Memory & Cognition 1996,24(6), 744-755 Cues to speech segmentation: Evidence from juncture misperceptions and word spotting JEANVROOMEN and MONIQUE VAN ZON Tilburg University, Tilburg, The Netherlands and BEATRICE DE GELDER Tilburg University, Tilburg, The Netherlands and Unioersite Libre de Bruxelles, Brussels, Belgium The question of whether Dutch listeners rely on the rhythmic characteristics of their native language to segment speech was investigated in three experiments. In Experiment 1,listeners were induced to make missegmentations of continuous speech. The results showed that word boundaries were inserted before strong syllables and deleted before weak syllables. In Experiment 2, listeners were required to spot real CVC or CVCC words (C = consonant, V = vowel) embedded in bisyllabic nonsense strings. For CVCC words, fewer errors were made when the second syllable of the nonsense string was weak rather than strong, whereas for CVC words the effect was reversed. Experiment 3 ruled out an acoustic explanation for this effect. It is argued that these results are in line with an account in which both met­ rical segmentation and lexical competition playa role. Understanding spoken language requires that listeners Intermediating Units segment a spoken utterance into words or into some One approach, which has been adopted by several psy­ smaller unit from which the lexicon can be accessed. A chological models of spoken word recognition, is to as­ major difficulty in speech segmentation is the fact that sume that the speech signal is classified into some inter­ speakers do not provide stable acoustic cues to indicate mediate prelexical linguistic unit. The notion is that the boundaries between words or segments. At present, it is acoustic signal is categorized into segments, and once seg­ therefore unclear as to how to start a lexical access attempt ments have been identified, lexical access can proceed in the absence ofa reliable cue about where to start. Sev­ without major difficulties. While there is, as yet, no agree­ eral decades ofspeech research have not yet led to a widely ment among psycholinguists about the structure or size accepted solution for the speech segmentation problem. ofsuch a unit (e.g., phoneme, onset/rime, syllable, etc.), So far, three proposals have appeared in the literature that the syllable is clearly a segmentation unit that has cap­ are of direct relevance here. One is that the continuous tured attention. Several authors have claimed that speech speech stream is categorized into discrete segments which is segmented into syllable-sized units (for an overview, then mediate between the acoustic signal and the lexicon. see Segui, Dupoux, & Mehler, 1990). The basic idea of The second proposal is that there is an explicit mecha­ the "syllabic hypothesis" is that a lexical access attempt nism that targets locations in the speech stream where word is initiated at the beginning of each syllable. A seminal boundaries are likely to occur. The third is that word seg­ study by Mehler, Dommergues, Frauenfelder, and Segui mentation is a by-product of lexical competition. In the (1981) provided empirical evidence for such a syllable­ present study, these alternatives are considered. based speech segmentation procedure. In their study, lis­ teners detected a segment more quickly ifit corresponded exactly to the first syllable ofa word than ifit comprised more or less than the syllable. Typically, listeners detected This research was supported in part by a grant from the Human Fron­ tier of Science Programme "Processing consequences of contrasting ba more quickly in ba.lance (the dot indicates the syllable language phonologies" and from the Belgian Ministere de l'Education boundary) than in bal.con, and bal more quickly in bal.con de la Communaute Francaise ("Action de recherche concertee't-i-Lan­ than in ba.lance. The benefit of syllable-based segmen­ guage processing in different modalities: Comparative approaches), tation would be that the majority oflexical access attempts 1.V's participation in this research was made possible by a fellowship is successful, at least if contrasted with phoneme-based from the Royal Netherlands Academy ofArts and Sciences. M.v.Z. was supported by a grant from the Cooperation Center ofTilburg and Eind­ segmentation. However, an aspect that has put the syllabic hoven Universities (SOBU). We would like to extend our thanks to hypothesis in a broader context is that linguistic varia­ James McQueen, Anne Cutler, and Rene Collier, for their insightful tion appears to play an important role since perceptual comments on earlier versions of this paper, and to Theo Popelier for procedures may depend on the listener's native language. help in testing study participants. Correspondence concerning this ar­ ticle should be addressed to 1. Vroomen, Department of Psychology, The above-mentioned segment-detection results were Tilburg University, P.O.Box 90153, 5000 LE Tilburg, The Netherlands obtained with French listeners and French stimuli. Sub­ (e-mail: [email protected]). sequent studies showed that this pattern ofresults did not Copyright 1996 Psychonomic Society, Inc. 744 CUES TO SPEECH SEGMENTATION 745 hold up in English (Cutler, Mehler, Norris, & Segui, 1983, tern). Thus, in accordance with the MSS, listeners seemed 1986). With English listeners, no syllabic effects were ob­ to rely on a strategy of assuming that strong syllables tained; these listeners were equally fast in detecting ba or marked the beginning oflexical words. bal in balance and ba or bal in balcony. Cutler et al. (1986) A second line ofempirical evidence favoring the MSS attributed the asymmetric results to phonological differ­ came from a word-spotting study (Cutler & Norris, 1988). ences between French and English. A major phonological Listeners were required to monitor bisyllabic pseudo­ contrast between these languages believed to be critical words and to press a button as soon as they heard a real is the fact that English is a stress language with diverse word embedded at the beginning ofsuch a pseudoword. syllable structures and English speakers' intuitions about The listeners monitored for CVC (e.g., thin) or CVCC syllable boundaries are often vague. In contrast, French (e.g., mint) words (C = consonant, V = vowel) that were has less diverse syllable structures and syllable bound­ embedded in a pseudoword string that ended in either a aries are more clear. Cutler et a!' (1986) argued that these strong (e.g., thintayfor mintayf) or a weak syllable (e.g., factors made the syllable an appropriate segmentation thintefor mintef). In the case ofa strong syllable (thin­ unit for French but not for English. tayfand mintayf), the MSS predicts that the pseudoword will be segmented as thin_tayfand min_tayf(the under­ Explicit Segmentation score indicates the metrical segmentation boundary), The proposal made by Cutler et al. (1986) shifted atten­ whereas there is no segmentation at all in the case of a tion from the now somewhat dated question about "the size weak syllable ending (thintef and mintef). In line with ofthe intermediate unit" toward the issue ofwhere in the these predictions, the results showed that CVCC words speech signal word boundaries are likely to be perceived. like mint were harder to detect in mintayfthan in mintef, At the same time, it introduced the notion that segmenta­ whereas there was no difference for CVC words: thin tion strategies oflisteners were tuned to the phonology of embedded in thintayf was detected as quickly as thin the native language. The crucial aspect of the English embedded in thintef It was proposed that the CVCC tar­ phonology, and also the Dutch, is the metrical distinction get mint from mintayfwas divided across two segmenta­ between strong and weak syllables. Strong syllables have tion units into min], with the impeding consequence that full unreduced vowels, whereas weak syllables have re­ speech material had to be assembled across a segmenta­ duced vowels, which are usually realized as schwa. Words tion boundary. For CVC words (thin) there was no differ­ likefather, mother, or brother all start with a strong sylla­ ence between thintayfand thintefbecause the segmenta­ ble followed by a weak one, whereas words like abuse, ad­ tion trigger in thin_tayf did not penetrate thin. just, or believe start with a weak syllable followed by a strong one. Cutler and Norris (1988) proposed the metri­ A Language-Universal Account: cal segmentation strategy (MSS), which claims that English Rhythmic Segmentation listeners initiate lexical access attempts at the beginning of The metrical effects observed in English and the seem­ every strong syllable. The speech recognition system thus ingly different syllabic effects observed in French have takes the onset of strong syllables as the onset of lexical recently been combined in an approach that covers the words (i.e., content words, excluding functors). differences between these two languages. The more gen­ Prima facie evidence in favor of the MSS was ob­ eral proposal is that speech segmentation is based on lan­ tained from the lexical statistics of the English vocabu­ guage rhythm (Cutler, Mehler, Norris, & Segui, 1992; lary which, indeed, show that the success rate of the Cutler, Norris, & McQueen, in press). The rhythm ofEn­ MSS will be quite high: Content words begin three times glish can be characterized as stress-based, whereas French as often with strong syllables, and words beginning with has syllabic rhythm. This argument is in line with studies strong syllables are twice as frequent as those beginning showing that English listeners apparently use stress-based with weak syllables (Cutler & Carter, 1987). Words like segmentation (Cutler et al., 1986) and French use syllabic farther, mother, or brother thus have a more typical stress segmentation (Mehler et a!., 1981). Moreover, this more pattern than words like abuse, adjust, or believe. Subse­ general proposal led to the prediction that moraic segmen­ quent empirical evidence for the MSS came from two tation should be found in Japanese, which has moraic types ofstudies: juncture misperceptions and word spot­ rhythm.