 

C)

.’{\
.v

 

i LIBRARY
Michigan State
University

 

 

 

This is to certify that the
thesis entitled

EFFECTS OF L2 PROFICIENCY ON WORD
RECOGNITION: THE USE OF PHONOLOGY IN READING
BY L1 CHINESE LEARNERS OF ENGLISH

presented by

RYAN T. MILLER

has been accepted towards fulﬁllment
of the requirements for the

MA. degree in Teaching English to Speakers
of Other Languages

 

 

/2// 47an ”yr/(Cg

V v Major Professor 8 Signature
6 "/5" 253;?
Date

MSU is an Aﬁinnative Action/Equal Opportunity Employer

 

PLACE IN RETURN BOX to remove this checkout from your record.
TO AVOID FINES return on or before date due.
MAY BE RECALLED with earlier due date if requested.

 

DATE DUE DATE DUE DATE DUE

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

SIOB K:/Prolecc&Pres/ClRC/DateDuo.indd

 

EFFECTS OF L2 PROFICIENCY ON WORD RECOGNITION: THE USE OF
PHONOLOGY IN READING BY L1 CHINESE LEARNERS OF ENGLISH

By

Ryan T. Miller

A THESIS

Submitted to
Michigan State University
in partial fulﬁllment of the requirements
for the degree of

MASTER OF ARTS
Teaching English to Speakers of Other Languages

2009

ABSTRACT

EFFECTS OF L2 PROFICIENCY ON WORD RECOGNITION: THE USE OF
PHONOLOGY IN READING BY Ll CHINESE LEARNERS OF ENGLISH

By

Ryan T. Miller

Much of the way words are recognized and processed in any language is related to
and inﬂuenced by the writing system used in that language. A multitude of research has
shown that ﬁrst language orthography can have a signiﬁcant impact on the way that
words are recognized in a second language. However, is this transfer effect from the L1
orthography a static one that is insurmountable, or is it possible for cognitive processes to
be restructured for the beneﬁt of the L2 reader? The present study investigated this
question through the use of two independent measures of word recognition processes. A
lexical decision task incorporating an altemating-case text paradigm, and including word-
based independent variables of word frequency and spelling regularity indicated that at
higher proﬁciency levels, L1 Chinese learners of English have improved intra-word
processing, especially among words with regular grapheme-phoneme correspondence. In
addition, a semantic category judgment task using a homophone paradigm indicated that
high proﬁciency readers have a greater reliance on phonology, and a lesser reliance on

orthography, than do low proﬁciency readers.

Copyright by
Ryan T. Miller
2009

TABLE OF CONTENTS

LIST OF TABLES .............................................................................................................. vi
LIST OF FIGURES ........................................................................................................... vii
Introduction ......................................................................................................................... 1
Review of Research ............................................................................................................ 3
Introduction ............................................................................................................. 3
English Orthography ............................................................................................... 4
Chinese Orthography .............................................................................................. 5
Models of Word Recognition .................................................................................. 8
Visual Route ................................................................................................ 9

Phonological Route ..' ................................................................................. 11

Dual Route ................................................................................................ 15

Orthographic Depth Hypothesis ............................................................... 16

Transfer ................................................................................................................. 1 7
Proﬁciency ............................................................................................................ 20
Research Questions ........................................................................................................... 24
Hypotheses ........................................................................................................................ 25
Methodology ..................................................................................................................... 26
General Procedure ................................................................................................. 26
Background Questionnaire .................................................................................... 28
Proﬁciency Measure ............................................................................................. 29
Participants ............................................................................................................ 3O
Experiment 1 ..................................................................................................................... 35
Materials ............................................................................................................... 35
Procedure .............................................................................................................. 39
Results ................................................................................................................... 40
Reaction Times ......................................................................................... 41

Accuracy ............................................................................................................... 48
Discussion ............................................................................................................. 50
Experiment 2 ..................................................................................................................... 54
Participants ............................................................................................................ 55
Materials ............................................................................................................... 55
Procedure .............................................................................................................. 57
Results ................................................................................................................... 58

Accuracy ................................................................................................... 59

Reaction Times ......................................................................................... 62
Discussion ............................................................................................................. 65
General Discussion ........................................................................................................... 69
Limitations and Future Research .......................................................................... 71
Appendices ........................................................................................................................ 74
Appendix A ........................................................................................................... 74
Appendix B ........................................................................................................... 75
Appendix C ........................................................................................................... 77
Appendix D ........................................................................................................... 78
Appendix E ........................................................................................................... 80
Appendix F ........................................................................................................... 81
Appendix G ........................................................................................................... 83
References ......................................................................................................................... 84

LIST OF TABLES

Table 1 Means (and standard deviations) of proficiency measure results ....................... 34

Table 2 Means (and standard deviations) of participant age and length of residence in
English-speaking locations ............................................................................................... 35

Table 3 Descriptive statistics for frequency, length, and word class for each condition .39
Table 4 Means (and standard deviations) of reaction times and error rates ................... 42

Table 5 Summary of means, standard deviations, and coeﬂicients of variation for
alternating case stimuli ..................................................................................................... 45

Table 6 Examples of phonological and orthographic similarity ...................................... 57

Table 7 Means and standard deviations of error rates for each condition, in percents ...60

vi

LIST OF FIGURES

Figure 1. Chinese and English orthography. ..................................................................... 5
Figure 2. Direct or visual route to meaning (post-lexical phonology). ............................. 9
Figure 3. Phonology-mediated route to meaning (pre-lexical phonology). .................... 12
Figure 4. Dual-route model. ............................................................................................. 16
Figure 5. MTELP scores for all 34 participants (maximum possible score of 60). ........ 31

Figure 6. Interaction between spelling regularity and proﬁciency within case. .............. 44

Figure 7. Means of lower case and alternating case reaction times by proﬁciency

group .................................................................................................................................. 46
Figure 8. Interaction between case and proﬁciency within spelling regularity

conditions ........................................................................................................................... 47
Figure 9. Error rate interaction between proﬁciency group and frequency. ................... 49
Figure 10. Paradigm used in Experiment 2. .................................................................... 58

Figure I 1. Category judgment accuracies in the four conditions by proﬁciency group. .61

Figure 12. Error rates of phonological similarity in the similarly-spelled and less
similarly spelled conditions. ............................................................................................. 64

Figure 13. Mean reaction times for phonological similarity by proﬁciency group. ........ 65

vii

Introduction

Reading, in any language, is a complicated process that involves a wide range of
subprocesses in order to be successful, and a skill that has long been of interest to
researchers. These subprocesses of reading are typically divided into so-called higher-
level processes and lower-level processes. Higher-level processes involve such concepts
as how ideas within the text relate to each other (cohesion), the organization of the parts
of a text (rhetorical structure), and what the main ideas of a text are and how they relate
to the reader’s own knowledge of the world. Lower-level processes involve the
processing of the pieces that make up a text, such as letters, words, characters, and the
smaller units that these are comprised of. Which type of processing is more important is
a question that has long been debated. Top-down models (that is, models that investigate
how higher-level processes affect lower-level processes) have largely focused on the
importance of background knowledge of content and patterns of structure, while bottom-
up models have focused on automaticity of word recognition, or the process of
understanding meaning from the graphemic representation of a word.

While strong views of top-down and bottom-up processing, where one type of
processing feeds into the other in a serial manner, were once prevalent, it is now widely
accepted that there is interplay between the two. Stanovich (1981) proposed an
interactive-compensatory model in which deﬁciencies in either higher- or lower—level
processing can be made up for by the other. This can be exempliﬁed by the concept of
hypothesis-testing. Here, the reader makes a hypothesis about what is to come based on
background knowledge; when there is less background knowledge available for top-down

processing, the reader is less able to make an accurate hypothesis. In this case, the reader

is required to gain as much information as possible from the text itself. On the other hand,
deﬁcient lower-level processing can be facilitated by greater background knowledge that
would allow the reader to make a better hypothesis. However, this greater use of
background knowledge to compensate for inadequate lower-level processing consumes
attentional resources (Stanovich, 1980), and because the human brain has a limited
amount of cognitive capacity (LaBerge & Samuels, 1974; Schnieder & Shiffrin, 1977),
the greater consumption of cognitive resources to form hypotheses can mean that fewer
resources are available for comprehension (Stanovich, 1980). As such, deﬁcient lower-
level processes can greatly hinder comprehension in a way that it is difﬁcult for higher-
level processing to compensate for, and can even make comprehension impossible
(Segalowitz, 1986). It is for this reason that word recognition is often seen as the
foundation of reading comprehension (Gough, 1984; Stanovich, 1991).

For second language readers, the word recognition process can be further
complicated by differences inherent in the processing of their ﬁrst and second languages.
It has been suggested that different orthographies entail differences in word recognition,
and that word recognition strategies are formed in a reader based on the orthography of
their ﬁrst language (Durgunoglu, Nagy, & Hancin-Bhatt, 1993). First language word
recognition processes have been shown to inﬂuence second language learner reading,
especially when that second language uses an orthography that is typologically different
from that of the reader’s ﬁrst language (Koda, 1988, 1989, 1990).

This inﬂuence from the learner’s Ll has been considered to be a static element in
the previous research, with consideration being given almost exclusively to the

characteristics of the L1 and the L2. While this research has been fruitful in describing

the nature of word recognition in different languages and the inﬂuence of one language
upon another, there has been little attention paid to the role that the readers themselves
play, and how individual differences in the reader may affect word recognition. Recently,
however, more research has taken note of the possibility that word recognition processes
may actually be more ﬂuid than previously thought, and with increased exposure to and
greater proﬁciency in a second language, the reader’s word recognition and processing of
that language might change (e. g. Akamatsu, 2005; Chikamatsu, 2006). While much of
the previous research has focused on how the L1 affects L2 word recognition, in order to
have a more comprehensive understanding of L2 word recognition processes, we must
also investigate to what extent L1 orthography inﬂuences L2 word recognition in
individual readers. Koda (2005) emphasizes this point by stating that there is a need for
studies to investigate “developmental changes in the manner and extent to which L1
and L2 processing experiences coalesce in forming L2 word-recognition skills” (p.46;
emphasis added). The present study seeks to explore this important, but as of yet largely
neglected, aspect of reading by investigating the effect of proﬁciency on the use of
phonology during word recognition by L1 Chinese learners of English.
Review of Research

Introduction

Word recognition processes in any written language have been shown to be
shaped by the nature of the writing system of the language. At the most basic level, the
various writing systems of the world’s languages map spoken language to written
characters in three main ways (DeFrancis, 1989; Muljani, Koda, & Moates, 1998; Wang,

Koda, & Perfetti, 2003): alphabetic orthographies map phonemes, the smallest units of

sound; syllabic orthographies map syllables, or units of multiple sounds; and
logographic/morphographic orthographies map words or morphemes, the smallest unit of
meaning (Spencer, 1991). These fundamental differences in the orthographies of
languages have signiﬁcant implications for the way that words are recognized in each
language.

English Orthography

In English and other alphabetic languages, words consist of letters, which each
have a sound that they represent. In these languages, there are strong relationships
between letters and their sounds, called grapheme-phoneme correspondence (GPC) rules.
Because English words can be processed serially, letter by letter, GPC rules make the
phonological information of each constituent letter of a word accessible during decoding.
When reading English words, native English speakers assemble the sounds of the letters
to form words; this is how native-English speaking children learn to read (Ehri, 1998),
and is the fundamental process of word recognition in English.

However, even though the phonology of words is highly accessible in English,
very familiar words (also called sight-words) have been shown to be recognized based on
their visual representation, without need for activation of phonology. In fact, research
has shown that highly-skilled readers could name familiar, monosyllabic words (made up
of multiple letters) just as quickly as they could name a single letter (Doehring, 1976). In
addition, relying on GPC rules alone does not always give adequate information, as in the

case of irregularly spelled words such as ‘knight’ or ‘colonel’.

 

Chinese English

 

Orthographic representation 55 t

/t/

93

l...
Meaning \j

Figure 1. Chinese and English orthography.

\

Phonological representation / mao/ l

 

Chinese Orthography

The Chinese language has long been considered to be a logography. Although
English words have grapheme-phoneme correspondence rules that the reader can apply to
pronounce and assemble each letter of a word, this concept does not apply to Chinese, as
words are formed in a single block of a number of strokes that do not map out a full
phonetic representation of the word (see Figure 1).

At a‘ phonological level, rather than a written character representing a single
phoneme, each Chinese character represents a syllable. In addition, at a semantic level,
each character represents a morpheme, or a unit of meaning. This is different from
English in that, as a constituent letter of a word, the letter ‘1’ has no meaning other than to
indicate a labio-dental fricative in English, whereas a single Chinese character can

represent an entire concept, e. g. k (/huo/3 1), or ﬁre. Such a character cannot be broken

down into any smaller unit. Chinese words are generally considered to be

 

1 Throughout this paper, numbers will be used to represent the tones of standard Mandarin Chinese. The
numbers 1, 2, 3, and 4 will be used to represent the level (31?), rising (ﬁg), departing (ﬁg), and
entering (A?) tones, respectively.

monomorphemic, although polymorphemic (and thus multi-syllabic and multi-character)
words also exist. It is because of this relationship between the written form, sound, and
meaning that Chinese characters are often referred to as morphemic (e. g., Leong, 1986)
or morphosyllabic (e.g., DeFrancis, 1989).

In addition, modern standard Chinese consists of 4,574 characters (Language and
Teaching Institute of Beijing Linguistic College, 1986, as cited in Perfetti, Liu, & Tan,
2005). Also, in modern standard Chinese, there are roughly 420 different syllables
(disregarding tones), and thus for every syllable there are, on average, about eleven
characters that share the same syllable (i.e. homophones). If the Chinese writing system
had been a representation of the phonology of the language, the reader would then need
to disambiguate between these eleven homophones using syntactic or semantic context,
thus consuming more cognitive resources (Katz & Frost, 1992). Such a system would be
far from efﬁcient. This would seem to indicate that, during reading, greater reliance on
orthographic representation would be necessary as reliance on phonological
representation would only lead to confusion among the ten other words that have the
same pronunciation (Katz & Frost, 1992). Since these homophones are not homographs,
however, reliance on the orthographic representation would give sure-ﬁre information
about the intended meaning.

On the other hand, if Chinese characters had no systematic relation to phonology,
a reader would have to remember, without any cue, the phonology of each and every one
of the 4,5 74 characters individually, which would be a very daunting task. Although
many characters in Chinese consist of only a single part (such as ‘ﬁre’ above), a large

number of characters are actually an amalgam of smaller pieces put together; such

characters are called compound characters, and the smaller pieces are called radicals.
Most compound characters consist of at least two radicals, but can contain many more.
Radicals can be divided into two types: semantic radicals and phonetic radicals.
Semantic radicals provide a hint as to the meaning of the compound character, while

phonetic radicals provide a hint as to the pronunciation. For example, B is pronounced
/ri/4, and has the meaning of sun. The character Ea pronounced /qing/l , has the meaning
of green. A combination of El (/ri/4, sun) on the left and § (/qing/1, green) on the right
produce the character Hi, pronounced /qing/2 and meaning sunshine. In this character,

both a semantic radical (sun, indicating a semantic relationship between the radical [sun]
and the compound [sunshine]) on the left and a phonetic radical (/qing/1) on the right
(indicating a phonetic relationship between the radical [pronounced /qing/1] and the
compound [pronounced /qing/2]) are present (Perfetti, Liu, Fiez, Nelson, Bolger & Tan,
2007). Overall, there are 895 phonetic radicals and 214 semantic radicals in Chinese
(DeFrancis, 1989). Unlike the previous example, however, most compound characters
have only one radical that is either semantic or phonetic, but not both. The rest of the
character typically has no relation, either semantic or phonetic, to the character as a
whole. Although the greater number of phonetic radicals may seem to indicate that
reliable phonetic cues are embedded in Chinese characters, phonetic radicals are reliable
only 38% of the time (Zhou, 1978, as cited in Perfetti, Liu, & Tan, 2005). In addition,

there are many compound characters that contain neither a phonetic nor a semantic

radical. For example, % (/qing/ 1, green) is actually a compound character that consists
of two characters that provide neither a semantic nor a phonetic cue: i, meaning life

and pronounced /sheng/1, on the top and E , meaning moon and pronounced /yue/4, on

7

the bottom. Thus, although both semantic information and phonological information can
be indicated by the radicals within compound characters, this information is oﬂen
unreliable, and even when it is reliable, understanding it requires parsing that is much
different from the way that English words are parsed.

Models of Word Recognition

Before it is possible to investigate how writing systems affect word recognition
processes of L2 learners, there must ﬁrst be a description of the available models of word
recognition. Psycholinguistic models of word recognition will be discussed in terms of
the major features, strengths, weaknesses, and assumptions that underlie them, and
supporting research will be presented for each.

All models of word recognition incorporate, at some point, three basic elements:
orthography, phonology, and semantics (DeFrancis, 1989; Wang, Koda, & Perfetti, 2003),
and according to the Universal Phonological Principle (Perfetti, 2003; Perfetti, Zhang, &
Berent, 1992), there is a direct relationship between orthography and phonological
activation in the accessing of semantic information in all languages. What has been
under debate for a number of years is the order in which, and the degree to which, these
are processed or activated. Three main types of psycholinguistic models of phonological
involvement in word recognition have emerged: strong visual-route models, strong
phonological-route models, and dual-route models. Although these models will be
discussed here mostly in terms of ﬁrst language word recognition, they also have
important implications for L2 reading in the sense that transfer of L1 word recognition
processes has been shown to occur in second language readers. The issue of transfer will

be discussed later.

Visual Route

In the strong visual-route model of word recognition, the meaning of a word is
accessed directly from the orthography. That is, a written word is viewed and the overall
visual shape of the word is recognized, which then leads to access of the meaning of the
word, that is, lexical access (Figure 2). Unlike phonological models, where phonology is
activated between recognition of the orthography and semantic access, phonological
activation has been proposed by various visual-route models to occur after lexical access,
at the same time as lexical access, or not at all. These are also referred to as direct-route
models because semantic knowledge is accessed directly from the orthography, without
pre-lexical activation of phonology; this type of recognition is also referred to as
addressed recognition.

The biggest obvious advantage of the visual-route model is that the intermediary
step of accessing phonology that phonological routes propose is cutout of the process.
This streamlined method would seem to be more efﬁcient than a phonology-mediated
route of access, and, in fact, some studies (e.g. Brown and Haynes, 1985; Fender, 2003)

have suggested that accessing the lexicon through a direct route does result in faster word

 

Sound

Symbol I

Meaning

 

 

 

 

 

 

 

 

 

 

 

Figure 2. Direct or visual route to meaning (post-lexical phonology).

recognition.

Perhaps the strongest support for visual—route models is found in studies of the
effects of word frequency on word recognition. This is because, according to a strong
phonological-route model, words are assembled piece by piece, and therefore it should
not matter if an entire word is more frequently encountered or not; if anything, word
length or frequency of certain letter combinations should be more important. Baluch and
Besner (1991) conducted a study that supports a visual route of word recognition based
on word frequency effects, investigating word recognition in Persian. Persian was used
for this study because some words include vowel sounds that are speciﬁed in the
orthography, whereas other words include vowel sounds that are not, thus creating a
variation in the degree of orthographic transparency within the single script. Baluch and
Besner found that there were effects of word frequency not only in the words that were
more orthographically opaque, but also the orthographically more transparent words as
well. This evidence was interpreted as showing that even when a script does allow a one-
to-one matching with the phonology (and thus would be fairly easy for the reader to parse
letter by letter), words are still recognized by their overall form, thus resulting in effects
of frequency. Similarly, research has shown that phonology plays a lesser role in
stronger readers than weaker readers (through the existence of so-called ‘sight words’)
(Chen, Flores d’Arcais, & Cheung, 1995; Ehri, 1998) .

If a phonological route were to explain the existence of familiarity effects, it
would need to show that it is the frequency of letter combinations matching with sounds
that would assist in word recognition, rather than the frequency of the entire word (as, in

a strong phonological-route model, the word would be parsed serially). On the contrary,

10

Peereman, Content, and Bonin (1998) found that, for French speakers (a fairly
transparent orthography), spelling regularity (that is, the consistency of the relationship
between phonology and orthography, also known as feedback consistency) did not
correlate with word recognition speed in a lexical decision task.

In addition to word frequency effects, word superiority effects have been cited as
support for visual-route models. Word superiority refers to the phenomenon that actual
words are judged to be words more quickly than non-words are judged to be non-words
(Baluch & Besner, 1991; Tabossi & Laghi, 1992). According to a purely phonological-
route model, as long as non-words contain permissible combinations of letters (e. g. ‘plof’
rather than ‘nplf’), the reader should not know whether a word is a word or a non-word is
a non-word until they reach the end of the word, or at least until they reach a point in the
word where there are no more words that contain a particular combination of letters. If
this were the process used to recognize words, non-words should be recognized faster
than real words because a string of letters would be judged to be a non-word mid-way
through the word, while identiﬁcation as a real word could not be made until the entire
string is processed. Thus, this rapid judgment of a string of text as a word or non-word
requires addressed, whole-word visual-route processing rather than assembled,
phonological-route processing.

Phonological Route

11

In strong phonological-route models, after the orthographic code is recognized,
the related phonological code is activated. This activation then spreads to activate
meaning nodes that match that phonological code. That is, the activation of phonology is
a mediatory step in the recognition of words (see Figure 3). Two hypotheses have been
proposed to explain how phonology is activated during word recognition. In one, the
word is processed serially, activating the phonology of each letter and assembling the
word from these pieces; this process is, therefore, called “assembled phonology”. In the
other type of proposed phonological activation, the word is viewed as a whole, much the
same as the visual-route models propose, except that the phonology of the whole word is
activated at once. Again, this occurs prior to lexical access; this hypothesis is called
addressed phonology. Addressed phonology would seem to solve many of the
discrepancies between visual- and phonological-route models (such as the frequency and
word superiority effects described earlier). However, Lesch and Pollatsek (1998) showed
that, while addressed phonology would solve many problems, this is not how people read.
In this study, readers were shown a prime word (for example pillow), after which they

judged whether the next word (for example ‘bead’) was semantically related to the ﬁrst

 

 

 

 

 

/ Sound
Symbol 1

Meaning

 

 

 

 

 

 

 

Figure 3. Phonology-mediated route to meaning (pre-lexical phonology).

12

word or not. Results showed that readers were sensitive to the similarity of the beginning
of the word with the beginning of a word that is semantically related (to pillow, e.g. bgad
and bed) as well as the possibility of the end of the word to be pronounced the same as a
word that is semantically related (i.e. -ead as in head and the ed in bed). These results
were interpreted as showing that the reader had processed the word piece-by-piece in an
assembled manner, rather than addressed.

Strong phonological-route models are supported most strongly by research using a
homophone interference paradigm, in which the reading of a printed word activates
homophones of that word as well (Luo, 1996; Van Orden 1987). In Van Orden (1987),
native English-speaking participants saw a category prime (such as ﬂower) and had to
judge whether a word (for example rows) ﬁt the category or not. Similarly, in Luo
(1996), participants were presented with a prime (such as wolf) and were asked which of
two words (for example lion and bare) was semantically related to the prime word. In
both of these studies, it was hypothesized that the phonology of the word presented
would, due to the presentation of the prime, activate not only the word that was written
(rows and bare), but also the exemplar word (e. g. rose and bear, respectively) and thus
cause interference, resulting in slower and/or incorrect answers because of this confusion.
Both of these studies did ﬁnd longer reaction times and/or higher error rates with the
homophone words compared to visually similar (and phonologically dissimilar) controls
(for example robe and bean, respectively). Thus, these studies concluded that, in native
English speakers reading in English, the phonology of words was activated prior to

semantic activation (that is, phonological activation was pre-lexical).

13

Supporters of phonological-route models claim that the phonological route
argument makes sense based on the fact that meaning is gleaned from spoken language
using this same process, that of activating the phonological form and then the meaning
(Perfetti, 2003). Considering the fact that in all languages around the world, the learning
of spoken language precedes the learning of written language (Hoosain, 1991) and that,
for nearly all people, encounters with aural/oral language is much more common than
written language (Frost, 1998), it would stand to reason that the human mind is built to
(or at least trained to) recognize meaning through recognition of sound. This is not to say
that all written language has a direct one-to-one relationship with spoken language, but
rather that because our brains already have an efﬁcient system for using phonology to
access the lexicon, it does not make sense to abandon that system in favor of a system
that does not use phonology. At its most fundamental, written language is a derivative of
spoken language, and most certainly not the other way around.

Although the ﬁndings that support phonological-route models do seem to be
robust, the issues of word frequency and word superiority effects still remain. Aaron,

J oshi, Ayotollah, Ellsberry, Henderson, and Lindsey (1999) addressed the existence of
word frequency effects in a phonological-route model by showing that sight-word
reading is strongly correlated with decoding ability; that is, the ability to recognize words
as a whole may be dependent upon a prior ability to phonologically assemble those words.
These ﬁndings are, however, dependent upon people whose decoding skills are weak to
also be weak sight-readers. To the contrary, some studies have included participants who

were strong sight-readers, but weak decoders (Freebody and Byme, 1988). In addition,

14

so far, no study has explained word superiority effects in terms of a phonological-route
model.

It is due to the inability of either of these models to fully explain the word
recognition process that dual-route models have emerged.
Dual Route

In response to the previously-discussed shortcomings of both strong visual- and
strong phonological-route models, some researchers have proposed dual-route models
that incorporate both a visual and a phonological pathway to the lexicon. There have
been two separate interpretations of the term “dual-route” (cf. Van Orden, 1987). One
identiﬁes two separate routes that occur not in parallel, but rather in a serial, cascading
manner, with the result of visual-route processing feeding into phonological-route
processing. In essence, these are only visual-route models with a phonological route
tacked onto the end. On the other hand, most dual-route models specify two separate
routes, occurring in parallel, and it is this type of dual-route model that will be discussed
here.

Dual-route models have the advantage, due to the incorporation of both a
phonological and visual route running in parallel (Figure 4), of being able to account for
both homophone and word frequency effects. Dual-route models are frequently
presented as a metaphorical horse race (e.g. Paap & Noel, 1991) in which a phonological
horse and a visual horse are racing toward a common goal, lexical activation. In most of
these models, the horses do not interact, but rather the fastest horse (route) to reach the
lexicon wins. Dual-route models where the horses do interact have also been proposed

(e. g. Seidenberg, 1985; Seidenberg, 1992). These models think of activation in terms of

15

 

 

Grapheme to
phoneme
7 conversion

Direct access I , ’
to meaning I

Meaning }

 

 

 

Symbol

 

 

 

 

 

 

Figure 4. Dual-route model.

a threshold; information is gained from a combination of phonology and orthography
until a threshold is reached, resulting in lexical activation.
Orthographic Depth Hypothesis

Somewhat of a product of the dual-route models is the orthographic depth
hypothesis (ODH) (Frost, 1994; Frost & Katz, 1992; Frost, Katz, & Bentin, 1987). While
the ODH does incorporate both reliance on visual information and phonological
information, it discusses phonological activation not as a binary question of one route or
another, but rather in terms of degree of activation. The ODH proposes that the extent to
which the written form of a language corresponds with the phonological form determines
the process used to recognize words (Akamatsu, 2005; cf. Frost, 1994). That is, the more
orthographically shallow a language is, the greater extent phonology will be activated
during word recognition; on the other hand, more orthographically deep languages will
involve less phonological activation. In a study by Katz and Feldman (1981), it was
found that word naming depended more on phonology in Serbo-Croatian (an
orthographically very shallow language) than in English (orthographically deeper than
Serbo-Croatian), and word naming depended more on phonology in English than in

Hebrew (orthographically deeper than English). Similar effects have been demonstrated

16

in a number of studies in a variety of languages (Welsh: Ellis & Hooper, 2001; Hebrew:
Geva & Siegel, 2000; English, French, and Spanish: Goswami, Gombert, & De Barrera,
1998; Turkish: Oney & Durgunoglu, 1997; 13 European languages: Seymour, Arc, and
Erskine, 2003)

Regarding the languages involved in this study, the ODH indicates that, for native
speakers of the languages, reading in English would involve a much greater degree of
phonological activation than reading in Chinese. This is due to the existence of GPC
rules in English, and the lack of a reliable phonological cue in Chinese characters. This is
not to say that phonological activation does not occur while reading Chinese characters;
on the contrary, there have been a number of studies by Perfetti and colleagues that
demonstrate phonological activation during reading in Chinese (e. g., Perfetti & Tan,
1998; Perfetti & Zhang, 1995; Tan & Perfetti, 1997, 1998, 1999; Tan, Spinks, Gao, Liu,
Perfetti, Xiong, Stofer, Pu, Liu & Fox, 2000). While these studies do claim that
phonology is activated during word recognition in both English and Chinese, they also
state that whereas word recognition in English can involve cascaded activation of
phonology (which suggests assembled phonology), in Chinese, “phonology awaits the
completion of orthographic processes” (Perfetti et al, 2007, p. 133), suggesting addressed
recognition. The general conclusion of these studies is that, in any language, phonology
is activated to the extent “allowed by the writing system” (Perfetti, Zhang, & Berent,
1992, p.245). This statement would tend to be in agreement with the ODH in that the
ODH also states that the degree of phonological activation depends on the writing system
of the language.

Transfer

17

The degree to which phonological information is used affects more than just the
ﬁrst language. Much research has shown that L2 learners’ word recognition is heavily
inﬂuenced by the way that words are recognized in their L1 (Brown & Haynes, 1985;
Chikamatsu, 1996, 2006; Green & Meara, 1987; Koda 1988, 1989, 1990; Mori, 1998;
Wade-Woolley, 1999; Wang, Koda, & Perfetti, 2003). Such transfer includes the degree
to which phonological information is activated while reading. Koda (1989) investigated
L1 Japanese (a relatively deep orthography that is based on Chinese) learners of English
(which is comparatively much more shallow) and showed that these learners could recall
pronounceable strings of letters no better than unpronounceable strings of letters, and in
fact could sometimes recall the unpronounceable strings better. That is, they did not
make use of the phonological information that was available in order to make the letters
into words (albeit nonsensical words), and instead remembered the visual representation
of the letters. These results were interpreted as being a result of transfer from the L1
recognition process, which, according to the ODH, includes a lesser involvement of
phonology.

Mori (1998) investigated Ll English and L1 Chinese learners of Japanese in a
similar fashion. In this study, participants were presented with pseudoword Japanese
kanji characters, some of which were embedded with hiragana (Japanese syllabary)
characters. In a recognition task, L1 Chinese participants were able to recall an equal
number of pseudocharacters that had embedded phonologically-accessible elements (the
hiragana characters) as those that did not contain such elements. On the other hand, the
L1 English learners remembered a signiﬁcantly greater number of pseudoword kanji that

had embedded pronounceable parts than they did kanji that had no such element, thus

18

showing that the L1 English learners made use the sound element of the characters,
whereas the L1 Chinese did not. This was, again, interpreted to be an effect of the L1
word recognition process being transferred to the L2. Such a strategy would be more
similar to the way Ll English speakers would process English.

Although these studies showed L1 transfer of the use of phonology, they showed
it via ofﬂine recall tasks, which may be different from online reading. Chikamatsu
(1996) also investigated L1 English and L1 Chinese learners of Japanese, and found
similar results to the previous studies, although through a different methodology. This
study involved a script-manipulation task using the syllabic characters of Japanese
(hiragana and katakana). In this task, a number of words that students were more familiar
with seeing in hiragana were written in katakana and vice versa, thereby preventing
visual recognition of the words and forcing use of phonology. She found signiﬁcant
effects of visual familiarity (that is, the script that the words were written in) and also
word length, with familiarity affecting the L1 Chinese group more than the L1 English
group, thus showing that the L1 English learners were not as disturbed by changes in the
visual appearance of the words as long as phonological information did not change. In
addition, word length affected the L1 English group more than the L1 Chinese group,
showing that, for the L1 Chinese group, word length did not make as much of a
difference, which would be the result of addressed visual recognition.

In a study involving Ll Chinese and L1 Korean learners of English, Wang, Koda,
and Perfetti (2003) investigated learners’ relative reliance on phonological and visual
information during word recognition. Chinese was chosen because Chinese orthography,

according to the ODH, should involve a minimal amount of phonological activation,

19

while Korean was chosen because although Korean and Chinese are both syllabaries,
Korean orthography (hangul) is alphabetic in nature and thus, according to the ODH,
should involve more activation of phonology (Cho & Chen, 1999). This study involved a
semantic category judgment task, similar to that developed by Van Orden (1987), in
which, after the presentation of a category prime (such as a ﬂower), there were foils that
were phonologically-similar (e.g. rows) or controls that were orthographically-similar
(e. g. robe) to actual category exemplar words (e. g. rose), thus allowing for a measure of
relative use of phonological information and orthographical information during L2 word
recognition. In addition, to measure reliance on orthographical information, the
orthographic similarity to exemplar words was also manipulated. In this task, L1 Korean
learners of English were found to make more false-positive errors with regard to the
phonologically-similar foils than the orthographically-similar foils, whereas the L1
Chinese learners did not make more errors on one than the other. While this does suggest
that there is phonological activation for the L1 Chinese learners, there is less than for the
L1 Korean learners. This difference was theorized to have been a result of transfer from
the learners’ L1 processing. To conﬁrm these ﬁndings, a phoneme deletion task was also
conducted. In this task, L1 Chinese learners performed more poorly than the L1 Korean
learners, and also made more errors which, although they were orthographically
acceptable, were phonologically incorrect.
Proﬁciency

Many of the studies described above were conducted with intermediate to
advanced L2 learners. Very few studies have been conducted with lower-level L2

learners, and even fewer have been conducted comparing low proﬁciency and high

20

proﬁciency learners. A number of studies did not operationalize proﬁciency at all. Wang,
Koda, and Perfetti (2003) state that the results of previous studies investigating L1
transfer of word recognition skills are “clouded by the lack of attention to the level of
English proﬁciency in the L2 groups,” and further state that it is “important to separate
whether differences arise from the L1 writing system background or from the level of
skill attained in L2” (p. 134). In their study, the majority of participants were recruited
from intermediate ESL courses, with additional participants recruited from advanced ESL
classes and among ﬁrst-year graduate students. In their description of their results, which
support the ODH and L1 transfer, Wang et a1. further state, “Moreover, the logographic
Ll transfer effect on alphabetic L2 learning may occur only at the beginning stages of '
learning. With increasing proﬁciency in English, the effect of phonology on English
word processing will eventually prevail in Chinese L2 readers’ performance” (p.144). As
strong of a statement as this is, proﬁciency was not a variable in this study; rather it was
controlled due to the possibility of effects of proﬁciency on the degree to which
phonology is used during word recognition.

One important element to consider in research investigating changes in word
recognition is whether changes occur due to actual restructuring of the learner’s cognitive
processes, or if changes are simply due to higher proﬁciency learners processing words in
the same way, but faster than lower proﬁciency learners. Due to increased proﬁciency, it
is possible that better experimental results may be due to a speeding up of performance,
rather than actual restructuring of cognitive processes, leading to automatization. It is

important to include analyses of automatization in word recognition studies because

21

when learners’ word recognition processes are automatized, fewer cognitive and
attentional resources are consumed (Segalowitz & Hulstijn, 2005).

This phenomenon was investigated by Segalowitz and Segalowitz (1993) and
Segalowitz, Segalowitz, and Wood (1998). They implemented a statistical procedure
using the learners’ reaction times (RT) and standard deviations of those reaction times
(SDrt) to determine a coefficient of variation (CVrt). In the case of simple speed-up,
mean RT and SDrt decrease proportionally, which results in no change in CVrt. On the
other hand, when improvement in performance is due to increased efﬁciency of
processing (i.e. automatization), SDrt will decrease to a greater extent than mean RT,
thus resulting in a reduced CVrt (Akamatsu, 2007). Incorporating this procedure,
Segalowitz and Segalowitz (1993) using L1 French and L2 English, and Segalowitz,
Segalowitz, and Wood (1998) using L1 English and L2 French determined that learners’
L2 processing was, in fact, restructured as proﬁciency increased. Incorporating this
concept into cross-orthographic word recognition research involving intra-word
processing, Akamatsu (2007) found that, with practice, the word recognition processes of
L1 Japanese learners of English were restructured, although only in the case of low-
frequency words; in the case of high-frequency words, simple speed-up was shown.

Chikamatsu (2006) also took the difference between speed-up and automatization
into account during her investigation of the use of phonology during word recognition of
L1 English learners of Japanese. This study used a similar script manipulation paradigm
as Chikamatsu (1996), described earlier. A lexical decision task using decontextualized
words was used as well as a task where participants read paragraphs containing words in

changed scripts. The changed script forced the participants to recognize words not by

22

their overall visual appearance (a more visual route), but rather to process the words in a
serial manner, assembling the phonology of the word. In these experiments, it was found
that advanced students (fourth semester university students of Japanese) were more
affected by visual unfamiliarity than beginning students (second semester) in the
decontextualized lexical decision task. In addition, this was shown to be a restructuring
of the word recognition process, not merely speed-up, although no statistics relating to
this were reported. These results were interpreted as showing that as learners become
more proﬁcient in a language, the cognitive processes of word recognition adapt to the
L2.

While the previous study seems to suggest that with increased proﬁciency, word
recognition processes are indeed restructured, conﬂicting evidence has also been shown.
Akamatsu (2005) also investigated the effect of proﬁciency on the use of phonology
during the word recognition processes, although using L1 Japanese learners of English.
Similar to Chikamatsu (2006), this study used a type of script manipulation. In this study,
alternating case in English (c. g. aLtErNaTiNg CaSe) was used in a word naming task.
Akamatsu (2005) argues that this paradigm works well as a measure of L1 orthographic
effects because readers from L1 3 with non-phonological orthographies (e. g. Chinese and
Japanese) show a large impact in ability to process the constituent letters of words
(Brown & Haynes, 1985). Following the ODH, the processing of the individual letters of
words would indicate activation of the phonology of those letters (i.e., assembled
phonology). In this experiment, expected results were operationally the opposite of
Chikamatsu (2006), due to the languages involved, although paradigmatically the same.

That is, although case manipulation was expected to have a greater effect on the lower

23

proﬁciency group, which is opposite of Chikamatsu (2006), such a result would indicate
a change in processing to adapt to the script used in the L2, which is the same as
Chikamatsu hypothesized. However, the results showed that the high-proﬁciency
English learners did not perform any better than the low-proﬁciency group. This was
interpreted to mean that the effects of the L1 orthography on L2 word recognition were
so profound that any amount of increase in proﬁciency and/or exposure (the high-
proﬁciency group was also in an ESL context while the low-proﬁciency group was in an
EFL context) could not reduce them (Akamatsu, 2005).

The conﬂicting results of these studies show that more research is necessary in
order to determine the precise effect of proﬁciency on cross-orthographic word
recognition.

Research Questions

The various studies described have informed the present study, and have led to the
following research questions:

1) Does L2 proﬁciency affect L2 readers' processing of the constituent letters of words?

a) If so, do differences in readers' abilities to process constituent letters of words

depend on the frequencies of the words?

b) If so, do differences in readers' abilities to process constituent letters of words

depend on the spelling regularity (i.e., regularity of GPC rules) of the words?
2) When high L2 proﬁciency readers read English words for meaning, do they show
more reliance on phonological information compared to low proﬁciency readers? Do
lower L2 proﬁciency readers show heavier reliance on orthographic information than

high L2 proﬁciency readers?

24

Hypotheses

Regarding the ﬁrst research question, the processing of the constituent letters of
words was chosen as a measure of phonological involvement as previous studies
(Akamatsu, 1999, 2005) used this methodology to measure the same aspect of reading.
Intra-word processing is an area in which L2 readers whose L1 5 have non-phonological
orthographies often have a greater amount of L1 transfer (Brown & Haynes, 1985; Holm
& Dodd, 1996; Koda, 1999; Wang, Koda, & Perfetti, 2003); that is, because their L1
orthography does not allow for the concept of infra-word processing and assembled
phonology, L2 English readers whose L1 orthography is non-alphabetic often have
trouble with this part of reading English words. In the present study, it is hypothesized
that L2 proﬁciency will have an effect on L2 readers’ processing of the constituent letters
of words. Based on the research described previously, a high level of L2 reading
proﬁciency should reduce the effect that the L1 orthography has on L2 word recognition.
It is hypothesized that the effect of the L1 orthography will be greater for lower-
proﬁciency L2 readers than for higher-proﬁciency L2 readers.

Previous research has shown that word frequency does have an effect on word
recognition processes in both L1 and L2 (i.e., more frequent words are more oﬁen
processed via an addressed visual route, whereas less frequent words are more likely to
be processed via an assembled phonological route) (Forster & Chambers, 1973; Hino &
Lupker, 2000); in addition, previous research has shown that words that do not follow
regular grapheme-phoneme correspondence rules (termed exception spelling words in the
present study) are more likely to be processed via a visual route whereas words that do

have regular GPC rules are more likely to be processed via a phonological route (Hino &

25

Lupker, 1998; Plaut, McClelland, Seidenberg, & Patterson,l996). For those reasons, it is
hypothesized that word frequency and spelling regularity may have an effect on the
degree to which L1 orthography affects L2 word recognition.

In response to the second research question, it is hypothesized that high
proﬁciency readers will show more reliance on phonological information than low
proﬁciency readers. This is because it is hypothesized that as proﬁciency increases,
learners’ word recognition processes will be restructured to adapt to the L2 script. In the
context of the present study, according to the ODH, the script of English involves a
greater degree of phonological involvement than Chinese, and as such, high L2 English
proﬁciency readers whose L1 is Chinese will show greater reliance on phonology than
low L2 English proﬁciency readers, and low L2 English proﬁciency readers will rely
more on orthographic information than phonological.

Methodology
General Procedure

This study sought to investigate the effect of L2 proﬁciency on L2 word
recognition processes. This was achieved by measuring readers’ ability to process the
constituent letters of words and readers’ reliance on the phonology of words. According
to the orthographic depth hypothesis, reading in Chinese involves a lesser degree of
phonological activation, whereas reading in English involves greater activation; this
study was concerned with potential differences in the amount of phonological activation
at different levels of proﬁciency. In order to achieve this, participants’ proﬁciency was
measured at the time of testing via administration of portions of the Michigan Test of

English Language Proﬁciency (MTELP). In addition, participants completed a

26

background questionnaire. Two computerized experiments were conducted with all
participants; the order in which participants performed the experiments and the order of
the items within all experimental tasks were counterbalanced.

During recruitment, all participants were given an information sheet (Appendix
A) which gave a general description of the tasks, beneﬁts, and risks associated with
participating in the study, and contact information for the investigators. Because the
study involved some participants whose proﬁciency in English was expected to be low,
an attempt was made to make the form as simple and as clear as possible, while still
including necessary information. In addition, the consent form was exactly the same as
the advertising ﬂyer that participants received prior to participation, and as such,
participants had a chance to read the form at their own pace before coming to participate
in the study. Participants agreed orally to the consent formz. All participants received
compensation of $15 cash for participating in the study, after which4 they signed a
statement saying they had received the compensation. This form was, however, stored
separately from each participant’s data in order to protect participants’ privacy.
Participants were identiﬁed only by a subject ID number on all test materials and
analyses, except the signed statement of receipt of compensation; the subject ID numbers
were used only to correlate data from different data sets.

Each participant attended a single 90-minute session arranged with the author.
During this time, all participants completed: 1) consent form, 2) background

questionnaire (Appendix B), 3) proﬁciency test, 4) Experiment 1(lexical decision task)

 

2 A signed consent form was included in the original Institutional Review Board application. However, the
IRB reviewer recommended that the advertisement ﬂyer be used as the consent form instead of the original
signed form, as the advertisement ﬂyer contained all of the information appropriate for an ‘Exempt’ status
study, and by not having participants sign a consent form, the form would not need to be kept for the
required three years.

27

and Experiment 2 (semantic category judgment task) (the order of the experiments was
counterbalanced), and 5) debrieﬁng form (Appendix C) and compensation for
participation. The following will describe the background questionnaire used as well as
the proﬁciency measure, after which the two experiment will be described in detail.

Background Questionnaire

After agreeing to the consent form, all participants completed a two-page
background questionnaire (Appendix B). Approximately 12 minutes were allotted for
completion of the background questionnaire. The questionnaire “was partially based on
the pretest version of the Language Contact Proﬁle by Freed, Dewey, Segalowitz, and
Halter (2004). The questionnaire included questions about L1, gender, age, and current
and previous L1 and L2 use and exposure.

In addition, the background questionnaire asked about the participant’s use of
pinyin, the method of learning the pronunciation of Chinese characters through the use of
the roman alphabet. This system is used mainly in mainland China, and is taught to
native Chinese-speaking children from a young age (usually when they enter ﬁrst grade
in elementary school) and is used when teaching Chinese characters in school. This
system was adopted by the government of China in 195 8, and has been used in schools
since then (Yin & Felley, 1990). In Taiwan, this system is not used, and instead, the
sounds of Chinese characters are shown by using a speciﬁc set of other Chinese
characters. A number of studies have shown that the increased exposure to the roman
alphabet through pinyin can have an effect on the way Chinese speakers recognize
English words (e.g. Everson, 1998; Tan & Perfetti, 1998; Wang & Geva, 2003). Some

participants from Taiwan did ask about this question, as the system used in Taiwan is also

28

sometimes called pinyin, although the majority of Taiwanese participants did understand
pinyin to mean the romanized variant. The system that is most widely used in China is
called hanyu pinyin, and it may be possible that this confusion could have been
eliminated by using this term instead of only pinyin. This question was purposely left as
only pinyin in order to include other systems of romanization that are occasionally used
throughout China (such as tongyong pinyin).

Proﬁciency Measure

After completing the background questionnaire, participants took the grammar
and reading portions of the Michigan Test of English Language Proﬁciency (MTELP),
form P, acquired from the University of Michigan English Language Institute. This test
was chosen because it has a long history of use in language acquisition research as well
as as a valid and reliable placement test in language programs (Baldauf & Dawson, 1980).
In addition, it was chosen because participants were unlikely to have previous exposure
to the test, as it is not used by Michigan State University to test language proﬁciency, nor
is it accepted by the university as proof of language proﬁciency. It was decided to
administer only these two sections of the test because of time constraints, and the
vocabulary section provided only a minimal amount of information regarding the
participants’ proﬁciency. As participants took two out of three sections of the test, the
total amount of time allowed to take the test was reduced proportionally. The University
of Michigan English Language Institute recommends that a total of 75 nrinutes be
allowed to complete the entire test; therefore, participants in the current study were
allowed 50 minutes, which is two-thirds of the full time to complete two-thirds of the full

test.

29

Participants

All participants were native speakers of Chinese enrolled at Michigan State
University in one of the following: the MA program in Teaching English to Speakers of
Other Languages, the Second Language Studies Ph.D. program, Level 2 or Level 3
classes at the English Language Center (ELC), or Level 2 of a community outreach
English as a Second Language course taught at Michigan State University. Participants
were contacted either by email (MA TESOL and SLS Ph.D. students) or in their classes
(ELC and community outreach course students), with institution and instructor approval.
While age restrictions were not reported in many of the studies closely related with this
study (Akamatsu, 2005; Chikamatsu, 2006; Wang, Koda, & Perfetti, 2003), the
participants included in those studies were undergraduate university students, who would
most likely be from a much more limited age group. Because the present study included
a number of graduate students as well as community members taking the community
outreach course, an age restriction was instituted. Following a number of
psycholinguistic studies (e.g. McCann, Folk, & Johnston, 1992; Rumsey, Horwitz,
Donohue, Nace, Maisog, & Andreason, 1997), all participants included in this study were
between 18 and 40 years of age. All participants were from either mainland China or
Taiwan. Participants’ status as native speakers of Chinese was conﬁrmed by the

background questionnaire. All participants received $15 cash for their participation.

30

In total, 34 participants completed all parts of the study. Although participants
were recruited from populations that were expected to yield distinct high and low
proﬁciency groups, upon inspection of the results of the proﬁciency measure, it was
discovered that this was, in fact, not what was shown; rather, proﬁciency scores were
fairly continuous between the lowest score of 19 and the highest score of 55 (of a
maximum of 60) (Figure 5). Because this study sought to investigate the effect of
proﬁciency differences on word recognition, it was imperative that proﬁciency
differences be evident between the two groups. For this reason, it was decided to
implement an extreme groups approach (EGA). The 34 participants were divided into
thirds (a high proﬁciency group, a medium proﬁciency group, and a low proﬁciency
group), which yielded 11.3 participants per group. Because portions of this study
implemented two word lists, for the purpose of counterbalancing it was necessary that
there be equal numbers of participants who saw each list in each group, thus necessitating
an even number of participants in each group. Therefore, groups were limited to 10

participants each. The 10 participants with the highest total MTELP scores were labeled

 

 

 

 

 

Figure 5. MTELP scores for all 34 participants (maximum possible score of 60).

31

as the high proﬁciency group; scores in this group ranged from 47 to 55 out of 60. The
10 lowest scores were labeled as the low proﬁciency group; scores in this group ranged
from 19 to 29. Only the outer tertiles were included in the analysis. Although the use of
EGA is controversial because it produces artiﬁcially inﬂated effect size estimates, it is
beneﬁcial when the exact relationship between the variables involved is unknown, and
the goal of the research is to determine whether an interaction exists, regardless of the
magnitude (Preacher, Rucker, MacCallum, & Nicewander, 2005); it is for this reason that
it was felt that EGA was justiﬁed in the context of the current study. However, because a
result of the use of EGA is artiﬁcially inﬂated effect size estimates, these will not be
reported in the present study.

In addition to MTELP scores, as a part of the background questionnaire, all
participants rated their own abilities (in both Chinese and English) in the categories of
listening, speaking, reading, writing, and grammar (Table l). Self-evaluations have also
been collected in other psycholinguistic studies (e. g. Jiang, 2007; Midgley, Holcomb, &
Grainger, 2008), and can often provide insight into or corroboration of tested measures.
While all participants rated themselves similarly as native speakers of Chinese, the self-
ratings of the high-proﬁciency group were signiﬁcantly higher in all categories (listening,
speaking, reading, writing, and grammar) than the self-ratings of the low-proﬁciency
group.

The number of male and female participants was the same between proﬁciency
groups (female n = 8, male n = 2). There were more pinyin users in the low-proﬁciency
group (n = 8) than the high proﬁciency group (n = 6). More participants in the high

‘ proﬁciency group reported using some English at home (n = 4) than in the low

32

proﬁciency group (n = 1), and the same was seen in terms of pre-university education in
English (high proﬁciency n = 3; low proﬁciency n = 1). The mean age was higher for
the high proﬁciency group than for the low proﬁciency group although this difference
was not signiﬁcant (Table 2). Length of residence in English speaking places (including
Michigan and any other primarily English-speaking locations) was signiﬁcantly greater
for the high proﬁciency group than for the low proﬁciency group, although it seems
logical that those who have been immersed in the L2 environment would have higher

proﬁciency levels.

33

3.3 8.8 and 86 and 8.2
and 8.8 62v 8.2 and on:

 

.. * Lunch.

*eamﬂvwom 1185580
3.80m mag

._0>0_ sovnﬁa 50.8% tmuqaommdwmm 0.83 2305 @980 .13
._0>0_ EVE as «gnaw mug—Maw? 803 338 955 a ..
_0>0_ EVE? 50.8% bun—85:»? 0.83 885 955 ..

9808A $8.32 A333 $3.32 amazed 33
5.386 A8534.” 83:3 8:325 3.586 as:
£5.85 tween? inseam Balsam twins: 880

3.50m xoaomowoa votomcuum—ow «can

53me uxzacﬁ EESREAF «Eogﬁh 33333. V50» .33:
ﬂ 03d.

34

Table 2
Means (and standard deviations) of participant age and length of residence in English-

speaking locations

 

 

Age in years Residency in months*
High Proﬁciency 28.40 (4.90) 25.83 (19.37)
Low Proﬁciency 23.10 (7.19) 8.60 (4.97)

 

* Group means were signiﬁcantly different at p < .05 level

Experiment 1

Experiment 1 implemented a lexical decision task in order to address research
question 1: ”Does L2 proﬁciency affect L2 readers' processing of the constituent letters
of words?”, as well as the sub-questions: “If so, do differences in readers' abilities to
process constituent letters of words depend on the frequencies of the words?” and “If so,
do differences in readers' abilities to process constituent letters of words depend on the
spelling regularity of the words?” In a lexical decision task, a participant is presented
with a single word and must respond as quickly and accurately as possible whether that
word is a word in English or not. Research question 1 was addressed by manipulating the
case in which the stimuli words were written, which forced readers to process constituent
letters in the stimulus words manipulated in this way. The two sub-questions were
addressed by manipulating the frequency and spelling regularity of the stimuli words
(also described below).
Materials

Participants completed a lexical decision task that contained a total of 160 items,

including 80 critical words and 80 distractor words (Appendix D); in this task,

35

participants were to choose whether the word presented was a real English word or not.
Case alternation (e. g. cAsE aLtErNaTiOn) was implemented in this experiment in order
to force readers to process words in an assembled, serial way while reading the
manipulated words. Half of the words were presented in this visually unfamiliar form,
and half of the words were presented in a visually familiar form (all lower case). This
methodology has been used in a number of studies investigating L2 word recognition
(Akamatsu, 1999, 2003, 2005; Kinoshita, 1987) and L1 word recognition (Ehri & Wilce,
1982), as well as other word-shape effects on reading (e.g. Fisher, 1975). As in
Akamatsu’s studies, by presenting the words in a visually unfamiliar form, visual cues
become inaccessible to the reader, making a visual route of word recognition impossible.
The serial parsing of a word is theorized to activate the phonology of each letter (Paap,
Newsome, & Noel, 1984). Word case was included as a within-subject independent
variable.

Two word lists were used in this experiment; each list contained the same words,
but words that were in lower case in one list were in alternating case in the other list, and
vice versa. Half of the participants in each group saw List 1 and half of the participants
saw List 2. This was to avoid any effect of the words used; that is, if only one list were
used, it is possible that the list of lower case words and the list of alternating case words
may have had inherent differences. By using two word lists containing the same words
but in different cases, the possibility of inherent differences is removed because the lower
case words for one group and the alternating case words in the other group have exactly

the same attributes, except for the case in which they are presented.

36

In addition, previous studies have shown that word frequency and spelling
regularity have signiﬁcant effects on word recognition in both L1 English speakers (e. g.
Hino & Lupker, 2000; Jared, 2002) as well as L2 English speakers (Akamatsu, 1999),
and as such, those factors were also included as word-based variables.

Differences between the critical words were operationalized in terms of frequency,
spelling regularity, and letter case; length and word class were controlled (Table 3).
Frequency was determined using Kucera and Francis (1967). Because the Kucera and
Francis (1967) corpus was compiled from native—speaker data, and it was assumed that
English learners most likely had had different exposure to language than native speakers,
stimuli were selected based ﬁrst on frequency, and secondly on their appropriateness to
L2 speakers of English based on the author’s personal experience as an ESL/EFL teacher.
The stimulus words that were used in the experiment, however, still follow the frequency
counts determined using Kucera and Francis (1967). Frequency between the high
frequency conditions and the low frequency conditions was signiﬁcantly different as
shown by t-test, t(39) = 11.99, p<.0001.

Spelling regularity was determined using Bemdt, Reggia, and Mitchum’s (1987)
study of grapheme-phoneme correspondence probability as a guide. The regular spelling
words had grapheme-phoneme probabilities that were above 70% (or the highest
probability for that grapheme), whereas the exception spelling words contained
grapheme-phoneme probabilities of 30% or less, although most were below 10%.
Although a number of studies have used Venezky (1970) to determine spelling regularity
(e.g. Hino & Lupker, 2000; Holm & Dodd, 1996), Venezky’s rules are not based on

quantitative evidence (Kessler & Treiman, 2001), and as such, Berndt et al. (1987) was

37

used instead (and therefore the term “exception spelling” is used rather than “irregular
spelling”). In addition, Akamatsu (1999, 2005), on which this study is based, also used
Bemdt et a1. (1987) to make spelling regularity classiﬁcations.

Words in each condition were matched for length, with word length differences
being non-signiﬁcant between each of the word factor conditions. Word class was
matched within frequency levels (and therefore between spelling regularity levels),
however there were differences in word class numbers between frequency levels; the
rarity of low frequency adjectives and adverbs that are between three and ﬁve letters in
length and that that show extreme spelling regularity patterns contained in the Kucera and
Francis corpus necessitated this difference.

In addition, a number of the stimuli from Akamatsu (1999, 2005) were, with
permission, included. The stimuli set for those studies had also been categorized using
Bemdt et a1. (1987) for spelling regularity and Kucera and Francis (1967), among other
corpora, for frequency. However, because the current study necessitated twice the
number of stimulus words as Akamatsu’s, the stimuli from that study were mixed in
among other stimuli determined by the method described above. The stimuli from
Akamatsu’s studies were included with their frequencies as shown in Kucera and Francis
(1967)

Non-word distractors were constructed by changing the ﬁrst 1-2 letters of
common English words not used elsewhere in the stimuli. Because of this, the spelling
patterns of the non-word distractors followed common English spelling patterns. Half of
the non-word distractors were presented in alternating case and half were presented in

lower case.

38

Table 3

Descriptive statistics for frequency, length, and word class for each condition

 

Frequency (OPIVD Length iletters)

 

 

Condition M SD M SD Nouns Verbs Adj. Adv.
High-Reg 401.70 249.74 4.15 0.49 8 6 4 2
High-Exc 382.05 162.69 4.25 0.44 8 6 4 2
Low-Reg 10.45 6.98 4.20 0.62 12 6 2 0
Low-Exc 10.35 6.46 4.25 0.72 12 6 2 0

 

Note: OPM = Occurrences per million, as determined using the Kucera and Francis
(1967) corpus; High and Low refer to frequency categorization, and Reg and Exc refer
to spelling regularity categorization as either regular or exception spelling, according to
Bemdt et a1. (1987)
Procedure

The task was conducted using the DMDX software package (Forster & Forster,
2003) on Windows XP-based computers. Participants were seated approximately 50cm
from the monitors, with blinder walls on either side of the participant. Stimuli were
presented in 24-point font in black text on a white background. After reading the
instructions (Appendix E), each participant completed ten practice items, of which ﬁve
were presented in lower case and ﬁve were presented in alternating case, and ﬁve were
real words and ﬁve were non-word distractors (all practice items were, however,
presented in a randomized order). Before each stimulus item, participants were presented
with a ﬁxation mask, which consisted of ﬁve hash marks (#####), for 500ms.

Participants then made a decision as to whether they thought the presented word was a

real word in English or not. Participants responded using their left and right index ﬁngers.

39

To respond that a word was a real English word, participants pressed the ‘L’ key on the
keyboard (covered with green paper to indicate a positive response), and to respond that a
word was not a real word in English, participants pressed the ‘D’ key on the keyboard
(covered with red paper to indicate a negative response). Both response time and
accuracy were recorded for each word. Participants were given a break after half (80) of
the items. Participants were asked about their level of fatigue after each task and
informed that they could take a break if they needed to; however, no participants
responded that they were fatigued.
Results

The analysis utilized a 2 x 2 x 2 x 2 (two levels each of participant L2 proﬁciency,
stimulus word frequency, stimulus word spelling regularity, and stimulus word case)
repeated measures factorial design. Data were analyzed both by participant and by item.

In the participant analysis, L2 proﬁciency was the between-subjects independent
variable and the word conditions (frequency, regularity, and case) were within-subject
independent variables. In the item analysis, L2 proﬁciency and case were within-item
factors, and spelling regularity and frequency were between-item factors. Dependent
variables in both analyses were reaction time and error rate.

In order to preserve data, responses greater than two standard deviations from
the mean of each participant were replaced with a value equal to two standard deviations
from the mean rather than discarded. Trimming data at two standard deviations from the
mean of each participant in each condition is a common procedure in word recognition

research (e.g. Holmes, 1998; J iang, 1999). The trimming of data affected 2.34% of the

40

data. The analyses of the reaction time data were based only on correct responses.
Means for reaction times and error rates are presented in Table 4.

Reaction Times

Overall, there was a main effect of proﬁciency group, both by participant, F(1,18)
= 11.30,p < .01, and by item, F(1, 36) = 176.03,p < .001, showing that, on the whole,
the more proﬁcient readers recognized words faster than the less proﬁcient readers.

In addition, there was a main effect of case, with participants recognizing lower
case words signiﬁcantly faster than alternating-case words, by participant, F (1,1 8) =
59.72, p < .001, and by item, F(1,36) = 77.74, p < .001. There was also a main effect of
word frequency; high frequency words were recognized more quickly than low frequency
words, both by participant, F(1,18) = 99.40, p < .001, and by item, F (1,36) = 52.13, p
< .001.

The main effect of spelling regularity was not signiﬁcant by participant or by item.
However, there was a signiﬁcant three-way interaction between spelling regularity, case,
and proﬁciency that was shown only in the item analysis (Figure 6), F (36) = 4.98, p < .05.
Post-hoe paired-sample t-tests showed that the effect of spelling regularity between the
two proﬁciency groups was signiﬁcant only in the lower case condition, t(l 9) = 2.46, p
< .05.

There was a signiﬁcant interaction between frequency and proﬁciency group by
participant, F(1,18) = 8.59, p < .01 and by item, F(1,36) = 26.68, p < .001, showing that
word frequency had more of an impact on the low proﬁciency group than it did on the

high proﬁciency group.

41

Table 4

Means (and standard deviations) of reaction times and error rates

 

 

 

Reaction Times (ms) Error Rates (%)
High Prof. Low Prof. High Prof. Low Prof.
Condition M (SD) M (SD) M (SD) M (SD)

Alternating case (visually unfamiliar)
High, Reg. 888.62 (142.7) 1064.00 (157.0) 5.00 (6.3) 3.12 (4.8)
High, Exc. 854.19 (132.3) 996.88 (146.2) 1.88 (5.4) 4.64 (5.5)
Low, Reg. 1176.75 (214.4) 1392.75 (336.0) 10.07 10.6) 32.63 (21.2)
Low, Exc. 1072.12 (190.0) 1371.25 (240.3) 5.07 (6.4) 35.07 (17.5)
Lower case (visually familiar)
High, Reg. 727.88 (110.2) 775.44 (116.9) 0.62 (2.5) 0.00 (0.0)
High, Exc. 686.88 (77.5) 818.75 (121.6) 0.00 (0.0) 1.88 (5.4)
Low, Reg. 903.81 (108.1) 1023.06 (130.3) 10.38 (10.7) 21.25 (16.3)
Low, Exc. 859.12 (151.1) 1068.69 (131.3) 9.38 (16.5) 28.53 (12.6)

 

Note: ‘High’ and ‘Low’ refer to frequency as determined using Kucera and Francis
(1967); ‘Reg.’ and ‘Exc.’ refer to regular spelling and exception spelling, as determined

using Bemdt et al. (1987).

Most directly related to research question one, however, was the interaction

between proﬁciency and case. The results did reveal a signiﬁcant interaction between

case and proﬁciency group, by participant: F (1,1 8) = 5.47, p < .05, and by item: F(1,36)

= 11.85, p < .005 (Figure 7). However, this may be qualiﬁed by a signiﬁcant three-way

42

interaction between proﬁciency group, spelling regularity, and case, signiﬁcant in the
item analysis, F (1,36) = 4.98, p < .05, but non-signiﬁcant in the participant analysis.
Post-hoe paired-sample t-tests showed that the effect of alternating case text between the
two proﬁciency levels was signiﬁcant in the regular spelling word condition, t(19) = -
4.01, p < .001 , but non-signiﬁcant in the exception word condition, t(19) = -.875, p > .3.
This difference can be seen in Figure 8.

Even though the interaction between proﬁciency and case was signiﬁcant, the
question remains about whether this effect is due to speed-up (i.e., more proﬁcient
readers just read faster), or whether actual restructuring is occurring. As discussed earlier,
Segalowitz and Segalowitz (1993) developed a statistical measure of automatization to
investigate differences between speed-up and restructuring. This method has been
subsequently been used in a number of studies investigating restructuring of cognitive
processes (e.g. Akamatsu, 2007; Chikamatsu, 2006; Fukkink, Hulstijn, & Simis, 2005;
Jiang, 2007). According to Fukkink et a1. (2005), “In the case of true automatization. ..
component processes become more routinized or are eliminated altogether. Hence, not
only will mean RT and SDrt reduce, but so will CVrt” (p.57). CVrt is the coefﬁcient of
variation of the reaction times, which is found by dividing the standard deviation of the
reaction time by the mean reaction time. The data from the present study are presented in
Table 5. The mean reaction time, SDrt, and CVrt all decrease from low proﬁciency to
high proﬁciency readers. Paired-sample t-tests on the CVrt of low proﬁciency and the
CVrt of high proﬁciency participants showed signiﬁcant differences both in the case of

exception recognition process, and not merely speed-up.

43

Lower Case

 

 

 

 

 

 

950 Spelling Regularity
— Regular
' - - Exception
’5?
E, 900-
or
.E
[.—
c
.2 850-
o
to
m
n:
E
a) 800‘ ,I'
2 ,1
.l
750“
If F
High Proﬁciency Low Proﬁciency
Proficiency Group
Alternating Case
1300‘ Spelling Regdanty
— Regular
' ' ' Exception

—l
N
O
9

1 000‘

Mean Reaction Time (ms)
3
C?

 

900'

 

 

 

TI 1
High Proﬁciency Low Proﬁciency
Proficiency Group

Figure 6. Interaction between spelling regularity and proﬁciency within case.

44

Table 5

Summary of means, standard deviations, and coefficients of variation for alternating case

stimuli

 

 

Mean RT SDrt CVrt
Both Spelling Conditions
Low Proﬁciency 1343.26 688.81 .4941
High Proﬁciency 951.48 291.85 .3051
Regular Spelling Conditions
Low Proﬁciency 1369.11 674.31 .471 1
High Proﬁciency 974.56 310.15 .3181

 

45

 

Lower Case
EAlternating Case

1 200.00-

1 ,000.00‘

800.00—

Mean

600.00—

400.00‘

200.00“

 

 

 

High Proﬁciency Low Proﬁciency
Proficiency Group

Figure 7. Means of lower case and alternating case reaction times by proﬁciency

group.

46

Regular Spelling

 

 

 

 

 

 

 

 

1300‘ Case
—Altemating Case
A ' ' ' Lower Case
w _.
E, 1200
0
.§
l- 1100-
c
o
"is
o 1000-
t!
r:
m
g 900‘
_____ e
. — — ......
800—
l I
High Proﬁciency Low Proﬁciency
Proficiency Group
Exception Spelling
1200‘ Case
—Alternating Case
‘ ' ' Lower Case

E 11001
g 1000-
E’
g ,.
:8 900'
to
.E
LY} soo-

.,

700‘
l I
High Proﬁciency Low Proﬁciency
Proficiency Group

Figure 8. Interaction between case and proﬁciency within spelling regularity conditions.

47

Accuracy

Similar to the results of the reaction time analysis, there was a main effect of
proﬁciency group when participants’ error rates were analyzed both by participant,
F(1,18) = 63.63,p < .001, and by item, F(1,36) = 27.58,p < .001. In addition, there
was a main effect of case when error rates were analyzed by participant F (1,1 8) = 10.45,
p < .01, and by item F(1,36) = 6.75, p < .05. The main effect of frequency was also
signiﬁcant by both participant, F(1,18) = 93.89, p < .001, and by item, F(1,36) = 46.44, p
< .001.

The interaction between frequency and proﬁciency group was also signiﬁcant, by
participant: F(1,18) = 37.82,p < .001, by item: F(1,36) = 21.81,p < .001. Post-hoc
paired samples t-tests showed that the lower proﬁciency group showed signiﬁcantly
different error rates between high (2.89%) and low frequency words (31.70%), t(9) = -
6.87, p < .001; this factor did also have an effect on the high proﬁciency group (1.50%
for high frequency and 8.03% for low frequency), although this difference did not reach
signiﬁcance (Figure 9).

The interaction between proﬁciency group and spelling regularity was also found
to be signiﬁcant, by participant: F (1,18) = 4.84, p < .05, and by item: F(1,36) = 4.14, p
< .05. Post-hoc paired-sample t-tests revealed a signiﬁcant difference between the high
proﬁciency readers’ error rate for regular spelling words (M = 7.5%) and that for
exception spelling words (M = 2.0%), t(9) = 3.50, p < .01. This difference, M = 16.2%
for regular words, M = 18.4% for exception words, was not signiﬁcant in the case of the

low proﬁciency readers, t(9) = -.756, p > .4.

48

Interaction between proficiency and frequency

 

 

 

 

 

40— Proﬁciency Group
— High Proﬁciency
' ' ' Low Proﬁciency
w
c ,o
8 30-
E ,'
E ,’
.E ,’
g 20- ’I
2
'o ,I
9 ,
to .'
.E _
7;, 10 ,
UJ
0—
l I
High Low
Frequency

Figure 9. Error rate interaction between proﬁciency group and frequency.

The interaction between proﬁciency group and case was signiﬁcant in the item
analysis, F (1,36) = 4.51, p < .05; in the participant analysis, however, it approached but

did not reach signiﬁcance, F(1,18) = 4.20, p = .055. In addition, none of the three-way or

49

four-way interactions were signiﬁcant (all F(1,18) < 1.38, p > .2 by participant, F (1,36) <
0.84, p > .3 by item).
Discussion

The lexical decision task was designed to measure the effect of proﬁciency on
readers’ ability to process the constituent letters of words (i.e., in response to Research
Question 1). The processing of individual letters in words is a process that does not exist
in the native language of the participants, Chinese, but is prevalent in English. In
addition, the processing of constituent letters in words can be seen as a measure of
assembled phonology, which is theorized to be the primary way that words are
recognized in English (Lesch & Pollatsek, 1998); however, assembled phonology is also
a concept that does not apply to Chinese orthography. Therefore, this experiment sought
to determine whether the processing of constituent letters and assembled phonology
while reading L2 English are developmental; that is, whether they change along with L2
proﬁciency.

Based on the literature, it was hypothesized that L2 proﬁciency would have an
effect on L2 English word recognition in terms of processing the constituent letters of
words. Overall, the results of this experiment support this hypothesis, thus answering
afﬁrmatively to the ﬁrst research question. Both the reaction time results as well as the
accuracy results showed that alternating case text had a signiﬁcantly greater effect on the
low proﬁciency readers than on the high proﬁciency readers. That is, the high
proﬁciency readers showed less interference from loss of visual word shape information
and the processing of the individual letters of words. In the item analysis of the reaction

time data, however, this effect was constrained to the words that were regularly spelled

50

(research question 1b), and did not appear in the words that had exception spelling. In
the analysis of all alternating case stimuli as well as regular spelling alternating case
stimuli, restructuring of cognitive processes, rather than mere speed-up, was shown.
Through the analysis of the reaction time data, SDrt, and CVrt, it was shown that CVrt
decreased signiﬁcantly, indicating restructuring.

The interaction between proﬁciency, case, and spelling regularity may be
consistent with the theory that phonology was activated while the alternating case words
were parsed. In regular spelling words, grapheme-phoneme correspondence rules are
more readily accessible to the reader. If an effect of case were found across both regular
and exception word conditions, then it could be said that readers were not affected by
these differences in the relationships between the letters and their sounds, since spelling
regularity is a measure of the relationship between letters and their sounds. This three-
way interaction was, however, only found in the item analysis of the reaction time data,
and in none of the other analyses. In all other analyses of the reaction time data, there
was a two-way interaction between proﬁciency and case, without inﬂuence ﬁ'om spelling
regularity.

The absence of a three-way interaction between proﬁciency, case, and ﬁequency
shows that the effect of proﬁciency on reading constituent letters of words did not depend
on the ﬁequency of the words. In answer to research question 1a, frequency did not have
an effect on the relationship between case and proﬁciency group.

These ﬁndings suggest that word recognition processes are, in fact, restructured as
L2 readers gain in L2 proﬁciency, perhaps in a number of ways. The results of this

experiment may provide an extension to previous research regarding intra—word

51

processing/decoding (e.g., Brown & Haynes, 1985; Holm & Dodd, 1995; Koda, 1999;
Wang, Koda, & Perfetti, 2003) to show that intra-word processing can develop together
with proﬁciency. In addition, assuming that the processing of the constituent letters of a
word does lead to assembled phonology during word recognition as theorized, these
ﬁndings may also demonstrate that the degree of phonological activation during visual
word recognition may be developmental as well.

In addition, the signiﬁcant interaction between proﬁciency and spelling regularity
found in the error rate analysis and the three-way interaction between proﬁciency group,
spelling regularity, and case found in the reaction time analysis may also have relevance
to the issue of activation of phonology while reading. The low proﬁciency group showed
slower reaction times for the exception words than the regular words (as was expected
based on the literature) and no signiﬁcant difference in accuracy between the regular
spelling words and the exception spelling words (found in the error rate analysis),
suggesting a lack of sensitivity to spelling-sound relationships. In effect, the low
proﬁciency readers were either reading words the same, no matter their differences in
grapheme-phoneme correspondence regularity, or were possibly affected by the
compounding of weaker infra-word processing abilities with exception spelling of words.
The fact that the high proﬁciency readers took longer to read words with regular GPC
rules and made a signiﬁcantly larger number of errors on these words was unexpected, as
it is typically exception words on which readers make more errors. However, this result
may be evidence of a sensitivity to grapheme-phoneme correspondence rules that the
high proﬁciency readers have developed, and which the low proﬁciency readers do not

have. It is possible that because of their larger vocabulary and greater exposure to printed

52

language, high proﬁciency readers have identiﬁed which words need to be processed in
an addressed manner, and which words can be processed in an assembled manner; i.e.,
high proﬁciency readers processed exception words in an addressed manner because of
the inconsistent GPC rules contained within them, while they processed words with
regular GPC rules in an assembled manner. Assembled processing typically takes longer
than addressed processing (Fender, 2003), which would explain why the assembled
(regular spelling) words had longer reaction times for the high proﬁciency readers.

In addition, other robust main effects and interactions are also worth noting. In all
analyses, there were signiﬁcant main effects of proﬁciency group. This shows that, in
general, high proﬁciency readers read words more quickly and more accurately than low
proﬁciency readers did. As was predicted, word frequency also had a large effect on both
groups. However, both the reaction time analyses and the error rate analyses showed that
frequency had an even greater effect on the low-proﬁciency group than the high
proﬁciency group. This is to be expected, as higher-proﬁciency readers would tend to
have larger vocabularies, which would typically include greater exposure to a larger
number of low-frequency words, than lower-proﬁciency readers would.

Case also had a major effect on both groups. Even though alternating case had a
signiﬁcantly greater effect on the low proﬁciency readers than the high proﬁciency
readers, the effect on the high-proﬁciency readers was signiﬁcant as well. This may
show that although readers’ processing of the constituent letters improves with
proﬁciency, there is still an L1 effect even at higher L2 proﬁciency levels. This effect
could, however, also be a task effect in that it can be assumed that even native English

speakers would slow down to some extent when reading text in alternating case versus all

53

lower case, no matter their ability to process the constituent letters of words. Text in
alternating case introduces an added effect in that letters are encountered in places where
they are not expected to be (that is, capital letters are in the middle or at the end of words).
Thus, the ‘translation’ of the capital letters into their lower case counterparts may also
introduce a certain degree of latency that is unavoidable, and most likely would occur
with native speakers of English as well.

Although this measure of the processing of constituent letters of words gives a
good approximation of the activation of the phonology of words, it does so in a somewhat
indirect way. In addition, lexical decision tasks may involve a minimal amount of
semantic access, as it may be possible to make a decision as to a string of letters’ status as
a word or not without activating the meaning of the word. For that reason, a second
experiment was conducted that measures reliance on phonological information in a more
direct manner and involves explicit access to meanings of words, although without the
ability to incorporate the word factors investigated in the lexical decision task.

Experiment 2

Experiment 2 addresses the second research question: “When high L2 proﬁciency
readers read English words for meaning, do they show more reliance on phonological
information compared to low proﬁciency readers? Do lower L2 proﬁciency readers show
heavier reliance on orthographic information than high L2 proﬁciency readers?”

Experiment 1 used text in alternating case to investigate the effect of visual
familiarity; as in previous studies using the same paradigm, these tasks were designed
based on the assumption that visually-unfamiliar words (altemating-case words, in the

above experiments) would be processed in an assembled manner, which would activate

54

the phonology of the constituent letters. However, lexical decision tasks do not provide a
way to compare relative reliance on visual and phonological processing, but instead force
the reader to process the constituent letters of words. There may be a difference between
a reader’s ability to read a certain way when forced, and their tendencies when they
actually read. In addition, Experiment 1 may show changes in how readers process the
constituent letters of words, but there is the possibility that high-proﬁciency readers may
have been less affected by the alternating case because of greater familiarity with the
letters themselves. Therefore, a third task was also conducted that attempted to gauge
how much participants relied on phonology during word recognition without forcing
them to process words in a speciﬁc way. This task is based on a task developed by Van
Orden in 1987 for use with L1 English speakers, and used subsequently by Wydell,
Patterson, & Humphreys (1993) with L1 Japanese, and by Wang, Koda, and Perfetti
(2003) in an L2 (English) context with L1 Chinese and L1 Korean. This experiment is a
partial replication of a part of Wang et al. (2003) using the same stimuli, but with
different populations (and therefore a different between-subjects independent variable).
Participants

The same extreme groups of participants that were included in Experiment 1 were
included in Experiment 2 as well.
Materials

The stimuli from Wang, Koda, and Perfetti (2003) were used, with permission, in
this experiment. The two contrasts that were included in the stimuli were phonological
similarity and spelling similarity (Table 6). Phonological similarity had two levels:

phonologically similar (homophonic) foils of category exemplar words, and

55

orthographically similar controls. Spelling similarity also had two levels: similarly
spelled, and less similarly spelled. That is, half of the phonological foils and
orthographic controls were spelled similarly to the exemplar word that was being primed
for, while the other phonological foils and orthographic controls were words with
spellings that were less similar to the word that had been primed for. Therefore, in total,
there were four experimental conditions: 1) similarly spelled phonological foils (e.g.
“feat” for the category “a body part”), 2) similarly spelled orthographic controls (e. g.
“feel”), 3) less similarly spelled phonological foils (e. g. “rows” for the category “a
ﬂower”), and 4) less similarly spelled orthographic controls (e.g. “robe”). Stimuli
consisted of 9 category words or phrases for similarly spelled words, and 8 categories for
the less similarly spelled words. Categorizations of “more” or “less” similarly spelled
were made by Wang et a1. (2003). All stimuli are listed in Appendix F.

When the stimuli were developed by Wang et a1. (2003), the stimuli were also
judged by ESL instructors as to the words’ familiarity to ESL learners. This was because
of effects that have been found in terms of frequency in previous studies. Wang et al.
chose phonological foils and orthographic controls of moderate familiarity on the basis
that more familiar words would be so-called “sight words”, which would typically be
recognized with a lesser degree of phonological activation, and less familiar words would
be more likely to be recognized with a higher degree of phonological activation. It can

be assumed that the familiarity judgments of the ESL instructors in that study, if

56

accurate for the participants in Wang et al. (2003), would most likely also be accurate for
the population included in the present study.

In total, all participants completed 103 trials, of which 51 were critical trials and
52 were distractors. Distractors were included so that participants would not guess the
purpose of the study because a high number of stimuli were similar to actual category
words. Of the 51 critical trials, 17 were category exemplar words, 17 were homophones
of the exemplar words, and 17 were words that were orthographically similar to the
exemplar words. Of the 52 distractors, 34 were “Yes” trials (i.e. words that did ﬁt the
category) and 18 were “No” trials. As in Wang et a1. (2003), this was in order to offset
the larger number of “No” trials among the critical trials. Participants completed 12
practice items before the real test began, receiving feedback on the ﬁrst two. This task
was also presented using the DMDX software package (Forster & Forster, 2003).
Procedure

Stimuli were presented on a Windows XP-based computer in black text on a
white background in 24-point font; participants were seated approximately 50cm from

the monitor with blinder walls on either side. Instructions are in Appendix G.

Table 6

Examples of phonological and orthographic similarity

 

 

Category Prime Phonological Foil Orthographic Control
Similar A body part feet feat
Less Similar Part of a person’s face knows snobs

 

57

 

Fixation point (+)
(500ms)

   
  

 

 

 

 

Category stimulus
presented
(1500ms)

 

 

 

 

Target word
presented

 

Response

 

 

Fixation point (+)
(500ms)

 

 

 

Figure 10. Paradigm used in Experiment 2.

Participants were ﬁrst presented with a ﬁxation point (+) for 500ms, after which they
were shown a category stimulus for 1500ms. Next, the category prime disappeared, and
a target word (or a distractor) appeared. Participants then decided whether the target
word (or distractor) ﬁt the category or not by pressing the red (N 0) or green (Yes) keys
(described above). After making a decision, participants were again presented with a
ﬁxation point (+) for 500ms before presentation of the next category prime. Figure 10
shows a ﬂowchart of the procedure. All times are consistent with Wang et al. (2003).

It is theorized that if a participant makes more mistakes (or has higher reaction
times) on the phonological foils than on the orthographic controls, then the reader is
relying substantially on phonology while reading for meaning. If, on the other hand, a
participant makes more false-positives on more similarly spelled foils than less similarly
spelled foils, then the reader is relying on orthography.

Results

58

The experiment implemented a 2 (proﬁciency group) x 2 (phonological similarity)
x 2 (spelling similarity) repeated measures factorial design, with proﬁciency group as a
between-subject independent variable and phonological similarity and spelling similarity
within-subject independent variables in the participant analysis. In the item analysis, the
word factors were between-item independent variables, and proﬁciency was a within-
items independent variable. Dependent variables in both the participant and item
analyses were the error rates and reaction times of the category judgments. Analyses
were performed using repeated measures factorial AN OVAs.

Accuracy

Table 7 shows the means and standard deviations of the error rates for each of the
four conditions for both proﬁciency levels. The analysis showed a main effect of
proﬁciency group, reliable by participant, F (1 ,18) = 18.20, p < .001, and by item. F(1,30)
= 42.24, p < .001. Overall, high proﬁciency readers made fewer errors than low-
proﬁciency readers in all four experimental conditions (Figure 11).
In addition, there was a main effect of phonological similarity that was reliable by
participant: F(1,18) = 10.02, p < .01, indicating that, overall, more errors were made on
phonological foils than on orthographic controls. This was, however, not signiﬁcant in
the item analysis, F (1,30) = 2.49, p>.1. The main effect of spelling similarity was not
signiﬁcant by participant or by item.

More pertinent to the present study, however, were the interactions between the
word—based variables (phonological and orthographic similarity) and proﬁciency.
Although neither two-way interaction between phonological or orthographic similarity

and proﬁciency were signiﬁcant, there was a signiﬁcant three-way interaction, both by

59

participant, F(1,18) = 15.94,p < .001, and by item, F(1,30) = 5.97,p < .05. Post-hoe
paired sample t-tests revealed a signiﬁcant effect by participant of phonological similarity
for the high proﬁciency group; that is, the high proﬁciency group had signiﬁcantly higher
error rates on the phonological foils, showing a sensitivity to the phonology of the words,
t(19) = 2.77, p < .05; however, this was not reliable by item t(16) = 1.56, p > .1.
Phonological similarity did not have an effect on the low proﬁciency group, by
participant, t(19) = 1.03, p > .3, or by item, t(16) = 0.72, p > .4. Post-hoe paired sample t-
tests on the effect of spelling similarity did not show any signiﬁcant differences for either

high or low proﬁciency readers, by either participant or item.

Table 7

Means and standard deviations of error rates for each condition, in percents

 

 

 

 

Similarly Spelled Less Similarly Spelled
M SD M SD

High Proﬁciency

Phonological Foils 18.88 19.65 13.75 12.43

Orthographic Controls 7.77 9.14 6.25 10.62
Low Proﬁciency

Phonological Foils 32.21 26.43 51.25 19.94

Orthographic Controls 45.53 19.22 26.25 18.11

60

9.8m xuaﬂouoa 3 28238 .58 on: a $6888 .8833 bowogao .~ ~ Same

 

 

 

3:20:05 :91 5:30:95 26..
29950 259355 2.0“. .moagococd 29:80 ozawaoﬁo m__on_ 629.895
Ll . _ l?
to
j:
a m
m
e. . an:
ion m
8
w
O
13%
.o tom
B__E_m $3 . . .
5::5 II
25:56 8.:on .8

 

 

 

In addition, an unexpected result appeared within the similarly spelled conditions
(Figure 12). Whereas the high proﬁciency group made more false-positives on the
phonological foils, as was expected as well as what was shown across all conditions by
the Chinese group in Wang et al. (2003), the low proﬁciency group in the present study
actually made more false-positives on the orthographic controls within the similarly
spelled conditions. This result was not found in the less similarly spelled stimuli,
however, where low proﬁciency readers made more errors on phonological foils. High
proﬁciency readers, on the other hand, made more errors on phonological foils than
orthographic controls regardless of differences in spelling similarity.

Reaction Times

As is customary in analyses of reaction time data, only correct responses were
analyzed. Reaction time data were trimmed to two standard deviations from the mean of
each participant. As in the error rate analysis, a 2x2x2 repeated measures factorial
ANOVA was performed.

The analysis of the reaction time data showed a main effect of proﬁciency group,
which was signiﬁcant by participant, F(1,18) = 11.64, p < .01, and by item, F(1,30) =
73.11, p < .001. This shows that, overall, the high proﬁciency readers made faster
decision times than the low proﬁciency readers. There was a signiﬁcant interaction
between proﬁciency group and phonological similarity in the item analysis, F(1,30) =
7.63, p < .01. Although the high proﬁciency readers’ mean reaction time only increased
from 1157.18ms (SD = 187.69) on the phonological foils to 1167.06ms (SD = 163.62) on

the orthographic controls (an increase of less than 10ms), the low proﬁciency readers’

62

mean reaction time increased from 1477.71ms (SD = 289.247) on the phonological foils
to 1788.18ms (SD = 396.94) on the orthographic foils, an increase of over 300ms (Figure
13). Although this difference was signiﬁcant by item, it was not signiﬁcant by
participant, F(1,18) = 2.96,p > .1.

None of the interactions between spelling similarity and proﬁciency group,
phonological similarity and spelling similarity, or the three-way interaction between these

were signiﬁcant when analyzed by either participant or item (all F (1 ,18) < 2.96, p > .1).

63

Similar

 

 

 

 

 

50— Proﬁciency Group
' , r—-High Proﬁciency
’ ' ' - Low Proﬁciency
A40“ ’
35,
a .-
a
a: 30-
3
5
c 20-
a
o
E
10‘
0-
l T
Phonological Foils Orthographic Controls
Phonological Similarity
Lou Similar
60" Proﬁciency Group
‘—High Proﬁciency
50_ .g - - - Low Proﬁciency

b
O
I

Mean Error Rate (%)
M
O
I

 

 

 

201
O—
17 I
Phonological Foils Orthographic Controls

Phonological Similarity

Figure 12. Error rates of phonological similarity in the similarly-spelled and less
similarly spelled conditions.

64

Interaction between proficiency and phonological similarity

 

Proﬁciency Group
’ , I — High Proﬁciency

$1800“ - - - Low Proﬁciency
e
a
i: 1600- , . . ’ '
C C
.9
‘6
8 1400‘
a:
c
to
a)
2

1200-

 

 

 

 

I l
Phonological Foils Orthographic Controls
Phonological Similarity

Figure 13. Mean reaction times for phonological similarity by proﬁciency group.

Discussion

Experiment 2 set out to answer the questions, “When high L2 proﬁciency readers
read English words for meaning, do they show more reliance on phonological
information compared to low proﬁciency readers?” and, “Do lower L2 proﬁciency
readers show heavier reliance on orthographic information than high L2 proﬁciency
readers?” This experiment used differences in phonological similarity and spelling
similarity among the stimuli in order to provide answers to these questions. Through
identical methodology, Wang et al. (2003) showed that L1 Chinese speakers relied less

on phonological information than L2 English readers whose L1 orthography is also

65

alphabetic (Korean). In addition, that study showed that L1 Chinese speakers showed a
greater sensitivity to orthographic similarity when reading L2 English than alphabetic
(Korean) readers. The present study investigated the effects of these factors within a
group of L1 Chinese readers of L2 English, but with differing levels of proﬁciency. It
was hypothesized that processing strategies transferred from the L1 , those demonstrated
by Wang et a1. (2003), would be more evident in low proﬁciency readers, while high
proﬁciency readers would show patterns that reﬂect a lessened effect of L1 orthography.
In the analysis of the results, the error rate data is likely to be more useful than the
reaction time data. Wang, Koda, and Perfetti (2003) had few signiﬁcant results from the
reaction time data, similar to the present study. They state that a lack of signiﬁcant
interactions among reaction time data is a common occurrence in L2 research because of
low L2 accuracy levels (J uffs, 2001). In the present experiment, as well, there could have
been an effect of the very high error rates that occurred due to the paradigm of the
experiment, which induced errors through the use of words that were homophones of, or
spelled similarly to, words that had been primed for. Because the reaction time analyses
included only correct responses, in some cases there could have been a only small
number of accurate results remaining after eliminating incorrect responses, which may
not provide adequate information regarding the processes that occur in the populations
involved. Therefore, this limitation should be considered when interpreting any results
(signiﬁcant or non-signiﬁcant) from the analysis of the reaction time data.

Concerning the ﬁrst part of research question two, whether high proﬁciency
readers show more reliance on phonological information compared to low proﬁciency

readers, the results suggest that the answer is in the afﬁrmative. For the high proﬁciency

66

readers, phonological similarity to primed words (i.e. homophones) caused readers to
make more errors than when words were orthographically similar. On the other hand,
such an effect was only evident in low proﬁciency readers under certain orthography-
based conditions (orthographically less-similar words). Therefore, it appears high
proﬁciency readers had a more general reliance on phonology than low proﬁciency
readers did.

Regarding the second part of research question two, whether low proﬁciency
readers show heavier reliance on orthographic information than high proﬁciency readers
do, the results also indicate a positive answer. The low proﬁciency readers showed much
more variation as a result of spelling similarity than the high proﬁciency readers did,
even to the extent that it changed their reliance on phonology. The high proﬁciency
readers showed nearly identical patterns when reading similarly spelled stimuli and less
similarly spelled stimuli, thus showing a lesser degree of reliance on this orthographic
information. The low proﬁciency readers, on the other hand, showed reliance on
phonological information when interference from orthographic similarity was less, and
reliance on orthographic information when orthographic similarity was greater.

However, this is not to say that the low proﬁciency readers only showed reliance
on orthographical information, without reliance on phonological information; rather, the
low proﬁciency readers showed a very high degree of reliance on phonology within the
less similarly spelled conditions. That is, only when the inﬂuence of orthography was
lessened, as was the case in the less similarly spelled stimuli, did the effect of
phonological similarity appear. This could indicate that although low L2 English

proﬁciency readers whose L1 is Chinese do rely on phonological information to an extent

67

while reading for meaning, they rely primarily on orthographic information, and
secondarily on phonological information.

Overall, these results suggest that low proﬁciency L2 English readers whose L1
is Chinese show less reliance on phonological information, and greater reliance on
orthographic information, than high proﬁciency L2 English readers. The high
proﬁciency readers showed patterns that are nearly identical to those shown by the
alphabetic readers in Wang et al. (2003), suggesting that as proﬁciency increases,
phonological activation may become more prevalent, and therefore a larger factor, during
word recognition. The low proﬁciency readers, on the other hand, showed results that
were somewhat different from those shown by the Chinese group in Wang et a1. (2003).
In Wang et al., the participants were 15 out of 20 participants were recruited from an ESL
program, although from the intermediate and advanced level courses, whereas the low
proﬁciency readers in the present study were recruited from lower level courses. The
remaining 5 L1 Chinese participants in that study were ﬁrst year graduate students, which
would seem to be more similar to the high proﬁciency group in the present study.
However, the mean MTELP score for the L1 Chinese group in Wang et al. was 62.13%
(for all three sections of the MTELP), whereas the high proﬁciency group’s mean
MTELP score in the present study was 83.67%, while the low proﬁciency group’s mean
score was 37.67% (for the grammar and reading sections of the MTELP). Therefore, the
participants in Wang et al. could be seen as between the high and low proﬁciency groups
in the present study. In Wang et al., however, orthographic information had a greater
impact on the L1 Chinese group than did phonological information. In the present study,

however, both orthographic information and phonological information had large impacts

68

on word recognition by the low proﬁciency readers, suggesting that low proﬁciency
readers do have a certain level of reliance on phonological information as well.
General Discussion

The present study sought to investigate differences in word recognition processes
that arise from differences in L2 proﬁciency. Overall, the results of the two experiments
suggest that with increased L2 English proﬁciency, intra—word processing of constituent
letters of words increases, as does reliance on phonological information. In addition, the
results suggest that with proﬁciency, the cognitive processes of L2 English learners
whose L1 is non-alphabetic are restructured, rather than merely sped-up. These results
are consistent with results shown by Chikamatsu (2006), which investigated L2 Japanese
reading processes of L1 English speakers (i.e., alphabetic L1 and morphographic L2,
opposite of the present study) and thus restructuring of processes in order to ﬁt
(morphographic) Japanese orthography (i.e., a greater reliance on orthographic
information by the high proﬁciency group). The results of the present study are also
consistent with predictions made, although not investigated, by Wang, Koda, and Perfetti
(2003)

Taken together, the results of the two experiments suggest that the effects of L1
orthography on word recognition processes (i.e. lesser reliance on, and lesser activation
of, phonology) might diminish as L2 proﬁciency increases. This was seen in the present
study through the use of two very different methodologies in order to reduce any effects
that may be resultant from the type of task used. Experiment 1 investigated these
phenomena in terms of processing constituent letters of words and Experiment 2 did so in

‘ terms of reliance on phonological information.

69

In Experiment 1, it was shown that high L2 English proﬁciency readers showed
faster intra-word processing with fewer errors as well as greater sensitivity to grapheme-
phoneme correspondence than did low L2 English proﬁciency readers. High proﬁciency
readers processed the individual letters of words, theorized to activate the phonology of
those letters, more quickly and with a higher accuracy rate than lower proﬁciency readers.
An analysis of the differences between the coefﬁcients of variance of the participants in
both proﬁciency groups suggest that the differences in intra-word processing between
high and low proﬁciency readers may be due to the restructuring of cognitive processes,
rather than simple speed-up.

In Experiment 2, high L2 English proﬁciency readers relied more on phonological
information than did low L2 English proﬁciency readers. In addition, it was shown that
low L2 English proﬁciency readers may be more sensitive to orthography than high
proﬁciency readers. These results would seem to support the results of Experiment 1,
showing a decrease in L1 orthographic transfer in higher L2 proﬁciency readers.

These results may also be relevant to the application of the orthographic depth
hypothesis to L2 reading. The ODH hypothesizes degrees of phonological reliance due
to orthography, and previous cross-orthographic word recognition research has applied it .
in this way, without regard to other factors that may affect word recognition. Previous
research has applied the ODH to cross-orthographic word recognition in terms that are
absolute: that is, that the effect of the L1 orthography on L2 reading is constant and
unvarying. The results of the present study, however, suggest that the degree to which L1
orthography affects L2 reading may actually be more variable, with higher proﬁciency

readers showing less of an L1 orthographic effect than lower proﬁciency readers. Future

70

studies of L2 word recognition must pay closer attention to L2 proﬁciency as a potential
source of variation among learners.

These results also highlight differences between L1 and L2 literacy development.
When learning to read, native English-speaking children go through phases of literacy
development that include the amalgamation of constituent letters of words at the lower
end of development, and addressed recognition of entire words (i.e., the reading of “sight
words”) at the higher end of development (Ehri, 1995). In fact, for native English-
speaking children, the ability to process the constituent letters of words, and the ability to
phonologically recode those letters, is a requisite skill of sight word reading (Ehri, 1998).
The present study, however, suggests that it may not necessarily be this way in the
development of L2 English literacy, and in fact, depending on the orthography of the
reader’s L1, these skills may develop in an order opposite of L1 readers. Similar to L1
reading acquisition, however, is that highly proﬁcient readers have abilities to process
words both with reliance on phonological information as well as reliance on orthographic
information.
Limitations and Future Research

While the present study showed what appear to be consistent results triangulated
through the use of independent experimental paradigms, they did so with a very limited
sample size (N = 20). Future research investigating developmental changes in the
inﬂuence of L1 orthography on L2 word recognition should be performed incorporating
larger numbers of participants in order to ﬁnd results that are more generalizable.

In addition, as this study was concerned with the effect of L2 proﬁciency on L2

word recognition, a cross-section of proﬁciency levels was used. Ideally, a longitudinal

71

study should be performed in order to assess changes as a result of L2 proﬁciency within
individual learners.

Furthermore, the present study only investigated recognition of single words,
which is not how people usually read. Previous research (e. g. Chikamatsu, 2006; Fender,
2003) has shown that L2 learners may recognize words differently based on the context
in which the words appear. It would be beneﬁcial for future studies to investigate effects
of L2 proﬁciency on not only single words, but also words within a text (sentence,
paragraph, etc.) as well.

The present study assumes, based on the orthographic depth hypothesis, that
differences between the low proﬁciency readers and the high proﬁciency readers were in
response to the alphabetic nature of the orthography of the L2. It would be prudent for
future research to incorporate other alphabetic L2s as well in order to eliminate any
possible effects that are speciﬁc to English, and to ﬁnd which effects can be generalizable
to alphabetic orthographies as a whole. It would also be informative to compare results
of L2 English readers with L1 English readers to see if word recognition processes of
high proﬁciency L2 English readers become more similar to those of L1 English readers.

In the same vein, it would also be beneﬁcial for future studies to include L2
English learners of not only a variety of L2 proﬁciency levels, but Lls as well. It is
possible that the developmental results found in this study are common to all learners of
English, regardless of the nature of the L1 orthography. For this reason, high and low L2
proﬁciency readers of a greater number and variety of L1 s should be included in future
studies. In addition, it might be the case that high proﬁciency L2 English readers of any

type of L1 will read words in similar ways; that is, with increased L2 proﬁciency, L2

72

readers may converge on a speciﬁc type of processing that is optimal for the L2. This is
suggested by very close similarities between the high proﬁciency group in the present
study (who have a morphographic L1 orthography) and the alphabetic L1 group (Korean)
in Wang et a1. (2003).

In terms of pedagogical applications, if the results found in the present study are
found to be generalizable, future pedagogical studies of learner training that encourages
processing of constituent letters in words and reliance on phonology might be also useful
(a vein of research already begun by Akamatsu, 2007). If processing of constituent
letters of words and reliance on phonological information are, in fact, traits of high
proﬁciency readers, it may be useful to investigate whether these traits are an extraneous
result of increased proﬁciency only, or whether the acquisition of these traits can produce

higher-level readers.

73

Appendix A

Y ar invitedt a ici te in r s a h d r din in n Ian .

The study will investigate how people whose native language is Chinese read English. You are
able to participate in this study if:

1) Your native language is Chinese, and
2) You are between 18 and 40 years old.

Procedure:

The study will involve attending 1 session, arranged with Ryan Miller. During this session, you
will:

1) ﬁll out a short questionnaire about your language learning experiences,

2) take a test to measure your English language proﬁciency, and

3) read words and sentences on a computer screen and push buttons on the keyboard.

Session times:
<Session times, dates, locations>
Or, you can schedule a time with Ryan Miller by emailing ryantm@msu.edu.

Voluntary participation:
Your participation in this study is completely voluntary. You may refuse to answer any questions,
and you may withdraw from the study at any time.

Compensation:
For your participation in this study, you will receive $15 cash payment.

Risks:
Your participation in this study will not affect your enrollment in the university, and it will not
affect your grades in your classes.

Beneﬁts:
Your participation in this study will give you some extra practice in reading English. In addition, it
contributes to second language research.

Confidentiality:

You will not be required to give your name or any other personally identiﬁable information. You
will be identiﬁed only by a number throughout the data collection process as well as any further
publication.

Contact information:

If you have any questions or concerns, contact Ryan Miller at ryantm@msu.edu, by phone at
517-355-0800, or by mail at A-714 Wells Hall, Michigan State University, East Lansing, MI, 48824

74

1)
2)
3)
4)

5)

6)

7)

3)

9)

Background Information

Appendix B

 

Gender (circle one): Male Female
Age: years old

Country of birth:

What is your native language? Chinese
(Language: )

 

If your native language is Chinese, did you use pinyin ( fifia‘ ) when learning

Chinese?

Yes

English

No

Other

Subject ID:

a. If yes, how old were you when you started using pinyin ( fifief )?

years old

What language(s) do you speak at home now? (circle)

Chinese

English

Other (Language:

 

a. If more than 1, what percent of the time do you speak each:
Other:

Chinese: %

English:

%

In what languages did you receive your education before university?

Chinese English Other (What language?

a. If more than one, please give the approximate number of years for each

language
Chinese:

English:

Other

How many years and months have you lived in Michigan?

months

4

years

Have you ever lived in another English-speaking place besides Michigan?

Yes

75

No

%

 

 

 

 

 

 

a. If yes, where, and for how long? Place: , years __ months
Place: , years __ months
10) Please mark what you think your ability is in each of the following.
Use the scale: 1 = Poor, 2 = Fair, 3 = Good, 4 = Very good, 5 =
Native/Native-like
Listening Speaking Reading Writing Grammar
Chinese
English
Other
(Language: )

 

 

 

 

 

 

 

 

1 1) How long do you usually spend each week doing the following activities? How

long did you spend before last semester? Please write the number of hours and

 

 

 

 

 

 

 

 

minutes.

NOW BEFORE last semester
Reading in English _hours_minutes _hours_minutes
Writing in English _hours_minutes _hours_minutes
Listening in English _hours_minutes _hours_minutes
Speaking in English _hours_minutes _hours_minutes
Reading in Chinese _hours_minutes _hours_minutes
Writing in Chinese _hours_minutes _hours_minutes

 

 

 

 

12) What do you usually read in English, and how long do you spend reading them

each week? If ‘other’, please write what you read and for how long.

Newspapers: hours minutes
Magazines: hours minutes
Novels: hours minutes

Web sites: _hours_minutes

 

 

 

Other L ): hours—_minutes
Other ( ): hours—minutes
Other L ): hours—minutes
Other ( ): hours—minutes

 

76

 

 

Appendix C

Debﬂeﬁng

The study you have participated in concerned impact of second language proﬁciency on
the process of recognizing words while reading. Speciﬁcally, it measured how native Chinese
speakers use sound while reading in English. One task that you performed measured the time it
took you to recognize single words. Some words were more common words than other words.
This is because the brain tends to use the sounds of words less for common words than for
uncommon words. Also, some words had spellings that follow common spelling patterns and
some were more unusual. This is because the brain may store words that have less common
spelling patterns as a larger piece, rather than as a combination of sounds. Some words were
not real words. These words were included to test your ability to recognize real words as
opposed to unreal words. Some words were in lower case (lower case) and some words were in
alternating case (aLtErNaTiNg CaSe) because when you read words in alternating case, your
brain cannot recognize the words by their shape, and instead your brain must ﬁgure out the
sounds of the words. Another task was similar to the ﬁrst task, but had the words in sentences.
This was to measure changes that might result from reading to understand meanings of words
and sentences. The remaining task was to measure how you read words for meaning, but not in
sentences.

As compensation for my participation in the research study, I
acknowledge receipt of $15 cash from .

(researcher name)

 

Signature

77

Appendix D

 

High Frequency — Regular Spelling

High Frequency — Exception Spelling

 

such
men
place
come
home
gave

tell
help
soon
turn

 

state
still
fact
want
most
week
free
class

road
main

good
house
word
come
took
head
group
half
says
dead

 

work
again
does
water
once
knew
young
girl
love
move

 

Low Frequency - Regular Spelling

Low Frequency — Exception Spelling

 

 

pitch
beam
sank
ripe
tower
pump
slip
ﬂoat
slam
spark

 

neat
ﬂuid
rage
wax
dot
trunk
deed
weld

peel
fade

 

bowl
naive
earn
blew
knock
wool
knit
sew
worm
ivory

 

sweat
ﬂood
urge
fur
log
glue
aisle

swamp

swarm
dial

 

78

 

 

Non-word Distractors

 

 

balf
bant
boad
bood
bool
boose
bork
bove
cipy
clow

dase

deak
deght
dit

dolp
dool

 

dost

dramp

dreat
drep
fep
ﬂace
ﬂape
foll
forli
foth
fove
frake

fregh
frep

fushe

 

gial
gorrn

gream

haide
hepe
hupe
jart
kolm
kouf
lairrn
lare
lear
lere
lipht
lity

lor

 

mai g
mand
mave
meafe
mide
mif
mive
nad
nar
natch
nouse
nowl
orm
phint
plue
poes

 

pome
poom
poot
reat
rilp
rone
roshy
rull
starm
stook
sut
tays
tew
trom
tross
vank

 

79

 

Appendix E

In this task, you will see some text on the screen. You need to
decide if the text is a real English word or not. Some words will
be real words, and some will not be real words.

Press the SPACEBAR to continue.

If you think the text is a real word in English, press the GREEN
button as soon as you recognize the word. If you think the text
is not a real word in English, press the RED button.

Press the SPACEBAR to continue.

Try to answer as quickly and accurately as you can, but be sure
to read each word before you press a button.
Press the SPACEBAR to continue.

Some words will be written in a mixture of UPPERCASE letters and
lowercase letters. This means that ‘dog’ is still a word if it is
written ‘DoG’ or ‘dog'. For both of these, you would press the
GREEN button (meaning Yes, it is a word).

Press the SPACEBAR to continue.

First, you will have 10 practice questions. Remember, push the
green button if what you see is a word in English, and press the
red button if you think it is not a word in English.

Push the SPACEBAR to continue.

Press the SPACEBAR to go on to the 10 practice questions.
Ready? Here we go...

[Practice items]
big
gorp
aPpLe
cup
pHoNe
wOfP
cidno
erE
fLoMp
klim

80

Appendix F

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Category name Exemplar Homophone Spelling Control
Similarly spelled foils

A feature of an ocean shore beach beech bench

Part of a sandwich bread bred bead

A small stream creek creak cheek

A body part feet feat fees

Transportation in the sky plane plain plans

Type of weather rain rein ruin

Used to get 13) or down stair stare stars

The end of your feet toe tow toL

An alcoholic drink wine whine wink

Less similarly spelled foils

An animal bear bare beat

A breakfast food cereal serial several

A part of a plant ﬂower ﬂour ﬂowed

Part of a person’s face nose knows snobs

A passage used by vehicles road rode rods

A ﬂower rose rows robs

Part of a boat sail sale salt

A sea animal whale wale wheel

 

 

 

81

 

 

Distractors

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Category Stimulus Category Stimulus
Used for communication phone A hot drink tea
Something you can read book A place on campus union
A place to live house A place on campus stadium
Something to hold liquids cg) A wort soccer
A person ﬁ'iend A sport tennis

A thing you wear on your shoe
Something you carry bag foot

A thing you wear on your sandal
Something you wear coat foot
A ﬁ'uit lemon A place you work ofﬁce
A fruit apgle A fruit book
A kind of furniture chair Somethingyou wear plane
A kind of furniture table A vehicle cloud
A place in a house kitchen A food book
A place in a house bedroom A person table
Something you watch TV An animal chair
A type of clothing shirt A computer part friend
A type of clothing hat A place in a house apple
A color blue A hot drink light
A color green An animal wire
An animal mouse A type of weather pig
An animal rabbit A food computer
Something you listen to music Used to play sports kangaroo
Something you listen to radio A feature of a house CD
A vehicle train Part of a house arrow
A computer part mouse A place mother
A thing you read letter A family member sand
A hot drink coffee A family member page

 

82

 

Appendix G

In this task, you will see 2 sets of text on the screen for
each item. First, you will see a phrase. These words are a
category. Then, those words will go away and you will see a
single word. You need to decide if the single word fits
the category or not. Press the GREEN button if the word
fits the category. Press the RED button if it does not fit
the category

Press the SPACEBAR to continue.

Next, you will see some examples. Remember, press the GREEN
button if the word fits the category, and press the RED
button if it does not fit the category. Try to press the
buttons as quickly and accurately as you can.

Press the SPACEBAR to continue.

You will see 2 examples. Remember, first you will see a
category, then you will see a word. Press the GREEN button
if the word is something that fits in the category, and
press the RED button if it does not fit.

Press the SPACEBAR to go to the first example.

A pet [1500ms - then disappear] / dog

In this example, the word ‘dog’ fits the category ‘a pet’,
so you should have pressed the GREEN (Yes) button.
Press the SPACEBAR to do another example.

A place [1500ms — then disappear] / egg

In this example, ‘egg’ does not fit the category ’a place',
so you should have pressed the RED (No) button.
Press the SPACEBAR to do 10 practice questions.

Ready? Here we go. Press the SPACEBAR to start the practice.
[Practice items]

A kind of clothing / sweater

A food / bread

A family member / bicycle

A drink / juice

Something you listen to / paper
A type of transportation / car
An electronic device / orange
Something you watch / TV

A person / desk

A food / mother

83

References

Aaron, R, Joshi, R., Ayotollah, M., Ellsberry, A., Henderson, J., & Lindsey, K. (1999).
Decoding and sight-word naming: Are they independent components of word
recognition skill? Reading and Writing, 11(2), 89-127.

Akamatsu, N. (1999). The effects of ﬁrst language orthographic features on word

recognition processing in English as a second language. Reading and Writing,
11(4), 381-403.

Akamatsu, N. (2003). The effects of ﬁrst language orthographic features on second
language reading in text. Language learning, 53(2), 207-231.

Akamatsu, N. (2005). Effects of second language reading proﬁciency and ﬁrst language
orthography on second language word recognition, In V.J. Cook and B. Bassetti,
Second Language Writing Systems, (pp. 238-259), Clevedon, England:
Multilingual Matters.

Akamatsu, N. (2007). The effects of training on automatization of word recognition in
English as a foreign language. Applied Psycholinguistics, 29(2), 175-193.

Baldauf, R., & Dawson, R. (1980). The predictive validity of the Michigan test of English
language proﬁciency for teacher trainees in Papua New Guinea. Educational and
Psychological Measurement, 40(4), 1 201-1 205.

Balota, D., & Chumbley, J. (1984). Are lexical decisions a good measure of lexical
access? The role of word frequency in the neglected decision stage. Journal of
Experimental Psychology: Human Perception and Performance, 10(3), 340-357.

Baluch, B., & Besner, D. (1991). Visual word recognition: Evidence for strategic control
of lexical and nonlexical routines in oral reading. Journal of Experimental
Psychology: Learning, Memory, and Cognition, 17(4), 644-652.

Bemdt, R., Reggia, J ., & Mitchum, C. (1987). Empirically derived probabilities for
grapheme-to-phoneme correspondences in English. Behavior research methods,
instruments & computers, 19(1), 1-9.

Brown, T., & Haynes, M. (1985). Literacy background and reading development in a
second language. In T.H. Carr (Ed.), The development of reading skills, (pp. 19-
34). San Francisco: Jossey-Bass.

Chen, H., & Flores, A. GB, & Cheung, SL (1995). Orthographic and phonological
activation in recognizing Chinese characters. Psychological Research, 58, 144-
153.

84

Chikamatsu, N. (1996). The Effects of L1 Orthography on L2 Word Recognition: A
Study of American and Chinese Learners of Japanese. Studies in Second
Language Acquisition, 18(4), 403-432.

Chikamatsu, N. (2006). Developmental Word Recognition: A Study of L1 English
Readers of L2 Japanese. The Modern Language Journal, 90(1), 67-85.

Cho, J ., & Chen, H. (1999). Orthographic and phonological activation in the semantic

processing of Korean Hanja and Hangul. Language and Cognitive Processes,
14(5), 481-502.

Corrigan, A., Dobson, B., Kellman, E., Spaan, M., & Tyma, S. (1979). Michigan test of
English language proﬁciency (Form P). Ann Arbor: Michigan University Press.

DeFrancis, J. (1989). Visible speech: The diverse oneness of writing systems, Honolulu:
University of Hawaii Press.

Doehring, D. (1976). Acquisition of rapid reading responses. Monographs of the society
for research in child development, 1-54.

Durgunoglu, A., Nagy, W., & Hancin-Bhatt, B. (1993). Cross-language transfer of
phonological awareness. Journal of Educational Psychology, 85(3), 453-465.

Ehri, L. (1995). Phases of development in learning to read words by sight. Journal of
Research in Reading, 18(2), 116-125.

Ehri, L. (1998). Grapheme-phoneme knowledge is essential for learning to read words in
English. In J .L. Metsala and L. Ehri (Eds.), Word recognition in beginning
literacy, (pp3-40). Hillsdale, NJ: Erlbaum.

Ehri, L., & Wilce, L. (1982). Recognition of spellings printed in lower and mixed case:
Evidence for orthographic images. Journal of Reading Behavior, 14(3), 219-230.

Ellis, N. C., & Hooper, A. M. (2001). Why learning to read is easier in Welsh than in
English: Orthographic transparency effects evinced with frequency-matched tests.
Applied Psycholinguistics, 22, 571—599.

Everson, M. (1998). Word recognition among learners of Chinese as a foreign language:
Investigating the relationship between naming and knowing. Modern Language
Journal, 194-204.

Fender, M. (2003). English word recognition and word integration skills of native
Arabic-and J apanese-speaking learners of English as a second language. Applied
Psycholinguistics, 24(02), 289-315.

Fisher, D. (1975). Reading and visual search. Memory and Cognition, 3, 188-196.

Forster, K., & Chambers, S. (1973). Lexical Access and Naming Time. Journal of Verbal
Learning and Verbal Behavior, 12(6), 627-635.

85

Forster, K., & Forster, J. (2003). DMDX: A Windows display program with millisecond
accuracy. Behavior Research Methods Instruments and Computers, 35(1), 116-
124.

F reebody, P., & Byme, B. (1988). Word-reading strategies in elementary school children:
Relations to comprehension, reading time, and phonemic awareness. Reading
Research Quarterly, 441 -453.

Freed, B., Dewey, D., Segalowitz, N., & Halter, R. (2004). The language contact proﬁle.
Studies in Second Language Acquisition, 26(02), 349-356.

Frost, R. (1994). Prelexical and Postlexical Strategies in Reading: Evidence From a Deep
and a Shallow Orthography. Journal of Experimental Psychology: Learning,
Memory, and Cognition, 20(1), 116-129.

Frost, R. (1998). Toward a strong phonological theory of visual word recognition: True
issues and false trails. Psychological Bulletin, 123, 71-99.

Frost, R., & Katz, L. (1992). Orthography, phonology, morphology, and meaning,
Amsterdam: North-Holland.

Frost, R., Katz, L., & Bentin, S. (1987). Strategies for visual word recognition and
orthographical depth: A multilingual comparison. Strategies, 13(1),104-115.

F ukkink, R., Hulstijn, J ., & Simis, A. (2005). Does training in second-language word
recognition skills affect reading comprehension? An experimental study. Modern
Language Journal, 89(1),54-75.

Geva, E., & Siegel, L. (2000). Orthographic and cognitive factors in the concurrent
development of basic reading skills in two languages. Reading and Writing, 12(1),
1-30.

Goswami, U., Gombert, J ., & de Barrera, L. (1998). Children's orthographic
representations and linguistic transparency: Nonsense word reading in English,
French, and Spanish. Applied Psycholinguistics, 19(1), 19-52.

Gough, P. (1984). Word recognition. Handbook of reading research, 1, 225-253.

Green, D., & Meara, P. (1987). The effects of script on visual search. Second Language
Research, 3(2), 102.

Hino, Y., & Lupker, S. (1998). The effects of word frequency for Japanese Kana and
Kanji words in naming and lexical decision: Can the dual-route model save the
lexical-selection account? Journal of Experimental Psychology, Human
Perception, and Perforamnce, 24, 1431-1453.

Hino, Y., & Lupker, S. (2000). Effects of word frequency and spelling-to-sound
regularity in naming with and without preceding lexical decision. Journal of
experimental psychology. Human perception and performance, 26(1), 166.

86

Holm, A., & Dodd, B. (1996). The effect of ﬁrst written language on the acquisition of
English literacy. Cognition, 59(2), 119-147.

Holmes, V., & Carruthers, J. (1998). The relation between reading and spelling in skilled
adult readers. Journal of Memory and Language, 39(2), 264-289.

Hoosain, R. (1991). Psycholinguistic implications for linguistic relativity: A case study of
Chinese: L. Erlbaum Associates.

Jared, D. (2002). Spelling-sound consistency and regularity effects in word naming.
Journal of Memory and Language, 46(4), 723-750.

J iang, N. (1999). Testing processing explanations for the asymmetry in masked cross-
language priming. Bilingualism: Language and Cognition, 2(01), 59-75.

J iang, N. (2007). Selective integration of linguistic knowledge in adult second language
learning. Language Learning, 57(1), 1-33.

Juffs, A. (2001). Psycholinguistically Oriented Second Language Research. Annual
Review of Applied Linguistics, 21, 207-220.

Katz, L., & Feldman, L. (1983). Relation between pronunciation and recognition of
printed words in deep and shallow orthographies. Relation, 9(1), 157-166.

Katz, L., & Frost, R. (1992). The reading process is different for different orthographies:
The orthographic depth hypothesis. Orthography, phonology, morphology, and
meaning, 67-84.

Kessler, B., & Treiman, R. (2001). Relationships between sounds and letters in English
monosyllables. Journal of Memory and Language, 44(4), 592-617.

Kinoshita, S. (1987). Case alternation effect: Two types of word recognition? The
Quarterly Journal of Experimental Psychology Section A, 39(4), 701 -720.

Koda, K. (1988). Cognitive process in second language reading: Transfer of L1 reading
skills and strategies. Second Language Research, 4(2), 133.

Koda, K. (1989). Effects of L1 orthographic representation on L2 phonological coding
strategies. Journal of Psycholinguistic Research, 18(2), 201-222.

Koda, K. (1990). The Use of L1 Reading Strategies in L2 Reading: Effects of L1
Orthographic Structures in L2 Phonological Recoding Strategies. Studies in
Second Language Acquisition, 12(4), 393-410.

Koda, K. (1999). Development of L2 intraword orthographic sensitivity and decoding
skills. Modern Language Journal, 51-64.

Koda, K. (2005). Insights into second language reading: A cross-linguistic approach,
Cambridge: Cambridge University Press.

Kucera, H., & Francis, W. (1967). Computational analysis of present-day American

87

English, Providence: Brown University Press.

LaBerge, D., & Samuels, S. (1974). Toward a theory of automatic information processing
in reading. Cognitive Psychology, 6(2), 293-323.

Leong, C. (1986). What does accessing a morphemic script tell us about reading and
reading disorders in an alphabetic script? Annals of Dyslexia, 36(1), 82-102.

Lesch, M., & Pollatsek, A. (1998). Evidence for the use of assembled phonology in
accessing the meaning of printed words. Journal of experimental psychology.
Learning, memory, and cognition, 24(3), 573.

Luo, C. (1996). How is word meaning accessed in reading? Evidence from the
phonologically mediated intereference effect. Journal of experimental psychology.
Learning, memory, and cognition, 22(4), 883-895.

McCann, R., Folk, C., & Johnston, J. (1992). The role of spatial attention in visual word
processing. Journal of Experimental Psychology, 18(4), 1015-1029.

Midgley, K., Holcomb, P., & Grainger, J. (2009). Language effects in second language
learners and proﬁcient bilinguals investigated with event-related potentials.
Journal of Neurolinguistics, 22(3), 281-300.

Mori, Y. (1998). Effects of ﬁrst language and phonological accessibility on kanji
recognition. Modern Language Journal, 69-82.

Muljani, D., Koda, K., & Moates, D. (1998). The Development of Word Recognition in a
Second Language. Applied Psycholinguistics, 19(1), 99-113.

Oney, B., & Durgunoglu, A. (1997). Beginning To Read in Turkish: A Phonologically
Transparent Orthography. Applied Psycholinguistics, 18(1), 1-15.

Paap, K., Newsome, S., & Noel, R. (1984). Word shape’s in poor shape for the race to the
lexicon. Journal of Experimental Psychology: Human Perception and
Performance, 10(3), 413-428.

Paap, K., & Noel, R. (1991). Dual-route models of print to sound: Still a good horse race.
Psychological Research, 53(1), 13-24.

Peereman, R., Content, A., & Bonin, P. (1998). Is perception a two-way street? The case
of feedback consistency in visual word recognition. Journal of Memory and
Language, 39, 151-174.

Perfetti, C. (2003). The universal grammar of reading. Scientiﬁc Studies of Reading, 7(1),
3-24.

Perfetti, C., Liu, Y., Fiez, J ., Nelson, J ., Bolger, D., & Tan, L. (2007). Reading in two
writing systems: Accommodation and assimilation of the brain's reading network.
Bilingualism: Language and Cognition, 10(02), 131-146.

88

Perfetti, C., Liu, Y., & Tan, L. (2005). The lexical constituency model: Some

implications of research on Chinese for general theories of reading. Psychological
Review, 112(1), 43-59.

Perfetti, C., & Zhang, S. (1995). Very early phonological activation in Chinese reading.
Learning, Memory, 21(1), 24-33.

Perfetti, C., Zhang, S., & Berent, I. (1992). Reading in English and Chinese: Evidence for
a universal phonological principle. Orthography, phonology, morphology, and
meaning, 94, 227-248.

Plaut, D., & McClelland, J. ,,Seidenberg, MS, & Patterson, ?.(1996). Understanding
normal and impaired word reading: Computational principles in quasi-regular
domains. Psychological Review, 103, 56-115.

Preacher, K., Rucker, D., MacCallum, R., & Nicewander, W. (2005). Use of the extreme
groups approach: A critical reexamination and new recommendations.
Psychological methods, 10(2), 178-192.

Rumsey, J ., Horwitz, B., Donohue, B., Nace, K., Maisog, J ., & Andreason, P. (1997).
Phonological and orthographic components of word recognition. A PET-rCBF
study. Brain, 120(5), 739-759.

Schneider, W., & Shiffrin, R. (1977). Controlled and Automatic Human Information
Processing: 1. Detection, Search, and Attention. Psychological Review, 84(1), 1-
66.

Segalowitz, N. (1986). Skilled reading in the second languageln J. Vaid (Ed.), Language
processing in bilinguals: Psycholinguistic and neuropsychological perspectives,
(pp.3-19). Hillsdale, NJ: Lawrence Erlbaum.

Segalowitz, N., & Hulstijn, J. (2005). Automaticity in bilingualism and second language
learning. Handbook of bilingualism: Psycholinguistic approaches, 371-3 88.

Segalowitz, N., & Segalowitz, S. (1993). Skilled Performance, Practice, and the
Differentiation of Speed-Up from Automatization Effects: Evidence from Second
Language Word Recognition. Applied Psycholinguistics, 14(3), 369-3 85.

Segalowitz, S., Segalowitz, N., & Wood, A. (1998). Assessing the Development of

Automaticity in Second Language Word Recognition. Applied Psycholinguistics,
19(1), 53-67.

Seidenberg, M. (1988). Cognitive neuropsychology and language: The state of the art.
Cognitive Neuropsychology, 5(4), 403-426.

Seidenberg, M. (1992). Connectionism without tears. In S. Davis (Ed.), Connectionism:
Theory and practice, (84-122). New York: Oxford University Press.

Seymour, P., Aro, M., & Erskine, J. (2003). Foundation literacy acquisition in European
orthographies. British Journal of Psychology, 94(2), 143-174.

89

Spencer, A. (1991). Morphological theory: An introduction to word structure in
generative grammar: Oxford: Blackwell.

Stanovich, K. (1980). Toward an interactive-compensatory model of individual

differences in the development of reading ﬂuency. Reading research quarterly,
16(1), 32-71.

Stanovich, K. (1991). Word recognition: Changing perspectives. Handbook of reading
research, 2, 418-452.

Tabossi, P., & Laghi, L. (1992). Semantic priming in the pronunciation ofwords in two
writing systems: Italian and English. Memory & Cognition, 20(3), 303-313.

Tan, L., & Perfetti, C. (1997). Visual Chinese character recognition: Does phonological
information mediate access to meaning? Journal of Memory and Language, 3 7(1),
41-57.

Tan, L., & Perfetti, C. (1998). Phonological codes as early sources of constraint in
Chinese word identiﬁcation: A review of current discoveries and theoretical
accounts. Reading and Writing, 10(3), 165-200.

Tan, L., & Perfetti, C. (1999). Phonological activation in visual identiﬁcation of Chinese

two-character words. Journal of Experimental Psychology: Learning, Memory,
and Cognition, 25(2), 382-393.

Tan, L., Spinks, J ., Gao, J ., Liu, H., Perfetti, C., Xiong, J ., et al. (2000). Brain activation
in the processing of Chinese characters and words: a functional MRI study.
Human Brain Mapping, 10(1), 16-27.

Van Orden, G. (1987). A ROWS is a ROSE: spelling, sound, and reading. Memory &
Cognition, 15(3), 181-198.

Wade-Woolley, L. (1999). First language inﬂuences on second language word reading:
All roads lead to Rome. Language Learning, 49(3), 447-471.

Wang, M., & Geva, E. (2003). Spelling performance of Chinese children using English as
a second language: Lexical and visual—orthographic processes. Applied
Psycholinguistics, 24(01), 1-25.

Wang, M., Koda, K., & Perfetti, C. (2003). Alphabetic and non-alphabetic L1 effects in

English semantic processing: A comparison of Korean and Chinese English L2
learners. Cognition, 87, 129-149.

Wydell, T., Patterson, K., & Humphreys, G. (1993). Phonologically mediated access to
meaning for Kanji: is a Rows still a Rose in Japanese Kanji? Journal of
experimental psychology. Learning, memory, and cognition, 19(3), 491-514.

Yin, B., & Felley, M. (1990). Chinese Romanization: Pronunciation and Orthography:
Beijing: Sinolingua.

9O

ml
A"
R”
H
U"
”I
R"
E”
Vil
Ni”
U"
E“i
"
"
“1
H
H
H

3 293 03063 3626