Assimilation of Place of Articulation: Evidence from English and Japanese

This study investigated place of articulation assimilation in English and Japanese using electropalatography. Three speakers of each language were given nonwords and were asked to produce a novel word blend. These blends were constructed in such a way as to result in consonant clusters with different places of articulation, and could therefore potentially assimilate. Patterns of assimilations for the two languages were investigated using durational and articulatory information. The results showed that Japanese speakers tended to assimilate the blend whereas for Australian English speakers, the process of assimilation was more varied. These results are explained in terms of differences between the word-internal phonotactics of Japanese, in which homorganic consonant clusters are mandatory as opposed to English in which they are not. INTRODUCTION Assimilation of place of articulation (POA) has been found to differ across languages. This experiment aims to further investigate assimilation in English and Japanese, using electropalatographic and acoustic analysis. Past research has shown post-lexical assimilation is tightly constrained by the phonological representation of words in the lexicon (Cutler & Otake, 1998; Otake, Yoneyama, Cutler & van der Lugt, 1996). Otake et al. (1996) reported that responses by Japanese listeners to a target phoneme were slowed when it was preceded by another phoneme with a different place specification. This pattern of reaction times was not found for native Dutch listeners, who are not disadvantaged in this way. Otake et al.’s (1996) results suggest that the phonotactics of a language form a part of how language is processed. Differences in whether or not speakers assimilate within a word could be explained by the fact that Germanic languages, such as Dutch, German and English, permit heterorganic consonant clusters at a lexical level of phonological representation (e.g. for Dutch and German /mt/ hemd/Hemd ‘shirt’), or English /ng/ in sunglasses). On the other hand, Japanese consonant clusters are lexically homorganic (e.g. /mp/ tempura ‘battered fried food’ or /nt/ bento ‘box lunch’). Language differences were also found in an experiment in which native Dutch and Japanese speakers produced word blends that could potentially result in assimilated consonant clusters from heterorganic consonant sequences (Cutler & Otake, 1998). The blends were derived from the first and last syllable of two hypothetical town names (e.g. for Dutch: /zondorp/ + /veepa:l/ > /zonpa:l/; for Japanese: /rundo/ + /makapa/ > /runkapa/). Their analysis suggested the resulting blends contained fully assimilated consonants in Japanese (/ruNkapa/) but that the Dutch forms were unassimilated (/zonpa:l/). Cutler and Otake’s (1998) study reveals assimilation differences consistent with a language’s lexical representation of consonant clusters. However, to address whether these assimilation patterns are categorical cannot be determined from auditory analysis alone, because gestures of word-final and word-initial consonants may have overlapped (resulting in a partial assimilation). Using electropalatography (EPG), Ellis and Hardcastle (2002) have investigated nasal POA assimilation in English. They found that assimilation strategies may differ between speakers, where some speakers always do or do not assimilate and other speakers vary between the two. Also, they noted that some speakers changed strategy after a number of repetitions. For example, although a speaker’s initial productions were unassimilated, their later productions, after multiple repetitions, were assimilated. Stephenson et al. Assimilation of place of articulation Proceedings of the 9th Australian International Conference on Speech Science & Technology Melbourne, December 2 to 5, 2002.  Australian Speech Science & Technology Association Inc. Accepted after abstract review page 593 Gradient assimilation, without categorical modification of the segments, has been suggested by other experimental research (Holst & Nolan, 1995). Holst and Nolan investigated sibilant assimilation (e.g. “...restocks shelves...”) and reported that the degree of overlap between adjacent segments is variable. Productions varied between two separate segments, segments showing partial overlap and also others that showed complete assimilation. They found that even when complete assimilation occurred, this consonant cluster was significantly longer in duration than a single consonant. This suggests that even when complete assimilation occurs the resulting production is not identical to a singleton. To explore the issue of categorical versus gradient, we investigated word blends produced by native Japanese and English speakers, using both acoustic and electropalatographic techniques. EXPERIMENT 1 AUSTRALIAN ENGLISH