A cross-linguistic study of diphthongs in spoken word processing in Japanese and English

Kiyoko YoneyamaInterspeech 2004

TL;DR

This study investigates how Japanese and English speakers process diphthongs, revealing that Japanese listeners are sensitive to the moraic structure of their native language, treating diphthongs as two separate units. English listeners, conversely, treat diphthongs as single units, a pattern that persists even when processing Japanese.

What Problem Does This Paper Address?

Spoken language processing is influenced by the rhythmic structure of a listener's native language. Japanese, for instance, has a moraic structure, where each mora is a unit of rhythm. This study explores how this moraic sensitivity affects the processing of diphthongs (vowel sequences like "ai" or "oi") in both Japanese and English. Specifically, it questions whether Japanese listeners treat the two vowels in a diphthong as distinct moraic units, and how this contrasts with English listeners who may perceive diphthongs as single phonetic units. The research also examines whether bilingualism alters these processing strategies.

Methods

The study employed three phoneme-detecting experiments. Participants were presented with sequences of spoken words and asked to detect a specific target sound. The core methodology involved comparing detection accuracy and reaction times for two types of words: CV₁V₂ words (containing diphthongs) and CV₁ words (containing single vowels). The key innovation lies in using these specific word types and comparing performance across different language groups (native Japanese, native English, and semi-bilingual Japanese speakers of English) when listening to both Japanese and English stimuli.

Datasets

The study utilized newly constructed sets of language materials for both Japanese and English. These materials consisted of 24 content words each, divided equally between CV₁V₂ words (containing diphthongs) and CV₁ words. These experimental words were mixed with filler words into 48-word sequences. The stimuli were recorded by native speakers of Tokyo Japanese and a North Kentucky dialect of American English.

Key Insights and Ideas

  • Japanese Moraic Sensitivity: Japanese listeners demonstrate a sensitivity to the moraic structure of their language, even when the moraic vowel is the second part of a diphthong. This suggests they can break down diphthongs into their constituent moraic units.
  • English Unitary Perception: English listeners, when processing English, treat diphthongs as single units, making it harder to detect individual vowel sounds within them.
  • Cross-Linguistic Influence: Interestingly, Japanese listeners tend to treat English diphthongs as sequences of two units, similar to how they process Japanese diphthongs. Conversely, English listeners maintain their unitary perception of diphthongs even when listening to Japanese.
  • Bilingualism's Role: Semi-bilingual Japanese speakers of English show a fascinating adaptation. They treat English diphthongs as single units (like English speakers) but retain their moraic sensitivity for Japanese diphthongs. This suggests that extensive second-language exposure can lead to flexible listening strategies.

Key Results

Experiment 1 (Japanese Listeners): * Miss Rate: Targets were detected significantly more accurately in Japanese materials than English. Targets in CV₁V₂ words were detected more accurately than in CV₁ words. In English materials, targets were detected more accurately in CV₁V₂ words than CV₁ words. * Response Times: Targets were detected significantly more quickly in Japanese materials than English. Targets in CV₁V₂ words were detected more quickly than in CV₁ words. In English, targets were detected significantly more quickly in CV₁V₂ words than CV₁ words.

Experiment 2 (English Listeners): * Miss Rate: Targets were detected more accurately in CV₁ words than CV₁V₂ words in both English and Japanese. * Response Times: Targets were detected significantly more quickly in Japanese words than English. Targets were detected significantly more quickly in CV₁ words than CV₁V₂ words. Detection time differences were significant in English but not Japanese.

Experiment 3 (Semi-bilingual Japanese Speakers): * Miss Rate: Targets in CV₁ words were detected significantly more accurately than in CV₁V₂ words. An interaction showed a tendency for semi-bilinguals to be sensitive to moraic structure in Japanese. * Response Times: No significant main effect or interaction was observed.

Limitations and Future Work

The study acknowledges that the findings regarding Japanese listeners' processing of English diphthongs might suggest they are "listening to Japanese" even when presented with English. This could be due to the participants' limited exposure to English or their strong native language processing habits. Future work could explore how different levels of English proficiency in semi-bilinguals might influence these processing strategies. The authors also suggest that the development of a flexible listening strategy in semi-bilinguals warrants further investigation.

Summary updated

BibTeX
@inproceedings{yoneyama2004cross,
  author = {Kiyoko Yoneyama},
  title = {{A cross-linguistic study of diphthongs in spoken word processing in Japanese and English}},
  booktitle = {Interspeech},
  year = {2004},
}
Figure