The role of vowel phonotactics in native speech segmentation Katrin Skoruppa a, ⁎ , Andrew Nevins b , Adam Gillard b , Stuart Rosen c a German Seminar, University of Basel, Switzerland b Department of Linguistics, UCL, UK c Department of Speech, Hearing and Phonetic Sciences, UCL, UK Keywords: Speech segmentation ; Phonological cues ; Phonotactics ; Lax vowel constraint ABSTRACT Numerous studies have shown that listeners can use phonological cues such as word stress and consonant clusters to find word boundaries in fluent speech. This paper investigates whether they can also use language-specific restrictions on vowel positioning for native speech segmentation. We show that English adults can exploit the fact that typical English words do not end in a lax vowel (e.g. [*diːtʊ]) in order to segment unknown words in a nonsense phrase-picture matching task, in contrast to the null results in prior studies using lexical tasks. However, they only used this cue in quiet listening conditions, and not in the presence of background noise. Thus, like consonant clusters, the lax vowel constraint is vulnerable in adverse listening conditions. 1. Introduction Breaking down continuous speech into word units is challenging because there are no clear acoustic correlates for word boundaries (Klatt, 1980). Recent work has shown that listeners are sensitive to a variety of cues signalling word boundaries, including lexical viability (e.g. Norris, McQueen, Cutler, Butterfield, & Kearns, 2001), transitional probabilities (e.g. Saffran, Newport, & Aslin, 1996) and phonological cues, including stress (e.g. Cutler & Norris, 1988) and phonotactic (i.e. sound positioning) regularities (e.g. McQueen, 1998; Mattys, White, & Melhorn, 2005). With regard to phonological cues, English listeners will generally assume that words begin with stressed syllables (Cutler & Norris, 1988), and therefore detect embedded words (like mint) more easily before weak syllables (as in the string ˈmintesh) than before stressed syllables (as in the string minˈtayve). Mattys et al. (2005) show that in a cross-modal fragment priming task, English listeners also take consonant clusters into account for the purposes of word segmentation, and weigh them more heavily than stress cues. In their study, fragments (e.g. [kʌstə]) primed corresponding words (e.g. customer) more effectively when they were embedded in consonant clusters such as [mk] that rarely appear within words (here,[ɡɑstemkʌstə]) than in clusters such as [ŋk] in [ɡɑsteŋkʌstə]. Most of this work on phonotactic constraints in word segmentation has focused on consonants, whereas vowels have received less attention. On the one hand, consonants and vowels may indeed play different roles in speech segmentation. Nespor, Pena, and Mehler (2003) claim that consonants are more important for lexical access, whereas vowels are crucial for phrasal intonation and its relation with syntactic structure. Indeed, English adults seem to use consonants, but not vowels, during lexical access in tasks involving word reconstruction (Van Ooijen, 1996; Sharp, Scott, Cutler & Wise, 2005), word learning (Creel, Aslin, & Tanenhaus, 2006) and auditory priming (Delle Luche et al., 2014). Thus, one may indeed hypothesize that native language speech segmentation can be influenced by consonant phonotactics, but not by vowel phonotactics. However, given that some researchers have found vowel effects under certain conditions (e.g. for vowel-initial bisyllabic words, Delle Luche et al., 2014), the distinction between vowels and consonants may be less categorical, and there may be some influence of vowels in tasks involving lexical processing. The present study focuses on positional restrictions on English vowels that listeners can potentially exploit for the purposes of word segmentation. Specifically, they could make use of the fact that English words can end in tense vowels like [u] in shoe or [iː] in tea, ⁎ Correspondence to: German Seminar, University of Basel, Nadelberg 4, CH-4051 Basel. E-mail address: kskoruppa@gmail.com (K. Skoruppa). 1 Published in Journal of Phonetics 49, 67-76, 2015, which should be used for any reference to this work