of orthography in processing
spoken words. Specifically, they suggested that orthographic information is automatically
activated in word recognition and rhyme-judgment tasks. If orthographic information is
activated in word recognition and rhyme-judgment tasks, it is plausible that it will also be
activated in the acquisition of L2 phonology.
Further evidence that conflicting orthography and phonology play a role in speech
processing was provided by Taft and Hambly (1985). In an effort to determine whether words
are represented morpho-phonemically as opposed to orthographically, they conducted a syllable
monitoring task with 12 adult English-speaking participants. Two groups of stimuli items were
created. The first vowel of each of the words was the reduced vowel /ə/. Each word was
preceded by a target phoneme string which consisted of the first two consonants of the word
surrounding a full vowel. This vowel was either consistent with the spelling of the reduced
23
vowel (e.g., <val>-/væl/ and <validity>-/vælɪdəti:/) or inconsistent (e.g., <vol>-/vɔl/ and
<validity>-/vælɪdəti:/). In the first group of items, the full value of the reduced vowel could be
determined from morpho-phonologically related forms (e.g., /væl/ from /væl/ of /væləd/). For
the other group, orthography was the only indicator of the full vowel (e.g., <lag>-/læg/ and
<lagoon>- /ləgu:n/). Participants were instructed with consistent and inconsistent stimuli and
were instructed to say ‘yes’ if they thought they heard the string of phonemes that was presented
to them in the word that followed it (e.g., /læg/ in /ləgu:n/) and to say no if they did not hear the
string of phonemes which they were presented with in the word that followed it. It was predicted
that, if orthography were to affect lexical representations, in addition to stating that they heard
strings with a full vowel that were a morpheme in the words with which they were presented
(e.g., /hɔr/ was given as a morpheme in /həɹajzən/), participants would also identify syllables
that contained a full vowel but that were not a morpheme in the word with which it was
presented (e.g., /læg/ is not a morpheme in /ləgu:n/), as the beginning strings of the words. The
results showed that there was a highly significant main effect of consistency in both the
participant and item analysis. In other words, the full vowel status of the word was determined
by orthographic similarity and not by morpho-phonological similarity. They proposed that
orthography is involved at the post-lexical access level of processing.
Ziegler and Ferrand (1998) also examined the effect of orthographic consistency via an
auditory lexical decision task with 82 French-speaking university students. Participants were
presented with two types of words: ‘consistent’ words whose rhymes could be spelled in only
one way (e.g., <stage> in which the rhyme can only be spelled with <age> as in <stage>,
<rage>, <cage>) and ‘inconsistent’ words whose rhymes could be spelled multiple ways (e.g.,
the rhyme in <plomb>- /plɔ̃/ can be spelled as in <nom>, <prompt>, <ton>, <tronc>, <long>).
After the presentation of each stimulus, the participants were asked to decide whether the word
24
was a real word in French. The results clearly showed a strong consistency effect for words.
Reaction times were significantly longer and the error rates were higher for inconsistent words
(21%) in comparison with consistent words (8%). The authors concluded that the existence of a
consistency effect in auditory word recognition provides further evidence for the claim that
orthographic information affects the perception of spoken words.
Halle et al. (2000) also studied the effect of orthographic and phonological inconsistency
on the perception of /b/ and its devoiced counterpart [p] with 14 French-speaking adult
undergraduate students. Words in which voicing assimilation alters the production of a
preceding phoneme were used to test the effect of orthography on phonetic judgments via
phoneme gating and monitoring tasks. French words including the voiced bilabial stop and its
corresponding devoiced counterpart (e.g., <absurd>-[apsyʀd]) were tested. In the phoneme
gating task, participants were presented with a portion of the real words such as [psyʀd] from
<absurd> and had to write down what they heard. There was a highly significant response by
item type interaction: when the lexical content of the word was not provided, listeners most
often heard [p] than [b]. In a follow up phoneme monitoring task, participants were presented
with words whose orthographic code and phonological forms were consistent such as
<capsule>-[kapsyl] and those whose orthographic codes and phonological forms were
incongruent such as <absurd>-[apsyʀd]. The participants then had to identify the target
phonemes by pressing a key on the computer. Participants were largely influenced by the
graphic codes of the words since [b] was detected more often than [p] in words such as
<absurd>. The authors proposed that, when interpreting speech sounds, listeners exploit a
language’s orthographic regularities. They suggested that the interference was not strictly postlexical and did not result from the recognition of a specific lexical entry but was rather more
likely conveyed by a cohort of similar words.
25
The effect of grapheme-to-phoneme congruency on speech processing has also been
shown to play a central role in phonological awareness tasks in Tyler and Burnham (2006). In
this study, the researchers examined the effect of orthographic congruency on phoneme deletion
in 48 adult English-speaking participants. They conducted four experiments in which
participants were presented with orthographically matched stimulus-response pairs (e.g.,
<wage>-<age>) and mismatched pairs (e.g., <worth>-<earth>) and had to delete the initial
sound in the word with which they were presented (e.g., /w/ in [wejdʒ]). In the first experiment,
the participants were presented with orthographically congruent and incongruent stimuli and
were instructed to delete a sound from the stimuli. The results suggested an effect of
orthographic incongruency with reaction times for incongruent orthographic items (1,378ms)
being significantly longer than the congruent ones (1,034ms). In order to determine whether
orthographic influence is automatic and unavoidable as opposed to an optional strategy, in the
second experiment, participants were specifically instructed not to use spelling. As in the first
experiment, there was a significant interaction between congruency and experiment by
participant; reaction times for incongruent items were longer than for congruent ones, albeit the
differences in the second experiment were smaller (between 900-1000ms and about 1100ms for
the congruent and incongruent stimuli respectively). Based on the results of the first and second
experiments, the authors speculated that alphabetically literate adults perform well on phoneme
deletion tasks because orthographic processes play a role. The same experiment was replicated
without the use of carrier sentences which the authors thought might have biased the use of
orthography in the first two experiments. Nonetheless, similar results were obtained where
reaction times were significantly longer for incongruent stimuli (1.422ms) than congruent ones
(1,042ms). In a fourth experiment, the effect of type of onset was controlled with half of the
stimuli having a single consonant onset and the other half a consonant cluster (e.g., <pl>). As in
26
the previous experiments, there was a main effect of congruency, with the incongruent mean
reaction times (1,000ms) being significantly longer than congruent mean reaction times
(992ms). In addition, there was no effect of onset complexity. Based on these results as well as
participants’ remarks that when performing the phoneme deletion tasks, they imagined the letter
to be deleted, Tyler and Burnham (2006) proposed that, at the very least, the orthographic forms
of words are active during phoneme deletion tasks. In addition, in line with Ehri’s (1980, 1984)
orthographic image hypothesis which states that phonological awareness is enabled using
orthographic images of speech sounds, it was proposed that alphabetically literate learners can
perform phoneme deletion tasks not because they have direct phonological awareness but
because they have learnt about speech sounds via learning grapheme-phoneme correspondences.
The effect of orthographic consistency has also been examined using spoken word
recognition tasks in Ranbom and Connine (2011) who tested 24 English-speaking undergraduate
students. The effect of silent letters such as the <t> in <castle> on speech processing was tested
in both discrimination and priming tasks. For example, in a same/different discrimination task,
participants were presented with stimuli including pairs consisting of a correct pronunciation of
the word and a spelling-based pronunciation of the word in which the silent letter was realized
(e.g., [kæsl̥ ]and [kæstl̥ ] for <castle>) and another set of stimuli including pairs consisting of a
correct form of the word and a mispronunciation that was not related to the spelling of the word
([hæstl̥ ] for [hæsl̥ ], <hassle>). The rate of ‘same’ responses for mispronunciations for the stimuli
with a silent letter in their orthographic form (e.g., [kæstl̥ ] for <castle>) was significantly higher
than for those that were not related to the words’ orthographic forms (e.g., [hæstl̥ ] for <hassle>).
The results of the priming tasks also showed that whereas priming for orthographically related
misproductions (e.g., [kæstl̥ ]) and their citation form counterparts (e.g., [kæsl̥ ]) was equivalent,
with orthographically unrelated misproductions (e.g., [hæstl̥ ]), there was a significantly reduced
27
priming relative to the citation form (e.g., [hasl̥ ]). The authors subsequently performed a
frequency analysis for a subset of the silent letters in their stimuli (e.g., <st>, <mb>, <sw>,
<sth>, <ght>, <pb> and <bt>). The analysis showed that only 18.3% of the silent-letter strings
were produced with a silent letter in English. Due to the prevalence of the pronounced forms
and the rarity of the silent forms in English, it was proposed that the default form in English is
the pronounced form and the non-silent letter phonological form is lexicalized. In line with the
phonological restructuring view which proposes that learning to read modifies existing
phonological representations in the lexicon, where orthographic neighborhood density
(frequency) can affect a phonological representation (Metsala & Walley, 1998; Ziegler,
Muneaux, & Grainger, 2003), the authors suggested that combined experience with written and
spoken forms results in two phonological representations: an orthographic-based representation
resulting from learning to read and another phonological representation consistent with the
spoken form. In other words, learning to read not only results in a more fine grained
phonological representation, but also creates an additional phonological representation that
contains features or segments not present in the spoken form. If there are orthographic-based
representations in L1 phonology, it is also plausible that there would be orthographic-based
representations in L2 phonology.
Whereas the above studies have focused on the effect of orthography in the perception of
L1 sounds, Mazzaro (2011) has investigated the role of literacy in both perception and
production of L1 sounds. In a sociolinguistic study of the Spanish fricatives /f, x/ and
the approximants [β, γ], Mazzaro (2011) found that one of the most significant factors that
affects the perception and production of such sounds was participants' level of formal education.
The research included 22 native speakers of Argentine (Corrientes) Spanish (7 (semi)illiterate
and 15 literate), who performed sociolinguistic interviews and perception (AX discrimination
28
task)
and
production (picture
naming
task)
experiments.
Her
results
showed
a
significantly higher rate of labio-velar alternation ([f]>[x] and [β]>[ɣ]) in the speech of
illiterates. More importantly, the results of the perception experiment showed that illiterate
participants had a significantly higher rate of non-target percepts. Specifically, [f] was
misperceived as [x] and [β] as [ɣ] in the context of following [u] and [w]. Mazzaro's (2011)
study shows that literacy is an important factor that can block the spread of perceptually driven
sound variation. In other words, the study indicates that literate speakers’ production of
approximants ([ɣ] and [β]) and fricatives ([f] and [x]) tends to be more in line with the graphic
representation of such sounds. On the other hand, illiterate speakers’ perception and production
of fricatives and approximants was variable ([ɣ] ~ [β], [f] ~ [x]) in certain phonetic contexts.
Given the proposed effect of orthography in L1 categories, it is plausible that literacy in the L1
and L2 may also affect the formation of L2 categories.
In sum, the studies reviewed in this section underline the importance of the role of
literacy in L1 perception, for the most part, and point to the possibility that (1) inconsistency
between orthography and phonology can affect phonological processing in adults (e.g.,
Seidenberg et al., 1979; Taft & Hambly, 1985; Zeilger & Ferrand., 1998; Halle et al., 2000;
Tyler & Burnham, 2006; Ranbom & Connine, 2011); (2) onset of reading acquisition,
specifically learning grapheme-to-phoneme correspondences promotes language specific
perception in children (Burnham et al., 1991; Burnham, 2003) and (3) literacy affects both L1
perception (e.g., Morais et al., 1979; Reid et al, 1986, Mazzaro, 2011) and production (Mazzaro,
2011). As mentioned previously, considering the evidence in support of the role of orthography
in L1 phonological category formation, one would expect orthography to also play a role in the
development of L2 phonology. For example, if there are orthography-based categories in L1
phonology, it is plausible that inconsistency between L1 and TL grapheme-to-phoneme
29
correspondences would also interfere with the acquisition of Spanish phonemes in literate
learners and promote non-target-like category formation. Therefore, it is important for the
perception-based models of L2 which investigate L2 acquisition in the literate population, to
incorporate the role of orthography in transfer and category formation. I now turn to the relevant
studies that have investigated the role of orthography in the acquisition of L2 phonology.
2.3 The role of orthography in L2 phonological acquisition
The review of some of the prominent models of the acquisition of L2 phonology in Section 2.1
revealed that, with the exception of Best and Tyler’s (2007) PAM-L2, none of the models
explicitly consider the effect of orthography on L1-based phonological transfer and category
formation. However, there is some evidence in previous research to suggest that orthography
can both promote and hinder the acquisition of L2 phonology by affecting the perception and/or
production of TL structures. In this section, I will review both the positive then the negative
effects of orthography.
2.3.1 Positive
effects
of
orthography on
L2
phonological
acquisition
The first of these is Steele (2005), which studied the acquisition of stop-liquid clusters by
beginning Mandarin-speaking learners of French via a word-learning task. One group of
learners (13 participants) was required to complete the task in the absence of orthography and
the other (10 participants) in the presence of orthography. The stimuli consisted of French words
with stop-liquid clusters. The target clusters were controlled for the place and voicing of the stop
as well as the liquid type (e.g., voiceless stop-rhotic cluster as in <patronat> /patʁɔnɑ/; voiced
30
stop-rhotic clusters such as <drapeau> /dʁapo/; voiced stop-lateral clusters such as <blennie>
/bleni/; voiceless stop-lateral clusters such as <plateau> /plato/). The results showed that two
asymmetries were observed in the non-orthographic condition. First, whereas the learners had
target-like and epenthetic realizations of both the stop-liquid and stop-rhotic clusters (e.g., 51%
and 16%, respectively), deletion of the liquid was restricted to stop-rhotic clusters (34%).
Second, whereas deletion was almost non-existent in the voiced stop-rhotic clusters (5%), it
occurred at the rate of 53% in the voiceless stop-rhotic clusters. The results for the orthographic
group suggested that not only was the rate of rhotic deletion in voiceless clusters (33%) lower
than with the non-orthographic group but there was also a greater proportion of target-like
voiceless stop-rhotics realizations (46% versus 26%). Steele’s results also showed that when
deletion occurred, the voiceless stop-rhotic clusters were produced as a single aspirated stop
(e.g., target préfet /pʁefɛ/ was produced as [pχefɛ] or [pʰefɛe]). In light of the phonetic
similarity between aspirated stops in Mandarin and French voiceless stop-rhotic clusters in
which the rhotics tends to be realized as a voiceless fricative, Steele proposed that, in the
absence of orthography learners perceive these clusters as an aspirated stop. He further argued
that the presence of orthography leads to the realization that the target forms involve two
segments and not one and thus results in a higher proportion of target-like realizations.
There is also evidence that positive effects of orthography on production can be
modulated by L1 orthographic depth. The depth of an alphabetic orthography refers to the
degree to which it deviates from one-to-one grapheme-to-phoneme correspondences. Languages
in which one-to-one grapheme-to-phoneme correspondences are more common are considered
transparent while those in which one grapheme can correspond to more than one sound and vice
versa are considered opaque. Erdener and Burnham (2005) studied the effect of exposure to
different auditory, orthographic, and visual input on pronunciation. In this study, 32 adult
31
Australian speakers of English whose L1 is opaque and 32 adult Turkish speakers whose L1 is
transparent were required to do repetition tasks in which they heard Irish which is considered
opaque and Spanish which is considered transparent. The participants did not have any
knowledge of the target languages tested in this study. The results showed a significant
facilitative effect of the orthographic condition in comparison with the auditory-only condition
for both language groups. For example, Spanish bilabial confusion rates were significantly
lower in the orthographic condition than in the auditory only condition for the Turkish speakers
(4.55% and 14.49% respectively). This was also true for the Australian speakers (15% and 21%)
in the orthographic and auditory only conditions respectively. More importantly, group
interactions also suggested that, when orthography was presented and transparent (e.g., in
Spanish), Turkish speakers made significantly fewer errors than Australian speakers. For
example, the percentage of Spanish /p/ and /b/ confusions in the orthographic condition was
4.55 % and 15% for Turkish and Australian participants respectively. On the other hand, group
interactions revealed that Turkish participants performed consistently worse than their
Australian counter-parts in the orthographic condition when the TL was an opaque language,
namely Irish. The authors speculated that because Turkish is transparent, Turkish speakers
might process orthographic information on a grapheme-to-phoneme correspondence basis which
leads to them being more prone to the facilitative effects of orthography in a transparent
language. They suggested, on the other hand, that Australian speakers process orthographic
information differently. That is, due to the fact that English is opaque, they develop a whole
picture-orthographic representation of lexical items. Hence, they are less affected by the
facilitative effects of orthography in comparison to Turkish speakers. Although English speakers
appear to be less affected by orthography, nonetheless they are affected by it. Therefore, it is
likely that they will also be affected by orthography in learning Spanish as a second language.
32
There is also evidence that exposure to orthography can help to establish lexical
contrasts for auditorily confusable novel words. Escudero et al., (2008) tested the effect of
training with an auditory versus auditory-orthographic task on the acquisition of auditorily
confusable English words that contain the vowels /æ/ and /ɛ/ (e.g., /bæskət/ and /bɛstət/), a
contrast which does not exist in Dutch, with 50 highly proficient native Dutch speakers. In the
auditory only task, participants learned English non-words by matching their auditory forms to
pictured meanings whereas in the orthographic task, the participants additionally saw the written
forms (e.g., <bestet> for /bɛstət/ and <basket> for /bæskət/). Immediately after the completion
of the tasks they tested the learners’ ability to distinguish between the pictures of the words
containing the target competitor contrasts using an eye-tracking paradigm. The results showed
that the group which had received auditory training fixated on both stimuli with /æ/ and /ɛ/
when presented with either /æ/ or /ɛ/ auditory stimuli. In other words, training with auditory
only input did not lead the participants to establish two separate categories for /ɛ/ and /æ/.
However, the group that had received auditory-orthographic training fixated on /ɛ/, when
presented with auditory /ɛ/, and on /æ/, when presented with auditory /æ/. The mean fixation
proportion on target for /æ/ and /ɛ/ in the absence of orthography were 21.6% and 29.8%,
respectively. However, in the presence of orthography, the mean fixation proportion on target
for /æ/ and /ɛ/ were 26.4% and 25.2%, respectively. The authors thus concluded that learners
make use of abstract knowledge (grapheme-to-phoneme correspondences) to establish new
phonological lexical representations.
More evidence on the evidence on positive effects of orthography was also attested in
Escudero and Wanrooij (2010). They examined the effect of orthography on the perception of
Dutch vowels by 204 Spanish learners of Dutch. The participants included two groups of adult
beginner and advanced learners of Dutch. All participants were required to perform a purely
33
auditory categorization task and another auditory-orthographic categorization task in which both
of the items to be classified and the response alternatives were presented auditorily. The
categorization task had an XAB format involving the Dutch contrasts /a-ɑ/, /i-ɪ/, /y-ʏ/, /ɪ-ʏ/ and
/i-ʏ/. In the orthographic task, the participants were presented with the same auditory stimuli but
were required to chose from the orthographic representations of the twelve Dutch monophthong
vowels including <aa>, <a>, <ie>, <i>, <uu>, and <u>. For example, when presented with
auditory /a/, they had to choose between <a> and <aa>. The results showed that, whereas /ɑ/
and /a/ were the most difficult contrast in the auditory-only task, this contrast was the easiest to
identify in the orthographic task for both beginners and advanced learners. In other words, other
vowels were significantly better identified for /a/ and /ɑ/ in the auditory only condition but /a/
and /ɑ/ were identified significantly better than other vowels in the orthographic condition. For
example, the percentage correct answers in the auditory only condition for advanced learners for
the contrast /a-ɑ/ and /i-y/ was 59.5 % and 80.1 % respectively. However, the percentage
classification rate of /a/, /ɑ/, /i/, and /y/ in the auditory-orthographic condition for the advanced
group was 86.7%, 68.1%, 38.3%, and 17.3% respectively. The researchers proposed that the
advantage exhibited for the /a-ɑ/ contrast in the auditory-orthographic condition was due to the
confound effect of Spanish speakers relying on durational cues to discriminate between the
Dutch /a/ which is longer than the Dutch /ɑ/ (Escudero, Benders, & Lipski, 2009) as well as the
number of orthographic symbols. That is, providing the double consonant <aa> and single
consonant <a> in the orthographic condition lead the learners to pay more attention to the
durational cues in the auditory stimuli and resulted in the better identification of /a/ and /ɑ/.
All in all, the above studies suggest that orthography can potentially affect positively the
production of target sequences that are prone to misperception (Steele, 2005) and help
distinguish between confusable vowels (Escudero et al., 2008; Escudero & Wanrooij, 2010). In
34
addition, these studies point to the fact that the degree of L1 opacity determines the extent to
which learners may benefit from the facilitative effects of orthography (Erdener & Burnham,
2005). Specifically, orthography will exert a stronger influence on learners with a transparent
orthography than those with an opaque orthography. In the next section, I will review the
negative effects of orthography, which are central to the research questions regarding the effect
of auditory-orthographic condition and grapheme-to-phoneme inconsistency.
2.3.2 Negative effects of orthography on L2 phonological
acquisition
In the previous section, I discussed the results from a number of studies that demonstrate
the positive impact of exposure to orthography on L2 perception and production. In this section,
I will provide an overview of some of the studies that have highlighted orthography’s potential
negative effects. In doing so, I will review some of the factors proposed in the literature as
responsible for shaping the rate of orthography-induced transfer leading to non-target-like
productions. Some of the studies reviewed in this section demonstrate that the prerequisite for
orthography-induced transfer is a discrepancy between the sounds to which a shared grapheme
in the TL and L1 corresponds. Specifically, in cases when a shared grapheme corresponds to
different phonemes in the TL and L1, learners may substitute their L1 phoneme for the TL
phoneme. In other words, if a hypothetical grapheme <x> corresponds to a hypothetical
phoneme /y/ in the learners’ L1 and to a hypothetical phoneme /z/ in the TL, learners may
produce /x/ instead of /y/. For example, the grapheme <j> which corresponds to /x/ in Spanish
may be produced as [dʒ] by English speakers because it corresponds to /dʒ/ in English (e.g.
Erdener & Burnham, 2005). In addition to pointing out the inconsistency between the TL and
35
L1 as the main prerequisite for orthography-induced transfer, these studies also provide some
evidence that there are other factors that modulate the effect of orthography-induced transfer
such as auditory-orthographic condition of learning and/testing, and the amount of exposure to
TL orthographic input over time.
The first of these studies is Young-Scholten (2000) who proposes that the amount of
exposure to TL orthographic input shapes the rate of transfer in the acquisition of the final
devoicing rule in German. In German, obstruents are devoiced in syllable-final position, though
this is not signaled in the orthography. For example, /bʊnd/ is realized as [bʊnt] and written as
<bund> not as *<bunt>. Young-Scholten (2000) carried out an 11-month, longitudinal study of
three American English-speaking exchange students (aged 15, 16 and 17) studying in Germany
who had had no exposure to German prior to their arrival in the host country. She collected the
data on a monthly basis by asking learners to say the German word for adjectives and nouns
written in English on cards. During these sessions, learners were also required to provide
background information by providing ratings on their own amount of exposure to aural and
auditory exposure. These results suggested that exposure to orthography played a crucial role in
the non-acquisition of the final devoicing rule in German (e.g., the German voiceless /t/ in
/bʊnt/, written as <bund> was erroneously produced as *[bʊnd]). The results also showed that
the learner who reported the highest amount of exposure to written German also had the highest
percentage of orthography-induced transfer in production. This study, to the best of my
knowledge, is the only study that considers the effect of quantity of orthographic input on L2
development. However, as acknowledged by the author, the particular measure of amount of
exposure to orthographic input used had its limitations, given that self-reports may not
accurately reflect a learner’s intake of input. Young-Scholten (2002) states that, while there is a
need for more ‘strictly-controlled’ studies where the learners’ amount of exposure to
36
orthographic input can be manipulated, in practical terms, feasibility issues may interfere with
carrying out such studies. Similarly, Flege (2009) also acknowledges that the amount of input
tends to be estimated and not measured because the information collected regarding the amount
of L1 and TL input a learner might have received is based on self-reports and is therefore
subject to error. He explains that this methodological limitation is due to the fact that collecting
background information on a regular basis from participants is impractical and even unethical
which renders the task almost impossible.
The effect of orthographic inconsistency was also investigated by Bassetti (2007). Her
study differed from those of Young-Scholten (2000) and Erdener and Burnham (2005) in that
she examined the effect of orthographic inconsistency within the TL and not between the TL
and L1. She studied the production of triphthongs in Mandarin in 8 final-year adult Italianspeaking learners of Mandarin studying at a university in Italy who used the alphabetic pinyin
writing system. The learners were required to do a picture-naming task. This study showed that
despite the fact that the learners were not exposed to pinyin orthography during the picturenaming task, their production of triphthongs was affected by the representation of these
tripthongs in pinyin. That is, for example, whereas the vowel /o/ in the triphthongs that were
orthographically represented with three graphemes in pinyin (e.g., /iou/ spelled as <you/>) was
phonetically realized 100% of the time (e.g., [iou]), the same vowel was only produced 57% of
the time when the corresponding triphthongs were spelled with only two graphemes (e.g., <iu>
for /iou/ was erroneously produced as *[iu]). These differences were highly significant and the
effect size was large (η2 = .74). The author attributed these results to pinyin being a transparent
orthography, where one grapheme corresponds to one phoneme in most cases. Specifically, it
was reasoned that the learners’ knowledge of the transparency of pinyin could have lead to the
overgeneralization of this characteristic to triphthongs represented with three graphemes.
37
The presence of orthography at learning and/or testing has also been proposed as a
moderating factor on transfer. For example, task effects were evident in a study of the
production of Polish clusters by 24 novice English-speaking and 14 Japanese-speaking learners
in Young-Scholten et al. (1999). The participants’ ages ranged between 13-44. Japanese and
English participants were chosen because, whereas English allows for more complex syllable
structures (e.g., consonant clusters), Japanese does not do so. The structure of the Polish words
tested were as follows: (i) CC(C)(C)V(C); (ii) C(C)(C)VCC(C); and (iii) CCC(C)VC(C)(C)V.
The experiment consisted of two conditions. In both conditions, participants were given a tape
with recordings of Polish stimuli along with a book that contained their corresponding pictures
and were required to learn the words. The experiment was conducted in three sessions over
several days. Upon finishing the last session, the participants were required to first recite all the
words they had learnt using the picture-book and then perform a picture-naming task in which
they were exposed to the written words. The results showed that the rate of epenthesis in
contrast to deletion was considerably higher for English-speaking participants when learners
were trained with orthography. The overall epenthesis and deletion rates for the Englishspeaking group reported in Young-Scholten (2002) were as follows:
26 % and 30%,
respectively, when they were not exposed to orthography at all; 37% and 21%, respectively,
when only exposed to orthography at testing; 28% and 17%, respectively, when exposed to
orthography at learning but not at testing, and 33% and 12%, respectively, when they were
exposed to orthography at learning and testing. Japanese-speaking learners, in contrast,
exhibited a lower rate of deletion when they did not see the words. Young-Scholten (2002)
attributed the higher rate of epenthesis in comparison with the rate of deletion to the memoryenhancing effect of orthography both during learning and testing.
38
Evidence of input-based task effects on pronunciation was also found in Erdener and
Burnham (2005). As discussed earlier in Section 2.3.1, using a repetition task, these researchers
tested the role of exposure to audio-visual speech and orthographic information in the
production of native speakers of Turkish and Australian English. The experimental design
involved four conditions during training in terms of the presentation of the stimuli: auditory
only; auditory-visual; auditory-visual-orthographic; and auditory-orthographic. The participants
were tested with Irish English, an opaque language, and Spanish, a transparent language.
Although exposure to orthographic input did have a facilitative effect for both groups, albeit
more so for the Turkish learners whose L1 is more transparent than for Australian speakers
whose L1 is opaque, it also had a hindering effect. Specifically, Turkish speakers, whose L1 is
transparent, were affected by the effect of orthographic incongruency to a greater extent than
their Australian counter-parts, whose L1 is opaque. For example, whereas the Turkish speakers
did not confuse /x/ and /ʒ/ in the auditory only condition, they did so in the auditoryorthographic condition at a significantly high rate (45.83%). With respect to the Australian
speakers of English, their productions were not significantly affected by the incongruency
between the Spanish and English grapheme-to-phoneme correspondences. That is, they did not
make any errors in the production of [x] in the auditory only condition and their error rate
consisted of only 3.13 % in the auditory-orthographic condition. The production of [x] in
Turkish speakers was attributed to the presence of the grapheme <j> which corresponds to /ʒ/ in
Turkish but to /x/ in Spanish. However, there is no indication in the study as to whether the TL
sound actually exists in the Turkish learners’ L1 or not. The general effects of orthography on
non-native speech production in this study were proposed to be double. First, orthography may
interfere with audio-visual speech perception. Specifically, orthographic input may have
affected the perception of TL sounds because it is connected to speech via symbols. Second,
39
participants, especially the Turkish ones who seemed to rely on orthographic input to a greater
extent, may have focused on the orthographic input and ignored the auditory information.
The effect of orthographic inconsistency has also been examined in Hayes-Harb et al.
(2010). The researchers investigated the relationship between orthographic and phonological
representations in adult English speakers learning bi-syllabic English-like pseudo-words. Their
experiment consisted of both training and testing phases. The participants were assigned to three
different conditions at training (auditory only, congruent, and congruent/incongruent
orthographic conditions) but to one single condition at testing. The training task was a picturelearning task, where participants saw an image of a concept and heard the word associated with
it from which they had to learn the associations. The three conditions differed in terms of the
presence of orthography and the type of orthographic information available at training. In the
congruent condition, the pictures were accompanied with a written form of the word that
corresponded to the auditory form of the word (<fasha>-[faʃə]), and the incongruent group was
trained with three types of items: (i) congruent control items (<fasha>-[faʃə]); (ii) incongruent
items with a wrong letter (e.g., <faza>-[faʃə]); and (iii) incongruent items with an extra letter
(e.g., <kamand>-[kaməd]). The auditory-only group was trained without any exposure to
orthography. At testing, the participants were shown a picture and heard a word and were asked
whether the word was the correct word for the picture. Participants were shown pictures that
either matched or mismatched the auditory labels that they had been presented with during
training. Both the matched and the mismatched group consisted of congruent, incongruent-extra
letter, and incongruent-wrong letter stimuli. In the matched items, the auditory form that the
participants were presented which matched what they had been presented with auditorily at
training. For example, for the matched congruent stimuli, they heard /[aməg] for the word
‘thumb’ which at training they had also heard as [faməg] and saw as <famog>. For the
40
incongruent-extra letter items, they saw the word ‘envelope’, for example, and heard the word
[kaməd] which they had also heard as [kaməd] at training for the orthographic forms
<kamad>/<kamand>. For the incongruent-wrong letter context/condition, for the word ‘apple’,
they heard for example [faʃə], which they had also heard as [faʃa] at training and seen as
<fasha>/<faza>. The mismatched incongruent items also consisted of three types of stimuli:
congruent, incongruent-extra-letter, and incongruent-wrong-letter which did not correspond to
the auditory forms or pictures at training. For example, for the mismatched congruent items, the
participants would see the picture of a tiger and hear [faməg] which corresponded to ‘thumb’ at
learning. The mismatched incongruent extra-letter and the incongruent-wrong-letter items
consisted of the auditory label at testing corresponding to the incongruent orthographic forms,
instead of the auditory forms at training. For example, learners were presented with a picture of
an envelope, heard the word [kamənd] at testing which they had heard as [kaməd] and seen as
<kamad> and <kamand> at training, or they were presented with the word ‘apple’, heard the
word and heard /faza/ at testing which they had heard the auditory input [faʃa] and seen the
words, <faza> and <fasha>. The results suggested that there was a significant interaction
between condition and item type with participants being relatively lower in accuracy in the
incongruent/congruent orthography condition on incongruent items. This was interpreted as the
written forms impacting the learners’ memory of the phonological forms of new words. On the
other hand, while Hayes-Harb et al. (2010) did not find an interaction of group and item type for
incongruent-extra-letter items, there was an interaction of group and item type between
incongruent/congruent group and incongruent-wrong-letter items. This showed that, whereas
silent letters do not have a detrimental effect on learning new words, discrepancies between the
grapheme-to-phoneme correspondences in the L1 and those in new words will affect learning of
new words negatively. Given that English orthography is opaque and both silent letters and
41
phonemes with multiple written forms exist the authors proposed that the results with respect to
the silent letters were unusual.
The above studies show that a mismatch between the L1 and TL grapheme-to-phoneme
correspondences may lead to orthography-induced transfer (Young-Scholten, 2000; Erdener &
Burnham, 2005). In addition, other factors such as the quantity of exposure to orthographic
input over time (Young-Scholten, 2000) and the presence of orthographic input at learning
and/or testing may modulate the effect of orthography on L2 production (Young-Scholten et al.,
1999; Erdener & Burnham, 2005). While these studies show that exposure to orthographic input
can lead to non-target-like productions, there is a need for more studies that would further our
understanding of the factors that affect orthography-induced transfer. Such a study, focusing on
the role of auditory-orthographic condition, grapheme-to-phoneme inconsistency between the
L1 and TL and different aspects of PM in novice English-speaking learners of Spanish, will be
presented in Chapters 3 and 4.
In Section 2.2, I reviewed studies that have specifically addressed the interfering effect
of orthography in L1 language processing including category formation of L1 phonemes.
Similarly, in this section (2.3), I have reviewed a number of studies that have specifically
addressed the impact of orthography on the acquisition of L2 production. In the next section, I
will move on to discussing PM capacity in relation to L2 acquisition, in particular, the
acquisition of new vocabulary. Because learning TL grapheme-to-phoneme correspondences in
the context of vocabulary learning involves storing a string of sounds in the PM and also
comparing the TL sounds with the L1 sounds which correspond to the shared grapheme, it is
plausible that PM effects will be observed in this study in orthography-induced transfer.
42
2.4 PM capacity and L2 acquisition
Another question in the study to be presented in Chapters 3 and 4 is whether PM affects the
quantity of orthography-induced transfer. As previously mentioned, the acquisition of new
grapheme-to-phoneme correspondences embedded in new words, in a vocabulary learning task,
will require learners to store the TL sound among a string of other sounds (e.g., an entire word)
so that they can notice the difference between the L1 grapheme-to-phoneme correspondences
and TL grapheme-to-phoneme correspondences. Therefore, it is plausible that PM will impact
the effect of orthography-induced transfer. Whereas Sections 2.2 and 2.3 reviewed previous
studies that have focused on the effect of orthography on L1 perception and production and L2
acquisition, given that the effect of PM has not been previously investigated on orthographyinduced transfer, this section will provide a broader picture of how PM may affect L2
acquisition instead.
In this section, I will provide an account of the PM component in Baddeley and Hitch’s
(1974, 2000) model of working memory (2.4.1). Because this study also addresses primacy and
recency effects in orthography-induced transfer, I will review primacy and recency effects
reported in the memory literature (2.4.2). Section 2.4.3 will describe repetition effects in
vocabulary learning, another factor whose effect will be examined in relation to orthographyinduced transfer in the context of vocabulary learning in the present work. Finally, Section 2.4.4
will review the relationship between individual variation in PM and L2 acquisition, another
variable of interest in the experiment to be presented in Chapters 3 and 4.
43
2.4.1 PM
In this section, I will describe PM in Baddeley and Hitch’s (1974, 2000, 2003) highly influential
working memory model. In order to provide a complete view of the model, I will also briefly
describe the other components of the model, even though the effect of the other components is
not examined in this study. In Baddeley and Hitch’s revised (2003) model of working memory,
(see Figure 1 below), PM, also called the phonological loop, is one of the four components of
working memory. The three other components of this model are the central executive, the visuospatial sketchpad, and the episodic buffer. PM consists of two sub-systems, the phonological
store and the articulatory rehearsal system. The former temporarily stores verbal/acoustic
information for approximately two seconds whereas the latter is responsible for covert or overt
rehearsal of the materials, thereby prolonging the retention of information in the phonological
store.
Figure 1. The revised working memory model, (Baddeley 2003, p.196)
The phonological loop can be considered an on-line capacity for processing and analyzing new
information (Baddeley & Hitch, 1974; Baddeley, Gathercole & Papagno, 1998; Baddeley, 1999,
44
2003). Whereas PM is associated with storing auditory information, the visuo-sketchpad
temporarily restores visual and spatial information. The PM and sketchpad are also sometimes
referred to as slave systems because they are controlled by the central executive. The central
executive, in contrast, is responsible for processing operations such as dividing attention,
focusing, and switching. It is responsible for overseeing basic working memory subsystems and
long-term memory. The episodic buffer, a more recent addition to the model, creates integrated
episodes by combining information from the specialized subsidiary storage systems and longterm memory.
In this section, I have provided an overview of one of the most prominent working
memory models. In doing so, I have focused on defining PM. In the next section, I will describe
two of the well-known aspects of working memory, primacy and recency effects as well as
repetition effects. The effect of primacy and recency effects as well as repetition on
orthography-induced transfer will be tested in the present study.
2.4.2. Primacy and recency effects
There is ample evidence to suggest that vocabulary learning, non-word repetition, and serial list
recall (recalling items in the order in which they occurred), a well-known characteristic of PM,
are related abilities (e.g., Gathercole & Baddeley, 1993; Baddeley et al., 1998). Hence,
assuming that PM will play a role in orthography-induced L1-based phonological transfer, it is
also plausible that primacy and recency effects will also shape orthography-induced transfer.
Specifically, this effect should be observed both at the word and list levels. That is grapheme-tophoneme correspondences presented to learners word or list initially or word or list finally
should be better stored and recalled and exhibit a lower proportion of transfer. Currently, there
45
are no previous studies that have examined these effects with respect to orthography-induced
transfer. Therefore, in this section, I will provide a brief overview of some of the findings
concerning primacy and recency effects in recall tasks, for both word and non-word list recall as
well as for word and (non)-word recall.
A robust finding in memory studies is that, when participants are presented with a list of
single items, there is a higher probability of recall for initial and final list items (Deese &
Kaufman, 1957; Murdock, 1962; Waugh & Norman, 1965; Foreit, 1976). Studies have
repeatedly shown U-shaped or S-shaped serial position curves in which recall is initially
accurate, then decreases throughout the list, and finally improves toward the end of the task
(e.g., Deese & Kaufman., 1957; Murdock, 1962; Waugh & Norman, 1965). This is normally
referred to as primacy recency effect. The initial position recall advantage has been attributed to
possible elevated attention and recapitulation/rehearsal leading to a greater probability of
transfer of the information and consequent consolidation into long-term memory (Deese &
Kaufman, 1957; Atkinson & Shifrin, 1968; Craik, 1970; Rundus & Atkinson, 1970; Rundus,
1971). The recall advantage exhibited with final items, on the other hand, has been suggested to
be due to the end-of-list items still residing in a limited-capacity short-term buffer from which
they are reported without error (Atkinson, & Shiffrin, 1968). Given that the study to be
presented also tests the effect of primacy and recency effects on orthography-induced transfer, I
will provide a brief overview of some of the previous studies in the memory literature.
Murdock (1962), one of the pioneering studies on serial position effects, demonstrated
that, when participants were asked to recall list items with varying list lengths and presentation
rates, a primacy effect appeared to extend over the first three or four serial positions, a recency
effect spanned the last eight words, and a horizontal asymptote (a straight line) spanned position
46
5 up to the last eight serial positions. Figure 2 shows typical serial position curves. As shown,
primacy and recency effects are observed for lists varying between 10-40 items. In other words,
items that are presented at the beginning of a list and/or at the end of a list have a higher
probability of recall.
Figure 2. Typical serial position curves observed for different list lengths and presentation rates
in a free recall task (Murdock 1962, p.483)
Empirical evidence suggests that positional privileges are also apparent at retrieval and
recognition when single words (as opposed to word lists) are concerned. For example, Brown
and McNeill (1966) examined working memory effects in adult native English speakers. They
attempted to induce a tip-of-the tongue state (when information is available but not fully
accessible in memory) in English undergraduate students by presenting them with definitions of
infrequent words. For example, they presented participants with the definition ‘to leave the
throne’ for the word ‘abdicate’. The typical bow-shaped serial recall curve was present in the
recall of low frequency words by English speaking undergraduate students. Horowitz et
al.(1968) also found that, when words of 6-9 syllables were read out to English-speaking
participants, when showing the participants only fragments of words prior to eliciting recall, the
47
beginnings of words or the endings of words resulted in a correct response more readily and
with the shortest latency than with the middle of words. However, showing the participants the
beginnings of words elicited more correct responses and with shorter latencies than with word
endings.
Serial position effects have also been investigated in non-word repetition tasks and nonword list recall tasks, albeit to a lesser extent as pointed out by Gupta (2005). This researcher
examined 20 English-speaking undergraduate students and found significant primacy and
recency effects in non-word repetition tasks in both non-word repetition of words of 4 and 7
syllables in length. It was noted that serial position effects in non-words decreased in a similar
manner as serial position effects in lists of known words or digits decrease steadily from list
length 7–2. It was therefore speculated that the same underlying mechanisms that are operative
in serial recall at the list level are also operative at the word level.
In summary, all in all, there is evidence of primacy and recency effects both at the list
and (non)-word level. Given the prevalence of primacy recency effects in (serial) list recall,
(non-)word repetition, it is plausible that similar effects would be observed in orthographyinduced transfer, given that it is plausible that learners will have to rely on their PM to store the
TL sound and notice the differences between the TL and L1 grapheme-to-phoneme
correspondences. If PM is indeed involved in orthography-induced transfer, then it is plausible
that phenomena that characterize its working including primacy and recency effects will also
shape orthography-induced transfer. Although such a hypothesis seems plausible, there have
been no previous studies that have examined the effect of memory biases in relation to
orthography-induced transfer. The effect of these phenomena on orthography-induced transfer
will be tested both at the word and list levels in this study. In this section, I have discussed
48
primacy and recency effects. In the next section, I will discuss another working memory
phenomenon, namely repetition effects.
2.4.3 Effect of repetition on L2 vocabulary learning
Another basic phenomenon of PM is the facilitative effect of repetition of the to-be-remembered
item on its recall (e.g., Hebb, 1961; Atkinson & Shiffrin, 1968; Mathews & Tulving, 1973). For
example, Hebb (1961) presented participants with 24 lists of 9 digit items and measured the
impact of repetition of items on individual lists on immediate serial recall. The participants were
not informed, however, that the list on Trial 3 was repeated on every subsequent third trial
among a number of lists. The results showed that performance improved with repetition,
compared to that measured in non-repeating trials. In other words, as trials progressed, performance systematically improved on the repeating string relative to the novel ones.
There is also considerable evidence to suggest that repetition affects implicit vocabulary
learning in L2 acquisition (e.g., Saragai et al., 1978; Horst et al., 1998; Rott, 2000; Waring &
Takaki, 2003; Webb, 2007). Whereas previous research suggests that repetition is beneficial for
learning new lexical items, studies differ in terms of their proposals concerning the number of
encounters with new words necessary for acquisition to take place. Given that the study to be
presented in Chapters 3 and 4 also tests the effect of repetition on orthography-induced transfer
in a vocabulary-learning task, in this section, I will provide a brief account of some of the
prominent L2 vocabulary studies which have examined the effect of repetition on vocabulary
learning.
One of the first studies to investigate the impact of repetition on L2 vocabulary learning
through reading is Saragai et al. (1978). Twenty advanced English-speakers were asked to read
49
Anthony Burgess’s (1962) A Clockwork Orange. The novel was in English but contained words
of Russian slang which were unknown to the participants. The participants were required to read
the book at home and were tested on the vocabulary within a few days of their finishing reading.
The researchers did not originally disclose the real aim of the experiment to the participants.
Instead they told them that, upon finishing the book, they would have to do a comprehension
and literary criticism of the book. The researchers found a significant positive correlation
between word frequency in the novel and vocabulary learning. The size of the correlation
confirmed that repetition positively affected learning. Additionally, they suggested that other
factors such as meaningfulness of the context, and the degree of similarity between the TL and
L1 form of the words may also affect acquisition. Learners were then tested with a surprise
multiple-choice test which required them to recognize a correct definition for each word. The
results showed that words presented to learners fewer than six times were learnt by half of the
learners whereas words presented six or more times were learnt by 93% of participants. Based
on these findings, these authors proposed that, in general, ten encounters were required for the
acquisition of an unknown word, although different words may be acquired at different rates.
Horst et al. (1998) studied the effect of repetition on the acquisition of novel words in
English by 34 low-intermediate Arabic-speaking learners. Participants were required to read
Thomas Hardy’s (Jones 1979) The Mayor of Casterbridge, over six class-room sessions, over a
period of ten days. After a lapse of one week, the participants were tested with a 45-item
multiple-choice test which required recognizing a correct definition for each word as well as an
association test in which participants were required to make a semantic association between two
words by rejecting a third one; the same two tests had been administered as a pre-test about a
week before the reading sessions had started. There was a significant mean vocabulary gain
when pre- and post-reading multiple-choice test results were compared. Performance on the
50
word association test also improved significantly (16%). In addition, there was a significant
correlation between the number of times each word occurred in the book and relative learning
gains. The authors highlighted that, in general, the mean vocabulary gain value (5 words) was
low and attributed this to the fact that learners were of low proficiency. They speculated that
higher vocabulary gains would have been observed, had the learners been at an advanced level.
In addition, the size of gains for the effect of repetition on vocabulary learning varied
considerably from word to word. Horst et al. (1998) suggested that the differences in the size of
gains between the target words could have been due to the presence or absence of corresponding
pictures in the book as well as to the category of the target words; there were higher gains for
nouns in comparison with other lexical categories. The frequency analysis results suggested that
eight or more encounters with a given new lexical item were required for acquisition to take
place.
The effect of frequency of exposure on new vocabulary acquisition was also tested by
Rott (2000). Over a period of 13 weeks, she tested the effect of the difference between 2, 4, and
6 exposures to unknown nouns and verbs related to everyday life in 95 fourth-semester Englishspeaking learners of German at the University of Illinois. In weeks 1-4, tests were conducted to
ensure that the target words were unfamiliar to the participants. Vocabulary learning sessions
took place in weeks 4-13 with participants being divided into 3 exposure groups: (i) the 2exposure group; (ii) the 4-exposure group; and (iii) the 6-exposure group. Each group was
exposed to the same vocabulary set once a week. At each session/exposure, learners were
required to read 2 sets of passages, each of which contained 6 different target words. Learners
were tested immediately after, a week after, and a month after reading. To assess word
acquisition and retention, a recognition and a production task were administered. In the
recognition task, learners were provided with a list of words (12 target stimuli and 8 distracters)
51
as well as 4 definitions for each word and were requested to choose the correct definition. The
production task required learners to translate the list of words into their native language. The
tests were administered three times: (i) immediately after reading; (ii) after 1 week; and (iii)
after 4 weeks. Learners who had encountered unfamiliar words two, four or six times during
reading demonstrated significantly more word knowledge than students who had not
encountered the words during reading. The effect of reading for word learning had measurable
effects immediately after testing, after reading exposure, and 1 week later. In addition, all
learners retained a significant amount of passive vocabulary knowledge over a period of 4
weeks. However, half of the learners showed a significant decrease in productive word
knowledge over 4 weeks. Moreover, whereas the effect of two and four encounters was similar
for both the productive and receptive word knowledge, six encounters resulted in a significantly
larger vocabulary gain. Hence, it was concluded that six encounters are needed for considerable
lexical gains to occur and that vocabulary growth through reading has a stronger effect on
passive than active vocabulary knowledge.
In another study, Waring and Takaki (2003) examined the effect of repetition on the
acquisition of new words in 15 lower-intermediate Japanese-speaking learners of English
ranging from 19 to 21 years of age. The participants were asked to read a graded (easier) version
of Antonione de St. Exupery’s The Little Prince (1943). 25 English-like nonsense words were
substituted for the existing nouns in the text. For example, <week> was changed into <prink>.
The meanings of verbs and adverbs were deemed more difficult to guess from the context and so
only nouns and adjectives were chosen. In addition, words that differed in their frequency of
occurrence were tested: those which appeared once, four to five times, eight to ten times,
thirteen to fourteen, and fifteen to eighteen times in the entire book. Participants were tested
immediately after reading the book, a week to ten days later, and three months after reading the
52
book. A word form recognition test, multiple choice recognition test, and a translation test were
administered. The word recognition test required the participants to circle any words that they
recognized from the text. The multiple-choice recognition test was a four-choice test including a
single correct meaning and one distracter. The translation test presented the 25 words in a list
and the participants were asked to write the meanings in Japanese. Few new words were learnt
and learners were more successful at learning new words when encountered more frequently.
For example, words encountered more than eight times had approximately 50% chance of recall
three months after the test in the word recognition and multiple-choice translation tasks. Given
that the unprompted translation task was considered to be the most representative of learning, it
was speculated that it may take between 25-30 encounters to acquire new vocabulary.
The effect of repetition on vocabulary acquisition was also tested in Webb (2007) which
shows that orthographic knowledge of the word benefits most from repetition. He examined the
effects of 1, 3, 7, and 10 encounters with nonsense words with 98 intermediate Japanesespeaking learners of English. 23 first year students were assigned to the control group and did
not complete any of the learning tasks. The learners were assigned to four experimental groups
in which the first group encountered each target word once, the second group encountered each
target word three times, and the third group encountered each target word seven times. The task
involved reading a set of number of pages with each page presenting ten contexts and each
context containing a different target word. Thus, after reading one page, the participants had
seen only one new word. With regards to the stimuli, ten target words were chosen (six nouns,
four verbs). The participants were required to perform ten different tests including productive
knowledge of orthography, receptive knowledge of orthography, and receptive knowledge of
meaning. In the productive knowledge of orthography test, participants were prompted with the
auditory cue of a new word that they had heard and were asked to write the word in English.
53
The test for the receptive knowledge of orthography was a four-choice multiple-choice test that
provided the learners with a correct answer and three distracters. In this test, the learners were
required to circle the correct spelling of the word. The receptive knowledge of meaning and
form required learners to translate a list of words. Whereas intermediate learners were required
to take all the tests, the control group only had to take the productive knowledge of orthography
test. The size of correlations suggested that repetition does play a significant role in gaining
vocabulary knowledge. The results also showed that gains in all aspects of knowledge tended to
increase as the number of presentations increased and that, after 10 encounters, the gains were
significantly larger than after one encounter for all aspects of knowledge. Moreover, the results
revealed that after only one encounter, sizable gains were observed in both receptive and
productive knowledge of orthography (67% and 50% respectively). Although sizable gains were
also found for other aspects of vocabulary learning (e.g., 58% for meaning and form), the gains
were significantly higher for knowledge of orthography. Given these results, the authors
suggested that orthography is the first knowledge type to be acquired.
In sum, the general picture that emerges from these studies is that repetition positively
affects learning, even though there is no consensus as to how many encounters are required for
acquisition to take place (e.g., between 6 to more than 20 times). The wide range of frequency
of exposure necessary to acquire vocabulary via reading in a TL that was observed in the
various studies could be due to different factors including but not limited to differences in (i)
specific texts (ii) testing procedures; and (iii) differences in the stimuli. These differences could
have been due to different levels of difficulty both in the vocabulary learning tasks as well as the
tests that would explain the discrepancies in the results. For example, a simple word recognition
task is less demanding than a vocabulary test that taps into various aspects of word learning.
Along the same lines, reading a passage requires less focus than reading an entire book.
54
Moreover, differences in the type and number and type of stimuli might have resulted in the
different outcomes. All in all, differences in methodological approaches might have lead to
differences in the results in the studies reviewed here.
Another pattern that emerges is that the focus of these studies has been the effect of
repetition on the acquisition of meaning and less attention has been paid to the effect of
frequency of exposure on the active learning of word form. In particular, there appears to be a
gap with respect to the effect of frequency of exposure to new words on the acquisition of new
grapheme-to-phoneme correspondences. Webb (2007) appears to be the only study that has
examined spelling rules in general in the context of repetition of new vocabulary learning.
While it is clear that more exposure to new words in an L2 leads to a better encoding of these
words, it is not clear how and to what degree repetition of new words may impact the
acquisition of new grapheme-to-phoneme correspondences in the TL and subsequently lower
the proportion of transfer.
In this section I have reviewed some of the previous literature that suggests that
repetition positively affects the acquisition of new vocabulary in reading. It is plausible that
repetition will also affect orthography-induced transfer, for example, see Chapter 4. In the next
section, I will review some of the previous studies that have examined the effect of PM capacity
on learning various aspects of an L2, including vocabulary learning.
2.4.4 PM capacity and individual variation in L2 acquisition
Another question to be investigated here is whether there is a negative correlation between
individual PM capacity and orthography-induced transfer in the L2 acquisition of phonology.
Given the lack of previous literature on the relationship between PM capacity and orthography-
55
induced transfer, I will provide an overview of studies that have considered this relationship in
other domains of L2 learning. There is a considerable body of experimental research that has
examined the role of PM in L2 acquisition (e.g., Service, 1992; Service & Kohonen, 1995;
Cheung, 1996; Dufva & Voeten, 1999; Gathercole, Service, Hitch, Adams & Martin, 1999;
French, 2004; Mizera, 2006; O’Brien, Segalowitz, Collentine & Freed et al., 2006, 2007;
Hummel, 2009). A review of the studies in this section will show that apart from Mizera (2006)
who has not reported a correlation between PM and L2 learning, there is ample evidence that
suggests that there is a correlation between individual variation in PM and different aspects of
L2 learning (e.g., Service, 1992, Service & Kohonen, 1995; Cheung, 1996; Dufva & Voeten.,
1999; French, 2004; O’Brien et al., 2006 & 2007; Hummel, 2009). While the effect of PM on
orthography-induced transfer has not been investigated, as previously mentioned, given the
positive correlation between PM and different aspects of TL learning reported in some of the
studies, such a correlation is plausible and will be indeed tested in the present study. PM may be
a determining factor in modulating the rate of orthography-induced transfer given that, in order
for learners to notice the mismatches between their L1 and TL grapheme-to-phoneme
correspondences, they need to store the sounds corresponding to a shared grapheme in their L1
and TL, in order to compare them with the sound represented in their L1. That the ability to
store L2 sounds is impacted by learners’ PM capacity will become apparent in this section,
especially in studies that have demonstrated a link between L2 vocabulary learning and PM
capacity (e.g., Service, 1992; Service & Kohonen., 1995; Cheung, 1996; Dufva & Voeten, 1999;
French, 2004; Hummel, 2009)
A number of studies have examined the role of PM in L2 vocabulary and grammar
learning from a developmental perspective in both children and adolescents. Service (1992) was
one of the first studies that examined the relationship between PM and L2 proficiency. In a
56
longitudinal study, she tested Finnish children beginning to learn English in a school setting
(average age 9;4 years). PM was indexed using an English-based non-word repetition task as
well as a Finnish-based non-word English repetition task. English proficiency was measured in
terms of listening, reading comprehension, and production. PM was found to be a good
predictor of L2 proficiency in the first 2-3 years of schooling. Specifically, phonological scores
taken in the first testing session correlated with L2 proficiency in the final testing session
administered 2.5 years later. It was hypothesized that the correlation between PM and L2
proficiency may be due to the relationship between PM and vocabulary acquisition. This
hypothesis was tested in a follow up study (Service & Kohonen, 1995) that examined a subset of
the same group of children from the 1992 study. In this latter research, PM was again measured
using an English non-word repetition task. The L2 proficiency task, in addition to
comprehension and production tasks, included a vocabulary test. There was a correlation
between performance on the non-word repetition task and overall L2 proficiency, as well as a
strong correlation between PM and L2 vocabulary learning.
Whereas Service (1992) and Service and Kohonen (1995) established a link between PM
and L2 proficiency as measured by vocabulary learning, a number of studies have debated the
effect of PM on learning in low versus high proficiency learners. For example, Cheung (1996)
studied both high and low proficiency bilingual Chinese students whose average age was 12.5
years. PM measures were obtained using a non-word repetition task and a vocabulary test was
conducted measuring the total number of trials needed to learn the Mandarin translation of three
English words. Based on the results, PM was a predictor of L2 word learning ability, albeit in
the low proficiency group only. It was consequently hypothesized that high proficiency
individuals rely on long-term knowledge of English instead of PM. Given that PM affects
vocabulary learning in low proficiency learners, and because the learners in the present study
57
are also novice learners of Spanish as an L2, it is also plausible that PM capacity will affect the
quantity of orthography-induced transfer.
Another study that corroborates the impact of PM on learning at the beginning stages of
acquisition is Dufva and Voeten (1999). This longitudinal study investigated the effect of PM
and native language (NL) literacy in 160 child learners of English as a foreign language. These
researchers tested Finish school children from the first to the third grade. In their study, the level
of proficiency in English was measured in terms of active vocabulary, listening comprehension,
and communicative skills. The researchers measured PM using a pseudo-word repetition test
and literacy skills were measured in terms of word recognition and comprehension. The results
showed that both PM and NL literacy had positive effects on learning English as a foreign
language and explained much of the variance at the lowest level of English proficiency (grade
3).
That there is a correlation between low L2 proficiency and vocabulary learning has also
been demonstrated in French (2004). In a similar manner to Cheung (1996), French investigated
the association between PM, as measured by an Arabic-based non-word repetition task, and
vocabulary and grammatical knowledge acquisition in francophone children enrolled in a 5month intensive English program in Quebec. PM was tested at the beginning and end of the
program. As in Cheung (1996), PM was found to be the predictor of vocabulary gains but not
grammatical gains in participants with low rather than high proficiency levels, suggesting that
contributions from PM to L2 learning may become less important as familiarity with the TL
increases.
The relationship between PM and L2 learning has also been investigated in adult L2
learning from a developmental point of view. For example, in a longitudinal study, O’Brien et
58
al. (2006) examined the association between PM, operationalised in a serial English-based nonword repetition task, and the development of productive vocabulary (e.g., unique words,
neologisms such as ‘problemo’); narrative abilities (such as correct past participle production,
third person morphology); accuracy of elements of inflectional morphology; and the use of
complex grammatical structures (e.g., accurate use of prepositions and pronouns). These
researchers collected speech samples from 43 native English-speaking adults learning Spanish
as an L2 who had had at least two prior semesters of formal study of Spanish. The participants
were grouped into less and more proficient groups. PM was measured using a serial non-word
recognition task in which the participants judged whether the presentation order of two strings
of non-words was the same or different, both at the beginning and end of the semester. The
results showed that a higher PM score correlated with a larger repertoire of words at both test
times. However, PM did not correlate with vocabulary gain/use. The lack of correlation between
vocabulary acquisition and PM was partly attributed to the fact that a non-word recognition task
as opposed to a non-word repetition task had been used, suggesting that perhaps vocabulary
acquisition might be more closely related to the repetition abilities/articulatory component of
PM. On the other hand, PM was significantly correlated with the development of L2 narrative
skills for less proficient participants and with gains in correct use of function words for more
proficient participants. Overall, the results suggested that PM plays an important role in
narrative development at earlier stages of L2 learning and in the acquisition of grammatical
competence at later stages.
The role of PM in L2 adult learning has also been investigated by Hummel (2009) from
a developmental perspective with 77 advanced adult native-speakers of French who were
advanced learners of English. Hummel examined the relationship between PM, aptitude, and L2
proficiency as measured via vocabulary and grammatical knowledge as well as reading
59
comprehension as operationalised by traditional aptitude tests such as the Modern Language
Aptitude Test. L2 proficiency was tested in terms of vocabulary and grammatical knowledge
and reading comprehension as measured by the Michigan Test of English Language Proficiency
(MTELP). The participants were divided into two subsets: a less proficient advanced group and
a more proficient advanced group. There was a significant correlation between PM and
vocabulary learning, PM and L2 proficiency, and PM and grammatical knowledge. However,
there were no correlations between PM and reading comprehension. On the other hand, aptitude
test results were correlated with reading comprehension and grammatical knowledge. In
addition, while there was a correlation with the lower proficiency subset group of the advanced
learners and vocabulary knowledge, PM was not the predictor for the most advanced sub-set of
the advanced learners. The author stated that better PM leads to better processing abilities,
retention, and repetition of phonetic material which in turn leads to better processing of new
sound patterns in a second language.
While some studies point to PM mainly affecting the initial stages of acquisition, which
is the same level of proficiency tested in the present study, there are other studies that point to
the impact of PM with more proficient adults as well, albeit with respect to oral fluency and
perceptual categorization. For example, O’Brien et al. (2007) showed that PM is implicated in
developmental gains in L2 oral fluency. These researchers examined the relationship between
PM and L2 fluency gains in 43 novice and intermediate proficiency English-speaking adult
learners of Spanish at both the beginning and end of the semester. PM was operationalised as an
English-based serial non-word recognition task. Oral fluency included both general oral ability,
measured in terms of the total number of words spoken and the length in words of the longest
turn, and oral fluidity measures, measured in terms of rate of speech, mean length of speech runs
in words containing no silent pauses or hesitations greater than 400 ms, mean length of speech
60
runs in words containing no filled pauses, and longest speech run in words containing no silent
or filled pauses. PM played a role in the oral fluency gains of adults learning Spanish and the
PM scores predicted the learners’ oral fluency as measured by the amount of speech they
produced, the length of their longest turn, their speech rate, the amount of speech they produced
between filled pauses, and the length of their longest fluent runs at the end of the study in a
number of measures at the end of the semester. The study suggests that initial PM was highly
correlated with L2 oral fluency development.
PM has also been found to impact perceptual ability. Aliaga-García, Mora and CerviñoPovedano (2011) examined the effect of PM on the perception of vowels using high-variability
phonetic training. They tested bilingual Catalan-Spanish speakers studying English as foreign
language who had completed a semester of formal study of English Phonetics. PM capacity
measures were obtained using a Catalan-based serial non-word repetition task. Perceptual
accuracy was assessed at pre- and post-test through discrimination and identification tasks and
the perceptual scores. The differences between mean correct identification and discrimination
scores at pre-test and post-test showed that high PM capacity individuals outperformed low PM
capacity individuals in vowel categorization and had higher gains in perceptual ability after
training. The authors concluded that PM may have a role in learners’ use of cue weighting
which may lead to a better categorization of L2 sounds as well as an advantage in the
development of accurate long-term representations for L2 vowels.
While the above studies suggest that PM affects different domains of L2 learning, either
at the beginning stages of acquisition and/or later stages, there have also been studies that have
not found any evidence in support of the role of PM in L2 acquisition. For example, Mizera
(2006) studied 44 English-speaking university adult learners of Spanish and did not find an
61
effect of PM on oral proficiency. In his study, both low and high-proficiency learners were
tested. PM was measured in three different modes: (a) a verbal mode as in a speaking span test
in which the participants are presented with a list of letters and are required to recall them; (b) a
non-verbal mode as in a math span test which used numbers and arithmetic problems as
materials; and (c) a non-word repetition test. Oral fluency was analyzed in terms of speech rate,
inter-clausal pauses, and morpho-syntactic error rate of recorded speech samples. The results of
each of the PM tests were correlated with the fluency measures. However, none of the
correlations between PM and fluency measures yielded significant results. Therefore, it was
concluded that PM was not implicated in oral fluency.
Together, the above studies show that PM in L2 language learning has been examined in
relation to L2 vocabulary learning, the acquisition of the morphosyntax, oral fluency, and
perceptual categorization. The majority of the evidence points to a positive effect of PM on L2
acquisition (Service 1992; Service & Kohonen, 1995; Cheung, 1996; Dufva & Kohonen, 1999;
French, 2004; O’Brien et al 2006, 2007; Aliaga et al., 2011). In addition, there is considerable
evidence that points to the positive effect of PM on vocabulary learning in low proficiency
learners (Service 1992; Service & Kohonen, 1995; Cheung, 1996; Dufva & Voeten, 1999;
French, 2004). These studies are particularly relevant to the present work because they also
investigates low proficiency learners. The sole study that I am aware of that has not found any
correlation between PM capacity and L2 acquisition is Mizera (2006). A lack of a significant
correlation between PM capacity and L2 acquisition in Mizera (2006) might have been due to a
small sample size. Although a sample size of 44 participants is comparable to other studies that
have found a positive correlation between PM and L2 vocabulary learning, such as O’Brien et
al., (2006) which studied 43 learners, power analysis tests were not performed in Mizera (2006)
62
to show whether one could say with confidence that these results were reflective of the larger
population.
All in all, the majority of the evidence points to a positive correlation between PM
capacity and L2 learning, including vocabulary learning. However, there have been no previous
studies that have examined the effect of PM on orthography-induced transfer. Given the effect
of PM capacity on L2 vocabulary learning, in particular in low proficiency learners as well as
the fact that the learners in the experimental study to be presented in Chapters 3 and 4 were
novice learners and were required to perform a picture-naming task, it is plausible that PM
capacity will be correlated with the quantity of orthography-induced transfer.
2.5 Chapter summary
In this chapter, I have reviewed previous studies that have addressed the role of orthography on
the L2 acquisition of phonology as well as L1 perception and production. The review of
previous research on the influence of orthography on L2 phonological acquisition demonstrates
that orthography can affect both positively (e.g., Erdener & Burnham., 2005; Steele, 2005;
Escudero et al., 2008) and negatively (e.g., Young-Scholten, 2000; Young-Scholten, 2002;
Erdener & Burnham., 2005; Bassetti, 2007; Hayes-Harb et al., 2010) L2 production and
perception. With regards to the negative effect of orthography, some of the factors that have
been shown to lead to non-target-like productions in the presence of orthography are (a) the type
of auditory-orthographic input at the time of learning and production (Young-Scholten, 1999;
Erdener & Burnham, 2005); (b) inconsistency between the TL and L1 grapheme-to-phoneme
correspondences (Young-Scholten, 2000; Erdener & Burnham., 2005; Hayes-Harb et al., 2010);
(c) inconsistency in grapheme-to-phoneme correspondences in the TL (Bassetti, 2007); and (d)
63
the amount of input available to learners over the course of learning (Young-Scholten, 2000). A
review of these factors has shown that, although there is a growing body of literature on the
effect of orthography on L2 phonological acquisition not much is known about the factors that
shape orthography-induced transfer.
This chapter has also shown orthography effects in L1 perception. A review of previous
literature on the effect of orthography and literacy also suggested that (1) inconsistency between
orthography and phonology can affect phonological processing in adults (Seidenberg &
Tanenhaus, 1979; Taft & Hambly, 1985; Zeilger & Ferrand, 1998; Halle et al., 2000; Tyler &
Burnham, 2006; Ranbom & Connine, 2011); (2) onset of reading acquisition, specifically
learning grapheme-to-phoneme correspondences promotes language specific perception in
children (Burnham et al., 1991; Burnham, 2003) and (3) literacy affects both L1 perception
(Morais et al., 1979; Reid et al, 1986, Mazzaro, 2011) and production (Mazzaro, 2011). Given
these studies, it is plausible that orthography will also affect the population in the present study,
namely novice literate adult learners of Spanish.
This chapter also highlighted the fact that in spite of the current evidence on the effect of
orthography in L1 and L2 phonology, the role of orthography for the most part has been ignored
in the models of L2 phonological acquisition (e.g., Flege 1995; Brown, 1998, 2000; Best &
Tyler, 2007). In previous models, transfer has been formalized in light of phonetic (e.g., Flege,
1995, Best & Tyler, 2007) and phonological categories (e.g., Brown, 1998; 2000). The only
model that recognizes the potential influence of orthography on L1-based phonological transfer
is Best and Tyler’s (2007) PAM-L2. However, this model only hypothesizes an effect of
orthography on category assimilation in cases where the TL sound is different from the L1.
Therefore, there remains room for future models to consider the potential influence of
64
orthography on phonological transfer. The present study provides further empirical evidence for
the effect of orthography on phonological transfer and focuses on cases of grapheme-tophoneme inconsistencies between English and Spanish, where the TL is an existing sound in the
L1.
In addition to reviewing the role of orthography in the L2 acquisition of phonology, an
overview of the role of PM in L2 acquisition was provided. First, PM was described according
to a prominent model of working memory (Baddeley & Hitch, 1974, 2000) and then its primary
characteristics were reviewed. The two universal PM phenomena reviewed here were primacy
and recency effects (Deese & Kaufman, 1957; Murdock, 1962; Waugh & Norman, 1965; Brown
& McNeill, 1966; Horowitz et al., 1968; Craik, 1970; Rundus & Atkinson, 1970; Rundus, 1971
Foreit, 1976; Gupta 2005) and repetition effects (Hebb, 1961; Atkinson & Shiffrin, 1968;
Mathews & Tulving, 1973). Repetition effects were also reviewed in the context of L2
vocabulary learning (Saragai et al, 1978; Horst et al., 1998; Rott, 2000; Waring & Takaki, 2003;
Webb, 2007). Here it was seen that repetition positively affects the L2 acquisition of new words.
Following the review of the phenomena of PM working and the role of another aspect of PM,
namely PM capacity was examined in light of L2 acquisition. A review of previous studies that
have examined PM’s effect on the L2 acquisition of different domains showed that, in most
cases there is a positive impact on vocabulary learning with low proficiency learners (Service
1992; Service & Kohonen, 1995; Cheung, 1996; Dufva & Voeten, 1999; French, 2004).
Consequently, given that learning of new grapheme-to-phoneme correspondences most likely
requires the storing of the TL sound in order that learners may notice the mismatch between the
TL and L1 grapheme-to-phoneme correspondences, it was hypothesized that PM would also be
implicated in orthography-induced transfer leading to non-target-like productions. This
prediction, along with others, will be tested in the following chapters.
65
Chapter 3
Hypotheses & methodology
In the preceding chapter, I reviewed the literature related to the characterization of transfer in
models of L2 acquisition, the role of orthography in L1 perception and L2 acquisition as well as
the relationship between different aspects of PM and L2 learning. In this chapter, I will present
the hypotheses and the methodology of a new experimental study investigating the influence of
orthography in L2 phonological acquisition; the results of this study will be presented in Chapter
4.
The study in question involved novice English-speaking learners of Spanish who
completed two tasks designed to test the hypothesized effect of presence of orthography and the
following factors on L1-based phonological transfer: auditory-orthographic condition,
grapheme-to-phoneme inconsistency and PM aspects.. The main task, a Spanish picture-naming
task, and the secondary task, a Farsi based non-word repetition PM task, had the following
goals. The picture-naming task aimed to determine the effect of the following factors on the
proportion of orthography-induced transfer in production: (1) exposure to auditory-orthographic
input at the time of learning and at production (discussed in Section 2.3.2 in Chapter 2); (2)
inconsistency between English and Spanish grapheme-to-phoneme correspondences (discussed
in Section 2.3.2 in Chapter 2); (3) primacy and recency/ positional effects at the list level and
primacy effects at the word level (discussed in Section 2.4.2 in Chapter 2), and (4) repetition
effects/round (discussed in Section 2.4.3 in Chapter 2). The PM task, on the other hand, sought
to determine individual participants’ PM capacities so that individual scores could be correlated
with the mean proportion of orthography-induced transfer in order to investigate the potential
effect of PM on orthography-based transfer for the first time.
66
In the remainder of this chapter, I first provide my hypotheses (Section 3.1). I then turn
to the various methodological aspects of the experiment’s design including the participants
(Section 3.2), and details of the picture-naming task (Section 3.3) and the PM task (Section 3.4),
stating the motivation behind the particular experimental designs and describing both the stimuli
and procedures involved. Finally, the testing protocol in this experiment is explained in Section
3.5. I now turn to the study’s hypotheses.
3.1 Hypotheses
In Chapter 2, Section 2.3, the role of orthography in shaping L2 phonological acquisition was
examined. In addition, PM and some of its principal aspects, namely primacy and recency
effects and repetition effects, were reviewed and the potential effects of PM capacity on L2
acquisition, in particular vocabulary learning, were discussed. I will now present a series of
hypotheses motivated by the findings of the studies reviewed in Chapter 2.
1. General effect of orthography on transfer:
Based on Young-Scholten et al. (1999), Young-Scholten (2000, 2002) and Erdener and
Burnham (2005), exposure to orthography at learning and/or production will promote to L1based transfer leading to non-target-like productions.
2. Effect of different factors on orthography-induced transfer:
(a) Based on Young-Scholten et al. (1999) Erdener and Burnham (2005) the effect of each of the
auditory-orthographic conditions on transfer will differ. Specifically, the following hierarchy
will be observed where increased L1 influence is predicted moving from left to right:
67
orthography at learning & production > orthography at learning > orthography at production >
auditory input only (i.e., no orthography at learning nor production)
The prediction that the orthography at learning and production condition will exert a
greater influence on phonological transfer in comparison with when learners are presented with
orthography only at learning and orthography only at production conditions is motivated by the
presumption that presenting learners with orthographic input twice as opposed to once, namely
both at learning and at production, will negatively affect transfer. As to why orthography at
learning only condition is predicted to induce a higher proportion of transfer than orthography at
production condition only, it is argued that in the former condition learners are provided with a
better opportunity to form target-like representations because they are only presented with
auditory input at learning.
(b) Effect of inconsistency between TL and L1 grapheme-to-phoneme correspondences on
transfer: Given the findings in Young-Scholten (2000), when a shared grapheme corresponds to
two different phonemes in the TL and L1 (e.g., <ll> corresponds to /j/ as in <pollero>-[pojeɾo]
in Spanish but to /l/ as in <balloon> [bəlun] in English ), even when the TL phone exists in the
L1 inventory (e.g., [j] exists in English such as in <yes>-[jɛs]), exposure to orthographic input
will lead to learners substituting their L1 phoneme for the TL phoneme (e.g., [poleɾo] for
[pojeɾo]).
In contrast, grapheme-to-phoneme correspondences that are shared by English and
Spanish are not predicted to result in non-target-like productions. For example, because <m>
corresponds to /m/ in both Spanish and English (e.g., <macaca> [makaka] and <madam>
[mædəm], respectively), learners will produce <m> correctly as [m] (see Table 3.4 for a list of
grapheme-to-phoneme correspondences tested in the present study).
68
(c) Influence of different aspects of PM on orthography-based transfer: PM will affect the
proportion of orthography-induced transfer for two reasons. First, because when learning new
grapheme-to-phoneme correspondences, learners presumably rely on PM to store the TL sound
that corresponded to the shared grapheme in order to notice the discrepancy between the TL and
L1 grapheme-to-phoneme correspondences. Second, in this study, the effect of orthographyinduced transfer will be tested in the context of L2 vocabulary learning and there is much
evidence that PM is implicated in L2 vocabulary learning which requires the learning of new
strings of sounds (e.g., Service 1992; Service & Kohonen, 1995; Cheung, 1996; Dufva &
Voeten, 1999; French, 2004). Therefore:
(i)
Given the well documented existence of primacy and recency effects (e.g., Deese &
Kaufman, 1957; Murdock, 1962; Waugh & Norman, 1965; Craik, 1970; Rundus & Atkinson,
1970; Rundus, 1971; Foreit, 1976; Gathercole & Baddeley, 1993; Gupta, 2005), we predict that
primacy and recency effects will decrease the proportion of orthography-induced transfer. In
particular, it is predicted that the position of a given word in the learning and/or production
sequence will influence the proportion of orthography-induced transfer leading to non-targetlike productions at the list level/within the triplet. In this study, primacy and recency effects
were controlled for by creating permutations in which the target stimuli would appear in
different places in lists/triplets presented to learners during learning and production in the
picture-naming task. Specifically, each list contained three items and the target stimuli were
tested in the following positions: word learned first and tested first (1*1), word learned first and
tested last (1*3), word learned in the middle of the list and tested in the middle (2*2), word
learned last and tested first (3*1) and words learned last and tested first (3*3). Therefore, I
69
predict that the following hierarchy will be observed wherein the proportion of orthographyinduced transfer will increase from left to right in the orthographic conditions (orthography at
learning & production, orthography at learning and orthography at production) in the following
hierarchy:
(1*1), (3*1) > (2*2), (1*3), (3*3)
(ii) Given the well established primacy and recency effects at the word level (e.g., Brown &
McNeill., 1966; Horowitz et al., 1968; Gupta 2005), there will be a primacy effect at the word
level in the orthographic conditions (ortho-learning & production, ortho-learning and orthoproduction). That is, orthography will exert less influence on transfer for grapheme-to-phoneme
correspondences occurring word initially than medially. For example, with the grapheme-tophoneme correspondence <ll>-/j/ , a lower proportion of orthography-induced transfer will be
observed with the word <llanura>, where the grapheme-to-phoneme correspondence is wordinitial, in comparison with the word <pollero> in which the grapheme-phoneme correspondence
is word-medial.
(iii) Word repetition effects demonstrated in previous research (e.g., Saragai et al., 1978; Horst
et al., 1998; Rott, 2000; Waring & Takaki, 2003; Webb, 2007) will affect the proportion of
orthography-induced transfer in the orthographic conditions (orthography at learning &
production, orthography at learning and orthography at production). That is, the proportion of
orthography-induced transfer will decrease as the number of repetitions of the picture-naming
task increases. In this study, learners repeated the picture-naming task three times (3 rounds).
Therefore, in the following hierarchy, decreased L1 influence is predicted moving from left to
right:
70
Round 1 > Round 2 > Round 3
(iv) Individual differences in PM will be negatively correlated with the proportion of
orthography-induced transfer in the orthographic conditions (orthography at learning &
production, orthography at learning only and orthography at production only). I now turn to the
description of the participants in this study.
3.2 Participants
45 adult native speakers of Canadian English participated in this study. I conducted this study in
Toronto, a multicultural city, where French is taught in school and it is common for people to
speak a number of languages. Having these facts in mind, exposure to languages other than
English was strictly controlled for. Minimal knowledge of French was accepted since all
participants were being recruited in Ontario where French is mandatory in the education system.
For the same reasons for controlling for the effect of interference of knowledge of other
languages, only those learners whose parents’ native language was English and who had been
raised in households where only English was spoken were recruited. The criterion of language
background was controlled for as strictly as possible, as it has been proposed that knowledge of
a second or third language can interfere with learning a new language (e.g., De Angelis, 2007).
For example, it has been noted that literacy in a second language may result in a higher degree
of meta-linguistic awareness which may in turn result in a faster acquisition of the written code
of a new language (e.g., Ceñoz & Genesee, 1998; Ibrahim, Eviatar & Ahron-Peretz, 2002;
Ceñoz & Hoffmann, 2003; Errasti, 2003; Lesaux & Geva, 2006). Although the learners in this
study were adults, I was concerned about the possibility that knowledge of other languages
would interfere or help with the development of decoding skills for the Spanish written system.
71
Given the stringent language background requirements, the multicultural nature of Toronto, and
time constraints, gender was not controlled for, in this study. With regards to age, all
participants were over eighteen and their mean age was 21.8. Moreover, since the goal of this
study was to test the effect of orthography on phonological transfer in pronunciation, it was also
essential for participants to be literate. Hence, a minimum of high school education was
required. In addition it was essential for participants not to have any speech or cognitive
impairments. Participants were recruited via advertisement on a University of Toronto website
as well as through friends and acquaintances.
As part of the experiment, a detailed background questionnaire was administered (see
Appendix A). The purpose of the questionnaire was to gather the relevant information from the
participants concerning the above mentioned criteria. A question regarding playing an
instrument as well as some questions with respect to variables associated with PM such as (e.g.,
self-estimated vocabulary size, speaking speed and reading speed) were also included in the
questionnaire as a starting point for future studies. The background information collected
through the questionnaire revealed that 5 participants had been exposed to Spanish or other
languages through friends, media and/or travelling which did not conform to the requisite
criteria for participant selection. Therefore, their data was discarded and only the data for the
remaining 40 participants were analyzed. I now turn to task 1, the Spanish picture-naming task.
3.3 Task 1: Spanish picture-naming task
In order to test the effect of exposure to different combinations of auditory-orthographic input at
the time of learning and/or production on phonological transfer during L2 acquisition, it was
72
necessary to elicit production. To this end, the picture-naming task from Steele (2002) was
adapted. In the following sections, I describe the nature of the task and the stimuli used.
3.3.1 Task design
As mentioned previously, the participants were required to perform a picture-naming
task. This task was designed to test hypotheses regarding the effect of auditory-orthographic
condition,
inconsistency
between
the
Spanish
and
English
grapheme-to-phoneme
correspondences, primacy and recency effects as well as repetition effects. The task design in
the picture-naming task was as follows. In one session, via a PowerPoint presentation,
participants were asked to learn the Spanish words with which they were presented and then
name them. During learning, Spanish words were presented to them in triplets consisting of two
target stimuli and a distracter. Each word was auditorily presented three times in a row
accompanied by an image illustrating its meaning. The image – with or without the orthography,
depending on the auditory-orthographic condition – remained on the screen for the duration of
the presentation of each word, about 3 to 7 seconds. This process was then repeated for the other
two words in the group. Testing the participants’ production followed immediately on the
learning phase for each triplet. In this second phase, participants once again saw the images
corresponding to each word – once again with or without orthography depending on the
condition – and had to name them in Spanish. At testing, each image remained on the screen for
3 seconds and was followed by the presentation of the next image. The order of presentation of
these images was not randomly assigned. Instead, a particular order of presentation, described in
Section 3.2.2, was chosen to test for primacy and recency effects.
73
As stated before, one of the purposes of this study was to examine the effect of auditoryorthographic condition at the time of learning and production on the proportion of orthographyinduced transfer. In order to do so, as illustrated in Table 3.1, 4 auditory-orthographic conditions
were created for the word-learning task and ten participants were assigned to each auditoryorthographic condition. These conditions differed in the combination of exposure to auditory
and orthographic input during learning and production.
Table 3.1
Modality of Presentation of Input at Learning and Production (Auditory-orthographic
Condition)
Type of input
Auditory
Condition
Ortho-learning & production
Ortho-learning
Ortho-production
Auditory only
Orthographic
Learning
Production
Learning
Production
Yes
Yes
Yes
Yes
No
No
No
No
Yes
Yes
No
No
Yes
No
Yes
No
In all of the conditions, learners were exposed to auditory input at the time of learning
but not at production. The four conditions differed in terms of the presence/absence of
orthographic input at the time of learning and production (2 x 2 = 4 permutations). In the ortholearning & production condition, participants were presented with orthographic input at both
learning and production. In contrast, the ortho-learning and ortho-production conditions exposed
the participants to orthographic input during learning or production only respectively. Finally,
the auditory only condition involved only auditory and no orthographic input.
In order to see whether the factor ‘round’ (i.e., repetition) had an effect on orthographyinduced transfer, participants were asked to do the entire picture-naming task (108 words) 3
74
times. Participants were told that they could have a 2-3 minute or longer breaks (if needed) in
between each round. I now turn to the particulars of the word stimuli.
3.3.2 Stimuli
72 Spanish words were selected and paired with black and white images downloaded from the
Center for Research in Language’s website (1). High frequency Spanish words (e.g., <hola>[ola] ‘hello’) were avoided in order to ensure that there were no words that participants might
have come across previously. As illustrated in Table 1 (Appendix B), the target stimuli and the
distracters were assigned meanings that were different from their real meanings given that the
stimuli’s true meanings were difficult to depict (see Appendix C) These randomly assigned
meanings, illustrated in Table 1 (Appendix B), spanned a number of semantic fields such as
names of household items and animals as well as vegetables, as they were relatively frequent
and familiar. Frequent and familiar meanings were chosen for two reasons. First, this would
facilitate the visual presentation of the stimuli. Second, it would facilitate learning with novice
learners.
In order to test the hypothesis regarding the effect of grapheme-to-phoneme
inconsistency between Spanish and English, the stimuli included examples of 2 types of Spanish
grapheme-to-phoneme correspondences illustrated in Table 3.2, namely those that are the same
in Spanish and English including <m>-/m/, <n>-/n/, <b>-/b/, <d>-/d/ and <h>-/Ø/ VCV
(intervocalically) and those that are different (<v>-/b/, <d>-/ð/, <z>-/s/, <h>-/Ø/ # (word
initially) and <ll>-/j/); <v> corresponds to /b/ in Spanish but to /v/ in English (e.g., <vireca>
[biɾeka], and <vote>-[vot], respectively), <d>- corresponds to /ð/ in Spanish but to /d/ in English
(e.g., <darico> [daɾiko] and <madam>-[mædəm], respectively), <z> corresponds to /s/ in
75
Spanish but <z>-/z/ in English (e.g., <zatara>-[sataɾa] and <zoo>-[zu], respectively), <h> is
silent word initially in Spanish but realized as /h/ in English in this position (e.g., <hanega>[aneɣa] and <hand>-[hænd], respectively) and <ll> corresponds to [j] in some varieties of
Spanish but to [l] in English (e.g., <pollero>-[pojeɾo], <balloon>-[bəlun], respectively). Only
grapheme-to-phoneme correspondences that were believed not to contain any sounds that do not
exist in English were included in the study in order to minimize the possibility of misperception
and/or difficulty in production.
In order to test for primacy effects within the word, the stimuli were balanced for
position in the word, when a grapheme-to-phoneme correspondence remained the same across
position (e.g., word-initially and word-medially/intervocalically) in English and in Spanish. The
only ‘different’ grapheme-to-phoneme correspondence for which this criterion was met was
<ll>-/j/ because <ll> corresponds to /l/ both word initially and word-medially in English (e.g.,
<Lloyd>-[lojd] vs. <balloon>-[bəlun]) and to /j/ both word initially and word-medially in
Spanish (e.g.,<lloreta>-[joɾeta] ‘crying fit’ vs. <pollero>-[pojeɾo] ‘one who keeps fowls’). The
other grapheme-to-phoneme correspondences in this study could not satisfy this criterion. For
example, whereas the grapheme <s> corresponds to /s/ word initially in English (e.g., <Sue>[su]), intervocalically, it may correspond to either /s/ or /z/ (e.g., <roses>-[ɹozəz], <mason>[mesən]).
76
Table 3.2
Picture-naming Task: Target Stimuli
Types of graphemeto-phoneme
correspondences in
relation to English
Same
<m>-/m/
<n>- /n/
<b>-/b/
<d>-/d/
Position of grapheme-to-phoneme correspondence in the word
Initial
<macaca> [makaka], <metopa>
[metopa] , <macana> [makana]
<nerita> [neɾita],<namoro>[namoɾo],
<nacrita> [nakɾita]
<botina> [botina], <bacana> [bakana],
<boruca> [boɾuka], <bofena> [bofena],
<batata> [batata], <bimana> [bimana]
<darico> [daɾiko], <derogo> [deɾoɣo]
<dimana> [dimana], <degano>
[deɣano], <dagame> [daɣame],
<detenga> [deteŋga]
<h>-/Ø/ VCV
<s>-/s/
Different
<v>-/b/
<d>-/ð/
<z>-/s/
<h>-/Ø/ #
<ll>-/j/
Word-medially
<omino> [omino], <tomen̪to>
[tomento], <amago> [amaɣo]
<anata> [anata], <anafe> [anafe],
<anego> [aneɣo]
<ahumar> [aumaɾ], <aherir> [aeɾiɾ],
<rehogar> [reoɣaɾ], <ahitar> [aitaɾ],
<ahotar> [aotaɾ], <ahincar> [aiŋkaɾ]
<somato> [somato], <socapa>
[sokapa], <sotera> [soteɾa], <sicono>
[sikono], <sigogo> [siɣoɣo], <sarama>
[saɾama]
<vireca> [biɾeka], <veneno> [beneno],
<verato> [berato], <vagante>
[baɣante], <vigota> [biɣota],
<vegana> [beɣana]<zatara> [sataɾa],
<zatico> [satiko], <zapito> [sapito],
<zafero> [safeɾo] , <zanate> [sanate],
<zarina> [saɾina]
<harapo> [aɾapo], <harina> [aɾina],
<hanega> [aneɣa], <horita> [oɾita]
<horaco> [oɾako], <hontana> [ontana]
<llanero> [janeɾo], <llanito> [janito]
<llamingo> [jamiŋgo], <lloreta>
[joɾeta], <llanura> [januɾa], <llorona>
[joɾona]
<codena> [koðena], <pidona>
[piðona], <adentro> [aðen̪tɾo],
<adono> [aðono], <adormo>
[aðoɾmo], <tudanco> [tuðaŋko]
<mallugo> [majuγo], <malllera>
[majeɾa], <malleto> [majeto],
<pollero>[pojeɾo], <pallete>,
[pajete],<collete> [kojete]
77
In order to test for primacy effects within the word, the stimuli were balanced for
position in the word, when the corresponding sounds for the shared graphemes remained the
same across position in each language. The only ‘different’ grapheme-to-phoneme
correspondence for which this criterion was met was <ll>-/j/ because <ll> corresponds to /l/
both word initially and word medially in English (e.g., <Lloyd>-[lojd] vs. <balloon>-[bəlun])
and to /j/ both word initially and word medially in Spanish (e.g.,<lloreta>-[joɾeta] “crying fit”
vs.
<pollero>-[pojeɾo]
“one
who
keeps
fowls”).
The
other
grapheme-to-phoneme
correspondences in this study could not satisfy this criterion. For example, whereas the
grapheme <s> corresponds to /s/ word initially in English (e.g., <Sue>-[su]), intervocalically, it
may correspond to either /s/ or /z/ (e.g., <roses> [ɹozəz], <mason> [mesən]). The stimuli were
also controlled for the number of words assigned to each position. 6 words were assigned to
each position with the exception of <m> and <n> for which only 3 words were assigned to each
cell since it was believed that – due to positive transfer – these grapheme-to-phoneme
correspondences would not be problematic and positional effects would not play a role. Having
a smaller number of these stimuli type allowed for a shorter overall task.
For the sake of consistency, the stimuli were also controlled for the number of syllables,
the position of stress with respect to the target grapheme/phoneme, and words with morphemes
of Latin origin. All stimuli were trisyllabic words with primary stress on the penult (e.g.,
<nerita> [neˈɾita] ‘a type of mollusk’). The only exception where stress patterns differed was in
words with /h/ given that its realization in English is conditioned by stress: <h> is pronounced
when stressed word initially such as in <hero>, <her>, <hat> but is silent in unstressed positions
inter-vocalically, such as in <vehicle> [ˈvi;jəkəl]. Hence, in an effort to test the effect of
‘sameness’ on words with <h> in intervocalic position, only those words that were stressed on
their final syllables where <h> would be in an unstressed position, such as <ahumar> [aumˈaɾ]
78
‘to smoke’ were chosen. The complete stimuli list was vetted by two native speakers of English
for words with morphemes of Latin origin, given the possibility that these types of words could
have led to a higher proportion of transfer. For example, Spanish words ending in <-ción>,
which is the equivalent of English <-tion>, were not included.
The distracters consisted of the 36 Spanish words in Table 3.3. For the same reasons
mentioned above, these words were also assigned new meanings (see Appendix C for real
meanings and Table 1 in Appendix B for assigned meanings). These words differed from the
target stimuli in the number of syllables (monosyllables: e.g., <a> “the letter a”; bisyllables:
e.g., <toco>-[took] ‘I touch’) and the complexity of their syllable structure (e.g., presence of
clusters <gofre> [gofɾe] ‘a kind of cake’). In addition, new grapheme-to-phoneme
correspondences such as <rr> representing /r/, and <gue> pronounced [ge] were included.
Table 3.3
Picture-naming Task: Distracters
Bisyllabic
<chorro> [ʧoro], <agá> [aɣa], <croe> [kɾoe], <guiri> [giɾi], <troque> [tɾoke],
<trinche> [tɾiʧe], <grúa> [gɾua], <gofre> [gofɾe], <trucha> [tɾuʧa], <aquí> [aki],
<trina> [tɾina], <fea> [fea], <toco> [toko], <pecho> [peʧo], <oca> [oka], <toga>
[toɣa], <acá> [aka], <efe> [efe], <poa> [poa], <tía> [tia]
Monosyllabic
<a> [a], <te> [te], <e> [e], <tú> [tu], <o> [o], <ti> [ti], <fui> [fui], <to> [to], <pe>
[pe], <cha> [ʧa], <ta> [ta], <cu> [ku], <fo> [fo],<ca> [ka], <che> [ʧe], <u> [u]
The stimuli were presented as follows. 36 triplets, each containing two target stimuli and
one distracter, were formed from the 108 words (72 stimuli plus 36 distracters; see Table 3.4).
In an effort to keep the level of difficulty of each pair of grapheme-to-phoneme correspondences
in each triplet as consistent as possible, as shown in Table 3.4, each group consisted of one
grapheme-to-phoneme correspondence that is identical in Spanish and English (e.g., <m>-/m/)
and one that is different (e.g., Spanish <v>-/b/ whose English counterpart is <v>-/v/), except for
the triplets 31-36 which included two grapheme-to-phoneme correspondences from the
79
‘different’ category (e.g., <ll>-/j/ and <z>-/s/). The stimuli were also pseudo-randomized so that
primacy and recency effects with respect to position in triplets could be controlled for. The
positional permutations were as follows: first in learning and first in production (1*1), first in
learning and last in production (1*3), last in learning and first in production (3*1), second in
learning and second in production (2*2), and last in learning and last in production (3*3). This
order remained the same for all three rounds.
Table 3.4
Picture-naming task: Positional Composition of Grapheme-to-sound Correspondences at
Learning and Production per triplet (continued)
Triplet
Position of grapheme-to sound correspondences
Learning
Production
1
2
3
1
2
3
1
<m>-/m/
Distracter
<v>-/b/
<v>-/b/
Distracter <m>-/m/
2
<m>-/m/
<v>-/b/
Distracter
<m>-/m/
<v>-/b/
Distracter
3
<v>-/b/
Distracter
<m>-/m/
<m>-/m/
Distracter <v>-/b/
4
<v>-/b/
<m>-/m/
Distracter
<v>-/b/
<m>-/m/
<m>-/m/
5
Distracter <v>-/b/
<m>-/m/
Distracter <v>-/b/
<m>-/m/
6
Distracter <m>-/m/
<v>-/b/
Distracter <m>-/m/
<v>-/b/
7
<n>-/n/
Distracter
<d>-/ð/
<d>-/ð/
Distracter <n>-/n/
8
<n>-/n/
<d>-/ð/
Distracter
<n>-/n/
<d>-/ð/
Distracter
9
<d>-/ð/
Distracter
<n>-/n/
<n>-/n/
Distracter <d>-/ð/
10
<d>-/ð/
<n>-/n/
Distracter
<d>-/ð/
<n>-/n/
Distracter
11
Distracter <d>-/ð/
<n>-/n/
Distracter <d>-/ð/
<n>-/n/
12
Distracter <n>-/n/
<d>-/ð/
Distracter <n>-/n/
<d>-/ð/
13
<d>-/d/
Distracter
<h>-/Ø/ #
<h>-/Ø/ # Distracter <d>-/d/
14
<d>-/d/
<h>-/Ø/ #
Distracter
<d>-/d/
<h>-/Ø/ # Distracter
15
<h>-/Ø/ # Distracter
<d>-/d/
<d>-/d/
Distracter <h>-/Ø/ #
16
<h>-/Ø/ # <d>-/d/
Distracter
<h>-/Ø/ # <d>-/d/
Distracter
17
Distracter <h>-/Ø/ #
<d>-/d/
Distracter <h>-/Ø/ # <d>-/d/
18
Distracter <d>-/d/
<h>-/Ø/ #
Distracter <d>-/d/
<h>-/Ø/ #
19
<b>-/b/
Distracter
<h>-/Ø/ V
<h>-/Ø/ V Distracter <b>-/b/
20
<b>-/b/
<h>-/Ø/ V
Distracter
<b>-/b/
<h>-/Ø/ V Distracter
21
<h>-/Ø/ V Distracter
<b>-/b/
<b>-/b/
Distracter <h>-/Ø/V
22
<h>-/Ø/ V <b>-/b/
Distracter
<h>-/Ø/ V <b>-/b/
Distracter
23
Distracter <h>-/Ø/
<b>-/b/
Distracter <h>-/Ø/ V <b>-/b/
24
Distracter <b>-/b/
<h>-/Ø/ V
Distracter <b>-/b/
<h>-/Ø/V
(continued)
80
Table 3.4
Picture-naming task: Positional Composition of Grapheme-to-sound Correspondences at
Learning and Production per triplet (continued)
Triplet
Position of grapheme-to sound correspondences
Learning
Production
1
2
3
1
2
3
25
<s>-/s/
Distracter
<ll>-/j/
<ll>-/j/
Distracter <s>-/s/
26
<s>-/s/
<ll>-/j/
Distracter
<s>-/s/
<ll>-/j/
Distracter
27
<ll>-/j/
Distracter
<s>-/s/
<s>-/s/
Distracter <ll>-/j/
28
<ll>-/j/
<s>-/s/
Distracter
<ll>-/j/
<s>-/s/
Distracter
29
Distracter <ll>-/j/
<s>-/s/
Distracter <ll>-/j/
<s>-/s/
30
Distracter <s>-/s/
<ll>-/j/
Distracter <s>-/s/
<ll>-/j/
31
<z>-/s/-/s/ Distracter
<ll>-/j/
<ll>-/j/
Distracter <z>-/s/
32
<z>-/s/
<ll>-/j/
Distracter
<z>-/s/
<ll>-/j/
Distracter
33
<ll>-/j/
Distracter
<z>-/s/
<z>-/s/
Distracter <ll>-/j/
34
<ll>-/j/
<z>-/s/
Distracter
<ll>-/j/
<z>-/s/
Distracter
35
Distracter <ll>-/j/
<z>-/s/
Distracter <ll>-/j/
<z>-/s/
36
Distracter <z>-/s/
<ll>-/j/
Distracter <z>-/s/
<ll>-/j/
Given that the particular phonetic realizations that I was looking for (e.g., <ll>-/j/ and
<z>-/s/) are not present in all varieties of Spanish and that Mexican Spanish has these surface
realizations, a 36 year old female Mexican (Chihuahua) speaker of Spanish was recorded.
Another reason for choosing this particular speaker for the recording of the stimuli was that she
had professional voice training. She had knowledge of the goals of the experiment and was
asked to read the stimuli naturally, pausing for a count of three before between the production of
each word. With all relevant characteristics of the word learning task now reviewed, I turn to the
second task that tested the participants’ PM.
3.4 Task 2: PM task
This task aimed to measure learners’ PM in order to test the possibility of a correlation with
individual differences in the word learning task. Specifically, this task served to verify the
hypothesis that individual PM scores will correlate with the proportion of orthography-induced
81
transfer in the word learning task based on the assumption that learners with a superior PM
would be able to better store the auditory input and notice the mismatch between the two
conflicting sources of input (auditory versus orthographic), as was stipulated in 1 (c) (iv).
In general, two kinds of PM tasks have been used in L2 vocabulary acquisition studies,
non-word repetition tasks and/or serial non-word recognition tasks. In non-word repetition tasks,
participants are asked to repeat non-words of various syllable lengths. A serial non-word
recognition task, in contrast, is a discrimination task where participants are required to judge
whether the second presentation of a list of items differs from the first. Whereas both types of
tasks involve phonological storage (Baddeley, 2003), non-word repetition tasks are thought to
involve both phonological storage and an articulatory component (Snowling, Chiat, & Hulme,
1991; Bowey, 2001; Baddeley, 2003). Hence, it has been argued that non-word repetition is
possibly a more appropriate test for measuring PM, as vocabulary learning is more closely
related to the articulatory component of PM than the non-articulatory storage component
measured by serial non-word recognition (O’Brien et al., 2007). I opted for a non-word
repetition task, as my study aimed to determine whether differences in PM are correlated with
the degree of effect of orthography on transfer in pronunciation. In other words, given that my
audio-orthographic picture-naming task involved production, it made more sense to use a PM
task that also involved the articulatory component.
3.4.1 Stimuli
The stimuli for the PM task consisted of Farsi words and short phrases, as illustrated in Table
3.6. I chose Farsi for two reasons. First, being a native speaker of this language, it was relatively
easy to control for the linguistic factors such as the inclusion of phonemes that exist in both
82
Farsi and English. Second, it was unlikely for Canadian speakers of English to have knowledge
of this language; this was confirmed in the background questionnaire. While some previous
studies have used synthesized (e.g., Speciale, Ellis & Bywater, 2004) as opposed to natural
speech (e.g., Archibald & Gathercole, 2007; Hummel, 2009) in non-word repetition tasks, in
this study participants were presented with natural speech as it has been suggested that synthetic
stimuli may fail to represent certain perceptually relevant properties of the signal (Beddor &
Gottfried, 1995).
With respect to stimuli length, as shown in Table 3.5, the stimuli comprised non-words
composed of sequences of CV syllables. The stimuli varied from three to nine syllables in count,
with four examples of each count presented in order of increasing length. Some previous studies
such as Speciale et al. (2004) have used syllables varying from one to eight counts and others
such as Hummel (2009) have used syllables varying from three to nine. Given human storage
capacity limits, namely 7 plus or minus two items (Miller, 1956), initially, I considered
including stimuli of 7-9 syllables in length only. However, I noted that in previous studies,
learners had made errors even with word of three syllables in length. Hence, the stimuli in this
experiment ranged from 3 to 9 syllables in length.
Non-words were composed of Farsi sounds that exist in English in order to minimize the
possibility of misperception on the learners’ part. Moreover, this ensured that articulatory
difficulty would not interfere with the learners’ recall capacities. The stimuli were made up of
single words such as [næzæde] ‘not hit’ and adjectival and possessive phrases such as [tæɾɑneje
zibɑ] “beautiful melody”, [kælæmɑte ʃomɑlijɑ] “northerners’ words”, as well as first and last
names [sæmiɾɑje kɑʃɑni] in order to build up longer non-word sequences. In an effort to ensure
consistency in stress placement, as the latter is known to exert a powerful influence on non-word
83
repetition (Roy & Chiat, 2004), only words that were stress final were chosen. The short phrases
were believed not to pose any difficulties given that they were also stress final. Prior to the
running of the experiment, two native English speakers confirmed that the list of words did not
resemble any English words and/or phrases.
Table 3.5
PM Non-word Repetition Task: Farsi Stimuli
# Syllables
Farsi word
Translation
3
4
5
6
7
8
9
[næzæde]
[ʃekæmu]
[sælume]
[bikolɑ]
[zomoɾodi]
[mosibætɑ]
[sɑdeɾɑti]
[pesæɾæmu]
[bitɑ næsiſi]
[mɑɾe ʧæmæni]
[zibɑʃenɑsi]
[nɑbesɑmɑni]
[molɑnɑ pujɑʤu]
[tæɾɑneje zibɑ]
[nilu metɑnæti]
[nimɑ sælɑmæti]
[ʧekɑme gisutælɑ]
[ʃekɑjæte bimænɑ]
[sæmiɾɑje kɑʃɑni]
[nɑzilɑje dolæti]
[kælæmɑte ʃomɑlijɑ]
[ʤæmile gilækizɑde]
[zemɑnæte ʤonubijɑ]
[ʃɑnesɑzije dæɾæke]
[nɑzænine ʃokufɑpænɑ]
[nɑsɑzegɑſije moniɾe]
[ʃiɾinisærɑje sepide]
[dɑɾusɑziye dæɾækɑni]
‘not hit’
‘gluttonous’
‘name’
‘hatless’
‘emerald color’
‘problems’
‘exported’
‘cousin’
Proper name
‘grass snake’
‘aesthetics’
‘turmoil’
Proper name
‘beautiful melody’
Proper name
Proper name
Proper name
‘irrelevant complaint’
Proper name
Proper name
‘northerners’ words’
Proper name
‘southerners’ guarantee’
‘Darake’s comb-making’
Proper name
‘Monire’s unsociability’
‘Sepide’s pastry shop’
‘Darakani’s pharmacy’
Participants were trained prior to the non-word repetition task with three syllable words
read by the same native speaker of Farsi, a twenty nine-year-old male native speaker with
84
university education. The speaker was familiar with the goals of the experiment and was
instructed to read the stimuli at a natural rate, pausing for three seconds between each stimuli.
3.4.2 Task design
Participants were asked to listen to each non-word and repeat it as accurately as possible as soon
as it was presented. There was a 7 second gap between the presentation of each stimulus in order
to give the participants some time before they could concentrate on the next word. The number
of seconds assigned to the gaps between the stimuli was chosen randomly. The entire task took
six minutes per person.
3.5 Testing Protocol
Participants were tested individually in a quiet room. Each recording session took approximately
one and a half hours. All participants were informed orally about the experimental procedures
and general goals (i.e., vocabulary learning in L2 acquisition) and confidentiality issues. At the
beginning of the experiment, they were not told that their pronunciation was going to be
evaluated in this study. In addition, in an effort to ensure that participants would stay motivated
to learn the vocabulary with which they were presented in Spanish, initially, the fact that new
meanings had been assigned to the stimuli was not disclosed to them. Consent forms were
obtained from each individual.
Participants were first presented with the word learning task and then with the PM task.
This order was established based on two assumptions. First, learners could perceive the PM task
as more difficult than the word learning task due to the fact that the latter was based on Farsi
and Farsi is more distant from English in comparison with Spanish. Second, there was a
85
possibility that their perception of language distance would lead to a decrease in self-confidence
and motivation and would adversely affect the results were the Spanish word learning task
administered second.
The word learning task took approximately 45 minutes as each round (a total of 108
words) took 12 minutes and participants took a 3 minute break after each round. The PM nonword repetition task, on the other hand, took 6 minutes. All participants were provided with a 5minute break between the two tasks. Longer breaks were provided if requested by the
participants.
Prior to each task, participants were given instructions and briefly received some
training. The instructions were formulated to minimize the possibility of directing participants’
attention to either the orthographic or the auditory input. For example, in the auditoryorthographic conditions where the participants were exposed to orthographic input, they were
told that they were going to be presented with words in groups of threes, where they would hear
each word in Spanish three times and at the same time see their corresponding pictures and
writing in Spanish. They were instructed to learn the words and then name them in Spanish after
the presentation of each group of words. Instructions such as listen to the words or read the
words were avoided based on the assumption that they would produce greater orienting of
attention to one type of input. Participants were also reminded of the fact that they would have
to repeat the entire task three times and could take a three-minute break between each round.
The training for the word learning task involved simulating the experiment with one triplet with
the same auditory and orthographic condition that they were going to be exposed to in the actual
task. The words in the triplet used for training were different from the stimuli in the actual task
and conformed to the criteria set for stimuli creation.
86
For the PM task, on the other hand, participants were told that they were going to be
presented with a list of 28 Farsi words, each separated by a 7-second gap. They were instructed
to listen to and immediately repeat each word after hearing it. The training for the PM non-word
repetition task included simulating the experiment by presenting each participant with four Farsi
words other than those used in the task. These words, as in the actual task, also increased in
syllable counts (i.e., from 3-4).
Subsequent to the completion of the PM task, I filled out a questionnaire for each
participant. Upon the completion of the questionnaire, the participants were fully debriefed
about the initial lack of disclosure with respect to the assignment of new meanings to the
Spanish stimuli and the specific goals of the experiment as well as the reasons for this approach.
They were then offered a complete list of these Spanish words with their real meanings and
were provided with the opportunity to re-consent to the use of their data in this study. All
participants consented to the continued use of their data. Each session ended with compensating
the participants with 20 Canadian dollars.
The participants were recorded individually in Toronto. The recording equipment used
included an M-Audio Micro-track 24/96 professional 2-channel mobile digital recorder and a
lavaliere unidirectional microphone. The recordings were made at a sampling rate of 44.2 kHz
and a quantization rate of 16 bits; the audio files containing the extracted tokens were downsampled at 22.1 kHz and saved in wave format.
All in all, in this chapter I first provided my hypotheses. In sum, it was hypothesized that
presence of orthography will promote L1-based transfer in the acquisition of TL sounds by
novice learners of Spanish where the following factors will influence the proportion of transfer:
(i) auditory-orthographic condition at learning and production; (ii) grapheme-to-phoneme
87
inconsistency between Spanish and English as well as (iii) different aspects of PM, including
primacy and recency effects, repetition effects and individual PM capacity. In addition to stating
the hypotheses, I explained the two tasks included in this experiment, namely the Spanish based
picture-naming task and the Farsi-based PM non-word repetition task. In the next chapter, I will
provide the results for both these tasks.
88
Chapter 4
Data analysis and results: Word learning and PM tasks
In the preceding chapter, I outlined the hypotheses and described the methods employed in the
present experiment designed to test (a) the effect of orthography on L1-based phonological
transfer and (b) the effect of factors that might affect the proportion of orthography-induced
transfer, namely auditory-orthographic condition, grapheme-to-phoneme inconsistency and the
memory related factors of primacy and recency, repetition and individual PM. In this chapter, I
will describe the data analysis and report the results and evaluate the hypotheses outlined in
Chapter 3.
The remainder of this chapter is structured as follows. In Section 4.1 I will report the data
analysis and the results of the picture-naming task and in Section 4.2, I will do the same for the
PM task. The picture-naming task results in Section 4.1.2 show that orthography clearly triggers
L1-based phonological transfer. In addition, a strong effect is noted for auditory-orthographic
condition and grapheme-to-phoneme inconsistency between Spanish and English in shaping the
proportion or orthography-induced transfer. The results for the effect of phenomena
characterizing PM working, on orthography-induced transfer were mixed. There was a weak
recency pattern at the list level, a substantial primacy effect at the word level and some effect of
repetition/round. In addition, as shown in Section 4.2.2, there was no correlation between the
variation in individual PM and proportion of orthography-induced transfer. I now turn to the
data analysis and results for the picture-naming task.
89
4.1 Picture-naming task
In this section I will first provide the data analysis, where I will describe the linguistic
background of the transcriber, inter-transcriber reliability and the coding procedure. I will then
report the findings with regard to the overall effect of orthography and the various factors that
were analyzed in this study: auditory orthographic condition, inconsistency between TL and L1
grapheme-to-phoneme inconsistency and phenomena that characterize PM working, namely
primacy and recency effects. I now turn to data analysis.
4.1.1 Data Analysis
The learners’ productions from the picture-naming tasks were transcribed by two individuals,
namely the author and another linguist with training in phonetics and L2 acquisition. Her native
language was English and she had near-native fluency in Spanish. The author was a native
speaker of Farsi with near-native fluency in English and Spanish. Inter-transcriber reliability
was 98% (there were disagreements on a total of 168 of 8385 tokens). The author resolved the
small number of disagreements both with the help of a native Spanish-speaking phonetician, and
by doing a visual inspection using PRAAT. For example, when determining whether a sound
had been produced as a [b] or a [v], faint formant patterns or aperiodic noise were interpreted as
[v] whereas a gap with a low frequency voicing bar of vertical striations was transcribed as [b].
As the hypotheses regarding orthography-based transfer focused on the particular
realization of individual phonemes, only the target sounds were transcribed. Nonetheless, any
unusual (mis)productions, such as production of combinations of L1 and TL sounds that I will
refer to as blends (e.g., when <ll> was produced as [lj] or [lij] as in [poljeɾo], [polijeɾo]), were
noted. The data were coded as (i) involving ‘transfer’ when a learner’s production consisted of
90
the non-target-like substitution of an L1 sound for the target sound or a combination of the L1
and TL sound. These included the production of [v] for [b] (e.g., <vireca>-[viɾeka] ), [d] for/ð/
(e.g., <codena>-[kodena] or [koɾena] instead of [koðena]), [z] for [s] (e.g., <zatara>-[zatara]
instead of [satara]), [h] for silent <h> (<harapo>-[haɾapo] instead of [aɾapo]), and [l], [lj] and
[lij] for [j] (e.g., <pollero>-[poleɾo], [poljeɾo], [polijeɾo] instead of [pojeɾo]); (ii) ‘correct’ if it
was the same as the target sound; (iii) as ‘not produced’ if the entire word was not produced;
and (iv) as ‘deleted’ if only the target sound was deleted. All other productions, such as a [g] for
target /b/, were coded as ‘other’.
In total, there were 8640 tokens. 1765 tokens (20.41%) were coded as ‘transfer’, 5696
(65.9%) as ‘correct’, 832 (9.63%) as ‘not produced’, 46 (0.5%) as ‘deleted’; and 301 (3.5%) as
‘other’. When analyzing the proportion of transfer, the tokens coded as transfer were given a
value of 1 and the correct ones a value of ‘0’. The tokens that were coded as ‘not produced’,
‘deleted’ and ‘other’ which all together comprised 1179 tokens (14% of the total number of
tokens) were not analyzed. Due to the binary nature of the variable, the data were aggregated
across rounds. Mean proportion of transfer was calculated for grapheme-to-phoneme
correspondences that resulted in transfer leading to non-target-like productions across rounds.
For example, for <z>-/s/, the total number of transfer productions (e.g., [z]) across the three
rounds by all learners was divided by the sum of the total correct (e.g., [s]) and transfer
productions. Kruskal-Wallis and Mann-Whitney tests (non-parametric tests) were conducted.
4.1.2 Results: Picture-naming task
In this section, first, I will report the results concerning overall effect of orthography on L1based phonological transfer. Second, I will report on the factors that were hypothesized to affect
91
the proportion of L1-based phonological transfer. As mentioned previously, these included (1)
auditory-orthographic condition at learning and production; (2) type of grapheme-to-phoneme
correspondence; and (3) memory related factors: (i) positional effects primacy and recency
effects at the list level; (ii) primacy effects at the word level, and (iii) repetition/ round. The
effect of these factors will be examined by collapsing the results across the grapheme-tophoneme correspondences that resulted in transfer, as well as by analyzing the effects for each
grapheme-to-phoneme correspondence independently. I begin with the first of these factors.
4.1.2.1 Overall effect of orthography and differences between
auditory-orthographic conditions
Hypothesis 1 based on Young-Scholten et al., (1999), Young-Scholten (2000, 2002) and
Erdener and Burnham (2005), predicted that orthography will promote phonological transfer
leading to non-target-like productions and Hypothesis (2a) based on Young-Scholten et al.
(1999) and Erdener and Burnham (2005), predicted that each auditory-orthographic condition
will affect the proportion of transfer differently. Specifically, it was predicted that the condition
ortho-learning & production would lead to the highest proportion of transfer followed by
orthography-learning, followed by ortho-production, and, having the least effect, the auditory
only condition.
Table 4.1 provides the mean proportion transfer scores and standard deviations for the
effect
of
auditory-orthographic
condition,
collapsing
across
grapheme-to-phoneme
correspondences that triggered transfer leading to non-target-like productions. In accordance
with the general prediction that orthography would promote phonological transfer, all the
orthographic conditions resulted in a higher proportion of transfer than the auditory only
92
condition. Indeed, the auditory-only condition exhibited a very low amount of transfer leading
to non-target-like production. In addition, as predicted in hypothesis 2, the proportion of transfer
differed between the orthographic conditions. Specifically, the ortho-production condition
resulted in a lower proportion of transfer than ortho-learning & production and ortho-learning
conditions. However, contrary to the predictions, the ortho-learning & production and ortholearning conditions resulted in almost the same proportion of transfer leading to non-target-like
production.
Table 4.1
Mean Proportion Transfer and Standard Deviations by Condition
Condition
Ortho-learning & production
Ortho-learning
Ortho-production
Auditory only
M
.53
.54
.43
.08
SD
.45
.45
.44
.25
A Kruskal-Wallis test revealed that auditory-orthographic condition at learning and
production indeed had a significant influence on the proportion of learner forms involving
transfer (χ2(df = 3) = 243.73, p = .000). Mann-Whitney tests were also conducted to see where
the differences lay (see Table 4.2). With the exception of the ortho-learning & production and
ortho-learning, all pair-wise condition comparisons were significant.
Table 4.2
Mann-Whitney Test Results for the Effect of Condition on the Mean Proportion transfer
Condition
Ortho-learning & production & Ortho-learning
Ortho-learning & production & Ortho-production
Ortho-learning & production & Auditory only
Ortho-learning & Ortho-production
Ortho-learning & Auditory only
Ortho-production & Auditory only
U
57722.50
52086.00
1325.50
49981.50
24781.50
29866.00
z
-.95
-2.95
-6.00
-2.99
-13.94
-12.19
p
.950
.003
.000
.003
.000
.000
93
In sum, the results clearly show that orthography does promote phonological transfer during
the L2 acquisition of phonology, at least at the very early stages of acquisition tested in the
present study. Specifically, when all grapheme-to-phoneme correspondences for which the
learners’ productions involved transfer that resulted in non-target-like production were
collapsed, the ortho-learning & production and ortho-learning conditions resulted in an equal
proportion of orthography-induced transfer leading to non-target-like production (.53 and .54
respectively), followed by the condition ortho-learning (.43). The auditory only condition, in
sharp contrast, resulted in a very low proportion of transfer leading to non-target-like
production. Having looked at the results for all grapheme-to-phoneme mappings collapsed
together, I now turn to an analysis of individual mappings.
4.1.2.2 Effect of grapheme-to-phoneme inconsistency
Hypothesis (2b) predicted that, while those grapheme-to-phoneme correspondences that are the
same in Spanish and English would not result in any transfer leading to non-target-like
productions in any orthographic condition, those grapheme-to-phoneme correspondences that
differ between Spanish and English would result in such transfer.
The grapheme-to-phoneme correspondences that are the same in Spanish and English were
as follows: <m>-/m/, <n>-/n/, <s>-/s/ word-initially, <d>-/d/ word-initially, <b>-/b/ wordinitially, and <h>-/Ø/ VCV (intervocalically) in unstressed position. None of the same
grapheme-to-phoneme correspondences involved transfer leading to non-target-like productions
with one exception, namely <h>-/Ø/ where <h> was realized by many learners as [h]. Mean
proportion transfer scores and standard deviations for <h>-/Ø/ VCV (e.g., <ahumar>-[aumaɾ]
produced as [ahumaɾ]) and other grapheme-to-phoneme correspondences discussed below are
94
presented in Table 4.3. Table 4.3 shows that exhibited transfer in every orthographic condition
(orthography-learning & production, orthography-learning and orthography-production).
Table 4.3
Mean Proportion Transfer and Standard Deviations for Spanish Grapheme-to-phoneme
correspondences by Condition
Spanish graphemeCondition
to-phoneme
Ortho-learning
correspondence
& production
Ortho-learning Ortho-production Auditory only
M
SD
M
SD
M
SD
M
SD
<v>-/b/
.99
.05
.92
.07
.77
.32
.13
.28
<d>-/ð/
.92
.29
.98
.07
.90
.25
.55
.46
<z>-/s/
.69
.35
.67
.40
.64
.37
.00
.00
<h>-/Ø/ #
.48
.42
.56
.41
.30
.36
.00
.00
<h>-/Ø/ VCV
.49
.41
.17
.29
.18
.33
.00
.00
<ll>-/j/
.01
.20
.21
.40
.09
.37
.00
.00
The other grapheme-to-phoneme-correspondences that triggered transfer in every
orthographic condition were those that were considered to differ from their English counterparts, namely: <v>-/b/, <d>-/ð/, <z>-/s/, <h>-/Ø/ # (word-initially), and <ll>-/j/; <v>
corresponds to /b/ in Spanish but to /v/ in English (e.g., <vireca> [biɾeka], and <vote>-[vot],
respectively), <d>- corresponds to /ð/ in Spanish but to /d/ in English (e.g., <adorno> [aðoɾno]
and <madam>-[mædəm], respectively), <z> corresponds to /s/ in Spanish but <z>-/z/ in English
(e.g., <zatara>-[sataɾa] and <zoo>-[zu], respectively), <h> is silent word initially in Spanish but
realized as /h/ in English in this position (e.g., <hanega>-[aneɣa] and <hand>-[hænd],
respectively) and <ll> corresponds to [j] in Spanish but to [l] in English (e.g., <pollero>[pojeɾo], <balloon>-[bəlun], respectively. The mean proportion transfer scores and standard
deviations for these grapheme-to-phoneme correspondences are also summarized in Table 4.3.
Table 4.3 shows that the mean proportion transfer differed between individual grapheme-tophoneme correspondences. In addition, contrary to the predictions, two of the different
grapheme-to-phoneme correspondences, namely <v>-/b/ and <d>-/ð/, also resulted in some
95
transfer in the auditory-only condition. A Kruskal-Wallis test conducted to see whether the
factor different grapheme-to-phoneme correspondence was significant. It revealed that there was
a significant difference in mean proportion transfer within the ortho-learning & production
condition, (χ2(df = 5) = 199.70, p = .000), ortho-learning condition, (χ2(df = 5) = 177.35, p =
.000), ortho-learning & production condition, (χ2(df = 5) = 174.64, p = .000), and auditory
only condition, (χ2(df = 5) = 139.74, p = .000).
An interesting pattern was noted with respect to grapheme-to-phoneme correspondences
that resulted in a certain proportion of transfer leading to non-target-like productions in the
auditory only condition, namely <v>-/b/ and <d>-/ð/. Table 4.4 shows that these grapheme-tophoneme correspondences resulted in a higher mean proportion transfer in all three orthographic
conditions than in the auditory only condition. These results suggest that TL sounds that are
prone to transfer, exhibit a higher proportion of transfer in the presence of orthographic input.
Another finding with respect to grapheme-to-phoneme correspondences that resulted in
transfer was that particular grapheme-to-phoneme correspondences led to different mean
proportion transfer scores as shown in Table 4.3. In other words, the mean proportion transfer
differed between grapheme-to-phoneme correspondences. When considering grapheme-tophoneme correspondences that are different in Spanish and English, the results point to a trend
where <v>-/b/ and <d>-/ð/ resulted in the highest mean proportion transfer, followed by <z>-/s/,
<h>- /Ø/ #, and <ll>-/j/. Mann-Whitney tests were conducted to see whether these differences
were significant. The results summarized in Table 4.4 show that with the exception of <v>-/b/
and <z>-/s/ (p = .052) in the orthography-production condition, and <z>-/s/ and <h>-[Ø] # (p =
.70), there was a significant difference in the proportion of transfer between all the other
‘different’ grapheme-to-phoneme correspondences in each auditory-orthographic condition.
96
Table 4.4
Mann-Whitney Test Results for Pair-wise Comparisons of Spanish Grapheme-to-sound
Correspondences by Condition
Condition
Spanish Grapheme-to-sound
U
z
correspondence
Ortho-learning &
<v>-[b] & <d>-[ð]
1123.00
-1.98
production
<v>-[b] & <z>-[s]
544.500
-6.01
<v>-[b] & <h>-[Ø] #
412.00
-6.68
<v>-[b] & <h>- [Ø] VCV
386.50
-3.39
<v>-[b] & <ll>-[j]
77.00
-10.94
<d>-[ð] & <z>-[s]
697.00
-4.48
<d>-[ð] & <h>-[Ø] #
531.00
-5.67
<d>-[ð] & <h>-[Ø] VCV
509.00
-5.67
<d>-[ð] & <ll>-[j]
201.00
-10.38
<z>-[s] & <h>-[Ø] #
925.00
-2.32
<z>-[s] & <ll>-[j]
478.00
-9.11
<z>-[s] & <h>-[Ø] VCV
908.00
-2.29
<h>-[Ø] I & <ll>-[j] #
1082.00
-6.73
<h>-[Ø] I & <h>-[Ø] VCV
1207.50
-.13
<h>-[Ø] VCV & <ll>-[j]
975.50
-6.73
Ortho-learning
<v>-[b] & <d>-[ð]
1177.00
-.44
<v>-[b] & <z>-[s]
496.50
-6.01
<v>-[b] & <h>-[Ø] #
490.50
-6.06
<v>-[b] & <h>- [Ø] VCV
54.00
-8.62
<v>-[b] & <ll>-[j]
337.50
-9.32
<d>-[ð] & <z>-[s]
508.00
-5.9
<d>-[ð] & <h>-[Ø] #
500.00
-5.93
<d>-[ð] & <h>-[Ø] VCV
55.00
-8.56
<d>-[ð] & <ll>-[j]
345.00
-9.26
<z>-[s] & <h>-[Ø] #
1148.50
-.38
<z>-[s] & <h>-[Ø] VCV
447.00
-4.90
<z>-[s] & <ll>-[j]
1189.50
-5.41
<h>-[Ø]I & <ll>-[j]
1224.00
-5.25
<h>-[Ø] I & <h>-[Ø] VCV
475.00
-4.66
<h>-[Ø] VCV & <ll>-[j]
2010.50
-.032
Ortho-production
<v>-[b] & <d>-[ð]
970.50
-2.41
<v>-[b] & <z>-[s]
989.00
-1.94
<v>-[b] & <h>-[Ø] #
448.00
-5.76
<v>-[b] & <h>- [Ø] VCV
296.00
-6.68
<v>-[b] & <ll>-[j]
385.00
-9.27
p
.048
.000
.000
.001
.000
.000
.000
.000
.000
.020
.000
.022
.000
.899
.000
.661
.000
.000
.000
.000
.000
.000
.000
.000
.700
.000
.000
.000
.000
.974
.016
.052
.000
.000
.000
(continued)
97
Table 4.4
Mann-Whitney Test Results for Pair-wise Comparisons of Spanish Grapheme-to-sound
Correspondences by Condition (continued)
Condition
Spanish Grapheme-to-sound
U
z
correspondence
<d>-[ð] & <z>-[s]
732.00
-4.10
<d>-[ð] & <h>-[Ø] #
292.50
-7.07
<d>-[ð] & <h>-[Ø] VCV
206.50
-7.59
<d>-[ð] & <ll>-[j]
303.50
-9.75
<z>-[s] & <h>-[Ø] #
653.00
-4.25
<z>-[s] & <h>-[Ø] VCV
446.50
-5.51
<z>-[s] & <ll>-[j]
630.50
-8.23
<h>-[Ø] I & <ll>-[j]
1560.00
-4.32
<h>-[Ø] I & <h>-[Ø] VCV
949.50
-1.84
<h>-[Ø] v & <ll>-[j]
1911.50
-1.97
Auditory only
<v>-[b] & <d>-[ð]
441.50
-4.17
<v>-[b] & <z>-[s]
931.00
-3.49
<v>-[b] & <h>-[Ø] #
931.00
-3.39
<v>-[b] & <h>- [Ø] VCV
874.00
-3.39
<v>-[b] & <ll>-[j]
1805.00
-4.78
<d>-[ð] & <z>-[s]
318.50
-6.28
<d>-[ð] & <h>-[Ø] #
318.50
-6.28
<d>-[ð] & <h>-[Ø] VCV
299.00
-6.12
<d>-[ð] & <ll>-[j]
617.50
-8.32
<z>-[s] & <h>-[Ø] #
1127.00
.000
<z>-[s] & <h>-[Ø] VCV
1200.50
.000
<z>-[s] & <ll>-[j]
2327.50
.000
<h>-[Ø] I & <ll>-[j]
931.00
.000
<h>-[Ø] I & <h>-[Ø] VCV
1127.00
.000
<h>-[Ø] VCV & <ll>-[j]
2185.00
.000
p
.000
.000
.000
.000
.000
.000
.000
.000
.066
.048
.000
.000
.001
.001
.000
.000
.000
.000
.000
1.00
1.00
1.00
1.00
1.00
1.00
As mentioned, a surprising finding was that <h>-/Ø/ VCV, which was predicted not to result
in transfer because it was considered to be the same in Spanish and English, also resulted in
transfer. In other words, <h> was produced as [h] in unstressed intervocalic position in Spanish
(e.g., <ahumar>-[aumaɾ] was produced as [ahumaɾ]) when it is silent in unstressed intervocalic
position in English (e.g., <vehicle>-[vi:jəkəl]). Tables 4.3 show that <h>-/Ø/ VCV lead to a
lower mean proportion transfer than <h>-/Ø/ # in the ortho-learning & production (.49% and
.48%, respectively) and ortho-learning conditions (.17% and .56%, respectively) but triggered a
98
higher mean proportion transfer in ortho-production (.18% and .30% respectively). However, as
shown in Table 4.4, based on Mann-Whitney tests, the differences in the ortho-learning &
production and ortho-production conditions were not significant. In other words, only in one
orthographic condition did <h>-/Ø/ VCV result in a lower proportion of transfer than <h>-/Ø/ #.
In sum, the factor ‘grapheme-to-phoneme inconsistency’ significantly affected the mean
proportion transfer in each condition. Those grapheme-to-phoneme correspondences that were
different in Spanish and English induced transfer leading to non-target-like productions in the
orthographic conditions whereas those that were the same, with the exception of <h>-/Ø/ VCV,
did not do so. Moreover, mean proportion transfer differed between those grapheme-tophoneme correspondences that resulted in transfer. Finally, <v>-/b/ and <d>-/ð/ resulted in some
transfer in the auditory only condition as well and, interestingly, the mean proportion transfer
for these two grapheme-to-phoneme correspondences was significantly higher in the
orthographic conditions than in the auditory only condition. I now turn to the effect of auditoryorthographic condition on individual grapheme-to-phoneme correspondences.
4.1.2.3 Effect of auditory-orthographic condition on individual
grapheme-to-phoneme correspondences
In section 4.1.2.2, I showed that auditory-orthographic condition at learning and production
significantly affected the mean proportion transfer when all the grapheme-to-phoneme
correspondences that resulted in transfer were collapsed. In section 4.1.2.3, I analyzed which
grapheme-to-phoneme correspondences resulted in transfer. I will now explore whether the
findings in section 4.1.2.2 regarding the effect of auditory-orthographic condition on mean
99
proportion transfer are generalizable to each and every single grapheme-to-phoneme
correspondence that led to transfer.
When looking at the overall effect of auditory-orthographic condition at learning and
production, we have seen that the orthographic conditions were significantly different from the
auditory only condition. However, within the orthographic conditions, significant differences
were only found between the ortho-learning & production and ortho-production conditions.
Table 4.3 summarizes the results for the mean proportion transfer for each grapheme-tophoneme correspondence resulting in transfer by auditory–orthographic condition.
Mann-Whitney tests were conducted to test whether auditory-orthographic condition was
also a significant factor for each grapheme-to-phoneme correspondence that resulted in transfer
(Table 4.5). The findings concerning the effect of auditory-orthographic condition when all
grapheme-to-phoneme correspondences were collapsed do not hold for all individual graphemeto-phoneme correspondences. Although for every grapheme-to-phoneme correspondence, the
orthographic conditions were significantly different from the auditory only condition, the
differences between the orthographic conditions were non-significant for some of the graphemeto-phoneme correspondences. As illustrated in Tables 4.6 and 4.7, when looking at the effect of
auditory-orthographic condition on individual grapheme-to-phoneme correspondences that
resulted in transfer, the following results were obtained. For <d>-/ð/ and <z>-/s/, there were no
significant differences between the ortho-learning & production, ortho-learning, and orthoproduction conditions. For <v>-/b/ and <h>-/Ø/ #, while the ortho-learning & production and
ortho-learning conditions did not differ significantly both the ortho-learning & production and
ortho-learning conditions resulted in a significantly higher mean proportion transfer than the
ortho-production condition. Moreover, <h>-/Ø/ VCV was the only grapheme-to-phoneme
100
correspondence for which the prediction that the condition ortho-learning & production would
result in a higher mean proportion transfer in comparison with ortho-learning was confirmed.
For <h>-/Ø/ VCV, the ortho-learning & production also resulted in a significantly higher mean
proportion transfer than the ortho-production condition, however, there were no significant
differences between the ortho-learning and ortho-production conditions. Finally, <ll>-/j/ was the
only grapheme-to-phoneme correspondence for which the ortho-learning & production
condition resulted in a lower mean proportion transfer than in the ortho-learning condition and
there were no significant differences between the conditions ortho-learning & production and
ortho-production. However, the ortho-learning condition did lead to a higher mean proportion
transfer than the ortho-production condition for <ll>-/j/.
Table 4.5
Mann-Whitney Results for the Effect of Condition for Spanish Grapheme-to-phoneme
Correspondences
Spanish
Grapheme-tophoneme
correspondence
<v>-/b/
<d>-/ð/
Condition
Ortho-learning & production & ortho-learning
Ortho-learning & production & ortho-production
Ortho-learning & ortho-production
Ortho-learning & production & auditory only
Ortho-learning & auditory only
Ortho-learning & production & auditory only
Ortho-learning & production & ortho-learning
Ortho-learning & production & ortho-production
Ortho-learning & ortho-production
Ortho-learning & production & auditory only
Ortho-learning & auditory only
Ortho-production & auditory only
U
1199.50
770.00
775.00
77.50
78.50
232.50
1146.00
1181.00
1072.00
503.00
442.00
533.00
z
p
-.601
-4.65
-4.33
-8.99
-8.86
-7.36
-1.11
-.77
-1.89
-4.04
-4.84
-3.55
.548
.000
.000
.000
.000
.000
.268
.443
.059
.000
.000
.000
(continued)
101
Table 4.5
Mann-Whitney Results for the Effect of Condition for Spanish Grapheme-to-phoneme
Correspondences (continued)
Spanish
Grapheme-tophoneme
correspondence
<z>-/s/
<h>-/Ø/ #
<h>-/Ø/ VCV
<ll>-/j/
Condition
U
z
p
Ortho-learning & production & ortho-learning
Ortho-learning & production & ortho-production
Ortho-learning & ortho-production
Ortho-learning & production & auditory only
Ortho-learning & auditory only
Ortho-production & auditory only
Ortho-learning & production & ortho-learning
Ortho-learning & production & ortho-production
Ortho-learning & ortho-production
Ortho-learning & production & auditory only
Ortho-learning & auditory only
Ortho-production & auditory only
Ortho-learning & production & ortho-learning
Ortho-learning & production & ortho-production
Ortho-learning & ortho-production
Ortho-learning & production & auditory only
Ortho-learning & auditory only
Ortho-production & auditory only
Ortho-learning & production & ortho-learning
Ortho-learning & production & ortho-production
Ortho-learning & ortho-production
Ortho-learning & production & auditory only
Ortho-learning & auditory only
Ortho-production & auditory only
1089.50
1176.50
1158.50
147.00
294.00
171.50
1095.00
941.00
777.50
416.50
294.00
612.50
553.00
649.00
956.50
345.00
644.00
759.00
4009.50
4559.50
3986.50
3990.00
3325.00
3800.00
-.99
-.53
-.48
-8.34
-7.43
-8.20
-.95
-2.23
-3.26
-6.77
-7.43
-3.99
-3.89
-3.96
-.072
-6.81
-4.10
-3.98
-2.51
-.396
-2.09
-3.81
-5.25
-4.02
.323
.569
.628
.000
.000
.000
.344
.026
.001
.000
.000
.000
.000
.000
.943
.000
.000
.000
.012
.692
.036
.000
.000
.000
In sum, when looking at individual grapheme-to-phoneme correspondences, although for
one grapheme-to-phoneme correspondence the prediction that the ortho-learning & production
would result in a higher mean proportion transfer than the ortho-learning condition held (e.g.,
<h>-/Ø/ VCV) and for another the ortho-learning & production condition resulted in a lower
mean proportion transfer than ortho-learning condition (e.g., <ll>-/j/ ), for 4 out of 6 cases (<v>/b/, <d>-/ð/, <z>-/s/ and <h>-/Ø/ #), there were no significant differences between the
conditions ortho-learning & production and ortho-learning. As for the predictions that the orthoproduction condition would result in a significantly lower mean proportion of transfer than the
102
ortho-learning & production condition, this was the case for correspondences, namely <v>-/b/,
<h>-/Ø/ #, and <h>-/Ø/ VCV. The grapheme-to-phoneme correspondences <v>-/b/, <h>-/Ø/ #
as well as <ll>-/j/, exhibited a significantly lower proportion of transfer in the ortho-production
condition in comparison with the ortho-learning condition. In the remaining cases, the
conditions ortho-learning & production and ortho-production as well as ortho-learning and
ortho-production did not differ significantly. Each of the orthographic conditions (ortho-learning
& production, ortho-learning and ortho-production) differed significantly from the auditory only
condition.
In this and the preceding sections, I have discussed the effect of auditory-orthographic
condition and grapheme-to-phoneme inconsistency. The overall results have for the most part
supported the hypotheses regarding these two factors. I now turn to the results regarding the
effect of PM on shaping orthography-induced transfer. These include the effect of primacy and
recency, repetition and PM capacity. I begin with the results for the effect of primacy and
recency.
4.1.2.4 Effect of primacy and recency
In this section, I will provide the results for primacy and recency at the list level by reporting the
effect of position within the triplet for grapheme-to-phoneme correspondences that resulted in
transfer in each of the orthographic conditions. Hypothesis (2ci) predicted that primacy and
recency effects (e.g., Deese & Kaufman, 1957; Murdock, 1962; Waugh & Norman, 1965; Craik,
1970; Rundus & Atkinson, 1970; Rundus, 1971; Foreit, 1976) would decrease the proportion of
orthography-induced transfer leading to non-target-like productions at the list level/within the
triplet., Specifically it was predicted that words learnt first and produced/recalled first within a
triplet (1*1) and/or words learnt last and produced first within a triplet (3*1) would result in a
103
lower mean proportion transfer than words learnt first and recalled last (1*3), words learnt last
and tested last (3*3), and words learnt in the middle and tested in the middle (2*2).
Table 4.6 provides the mean proportion transfer scores and standard deviations by position
within triplet for the conditions ortho-learning & production, ortho-learning, and orthoproduction.
Table 4.6
Mean Proportion Transfer and Standard Deviations for Position within Triplet by Condition
Condition
Ortho-learning & production
Ortho-learning
Ortho-production
Position within triplet
1*1
1*3
2*2
3*1
3*3
1*1
1*3
2*2
3*1
3*3
1*1
1*3
2*2
3*1
3*3
M
.31
.30
.27
.28
.33
.32
.30
.26
.31
.33
.27
.27
.22
.19
.25
SD
.31
.30
.27
.28
.34
.32
.30
.27
.31
.33
.27
.27
.22
.19
.25
In the ortho-learning & production and ortho-learning conditions, neither position (1*1) nor
(3*1) resulted in the lowest mean proportion transfer; it was rather position (2*2) that had this
effect. In other words, there was no evidence for the predicted primacy or recency effects in
these orthographic conditions. In the ortho-production condition, on the other hand, consistent
with the predictions, position (3*1) led to the lowest mean proportion transfer.
Kruskal-Wallis tests were conducted to determine whether the effect on the mean proportion
transfer of position within the triplet at the time of learning and production was significant. The
test revealed that the factor position within the triplet was not significant for any of the
104
orthographic conditions (ortho-learning & production (χ2(df = 4) = 1.56, p = .814); ortholearning (χ2(df = 4) = 1.15, p = .814); ortho-production (χ2(df = 4) = 3.29, p = .511)). In other
words, primacy or recency effects at the list level (within the triplet) did not affect the mean
proportion transfer when all grapheme-to-phoneme correspondences that resulted in transfer
were collapsed.
In sum, whereas position (3*1) led to the lowest mean proportion transfer in the orthoproduction condition, there was no evidence to support a significant primacy or a
recency/positional effect at the list level within the triplet (at the list level). I now turn to the
effect of primacy and recency on individual grapheme-to-phoneme correspondences.
4.1.2.5 Effect of primacy and recency on individual grapheme-tophoneme correspondences
In section 4.1.2.4, the results showed that there were no significant primacy or recency effects at
the list level when grapheme-to-phoneme correspondences were collapsed. In this section, I will
explore whether this is also true for each grapheme-to-phoneme correspondence that resulted in
transfer leading to non-target-like behavior in the orthographic conditions. The results for the
mean proportion transfer and standard deviations by position for the grapheme-to-phoneme
correspondence in the ortho-learning & production, ortho-learning, and ortho-production
conditions are summarized in Table 4.7.
105
Table 4.7
Mean Proportion Transfer and Standard Deviations for Position within Triplet by Spanish
grapheme-to-phoneme Correspondence by Condition
Spanish Grapheme-tophoneme
correspondences
<v>-/b/
Condition
Ortho-learning & production
Ortho-learning
Ortho-production
<d>-/ð/
Ortho-learning & production
Ortho-learning
Ortho-production
<z>-/s/
Ortho-learning & production
Ortho-learning
Position within
triplet
1*1
1*3
2*2
3*1
3*3
1*1
1*3
2*2
3*1
3*3
1*1
1*3
2*2
3*1
1*1
1*3
2*2
3*1
3*3
1*1
1*3
2*2
3*1
3*3
1*1
1*3
2*2
3*1
3*3
1*1
1*3
2*2
3*1
3*3
1*1
1*3
2*2
M
SD
.96
1.00
1.00
1.00
1.00
.97
1.00
1.00
1.00
1.00
.97
1.00
1.00
.97
1.00
.83
.93
.83
1.00
1.00
.96
.97
.97
1.00
1.00
.87
.97
.77
.90
.70
.70
.77
.53
.73
.50
.43
.61
.10
.00
.00
.00
.00
.10
.00
.00
.00
.00
.10
.00
.00
.10
.00
.36
.14
.36
.00
.00
.11
.08
.10
.00
.00
.32
.10
.77
.32
.37
.37
.23
.42
.34
.44
.44
.31
(continued)
106
Table 4.7
Mean Proportion Transfer and Standard Deviations for Position within Triplet by Spanish
grapheme-to-phoneme Correspondence by Condition (continued)
Spanish Grapheme-tophoneme
correspondences
Condition
Ortho-production
<h>-/Ø/ #
Ortho-learning & production
Ortho-learning
Ortho-production
<h>-/Ø/ VCV
Ortho-learning & production
Ortho-learning
Ortho-production
Position within
triplet
3*1
3*3
1*1
1*3
2*2
3*1
3*3
1*1
1*3
2*2
3*1
3*3
1*1
1*3
2*2
3*1
3*3
1*1
1*3
2*2
3*1
3*3
1*1
1*3
2*2
3*1
3*3
1*1
1*3
2*2
3*1
3*3
1*1
1*3
2*2
3*1
M
SD
.67
.77
.67
.76
.68
.37
.70
.46
.50
.00
.43
.53
.57
.65
.64
.43
.53
.33
.30
.32
.13
.33
.50
.45
.47
.50
.56
.39
.05
.07
.24
.15
.18
.31
.20
.20
.47
.32
.35
.27
.35
.43
.40
.48
.45
.00
.41
.42
.47
.47
.39
.39
.35
.41
.41
.38
.28
.29
.48
.42
.45
.33
.44
.44
.17
.07
.38
.23
.34
.43
.29
.35
(continued)
107
Table 4.7
Mean Proportion Transfer and Standard Deviations for Position within Triplet by Spanish
grapheme-to-phoneme Correspondence by Condition (continued)
Spanish Grapheme-tophoneme
correspondences
Condition
<ll>-/j/
Ortho-learning & production
Ortho-learning
Ortho-production
Position within
triplet
3*3
1*1
1*3
2*2
3*1
3*3
1*1
1*3
2*2
3*1
3*3
1*1
1*3
2*2
3*1
3*3
M
SD
.00
.50
.45
.47
.48
.57
.44
.17
.11
.38
.22
.33
.43
.29
.35
.00
.00
.47
.42
.48
.33
.44
.44
.17
.11
.38
.23
.34
.43
.29
.35
.00
When the results were analyzed for individual grapheme-to-phoneme correspondences,
position (3*1) was the position that induced the lowest proportion of transfer most frequently,
namely six times, on its own and twice with positions (1*1) and (3*1). In other words, a recency
pattern was the most frequent pattern. In addition, in contrast to the findings when all the
grapheme-to-phonemes resulting in transfer were collapsed, other positions also triggered a
lower proportion of transfer. For example, position (1*1) involved the lowest proportion of
transfer for <v>-/b/ in the orthography-learning & production condition.
Kruskal-Wallis tests were conducted to investigate the effect of position within the triplet at
the time of learning and production on individual grapheme-to-phoneme correspondences in the
ortho-learning & production, ortho-learning, and ortho-production conditions and are
summarized in Table 4.8. Although primacy and recency patterns resulted in a lower mean
proportion transfer for certain contexts, Kruskal-Wallis tests showed that these effects were not
108
significant for any of the grapheme-to-phoneme correspondences in any auditory-orthographic
condition. The only grapheme-to-phoneme correspondence for which the factor ‘recency effect’
approached significance in one of the auditory-orthographic conditions (ortho-production
condition) was <d>-/ð/. The position that exhibited the lowest mean proportion transfer for this
grapheme-to-phoneme correspondence was also (3*1).
Table 4.8
Kruskal-Wallis Test Results for Effect of Position within Triplet on Mean Proportion Transfer
by Spanish Grapheme-to-sound Correspondence by Condition
Spanish
Condition
df
p
χ2value
grapheme-tophoneme
correspondence
<v>-[b]
Ortho-learning & production
4.00
4
.406
Ortho-learning
2.96
4
.564
Ortho-production
3.23
4
.520
<d>-[ð]
Ortho-learning & production
4.47
4
.346
Ortho-learning
2.18
4
.702
Ortho-production
9.21
4
.056
<z>-[s]
Ortho-learning & production
1.847
4
.764
Ortho-learning
4.06
4
.398
Ortho-production
6.12
4
.190
<h>-[Ø] #
Ortho-learning & production
.253
4
.993
Ortho-learning
1.96
4
.743
Ortho-production
3.54
4
.472
<h>-[Ø] VCV
Ortho-learning & production
.415
4
.981
Ortho-learning
3.97
4
.410
Ortho-production
4.88
4
.300
<ll>-[j]
Ortho-learning & production
.415
4
.981
Ortho-learning
3.97
4
.410
Ortho-production
4.88
4
.300
In sum, the following hierarchy was established in terms of the frequency of a position
exhibiting the lowest mean proportion transfer: position (3*1) especially in the ortho-production
condition, followed by (1*3), (2*2), and (1*1). In other words, the most frequent pattern noted
was a recency effect in the ortho-production condition. However, the factor ‘primacy recency
effect’ was not significant for any of the grapheme-to-phoneme correspondences and only
109
approached significance for <d>-/ð/. In this section, I have reported the results for the effect of
primacy and recency on orthography-induced transfer at the list level. I now turn to the effect of
primacy within the word.
4.1.2.6 Effect of primacy within the word
In section 4.1.2.5, I reported the results for the effect of position within the triplet. In this
section, I will provide the results for the effect of position within the word for the orthographic
conditions. Hypothesis (2ciii) predicted that primacy effects (e.g., Brown & McNeill, 1966;
Horowitz et al., 1968; Gupta, 2005) would exert an influence on the mean proportion transfer at
the word level. Specifically, it was hypothesized that word-initial position would exhibit a lower
proportion of transfer in the case of <ll>-/j/ for which stimuli included both initial and
intervocalic positions and in which English and Spanish grapheme-to-phoneme correspondences
remain the same across positions. Table 4.9 summarizes the mean proportion transfer scores and
standard deviation for <ll>-/j/ in initial and intervocalic positions. There was a lower mean
proportion transfer in word-initial position than in intervocalic position in all three orthographic
conditions. A Mann-Whitney test was conducted to see if the differences were significant (Table
4.10): the factor ‘position in the word’ was significant in ortho-learning & production and ortholearning conditions but only approached significance (p=.063) in the ortho-production
condition.
110
Table 4.9
Mean Proportion Scores and Standard Deviations for <ll>-/j/ by Position by Condition
Position
Word-initial
Word-medial
Condition
M
SD
M
SD
Ortho-learning & production
.00
.05
.14
.27
Ortho-learning
.10
.30
.41
.41
Ortho-production
.03
.09
.11
.34
Table 4.10
Mann-Whitney Test Results for <ll>-/j/ by Position by Condition
Auditory-orthographic condition
Ortho-learning & production
Ortho-learning
Ortho-production
U
905.50
815.50
66.50
z
-3.45
-3.15
-1.86
p
.001
.002
.063
In analyzing the effect of position on the proportion of transfer with <ll>-/j/ stimuli, it was
also noted that the error patterns for these stimuli differed from the other grapheme-to-phoneme
correspondences. That is, whereas for all other grapheme-to-phoneme correspondences learners
either produced the target sound, substituted the English corresponding sound for the shared
grapheme in Spanish and English, produced another sound or deleted the target sound all
together, in the case of stimuli containing <ll>-/j/, learners also produced combinations of the
L1 and TL corresponding sounds (‘blends’) for the shared grapheme <ll>. For example, for
<pollero>-/pojeɾo/, learners productions included [poljeɾo] and [pojleɾo] and for <llanura>[januɾa], they produced [ljanuɾa] or [jlanuɾa]. Instances of blending were considered transfer
errors because they still involved substitution of the L1 phoneme Whereas transfer for all the
other grapheme-to-phoneme correspondences comprised 100% L1 sound substitution, for <ll>/j/ it comprised 64% blending (e.g., /lj, lij/) and 36% L1 substitution (/l/).
In sum, most of the evidence was in support of a primacy effect at the word level; wordinitial position resulted in a significantly lower mean proportion transfer than intervocalic
111
position in the ortho-learning & production as well as in the ortho-learning conditions. In the
ortho-production condition, on the other hand, the difference between word-initial and
intervocalic positions only approached significance. I now turn to the effect of repetition on
orthography-induced transfer.
4.1.2.7 Effect of round/repetition
Section 4.1.2.6 reported the effect of position, specifically primacy effects, within the word. In
this section, I will provide the results for the effect of ‘round’ on the mean proportion transfer
across orthographic conditions. Hypothesis (2ciii) predicted that repetition effects (e.g., Saragai
et al., 1978; Horst et al.1998; Rott, 2000; Waring &Takaki, 2003; Webb, 2007) would decrease
the proportion of orthography-induced transfer in the orthographic conditions. Specifically, it
was predicted that the proportion of orthography-induced transfer would decrease as the number
of rounds increased. Table 4.11 summarizes the results for the effect of the number of rounds on
the mean proportion transfer. As shown, the mean proportion transfer slightly decreased in the
ortho-learning & production from round 1 to round 2 to round 3. This was also true for the
ortho-production condition. In the ortho-learning condition, however, the proportion of transfer
increased in round 2 in comparison with round 1 and then decreased in round 3, still remaining
higher than in round 1. A Kruskal-Wallis test revealed that the factor rounds was significant in
the ortho-learning & production (χ2(df = 2,) = 6.93, p = .031) and ortho-production conditions
(χ2(df = 2) = 12.13, p = .002) but not in the ortho-learning condition 2 (χ2(df = 2) = 1.07, p =
.585).
112
Table 4.11
Cross Tabulations: Mean Proportion Transfer Scores for Round by Condition
Condition
Ortho-learning & production
Round
1
2
3
1
2
3
1
2
3
Ortho-learning
Ortho-production
M
.58
.55
.49
.50
.54
.53
.50
.42
.39
A Pearson chi-square test was conducted to determine if all rounds were significantly
different from one another in the ortho-learning & production, ortho-learning, and orthoproduction conditions. As per Table 4.12, only the following pair-wise comparisons were
significantly different: rounds 1 and 3 in the ortho-learning & production condition; rounds 1 &
2 in the ortho-production condition 3; and rounds 1 & 3 in the ortho-production condition.
Table 4.12
Pearson Chi-square Results: Rounds by Condition
Condition
Ortho-learning & production
Ortho-learning
Ortho-production
Rounds
1& 2
1&3
2&3
1&2
1&3
2&3
1&2
1&3
2&3
χ2
1.06
6.82
2.50
.980
.308
.229
5.32
11.30
1.109
df
1
1
1
1
1
1
1
1
1
p
3.30
.009
.114
.322
.579
.632
.021
.001
2.29
In sum, the factor ‘round’ was significant in the ortho-learning & production condition with
differences between rounds 1 & 3 as well as in the ortho-production condition with differences
between rounds 1 & 2 and rounds 1 & 3. In the ortho-learning condition, although the factor
‘round’ proved to be non-significant, the proportion of transfer increased in round 2 and then
113
decreased in Round 3 resulting in a higher proportion of transfer than in round 1. I now turn to
the results with respect to the effect of round/repetition on individual grapheme-to-phoneme
correspondences.
4.1.2.8 Effect of round/repetition on individual grapheme-tophoneme correspondence
The results for the effect of round on mean proportion scores and standard deviations for each
grapheme-to-phoneme correspondence that resulted in transfer are summarized in Table 4.12.
The effect of round on individual grapheme-to-phoneme correspondences differed from the
overall effect of round reported in Section 4.1.2.8. As shown in Table 4.13, a number of patterns
emerged. There was only one instance (<h>-/Ø/ #) in which a decrease in the mean proportion
transfer was associated with an increase in the number of rounds in every orthographic
condition. In addition, with the exception of <h>-/Ø/ # and <h>-/Ø/ VCV, in the ortho-learning
condition, there was always an increase in the mean proportion transfer in round 2. Moreover, as
seen with primacy and recency effects, the factor ‘round’ did not have a uniform effect on
individual grapheme-to-phoneme correspondences. For example, whereas for <d>-/ð/, the mean
proportion transfer decreased in round 2 and did not change in round 3, for <z>-/s/, the mean
proportion transfer increased slightly in round 2 and then decreased in round 3.
114
Table 4.13
Cross Tabulations: Mean Proportion Transfer Scores for the Effect of Round for Spanish
Grapheme-to-phoneme Correspondence by Condition
Spanish
Grapheme-to-phoneme
correspondence
<v>-/b/
Condition
Round
M
Ortho-learning & production
1
2
3
1
2
3
1
2
3
1
2
3
1
2
3
1
2
3
1
2
3
1
2
3
1
2
3
1
2
3
1
2
3
1
2
3
1
2
1
.98
1
.97
1
.98
.95
.72
.68
.97
.90
.90
.97
.98
.98
.90
.90
.93
.75
.78
.59
.51
.64
.60
.73
.65
.55
.58
.48
.38
.67
.60
.53
.35
.32
.23
.55
.48
Ortho-learning
Ortho-production
<d>-/ð/
Ortho-learning & production
Ortho-learning
Ortho-production
<z>-/s/
Ortho-learning & production
Ortho-learning
Ortho-production
<h>-/Ø/ #
Ortho-learning & production
Ortho-learning
Ortho-production
<h>-/Ø/ VCV
Ortho-learning & production
(continued)
115
Table 4.13
Cross Tabulations: Mean Proportion Transfer Scores for the Effect of Round for Spanish
Grapheme-to-phoneme Correspondence by Condition (continued)
Spanish
Grapheme-to-phoneme
correspondence
Condition
Ortho-learning
Ortho-production
<l>/j/
Ortho-learning & production
Ortho-learning
Ortho-production
Round
M
3
1
2
3
1
2
3
1
2
3
1
2
3
1
2
3
.47
.13
.13
.17
.25
.16
.15
.11
.07
.04
.18
.20
.18
.13
.06
.03
Kruskal-Wallis tests were performed to investigate the effect of round by individual
grapheme-to-phoneme correspondences on the mean proportion transfer (Table 4.14). Whereas
in Section 5.3.4.1, it was shown that the factor ‘round’ was significant in the ortho-learning &
production and ortho-production conditions, this was not the case for any of the grapheme-tophoneme correspondence that resulted in transfer in the ortho-learning & production and orthoproduction conditions. As indicated in Table 4.16, the factor ‘round’ was only significant in the
ortho-production condition and only for <v>-/b/ and <ll>-/j/.
116
Table 4.14
Kruskal-Wallis Test Results for the Effect of Round on Spanish Grapheme-to-phoneme
Correspondences by Condition
Spanish Grapheme-to-phoneme
correspondence
<v>-/b/
<d>-/ð/
<z>-/s/
<h>-/Ø/ #
<h>-/Ø/ VCV
<ll>-/j/
Condition
Ortho-learning & production
Ortho-learning
Ortho-production
Ortho-learning & production
Ortho-learning
Ortho-production
Ortho-learning & production
Ortho-learning
Ortho-production
Ortho-learning & production
Ortho-learning
Ortho-production
Ortho-learning & production
Ortho-learning
Ortho-production
Ortho-learning & production
Ortho-learning
Ortho-production
χ2
1.98
1.17
14.28
2.48
.097
.573
5.61
1.56
4.41
4.80
2.08
2.06
.796
.412
1.64
4.35
.217
9.07
df
2
2
2
2
2
2
2
2
2
2
2
2
2
2
2
2
2
2
p
.372
.558
.001
.290
.953
.751
.061
.459
.110
.090
.354
.356
.672
.814
.440
.114
.897
.011
Pearson chi-square tests were conducted to investigate the between-round effects on the
mean proportion transfer (Table 4.15). These post-hoc tests were conducted at a more
conservative level of p = .01 to adjust for the number of tests performed. Table 4.14 shows that,
in the ortho-production condition, significant differences were found only between rounds 1 & 2
and rounds 2 & 3 for <v>-/b/, and between rounds 1 & 3 for <ll>-/j/.
117
Table 4.15
Pearson Chi-square Results: Pair-wise Comparisons of Rounds by Spanish Grapheme-tophoneme Correspondence by Condition
Spanish Grapheme- Condition
Round
χ2value df
to-Phoneme
Correspondences
<v>-/b/
Ortho-learning & production
1&2
.975
1
1&3
2&3
1.01
1
Ortho-learning
1&2
1.294
1
1&3
.077
1
2&3
.877
1
Ortho-production
1&2
11.24
1
1&3
13.65
2&3
.159
1
<d>-/ð/
Ortho-learning & production
1&2
2.14
1
1&3
2.14
1
2&3
.00
1
Ortho-learning
1&2
.052
1
1&3
.089
1
2&3
.005
1
Ortho-production
1&2
.001
1
1&3
.436
1
2&3
.474
1
<z>-/s/
Ortho-learning & production
1&2
.187
1
1&3
3.10
1
2&3
4.76
1
Ortho-learning
1&2
1.46
1
1&3
.82
1
2&3
.106
1
Ortho-production
1&2
.977
1
1&3
4.38
1
2&3
1.25
1
<h>-/Ø/ #
Ortho-learning & production
1&2
1.20
1
1&3
4.80
1
2&3
1.22
1
Ortho-learning
1&2
.565
1
1&3
2.07
1
2&3
.469
1
Ortho-production
1&2
.150
1
(continued)
p
.323
.315
.255
.781
.349
.001
.000
.690
.143
.143
1
.820
.765
.944
.976
.509
.491
.665
.078
.029
.277
.365
.745
.323
.365
.264
.272
.028
.269
.452
.150
.493
.699
118
Table 4.15
Pearson Chi-square Results: Pair-wise Comparisons of Rounds by Spanish Grapheme-tophoneme Correspondence by Condition (continued)
Spanish Grapheme- Condition
Round
χ2value df
to-Phoneme
Correspondences
1&3
1.98
1
2&3
1.04
1
<h>-/Ø/ VCV
Ortho-learning & production
1&2
.445
1
1&3
.722
1
2&3
.032
1
Ortho-learning
1&2
.002
1
1&3
.294
1
2&3
.284
1
Ortho-production
1&2
1.00
1
1&3
1.32
1
2&3
.017
1
<ll>-/j/
Ortho-learning & production
1&2
.802
1
1&3
4.38
1
2&3
1.52
1
Ortho-learning
1&2
.122
1
1&3
.005
1
2&3
.193
1
Ortho-production
1&2
3.52
1
1&3
7.71
1
2&3
1.04
1
p
.160
.307
.505
.396
.859
.968
.588
.594
.316
.251
.897
.371
.036
.217
.727
.944
.661
.061
.005
.307
In sum, the factor ‘round’ did not have an equal effect on all grapheme-to-phoneme
correspondences and, thus, the predictions were not all confirmed. Specifically, there appears to
be an interaction between the rounds and condition. That is, the factor ‘round’ was only
significant for <ll>-/j/, where round 1 differed significantly from round 3 and, for <v>-/b/,
where round 1 significantly differed from rounds 2 and 3 in the ortho-production condition.
In the preceding sections, I presented the results for the picture naming task by collapsing
across participants and provided the results concerning the effect of auditory-orthographic
condition
at
learning
and
production,
inconsistency between
grapheme-to-phoneme
correspondences in the TL and L1, and the phenomena that characterize PM working, namely
119
primacy and recency effects at the list level and primacy effects at the word level as well as
repetition effects on the mean proportion orthography-induced transfer. In this section and in
Sections 4.1.2.5-4.1.2.7, I reported the effect of phenomena that characterize PM, namely
primacy and recency effect and round/repetition, on the proportion of orthography-induced
transfer. Overall, the results revealed a clear effect for the presence of orthography at
learning/and or production with differences between the auditory-orthographic conditions. The
results also clearly showed that every grapheme-to-phoneme correspondence examined in this
study resulted in transfer, albeit the proportion of transfer differed between individual
grapheme-to-phoneme correspondences. In addition, there was a weak recency pattern at the list
level, some evidence for primacy effects at the word level and some evidence for the effect of
repetition, on orthography-induced transfer. Moreover, each individual grapheme-to-phoneme
correspondence was affected by (auditory-orthographic) condition, primacy and recency effects
and repetition, to a different extent. I now turn to the results regarding the hypothesized
correlation between individual PM capacity and proportion of orthography-induced transfer.
4.2. Data analysis and results: PM task and individual
variation in proportion of transfer
The goal of the present section is to describe the data analysis and present the results of the
Farsi-based non-word repetition PM task for each participant in order to verify the hypothesized
negative correlation between PM and mean proportion transfer for individual learners
(Hypothesis (2c)) in the orthographic conditions (ortho-learning & production, ortho-learning,
and ortho-production).
120
The remainder of this section is structured as follows. In Section 4.2.1, I describe the data
analysis and scoring methodology; in Section 4.2.2, I report the results of the PM task and
examine any potential correlations with individual variation in the mean proportion transfer.
4.2.1 Data analysis
As previously outlined in the methodology chapter, PM capacity was tested with a Farsi-based
non-word repetition task in which learners had to listen to a list of words, one at a time, then
repeat each word immediately. The 28 words were assigned meanings (e.g., [næzæde] ‘not hit’
and [zomoɾodi] ‘emerald color’) and varied between 3-9 syllables in length (4 of each length).
Different methods for scoring PM capacity have been employed in previous research. For
example, in Archibald and Gathercole (2006), recall accuracy was scored at the syllable and
phoneme level using a strict serial order criterion according to which a phoneme was scored as
correct if the right phoneme was produced in the order presented and a unit was only scored as
correct if it contained all of the phonemes in the order presented and the syllable was recalled in
its original position within the sequence. In Service and Kohonen (1995), Dufva and Voeten
(1999) and French (2004), on the other hand, responses were scored according to the number of
words that were correctly repeated with no phoneme replacements, omissions, or additions. In
further contrast, in Gathercole, Willis, Emslie & Baddeley (1992), Gathercole and Adams
(1993, 1994), and Gathercole, Hitch, Service and Martin (1997), the number of entire non-words
that were correctly repeated were scored. To this criterion, Speciale, Ellis and Bywater (2004)
added the number of syllables correctly repeated.
For the present study, following Archibald & Gathercole (2006), recall accuracy scores in
the non-word repetition task were calculated based on phonemes that were produced correctly
121
and in the right sequence without any misplacements/migrations. For example, a misplacement
could consist of a metathesis (e.g., /ʃ/ misplaced with /n/ producing /zibaneʃasi/ for target
/zibaʃenasi/ ‘aesthetics’). Given that the purpose of the picture-naming task was to determine the
effect of grapheme-to-phoneme correspondences on the proportion of transfer in the
pronunciation of particular individual sounds in Spanish, a scoring methodology based on the
number of phonemes produced correctly was deemed more appropriate as opposed to a scoring
methodology based on entire words or syllables produced correctly.
Given that the scores for the PM task were going to be based on the number of phonemes
produced correctly and in the right place, a maximum PM score of 336 was established for each
person. This was based on the calculation of the possible total number of correct phonemes (4
words each of 8, 10, 12, 14, 16 and 18 phonemes). All together 10080 phonemes were analyzed.
Individual PM scores (phonemes produced correctly and in the target position) were then
correlated with the mean proportion transfer values. The three orthographic conditions (ortholearning & production, ortho-learning, and ortho-production) were collapsed.
2 native speakers of Farsi, the author and another individual with training in linguistics,
transcribed the results. There was 97% inter-transcriber agreement. That is, all together, there
were only disagreements concerning 302 phonemes. The disagreements were resolved by a third
native speaker of Farsi. Variations in production were allowed for as long as phonemic
boundaries were not crossed. For example, an English alveolar stop was scored as correct even
though the target was dental. I now turn to the results regarding the hypothesized correlation
between individual variation in PM and proportion transfer.
122
4.2.2 Results: PM scores and individual variation in
orthography-induced transfer
The Farsi-based non-word repetition PM scores range from 151 to 301 with the mean proportion
of transfer value obtained from the picture-naming task of 0.5 and a standard deviation of 0.15.
Variation in mean proportion transfer is shown against PM scores in Table 4.16.
Table 4.16
Individual PM and Mean Proportion Transfer Scores in Orthographic Conditions
Participant
PM score
Mean proportion transfer
1
151
.53
2
158
.32
3
172
.61
4
190
.47
5
193
.38
6
197
.55
7
200
.33
8
202
.73
9
206
.72
10
213
.33
11
220
.80
12
220
.59
13
221
.60
14
222
.67
15
223
.53
16
232
.45
17
239
.44
18
242
.76
19
247
.50
20
247
.51
21
247
.37
22
251
.31
23
256
.55
24
257
.14
25
260
.60
(continued)
123
Table 4.16
Individual PM and Mean Proportion Transfer Scores in Orthographic Conditions (continued)
Participant
PM score
Mean proportion transfer
27
267
.50
28
289
.50
29
291
.44
30
301
.43
The results showed that transfer values of 0.70 or higher were observed in participants with
PM scores below 242. A Pearson’s correlation analysis performed on the factors PM score and
mean proportion transfer yielded a sample correlation of -0.15 (single-tailed p = 0.22). This
result is not statistically significant at the 0.05 alpha level which is contrary to the prediction
that PM scores and transfer values are negatively correlated. The large variability in the data is
one of the reasons that a strong correlation effect is not detected. The 95% confidence interval
of the observed correlation is from -0.48 to 0.23 indicating that while the effect size includes
zero, the data supports population correlation effect as large as -0.48. It is not very plausible
though that the population effect is larger than 0.48.
When a statistical test is non-significant, it could be that the null hypothesis of no effect
(Ho: r=0) is supported; or it could be that the null hypothesis is not confirmed but the test
cannot detect the effect due to small effect size, small sample size and large variability in the
study (known as the Type II error). To discriminate between these two scenarios, a statistical
power analysis on the observed effect was conducted. The power of a test is the probability of
detecting an effect, if such effect indeed exists.
It is a function of the sample size, the
population effect size and the significance criterion. Statistical power is generally considered
adequate if its value is 0.80 or above (Cohen, 1988). With a power of 0.8, it is 4 times more
likely to reject the null hypothesis (when in fact the null hypothesis is not verified) than not
rejecting it. The analysis generated an observed power value of 0.2 (0.05 alpha level, single-
124
tailed test). The low power indicates that support of the null hypothesis of no correlation effect
is weak. In other words, there is almost an 80% chance of committing a Type II error. To be
able to detect the effect size of 0.15 at an alpha level of 0.05 (single-tailed test) with a power of
0.8, the sample size would have to be 274. If an effect size larger than the observed effect is
observed, a power analysis would have to follow. The power is 0.89 if the population effect size
is 0.5 (N = 30, alpha = 0.05, single-tailed test). However when the population effect size
decreases to 0.4, 0.3, and 0.2, the power drops to 0.72, 0.5, and 0.28 respectively. The power
analysis conducted here showed that this study has sufficient power (a good chance to produce a
significant result) to detect large correlation effects but lacks the power to detect small
or medium effects.
In sum, the results of the correlation analysis are inconclusive. It cannot be suggested with
confidence that the null hypothesis of no correlation effect between PM scores and transfer
values is verified. The wide span of the 95% confidence interval (-0.48 to 0.23) of the observed
correlation indicates a likelihood that a real population effect exists and that it can be far from
the left of zero. Furthermore, the power analysis shows that the study has adequate power to
detect a large effect but not a very small effect.
All in all, the results do not show whether the lack of a correlation between individual PM
scores and orthography-induced transfer are truly representative of the population at large.
However, it must be noted that even though the results with respect to a correlation between
individual PM and proportion transfer are inconclusive, nonetheless other PM-related effects
were previously noted in Sections 4.1.2.4-4.1.2.8.; there was a significant primacy effect at the
word level, a weak recency pattern at the list level and some effects of repetition. I now turn to
the summary of the results prior to moving on to the discussion in chapter 5.
125
4.3 Summary of results
As shown by the findings in this chapter, orthography does indeed promote transfer in
production leading to non-target-like behavior. In addition, it is clear that a number of factors
affected the proportion of transfer leading to non-target-like productions in the production of the
novice English-speaking learners of Spanish tested. The factors that shaped proportion of
transfer were as follows: (1) auditory-orthographic condition at learning & production; (2)
grapheme-to-phoneme inconsistency (3) primacy effect at the word level position within word
(primacy effects); and (4) round. Given the results in this chapter, the effect of these factors can
be ranked as per the following hierarchy with the effect decreasing from left to right:
auditory-orthographic condition and grapheme-to-phoneme inconsistency between the TL
and L1 > phenomena characterizing PM working (primacy and recency and repetition)
With respect to the auditory-orthographic condition at learning and production, when all
grapheme-to-phoneme correspondences were collapsed, all three orthographic conditions
(ortho-learning & production, ortho-learning, and ortho-production) exhibited a higher
proportion of transfer than the auditory only condition. In fact, there was very little transfer in
the auditory only condition at all (e.g., only for <v>-/b/ and <d>-/ð/). Moreover, in keeping with
the hypotheses, the ortho-production condition led to a lower proportion of transfer than the
ortho-learning & production and ortho-learning conditions. However, contrary to the
hypotheses, the proportion of transfer in the ortho-learning & production condition was not
significantly different from that of the ortho-learning condition.
When the effect of auditory-orthographic condition was analyzed for each grapheme-tophoneme correspondence separately, more nuanced patterns were observed for some of the
126
grapheme-to-phoneme correspondences. For example, while for 4 out of 6 cases (<v>-/b/, <d>/ð/, <z>-/s/, and <h>-/Ø/ #), there were no significant differences between the ortho-learning &
production and ortho-learning conditions, for one grapheme-to-phoneme correspondence (e.g.,
<h>-/Ø/ VCV), the ortho-learning & production condition led to a significantly higher
proportion of transfer than the ortho-learning condition whereas for another grapheme-tophoneme correspondence (e.g.,<ll>-/j/), the ortho-learning & production condition led to a
significantly lower proportion of transfer than ortho-learning condition. As for the prediction
that the ortho-learning & production condition would lead to a lower proportion of transfer than
the ortho-production condition, this was true in half of the cases, namely for <v>-/b/, <h>-/Ø/ #,
and <h>-/Ø/ VCV. The ortho-production condition also led to a lower proportion of transfer
than the ortho-learning condition with <v>-/b/, <h>-/Ø/ #, and <ll>-/j/.
The factor ‘inconsistency between L1 and TL grapheme-to-phoneme correspondences’ had
a significant effect on the proportion of transfer for each of the orthographic conditions. Those
grapheme-to-phoneme correspondences that differed between Spanish and English resulted in
transfer in the orthographic conditions whereas those that were the same, with the exception of
<h>-/Ø/ VCV, did not do so. Moreover, the proportion of transfer differed between the
grapheme-to-phoneme correspondences that resulted in transfer. Furthermore, contrary to the
predictions, <v>-/b/ and <d>-/ð/ resulted in some transfer in the auditory only condition.
In addition, in analyzing the tokens whose TL and L1 grapheme-to-phoneme
correspondences were different, a different pattern of transfer for the tokens with <ll>-/j/ was
observed. Specifically, errors consisting of a combination of the TL and L1 phonemes (e.g., /lj/),
referred to as ‘blends’, were noted. This pattern was not found for any other grapheme-to-
127
phoneme correspondence. In all the other cases, transfer consisted of the substitution of the L1
sound.
With respect to the PM- related factors, results were mixed. For example, the factor
‘primacy and recency effects’ was not significant when all grapheme-to-phoneme
correspondences were collapsed. At the list level, there was no evidence to support either
primacy or recency effects for transfer when all the grapheme-to-phoneme correspondences
were collapsed in each orthographic condition. When the effect of this factor was analyzed for
each grapheme-to-phoneme correspondence separately, the same results were obtained with the
difference that results approached significance for <d>-/ð/, where position (3*1) led to the
lowest proportion of transfer. Although the factor ‘position’ was not significant for any of the
grapheme-to-phoneme correspondences, position (3*1), especially in the ortho-production
condition, exhibited the lowest proportion of transfer most frequently. On the other hand, with
respect to primacy effects at the word level, the results were in accordance with the predictions
for the most part. Specifically, word-initially, there was a lower proportion of transfer than
word-medially with all three orthographic conditions. Furthermore, in two conditions (ortholearning & production and ortho-learning), the differences were significant and in the orthoproduction condition, the differences approached significance. In all, the majority of evidence
supported a primacy effect.
Regarding the other PM-related factor, namely ‘round’, when all grapheme-to-phoneme
correspondences that resulted in transfer were collapsed, this factor was significant in the ortholearning & production condition, where the differences were significant between rounds 1 & 3,
as well as in the ortho-production condition, where the differences were significant between
rounds 1 & 2 as well as between rounds 1 & 3. In the ortho-learning condition, although the
128
number of rounds proved to be insignificant, the proportion of transfer increased in round 2 and
then decreased in round 3 resulting in a higher proportion of transfer than round 1. The results
for the effect of the factor ‘round’ were somewhat different when grapheme-to-phoneme
correspondences were analyzed individually. That is, round was only significant in the orthoproduction condition, albeit only for <ll>-/j/ and <v>-/b/, where significant differences between
rounds 1 and 3 were for the former and significant differences between rounds 1 & 3 and rounds
1 & 2 were found for the latter.
While the results with respect to the universal PM phenomena namely, primacy and recency
effects and round/repetition were mixed, the results with respect to the hypothesized correlation
between individual variation in PM and orthography-induced transfer were not characterized by
a significant correlation. Further power analysis suggested that a larger sample is needed to
verify the results with respect to the effect of individual differences in PM scores on
orthography-induced transfer leading to non-target-like productions.
In this chapter, I have examined the results for both the picture-naming task and PM task in
detail. All in all, there was a clear effect of orthography on L1-based transfer leading to nontarget-like productions. In addition the results showed strong effects for auditory-orthographic
condition and inconsistency between the TL and L1 grapheme-to-phoneme correspondences and
mixed effects for factors related to PM. These results will be considered in light of previous
research on the effect of orthography on phonological transfer in Chapter 5. In Chapter 5, I will
also highlight the contribution of the findings in present work and suggest future directions.
129
Chapter five
Discussion and conclusions
In Chapter 4, I presented the results which showed how exposure to orthography at the very
early stages in the acquisition process of an L2 can induce non-target-like productions.
Specifically, I presented the results with regards to the factors hypothesized to bring about L1based phonological transfer. These factors included auditory-orthographic condition at learning
and production, inconsistency between grapheme-to-phoneme correspondences in the TL and
L1, and PM. The latter was analyzed with respect to the general aspects of PM as well as
between-individual variation. In this chapter, I will discuss the findings in light of previous
research, highlight some of the contributions and propose further research.
The structure of the remainder of the chapter is as follows. In Section 5.1 I address the
issue of the effect of auditory-orthographic condition at learning and at production. In Section,
5.2, I remark upon the results regarding the effect of type of grapheme-to-phoneme
correspondences. In Section, 5.3, the effect of PM on orthography-induced transfer is
considered. Specifically, the effect of the following in relation to orthography-induced transfer
are debated: (i) different characteristics of PM, namely primacy and recency effects both at the
list and word levels as well as repetition, on controlling the proportion of orthography-induced
transfer and (ii) individual PM capacity. Finally, in Section 5.4, I will conclude by highlighting
the contributions and implications of this study for our understanding of the role of orthography
in transfer in L2 acquisition as well as potential pedagogical explanations and propose some
future studies.
130
5.1 Effect of auditory-orthographic condition
It was hypothesized that orthography would promote L1-based transfer leading to non-targetlike production. In addition, it was also predicted that auditory-orthographic condition would
shape the proportion of transfer as per the following hierarchy: ortho-learning & production >
ortho-learning > ortho-production > auditory only. The hypotheses with respect to auditoryorthographic condition were born out for the most part, when grapheme-to-phoneme
correspondences that resulted in transfer were collapsed. Indeed, each orthographic condition
exhibited a higher proportion of transfer than the auditory only condition and the mean
proportion transfer was higher in each of the ortho-learning & production and ortho-learning
conditions in comparison with the ortho-production condition. The only aspect of this particular
hypothesis that was not borne out was that there were no significant differences between ortholearning & ortho-production, contrary to the above hierarchy. In other words, the presence of
orthography at learning appears to exert the same influence regardless of its presence or absence
at production. This could be because the presence of orthography at learning actually interferes
with the formation of the learners’ representation of the TL sounds at the initial stages of
learning Spanish.
The finding that presence of orthography at learning shapes phonological transfer
leading to non-target-like transfer in production is consistent with the previous findings. For
example, as was discussed in Chapter 2, Erdener and Burnham (2005), using a repetition task,
also found that the presence of incongruent orthographic input at the time of learning led to
significantly higher error rates in comparison with the auditory only condition for Turkishspeaking learners of Spanish. As in this study, the presence of graphemes that correspond to two
different sounds in the learners’ TL and L1 at the time of learning promoted transfer of the L1
131
structures resulting in non-target-like L2 production. For example, when Turkish learners were
presented with the grapheme <j> in training simultaneously with the auditory Spanish stimulus
[x], at production, instead of producing [x], they erroneously substituted their L1 structure /ʒ/
which corresponds to the grapheme <x> in Turkish. In another study, Young-Scholten et al.
(1999) showed that the inclusion of orthographic input at the time of learning affected English
learners’ repair strategies in dealing with illicit Polish consonant clusters. English learners who
were trained with orthography at learning and/or at testing were more likely to use epenthesis
and less likely to use deletion in comparison with the learners that were not exposed to
orthography at all. Her results also showed that Japanese-speaking learners also exhibited a
lower rate of deletion when they were not exposed to orthography. That exposure to
orthography is a potential source of non-target-like productions was also shown by Neufeld
(1978). This earlier study showed that training adult English speakers with auditory input alone
via 18 lessons presented over a period of four weeks led to some learners passing as native
speakers when judged on the acquisition of some of the phonetic and prosodic characteristics of
Chinese, Inuktitut, and Japanese. While the present study confirms that the presence of
orthography at the time of learning and/or production can affect learners’ production, it also
makes a new contribution. Specifically, this thesis allows for a more fine grained understanding
of the type of auditory-orthographic input that triggers transfer by showing that the presence of
orthography at learning tends to influence the proportion of transfer to a greater extent than the
presence of orthography at production only.
When examining the effect of auditory-orthographic condition on individual graphemeto-phoneme correspondences in this study, it was observed that each of the orthographic
conditions, namely ortho-learning & production, ortho-learning, and ortho-production, resulted
in a higher proportion of transfer than the auditory-only condition. On the other hand, pair-wise
132
comparisons of auditory-orthographic conditions revealed different results for individual
grapheme-to-phoneme correspondences. For example, whereas the difference between ortholearning & production and ortho-learning was not significant for <v>-/b/, the difference between
ortho-learning & production and ortho-learning were significant for <z>-/s/ and <d>-/ð/.There
are no previous studies that have reported differing degrees of the effect of the presence of
orthography at learning and/or production on individual grapheme-to-phoneme correspondences
and it is not clear why the presence of orthography at learning and/or production should affect
different grapheme-to-phoneme correspondences differently.
In sum, the findings in this study, consistent with previous studies, suggest that exposure
to orthography at learning and/or production shapes transfer and hinders target-like L2
production. The present study also adds to our understanding of the role of orthography in
shaping transfer by providing the following hierarchy in terms of the degree of influence of
auditory-orthographic condition: ortho-learning & production ~ ortho-learning > orthoproduction > auditory only. This study also shows that whereas the latter hierarchy held when
all grapheme-to-phoneme mappings were collapsed, it did not always hold when such
correspondences were analyzed individually.
5.2
Effect
of
grapheme-to-phoneme
inconsistency
between English and Spanish
Another aim of the thesis was to determine whether grapheme-to-phoneme correspondences that
differed between English and Spanish as opposed to same grapheme-to-phoneme
correspondences would trigger transfer leading to non-target-like production of Spanish phones
by
English-speaking
learners.
The
prediction
that
identical
grapheme-to-phoneme
133
correspondences would not result in any transfer leading to non-target-like production was
borne out in all cases except with <h>-/Ø/ VCV. Intervocalic <h> in unstressed position is silent
in Spanish (e.g., <ahotar>-/aotaɾ/) and it was also assumed to be silent in English words (e.g.,
<vehicle>-[vi:jəkəl]). Therefore, the realization of this grapheme as [h] in this position in L2
Spanish was not expected. In order to understand this phenomenon in the learner data, a
frequency analysis of words with <h> was conducted using the Canadian Oxford Dictionary.
Table 5.1 summarizes the results for the lexical frequency for silent <h> by position within the
word in English.
Table 5.1
Lexical Frequency for Silent <h> by Position within the Word in English
Position within word
Silent realization
Total
% Silent realization
Word onset
- Absolute word initial
19
2865
.006
- Post-consonantal
418
418
100
Total
437
3283
.13
Intervocalic
- Stressed
0
37
0
- Unstressed
6
47
13
Total
6
84
7
Coda
58
58
100
Overall total
501
3425
15
The results in Table 5.1 show that lexical items with a silent realization of <h> (e.g.,
<annihilate>, [əˈniəlait]) are less frequent both in unstressed intervocalic position (13%) as well
as when the results are collapsed across position within the word (15%). In other words, a
produced <h> (e.g., <beehive> [ˈbihaiv]) is more commonly found in the English lexicon than a
silent one. Previously, Ranbom and Connine (2011) have claimed that letters that are both
pronounced and have a silent counter-part (e.g. <t> in -<mortgage>, [mɔɹgədʒ] versus
<vortex>-[vɔɹtɛks]), may be represented phonologically based on the more frequent form (e.g.,
the pronounced forms); the infrequent forms are simply lexicalized. Given the infrequent silent
134
realization of <h> in English words demonstrated in Table 5.2, it may also be the case that the
grapheme <h> is phonologically represented by its pronounced counterpart, and the silent cases
are lexicalized. The assumption that <h> is mentally represented as /h/ in the lexicon would then
explain the error patterns of the learners in this study with respect to the pronunciation of silent
<h> in the unstressed intervocalic position. All in all, it appears that lexical frequency of a
particular grapheme-to-phoneme correspondence, in cases where there is variability in the
realization of a particular grapheme in the learners’ L1, may affect the proportion of transfer
leading to non-target-like productions.
Another prediction in this study was that grapheme-to-phoneme correspondences that
differed between Spanish and English but for which the corresponding Spanish sound existed in
English would result in transfer in the orthographic conditions but not in the auditory only
condition. For example, <ll>-/j/ was predicted to result in the production of [l] in the
orthographic conditions but of [j] in the auditory only condition. All of the ‘different’ graphemeto-phoneme correspondences involved transfer leading to non-target-like production in the
orthographic conditions and this was consistent with the previous studies (Young-Scholten,
2000; Erderner & Burnham., 2005). However, two of the grapheme-to-phoneme
correspondences, namely <d>-/ð/ and <v>-/b/, also resulted in transfer in the auditory only
condition (their mean proportion transfer scores were .55 and .13 respectively). The fact that in
the auditory only condition, there were instances in which /ð/ was produced as [d] and [ð] and
/b/ as [v], raises some questions regarding the initial assumption that these are the ‘same’ sounds
and/or are not problematic for English-speaking learners of Spanish. A more detailed phonetic
analyses revealed that /d/ is actually realized as an approximant [δ] and not a fricative in
Spanish (Martínez-Celdrán, 2008), which has not been reported to exist in English. Furthermore,
previous studies have shown that /δ/ is problematic for English learners and is erroneously
135
pronounced as /d/ (Zampini, 1994, 1997, Waltmunson, 2005) and as [ɾ] (Waltmunson, 2005).
With respect to the sound /b/, I am not aware of any previous research that has shown that the
realization or perception of this phoneme (putting aside the issue of correct voice onset timing)
is problematic for English speakers. However, given the results in this study, it is apparent that
this sound does pose some difficulty for English learners. On the other hand, the results with
respect to <v>-/b/ and <d>-/δ/ also showed that, in the orthographic conditions, these two
grapheme-to-phoneme correspondences exhibited a higher mean proportion transfer in
comparison with the auditory only condition. Given these results, the initial predictions that
orthographic inconsistency between the TL grapheme-to-phoneme correspondences results in
transfer when the TL sound already exists in the L1 can be extended to cases where the TL
sound is problematic in the L1 or, in SLM (Flege 1995) terms, is a ‘similar’ sound. In fact, in
the case of ‘similar’ sounds, exposure to orthographic input may induce a higher proportion of
transfer as shown by higher proportions of transfer in the orthographic conditions for <d>-/δ/
and <v>-/b/. Evidence in support of orthography involving a higher proportion of transfer for
the latter two grapheme-to-phoneme correspondences is also found in Zampini (1994). Zampini
(1994) showed that<d>-/δ/ and <v>-/b/ exhibited a lower proportion of transfer in the
conversation task, where the English-speaking learners of Spanish were not exposed to
orthography at production, in comparison with the reading task, in which the learners were
exposed to orthography at production.
In addition to showing that grapheme-to-phoneme correspondences that differ between
English and Spanish trigger phonological transfer, it has been shown that the mean proportion
136
transfer significantly differed between these grapheme-to-phoneme correspondences. Table 5.2
1
summarizes the results for individual grapheme-to-phoneme correspondences.
Table 5.2
Hierarchy of Spanish Grapheme-to-phoneme Correspondences in Accordance to Their
Corresponding Mean Proportion Transfer
Auditory-orthographic Condition Hierarchy of mean proportion transfer for Spanish
grapheme-to-phoneme correspondences in a descending
order
Ortho-learning & production
(1) <v>-/b/ (.99)
(2) <d>-/δ/ (.92)
(3) <z>-/s/ (.69)
(4) <ll>-/j/ (.07)
Ortho-learning
(1) <v>-/b/ (.99) ~ <d>-/δ/ (.98)
(2) <z>-/s/ (.60)
(3) <ll>-/j/ (.21)
Ortho-production
(1) <d>- /δ/ (.90)
(2) <v>-/b/ (.77) ~ <z>-/s/ (.64)
(3) <ll>-/j/ (.09)
Auditory only
(1) <d>-/δ/ (.55)
(2) <v>-/b/ (.13)
The main trend that emerges from Table 5.2 is that <v>-/b/ and <d>-/δ/ resulted in the
two highest mean proportion transfer scores followed by <z>-/s/ and <ll>-/j/. I therefore propose
that the observed difference in the mean proportion transfer between these grapheme-tophoneme correspondences may be due to the difference in the degree of perceptual salience of
the difference between the L1 and the TL sounds that a shared grapheme may correspond to. In
other words, the smaller the phonetic/acoustic difference between the TL and L1 sounds for a
shared grapheme in Spanish and English, the less robust the difference and higher the possibility
of phonological transfer. The claim that the phonetic/acoustic distance between the TL and L1
determines equivalence classification has been previously proposed in SLM with regards to the
1
Table 5.2 does not include <h> because <h> corresponds to an actual phoneme (e.g., /h/) in English but it is silent
in Spanish.
137
mapping of TL sounds that do not exist in the L1. What is new about this proposal is that, in the
presence of inconsistent TL and L1 grapheme-to-phoneme correspondences, the effect of
phonetic distance on equivalence classification not only applies to similar sounds, but also when
the TL sounds are ‘old’ sounds in the L1.
The proposal regarding the differences in the degree of salience between the TL and L1
sounds is intuitively based and it does not provide an instrument for measuring the degree of
salience. For example, /v/ and /b/ sound considerably more similar than /l/ and /j/. Given that
there are no phonetic studies that have measured the phonetic distance between the TL and L1
contrasts examined in this study, using the UCLA phonological segment inventory in
Maddieson (1984), I compared the number of languages that have a contrast that I argue is the
most salient contrast in the present study (e.g., /l/-/j/) with the number of languages that have a
/b/-/v/ contrast, a contrast which I argue is one of the least salient contrasts here. This approach
was based on the assumption that if there are a larger number of languages in which /l/ and /j/
are contrastive in comparison with /v/ and /b/, it would follow that /v/ and /b/ are perceptually
more similar than /l/ and /j/ and would provide indirect evidence for my claim. That is, a lower
number of languages with the /v/-/b/ contrast in comparison with the /l/-/j/ contrast would
provide evidence for the neutralization of the former, possibly due to perceptual similarity,
diachronically. I surveyed a total of 913 languages. Indeed, I found that whereas /l/ and /j/ are
contrastive in 231 (25%) languages, /v/ and /b/ are only contrastive in 51 (5.58%) cases. While,
this finding provides some evidence for the claim regarding the effect of perceptual similarity of
the TL and L1 sounds, in the future, this claim could be further validated with quantitative data
from acoustic and perceptual studies. For example, if a perceptual study showed that the
difference between [v] and [b] is less than the difference between [l] and [j], then this would
138
further support the claim regarding the effect of the acoustic/phonetic distance between the TL
and L1on orthography-induced transfer.
The results in this study have clearly shown that orthographic inconsistency interferes
with the acquisition of TL phonemes, both when the TL phonemes are ‘the same’ and when they
are ‘similar’. Given the findings in this study as well as the evidence for the persistence of some
of these orthography-induced transfer errors in more advanced learners (e.g., Zampini, 1994;
Young-Scholten, 2000), it is therefore proposed that exposure to orthography may interfere with
category formation in the process of L2 learning and lead to the formation of non-target-like
categories. Given that the learners in this study were novice learners, it is unlikely that they
would have developed L2 categories after completing the picture-naming task in this study.
However, because of the multi-modal nature of L2 learners’ input, it is probable that
orthography hinders category formation throughout the L2 learning process and its negative
impact would also be visible in end state grammars. Currently, Best and Tyler (2007) is the only
perceptual model of L2 phonological acquisition which has mentioned the role of orthography
in the acquisition of TL sounds that do not exist in the L1. Given the evidence provided
previously in Young-Scholten (2000), Zampini (1994) and the present study as well as the
reliance on orthographic input in classroom teaching (Erdener & Brunham, 2005), it is crucial
that future models incorporate the role of orthography when formulating their predictions
regarding transfer and category formation. They can do so by addressing the fact that typically,
learners’ input is multi-modal and recognize the role of salience in the process of acquisition.
The results discussed so far have been related to proportion of transfer. I will now
discuss another finding with respect to the type of errors noted in learners’ productions in the
orthographic conditions only, when performing an auditory transcription of the learners’ errors.
139
Specifically, it was noted that whereas in the case of all of the different grapheme-to-phoneme
correspondences, transfer consisted of the substitution of the L1 sound for the TL sound (e.g.,
*/satiko/ for /zatico/), for <ll>-/j/ this was not always the case. The learners’ error types, in
addition to pure L1 substitutions also consisted of a combinations of TL and L1 sounds (e.g.,
/lj/). The fact that learners did not produce any blends when they were exposed to other
grapheme-to-phoneme correspondences may be because whereas /lj/ is a legitimate onset in
English as in the word <million> /miljən/, other combinations such as <bv> and/or <sz> are not
found in English onsets. Previously, it has been reported that presenting learners with
conflicting auditory (e.g., /g/) and facial cues (e.g., /b/) may either lead to the integration of the
TL and L1 sounds (e.g., /bg/) or an in between sound (e.g., /d/) (e.g., McGurk & MacDonald,
1976; Welch &Warren, 1980). This phenomenon is called the McGurk effect. It is possible that
exposing novice learners to inconsistent grapheme-to-phonemes may similarly trigger
perceptual blending in which the conflicting TL and L1 sounds are perceived a single percept. A
perception study would be needed to test this proposal. Specifically, learners would hear words
with [j] and see their written form with <l> and would then be provided with a list of possible
options and would be asked to identify what they heard. A larger percentage of identification of
[lj] would show that blending occurs in perception.
In summary, a number of proposals regarding the factors involved in promoting
orthography-induced transfer and the effect of orthography on L2 production were made in this
section. First, based on the finding regarding the presence of transfer for <h>-/Ø/ VCV, it was
proposed that lexical frequency may be a factor that shapes orthography-induced transfer.
Second, given that the mean proportion transfer differed between grapheme-to-phoneme
correspondences and the nature of the grapheme-to-phoneme correspondences in this study, it
was proposed that the less robust the degree of salience between an L1 and TL phonemes for a
140
shared grapheme, the higher the proportion of transfer leading to non-target-like realizations.
Third, based on the fact that <d>-/δ/ and <b>-/b/ resulted in transfer in the auditory only
condition and resulted in a significantly higher mean proportion of transfer than the auditory
only condition, it was proposed that orthography not only promotes transfer when the TL
phonemes already exist in the L1 but also perpetuates the effect of transfer for phonemes that
may be problematic or ‘similar’ for learners. I now turn to the discussion of the effects of PM.
5.3 PM
Another question in this study was whether PM plays a role in promoting orthography-induced
transfer leading to non-target-like production. It was hypothesized that PM would shape
orthography-induced
transfer
because
(a)
in
learning
new
grapheme-to-phoneme
correspondences, learners would have to rely on PM to store the TL sound that corresponded to
the shared grapheme before they could notice the discrepancy between the TL and L1
grapheme-to-phoneme correspondences and; (b) there is considerable evidence to suggest that
PM is implicated in L2 acquisition learning in general (Service, 1992, Service & Kohonen.,
1995; Cheung, 1996; Dufva & Voeten, 1999; French, 2004; O’Brien et al., 2006 & 2007;
Hummel, 2009), specifically in L2 vocabulary learning which requires the acquisition of new
strings of sounds; (c) the effect of orthography-induced transfer was tested in the context of
vocabulary learning in an L2. Therefore, it was hypothesized that (a) the phenomena
characterizing PM working, namely primacy and recency effects (b) repetition effects would
influence the proportion of orthography-induced transfer and (c) individual differences in PM
would be negatively correlated with the proportion of orthography-induced transfer. The effect
of PM will be considered with respect to its universal characteristics in section 5.1.3.1 and
5.1.3.2 and with respect to individual differences in section 5.1.3.3. The discussion in these
141
sections will show that the results on the effect of PM on orthography-induced transfer are
mixed.
5.3.1 Primacy and recency effects
In this study, the effect of primacy and recency on the proportion of transfer was tested at the
list level and primacy effects were tested at the word level. I will discuss the finding with
respect to the list level first. In this study, based on previous memory literature (Deese &
Kaufman, 1957; Murdock, 1962; Waugh & Norman, 1965; Craik, 1970; Rundus & Atkinson,
1970; Rundus, 1971; Foreit, 1976), it was hypothesized that position within the triplet at the
time of testing and production would affect the proportion of orthography-induced transfer
leading to non-target-like production. Specifically, it was predicted that there would be primacy
and/or recency effects, namely position (1*1) (first at learning and first at production) and (3*1)
(third at learning and first at production) would result in the lowest proportion of transfer in
every orthographic condition. The results suggested that position (3*1) was the position that
most frequently resulted in the lowest mean proportion transfer and most frequently in the orthoproduction only. However, the results only approached significance for one grapheme-tophoneme correspondence, namely <d>-/δ/. There were no significant recency or primacy effects
either for the remaining individual grapheme-to-phoneme correspondences nor when graphemeto-phoneme correspondences were collapsed. The recency pattern found in this study is
consistent with the findings in list recall studies in the memory literature (Deese & Kaufman,
1957; Murdock, 1962; Waugh & Norman., 1965; Craik, 1970; Rundus & Atkinson, 1970;
Rundus, 1971; Foreit, 1976). However, the question remains as to why primacy and recency
effects did not reach significance in this study. Most studies in the memory literature that have
reported primacy and recency effects have used lists longer than seven items (e.g., Deese &
142
Kaufman, 1957; Murdock, 1962). Thus, the lack of significant primacy and recency effects is
most likely due to the word lists consisting of three items in the picture-naming task not having
been long enough to induce these working memory effects on orthography-induced transfer in
production. In other words, the task was too simple in terms of its demand on working memory
for there to be significant serial position effects. The latter hypothesis regarding the role of
shortness of the word lists employed in this study would be confirmed if primacy and/or recency
effects were obtained if this experiment were replicated with a picture-naming task in which the
learners would be presented with a longer list of words at learning and production.
With respect to the working-memory effects at the word level, as mentioned previously,
only primacy effects were tested and only the grapheme-to-phoneme correspondence <ll>-/j/
was considered in relation to orthography-induced transfer in production. In contrast to the
findings regarding serial position effects at the list level, the results at the word level, for the
most part, confirmed a primacy effect. That is, the prediction that word-initial position would
result in a lower proportion of transfer than intervocalic position for <ll>-/j/ was borne out for
the ortho-learning & production and ortho-learning conditions. These results are consistent with
previous studies that have also reported a primacy effect for word level recall (Brown &
McNeill, 1966; Horowitz et al., 1968; Gupta 2005). In the ortho-production condition, on the
other hand, although word-initial position did exhibit a lower mean proportion transfer than
word-medial position, the results only approached significance. This may be due to the fact that
mean proportion transfer for <ll>-/j/ in the latter condition was very low.
In comparing the primacy and recency results between the list and word levels, it
becomes apparent that, whereas the results did not reveal a significant primacy or recency effect
at the list level, there was a word-initial position advantage at the word level. The word-initial
143
advantage at the word level in comparison with the list level is somewhat comparable to the one
reported in Archibald and Gathercole (2007). In order to test whether the same sequencing
memory mechanisms were involved in recalling at the (non)-word and list levels, Archibald and
Gathercole (2007) presented school-aged children with lists of sequences of consonants as
mono-syllabic non-words (e.g., <fow>, <moy>, <chee> and as a single word (e.g.,
<fowmoychee>). When they compared serial position effects between repetition of individual
non-words and word lists, a greater primacy effect was yielded for consonants in the non-word
repetition task in comparison with the list recall task. They attributed the non-word repetition
advantage to the facilitative effects of the additional physical cues inherent in the connected
multisyllabic stimuli such as the prosodic contour (e.g., Roy & Chiat, 2004) and coarticulation
(e.g, Nijland, van der Meulen, Gabreels, Kraaimaat, & Scrhreuder, 2002). That is, they argued
that the cues available for consonants in a stressed position and/or coarticulated could have
allowed for better encoding and recall and been responsible for the differences in the results.
While the patterns observed in this study are similar to the ones found in Archibald and
Gathercole (2007), their argumentation cannot be applied here because in the present study the
same stimuli were used for testing both primacy effects at the word level and primacy and
recency effects at the list level. Hence, as previously mentioned, the lack of a significant finding
for primacy and recency effects at the list level in this study is not related to differences in
prosody and/or coarticulation but is most likely due to the simplicity of task demands on
working memory.
The word-initial advantage in this study has been attributed to primacy effects. It is
important to mention that from a phonological point of view, the word initial-word medial
asymmetry has been attributed to acoustic prominence rather than working memory biases
(Steriade, 1997; Beckman, 1998; Cho & Jun, 2000; Colantoni & Steele, 2008). Therefore, it is
144
plausible that the word initial advantage is a result of a confounding effect of acoustic
prominence and primacy effects. This claim could be further tested by training participants with
two sets of grapheme-to-phoneme correspondences with differing degrees of acoustic salience.
A lower proportion of transfer for graphemes that correspond to an acoustically less salient
sound would indicate that acoustic prominence also plays a role in inhibiting transfer.
Another factor which might have also had a confounding effect is grapheme frequency.
For example, <ll> is found in word-initial position only in two words in English, namely
<Lloyd> and <llama>, a loanword from Spanish. It may be that the lexical infrequency of <ll>
word-initially makes this position salient, and leads to better noticing (e.g., Schmidt, 1990) of
the inconsistency between the Spanish and English grapheme-to-sound correspondence (e.g.,
<ll> corresponding to /j/ in Spanish and to /l/ in English) in this position. Better noticing,
subsequently leads to better encoding and recall in this position. According to Schmidt (1990,
1995), noticing is the essential ingredient for the input to become intake. For example, if a
difference between the L1 and the TL is noticed, then acquisition of the TL structure may occur.
However, if the difference is not noticed by the learner, acquisition will not take place. The
proposal that a lower proportion of transfer in the word-initial position might be due to a
confounding effect of grapheme-to-phoneme frequency is a new one. This hypothesis could be
tested by comparing the results of the proportion of orthography-induced transfer found for
<ll>-/j/ in different positions in this study with those of a grapheme-to-phoneme correspondence
that would trigger transfer but unlike <ll>-/j/ would be equally frequent in different positions
across the word in the learners’ L1.
All in all, whereas this study did not support the existence of a significant positional
effect at the list level, it provided evidence in support of word-initial position effect leading to
145
lower mean proportion of transfer than word-medial position in <ll>-/j/. It is proposed that the
lack of positional effects at the list level may be due to the fact that lists were composed of three
items only. With respect to the word-initial advantage, it is not clear whether this inhibiting
effect on orthography-induced transfer is solely due to primacy effects or it is the result of
primacy effects, acoustic prominence and/or graphemic infrequency in word-initial position.
In this section, I have discussed the results of the experimental study with respect to the
hypothesis regarding one of the well-known characteristics of PM. In doing so, I highlighted
differences of these effects at the list and word level and also pointed out the possibility of other
interfering factors in controlling orthography-induced transfer in production. In the next section,
I will discuss the effect of another memory phenomenon, namely repetition.
5.3.2 Effect of round/repetition
Another question in this study was whether task repetition/round would affect the proportion of
transfer in production. In order to test this hypothesis, participants were required to perform the
picture-naming task three times in a row in a single session. It was hypothesized that the
proportion of orthography-induced transfer in production would decrease as the number of
rounds increased from 1 to 3. When considering the overall effect of rounds on the proportion of
transfer, this factor was only significant with ortho-learning & production and ortho-learning
conditions. Specifically, in the ortho-production, the differences between rounds 1 & 3 and
round 1 & 2 were significant and, in the ortho-learning condition, the differences between
rounds 1 & 3 were significant. When looking at the effect of round on individual grapheme-tophoneme correspondences, significant differences were only found for <ll>-/j/ (rounds 1 & 3)
and <v>-/b/ (rounds 1 & 3, and 2 & 3) in the ortho-production condition. The finding with
146
regards to the positive effect of ‘round’ in this study are consistent with the general notion that
repetition positively affects recall (Hebb, 1961 Atkinson & Shiffrin, 1968; Mathews & Tulving,
1973) and previous L2 studies that have found that repetition is predictive of an increase in
higher probability of recall (Saragai et al. 1978; Horst et al., 1998; Waring & Takaki, 2003;
Webb, 2007). However, to the best of my knowledge, there are no previous studies that have
reported on the effect of frequency of exposure in terms of task repetition in relation to
orthography-induced transfer. Therefore, the finding that task repetition/frequency of encounters
with new TL grapheme-to-phoneme correspondences in one session decreased orthographyinduced transfer-based errors, albeit not for all grapheme-to-phoneme correspondences and not
in all conditions, is a new one. It is not clear why the factor ‘round’ would only affect <ll>-/j/
and <v>-/b/ as opposed to all grapheme-to-phoneme correspondences. These two grapheme-tophoneme correspondences do not form a particular class, such as ‘the easiest’ or ‘the most
difficult grapheme-to-phoneme correspondences’ for learners. In fact, <ll>-/j/ was the easiest
grapheme-to-phoneme correspondence to acquire based on the fact that it tended to involve the
lowest proportion of transfer and <v>-/b/ was one of the most difficult grapheme-to-phoneme
correspondences because it tended to have one of the highest proportions of transfer in
comparison with the other grapheme-to-phoneme correspondences examined in this study.
In sum, the factor ‘round’ did lower orthography-induced transfer in two out of three
conditions when grapheme-to-phoneme correspondences were collapsed. This suggests that
repetition could potentially positively affect establishing underlying representations when
grapheme-to-phoneme correspondences differ between English and Spanish. However, it is not
clear, at this point why this was only true for <ll>-/j/ and <v>-/b/ and not the other grapheme-tophoneme correspondences when they were considered individually. In this section and the
previous section, I have discussed the findings with respect to the well-known universal PM
147
phenomena of primacy and recency effects and repetition. In the next section, I will discuss the
effect of PM on orthography-induced transfer at the individual level.
5.3.3 Individual variation in PM and orthography-induced transfer
The results with respect to the relationship between individual PM capacity and orthographyinduced transfer did not confirm the predictions. Specifically, there were no correlations
between individual PM capacity and proportion of transfer. However, a power analysis showed
that a larger sample size would be needed to establish the existence of a correlation. Therefore,
currently, it cannot be said with confidence whether there is or there is not a correlation between
individual variation in PM capacity and orthography-induced transfer.
It is worth mentioning that, whereas most of the L2 studies have provided evidence in
support of the role of PM in vocabulary learning in low proficiency learners (Service, 1992;
Service & Kohonen., 1995; Cheung, 1996; Dufva & Voeten; 1999; French, 2004, Mizera, 2006)
did not find a correlation between PM and oral proficiency in low proficiency adult Englishspeaking learners of Spanish. One of the issues Mizera (2006) raised as a possible explanation
for the lack of a significant correlation was the contribution of affective factors such as stress
and anxiety at the time of performance either during the PM task and/or the oral fluency task.
Another possibility that Mizera (2006) raised in order to explain the results, was that working
memory is determined by background knowledge and familiarity with the matter. In other
words, the effects of working memory do not become visible unless there are memory traces of
the information that learners are presented with in long-term memory. Were this study to be
replicated with a larger sample size and the results still did not yield a significant correlation
between individual variation in PM and orthography-induced transfer, it would be worth
148
considering Mizera’s explanations. In such circumstances, a lack of a significant correlation
could then be attributed to participants being novice learners and their lack of exposure to
Spanish before performing the picture-naming task, especially in a task with high processing
demands, such as the task at hand; the picture-naming task in this study entailed dealing with
much new information, associating meaning to the written, auditory and visual input, and
deciphering grapheme-to-phoneme correspondences, all in a very short amount of time. In
addition, it is possible that affective factors may also be a valid explanation for a lack of a
significant correlation between individual variation in PM and orthography-induced transfer.
That is, different learners may have different levels of enthusiasm for taking part in the Spanish
task as opposed to the Farsi task for various reasons. One possibility would be that some
participants may prefer Spanish over Farsi and vice versa. Even though learners may not have
been exposed to these languages, some learners might have preconceptions about the difficulty
of acquisition of these languages. Subsequently, the perception of the degree of difficulty of
these languages may affect the participants’ level of motivation in different degrees. This in turn
might lead to differing levels of performance on each task for each individual learner.
In addition to the above possibilities, it is possible that attentional division abilities
might be at play in controlling the effect of orthography-induced transfer. Here, PM capacity
was measured as an index of individual variation in this study and ‘attentional division abilities’
(e.g., Bonnel & Hafter, 1998) were not considered. Bonnel and Hafter (1998) propose that while
dividing attention between two different sources of modality (e.g., auditory and visual) is easier
than doing so in one modality (e.g., auditory only), the former can be difficult if the task at hand
is more demanding than detecting occasional stimuli in the two sources of input. In the picturenaming task in the present work, participants assigned to auditory-orthographic conditions were
also required to divide their attention between different sources of input, namely the auditory,
149
visual (e.g., images) and orthographic information. In addition, the task at hand was demanding
because it did not simply require detecting stimuli in the input but it required vocabulary
learning in a foreign language. Therefore, it is possible that individual differences in attentional
division abilities might have impacted the learners’ proportion of transfer. If individual
differences in attentional division abilities were to affect the proportion of orthography-induced
transfer, learners with superior attentional divisional abilities would be able to pay attention to
the auditory, the orthographic and visual information, and thus would be more likely to notice
the mismatches between the TL and the L1 sound for a shared-grapheme-to-phoneme
correspondence, which would result in a lower proportion of transfer for such individuals. This
explanation, however, is less applicable to the ortho-production condition where, at the time of
learning, only auditory and visual input were presented to the learners and orthographic input
was only presented at production. In other words, learners in the ortho-production condition did
not have to divide their attention between the auditory and orthographic input at the time of
learning because they were not presented with them simultaneously.
In sum, with respect to the effect of PM on orthography-induced transfer, when
considering the effects of the universal aspects of PM (e.g., primacy and recency effects and
repetition effects) and individual PM capacity, the results are mixed. First, at the list level, there
was only a weak recency pattern that was not significant. Second, whereas in two out of three
conditions, word-initial position exhibited a lower proportion of orthography-induced transfer, it
was speculated that other factors such as psycho-acoustic prominence and graphemic frequency
might have had a confounding effect on reducing orthography-induced transfer. Therefore, one
cannot definitively say that the results in this study provide evidence in support of a primacy
effect. Third, there was a lack of a significant correlation between individual PM capacity and
orthography-induced transfer. A post-hoc power analysis revealed that a larger sample size
150
would be needed to further examine the hypothesis regarding the correlation between individual
PM capacity and orthography-induced transfer. The other potential factors proposed to explain
the lack of a correlation between PM and orthography-induced transfer in individual participants
were as follows: (a) a combination of the demanding nature of the task and the lack of any
previous exposure to Spanish; (b) different levels of enthusiasm for the picture-naming task and
the non-word repetition task in different participants; and (c) the potential interfering effect of
individual variation in attentional division abilities. Given the mixed results, the role of PM in
orthography-induced transfer will have to be further examined in future studies. I now turn to
conclusions and future directions.
5.4 Conclusions and future directions
This experiment has contributed to our understanding of the effect of orthography on
phonological transfer in different ways. First, it has provided a snapshot of the effect of
orthography at the absolute initial stage of acquisition in order to add to the relatively sparse
body of research that has examined the effect of orthography on promoting transfer leading to
non-target-like productions. The results in this study have confirmed the findings in previous
studies that exposure to orthographic input at learning and/or production results in orthographyinduced transfer (Young-Scholten et al., 1999; Erdener & Burnham., 2005). Additionally, the
experimental design in this study lent itself to testing a number of auditory-orthographic
conditions and has provided a more fine grained account of the effect of auditory-orthographic
effects on proportion of transfer leading to non-target-like productions in novice Englishspeaking learners of Spanish.
151
The findings in this thesis confirmed the findings in Young-Scholten (2000) in that
inconsistencies between the TL and L1 grapheme-to-phoneme correspondences when the TL
sound exists in the L1 promote transfer leading to non-target-like productions. In addition, this
study has extended the findings to TL sounds which may be considered ‘similar’ sounds. By
examining the role of grapheme-to-phoneme inconsistency in transfer in production, this study
has highlighted the importance of the role of orthography in L2 acquisition of phonology and
suggested that exposure to orthography may lead to the formation of underlying representations
for TL sounds. Specifically, it has suggested that in the process of acquisition learners exposed
to orthography may form non-target-like L1 based categories when the TL and L1 grapheme-tophoneme correspondences differ.
This study also adds to the body of knowledge on the effect of grapheme-to-phoneme
inconsistency because it shows that different grapheme-to-phoneme correspondences may result
in different proportions of transfer. It was proposed that frequency of a particular L1 graphemeto-phoneme correspondence as well as the difference in terms of the degree of acoustic/phonetic
distance and salience between the TL and the L1 sounds corresponding to a shared grapheme-tophoneme mapping may be the reason for these differences. The claim regarding the effect of
frequency of grapheme-to-phoneme correspondences is important, because whereas this factor
has been previously discussed in a word recognition study by Ranbom and Connine (2010), this
issue has not been addressed in any previous L2 production studies on orthography-induced
transfer. Regarding the proposal about the effect of acoustic/phonetic distance between the TL
and L1 sounds although a survey of the world’s languages for the frequency of /b/-v/ and /l/-/j/
contrasts provided some evidence for the hierarchy regarding the phonetic distances between the
TL and L1 sounds proposed in this study, quantifiable acoustic/phonetic data are needed to
measure the proposed phonetic distances. If for example, an acoustic/phonetic study showed that
152
the distance between /v/ and /b/ is smaller than the difference between /l/ and /j, the claim
regarding the effect of the degree of perceptual salience between the L1 and TL orthographyinduced transfer would be further substantiated.
The finding that the presence of orthography at learning and/or production at the initial
stages of acquisition can trigger transfer and that different grapheme-to-phoneme
correspondences may result in different proportions of transfer also has pedagogical
implications for teaching Spanish in a classroom setting. Specifically, this study has pointed to
an advantage for the audio-lingual method. That is, this study has shown that training learners
with the auditory form of the language leads to a better acquisition of the TL phonology when
the TL grapheme-to-phoneme correspondences differ between Spanish and English. Given that
exposure to orthography when the Spanish and English grapheme-to-phonemes differ can
promote transfer, it is recommended that at the initial stages of acquisition, Spanish language
instructors expose learners to the auditory input (e.g., via vocabulary learning tasks) prior to
introducing the corresponding orthographic forms. This approach would be beneficial to
learners because it would lower the chances of the formation of incorrect categories for TL
sounds at the very beginning stages of acquiring a new language. It is important that learners
acquire the right pronunciation of the TL sounds of interest at the beginning stages, otherwise
they may run a risk of forming incorrect categories and becoming fossilized learners. While an
audio-lingual method is recommended for pronunciation teaching in cases where the TL and L1
grapheme-to-phoneme correspondences differ, this approach may not always be feasible in
class-room setting. If instructors chose to introduce orthography simultaneously with the
auditory input instead, it is recommendable that they explicitly do a contrastive analysis of the
grapheme-to-phoneme correspondences that differ between English and Spanish so that learners
become aware of these differences between the two systems. A contrastive analysis of the
153
Spanish and English grapheme-to-phoneme correspondences would raise the learners’
awareness of the differences between the two languages and could subsequently lower their
pronunciation errors. In providing learners with a contrastive analysis of the Spanish and
English grapheme-to-phoneme correspondences, instructors could spend more time on teaching
grapheme-to-phoneme correspondences that induced a higher proportion of transfer such as
<z>-[s] in comparison with those that lead to a lower proportion of transfer such as <ll>-[j].
In addition to showing that orthography induces transfer in production, this study has
also raised the possibility that orthography may induce transfer in perception. Therefore, in
addition to pronunciation training via vocabulary learning and explicit instruction of graphemeto-phoneme differences between English and Spanish, it may be worth training learners
perceptually. However, further research with respect to the effect of orthography on perception
is required before concrete instructions for language educators could be offered.
With respect to the effect of PM on orthography-induced transfer leading to non-targetlike production, the results were somewhat mixed in this study: (a) primacy recency effects were
not observed at the list level but (b) a significant word initial position advantage was observed at
the word level; (c) there was some evidence of a repetition effect. In addition, as concerns the
word-initial position advantage level, it was suggested that other possible factors such as
psycho-acoustic prominence and graphemic frequency might have interfered. Moreover,
regarding the relationship between individual PM capacity on orthography-induced transfer, the
results did not yield a significant correlation. Therefore, future studies could further examine the
effect of PM both as concerns its universal characteristics and individual variation. For example,
primacy and recency effects could be examined using a longer list of items. Furthermore, in this
study, only primacy effects were tested at the word level, future studies could explore both
154
word-medial and word-final positions too. This will show whether recency effects in addition to
primacy effects play a role in orthography-induced transfer at the word level. Future
experimental designs could also test for the differences between effects of PM, acoustic
prominence, and graphemic frequency to be teased apart as previously explained. With regards
to individual differences, future studies could consider examining a larger data sample,
investigating other potential individual differences such as attentional division abilities in novice
learners and the role of PM capacity in learners of different proficiencies including beginners,
intermediates, and advanced. In this study, consistent with Mizera (2006), it was proposed that
one of the reasons for the lack of a significant correlation between individual PM capacity and
the proportion of orthography-induced transfer might be lack of experience with the TL.
Examining the effect of PM capacity in learners of different proficiencies will test this
hypothesis and contribute to our understanding of the effect of PM on orthography in L2
development. Although the body of research on the effect of orthography on L2 acquisition of
phonology is growing, there is much room for future studies to investigate the role of
orthography and factors that influence orthography-induced transfer in L2 development.
Finally, this study has focused on the effect of orthography on phonological transfer in
the L2 production of English-speaking learners of Spanish. Future studies could also investigate
the role of orthography-induced transfer in perception. For example, via a discrimination task,
future research could explore whether there is evidence of integration of auditory and
orthographic information at the perceptual level and determine the degree to which auditory and
visual channels may be integrated when mediated by orthography. It would also be beneficial to
test the effect of grapheme-to-phoneme inconsistencies in both perception and production in
alphabetic languages other than Spanish and with participants with a different alphabetic
155
language background to see to what extent the results found in this study can be generalized to
across languages.
All in all, this study is important because it has conducted a rigorous examination of the
effect of orthography on transfer in L2 and the factors that influence orthography-induced
transfer in adult novice English-speaking learners of Spanish. By providing a systematic
analysis of the effect of grapheme-to-phoneme inconsistency between L1 and TL on
phonological transfer, it has shown that orthography triggers L1-based phonological transfer at
the absolute initial stages of acquisition. There are very few studies that have examined transfer
at the absolute initial stages of L2 acquisition. Moreover, by providing a systematic analysis of
the role of a number of factors on promoting orthography-induced transfer, including graphemeto-phoneme inconsistency, auditory-orthographic condition and different aspects of PM, this
study has added to the sparse empirical evidence, especially concerning Spanish, and provided
a more-fine-grained picture of the factors that promote orthography-induced transfer leading to
non-target-like production.
156
Appendix A
Background questionnaire
A. Personal Information
•
Gender: _____________________________
•
Year of Birth: ________________________
•
Place of Birth: City _____________________
•
Highest Level of Schooling: Elementary, Secondary, Professional College, University
•
Current occupation:___________________________________
•
Subject of specialization _________________________________
Country ______________________
B. Language use
Which language(s) were you formally educated in?
•
Primary/Elementary School ________________
High School __________________
College ______________________
University _____________________
•
Which language(s) do you use (Indicate approximate percentage, e.g. 0, 50, 100%):
At school __________________________________________________________________
At home __________________________________________________________________
At work
__________________________________________________________________
In social situations
157
___________________________________________________________________________
Did you learn English from birth? ________ If not please explain:
•
__________________________________________________________________________
__________________________________________________________________________
•
What is your mother’s first language? ______________________
•
What is your father’s first language? ________________________
Were you exposed to any other language while growing up?_______ If so explain:
•
___________________________________________________________________________
___________________________________________________________________________
•
Do you speak any Spanish? ___________________________________________________
•
Have you ever been in contact with Spanish at school, through friends or media or traveling?
________________________________________________________________
•
If so describe the type and duration of your social contact and/or trip(s):
___________________________________________________________________________
B. Other languages
I. French
•
At what age did you begin to learn this language?
158
__________________________________________________________________________
•
Where did you learn this language?
__________________________________________________________________________
•
How long did you learn it for?
__________________________________________________________________________
•
Were your teachers native speakers of this language?
__________________________________________________________________________
•
Did you learn this language as a subject or was it the principal medium of instruction?
__________________________________________________________________________
•
Have you ever spent time in an area where this language was the native language? ________
If so please state where and for how long.
•
__________________________________________________________________________
Approximately how many hours a week do you use this language?
________________________________________________________________________
•
Please specify how many hours a week you use this language for:
Speaking___________________________________________________________________
Listening___________________________________________________________________
Reading___________________________________________________________________
II. Other languages
159
•
At what age did you begin to learn this language?
__________________________________________________________________________
•
Where did you learn this language?
__________________________________________________________________________
•
How long did you learn it for?
__________________________________________________________________________
•
Were your teachers native speakers of this language?
__________________________________________________________________________
•
Did you learn this language as a subject or was it the principal medium of instruction?
__________________________________________________________________________
•
Have you ever spent time in an area where this language was the native language? ________
•
If so please state where and for how long.
__________________________________________________________________________
•
Approximately how many hours a week do you use this language?
________________________________________________________________________
•
Please specify how many hours a week you use this language for:
Speaking___________________________________________________________________
Listening___________________________________________________________________
160
Reading____________________________________________________________________
D. Language ability
I. French
•
In comparison with other native speakers of French, how would you describe:
(a) Your reading ability in French (very poor, poor, average, good, very good, excellent)
(b) Your listening ability in French (very poor, poor, average, good, very good, excellent)
(c) Your overall competence in French (very poor, poor, average, good, very good, excellent)
II. English
•
In comparison with other native speakers of English, how would you describe:
(a) The size of your English vocabulary (small, medium, large, very large)
(b) Your reading speed in English (very slow, slow, average, fast, very fast)
(c) Your talking speed in English (very slow, slow, average, fast, very fast)
E. Other
Do you play any instruments?
2
__________________________________________________________________________
2
The factor of playing an instrument will not be considered in this project to explain individual variability. I
included this question here as a starting point for future work. Based on (Gottfried, 2007) I predict that the rate of
orthography-induced transfer in learners with musical training will be lower than those without any musical
training.
161
Appendix B
Picture-naming task: Assigned meanings
Table 1
Picture-naming Stimuli: Assigned Meanings
Semantic field/category
Assigned meanings
Clothing items
Accessories
Buildings and related parts
Animals
Dress, pants, jacket
Wallet, purse
Church, castle, windmill, swimming pool, window, floor,
roof
Fish, camel, panda, peacock, whale, tiger, crab, parrot,
turkey, elephant, penguin, octopus, horse
Pumpkin, peas, pear, radish, leaf, rose, cactus, crackers,
cookie, ice cream, waffle, pop corn, egg, walnut, wheat
Rain, fire, star
Triangle, square, circle, cross, star
Fireman, waiter, priest, sailor,
Thumb, chest
Razor, comb, brush, tweezers, towel
Wheelchair, cane
Drums
Fan, pitchfork, shovel, ladder, pot, drill, chain, rope, fishing
pole
Curtains, plate, pillow, chair, glass, carpet, clock, tea pot,
painting, bed, lampshade, wine bottle
Car, missile, fire truck, tank
Trophy, box, kite, tape, puzzle, feather, package, paper,
cross, gas, arrow, garbage, paper, ghost, town, pirot, king,
tear, dentures, wig
Plants, fruits, vegetables
and food
Natural phenomena
Shapes
Professions
Body parts
Toiletry
Walking aid
Musical instruments
Tools and instruments
Household items
Vehicles
Other
162
Appendix C
Picture-naming: Stimuli Meanings
Target Stimuli
<adentro> [aðen̪tɾo] ‘ inside’
<adono> [aðono] ‘I comply’
<adormo> [aðoɾmo] ‘I sleep’
<aherir> [aeɾiɾ] ‘to contrast size and weight’
<ahincar> [aiŋkaɾ] ‘to hurry’
<ahitar> [aitaɾ] ‘to cause indigestion’
<ahotar> [aotaɾ] ‘to incite’
<ahumar> [aumaɾ] ‘to smoke ‘
<amago> [amaɣo] ‘a sign of beginning of something’
<anafe> [anafe] ‘ portable stove’
<anata> [anata] ‘a type of eclesiastical tax’
<anego> [aneɣo] ‘ I drown’
<bacana> [bakana] ‘ domineering’ (Chile)
<batata> [batata] ‘potato’
<bimana> [bimana] ‘ two-handed’
<bofena> [bofena] ‘ animal lung’
<boruca> [boɾuka] ‘uproar’
163
<botina> [botina] ‘a gaiter’
<codena> [koðena] ‘ body of thickness required in cloth’
<colleta> [kojeta] ‘a small cabbage’ (Rioja, Spain)
<dagame> [daɣame] ‘a type of tree’(CUBA)
<darico> [daɾiko] ‘Persian gold coin’
<degano> [deɣano] ‘a farm administrator’
<derogo> [deɾoɣo] ‘ I abolish’
<detenga> [deteŋga] ‘I detain/he detains.SUBJ’
<dimana> [dimana] ‘has.3rd.SG a different origin’
<hanega> [aneɣa] ‘an agricultural measurement’
<harapo> [aɾapo] ‘a piece of fabric’
<harina> [aɾina] ‘flour’
<hontana> [on̪tana] ‘fountain’
<horaco> [oɾako] ‘whole’
<horita> [oɾita] ‘right now’ (Cuba and Mexico)
<llamingo> [jamiŋgo] ‘llama’ (Ecuador)
<llanero> [janeɾo] ‘an inhabitant of plain lands’
<llanito> [janito] ‘an inhabitant of Gibraltar’
<llanura> [januɾa] ‘plain’
<lloreta> [joɾeta] ‘crying fit’ (Honduras)
164
<llorona> [joɾona] ‘ a person that cries a lot’
<macaca> [makaka] ‘ ugly woman’ (Chile and Cuba)
<macana> [makana] ‘a situation that produces discomfort’ (Bolivia, Colombia and Ecuador)
<mallera> [majeɾa] ‘armorer’
<malleto> [majeto] ‘a type of wooden hammer’
<mallugo> [majuɣo] ‘I bruise something ‘ (Venezuela)
<metopa> [metopa] ‘ the space between trigliphs in adoric frieze’
<nacrita> [nakɾita] ‘a type of talc’
<namoro> [namoɾo] ‘ I fall in love’
<nerita> [neɾita] ‘a type of molusc’
<omino> [omino] ‘I predict’
<pallete>[pajete] ‘fender’
<pidona> [piðona] ‘demanding ‘
<pollero> [pojeɾo] ‘one who keeps or rears fowls’
<rehogar> [reoɣaɾ] ‘to fry’
<sarama> [saɾama] ‘ dirt’
<sicono> [sikono] ‘failure to produce fruits in fig trees’
<sigogo> [siɣoɣo] ‘a type of bird’ (Honduras)
<socapa> [sokapa] ‘pretext’
<somato> [somato] ‘I hit’ (El Salvador, Guatemala, Honduras)
165
<sotera> [soteɾa] ‘a type of spade’ (Argentina)
<tomento> [tomen̪to] ‘ a hair coating that covers the surface of some plant organs’
<tudanco> [tuðaŋko] ‘originally from Tudanca, Cantabria’
<vagante> [baɣan̪te] ‘ vagrant’
<vegana> [beɣana] ‘originally from La Vega’
<veneno> [beneno] ‘poison’
<verato> [beɾato] ‘from the province of Cáceres’
<vigota> [biɣota] ‘dead-eye’
<vireca> [biɾeka] ‘squint-eyed’
<zafero> [safeɾo] ‘a type of furniture’
<zanate> [sanate] ‘a type of bird’ (Costa Rica, Honduras ,Mexico, Nicaragua, Guatemala)
<zapito> [sapito] ‘a wooden cup’ (Cantabria)
<zarina> [saɾina] ‘a Zar’s wife’
<zatara> [sataɾa] ‘a type of wooden frame’
<zatico> [satiko] ‘a piece of bread’
Distracters
<a> [a] ‘the letter a’
<acá> [aka] ‘here’
<aquí> [aki] ‘here’
<agá> [aɣa] ‘an official of the Turkish army’
166
<ca> [ka] ‘why’
<cha> [ʧa] ‘tea’ (Philipines)
<che> [ʧe] ‘the letter ch’
<chorro> [ʧoro] ‘spurt’
<croe> [kɾoe] ‘croaks.SUBJ’
<cu> [ku] ‘the letter q’
<e> [e] ‘the letter e’
<efe> [efe] ‘the letter f’
<fea> [fea] ‘ugly’
<fo> [fo] ‘interjection for expressing disgust’
<fui> [fui] ‘ I was’
<gofre> [gofɾe] ‘a kind of cake’
<grúa> [gɾua] ‘crane’
<guiri> [giɾi] ‘a foreign turist’
<o> [o] ‘or’
<oca> [oka] ‘a type of domestic goose’
<pe> [pe] ‘the letter p’
<pecho> [peʧo] ‘chest’
<poa> [poa] ‘Graminea’
<ta> [ta] ‘knock’
167
<te> [te] ‘the letter t’
<ti> [ti] ‘you-SG (prepositional object pronoun)’
<tía> [tia] ‘aunt’
<to> [to] ‘careful (interjection)’
<toco> [toko] ‘I touch’
<toga> [toɣa] ‘toga’
<trina> [tɾina] ‘something made of three elements or units’
<trinche> [tɾin̪ʧe] ‘carves.SUBJ’
<troque> [tɾoke] ‘exchange’
<trucha> [tɾuʧa] ‘trout’
<tú> [tu] ‘ you.SG’
<u> [u] ‘ the letter u’
168
References
Archibald, L.M.D., Gathercole, S.E (2006). Short-term and working memory in SLI. International
Journal of Language and Communication Disorders, 41, 675-693.
Archibald, L.M.D. & Gathercole, S.E. (2007). Nonword repetition and serial recall: equivalent
measures of verbal short-term memory? Applied Psyholinguistics, 28, 587-606.
Aliaga-García, C., Mora, J. C., & Cerviño-Povedano, E. (2011, Msy). L2 speech learning in adulthood
and phonological short term memory. Paper presented at the meeting of New Sounds, Poznań,
Poland.
Atkinson, R.C. & Shiffrin, R.M. (1968). Human memory: a proposed system and its control processes.
In K.W. Spence (Ed.), The psychology of learning and motivation: advances in research and
theory (Vol. 2) (pp. 89–195). New York: Academic Press.
Baddeley, A. D. (1986). Working memory. New York: Clarendon Press/Oxford University Press.
Baddeley, A. D. (1999). Memory. In: R. A. Wilson & F. C. Keil (Eds.) The MIT Encyclopedia of the
Cognitive Sciences (pp. 514-517). Cambridge, Mass.: MIT Press.
Baddeley, A. D. (2002). Is working memory still working? European Psychologist (2), 85-97.
Baddeley, A. D. (2003). Working memory and language: an overview. Journal of Communication
Disorders, 36(3), 189-208.
Baddeley, A.D. & Hitch, G.J. (1974). Working memory. In G.A. Bower (Ed.), Recent advances in
learning and motivation, (Vol. 8) pp. (47–89). New York: Academic Press.
Baddeley, A. D., & Hitch, G. (2000). Development of working memory: should the Pascual-Leone and
the Baddeley and Hitch models be merged? Journal of Experimental Child Psychology, 77(2),
128-137.
Baddeley, A. D., Gathercole, S. E., & Papagno, C. (1998). The phonological loop as a language
learning device. Psychological Review, 105(1), 158-173.
169
Bassetti, B. (2007). Effects of hanyu pinyin on pronunciation in learners of Chinese as a foreign
language. In A. Guder, X. Jiang & Y. Wan (Eds.), The cognition, learning and teaching of
Chinese characters (pp. 156-179). Beijing, China: Beijing Language and Culture University.
Bassetti, B. (2009). Effects of adding interword spacing on Chinese reading: a comparison of Chinese
native readers and English readers of Chinese as a second language. Applied Psycholinguistics,
30(4), 757-775.
Beckman, J. N (1998). Positional faithfulness (Doctoral dissertation). Retrieved from A&I database.
(AAT9823717)
Beddor, P. S., & Gottfried, T. L. (1995). Methodological issues in cross-language speech perception
research with adults. In W. Strange (Ed.), Speech perception and linguistic experience:
theoretical and methodological issues in cross-language speech research (pp.207-232).
Timonium, MD: York Press.
Bertelson, P., De Gelder, B., Tfouni, L. V., & Morais, J. (1989). Metaphonological abilities of adult
illiterates: New evidence of heterogeneity. European Journal of Cognitive Psychology, 1(3),
239-250.
Best,C. (1995). A direct realist view of cross-language speech. In W.Strange (Ed.), Speech perception
and linguistic experience (pp.171-204). Baltimore: York Press.
Best, C. T., & Tyler, M. D. (2007). Nonnative and second-language speech perception: Commonalities
and complementarities. In M. J. Munro & O.-S. Bohn (Eds.), Second language speech learning:
The role of language experience in speech perception and production (pp. 13-34). Amsterdam:
John Benjamins.
Bonnel, A. M., and Hafter, E. R. (1998). Divided attention between simultaneous auditory and visual
signals. Percepual Psychology, 60, 179–190.
Bowey, J. A. (2001). Nonword repetition and young children's receptive vocabulary: A lonitudinal
study. Applied Psycholinguistics, 22(3), 441-469.
Browman, C.P., & Goldstein, L. (1989). Articulatory gestures as phonological units. Phonology, 6(2),
201-251.
170
Browman, C.P., & Goldstein, L. (1990). Representation and reality: Physical systems and phonological
structure. Journal of Phonetics, 18(3), 411-424.
Browman, C.P., & Goldstein, L. (1992). Articulatory phonology: an overview. Phonetica, 49(3-4), 155180.
Browman, C.P., & Goldstein, L. (1995). Dynamics and articulatory phonology. In T. van Gelder & R.
Port (Eds.), Mind as motion (pp. 175-193). Cambridge, MA: MIT Press.
Brown, C. A. (1998). The role of the L1 grammar in the acquisition of segmental structure. Second
Language Research, 14(2), 139-193.
Brown, C. A. (2000). The interrelation between speech perception and phonological acquisition from
infant to adult. In John Archibald (Ed). Second Language Acquisition and Linguistic Theory.
(pp.4 – 63). Oxford: Blackwell.
Brown, R., & McNeill, D. (1966). The "tip of the tongue" phenomenon. Journal of Verbal Learning
and Verbal Behavior, 5(4), 325-337.
Burgess, A. (1972). A Clockwork Orange. Harmondsworth: Penguin.
Burnham, D. K. (2003). Language specific speech perception and onset of reading. Reading and
Writing: An Interdisciplinary Journal, 16(6), 573-609.
Burnham, D. K., Earnshaw, L. J., & Clark, J. E. (1991). Development of categorical identification of
native and non-native bilabial stops: infants, children and adults. Journal of Child Language,
18(2), 231-260.
Ceñoz, J., & Genesee, F. (1998). Beyond bilingualism: multilingualism and multilingual education.
Philadelphia: Multilingual Matters.
Ceñoz, J., & Hoffmann, C. (2003). Acquiring a third language: What role does bilingualism play? The
International Journal of Bilingualism, 7(1), 1-5.
Cheung, H. (1996). Nonword span as a unique predictor of second-language vocabulary learning.
Developmental Psychology, 32(5), 867-873.
171
Cho, T., & Jun, S. (2000). Domain-initial strengthening as featural enhancement: Aerodynamic
evidence from Korean. Chicago Linguistics Society, 36, 31–44.
Clements, G.N. (1985). The geometry of phonological features. Phonology yearbook, 2, 225-252.
Cohen, J. (1988). Statistical power analysis for the behavioral sciences. Hillsdale: Lawrence Earlbaum
Associates.
Colantoni, L., & Steele, J. (2008). Integrating articulatory constraints in models of L2 phonological
acquisition. Applied Psycholinguistics, 29(1), 1-46.
Craik, F. I. (1970). The fate of primary memory items in free recall. Journal of Verbal Learning and
Verbal Behavior, 9(2), 143-148.
de Angelis, G. (2007). Third or additional language acquisition. Clevedon, UK: Multilingual Matter.
Deese, J., & Kaufman, R. A. (1957). Serial effects in recall of unorganized and sequentially organized
verbal material. Journal of Experimental Psycology, 54(3), 180-187.
Dufva, M., & Voeten, M. J. M. (1999). Native language literacy and phonological memory as
prerequisites for learning English as a foreign language. Applied Psycholinguistics, 20(3), 329348.
Ehri, L. C. (1980). The development of orthographic images. In U. Frith (Ed.), Cognitive processes in
spelling (pp.311-338). London, England: Academic Press.
Ehri, L. C. (1984). How orthography alters spoken language competencies in children learning to read
and spell. In J. Downing & R. Valtin (Eds.), Language awareness and learning to read (pp.119147). New York: Springer-Verlag.
Erdener, V.D. & Burnham, D.K. (2005). The role of audiovisual speech and orthographic information
in non native speech production, Language Learning, 55(2), 191-228.
Errasti, M. P. S. (2003). Acquiring writing skills in a third language: The positive effects of
bilingualism. International Journal of Bilingualism, 7(1), 27-42.
172
Escudero, P., Benders, T., & Lipski, S. (2009). Native, non-native and L2 perceptual cue weighting for
Dutch vowels: the case of Dutch, German, and Spanish listeners. Journal of Phonetics, 37(4),
452-466.
Escudero, P., Hayes-Harb, R., & H. Mitterer (2008). Novel second-language words and asymmetric
lexical access. Journal of Phonetics, 36(2), 345-360.
Escudero, P., & Wanrooij, K. (2010). Learning the phonological forms of new words: effects of
orthographic and auditory input. Language and Speech, 53(3), 361-381.
Flege, J. E. (1995). Second language speech learning: Theory, findings, and problems. In W. Strange
(Ed.), Speech perception and linguistic experience: Issues in cross-linguistic research (pp. 233277). Timonium, MD: York Press.
Flege, J. E. (2009). Give input a chance!. In T. Piske & M. Young-Scholten (Eds.), Input matters in
SLA (pp. 175-190). Clevedon: Multilingual Matters.
Foreit, K. G. (1976). Short-lived auditory memory for pitch. Perception & Psychophysics, 19(4), 68370.
French, L. M. (2004). Phonological working memory and L2 acquisition: a developmental study of
Quebec Francophone children learning English (Doctoral Dissertation). Retrieved from
Dissertation Abstracts International. (65(2-A), 487)
Gathercole, S. E., & Baddeley, A. D. (1993). Phonological working memory: A critical building block
for reading development and vocabulary aquisition? European Journal of the Psychology of
Education, 8(3), 259-272.
Gathercole, S.E., Hitch, G.J., Service, E., & Martin, A.J. (1997). Phonological short-term memory and
new word learning in children. Developmental Psychology, 33, 966-979.
Gathercole, S. E., Service, E., Hitch, G. J., Adams, A. M. G., & Martin, A. J. (1999). Phonological
short-term memory and vocabulary development: Further evidence on. Applied Cognitive
Psychology, 13(1), 65-77.
173
Gathercole, S.E., Willis, C.S., Emslie, H., Baddeley, A.D. (1991). The influences of number of
syllables and wordlikeliness on children’s repetition of nonwords. Applied Pyscholinguistics,
12, 349-367.
Gathercole, S.E., Willis, C.S., Emslie, H., & Baddeley, A.D. (1992). Phonological memory and
vocabulary development during the early school years: a longitudinal study. Developmental
Psychology, 28, 887-898.
Gottfried, T. L. (2007). Music and language learning. Effect of musical training on learning L2 speech
contrasts. In O-S. Bohn & M. J. Munro (Eds.), Language experience in second language
learning: In honor of James Emil Flege (pp.221-237). Amsterdam: John Benjamins.
Gupta, P. (2005). Primacy and recency in nonword repetition. Memory, 13(3-4), 318-324.
Halle, P. A., Chereau, C., & Segui, J. (2000). Where is /b/ in "apsurde" [apsyrd]? It is in French
listeners' minds. Journal of Memory and Language, 48(4), 618-639.
Hebb, D. O. (1961). Distinctive features of learning in the higher animal. In J. F. Delafresnaye (Ed.),
Brain mechanisms and learning (pp. 37–51). London: Oxford University Press.
Hayes-Harb, R., Nicol, J., & Barker, J. (2010). Learning the phonological forms of new words: effects
of orthographic and auditory input. Language and Speech, 53(3), 367-381.
Horowitz, L. M., White, M. A., & Atwood, D. W. (1968). Word fragments as aids to recall: The
organization of a word. Journal of Experimental Psychology, 76(2), 219-226.
Horst, M. T., Cobb, T., & Meara, P. (1998). Beyond a clockwork orange: acquiring second language
vocabulary through reading. Reading in a Foreign Language, 11(2), 207-203.
Hummel, K. M. (2009). Aptitude, phonological memory, and second language proficiency in nonnovice
adult learners. Applied Psycholinguistics, 30(2), 225-249.
Ibrahim, R., Eviatar, Z., & Ahron-Peretz, J. (2002). The characteristic of Arabic orthography slow its
processing. Neuropsychology, 16(3), 322-326.
174
Jones, L. (1979). Simplified version of T. Hardy’s The Mayor of Casterbridge, 2000, basewords.
Walton-on-Thames: Nelson.
Kellerman, E. & Sharwood Smith, M. (Eds.). (1986). Crosslinguistic influence in second language
acquisition. New York: Pergamon Press.
Lesaux, N. K. & Geva, E. (2006). Synthesis: Development of literacy in language-minority students. In
D. August & T. Shanahan (Eds.), Developing literacy in second language learners: Report of
the National Literacy Panel on Language-Minority Children and Youth. Mahwah, NJ: Lawrence
Erlbaum Associates.
Maddieson, I. (1984). Patterns of sounds. Cambridge: Cambridge University Press.
Martínez Celdrán, E. (2008). Some chimeras of traditional Spanish phonetics. In L. Colantoni, & J.
Steele (Eds.), Selected proceedings of the 3rd conference on laboratory approaches to Spanish
phonology. (pp. 32-46). Somerville, MA: Cascadilla.
Mathews, R. C., & Tulving, E. (1973). Effects of three types of repetition on cued and noncued recall of
words. Journal of Verbal Learning and Verbal Behavior, 12(6), 707-721.
Mazzaro, N. (2011). Experimental approaches to sound variation: a sociophonetic study of labial and
velar fricatives and approximants in Argentine Spanish (Unpublished doctoral dissertation).
University of Toronto. Toronto.
McGurk, H., & MacDonald, J. W. (1976). Hearing lips and seeing voices. Nature , 264 (5588), 746748.
Metsala, J. L., & Walley, A. C. (1998). Spoken vocabulary growth and the segmental restructuring of
lexical representations: Precursors to phonemic awareness and early reading ability. In J. L.
Metsala & L. C. Ehri (Eds.), Word recognition in beginning literacy (pp.89-120). Hillsdale, NJ:
Erlbaum.
Miller, G. A. (1956). The magical number seven, plus or minus two: some limits on our capacity for
processing information. Psychological Review, 63 (2), 81-97.
175
Mizera, G. J. (2006). Working memory and L2 oral fluency (Unpublished doctoral dissertation).
University of Pittsburgh, Pittsburgh.
Morais, J. C., Cary, L., Alegria, J., & Bertelson, P. (1979). Does awareness of speech as a sequence of
phones arise spontaneously? Cognition, 7(4), 323-331.
Murdock, B. B. (1962). The serial position effect of free recall. Journal of Experimental Psychology, 64
(5), 482-488.
Neufeld, G. (1978). A theoretical perspective on the nature of linguistic aptitude. International Review
of Applied Linguistics, 16(1-4), 15-26.
Nijland, L., Maasen, B., Van der Meulen, S., Gabreëls, F., Kraaimaat, F. W., & Schreuder, R. (2002).
Coarticulation patterns in children with dvelopmental apraxia of speech. Clinical Linquistics
and Phonetics, 16(6), 461-483.
O' Brien, I., Segalowitz, N., Collentine, J., & Freed, B. (2006). Phonological memory and lexical,
narrative, and grammatical skills in second language oral production by adult learners. Applied
Psycholinguistics, 27(3), 377-402.
O' Brien, I., Segalowitz, N., Freed, B., & Collentine, J. (2007). Phonological memory predicts second
language oral fluency gains in adults. Studies in Second Language Acquisition, 29(4), 557-581.
Odlin, T. (1989). Language transfer: Cross-lingistic influence in language learning. Cambridge:
Cambridge University Press.
Ranbom, L. J., & Connine, C. M. (2011). Silent letters are activated in spoken word recognition.
Language and Cognitive Processes, 26(2), 236-261.
Reid, C., Zhang, Y., Nie, H., & Ding, B. (1986). The ability to manipulate speech sounds depends on
knowing alphabetic writing. Cognition, 24(1-2), 31-44.
Richard, J., & Rodgers, T. (2001). Approaches and methods in language teaching. Cambridge:
Cambridge University Press.
176
Rott, S. (2000). The effect of exposure frequency on intermediate language learners' incindental
vocabulary acqusition and retention through reading. Studies in Second Language Acquisition,
21(4), 589-619.
Roy, P., & Chiat, S. (2004). A prosodically controlled word and nonword repetition task for 2-to 4year-olds: evidence from typically developing children. Journal of Speech, Language, and
Hearing Research, 47(1), 223-234.
Rundus, D. (1971). Analysis of rehearsal processes in free recall. Journal of Experimental Psychology,
89 (1), 63-77.
Rundus, D., & Atkinson, R. C. (1970). Rehearsal processes in free recall: A procedure for direct
observation. Journal of Verbal Learning and Verbal Behavior, 9(1), 99-105.
Saragai, T., Nation, I. P., & Meister, G. F. (1978). Vocabulary learning and reading. System, 6(2), 7278.
Schmidt, R. (1990). The role of consciousness in second language learning. Applied Linguistics, 11(2),
129-158.
Schmidt, R. (1995). Consciousness and foreign language learning: A tutorial on the role of attention and
awareness in learning. In R.W. Schmidt (Ed.), Attention and Awareness in Foreign Language
Learning (pp. 1–63). Honolulu: Second Language Teaching and Curriculum Center, University
of Hawai’i.
Speciale, G., Ellis, N. C., & Bywater, T. (2004). Phonological sequence learning and short-term store
capacity determine second language vocabulary acquisition. Applied Psycholinguistics, 25(2),
293-321.
Seidenberg, M. S., & Tanenhaus, M. K. (1979). Orthographic effects in rhyme monitoring. Journal of
Experimental Psychology: Human Learning and Memory, 5(6), 546-554.
Service, E. (1992). Phonology, working memory, foreign-language learning. Quarterly Journal of
Experimental Psychology, 45(1), 21-50.
177
Service, E., & Kohonen, V. (1995). Is the relation between phonological memory and foreign language
learning accounted for by vocabulary acquisition? Applied Psycholinguistics, 16(2), 155-172.
Snowling, M., Chiat, S., & Hulme, C. (1991). Words, nonwords, and phonological processes: Some
comments on Gathecole, Willis, Emslie, and Baddeley. Applied Psycholinguistics , 12(3), 369373.
Steele , J. (2002). Representation and phonological licensing in the L2 acquisition of prosodic
complexity (Unpublished doctoral dissertation). McGill University, Montreal.
Steele , J. (2005, June). Assessing the role of orthographic versus uniquely auditory input in acquiring
new L2 segments. 7èmes recontre internationals du réseau français de phonologie, Aix-en
Provence, France.
St Exupery A. de (1943). The Little Prince. New York: Harcourt Inc.
Steriade, D. (1997). Phonetics in phonology: the case of laryngeal neutralization. Ms., UCLA.
Taft, M., & Hambly, G. (1985). The influence of orthography on phonological representations in the
lexicon. Journal of Memory and Language, 24(3), 320-335.
Trubetzkoy, N. S. (1939/1958). Grundzüge der phonologie. Göttingen: Vandenhoek & Ruprecht.
Tyler, M. D., & Burnham, D. (2006). Orthographic influence on phoneme deletion response times.
Quarterly Journal of Experimental Psychology, 59(11), 2010-2031.
Young-Scholten, M. (2000, March). Is there a fundamental difference between adults’ and children’s
metalinguistic awareness? Paper presented at the meeting of AAAL, Vancouver.
Young-Scholten, M. (2002). Orthographic input in L2 phonological development. In P. Burmeister, T.
Piske & A. Rohde (Eds.), An integrated view of language development: Papers in honor of
Henning Wode (pp. 263-279). Trier: Wissenschaftlicher Verlag Trier.
Young-Scholten, M., Akita, M., & Cross, N. (1999). Focus on form in phonology: orthographic
exposure as a promoter of epenthesis. In P. Robinson & N. O. Jungheim (Eds.), Pragmatics and
178
pedagogy: Proceedings of the Third PacSLRF (vol. 2, pp. 226-234). Tokyo: Aoyama Gakuin
University.
Waltmunson, J. C. (2005). The relative difficulty of Spanish /t,d/, trill and tap by L1 English speakers:
Auditory and acoustic methods of defining pronunciation accuracy (Unpublished doctoral
dissertation). University of Massachusetts, Amherst.
Waring, R., & Takaki, M. (2003). At what rate do learners learn and retain new vocabulary from
reading a graded reader? Reading in a Foreign Language, 15(2), 130-163.
Waugh, N. C., & Norman, D. A. (1965). Primary memory. Psychological Review, 72(2), 89-104.
Webb, S. (2007). The effects of repetition on vocabulary. Applied Linguistics, 28(1), 46-65.
Weinreich, U. (1953). Languages in contact. New York: Linguistic Circle of New York.
Welch, R. B., & Warren, D. H. (1980). Immediate perceptual response to intersensory discrepancy.
Psychological Bulletin , 88(3), 638-667.
Zampini, M. L. (1994). The role of native language transfer and task formalit in acquisition of Spanish
spirantization. Hispania, 77(3), 470-481.
Zampini, M. L. (1997). L2 Spanish spirantization, prosodic domains, and Interlanguage rules. In
Hannahs, S. J. & Young-Scholten, M. (Eds.), Focus on phonological acquisition (pp.209-234)
Amsterdam: John Benjamins.
Ziegler, J. C., & Ferrand, L. (1998). Orthography shapes the perception of speech: The consistency
effect in auditory word recognition. Psychonomic Bulletin & Review, 5(4), 68-689.
Ziegler, J. C., Muneaux, M., & Grainger, J. (2003). Neighborhood effects in auditory word recognition:
Phonological competition and orthographic facilitation. Journal of Memory and Language ,
48(4), 779-793.
`