Effects of Mora Deletion, Nonword Repetition, Rapid Naming, and Visual Search Performance on Beginning Reading in Japanese

Annals of Dyslexia, Jun 2005 by Kobayashi, Maya Shiho, Haynes, Charles W, Macaruso, Paul, Hook, Pamela E, Kato, Junko

Stepwise regressions were conducted to determine the best set of predictors of the reading measures (accuracy, speed, comprehension) among the processing variables (mora deletion, nonword repetition, object RAN, number RAN, hiragana RAN, kanji RAN, visual search errors, and visual search speed). Results with reading accuracy as the dependent measure are shown at the top left side of table IV. The overall R-square was .39 (p

DISCUSSION

There were two major goals in this study. The first was to examine relationships among selected processing variables and reading performance in Japanese kindergartners and first graders. The second goal was to examine whether or not the current theory of early reading acquisition developed primarily from studies of English also applies to youngsters learning to read Japanese. Measures commonly used to study English might predict early reading performance in Japanese. At the kindergarten level, the processing variables included tasks of phonological analysis (mora deletion), phonological memory (nonword repetition), RAN (object, number, hiragana), and visual processing (visual search). The dependent measures assessed oral reading accuracy and speed. For the Japanese first graders, kanji RAN and a test of reading comprehension were added to the battery.

Our findings are largely consistent with those for Englishspeaking children, but differ somewhat between kindergarten and first grade. In kindergarten, the measure that most effectively predicted reading accuracy and speed was hiragana RAN, indicating that rapid visual to auditory transfer is a critical process for kindergarten-level reading in Japanese. One explanation for the importance of hiragana RAN is that most Japanese children have been exposed to hiragana characters by the time they reach kindergarten, and thus can recognize them fairly automatically by the middle of the second kindergarten year. These have become over-learned symbols and, like symbol-based RAN stimuli used in studies of reading in other languages, speed in recognizing these symbols turns out to be a robust predictor of reading fluency (e.g., Breznitz, 2001; van Daal & van der Leij, 1999; Wimmer, Mayringer, & Landerl, 2002). One might argue that the relationship between hiragana RAN performance and kindergarten reading is circular and obvious; that is, hiragana naming, of course, would predict accuracy and speed of hiragana reading. However, confirmation that the predictive power lies in rapid naming of symbolic information can be found in the strong correlation between hiragana and number RAN (r = .86) and the finding that, when hiragana RAN is removed from the regression analysis, number RAN becomes the only significant predictor of both reading accuracy and reading speed (accounting for 45% and 38% of the variance, respectively).

In first graders, somewhat different predictors emerged for reading accuracy and speed. As often seen in the acquisition of basic reading skills in English, phonological analysis (mora deletion) was a significant predictor of reading accuracy. This contrasts with results for Japanese kindergartners in which mora deletion was not tied to reading accuracy and correlated with reading speed only at the trend level (p = .06). This contrast might be explained by differing requirements of kindergarten and first grade texts in Japanese. The kindergarten text consisted of almost entirely overlearned hiragana characters (98%; 2% katakana) whereas the first grade text contained a number of recently introduced katakana characters (22%), as well as some kanji (5%). Thus, the first grade text was more complex; the inclusion of katakana might have placed higher demands on phonological processing than the kindergarten text. The first grade results are consistent with previous findings from Japanese studies indicating strong relationships between mora deletion and hiragana reading in typically achieving children and children with reading difficulties (e.g., Mann, 1986; Amano, 1989; Oishi, 1998). Note that the kindergarten results tend to lend support to the "hypothesis of granularity and transparency" proposed by Wydell and Butterworth (1999). According to this hypothesis, languages with transparent orthographies as well as languages whose orthographic units represent sounds at the word level would predict only a weak relationship between reading difficulties and phonological analysis. This holds for the kindergarten readers in that their text consisted entirely of hiragana, which is transparent. The first grade results may seem to contradict this hypothesis; however, the inclusion of recently acquired katakana may create a brief demand on phonological processing that diminishes as these symbols become overlearned.