In this paper we argue that a comparison of vowel systems of L1 and L2 should not be limited to measuring formants and vowel duration in speech production but should also include a contrastive... Show moreIn this paper we argue that a comparison of vowel systems of L1 and L2 should not be limited to measuring formants and vowel duration in speech production but should also include a contrastive study of the perceptual representations of the vowel systems entertained by native and non‐native users of the target language. An incorrect perceptual representation of the target sounds often lies at the heart of pronunciation difficulties of L2 speakers. To facilitate such perceptual research the present paper offers a universal vowel space in which 43 artificial sounds are sampled at perceptually equidistant steps along the dimensions of vowel height (7 steps), backness/lip rounding (9 steps). Duration can be added as an additional variable in as many steps as required by the researcher. The facility was provisionally tested in a study of the perceptual representation of the monophthongs of American English by American native listeners and by Persian learners of English. Several ways of analyzing the results of such a study are presented. The results show that native listeners distinguish tense and lax members of vowel pairs in English primarily by differences in vowel quality, while the Persian L2 listeners use vowel duration as the primary cue and largely ignore the quality cue. Show less
There is increasing evidence that non-native speech is more readily understood by listeners who share the native-language background with the speakers. Mandarin-accented English can be expected to... Show moreThere is increasing evidence that non-native speech is more readily understood by listeners who share the native-language background with the speakers. Mandarin-accented English can be expected to be better understood by Mandarin listeners than by American native listeners. The most likely reason for the effect would be that the non-native listeners fruitfully use their (intuitive) knowledge of the interfering source language (Mandarin) to classify the sounds as intended by the speaker (Cutler 2012). This phenomenon has been called the Interlanguage Speech Intelligibility Benefit (or ISIB) in its weak version (Bent & Bradlow 2003). There is also a strong version of the ISIB hypothesis which states that any non-native speaker of a language will be more intelligible to any non-native listener, simply because foreigners tend to speak more carefully and slowly than native speakers of the target language. I will draw on several published intelligibility studies, in which speakers and listeners from a wide variety of native-language backgrounds (including L1 English speakers and listeners) communicate with one another in English (Smith & Rafiqzad 1979, Bent & Bradlow 2003, Wang 2007, Van Heuven & Wang 2007, Wang & Van Heuven 2014), to assess the validity of the ISIB claim. I will show that the ISIB effect is found only occasionally and inconsistently when it is quantified in an absolute way. Generally, native listeners of the target language outperform any L2 listener, even when the L2 listener has the same mother tongue as the L2 speaker. However, if we quantify the ISIB in a relative manner, where R-ISIB is defined as the discrepancy between the actual intelligibility and the score predicted from linear addition of main effects of speaker and listener language background, the notion of interlanguage benefit begins to make more sense. It then appears that the combination of a speaker and listener who do not share the same native language suffers from a negative R-ISIB (even if one interlocutor is a native speaker of the vehicle of communication), but that any combination of speakers and listeners sharing the same mother tongue (whether L1 or L2 speakers of the vehicle of communication) show a consistently positive R-ISIB. Show less
