53–19.41 sec) and 19.81 sec for passages click here (range = 18.31–20.59 sec). The average duration of the Canadian speaker’s stimuli was 17.33 sec for target word lists (range = 16.85–17.80 sec) and 20.24 sec for passages (range = 18.77–21.53 sec). An important consideration is how the speakers used in this work compare with those in the cross-accent experiments of Schmale and Seidl (2009). As noted earlier, the 9-month-olds’ failure to recognize words across a native and a Spanish-accented speaker in Schmale and Seidl may have been owing to the accents varying on several suprasegmental and subphonemic dimensions. In contrast, the speakers used here were predicted to deviate primarily on vowel implementation. Thus,
an examination of acoustic and perceptual differences between these speakers increases
our understanding of the type of variation present in these stimuli, and may shed light on the causes of the 9-month-olds’ failure in previous work. Acoustic measurements and analyses of variance (ANOVAs) with F1 and F2 in /ae/ and /I/ as dependent measures and talker (North Midland-American speaker [“MidW”], and either Spanish-accented speaker (“Span”) or Southern Ontario Canadian speaker [“Can”]) support the prediction that talkers would differ on vowel implementation, see Figure 1, particularly with respect to the backing of /ae/ by the Canadian speaker.2 These dialectal accents Talazoparib nmr were chosen because they should diverge minimally, unlike in nonnative speech, which should diverge at other levels (including general characteristics, such as fluency, and subphonemic characteristics, such as coarticulation). This claim is supported by an investigation of the rate of speech, voicing, and coarticulation of the three speakers, which show that Sitaxentan the MidW and Can speakers differ less than the MidW and Span speakers, as evident in Figure 2. First, nonnative speakers lack
the fluency that characterizes native speakers, which affects global characteristics, including speech rate (although individual variation exists; naturally, a comparison with someone who stutters would not reveal this native advantage). For example, Span exhibited a relatively constant speech rate, whereas the native speakers differ less from each other by talking much slower when uttering words in isolation (I) than within passages (P); ANOVAs with rate as outcome and talker (Midwestern and either Canadian or Spanish) and type (passage, isolation) as factors confirm that the interaction talker × type is much larger in the MidW-Span comparison, F(1, 156) = 32.01 for MidW-Span; 5.34 for MidW-Can. As for consonants, the Spanish-accented speaker produces the /k/ in candle and kingdom with a much shorter VOT than either of the English-speaking speakers, and the VOT differs more, F(1, 78) = 120.72, than in the comparison among the native talkers, F(1, 78) = 27.87.