Speakers of a non-stress language, such as Japanese, have difficulty producing contrastive lexical stress in languages such as English. In previous studies, we have investigated productions of 83 morphologically related English words that exhibit stress shift, e.g., phótogràph-phòtográphic-photógraphy, and conducted acoustic analysis of vowels in corresponding positions that have the same consonantal context but different stress levels. We found that American English (AmE) speakers grouped primary- and secondary-stressed syllables together in terms of F1 and duration to maintain a clear distinction between these and unstressed syllables. Meanwhile, Japanese learners of English (JE) grouped secondary-stressed and unstressed syllables together on the same acoustic dimensions. Slujiter and van Heuven (1996) proposed that spectral balance plays a crucial role in the production of stressed syllables in Dutch. Spectral properties of L2 speakers have not been fully investigated yet, however. Following Kreiman et al. (2015), we measured H1–H2, H2–H4, H4–2 kHz, and 2–5kHz regions of the vowel spectra in the 83 words. The H2–H4 measure showed the same pattern of stress-related groupings as duration and F1 for both AmE and JE speakers. This suggests that JE speakers exhibit yet another dimension of stress cues that diverge from those of AmE speakers. Research supported by JSPS.
Kitahara et al. (Wed,) studied this question.