Content uploaded by Tohru Matsuo
Author content
All content in this area was uploaded by Tohru Matsuo on Sep 07, 2019
Content may be subject to copyright.
Measuring Japanese Learners’
Lexical Accuracy
and Fluency Using A Lexical Decision Task
Tohru Matsuo
語彙性判断テストを使用した
日本人学習者の語彙の正確さと流暢さの測定
Abstract
The purpose of this study is to measure word recognition fluency and accuracy for
Japanese learners of English with a visual lexical decision task. This study focuses on
investigating how reaction time and accuracy of visual word recognition in the lexical decision
task differ depending on the participants proficiency levels (Japanese university students,
Japanese English teachers, and native speakers of English) and on the word frequency level
(1K, 2K, 3K, and 4K) of the stimulus words. The results of a one-way and two-way mixed
analysis of variance (ANOVA) showed that reaction time and lexical decision task accuracy
generally distinguish the participants proficiency. Also, a general frequency effect for reaction
time was found for all groups.
Keywords:
Received September ,
抄 録
キーワード：
1. Introduction
This study investigates Japanese learners lexical accuracy and fluency, which is specified
as the automaticity of written word recognition. In both L1 and L2 research, written word
recognition, which refers to the processes of extracting lexical information from graphic
displays of words (Koda, 2004), is widely considered to be one of the most important processes
contributing to skilled reading comprehension (Grabe, 2009; Perfetti, 1999, 2007; Perfetti,
Landi, & Oakhill, 2005). In addition, fluent reading is not possible without the automatic
recognition of a large number of words (Grabe, 2009).
The concept of automaticity or automatization is based on skill acquisition theories, such
as Andersons adaptive control of thought theory (Anderson, 1983; Anderson & Lebiere, 1998)
and Logans (1988) instance theory. Andersons theory regards automatization as starting with
the conscious, controlled processing of declarative knowledge (i.e., knowledge of facts and
rules, such as the knowledge of letter features, and letter-sound correspondences in the case
of word recognition). After large amounts of processing, learners develop rapid, attention-
free processing, which consists largely of routines characterized by chunks of elementary
operations and computations (Anderson & Lebiere, 1998, p. 5). Alternatively, Logans instance
theory (1988) views automaticity as the retrieval of information from memory: performance is
automatic when it is based on single-step direct-access retrieval of past solutions from memory
(p. 493). Performance is initially executed based on rules (e.g., in English, the letter string ph
corresponds to the phone /f/.); however, each time the individual utilizes a particular rule, the
outcome is stored as a higher order unit called an instance. Hence, automatized performance
is not based on the usage of a rule but on the retrieval of previously encountered instances.
Regardless of the differences in these two theories of skill acquisition, automatic processing is
viewed as effortless and fast, and is reflected in a decrease in latencies when performing a task.
Automatized word recognition is important for successful reading comprehension
because of working memory limitations (Perfetti, 1985). Text comprehension is highly
demanding and requires a high degree of cognitive control because readers must process
multiple levels of language simultaneously (e.g., orthography, lexis, and morpho-syntax). If
individuals have not automatized a large number of words, too many attentional resources
must be devoted to recognizing individual words and sufficient cognitive resources are
unavailable for higher order comprehension tasks, such as integrating new information with
older knowledge, making inferences, and developing a representation of the text as a whole.
In order to measure automaticity of word recognition, researchers frequently use a
computerized visual lexical decision task in which test-takers decide whether strings of letters
are real words or non-words as quickly as possible. The lexical decision task produces two
measures, reaction time and reaction accuracy. In Second Language Acquisition (SLA) studies,
Matsuo: Measuring Japanese Learners’ Lexical Accuracy and Fluency Using A Lexical Decision Task
these two measures have often been utilized independently. For example, lexical decision task
accuracy has been utilized to assess the vocabulary size of second language learners (e.g.,
Eyckmans 2004; Mochida & Harrington, 2006), whereas reaction time has been employed to
determine the degree to which L2 learners have developed automatic word recognition (e.g.,
Segalowitz & Hulstijin, 2005; Segalowitz & Segalowitz, 1993). However, few researchers have
combined the two measures to determine how measures of reaction accuracy and response time
simultaneously distinguish L2 learners lexical proficiency (e.g., Harrington, 2006; Kojima, 2010).
In this study I adapted Harringtons design (2006) and examined how well visual lexical decision
task accuracy and response time discriminate English proficiency levels and item difficulty on
the basis of word frequency with L1 Japanese English as a Foreign Language (EFL) learners.
2. Harrington＇s study
Harrington (2006) validated two measures of a lexical decision task, response accuracy
and response time, and examined how the two measures served to discriminate among
between-group proficiency levels and within-group levels of word frequency on an English
lexical decision task. His research questions were whether lexical decision task accuracy
improves as group proficiency and word frequency level increase, whether reaction time
and lexical decision task accuracy improve as group proficiency and word frequency level
increase, and whether lexical decision response stability improves as group proficiency and
word frequency level increase. Response stability was measured with coefficient of variance
of response time, which is calculated as the mean standard deviation of reaction times divided
by the mean (Segalowitz, & Segalowitz, 1993). Three groups participated in the study: 32
intermediate English as a Second Language (ESL) students, 36 advanced ESL students, and
42 native speakers of English, who served as a control group. The 150-item lexical decision
task contained 90 real words and 60 pseudo-words. The 90 real words consisted of 18 items
from each of Vocabulary Levels Tests (Schmitt, Schmitt & Clapham, 2001) four frequency
bands (2K, 3K, 5K, and 10K) and from the Academic Word List (Coxhead, 2000). The results
indicated that reaction times systematically decreased as the participants proficiency and
word frequency increased, and accuracy increased as learner proficiency increased. The
intermediate ESL group was less accurate and slower than the advanced ESL group, which
was less accurate and slower than the native speakers of English. Moreover, both accuracy
and reaction time measures discriminated among all word frequency levels (2K, 3K, 5K, and
10K word frequency levels) in both the intermediate and advanced ESL groups. Response
variability, as measured by the coefficient of variance, also decreased as performance
improved. However, the change was subtler than those of lexical decision task accuracy
and response time. Harrington concluded that lexical decision task accuracy and response
times can discriminate between proficiency levels and word frequency; therefore, these two
measures of lexical knowledge can serve as valuable tools for examining the development of
L2 proficiency.
3. The gaps and purpose of the study
This study is designed to address a gap in Harringtons study, which concerns the
proficiency level of the participants. The participants in Harringtons studies have been ESL
students whose English proficiency is higher than the Japanese EFL students in this study. In
contrast, the university undergraduate students in this study have lower English proficiency
and it is unclear to what extent the findings of previous researchers apply to them. Hence, the
purpose of this study is to investigate whether the same trends observed in previous studies can
be obtained with lower proficiency Japanese English learners.
The following research hypotheses were formed.
1. Lexical decision task accuracy of Japanese learners will improve as group proficiency level
increases.
2. Lexical decision task accuracy of Japanese learners will improve as word frequency
increases.
3. Lexical decision task reaction times of Japanese learners will decrease as group proficiency
level increases.
4. Lexical decision task reaction times of Japanese learners will decrease as word frequency
increases.
4. The study
4. 1. Participants
Three groups of participants took part in this study. The first group was 20 Japanese
university students (male = 15, female = 5) majoring in law, whose English proficiency ranged
from pre-intermediate to intermediate according to the universitys classification system.
They were first and second year students who had received formal English education for 9
or 10 years, and their approximate age was 19. Their mean Test of English for International
Communication Institutional Program (TOIEC IP) score was 495 (Range: 265-525; SD =
73.74). The second group was 20 high proficiency Japanese English high school or university
teachers (male = 5, female = 15), who have earned an Master of Art (MA) in Teaching
English to Speakers of Other Language (TESOL) or Applied Linguistics. The last group was 20
native speakers of English (male = 15, female = 5) teaching at Japanese universities. These
participants all had an MA in TESOL and served as a reference group.
Matsuo: Measuring Japanese Learners’ Lexical Accuracy and Fluency Using A Lexical Decision Task
4. 2. Instruments
The lexical decision task was made up of 144 items. In order to avoid response bias
(Jiang, 2012), which is a type of cognitive bias caused by asking for an unequal number of
positive and negative responses, the number of words and pseudo-words was balanced (72
words and 72 pseudo-words). In Harringtons (2006) study, the 2K, 3K, 5K, and 10K frequency
bands were utilized for real words. However, the Japanese university students in this study
had lower English proficiency than his ESL participants; therefore, the 72 real words consisted
of 18 items from each of the first four 1,000 word families (18 items four word frequency
bands = 72 total items) in the British National Corpus. Example items used in this study are high
and brief (from the first 1,000 word band), song and store (from the second 1,000 word band),
wild and storm (from the third 1,000 word band), and crew and slope (from the fourth 1,000
word band). Stimulus words in the first, second, and third 1,000 word families were selected
from the Japanese students familiarity rate list (Yokogawa, 2006). In the Japanese students
familiarity rate list, familiarity rate was defined as how often learners think they see or hear the
words, not whether they know the meaning of the words. The list shows the familiarity rate for
3,000 words on a scale that ranges from 1.48 (See or hear rarely) to 6.92 (See or hear frequently).
This list was developed through the following process. First, the most frequent 2,981 words
were selected from the Kilgarriff Lemmatised Frequency List (Kilgariff, 1997), and 19 words
(the days of the week and the names of the months) were added. The resulting 3,000 words
were administered to 810 Japanese EFL students from ten universities in western Japan. They
were asked to rate how frequently they think they see or hear the words using a Likert scale
ranging from 1 (Never see or hear) to 7 (Very frequently see or hear).
Because the Japanese students familiarity rate list (Yokogawa, 2006) does not include
beyond the fourth 1,000 word families, the 18 stimulus words (6 adjectives, 6 noun, and 6
verbs) in the fourth 1,000 word families were selected from JACET 8,000 word list (Aizawa,
Ishikawa, & Murata, 2005). The JACET 8,000 list contains 8,000 words considered to be
important for Japanese learners to communicate in English. These 8,000 words were selected
from the British National Corpus and sub-corpus, which contains 5.8 million words from the
following genres: newspaper, TV program transcripts, Junior and Senior High School English
textbooks, scientific articles, and literature for young people. These 8,000 words are divided
into eight 1,000-word frequency levels. Level 1, the first 1,000 words, covers an average of 88.6%
of the running words in a high school textbook, and the first eight 1,000 word bands cover
98.7% of the running words.
When the frequency of word family of the stimulus words is controlled, one common
problem is that low frequency words are generally longer than high frequency words.
Researchers have shown that longer words produce longer reaction times (RTs), whether
length is defined in terms of the number of letters or the number of syllables (e.g., De Groot,
Borgwaldt, & Van, 2002; Stenneken, Conrad, & Jacobs, 2007). In order to avoid confounding
the effects of frequency and word length on reaction times, two lexical properties, the number
of letters and syllables were controlled through choosing six sets of three stimulus words, each
of which consisted of four, five, and six letters in each frequency level. Besides controlling the
number of letters and syllables, the part of the speech of the word was also controlled. Among
six sets of three stimulus words, two sets are adjectives, other two are nouns, and the rest are
verbs so that each of the part of the speech includes the equal number. Pseudo-words were
randomly generated with ARC Non-Word Database (Rastle, Harrignton, & Coltheart, 2002)
controlling for number of letters and syllables and selecting orthographically existing onsets,
bodies, and legal bigrams so that they approximately mirror real words (e.g., luig, opie, reuth,
and thafe).
4. 3. Procedures
The Lexical decision task was conducted utilizing the computer program RT builder V4
in Lextutor (Cobb, 2012). The participants individually took the test on a laptop computer in
a quiet room. They were told they would see the strings of letters that were either real words
or non-words. These instructions were given in English to the native speakers of English and
in Japanese to Japanese English teachers and Japanese university students. The participants
were asked to judge as quickly and as accurately as possible whether they knew the word.
After completing approximately 10 practice items with an oral explanation by the researcher,
they began the test. On each trial, as soon as the participant pressed 2 on the keyboard a wait
appeared on the screen. After a specified interval a word or pseudo-word randomly appeared
on the screen. The participants responded Yes (i.e., It is a word in English) by pressing 1, and
No (i.e., It is not a word in English) by pressing 3 on the keyboard. No feedback was given. Most
participants completed the test in 7 to 10 minutes.
5. Results
The results are divided into two parts. In the first section, an overview of the results is
provided and the outlying responses are discussed. In the second part, the hypotheses are
examined.
5. 1. Primary Analysis
There was a Japanese university student whose accuracy rate was 68.06% (46 errors out
of 144 items). This rate was considered to be fairly low. In a lexical decision task, participants
with an error rate of 20 percent or higher are often excluded (Jiang, 2012). Moreover, the false
alarm rate of this participant, which identifies pseudo-words as real words, was 55.5% (40 non-
Matsuo: Measuring Japanese Learners’ Lexical Accuracy and Fluency Using A Lexical Decision Task
word items out of 72 were wrongly identified as real words). This rate was exceptionally high
compared with the overall mean false alarm rate of 8% (approximately 5 items out of 72 non-
words). Therefore, this participant was excluded from further data analysis. The remaining 59
participants overall mean accuracy rate on the lexical decision task was 94.18% with accuracy
rates ranging from 81.94% to 100%.
Test scores for correctly identified real words were calculated for each level (1K, 2K, 3K,
and 4K) and for overall performance. In lexical decision experiments, outliers commonly
occur due to attention lapses, false-starts, and anticipatory responses. Generally there are two
widely used methods to define and identify outliers for visual lexical decision task (Jiang,
2012). The first is based on the absolute value, which defines cutoff points as any RT that is
shorter than 300 milliseconds (msec), which is known as the low cutoff and longer than 2,500
msec (known as the high cutoff). None of the RTs in this study fell below the low cutoff or
above the high cutoff point. An alternative method was to identify outliers is the use of standard
deviation. Outliers are usually defined as any RT that is outside 2, 2.5, or 3 standard deviations
of the mean RT of the same participant (Jiang, 2012). In order to avoid distorted results, outliers
are customarily replaced (e.g., Segalowitz & Segalowitz, 1993; Harrington, 2006). In this study,
responses more than 2.5 SDs beyond individual mean RTs were replaced with the value at
the 2.5 SD point. This affected less than 2% of the data across all the groups. Only correctly
identified real words were included in the final reaction time analyses.
5. 2. Secondary Analysis
5. 2. 1. Hypothesis 1: Lexical decision task accuracy will improve as group
proficiency level increases.
Table 1 shows the overall accuracy rate (both correctly identified real words and correctly
rejected pseudo words) by group. As Table 1 shows, the native English speakers accuracy
rate was higher than that of Japanese English teachers, and the accuracy rate of the Japanese
English teachers was higher than that of the Japanese university students. Cronbachs alpha
reliability estimates were calculated for each Japanese L1 group: Japanese university students =
.70 and Japanese English teachers = .82. Reliability estimates were not applicable to the native
English speakers due to the lack of variance, that is, almost of all the native speakers responded
accurately to the majority of the items.
Table 1. Descriptive statistics of overall accuracy performance rates by group.
Group N M SD Min Max
Japanese University Students .% . .% .%
Japanese English Teachers .%. .% .%
Native Speakers of English .% . .%.%
A one-way analysis of variance (ANOVA) was conducted to examine differences among
the groups accuracy performance. Group differences were significant, F(2,56) = 27.21,
p < .001. The strength of the relationship between overall performance accuracy and three
groups, assessed by partial 2, was very strong, with grouping factor accounting for 49% of
variance of the dependent variable.
Follow-up tests were conducted to evaluate pairwise differences among the means. The
variances among three groups ranged from 1.54 to 4.39, and the homogeneity of variance test
was significant, p < .001. Therefore, Dunnetts C was used for the post-hoc comparisons. There
were significance differences between the Japanese university students and Japanese English
teachers, and between Japanese university students and native speakers of English, p < .05.
However, the pairwise comparison between Japanese English teachers and native speakers
of English was not significant. These results partially supported hypothesis 1: lexical decision
task accuracy improved as proficiency level increased, but the difference between Japanese
English teachers and native speakers of English was not statistically significant.
5. 2. 2. Hypothesis 2: Lexical decision task accuracy will improve as word
frequency increases.
Table 2 shows the proportion of mean accuracy rate (i.e., 1 is perfect) and standard
deviation for correctly identified real words by frequency and group. As for Japanese university
students and Japanese English teachers, the accuracy means discriminate between the two
Table 2. Accuracy rate for correctly identified real words by frequency level and group.
Accuracy rate
M SD
K Japanese University Students . .
Japanese English Teachers . .
Native Speakers of English . .
K Japanese University Students . .
Japanese English Teachers . .
Native Speakers of English . .
K Japanese University Students . .
Japanese English Teachers . .
Native Speakers of English . .
K Japanese University Students . .
Japanese English Teachers . .
Native Speakers of English . .
Overall Japanese University Students . .
Japanese English Teachers . .
Native Speakers of English . .
Matsuo: Measuring Japanese Learners’ Lexical Accuracy and Fluency Using A Lexical Decision Task
groups and across the word frequency levels, although the high standard deviation of the
Japanese university students implies considerable individual variation in their responses.
On the contrary, the accuracy scores of the native speakers group did not change across all
frequency bands, which clearly indicates a ceiling effect.
In Figure 1, the Japanese university students group clearly shows the frequency effects
on accuracy for correctly responding to real words because their accuracy decreased as the
word frequency level decreased. The same tendency can be observed for the Japanese English
teachers, as they moved from the 2K to the 3K to the 4K word frequency levels; however, the
decrease in accuracy was smaller than that of the Japanese university students.
The accuracy scores were analyzed using a mixed ANOVA for subjects and
items. Group was the between-subjects factor (Japanese university students
Japanese English teachers native speakers) and frequency level was the repeated
measure factor (1K 2K 3K 4K). The sphericity assumption, which hypothesizes
that the variances of the data taken from the same participant are equal, was
violated; hence, the results were reported using the Greenhouse-Geiser correction.
Frequency effects on accuracy measure were significant, F = 14.74 (3, 168),
p < .001, partial 2 = .21. Tests of within-subjects contrast showed a linear relationship. Moreover,
there were significant interactions between word frequency and group.
As a post-hoc analysis, all the pairwise comparisons for mean accuracy scores by
frequency were conducted separately in each Japanese learner group. For the Japanese
Figure 1. Mean accuracy scores by group proficiency level and word frequency level
university students, F = 12.64 (3, 54), p < .001, partial 2 = .41, significant differences were
found for the 1K-3K, 1K-4K, and 2K-4K comparisons (p < .05, Bonferoni adjusted for multiple
comparisons). However, even though the raw data showed a decrease in accuracy by word
frequency levels for the Japanese English teacher, the effects were not significant, p = .061, partial
2 = .143. Therefore, hypothesis 2 was supported only for the Japanese university students.
5. 2. 3. Hypothesis 3: Reaction times in the lexical decision task will decrease as
group proficiency level increases.
Table 3 shows the means and standard deviations for the reaction times for correct
responses to real words by group and frequency. As the overall column in Table 3 shows, the
mean RT of Native English Speakers was faster than that of Japanese English Teachers, and the
Japanese English Teachers were faster than Japanese University Students. A one-way ANOVA
was conducted to examine group differences of overall reaction time to correctly respond to
real words. The ANOVA was significant, F(2,56) = 11.73, p < .001. Moreover, the strength of the
relationship between overall reaction time and the three groups, assessed by partial 2 was
strong, as the grouping factor accounted for 30% of the variance of the dependent variable.
Hence, hypothesis 3 was fully supported.
Table 3. Means and standard deviations for reaction time by group and frequency levels.
Reaction time (msec)
M SD
K Japanese University Students
Japanese English Teachers
Native Speakers of English
K Japanese University Students
Japanese English Teachers
Native Speakers of English
K Japanese University Students
Japanese English Teachers
Native Speakers of English
K Japanese University Students
Japanese English Teachers
Native Speakers of English
Over all Japanese University Students
Japanese English Teachers
Native Speakers of English
Matsuo: Measuring Japanese Learners’ Lexical Accuracy and Fluency Using A Lexical Decision Task
5. 2. 4. Hypothesis 4: Reaction times in the lexical decision task will decrease as
word frequency increases.
As Table 3 shows, reaction times for the Japanese university students ranged from
744 msec at the 1K level to 884 msec at the 4K level. For Japanese English teachers, mean
RTs increased from 679 msec at the 1K level to 826 at the 4K level. The Native Speakers of
English had a mean RT of 650 msec at 1K and 702 msec at 4K. The scores were analyzed
in a one-way mixed ANOVA for subjects and items. Group was the between-subjects factor
(Japanese university students Japanese English teachers native speakers of English) and
frequency levels was the repeated measure factor (1K 2K 3K 4K). Frequency effects were
significant, F (3,54) = 51.44, p < .001. Moreover, tests of within-subject contrasts were significant
(p < .001), indicating the linear relationship among the four frequency bands. See Figure 2 for
reaction time by group and word frequency.
In order to investigate the differences in reaction time among the four frequency bands,
pairwise t-tests were conducted separately for each group. For the Japanese university students
group, the results were significant, F = 21.66 (3, 54), p < .001, partial 2 = .55. Significant differences
were observed in all the pairwise comparisons except 1K-2K and 3K-4K. As for Japanese
English teachers group, the results were also significant, F = 28.83 (3, 57), p < .001, partial
2 = .60. Frequency effects on reaction time were observed in all the pairs except for 2K-3K and
3K-4K. For the native English speakers, F = 15.52 (3, 57), p < .001, partial 2 = .45. Significant
differences were observed at 1K-4K, 2K-3K, and 2K-4K. Table 4 shows the pairwise comparisons
Figure 2. Mean reaction time by group proficiency level and word frequency level
for reaction time by word frequency levels and group proficiency levels. These results generally
supported hypothesis 4.
6. Discussion
Overall accuracy (correct responses to real words and correctly rejected pseudo-
words) generally improved as the participants proficiency increased. Moreover, the standard
deviation of the accuracy scores decreased as group proficiency increased. This indicated
that the Japanese university students responded less consistently than the Japanese English
teachers, who in turn were less consistent than the native English speakers.
Frequency effects on accuracy measures were not statistically significant for neither
the Japanese English teachers nor the native English speakers. Native English speakers had
approximately the same accuracy rate across the four frequency bands, a finding that indicated
that their lexical knowledge was fully developed at least up to 4K. However, for Japanese
English teachers, frequency effects on accuracy might become significant with a large sample
size as the p-value was approaching significance (p = .06). On the other hand, frequency effects
were observed for the Japanese university students at the 3K and 4K levels. This suggests that
their lexical knowledge at 1K and 2K levels was fairly well developed; however, their lexical
knowledge was not yet fully developed at the 3K or 4K levels.
Overall reaction time when correctly identified real words was significantly different
among the three groups, a finding that clearly showed proficiency effects on reaction time.
Japanese English teachers responded to real words faster than Japanese university students,
and Native English speakers responded faster than the Japanese English teachers. This
indicates that reaction time can perhaps discriminate the lexical proficiency of English
learners, which supported the previous study (Harrington, 2006).
Table 4. Pairwise comparisons for reaction time by frequency and groups.
Level Reaction Time
differences JS JET NS
K-K ns * ns
K-K * * ns
K-K * * *
K-K * ns *
K-K * * *
K-K ns ns ns
Note
* = Difference significant at <., Bonferoni adjusted for multiple comparisons.
ns = Difference not significant. JS = Japanese university students.
JET = Japanese English teachers, NS = Native speakers of English.
Matsuo: Measuring Japanese Learners’ Lexical Accuracy and Fluency Using A Lexical Decision Task
Frequency effects on mean reaction time were generally observed in all three groups, as
all groups tended to respond more quickly to high frequency words than to low frequency
words. The lack of a significant difference in mean reaction times at the 3K-4K levels for the
Japanese university students group suggests that the 3K and 4K words were equally unfamiliar
to them. For the Japanese English teachers, the insignificant value at 2K-3K would possibly
become significant with more participants because the value was very close to significance
(p = .06).
7. Conclusion
This study measured Japanese L1 EFL learners accuracy response and reaction times,
specifically to examine group proficiency and word frequency effect (up to the 4K levels)
on accuracy and RT performance. Overall the results were aligned with the previous study
(Harrignton, 2006), which showed that reaction times and lexical decision task accuracy
generally distinguished proficiency. Moreover, a general frequency effect for reaction time
was found for all groups. Even though this study was conducted with a relatively small sample
size, RT research does not require a large number of participants scores (Jiang, 2012). There
have been many RT studies in which no more than 20 participants were tested (e.g., Costa &
Santesteban 2004; Meunier & Segui, 1999). Jiang (2012) argued that in the case of single-
presentation-list study, adding 10 participants did not change the pattern of the results from the
ten people already tested except for increasing the power of the design. Moreover, Jiang further
argued that a robust effect such as frequency effect takes a small number of participants to
materialize. However, in future research, more participants with larger variance are needed in
order to find clearer proficiency and frequency effects on accuracy and reaction time.
References
Anderson, J. R. (). The architecture of cognition. Mahwah, NJ: Erlbaum.
Anderson, J. R., & Lebiere, C. (). The atomic components of thought. Mahwah NJ: Erlbaum.
Aizawa, K., & Ishikawa. S., & Murata, T. (). JACET words. Tokyo: Kiriharashoten.
Balota, D. A., & Chumbley, J. I., (). 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(), -.
Cobb, T. (). Complete Reaction Timer (Version ) [Computer software]. Retrieved from www.
lextutor.ca/rt/
Costa, A. & Santesteban, M. (). Lexical access in bilingual speech production: Evidence from
language switching in highly proficient bilinguals and L learners. Journal of Memory and Language,
50(), -.
Coxhead, A. (). A new academic word list. TESOL Quarterly 34, -
De Groot, A. M. B., Borgwaldt, S., Bos, M., & Van den Ejinden, E. (). Lexical decision and word naming
in bilinguals: Language effects and task effects. Journal of Memory and Language, 47(), -.
Eyckmans, J. (). Measuring Receptive Vocabulary Size. LOT, Utrecht.
Grabe, W. (). Reading in a second language. Moving theory to practice. Cambridge: Cambridge
University Press.
Harrington, M. (). The lexical decision task as a measure of L lexical proficiency. In: Foster-Cohen, S.,
Kranjnovic, M.M., Digunovic J.M. (Eds.), EUROSLA Yearbook, pp. -.
Jiang, N. (). Conducting reaction time research in second language studies. New York and London,
Routledge
Kilgarriff, A. (). Putting frequencies in the dictionary. International Journal of Lexicography, 10, -.
Koda, K. (). Insights into Second Language Reading. New York, Cambridge University Press.
Kojima, M. (). Effects of word recognition speed, accuracy, and automaticity on reading ability.
Annual Review of English Language Education in Japan, 21, -.
Kilgariff, A., . Putting frequencies in the dictionary. International Journal of Lexicography, 10(), -
.
Logan, G. D. (). Toward an instance theory of automatization. Psychological Review, 95(), -.
Meunier, F. & Segui, J. () Frequency effects in auditory word recognition: The case of suffixed words.
Journal of Memory and Language, 41(), -.
Mochida, A., & Harrington, M. (). The Yes/No test as a measure of receptive vocabulary knowledge.
Language Testing 26, -.
Perfetti, C. (). Reading ability. New York, NY: Oxford University Press.
Perfetti, C. (). Comprehending written language. A blueprint of the reader. In: C. Brown, P. Hagoort
(Eds.), Neurocognition of language. Oxford: Oxford University Press, pp. -.
Perfetti, C. (). Reading ability. Lexical quality to comprehension. Scientific Studies of Reading, 11,
-.
Perfetti, C., Landi, N., Oakhill,J., (). The acquisition of reading comprehension skill. In: M. Snowling., C.
Hulme (Eds.), The science of reading. Malden, MA: Blackwell, pp. -.
Rastle, K., Harrington, J., & Coltheart, M. (). , nonwords: The ARC Nonword Database.
Quarterly Journal of Experimental Psychology, A, -.
Schmitt, N., Schmitt, D., & Clapham, C. (). Developing and exploring the behavior of two new
versions of the Vocabulary Levels Test. Language Testing, 18, -.
Segalowitz, N., & Hulstijn, J. (). Automaticity in bilingualism and second language learning. In F. F.
Kroll & A. M. B. De Groot (Eds.), Handbook of bilingualism: Psycholinguistic approaches (pp. -
). Oxford: Oxford University Press.
Segalowitz, N. S., & Segalowitz, S. J. (). Skilled performance, practice, and the differentiation of
speed-up from automatization effects: Evidence from second language word recognition. Applied
Psycholinguistics, 14, -.
Stenneken, P., Conrad, M., & Jacobs, A.M. (). Processing of syllables in production and recognition
tasks. Journal of Psycholinguistic Research, 36(), -.
Yokogawa, H., (). [Familiarity rate of the English words
for Japanese learners of written English word version ], Tokyo: Kuroshioshuppan.
