Exploring the Lombard Effect in first language Japanese speakers of English

Speakers adjust pronunciations in noise to make communication more effective. This phenomenon, known as the Lombard Effect after Etienne Lombard who first described it in 1911, is characterized by increased intensity, elongation of vowels, emphasis on stress syllables, increased pitch range, and changes to the vowel space (see Brum and Zollinger, 2011). Because the Lombard Effect interacts with linguistic factors (e.g., Wassink et al., 2006 and Patel and Schell, 2008) we hypothesize that a speaker’s native language will interact with the Lombard Effect and that second language speakers of English will carryover these effects. To test this prediction, we recorded native speakers of North American English and L1 Japanese speakers of L2 English participating in a modified map task to illicit natural communication between the speakers and the experimenter, in two conditions: noisy and quiet (presented over headphones). Measurements of vowel duration, vowel quality, and intensity were extracted from key words representing monosyllabic and trochaic words with six vowel qualities: /i ɪ a ae u ʊ/. Preliminary results indicate some language specific differences, but also unexpectedly that vowel space is more compressed in the Lombard condition than in quiet, which has connections to findings of Zhao and Jurafsky (2009) for Mandarin.Speakers adjust pronunciations in noise to make communication more effective. This phenomenon, known as the Lombard Effect after Etienne Lombard who first described it in 1911, is characterized by increased intensity, elongation of vowels, emphasis on stress syllables, increased pitch range, and changes to the vowel space (see Brum and Zollinger, 2011). Because the Lombard Effect interacts with linguistic factors (e.g., Wassink et al., 2006 and Patel and Schell, 2008) we hypothesize that a speaker’s native language will interact with the Lombard Effect and that second language speakers of English will carryover these effects. To test this prediction, we recorded native speakers of North American English and L1 Japanese speakers of L2 English participating in a modified map task to illicit natural communication between the speakers and the experimenter, in two conditions: noisy and quiet (presented over headphones). Measurements of vowel duration, vowel quality, and intensity were extracted from key words ...