A computational analysis of uniqueness points for Japanese auditory word recognition

Abstract

This paper discusses uniqueness points for all words in a 65<th>000-word computerized lexicon in Japanese. The uniqueness point is the point in a left-to-right scan where the word is distinguished from all other words in the lexicon. This is an important concept in the Cohort Theory of auditory word recognition. Four different analyses were conducted based on two variables: the processing unit (segment or mora) and the phonetically transcribed representation (with or without pitch information). The probability of the uniqueness point being prior to the end of the word was highest when pitch information was included and/or when words were represented in segments rather than in moras. Still, in the best combination of variables (segments with pitch information) only 48% of the words reached a uniqueness point before the end of the word. This study predicts that if Japanese listeners use small processing units and attend to pitch, recognition performance will be maximized, but for most words this optimal processing strategy will still result in a Cohort size greater than one at word’s end.