Developmental Word Recognition: A Study of L1 English Readers of L2 Japanese NOBUKO CHIKAMATSU Department of Modern Languages DePaul University 802 West Belden Avenue Chicago, IL 60614 Email: [email protected] This study focused on developmental word recognition strategies used by first language (L1) English readers of second language (L2) Japanese. There were two proficiency groups of Japanese learners. The study considered whether or not word recognition strategies are developmental and whether or not L1 orthographic interference (i.e., involvement of phonological coding) diminishes as learners gain L2 proficiency. In Experiment 1, consisting of “contextfree” lexical judgment tests controlled by visual familiarity, the higher-proficiency group showed more visual reliance and diminishing L1 English orthographic effects at the beginning stages of instruction. However, this developmental difference was not apparent in Experiment 2, in which “contextual” passage reading tests were controlled by word visual familiarity. The higher proficiency learners showed a trend toward greater reliance on visual information; however, no significant difference in visual familiarity effects was observed between the two groups. These results imply that (a) the L2 word recognition strategy is developmental and reconstructed as proficiency advances, yet (b) automaticity takes time to develop, and (c) developmental effects may be involved differently between prelexical and postlexical phonology.

WORD RECOGNITION IS THE MOST BASIC and most critical process in reading. It is a process by which a reader identifies a string of printed letters as a meaningful unit (Wolf, Vellutino, & Gleason, 1998). A number of previous studies involving first language (L1) reading have produced strong evidence of a direct correlation between reading ability and word recognition skills, from the early stages of reading in children to advanced levels of reading in adults (Cunningham, Stanovich, & Wilson, 1990; Stanovich, 1982, 1991a, 1991b). For instance, Cunningham et al. (1990) examined L1 English college readers and discovered word-recognition skills to be one of the strongest predictors of reading success or difficulty among readers having different cognitive, verbal, and nonverbal component skills.

The Modern Language Journal, 90, i, (2006) 0026-7902/06/67–85 $1.50/0
C 2006 The Modern Language Journal

Word recognition also was a significant factor in accounting for individual variance in reading success (e.g., reading comprehension). Thus, efficient word recognition appears to be an essential skill for achieving fluency among L1 readers. Word recognition also plays a significant role in second language (L2) reading. Over the last decade, L2 reading researchers have paid more attention to word recognition, in contrast to the schema-theoretic, top-down processing views of reading that dominated L2 reading research during the 1970s and the 1980s. Recent research has demonstrated the essential role of word recognition for successful L2 reading (Chikamatsu, 2003; Grabe, 1991; Haynes & Carr, 1990; Koda, 1992, 1994, 1996). However, the role of word recognition may differ for L1 and L2 reading. Unlike L1 readers, L2 readers often lack automatic lexical access (automaticity), which impairs their ability to make cognitive capacity available for higher level processing, and, thereby, contributes to poor reading comprehension. As Koda (1996)

68 pointed out, one probable source of this impairment could be cross-linguistic differences between the L1 and L2. For instance, the difference in L1 orthographic “depth,” such as a shallow or deep script, appears to affect phonological coding reliance in L2 word recognition (Chikamatsu, 1996; Koda, 1990; Wang & Geva, 2003). Differences in L1 orthographic “type,” that is, representational units, such as alphabetic, syllabic, or logographic, also contribute to different degrees of phonemic awareness in L2 word recognition (Akamatsu, 1999; Haynes & Carr, 1990; Koda, 1999; Wang, Koda, & Perfetti, 2003). Thus, many studies, conducted to examine such crosslinguistic differences, have shown significant degrees of transfer of L1 word recognition strategies to L2 word recognition. Recent studies demonstrate that such crosslinguistic orthographic effects may vary according to a learner’s L2 reading experience, and suggest L2 word recognition may be developmental (Koda, 1998, 1999). A clear L1 orthographic effect on L2 word recognition often surfaces in the early stages of L2 reading acquisition, with the effect diminishing as proficiency improves (Akamatsu, 2002; Gholamain & Geva, 1999; Wang & Geva, 2003). However, other studies indicate that such an L1 effect does not seem to diminish with increased proficiency, but may interfere even at advanced proficiency levels (Akamatsu, 1998; Chitiri, Sun, Willows, & Taylor, 1992; Chitiri & Willows, 1997; Haynes & Carr, 1990). Although these observations may seem to be contradictory, the discrepancies may be explained through a close examination of the factors involved, such as L1–L2 cross-linguistic distance, task-related factors, and learner proficiency, all of which sometimes varied across previous studies, and sometimes not. Thus, it remains unclear how a learner develops L2 word recognition strategies in relation to L1 orthographic features and L2 experience. Furthermore, it is difficult to draw uniform conclusions from previous studies because they have focused extensively on English as a second language (ESL) and on Indo-European languages, with little attention paid to languages such as Japanese. Accordingly, the present study addressed developmental word recognition issues in the context of L2 Japanese involving L1 English readers of differing proficiency in both contextfree and contextual word recognition settings. L1 EFFECTS ON L2 WORD RECOGNITION Ever since the introduction of the dual code processing model in the 1970s, scholars have

The Modern Language Journal 90 (2006) viewed word recognition as involving two main coding strategies, phonological and visual (Baron, 1973; Coltheart, 1978; Meyer, Schvaneveldt, & Ruddy, 1974). Subsequent studies have shown that the relative timing of and degree of involvement in the use of each strategy, or the interaction between the two strategies, seem to vary according to orthographic characteristics of the languages in question. For instance, the orthographic depth hypothesis (Durgunoglu & Oney, 1999; Frost, 1994; Frost, Katz, & Bentin, 1987; Tabossi & Laghi, 1992) proposes that “shallow” orthographies involve the activation of phonological information before lexical access (i.e., prelexical or assembled phonology), whereas “deep” orthographies involve phonological activation at or after lexical access (i.e., postlexical or addressed phonology). According to this hypothesis, alphabetic languages with regular grapheme–phoneme correspondence (GPC), such as Spanish, SerboCroatian, and Korean, involve prelexical phonology or a greater reliance on phonological information, as each visual sound constituent of a word is converted into a sound letter by letter and then into a complete phonetic internal representation of the word, prior to semantic recoding (Chitiri & Willows, 1994; Cho & Chen, 1999; Frost et al., 1987; Tabossi & Laghi, 1992). In contrast, nonalphabetic languages without systematic GPC, such as logographic Japanese or Chinese, involve postlexical phonology, or a greater reliance on visual coding where one processes the visual representation of a word as a whole unit directly related to its meaning without phonological mediation (Biederman & Tsao, 1979; Hoosain, 1991; Huang & Hanley, 1994; Huang & Wang, 1992; Ju & Jackson, 1995; Leck, Weekes, & Chen, 1995; Weekes, Chen, & Lin, 1998). Another view of word recognition, known as the universal phonological principle, predicts a primary automatic activation of phonological information, regardless of orthography (Lesh & Pollatsek, 1993; Lukatela & Turvey, 1994; Van Orden, 1987). In other words, according to this principle, prelexical phonological coding dominates even in deep orthographies, such as in Chinese (Chua, 1999; Perfetti & Tan, 1998; Perfetti & Zhang, 1995; Tan, Hoosain, & Peng, 1995; Tan & Perfetti, 1998; Xu, Pollatsek, & Potter, 1999). Although seemingly contradictory, some scholars claim that the two theories could be compatible. Even though universal automatic phonological coding is inevitable and involved in all languages including logographs, variation in orthographic effects is observed in the degree of involvement or timing of phonological coding across languages

Nobuko Chikamatsu (Perfetti, Zhang, & Berent, 1992; Sakuma, Sasamura, Tatsumi, & Masaki, 1998; Wydell, Patterson, & Humphreys, 1993). In view of these differences, L2 word recognition strategy may be affected by a learner’s L1 orthography. In other words, in L2 learning, adult learners may adopt or transfer their L1 word recognition strategy, which has already been developed on the basis of their L1. The best-known example of such L1 orthographic transfer is the L1 logographic effect observed in an L2 alphabetic language. There is a heavy visual coding reliance among L1 logographic groups performing in ESL. Koda (1987, 1988, 1990) reported that in lexical decision and word memory tests ESL learners with L1 logographic Chinese and Japanese were either more impaired by the unavailability of visual information or less impaired by phonological inaccessibility than shallower L1 alphabetic groups (e.g., English, Spanish, or Arabic). Akamatsu (1999) also determined that L1 Chinese and Japanese learners slowed down more in case alternation conditions in English word naming than did L1 alphabetic Persian learners, presumably because of the inaccessibility of visual information in words. Similarly, Wang et al. (2003) observed only phonological interference in semantic category judgment tests with homophones and orthographic foils among L1 alphabetic Korean learners, whereas orthographic interference surfaced only among L1 Chinese learners. Other studies support the finding of L1 orthographic depth effects on L2 English word recognition among L1 Chinese learners of English (Haynes & Carr, 1990; Wang & Geva, 2003) and across L1 Japanese, Spanish, or Arabic learners (Brown & Haynes, 1985). Similar results occurred for nonalphabetic L2s, such as Japanese. Chikamatsu (1996) observed more reliance on visual information among L1 Chinese learners than among L1 English learners of Japanese in Japanese syllabic kana word lexical judgment tests controlled for visual familiarity and word length. Mori (1998) reported L1 orthographic effects on L2 Japanese logographic kanji memory tests. In her study, the memory of L1 English participants deteriorated significantly due to phonologically inaccessible characters (i.e., pseudo kanji), compared with the memory of L1 Chinese participants, whose performance showed strong visual processing strategies in memorizing pseudo kanji characters. These cross-linguistic studies, then, revealed different degrees of visual (or phonological) coding involvement in L2 word recognition as a function of a learner’s L1 orthographic depth.

69 An L1 orthographic type effect (i.e., representational units, such as alphabet, syllabary, and logograph) was also apparent in the area of phonemic awareness or sensitivity to intraword structure in L2 alphabetic languages (Chitiri et al., 1992; Chitiri & Willows, 1997; Haynes & Carr, 1990; Muljani, Koda, & Moates, 1998; Wang et al., 2003). Koda (1999) noted that L1 Korean learners of English were more accurate than L1 Chinese learners in English orthographic acceptability judgment tests controlled by letter sequence legality. This result indicates that Korean learners were perhaps more aware of phonemic structures and sensitive to orthographic regularity in L2 English than Chinese learners because of Korean alphabetic features. Holm and Dodd (1996) detected a L1 effect on phonemic awareness in English even between two groups of L1 Chinese learners, one from the People’s Republic of China with experience with pinyin (an alphabetic aid designed to facilitate learning logographic Chinese) and the other from Hong Kong without pinyin experience. The latter group’s poor performance in L2 English word recognition was attributable to the lack of alphabetic experience (without pinyin) at the early stages of L1 learning. Whereas a number of the previous studies reported such strong evidence of L1 orthographic transfer, some recent ones have shown a weaker version of L1 orthographic effects, finding a universal cognitive developmental similarity over cross-linguistic effects in L2 word recognition, regardless of the L1 ( Jackson, Lu, & Ju, 1994). For instance, Gholamain and Geva (1999) determined that L1 English–L2 Persian bilingual children performed similarly in word and nonword naming tasks with high correlation between the two languages. This outcome supported the central processing hypothesis, which claims that cognitive factors, such as verbal working memory, control word recognition regardless of orthography. Nevertheless, in Gholamain and Geva’s study, L2 Persian word recognition approximated L1 English recognition by Grade 2, right after the completion of the introduction to the Persian alphabet. This rapid acquisition in Persian was allegedly attributable to Persian’s shallow orthographic features (with vowels marked in children’s texts) and supported the researchers’ second hypothesis, the script dependent hypothesis. Thus, they concluded that both universal cognitive factors and language-specific orthographic features were involved in L2 word recognition. Koda (1998) found that neither alphabetic Korean nor nonalphabetic Chinese L1 learners exceeded each other in either L2 English

70 phonological awareness tasks or decoding tasks. However, correlational and regression analysis revealed significant connections among three factors: phonological awareness, decoding, and reading comprehension, though only among the Korean and not among the Chinese participants. Thus, Koda concluded that L1 alphabetic Korean experience does not lead to better phonological awareness or decoding performance in L2 alphabetic English, but rather to more phonological awareness and phonological involvement in word recognition and reading comprehension, compared to that demonstrated by L1 logographic Chinese learners. In other words, L1 orthography affects L2 word recognition qualitatively rather than quantitatively. In the previous studies, considerable evidence for L1 orthographic effects on L2 word recognition was observed, possibly either facilitating or interfering with L2 skill development. When such L1–L2 orthographic difference interferes with L2 learning in its beginning stages, can the initial recognition strategy be transformed into a more efficient one over the course of L2 development? To answer this question, we need to examine the development of L2 word recognition strategies in relation to both L1 orthographic transfer and L2 proficiency. DEVELOPMENTAL ISSUES IN L2 WORD RECOGNITION Automaticity in word recognition, that is, automatic lexical access, is an essential skill for fluent reading. Although automaticity may develop naturally among adult L1 readers through extensive exposure and experience in their L1 reading, L2 readers seem to encounter more challenges as a result of limited L2 reading experience and cross-linguistic effects, such as those of L1 orthography. This lack of automaticity restricts cognitive capacity for higher-level processing and often results in poor comprehension or slow reading rates among L2 readers (Grabe, 1991). The challenge seems to be that increased automaticity is not a simple speed-up process (improvement without increased automaticity), but rather one of restructuring (improvement with increased automaticity) the mechanism of L2 word recognition. Segalowitz, Segalowitz, and Wood (1998) distinguished between these two processes, as follows: If word recognition becomes faster and recognition time becomes more stable, then surely there has been a shift toward increased automaticity. However, this is not necessarily the case. It is possible for practice to bring about a speeding up of some of the component

The Modern Language Journal 90 (2006) mechanisms without actually eliminating or reducing reliance on less efficient, controlled processes. In other words, practice may bring about a quantitative change for speed-up without reconstructing the underlying constellation of processes. (p. 55)

Needless to say, a learner becomes more facile in recognizing a word as proficiency increases and as he or she becomes visually more familiar with written scripts. However, to determine whether more rapid word recognition is caused by speed-up or restructuring, Segalowitz et al. (1998) compared reaction speed in L2 French lexical judgment tests of fast and slow readers, who were taking a first-year college French course. Lexical judgment tests were conducted throughout the academic year. The reaction time (RT) and the coefficient of variation of RT (C Vrt = SDrt/RT, which is the standard deviation of the RT [SDrt] divided by the RT) were subjected to correlational analysis. Although both groups’ RT decreased throughout the year, the CVrt declined significantly only among faster readers, not among slower readers. This result indicates that among faster readers RT reduction was not simply attributable to a speed-up process, but to ongoing restructuring of the L2 word recognition mechanism. However, slower learners also showed signs of restructuring; a significant correlation for CVrt and RT surfaced toward the end of the year. Thus, the study suggested that L2 French learners with L1 English backgrounds could start developing automaticity from a relatively early stage. However, important questions remain unanswered. To develop automaticity, what kind of restructuring process is taking place among L2 readers? How do learners’ word recognition mechanisms change as they increase in proficiency? What is a developed, effective strategy in L2 word recognition? Are cross-linguistic factors involved heavily in such a developmental stage? One possible restructuring model is to switch coding reliance from the phonological to the visual as a learner is exposed to more L2 written text and becomes more visually familiar with L2 scripts. One piece of evidence to support this view relates to word familiarity effects. In previous L1 word recognition studies, visual coding without sound mediation occurred more often when words were visually familiar than unfamiliar, and automaticity was already developing even in sound-based scripts (Seidenberg, 1985). For instance, L1 Japanese readers used a direct visual route without phonological mediation in syllabic kana word recognition when the words were familiar (Besner & Hildebrandt, 1987; Hirose, 1992;

Nobuko Chikamatsu Yamada, Imai, & Ikeba, 1990). Thus, as L2 readers become more familiar with printed words, they may start to depend on visual coding and recognize words as a whole without any letter-by-letter sound mediation. This effect was clear among L1 alphabetic participants—who tended to rely heavily on phonological coding initially—in the course of L2 reading skill development (Chikamatsu, 1996; Mori, 1998). Second language Japanese kana word recognition in two L1 groups, English and Chinese, followed a similar pattern (Chikamatsu, 1996). There was more phonological reliance in the L1 English group and more visual reliance in the L1 Chinese group in the syllabic kana word lexical judgment tests, when controlled for visual word familiarity. This cross-linguistic difference was attributable to L1 orthographic effects on L2 word recognition; that is, the L1 English participants transferred their L1 word recognition strategy to their L2, with heavy phonological reliance at the beginning stages of learning (approximately 1 year of Japanese learning experience). Although no superiority of visual coding was evident or significant as a L1 main effect, there was an implication in the significant interaction between visual familiarity and L1 that L1 Chinese learners of Japanese may be more efficient L2 readers than L1 English learners. In short, the L1 Chinese learners recognized visually familiar words written in the conventional kana script more quickly than the L1 English learners, even though both L1 groups recognized at a similar speed visually unfamiliar words, which were pronounced as a real word but written in an unconventional kana script. This finding suggests that in regular Japanese reading (where words are written in the conventional script), Chinese learners could be more effective readers as a result of their heavy dependence on visual coding, that is, the superiority of their visual coding in L2 kana word recognition. Chikamatsu, therefore, suggested that if the language skills of L1 English learners of Japanese improve and they become increasingly familiar with Japanese words, they might develop automaticity by relying less on phonological information and more on visual coding for recognizing words in Japanese kana. Diminishing L1 negative effects of heavy reliance on phonological coding and switching to a more efficient L2 strategy can possibly be viewed as developmental restructuring in L2 word recognition. Such a pattern was indeed evident in some previous L2 word recognition studies. Sun (1991), for instance, examined L2 Chinese word recognition in learners of two different profi-

71 ciency groups—high and low—each of whose native language was English. In a word matching task, the low-proficiency group made more errors with graphic foils than either the high-proficiency groups or the L1 Chinese group; that is, the lower proficiency group could not discriminate among the minor visual differences between characters. These findings indicated that the less proficient readers had not yet automatized the visual information processing that is crucial for Chinese word processing, possibly because of L1 English effects. In short, the difference observed between the low and high groups indicated significant L1 English effects at the early stages, which may nevertheless diminish as learner proficiency advances. Akamatsu (2002) reported no L1 orthographic effects among advanced ESL learners with different L1 orthographic backgrounds (alphabetic Persian and nonalphabetic Chinese and Japanese). In the real word naming tasks controlled for spelling regularity, all groups performed faster with frequently seen words than with infrequently seen words, and showed no significant difference in visual coding dependency. Rather, all groups showed similar word frequency effects; that is, they recognized high-frequency words more quickly than low-frequency words, which reflected a significant effect of L2 reading experience on L2 word processing development, regardless of the L1. Thus, Akamatsu suggested that the L1 effect may disappear as L2 processing experience increases and decoding skill development takes place more independently from crosslinguistic orthographic effects. Other studies have also shown a similar universal word frequency effect as opposed to a cross-linguistic effect on L2 English word recognition, which reinforces the importance of L2 processing experience to build automaticity in L2 word recognition (Koda, 1999; Muljani et al., 1998). There is, however, evidence that L1 orthographic transfer effects remain and that L2 automaticity cannot be developed even at advanced levels. Haynes and Carr (1990) examined L2 English word recognition and reading performance among college first-year and fourth-year students in Taiwan and compared their performance with that of L1 English speakers. Although there were no significant differences in reading comprehension scores among the three groups, an orthographic regularity effect occurred among the L1 English learners, but not among the other groups. This result means that although the L1 English group benefited significantly from phonological accessibility, such sensitivity to visual processing did not develop among the L2

72 learners, even among fourth-year students who had read extensive amounts of English documents daily. Haynes and Carr attributed this gap between L1 and L2 readers to L1 Chinese orthographic effects. Akamatsu (1998, 1999) observed significant L1 interference among advanced ESL readers who were exposed to English documents daily during their graduate study in Canada. Japanese and Chinese learners of ESL showed heavy reliance on visual information and little phonological awareness compared with L1 Persian learners, that is, L2 proficiency did not help nonalphabetic L1 learners overcome their disadvantages reading English. Chitiri and Willows (1997) studied fluent bilinguals of L1 Greek and L2 English (who were 10th graders in a Greek-American School in Greece and had used English in school for at least 10 years). These bilingual students exhibited a different pattern from L1 English monolinguals in sensitivity to phonological and syntactic information in English letter cancellation tasks, which reflected an influence of L1 Greek orthographic and syntactic features. Furthermore, unlike English monolinguals, the bilingual participants showed no word frequency effects, which led to the conclusion that bilinguals had not yet developed automaticity or efficiency in L2 word recognition like monolinguals even after extensive usage of English during most of their academic life. In summary, previous studies have shown mixed results concerning L1 orthographic transfer effects and their interaction with L2 proficiency. To determine whether or not L2 word recognition is developmental and L1 orthographic interference diminishes with increasing proficiency, we need to consider several key factors, such as L1–L2 orthographic distance, the word recognition strategy most efficient in a given L2 (phonological or visual information dependency, phonemic awareness, etc.), and the learner’s proficiency level. The present study addresses these concerns in the context of L2 Japanese kana word recognition, among learners with a L1 English background at two different proficiency levels in Japanese. CONTEXT–FREE VERSUS CONTEXTUAL WORD RECOGNITION The majority of word recognition studies thus far have been conducted in context-free settings (words in isolation). The L1 orthographic and developmental effects already noted, however, may manifest themselves if word recognition is tested in contextual settings (words embedded

The Modern Language Journal 90 (2006) in sentences), because the activation of phonological information in words (i.e., postlexical phonology) may play a crucial role in sentence interpretation or reading comprehension. Stanovich (1991b) claimed that phonological representation serves as an access code in short-term working memory for the text integration process. If so, different proficiency levels of L2 Japanese learners may show different degrees of phonological dependency between context-free and contextual word recognition tasks. Sun (1991) examined contextual L2 Chinese word recognition performance, in addition to context-free word recognition, within the same group of L1 English learners. In the contextembedded word recognition, the participants had to decide the validity of a sentence in which one word was replaced by another Chinese word resembling the target word either graphically, phonologically, or semantically (i.e., a sentence verification task). The low-proficiency group made more errors with phonological or semantic foils than the high-proficiency and L1 groups. This outcome did not coincide with those of the context-free word recognition, where the low-proficiency group made more errors with graphic foils than either the high-proficiency or L1 English groups. Phonological coding in the contextual word recognition task, that is, postlexical phonology, was ostensibly required especially in short-term verbal memory for information integration in sentence comprehension, but less proficient readers had not yet developed such skills. The high-proficiency group, however, exhibited similar error patterns as the L1 group, that is, more errors with graphic than phonological or semantic foils in contextual word recognition, possibly due to more attention being paid to phonological information for text reading. Thus, the two proficiency groups showed different degrees of phonological dependency in the two different L2 word recognition settings, perhaps because they required different timing of phonological coding, that is, prelexical versus postlexical phonology. If so, L1 orthographic effects and restructuring of L2 word recognition may be controlled differently by L2 proficiency and by word recognition setting. Another important reason for examining L2 word recognition in contextual conditions, especially passage reading, is that word recognition is one part of a complex process of reading, yet crucial for the ultimate purpose of reading, that is, comprehension. L1 reading studies have provided substantial evidence that the lack of good word recognition skills is highly correlated with poor performance in reading

73

Nobuko Chikamatsu comprehension, and word recognition ability serves as a crucial predictor of difficulties in developing reading comprehension (Cunningham et al., 1990; Stanovich, 2000). Although there are other significant factors that distinguish good versus poor readers, such as vocabulary and grammar knowledge or sentence memory, automatic word recognition could be even more crucial, because it occurs at an early stage of the reading (i.e., lower processing), and leads to greater variance at later stages of reading (i.e., higher processing). The automaticity of word recognition has been discussed in relation to the context effect in contextual word recognition settings in L1 English reading research. One recognizes a word more rapidly or more accurately when it appears in a semantically congruent, syntactically legal sentence, or both, than when it appears alone, in a semantically incongruent, or in a syntactically illegal context ( Jordan & Thomas, 2002; Morris & Harris, 2002). Such a context facilitation effect is more evident among younger, unskilled readers or with low-frequency words than among older, more skilled readers or with high-frequency words (Borowsky & Besner, 1993, 2000; Stanovich, 1980, 1984, 1990; West & Stanovich, 1978). When word recognition is automatic among skilled readers or for familiar words, there is no need to pay attention to the context as a source of additional information, and the visual information, phonological information, or both, in the word is sufficient for lexical access. In summary, word recognition interacts with context in the reading process, and developmental automaticity in word recognition affects reading behavior beyond the word level. The use of contextual word recognition settings in which words are embedded in sentences, such as passage reading, may make it possible to examine developmental difference in automaticity of L2 word recognition more clearly than traditional context-free settings. Unfortunately, no study has yet been conducted to examine L2 developmental word recognition in contextual settings. To acquire a holistic picture of developmental L2 word recognition, the present study comprised two experiments. Experiment 1 assessed contextfree lexical decision tests, and Experiment 2 considered contextual passage-reading tests, in which Japanese kana words were controlled for visual familiarity. THE PRESENT STUDY: RESEARCH QUESTIONS The present study addressed the following research questions:

1. Is skill in L2 cognitive word recognition developmental, and, if so, does L1 orthographic interference diminish as L1 English learners of Japanese progress from phonological to visual coding dependency as L2 proficiency advances (i.e., the restructuring model)? Alternately, do they simply speed up, becoming more efficient with phonological coding strategies originally transferred from L1 (i.e., the speed-up model)? 2. Is any developmental difference apparent in word recognition strategies beyond the word level in passage-reading? In other words, does phonological coding dependency differ for the two proficiency groups in contextual word recognition? EXPERIMENT 1: LEXICAL JUDGMENT KANA WORD TEST Participants A total of 34 college-level L1 English learners of L2 Japanese from two proficiency levels participated in the study. Of these participants, 18 were novice learners of Japanese enrolled in Japanese 102 (the second semester of first-year Japanese), and 16 were intermediate learners enrolled in Japanese 202 (the second semester of second-year Japanese) at a large Midwestern university. The novice group had received 50 minutes of instruction, 5 days per week, for nearly two semesters, and the intermediate group had the same amount of instruction for nearly four semesters. There was exactly a 1-year difference in Japanese learning experience between the two groups. In the following discussion, the novice group and intermediate group are referred to as Level 1 and Level 2, respectively. Materials The lexical judgment test used two types of Japanese syllabic kana letters, hiragana and katakana. In each kana system, there are 46 basic symbols and 25 additional symbols with diacritic marks. Hiragana and katakana share the same syllabic sound representation and can be transcribed in either system, but only one is conventionally used for a given word. Hiragana is used primarily for grammatical or function words but also for some content words. Katakana is used for loan words, mainly from Western languages. In Japanese language classes, both hiragana and katakana are introduced at an early stage of instruction. At the university where the present study was conducted, hiragana and katakana were introduced during the first week and third month

74

The Modern Language Journal 90 (2006)

of the first semester (i.e., Japanese 101), respectively, and each had been used extensively in writing and reading materials in class. The third script in Japanese, kanji, is a logographic script, which is primarily used for content words, such as nouns, verbs, and adjectives. Although a few kanji words had been introduced in the second semester of the first year, the learners had limited knowledge of kanji, and kanji vocabulary was not included in the study. Furthermore, the hiragana words for which equivalent kanji transcriptions had been introduced in class were not included in the kana word test.1 The lexical judgment test followed the design of Chikamatsu (1996) and included 240 Japanese items, all written in either katakana or hiragana. Based on visual familiarity the items were categorized into three types of stimuli: visually familiar words, visually unfamiliar words, and nonwords. Visually familiar words were kana words written in the conventional script. For ex[terebi] ‘a television’ is conventionample, ally written in katakana (see Table 1). Therefore, [terebi] in katakana is visually familiar. Visually unfamiliar words, however, were words written in an unconventional script although the pronunciation of the words written in a conventional script was maintained. For instance, [terebi] ‘television’ in hiragana is a visually unfamiliar form because it is not usually written in hiragana. All of the test words had been introduced and used in Japanese classes before the experiment, except for nonword items. Nonwords were all pronounceable but were not actual Japanese words. The word length of the stimuli ranged from 2- to 5-letter words. One kana letter differentiated nonwords from real words among 2-, 3-, and 4-letter words, and two letters differentiated nonwords from real words in 5-letter words, TABLE 1 Examples of Visual Familiarity Items Pronunciation Meaning Katakana Block (120 Words) Familiar [terebi] Words (30) Unfamiliar [denwa] Words (30) Nonwords (60) [temabi] Hiragana Block (120 Words) Familiar [denwa] Words (30) Unfamiliar [terebi] Words (30) Nonwords (60) [deyawa]

television

such as [temabi] or [deyawa]. Thus, the sound information in words was familiar in both visually familiar and unfamiliar stimuli, but the visual information was familiar only in visually familiar words. For nonword items, neither the sound nor the visual information was familiar. The test times consisted of 60 visually familiar words (30 hiragana and 30 katakana), 60 visually unfamiliar words (30 hiragana and 30 katakana), and 120 nonwords (60 hiragana and 60 katakana). These items were divided into two blocks—a katakana block and a hiragana block (see Table 1). All items were randomized in each block for each participant. Furthermore, the order of the blocks was counterbalanced. In short, all participants received both blocks, but half of them in each proficiency group viewed the katakana block first and then the hiragana block, and the other half viewed the hiragana block first and then the katakana block.2 Procedure Pretest and Training. The participants presumably were familiar with hiragana and katakana, which had been introduced during the first semester of first-year Japanese. However, the study by Chikamatsu (1996) suggested that among novice learners familiarity with katakana may not be as well established as with hiragana, given that katakana was introduced several weeks later than hiragana, and fewer katakana words were used throughout instruction. To address this discrepancy, 1 week prior to the lexical judgment test, all participants completed a pretest and participated in a pretraining session to ensure an adequate knowledge of katakana. In the pretest, participants filled in a katakana chart with 46 symbols without referring to written information. Then they received a complete katakana chart filled with all the letters and practiced katakana using the software, Hiragana & Katakana Ver. 1.0 (Hatasa, Kaga, & Henstock, 1993) for 20 minutes. After the practice, they again completed a katakana chart. The low rate of errors confirmed that everyone had enough familiarity with katakana letters.3

telephone

telephone television

Lexical Judgment Kana Word Test. A week following the kana pretest, all participants were tested in one session in a college computer laboratory. They had to decide whether or not they recognized the visually presented test item (i.e., a string of kana letters) as a Japanese word. Items appeared on the screen of a Macintosh computer, and participants responded by pressing the

75

Nobuko Chikamatsu appropriate key (1 for yes and 2 for no) as quickly as possible. The test program was implemented in HyperCard. Each target item remained visible until the participant entered a keyboard response. A 1-second interval was included between items. The test program recorded reaction times (RTs) to within 1 millisecond, as well as the learners’ keyboard responses. The RTs were recorded from the time a test item appeared on the screen to the time the participant responded. Although visually unfamiliar items were written in an unconventional script, the correct response for such test items was yes if the participants knew them as Japanese words because their pronunciation was identical to visually familiar words they had learned and used in class. The correct response for nonword items was no, as it was believed that the participants could not have learned such test items as a word (as indicated in Table 2). In short, the proportion of correct positive and negative responses was even, 50% yes and 50% no. This answering procedure was clearly explained with examples during a test instruction sequence. To ensure that participants understood the procedure, they had five practice items prior to the actual test. They received no feedback. Predictions. If word recognition strategies are developmental and restructured, that is, if words become more visually familiar as proficiency develops, then a L1 English learner of Japanese should start to process a stimulus as a whole unit and to rely more on its visual representation by moving away from heavy reliance on phonological information, which Chikamatsu (1996) observed as a major word recognition strategy affected by L1 English orthographic nature among novice L1 English learners of Japanese. Based on this restructuring model, the Level 2 group in the present study would likely rely on visual informa-

Visual Information Sound Information Correct Responses

(longer)

Level 1 Level 2

RTs

(shorter) Familiar

Unfamiliar Familiarity

(longer)

Familiar Unfamiliar Words Words Nonwords (60 items) (60 items) (120 items) Familiar

FIGURE 1 Prediction with “Restructuring Model” in the Lexical Judgment Test

FIGURE 2 Prediction with “Speed-up Model” in the Lexical Judgment Test

TABLE 2 Familiarity of Visual and Sound Information of Stimuli

Stimuli

tion more than the Level 1 group; therefore, it was predicted that Level 2 participants would demonstrate a more noticeable visual familiarity effect, that is, a marked slow down under visually unfamiliar word conditions versus familiar word conditions. This prediction appeared logical because visual information is available in visually familiar words, but not in unfamiliar words as indicated in Table 2. Therefore, if Level 1 participants depended on the sound information in a word and processed it letter by letter, they should not slow down in the visually unfamiliar condition as much as Level 2 participants; in short, the visual familiarity effect would be weak. This prediction is depicted in Figure 1. However, if L1 English learners of Japanese remained dependent on phonological information, and simply sped up in word recognition with a more efficient phonological coding strategy as proficiency developed, no visual familiarity effect should be observed. In other words, since a word would be processed letter by letter, but not as a whole unit, more proficient phonological readers

Unfamiliar Unfamiliar

Familiar

Familiar

Unfamiliar

YES

YES

NO

Level 1 Level 2

RTs (shorter)

Familiar

Unfamiliar Familiarity

76

Results Mean accuracy rates (i.e., mean correct response rates) were 91.99% and 91.74% for Level 1 and Level 2 participants, respectively, and no significant difference between the two groups was observed, F (1, 32) = 1.27, n.s.; overall, Level 1 and Level 2 participants performed with equal accuracy on the word tests. To test the aforementioned predictions, mean RTs of correct responses of participants were subjected to logarithmic transformation and submitted to a two-way 2 × 3 partially repeated-measures ANOVA (Level × Familiarity), with unbalanced group size using SPSS GLM program.4 Planned multiple comparison tests were also conducted to determine whether there were significant main effects or interactions. Main Effects of Levels and Familiarity. There was a significant main effect for Level, F (1, 32) = 18.98, p < .0001. In general, Level 2 participants responded significantly more rapidly than Level 1 participants. The mean RTs were 3.970 and 5.503 seconds, respectively. There was a significant main effect for Familiarity, F (2, 64) = 282.29, p < .0001; overall, participants recognized visually familiar words (mean RT = 2.907 seconds) more rapidly than they did visually unfamiliar words (mean RT = 3.644 seconds), and visually unfamiliar words more rapidly than nonwords (mean RT = 6.298 seconds). A planned multiple comparison test revealed a significant difference between familiar and unfamiliar conditions, F (1, 32) = 372.494, p < .0001, and between unfamiliar and nonword conditions, F (1, 32) = 44.361, p < .0001. Visual Familiarity: Level × Familiarity Interaction. There was a significant interaction between Level and Familiarity, F (2, 64) = 3.55, p < .05. A planned multiple comparison test revealed a significant level effect between familiar and unfamiliar conditions, F (1, 32) = 4.783, p < .05, but not between unfamiliar and nonword conditions, F (1, 32) = 0.260, n.s. In short, Level 2 partici-

pants slowed down in the unfamiliar condition at a greater rate than did those at Level 1 even though in general, Level 2 learners responded significantly more rapidly than their Level 1 counterparts (see Figure 3). The differences in the mean RTs between familiar and unfamiliar conditions were 0.802 seconds and 0.676 seconds for Level 2 and Level 1, respectively (see Table 3). The performance of Level 2 participants was affected by visual inaccessibility in unfamiliar word conditions to a greater degree than that of the Level 1 participants. The greater effect of visual familiarity observed among Level 2 participants supported the restructuring model, that is, Level 2 participants relied on visual information of the words more than Level 1 participants did, as predicted in Figure 1. DISCUSSION In Experiment 1—lexical judgment kana tests—the Level 2 learners responded significantly faster than the Level 1 learners. But what accounts for this difference? There was a significant interaction between Level and Familiarity; the performance of Level 2 participants slowed down in the visually unfamiliar condition at a greater rate than that of the Level 1 participants. There was a main effect for level, that is, in general, Level 2 particiFIGURE 3 Level x Familiarity (RTs) in Lexical Judgment Level x Familiarity 5 Level 1 Level 2

4 RTs (sec)

would process words efficiently even in the visually unfamiliar condition because each letter in a given word is familiar enough for learners to sound it out in both visually familiar and unfamiliar word conditions. Based on this speed-up model, in the present experiment, Level 2 participants would seemingly respond faster in general and slow down less in the visually unfamiliar condition than Level 1 participants. This prediction is depicted in Figure 2.

The Modern Language Journal 90 (2006)

3 2 1 Familiar Unfamiliar Visual Familiarity

TABLE 3 Mean Reaction Times in Seconds (and Standard Deviations) of Level × Familiarity

Level 1 Level 2

Familiar

Unfamiliar

Difference

3.516 (.992) 2.227 (.480)

4.192 (1.182) 3.029 (.732)

0.676 0.802

77

Nobuko Chikamatsu pants responded much more quickly than Level 1 participants. However, Level 2 participants slowed down in the unfamiliar word recognition condition significantly more at 0.802 seconds compared to Level 1 participants at 0.676 seconds. Based on the speed-up model, Level 2 learners would have simply been more familiar with Japanese letters or words and responded more rapidly; hence, the extent of response time deterioration in the unfamiliar condition should have been less than that of the Level 1 group. In contrast, the results of the lexical test showed that Level 2 learners slowed down at a greater rate than Level 1 learners, as the restructuring model predicted. Thus, we can conclude that cognitive strategies in L2 word recognition are developmental, and that the heavy phonological dependency affected by L1 orthographic features diminishes as proficiency advances from novice to intermediate levels— a relatively short course of L2 Japanese reading acquisition. In Experiment 1, the developmental restructuring change was observed in context-free word recognition, that is, words in isolation. However, given the assumption that (postlexical) phonology in word recognition may be crucial or required even more for information integration, different patterns of phonological reliance might be observed in sentence comprehension conditions between the two groups. Furthermore, if automaticity has started to develop in higher proficiency groups, the strategic difference observed in context free word recognition could create a greater gap in a higher processing task, such as a passage reading comprehension. Experiment 2 was conducted with the same groups to assess reading comprehension performance controlled for word visual familiarity where word recognition took place in sentential contexts in the interests of testing this possibility. EXPERIMENT 2: PARAGRAPH READING TEST Participants The same 34 individuals participated in the second experiment as in the first experiment.

contained kana words written in a conventional script. In unfamiliar conditions, the kana words were in an unconventional script. For instance, in the unfamiliar condition the conventional [resutoran] ‘restaurant’ katakana word in hiragana and the was written as [yasai] ‘vegetable’ was writhiragana word in katakana.5 The length of the ten as paragraphs varied from 143 to 157 letters.6 Each contained 10 to 12 visually controlled kana words, including five or six in hiragana and another five or six in katakana. The vocabulary and grammar in each passage had been introduced in the first-year course and subsequently used in class. Four multiple-choice reading comprehension questions were prepared for each paragraph (see Appendix). Procedures All participants were tested in one session in a college computer laboratory after a short break following the completion of the lexical judgment tests discussed in Experiment 1. Each participant read the four visually controlled passages, consisting of two visually familiar and two visually unfamiliar texts with four different contents. The participants started with a familiar passage followed by an unfamiliar one with a different content; the order of the four contents was randomized.7 In the computerized reading test, the participants read a passage silently as quickly as possible. Each passage appeared on the computer screen after the respondent hit the space bar. Immediately upon completing the passage, the participant had to respond to four reading comprehension questions in a multiple choice format, with four choices of answers. Participants were to answer by hitting a number key corresponding to their choice of the correct answer. This procedure occurred four times for each passage. It was clearly explained in a test instruction sequence, followed by a single trial passage with comprehension questions before each participant started the test. Reading times (in milliseconds) and reading comprehension responses were recorded. Predictions

Materials In the paragraph reading tests, four short Japanese paragraphs written in kana were prepared on familiar topics (“Mr. Brown,” “Yesterday,” “Weekend,” and “My sister”). Each passage had two levels of visual familiarity—familiar and unfamiliar. In familiar conditions, the paragraph

If word recognition is developmental and if a more proficient reader depends more heavily on visual information from the words in passage reading tests than a less proficient reader, reading comprehension performance should deteriorate more among Level 2 than among Level 1 learners when visual information is not available.

78

The Modern Language Journal 90 (2006)

FIGURE 4 Prediction of Visual Familiarity in the Reading Comprehension Test (higher)

AR (%)

Level 2 Level 1

(lower) Familiar

Unfamiliar Familiarity

In short, in the visually unfamiliar condition the accuracy of comprehension among Level 2 learners should decrease more drastically than among Level 1 learners. Although the comprehension of Level 1 learners should also deteriorate due to visual unfamiliarity of the words in the passage in the unfamiliar condition, the decrease in accuracy rates should not be as great as that of Level 2 learners (see Figure 4). Results The passage reading tests were designed to elucidate how the visual familiarity of the words embedded in a text would affect reading comprehension in the two proficiency groups. Therefore, the following analysis was conducted based on correct response rates on the reading comprehension questions (i.e., accuracy rates or ARs) although the results of reading times (RDTs) are briefly mentioned first. Participants having error rates equal to or greater than two standard deviations above the group mean for the reading test were excluded from the present analysis.8 Reading Speed: Reading Times. For each passage RDTs were subjected to logarithmic transformation and submitted to a two-way 2 × 2 ANOVA (Level × Familiarity). There was a significant main effect for Level, F (1, 29) = 16.42, p < .0001. Overall, Level 2 learners read passages more quickly than Level 1 learners. Mean RDTs for Level 1 and Level 2 were 122.361 and 73.900 seconds, respectively. There was also a significant main effect for Familiarity, F (1, 29) = 19.14, p < .0001. In short, learners read visually familiar passages significantly faster (a mean RDT of 88.757 seconds) than unfamiliar passages (M = 109.068 seconds). There was no significant interaction of Level and Familiarity, F (1, 29) = .21, n.s. Mean RDTs in

the Level 1 group were 108.862 seconds (SD = 37.02) for the familiar passages, and 135.861 seconds (SD = 57.51) for the unfamiliar passages, a 27-second increase. Mean RDTs of the Level 2 group were 67.312 seconds (SD = 14.46) for familiar passages and 80.489 seconds (SD = 16.22) for unfamiliar passages, a 13-second increase. Reading Comprehension: Accuracy Rates The ARs for reading comprehension questions were transformed into arcsine and submitted to a two-way 2 × 2 ANOVA (Level × Familiarity). Main Effects for Level and Familiarity (ARs). There was no significant main effect for Level, F (1, 29) = 2.12, n.s. Overall, Level 1 and Level 2 learners responded equally accurately. Mean ARs were 74.79% and 80.78% for Level 1 and Level 2, respectively. There was no significant main effect for Familiarity, F (1, 29) = 2.09, n.s. In short, visual familiarity of words did not appear to affect comprehension differentially. Mean ARs were 79.58% and 75.58% in familiar and unfamiliar conditions, respectively. Level × Familiarity Interaction (ARs). The main point of interest in Experiment 2 was the interaction between Level and Familiarity. The results indicated no significant interaction, F (1, 29) = 1.55, n.s. Counter to the prediction, no visual familiarity effect was observed in the reading comprehension of the two groups. Mean ARs for the Level 1 group were 74.22% for the familiar passages and 75.04% for the unfamiliar passages, a 0.82% increase. Mean ARs of the Level 2 group were 85.71% for familiar and 76.17% for unfamiliar passages, a 9.54% decrease (see Table 4 and Figure 5). DISCUSSION There was no significant main effect in ARs for Level or Familiarity, nor was there significant interaction between Level and Familiarity. Thus, no visual familiarity effect was observed in either TABLE 4 Mean Accuracy Rates (and Percentages) of Level × Familiarity Familiar Level 1 Level 2

74.22 (22.58%) 85.71 (16.16%)

Unfamiliar 75.04 (18.69%) 76.17 (16.03%)

Difference +.82 −9.54

79

Nobuko Chikamatsu FIGURE 5 Level x Familiarity (ARs) in Passage Reading Level x Familiarity (AR) 100

AR (%)

80

Level 1 Level 2

60 40 20 0 Familiar

Unfamiliar Familiarity

group, and no developmental difference in coding dependence was observed between the two groups. These findings, as already mentioned, were counter to the expectations. In short, the heavy visual dependence among Level 2 learners observed in the context-free lexical judgment test in Experiment 1 disappeared in the passage reading test in Experiment 2. Despite these findings, the descriptive data showed an interesting trend in behavioral difference between the two groups. In the Level 2 group, the accuracy rate decreased 9.54% from familiar to unfamiliar conditions, as expected. In contrast, Level 1 learners showed almost no difference between the two conditions, with a slightly better accuracy rate in the unfamiliar condition (a 0.82% increase than in the familiar condition). Although this interaction was not statistically significant, the performance of Level 2 participants may have deteriorated as a result of visual inaccessibility in the unfamiliar condition, whereas that was not the case among Level 1 participants. Thus, there was an indication that Level 2 learners may have started developmental restructuring by switching to more visual reliance in contextual word recognition, although not yet to a significant degree. One possible interpretation for the nonsignificant interaction (i.e., no developmental difference) involves differences in types and timing of phonological coding involved in word recognition. Previous research showed different degrees of phonological coding involvement between context-free and contextual word recognition. For instance, Segalowitz and H´ebert (1990) detected different patterns between skilled and unskilled French–English bilingual readers in phonological reliance in the two contexts. The bilingual participants consisted of two groups: flu-

ent balanced bilinguals, whose word recognition speed was the same in both the L1 and the L2 (i.e., skilled L2 readers), and fluent, yet unbalanced bilinguals, whose L2 recognition speed was slower than their L1 speed (i.e., unskilled L2 readers). Although the two groups showed no difference in phonological reliance in lexical judgment tests, the unskilled L2 readers revealed more phonological reliance in sentence verification tasks controlled by homophones. Segalowitz and H´ebert interpreted this outcome to mean that unskilled L2 readers relied on phonological coding for text comprehension that was not yet automatic, unlike that of skilled readers. The result was that unskilled participants made more errors with homophones in sentence verification tasks than skilled readers. Thus, regardless of whether phonological coding affects lexical access, phonological coding is crucial postlexically in working memory for higher processing, such as syntactic parsing, propositional encoding, and semantic integration, regardless of orthographic types (Glanzer, Fischer, & Dorfman, 1984; Morita & Tamaoka, 2002). This view of phonological coding in contextfree versus contextual word recognition can be applied to explain the participants’ behaviors in the present study. Unlike the skilled French–English readers in Segalowitz and H´ebert’s (1990) study, Level 2 participants in the present study were still in the process of developing their L2 word recognition strategy with a relatively short learning experience. Even though there was a 1-year difference in learning experience between the two groups, Level 2 learners had only been studying Japanese for 2 years. Compared to learners of Indo-European languages, native English speakers learning L2 Japanese require approximately three times longer to reach the intermediate level (Liskin-Gasparro, 1982).9 In this sense, then, the Level 2 learners were still at an early stage of development; as a result, they showed no difference from Level 1 learners in postlexical phonological coding in passage reading, an area in which phonological coding may play a crucial role for sentence comprehension even at a higher proficiency level. Thus, the discrepancy in phonological reliance between the present lexical decision tasks and reading passage tests implies that phonological coding may be involved in different ways in L2 word recognition (pre- vs. postlexically), and develops differently as learner proficiency improves. Such a developmental argument is enhanced with the role of automaticity of word recognition for high processing in passage reading. Automatic

80 word recognition is crucial for efficient reading comprehension (Stanovich, 2000). For the Level 2 learners who showed more visual dependency and signs of restructuring processing in contextfree word recognition (as seen in Experiment 1), we might assume that familiar word recognition would be automatic and facilitate reading comprehension. Yet, with unfamiliar word recognition, which could not be automatic as a result of linguistic features, reading comprehension behavior would deteriorate. Consequently, the gap between familiar and unfamiliar word recognition conditions could be substantial in reading comprehension among Level 2 learners, compared to Level 1 learners. However, this was not the case in the present study, as the reconstruction observed in Level 2 learners’ context-free word recognition may not have become automatic enough to make a significant difference in reading comprehension behavior beyond the word level. Another possible interpretation for the insignificant interaction between proficiency and the visual familiarity effect relates to the design of the test materials in the passage reading tests. Some hiragana words used in the passages (which were not visually controlled words), such as [watashi] ‘I,’ [gakusei] ‘student,’ and [daigaku] ‘university,’ are usually written in kanji from the early stages of Japanese learning , and , because their kanji equivalents ( , respectively) are introduced in the middle of the first year. As a result, for the present participants, the kanji equivalents may actually have been more familiar than the hiragana forms. The paragraphs in the experiment were entirely in kana, which was somewhat unnatural, especially for Level 2 learners who had been exposed to kana-kanji combined texts longer and more often than Level 1 learners by the time of the experiment. Thus, the test task of reading a text written entirely in kana may have encouraged both groups to rely more on a phonological coding strategy, as a consequence of the orthographic features of a syllabic sound-based kana script. Consequently, visual reliance may have diminished—something that could have occurred among high-proficiency learners in more natural reading settings. CONCLUSION The present study examined developmental word recognition strategies in L2 Japanese with two learner proficiency groups of L1 English speakers. The main question was whether or not L2 word recognition is developmental, and

The Modern Language Journal 90 (2006) whether L1 orthographic interference diminishes as proficiency improves. In Experiment 1, consisting of lexical judgment tests controlled for visual familiarity, the higher proficiency group showed more visual reliance, whereas the lower proficiency group exhibited more phonological reliance. In other words, L2 learners appear to switch to a more efficient word recognition strategy for reading Japanese, as L1 English orthographic effects observed at the beginning stages diminish with increased proficiency. This word recognition strategy appears to be developmental, and the restructuring process is affected by the amount of L2 word processing experience. However, this developmental difference did not surface in Experiment 2, in passage reading tests controlled for word visual familiarity. Although the higher proficiency learners displayed a trend toward greater reliance on visual information, no significant difference in visual familiarity effects for the two proficiency groups appeared in contextual word recognition. This discrepancy implies that developmental effects may be involved differently between prelexical and postlexical phonology in L2 word recognition skill development. That is, prelexical phonology (which is involved mainly in lower-level processing) may be more developmental among L1 English learners of Japanese, but postlexical phonology (which is required more for higher-level processing in sentence comprehension) may not be as developmental. Furthermore, restructuring in L2 word recognition may commence at a relatively early stage, as the Level 2 group showed, yet it takes time to build the automaticity that results in efficient reading comprehension. Although the results obtained in the present study have significant implications for developmental L2 word recognition, several additional issues remain to be investigated. First, the subtle yet insignificant difference observed between the two groups of learners needs to be examined further with more advanced learners having a longer L2 reading experience. More advanced learners, who can be assumed to be more familiar with Japanese words, could show more significant visual reliance in both context-free and contextual word recognition. Nonetheless, phonological coding could still play an important role in contextual settings, even among advanced L2 readers. Morita and Tamaoka (2002), for example, observed strong phonological reliance both in context-free and contextual kanji recognition even among fluent adult L1 Japanese readers. It would be ideal to investigate cross-linguistic subjects with different L1 ortho-

81

Nobuko Chikamatsu graphic backgrounds, such as English, Spanish, Chinese, and Korean, to assess the possible interaction between the L1 orthographic depth effects and L2 processing experience. Fortunately, it is becoming easier to include more advanced readers and cross-linguistic participants, as Japanese is being studied more widely in the world and learner proficiency has advanced. Second, logographic kanji word recognition should be investigated further with respect to developmental issues. Because kanji is logographic and may require different levels of visual dependency from sound-based alphabetic or syllabic scripts, learners of alphabetic L1s may show a different pattern from that displayed in syllabic kana word recognition, such as a delay of developmental change in kanji recognition due to L1 effects. Kanji research is also important to create a comprehensive picture of L2 Japanese word recognition development, in view of the fact that numerous kanji characters10 comprise a substantial portion of authentic Japanese language publications.11 Third, more studies of contextual word recognition are required. Given that discussions in word recognition research have mostly been based on context-free word recognition, the distinction between pre- and postlexical phonology has been an object of neglect, and a discrepancy in L1 effects or developmental behaviors is often observed across studies. Thus, there is a need for closer examinations of contextual word recognition because words embedded in sentences form a more natural context for word recognition to occur in reading. Finally, it is important to examine the relationship between word recognition and higherlevel processing in L2 Japanese reading. Reading is an interactive cognitive process between bottom-up and top-down processing. How L2 word recognition interacts with other global verbal skills, such as vocabulary and grammar knowledge, sentence parsing, information integration, and reading comprehension, remains unknown, but is an important issue for study, particularly for holistic L2 reading research and its pedagogical applications.

ACKNOWLEDGMENTS This project was funded by a University Research Council Grant at DePaul University. The author thanks the anonymous reviewers for helpful suggestions.

NOTES 1 For instance, the word [watashi] ‘I’ was first introduced as in hiragana, but later the kanji equivalent was introduced and used in class. Such words whose kanji equivalent had been introduced prior to the experiment were not included in the kana word test. 2 Familiar and unfamiliar words in both katakana and hiragana share the same pronunciation. Therefore, the test items presented first (e.g., the katakana familiar word, [terebi] ‘television’) might produce a phonological or even semantic priming effect on the other words presented later (e.g., the hiragana unfamiliar word [terebi] ‘television’), regardless of visual familiarity. The blocks were devised to reduce the likelihood that such an effect might bias data in the present study and to maintain sufficient distance between the visually familiar and unfamiliar words having the same pronunciation. Another reason for this design was so that participants would use one script throughout the block. 3 After the errors were counted, the participants who had error rates equal to or greater than two standard deviations above the group mean were to be excluded from the study. However, all participants answered correctly at the satisfactory level; thus no one was excluded on the basis of the pretest results. 4 All word and nonword test items were included in the analysis given that there was no outlier with error rates equal to or greater than two standard deviations above the grand mean of the items. 5 Some kana words included in the passage tests had first been introduced in hiragana in the first-year course. Their kanji equivalents, such as for [kuruma] ‘car,’ (in hiragana), had been introduced later in the second-year course. Words were considered visually familiar when they were written in hiragana (e.g., ) in this test condition because the Level 1 participants were never exposed to the kanji equivalents (e.g., ), and the Level 2 participants were exposed for longer periods and were more familiar with the hiragana version than the kanji form. The katakana versions (e.g., ) were used in visually unfamiliar paragraphs given that none of the subjects had seen them in katakana. 6 The length of each paragraph was as follows. Passage 1: “Mr. Brown” with 142 letters; Passage 2: “Weekend” with 153 letters; Passage 3: “Yesterday” with 157 letters; and Passage 4: “My sister” with 150 letters. The title of each paragraph was not given in the test. 7 Four sets (A, B, C, and D) of the four passages (Passages 1 to 4) were prepared with two types of familiarity (familiar and unfamiliar). Each participant was randomly assigned one of the four sets and completed them in the following order.

Set A: Passage 1 familiar, Passage 3 unfamiliar, Passage 2 familiar, Passage 4 unfamiliar; Set B: Passage 4 familiar, Passage 2 unfamiliar, Passage 3 familiar, Passage 1 unfamiliar; Set C: Passage 3 familiar, Passage 4 unfamiliar, Passage 1 familiar, Passage 2 unfamiliar;

82 Set D: Passage 2 familiar, Passage 1 unfamiliar, Passage 4 familiar, Passage 3 unfamiliar. 8 Two Level 1 participants and one Level 2 participant were excluded from the analysis as a result of the reading comprehension test. These individuals appeared to have misunderstood the test procedure. Therefore, the total number of participants was 31 (16 at Level 1 and 15 at Level 2). 9 According to the Princeton ETS Oral Proficiency Testing Manual (Liskin-Gasparro, 1982), Japanese is rated as one of the Group IV languages (Arabic, Chinese, Japanese, etc.), which require approximately 720 hours for English L1 speakers to attain intermediate proficiency, or three times longer than for Group 1 languages (Spanish, French, etc.), which require 240 hours. Level 1 participants had approximately 160 hours of Japanese language instruction in college and those at Level 2 had 320 hours. 10 According to Chikamatsu (2003), the number of existing kanji characters in the Japanese language is estimated at between 12,000 and 85,000. However, a recent corpus linguistic study (Chikamatsu, Yokoyama, Nozaki, Long, & Fukuda, 2000) revealed that among the 4,476 entries in a corpus of 1 year’s editions of a major newspaper, the 500 most frequent kanji characters accounted for approximately 80% of the total kanji usage. 11 Chikamatsu et al. (2000) reported that kanji comprised 41.38% of all printed characters in the corpus of 1 year’s editions of a major newspaper, which made kanji the most frequently occurring single type of printable characters among all other Japanese characters (i.e., hiragana, katakana, numerals, and alphabetic characters).

REFERENCES Akamatsu, N. (1998). L1 and L2 reading: The orthographic effects of Japanese on reading in English. Language, Culture and Curriculum, 11, 9–27. Akamatsu, N. (1999). The effects of first language orthographic features on word recognition processing in English as a second language. Reading and Writing: An Interdisciplinary Journal, 11, 381–403. Akamatsu, N. (2002). A similarity in word-recognition procedures among second language readers with different first language backgrounds. Applied Psycholinguistics, 23, 117–133. Baron, J. (1973). Phonemic stage not necessary for reading. Quarterly Journal of Experimental Psychology, 25, 241–246. Besner, D., & Hildebrandt, N. (1987). Orthographic and phonological codes in the oral reading of Japanese kana. Journal of Experimental Psychology: Learning, Memory, and Cognition, 13, 335–343. Biederman, I., & Tsao, I. C. (1979). On processing Chinese ideographs and English words: Some implications from Stroop-test results. Cognitive Psychology, 2, 125–132. Borowsky, R., & Besner, D. (1993). Visual word recognition: A multistage activation model. Journal of

The Modern Language Journal 90 (2006) Experimental Psychology: Learning, Memory, and Cognition, 19, 813–840. Borowsky, R., & Besner, D. (2000). Lexical access codes in visual word recognition: Are the joint effects of context and stimulus quality diagnostic? Canadian Journal of Experimental Psychology, 54, 196–207. Brown, T. L., & 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, CA: Jossey-Bass. Chikamatsu, N. (1996). The effects of L2 orthography on L2 word recognition: A study of American and Chinese learners of Japanese. Studies in Second Language Acquisition, 18, 403–432. Chikamatsu, N. (2003). Recent research in reading Japanese as a foreign language. In Y. A. Hatasa (Ed.), An invitation to second language acquisition research in Japanese (pp. 187–206). Tokyo: Kuroshio Shuppan. Chikamatsu, N., Yokoyama, S., Nozaki, H., Long, E., & Fukuda, S. (2000). A Japanese logographic character frequency list for cognitive science research. Behavior Research Methods, Instruments, and Computers, 32, 482–500. Chitiri, H.–F., Sun, Y., Willows, D., & Taylor, I. (1992). Word recognition in second language reading. In R. Harris (Ed.), Cognitive processing in bilinguals (pp. 287–297). Amsterdam: Elsevier. Chitiri, H.–F., & Willows, D. M. (1994). Word recognition in two languages and orthographies: English and Greek. Memory & Cognition, 22, 313–325. Chitiri, H.–F., & Willows, D. M. (1997). Bilingual word recognition in English and Greek. Applied Psycholinguistics, 18, 139–156. Cho, J. R., & Chen, H. C. (1999). Orthographic and phonological activation in the semantic processing of Korean Hanja and Hangul. Language and Cognitive Processes, 14, 481–502. Chua, F. K. (1999). Phonological recoding in Chinese logograph recognition. Journal of Experimental Psychology: Learning, Memory, and Cognition, 25, 876– 891. Coltheart, M. (1978). Lexical access in simple reading tasks. In G. Underwood (Ed.), Strategies of information processing (pp. 151–216). San Diego, CA: Academic Press. Cunningham, A. E., Stanovich, K. E., & Wilson, M. R. (1990). Cognitive variation in adult students different in reading ability. In T. Carr & B. A. Levy (Eds.), Reading and development: Component skills approaches (pp. 129–160). New York: Academic Press. Durgunoglu, A. Y., & Oney, B. (1999). A cross-linguistic comparison of phonological awareness and word recognition. Reading and Writing: An Interdisciplinary Journal, 11, 281–299. 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, 29, 116– 129.

Nobuko Chikamatsu Frost, R., Katz, L., & Bentin, S. (1987). Strategies for visual word recognition and orthographical depth: A multilingual comparison. Journal of Experimental Psychology: Human Perception and Performance, 13, 104–115. Gholamain, M., & Geva, E. (1999). Orthographic and cognitive factors in the concurrent development of basic reading skills in English and Persian. (1999). Language Learning, 49, 183–217. Glanzer, M., Fischer, B., & Dorfman, D. (1984). Shortterm storage in reading. Journal of Verbal Learning and Verbal Behavior, 23, 467–486. Grabe, W. (1991). Current development in second language reading research. TESOL Quarterly, 25, 375– 406. Hatasa, K., Kaga, M., & Henstock, P. (1993). Hiragana & Katakana (Ver. 1.0). [Computer software]. Purdue, IN: Purdue University Foreign Languages Department. Haynes, M., & Carr, T. H. (1990). Writing system background and second language reading: A component skills analysis of English reading by native speaker-readers of Chinese. In T. H. Carr & B. A. Levy (Eds.), Reading and its development: Component skills approaches (pp. 375–421). Toronto, Canada: Academic Press. Hirose, T. (1992). Recognition of Japanese kana words in printing tasks. Perceptual and Motor Skills, 75, 907–913. Holm, A., & Dodd, B. (1996). The effect of first written language on the acquisition of English literacy. Cognition, 59, 119–147. Hoosain, R. (1991). Psycholinguistic implications for linguistic relativity: A case study of Chinese. Hillsdale, NJ: Erlbaum. Huang, H. S., & Hanley, R. (1994). Phonological awareness and visual skills in learning to read Chinese and English. Cognition, 54, 73–98. Huang, J. T., & Wang, M. Y. (1992). From unit to gestalt: Perceptual dynamics in recognizing Chinese characters. In H. C. Chen & O. J. L. Tzeng (Eds.), Language processing in Chinese (pp. 3–36). Amsterdam: North Holland. Jackson, N. E., Lu, W.–H., & Ju, D. (1994). Reading Chinese and reading English: Similarities, differences, and second language reading. In V. W. Berninger (Ed.), The varieties of orthographic knowledge: Vol. 1. Theoretical and developmental issues (pp. 73–109). Dordrecht, The Netherlands: Kluwer. Jordan, T. R., & Thomas, S. M. (2002). In search of perceptual influences of sentence context on word recognition. Journal of Experimental Psychology: Learning, Memory, and Cognition, 28, 34–45. Ju, D., & Jackson, N. E. (1995). Graphic and phonological processing in Chinese character identification. Journal of Reading Behavior, 27, 299–313. Koda, K. (1987). Cognitive strategy transfer in second language reading. In J. Devine, P. L. Carrell, & D. E. Eskey (Eds.), Research in reading in English as a second language (pp. 127–144). Washington, DC: TESOL.

83 Koda, K. (1988). Cognitive process in second language reading: Transfer of L1 reading skills and strategies. Second Language Research, 4, 133– 156. Koda, K. (1990). The use of L1 reading strategies in L2 reading: Effects of L1 orthographic structures on L2 phonological recoding strategies. Studies in Second Language Acquisition, 12, 393–410. Koda, K. (1992). The effects of lower-level processing skills on FL reading performance: Implications for instruction. Modern Language Journal, 76, 502– 512. Koda, K. (1994). Second language reading research: Problems and possibilities. Applied Psycholinguistics, 15, 1–28. Koda, K. (1996). L2 word recognition research: A critical review. Modern Language Journal, 80, 450– 460. Koda, K. (1998). The role of phonemic awareness in second language reading. Second Language Research, 14, 194–215. Koda, K. (1999). Development of L2 intraword orthographic sensitivity and decoding skills. Modern Language Journal, 83, 51–64. Leck, K. J., Weekes, B. S., & Chen, M. J. (1995). Visual and phonological pathways to the lexicon: Evidence from Chinese readers. Memory and Cognition, 23, 468–476. Lesh, M. F., & Pollatsek, A. (1993). Automatic access of semantic information by phonological codes in visual word recognition. Journal of Experimental Psychology: Learning, Memory, and Cognition, 19, 285–294. Liskin-Gasparro, J. E. (1982). ETS Oral proficiency testing manual. Princeton, NJ: Educational Testing Service. Lukatela, G., & Turvey, M. T. (1994). Visual lexical access is initially phonological: I. Evidence from associative priming by words, homophones, and pseudohomophones. Journal of Experimental Psychology: General, 123, 107–128. Meyer, D. E., Schvaneveldt, R. W., & Ruddy, M. G. (1974). Functions of graphemic codes in visual word recognition. Memory and Cognition, 2, 309– 321. Mori, Y. (1998). Effects of first language and phonological accessibility on kanji recognition. Modern Language Journal, 82, 69–82. Morita, A., & Tamaoka, K. (2002). Phonological involvement in the processing of Japanese at the lexical and sentence levels. Reading and Writing: An Interdisciplinary Journal, 15, 633–651. Morris, A. L., & Harris, C. L. (2002). Sentence context, word recognition, and repetition blindness. Journal of Experimental Psychology: Learning, Memory, and Cognition, 28, 962–982. Muljani, D., Koda, K., & Moates, D. R. (1998). The development of word recognition in a second language. Applied Psycholinguistics, 19, 99–113. Perfetti, C. A., & Tan, L. H. (1998). The time course of graphic, phonological, and semantic activation

84 in visual Chinese character identification. Journal of Experimental Psychology: Learning, Memory, and Cognition, 24, 101–118. Perfetti, C. A., & Zhang, S. (1995). Very early phonological activation in Chinese reading. Journal of Experimental Psychology: Learning, Memory, and Cognition, 21, 24–33. Perfetti, C. A., Zhang, S., & Berent, I. (1992). Reading in English and Chinese: Evidence for a “universal” phonological principle. In R. Frost & L. Katz (Eds.), Orthography, phonology, morphology, and meaning (pp. 227–248). Amsterdam: Elsevier. Sakuma, N., Sasamura, S., Tatsumi, I., & Masaki, S. (1998). Orthography and phonology in reading Japanese kanji words: Evidence from the semantic decision task with homophones. Memory & Cognition, 26, 75–87. Segalowitz, N., & H´ebert, M. (1990). Phonological recoding in the first and second language reading of skilled bilinguals. Language Learning, 40, 503– 538. Segalowitz, S. J., Segalowitz, N. S., & Wood, A. G. (1998). Assessing the development of automaticity in second language word recognition. Applied Psycholinguistics, 19, 53–67. Seidenberg, M. S. (1985). The time course of phonological code activation in two writing systems. Cognition, 19, 1–30. Stanovich, K. E. (1980). Toward an interactive-compensatory model of individual differences in the development of reading fluency. Reading Research Quarterly, 16, 32–71. Stanovich, K. E. (1982). Word recognition skill and reading ability. In M. H. Singer (Ed.), Competent reader, disabled reader: Research and application (pp. 81– 102). Hillsdale, NJ: Erlbaum. Stanovich, K. E. (1984). The interactive-compensatory model of reading: A confluence of developmental, experimental, and educational psychology. Remedial and Special Education, 5, 11–19. Stanovich, K. E. (1990). Concepts in developmental theories of reading skill: Cognitive resources, automaticity, and modularity. Developmental Review, 10, 72–100. Stanovich, K. E. (1991a). Changing models of reading and reading acquisition. In L. Rieben & H. Perfetti (Eds.), Learning to read: Basic research and its implications (pp. 19–32). Hillsdale, NJ: Erlbaum. Stanovich, K. E. (1991b). Word recognition: Changing perspectives. In R. Barr, M. Kamil, P. B. Mosenthal, & P. D. Pearson (Eds.), Handbook of reading research (Vol. 2, pp. 418–452). New York: Longman. Stanovich, K. E. (2000). Progress in understanding reading: Scientific foundations and new frontiers. New York: The Guilford Press.

The Modern Language Journal 90 (2006) Sun, Y. (1991, July). Word recognition processes of second language readers of Chinese. Paper presented at the meeting of the International Conference on Second Language Acquisition in the Chinese Context, Hong Kong. Tabossi, P., & Laghi, L. (1992). Semantic priming in the pronunciation of words in two writing systems: Italian and English. Memory and Cognition, 20, 303–313. Tan, L. H., Hoosain, R., & Peng, D. (1995). Role of early presemantic phonological code in Chinese character identification. Journal of Experimental Psychology: Learning, Memory, and Cognition, 21, 43– 54. Tan, L. H., & 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: An Interdisciplinary Journal, 10, 165–200. Van Orden, G. C. (1987). A ROWS is a ROSE: Spelling, sound and reading. Memory and Cognition, 15, 181–198. Wang, M., & Geva, E. (2003). Spelling performance of Chinese children using English as a second language: Lexical and visual-orthographic processes. Applied Psycholinguistics, 24, 1–25. Wang, M., Koda, K., & Perfetti, C. A. (2003). Alphabetic and nonalphabetic L1 effects in English semantic processing: A comparison of Korean and Chinese English L2 learners. Cognition, 87, 129– 149. Weekes, B. S., Chen, M. J., & Lin, Y. B. (1998). Differential effects of phonological priming on Chinese character recognition. Reading and Writing: An Interdisciplinary Journal, 10, 201–222. West, R. F., & Stanovich, K. E. (1978). Automatic contextual facilitation in readers of three ages. Child Development, 49, 717–727. Wolf, M., Vellutino, F., & Gleason, J. B. (1998). A Psycholinguistic account of reading. In J. B. Gleason & N. B. Ratner (Eds.), Psycholinguistics (pp. 409–451). Fort Worth, TX: Harcourt Brace College Publishers. Wydell, T. N., Patterson, K. E., & Humphreys, G. W. (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, 491–514. Xu, Y., Pollatsek, A., & Potter, M. C. (1999). The activation of phonology during silent Chinese word reading. Journal of Experimental Psychology: Learning, Memory, and Cognition, 25, 838–857. Yamada, J., Imai, H., & Ikeba, Y. (1990). The use of the orthographic lexicon in reading kana words. Journal of General Psychology, 117, 311–323.

Nobuko Chikamatsu

85

APPENDIX A Sample of the Paragraph Reading Test The underlined words were controlled by visual familiarity although they were not underlined in the test materials. Topic: Weekend Visually Familiar Condition

Visually Unfamiliar Condition

English translation (The translation was not given in the test.) I went shopping with Ms./Mr. Kim by car on the weekend. At the department store, I/We bought a camera and a watch. Then, we had dinner at the restaurant. Because I like vegetables, I ate salad and bread. Ms./Mr. Kim ate fish. And, we had Canadian wine a little bit. a. What did they do on the weekend? 1. cooking & drinking 2. shopping & cooking 3. driving & photographing 4. shopping & eating b. How did they go? 1. by bus 2. on foot 3. by car 4. by train c. What did Mr. Kim eat at the restaurant? 1. salad 2. bread 3. vegetable 4. fish d. What kind of wine did they drink? 1. California 2. Canadian 3. French 4. Italian