An intonational cue to word segmentation in phonemically identical sequences

Abstract

How do listeners accomplish the task of word segmentation, given that, in spoken language, there are no clear and obvious cues associated with word beginnings and ends? There is now a vast body of evidence showing that listeners use their tacit knowledge of a wide range of patterns in their native language to help them segment speech, including cues from allophonic variation, phonotactic constraints, transitional probabilities, and lexical stress (e.g., Cutler & Norris, 1988;McQueen, 1998;Quené, 1992;Saffran, Newport, & Aslin, 1996). Here, we examine the possibility that cues from intonation or the melodic structure of a language can help listeners find word beginnings in the speech stream.A given stretch of speech can be consistent with multiple lexical hypotheses, and these hypotheses can begin at different points in the input. In the French sequence l'abricot / / "the apricot," segmental information could be compatible with several competing hypotheses, such as l'abri / / "the shelter," la brique / / "the brick," and la brioche / / "the brioche." Listeners are routinely confronted with such transient segmentation ambiguities, and some ambiguities are total, as in Il m'a donné la fiche/l'affiche / / "He gave me the sheet/the poster," where there is no contextual information favoring one hypothesis over another, and the lexical hypotheses fiche "sheet" and affiche "poster" both seem to be equally supported by the (lack of ) contextual information in the input.Traditional psycholinguistic models, such as TRACE (McClelland & Elman, 1986) and Shortlist (Norris, 1994), have considered the processes underlying the mapping of sensory information from the acoustic input to the stored entries in the lexicon from such a phonemic approach. In these models, segmentation can be considered as a by-product of lexical competition, in the sense that it is achieved by a process of competition between candidate words. Lexical hypotheses that are consistent with the input are activated at any moment in time, regardless of their location in the input. Bottom-up activation and lateral inhibition among competing candidates allow listeners to resolve transient ambiguities that arise by selecting candidates that account for the entire input (such as l'abricot rather than l'abri or la brique in / /). In addition, a "modified version" of the 1994 Shortlist model incorpo- 775© We investigated the use of language-specific intonational cues to word segmentation in French. Participants listened to phonemically identical sequences such as / /, C'est la fiche/l'affiche "It's the sheet/poster." We modified the f 0 of the first vowel / / of the natural consonant-initial production la fiche, so that it was equal to that of the natural vowel-initial production l'affiche (resynth-consonant-equal condition), higher (resynthconsonant-higher condition), or lower (resynth-consonant-lower condition). In a two-alternative forced choice task (Experiment 1), increasing the f 0 in the / / of la fiche increased the percentage of vowel-initial (affiche) resp...

Differential contribution of prosodic cues in the native and non-native segmentation of French speech

“…French listeners could use this acoustic information (among other cues) not only in offline tasks, but also in online ones, with target words being activated more when they matched the intended segmentation than when they did not. In a follow-up study, Spinelli et al (2010) demonstrated that raising the F 0 of /la/ in la fiche resulted in the greater selection and easier recognition of vowel-initial words (e.g., affiche) than if the F 0 had not been manipulated. In their natural stimuli, the disyllabic sequences contained the L tone of the phrase-initial accent (aligned with the offset of /la/ in la fiche and with the onset of /la/ in l'affiche) and the H* tone of the pitch accent (aligned with the offset of the vowel in /fiʃ/ ), indicating that increasing the F 0 i mmediately preceding or following the L tone resulted in more segmentation of word-initial boundaries.…”

“…Bagou and Frauenfelder (2006) report that French listeners exposed to an artificial language benefited from an increase in F 0 in word-initial syllables only when the word also contained an i ncrease in both F 0 and duration in its final syllable. This discrepancy may be due to the fact that the artificial language in Bagou and Frauenfelder (2006) did not contain function words, thus making it difficult to examine the effect of phraseinitial F 0 cues, which tend to occur at the boundaries between function and content words (for discussion, see Welby 2007;Spinelli et al 2010). Alternatively, these results might indicate that pitch accents are more reliable cues to word-final boundaries than phrase-initial accents are to word-initial boundaries in French.…”

Diphones-fr: A French database of diphone positional frequency

“…Note that Italian listeners do not seem to be sensitive to allophones that distinguish syllables (allargo) from words (al largo), either. This kind of segmentation study should be done in French, another Romance language in which listeners are sensitive to syllabic structure, but also to these kinds of allophonic cues (l'affiche vs. la fiche; Spinelli, Grimault, Meunier, & Welby, 2010;Spinelli, Welby, & Schaegis, 2007).…”

Effects of rhythm and phrase-final lengthening on word-spotting in Korean

“…For example, intonational phrases are cross-linguistically associated with acoustic disjuncture, which can be manifested by means of a pause, the presence of specific boundary tones, or significant phrase-final lengthening (Shattuck-Hufnagel & Turk, 1996). Listeners exploit these local cues to identify the boundaries of prosodic units; for instance, Welby (2007) and Spinelli et al (2010) found that French speakers use a local pitch rise as a cue to the boundary between content and function words. Local cues are exploited by listeners in artificial language learning tasks as well: Tyler & Cutler (2009) have shown that English, French and Dutch speakers use lengthening as a cue to phrase-finality and this helps them detect upcoming word onsets.…”

“…In addition, research shows that rhythmic consistency (e.g., the regular grouping of syllables into prosodically defined groups of equal length) can aid word segmentation by building up listeners' expectations on when future word boundaries will occur in speech (e.g., Dilley & McAuley, 2008, Morrill et al 2014. The overall role of prosody has been reported for a variety of languages, such as French (Christophe et al 2004), Dutch (Salverda et al 2003), and Italian (Shukla et al 2007), but studies have also shown that the effects of prosody vary by language (see e.g., Welby, 2007, andSpinelli et al 2010 on French, and Warner et al 2010 on Japanese).…”