Laurence White

Integration of Multiple Speech Segmentation Cues: A Hierarchical Framework.

A central question in psycholinguistic research is how listeners isolate words from connected speech despite the paucity of clear word-boundary cues in the signal. A large body of empirical evidence indicates that word segmentation is promoted by both lexical (knowledge-derived) and sublexical (signal-derived) cues. However, an account of how these cues operate in combination or in conflict is lacking. The present study fills this gap by assessing speech segmentation when cues are systematically pitted against each other. The results demonstrate that listeners do not assign the same power to all segmentation cues; rather, cues are hierarchically integrated, with descending weights allocated to lexical, segmental, and prosodic cues. Lower level cues drive segmentation when the interpretive conditions are altered by a lack of contextual and lexical information or by white noise. Taken together, the results call for an integrated, hierarchical, and signal-contingent approach to speech segmentation.

Calibrating rhythm: First language and second language studies

Abstract This paper presents a comparative evaluation of metrics for the quantification of speech rhythm, comparing pairwise variability indices (nPVI-V and rPVI-C) and interval measures (ΔV, ΔC, %V), together with rate-normalised interval measures (VarcoV and VarcoC). First, we examined how well these metrics discriminated “stress-timed” English and Dutch and “syllable-timed” Spanish and French. Metrics of interval standard deviation such as ΔV and ΔC were strongly influenced by speech rate, but rate-normalised metrics of vocalic interval variation, VarcoV and nPVI-V, were shown to discriminate between hypothesised “rhythm classes”, as did %V, an index of the relative duration of vocalic and consonantal intervals. Second, we applied these metrics to quantifying the influence of first language on second language rhythm, with the expectation that speakers switching “rhythm classes” should show rhythm scores different from both their native and target languages. VarcoV offered the most discriminative analysis in this part of the study, with %V also suggesting insights into the process of accommodation to second language rhythm.

How stable are acoustic metrics of contrastive speech rhythm

Acoustic metrics of contrastive speech rhythm, based on vocalic and intervocalic interval durations, are intended to capture stable typological differences between languages. They should consequently be robust to variation between speakers, sentence materials, and measurers. This paper assesses the impact of these sources of variation on the metrics %V (proportion of utterance comprised of vocalic intervals), VarcoV (rate-normalized standard deviation of vocalic interval duration), and nPVI-V (a measure of the durational variability between successive pairs of vocalic intervals). Five measurers analyzed the same corpus of speech: five sentences read by six speakers of Standard Southern British English. Differences between sentences were responsible for the greatest variation in rhythm scores. Inter-speaker differences were also a source of significant variability. However, there was relatively little variation due to segmentation differences between measurers following an agreed protocol. An automated phone alignment process was also used: Rhythm scores thus derived showed good agreement with the human measurers. A number of recommendations for researchers wishing to exploit contrastive rhythm metrics are offered in conclusion.

English words on the Procrustean bed: Polysyllabic shortening reconsidered

Abstract The polysyllabic shortening hypothesis holds that the duration of a primary stressed syllable is inversely proportional to the number of additional syllables within the word. We examine the evidence for this process in British English speech by measuring the duration of primary stressed syllables in monosyllabic, disyllabic and trisyllabic words, both right-headed series – e.g. mend , commend , recommend – and left-headed series – e.g. mace , mason , masonry . In contrast with some of the original studies of polysyllabic shortening (e.g. Lehiste, 1972 ), we record target words both when carrying nuclear pitch accent and when unaccented. As in previous studies, we find strong evidence of polysyllabic shortening in accented words, an effect of comparable magnitude in right-headed and left-headed words. In unaccented words, polysyllabic shortening is minimal or absent, but there is evidence, supporting previous studies, of domain-edge effects localised to specific sub-syllabic constituents. Unlike these effects, which occur on both pitch-accented and unaccented words, polysyllabic shortening of the primary stressed syllable in these data is confined to pitch-accented words.

Structural and dialectal effects on pitch peak alignment in two varieties of British English

We report three experiments, based on test sentences read aloud, on the influence of sentence position and phonological vowel length on the alignment of accent-related f0 peaks in Scottish Standard English (SSE) and Southern British English (RP). One experiment deals with prenuclear accent peaks and the other two with nuclear accent peaks. Three findings confirm reports in the recent literature on several other European languages. First, as has been reported for Dutch [Ladd, D.R., Mennen, I., & Schepman, A. (2000). Phonological conditioning of peak alignment in rising pitch accents in Dutch. Journal of the Acoustical Society of America, 107, 2685–2696], the alignment of prenuclear peaks is later with phonologically short vowels than with long ones, and the effect cannot be explained by actual vowel duration but appears to reflect syllable structure. Second, nuclear peaks are aligned much earlier (relative to the accented vowel) than prenuclear peaks, and, as in Dutch [Schepman, A., Lickley, R., & Ladd, D.R. (2006). Effects of vowel length and ‘right context’ on the alignment of Dutch nuclear accents. Journal of Phonetics, 34, 1–28], the effect of syllable structure appears to be absent in nuclear accents; instead, their alignment is strongly influenced by whether the accented syllable is in utterance-final position. Third, as in a number of other studies, we find evidence for differences of phonetic detail between languages or language varieties: both nuclear and prenuclear peaks are aligned later in SSE than in RP, and nuclear peaks appear to be aligned earlier in English than in Dutch.

Segmentation by lexical subtraction in Hungarian speakers of second-language English

Using cross-modal form priming, we compared the use of stress and lexicality in the segmentation of spoken English by native English speakers (L1) and by native Hungarian speakers of second-language English (L2). For both language groups, lexicality was found to be an effective segmentation cue. That is, spoken disyllabic word fragments were stronger primes in a subsequent visual word recognition task when preceded by meaningful words than when preceded by nonwords: For example, the first two syllables of corridor were a more effective prime for visually presented corridor when heard in the phrase anythingcorri than in imoshingcorri. The stress pattern of the prime (strong–weak vs. weak–strong) did not affect the degree of priming. For L1 speakers, this supports previous findings about the preferential use of high-level segmentation strategies in clear speech. For L2 speakers, the lexical strategy was employed regardless of L2 proficiency level and instead of exploiting the consistent stress pattern of their native language. This is clear evidence for the primacy and robustness of segmentation by lexical subtraction even in individuals whose lexical knowledge is limited.

British English infants segment words only with exaggerated infant-directed speech stimuli

The word segmentation paradigm originally designed by Jusczyk and Aslin (1995) has been widely used to examine how infants from the age of 7.5 months can extract novel words from continuous speech. Here we report a series of 13 studies conducted independently in two British laboratories, showing that British English-learning infants aged 8-10.5 months fail to show evidence of word segmentation when tested in this paradigm. In only one study did we find evidence of word segmentation at 10.5 months, when we used an exaggerated infant-directed speech style. We discuss the impact of variations in infant-directed style within and across languages in the course of language acquisition.

A behavioral database for masked form priming

Reading involves a process of matching an orthographic input with stored representations in lexical memory. The masked priming paradigm has become a standard tool for investigating this process. Use of existing results from this paradigm can be limited by the precision of the data and the need for cross-experiment comparisons that lack normal experimental controls. Here, we present a single, large, high-precision, multicondition experiment to address these problems. Over 1,000 participants from 14 sites responded to 840 trials involving 28 different types of orthographically related primes (e.g., castfe–CASTLE) in a lexical decision task, as well as completing measures of spelling and vocabulary. The data were indeed highly sensitive to differences between conditions: After correction for multiple comparisons, prime type condition differences of 2.90 ms and above reached significance at the 5% level. This article presents the method of data collection and preliminary findings from these data, which included replications of the most widely agreed-upon differences between prime types, further evidence for systematic individual differences in susceptibility to priming, and new evidence regarding lexical properties associated with a target word’s susceptibility to priming. These analyses will form a basis for the use of these data in quantitative model fitting and evaluation and for future exploration of these data that will inform and motivate new experiments.

Communicative function and prosodic form in speech timing

Abstract Listeners can use variation in speech segment duration to interpret the structure of spoken utterances, but there is no systematic description of how speakers manipulate timing for communicative ends. Here I propose a functional approach to prosodic speech timing, with particular reference to English. The disparate findings regarding the production of timing effects are evaluated against the functional requirement that communicative durational variation should be perceivable and interpretable by the listener. In the resulting framework, prosodic structure is held to influence speech timing directly only at the heads and edges of prosodic domains, through large, consistent lengthening effects. As each such effect has a characteristic locus within its domain, speech timing cues are potentially disambiguated for the listener, even in the absence of other information. Diffuse timing effects – in particular, quasi-rhythmical compensatory processes implying a relationship between structure and timing throughout the utterance – are found to be weak and inconsistently observed. Furthermore, it is argued that articulatory and perceptual constraints make shortening processes less useful as structural cues, and they must be regarded as peripheral, at best, in a parsimonious and functionally-informed account.

Disambiguating durational cues for speech segmentation

Vowels are lengthened in lexically stressed syllables and also in word-final syllables. Both stress and final-syllable lengthening can assist in word segmentation from continuous speech, but in languages like English, with a preponderance of stress-initial words, lengthening cues may conflict for indicating word boundaries. An analysis of a large corpus of English speech demonstrated that speakers provide distributional information sufficient to potentially allow listeners to determine whether vowel lengthening is associated with lexical stress or word finality without relying on a congruence of multiple suprasegmental cues to make the distinction.