I just finished collecting data for a new project (read: distractor from dissertation writing) that looks at phonological interference in Japanese/English cross-script priming. Put simply, priming studies look at how the prior presentation of a word or sound (the prime) affects the time needed to make a decision on a target word that follows. By manipulating the prime, and observing the resulting effect on the target word, we can make predictions about the organization of the first and second language lexicons in the case of cross-language priming.
Now, considering the huge number of English loanwords in Japanese, I wanted to see if manipulating the relationship between katakana script primes either helped or hindered the recognition of certain English words by Japanese learners of English. To do this I selected English words that can be read as “false friends” in Japanese—words that orthographically adhere to the Japanese mora structure like English “take” as タケ, “mine” ミネand “bike” ビケ. I used a masked priming technique where the katakana primes were flashed onscreen very briefly (50 ms—too quick to really recognize). This masking assumes that the visual or sound properties of the primes, at some low level of processing, will somehow influence target decision times. I used three types of primes. The first were “identity” primes, which are the correct katakana reading of the English target (カム—come). The second were the false friend-pairs (コメ—come); the third were unrelated (ボール—come). I predicted that the identity primes would be the fastest. But, the false friends were my real interest; I thought they would be the slowest, and least accurate, of the three conditions because once the katakana prime activates the Japanese moraic reading, participants must make a greater effort to recalculate the correct English pronunciation. Essentially, competition between the first/second language phonology should yield longer reaction times and lower accuracy.
The results so far show that ID primes were clearly the fastest, replicating previous J-E priming results. But, the false friend condition, while much slower than the ID primes, was no different from the unrelated primes. From what I can tell at this point, it looks like participants weren’t being “duped” by the L1 phonology in the (コメ—come) pairs. But, interestingly, the accuracy rates for the false friend-primed targets were significantly lower than even the unrelated pairs. So, the false-friend primes affected accuracy, but not speed. I still need to consider the effects of English proficiency (TOEIC score) on RTs and accuracy. Well, be on the lookout for a more detailed analysis at the upcoming BEAL forum!