Publications

Abstract

Large auditory-visual (AV) speech corpora are the grist of modern research in speech science, but no such corpus exists for Australian English. This is unfortunate, for speech science is the brains behind speech technology and applications such as text-to-speech (TTS) synthesis, automatic speech recognition (ASR), speaker recognition and forensic identification, talking heads, and hearing prostheses. Advances in these research areas in Australia require a large corpus of Australian English. Here the authors describe a blueprint for building the Big Australian Speech Corpus (the Big ASC), a corpus of over 1,100 speakers from urban and rural Australia, including speakers of non-indigenous, indigenous, ethnocultural, and disordered forms of Australian English, each of whom would be sampled on three occasions in a range of speech tasks designed by the researchers who would be using the corpus.

Abstract

English listeners largely disregard suprasegmental cues to stress in recognizing words. Evidence for this includes the demonstration of Fear et al. [J. Acoust. Soc. Am. 97, 1893–1904 (1995)] that cross-splicings are tolerated between stressed and unstressed full vowels (e.g., au- of autumn, automata). Dutch listeners, however, do exploit suprasegmental stress cues in recognizing native-language words. In this study, Dutch listeners were presented with English materials from the study of Fear et al. Acceptability ratings by these listeners revealed sensitivity to suprasegmental mismatch, in particular, in replacements of unstressed full vowels by higher-stressed vowels, thus evincing greater sensitivity to prosodic goodness than had been shown by the original native listener group.

Abstract

Two phoneme goodness rating experiments addressed the role of orthographic knowledge in the evaluation of speech sounds. Ratings for the best tokens of /s/ were higher in words spelled with S (e.g., bless) than in words where /s/ was spelled with C (e.g., voice). This difference did not appear for analogous nonwords for which every lexical neighbour had either S or C spelling (pless, floice). Models of phonemic processing incorporating obligatory influence of lexical information in phonemic processing cannot explain this dissociation; the data are consistent with models in which phonemic decisions are not subject to necessary top-down lexical influence.

Abstract

Three experiments, in which Japanese listeners detected Japanese words embedded in nonsense sequences, examined the perceptual consequences of vowel devoicing in that language. Since vowelless sequences disrupt speech segmentation [Norris et al. (1997). Cognit. Psychol. 34, 191– 243], devoicing is potentially problematic for perception. Words in initial position in nonsense sequences were detected more easily when followed by a sequence containing a vowel than by a vowelless segment (with or without further context), and vowelless segments that were potential devoicing environments were no easier than those not allowing devoicing. Thus asa, “morning,” was easier in asau or asazu than in all of asap, asapdo, asaf, or asafte, despite the fact that the /f/ in the latter two is a possible realization of fu, with devoiced [u]. Japanese listeners thus do not treat devoicing contexts as if they always contain vowels. Words in final position in nonsense sequences, however, produced a different pattern: here, preceding vowelless contexts allowing devoicing impeded word detection less strongly (so, sake was detected less accurately, but not less rapidly, in nyaksake—possibly arising from nyakusake—than in nyagusake). This is consistent with listeners treating consonant sequences as potential realizations of parts of existing lexical candidates wherever possible.

Abstract

Recognizing word boundaries in continuous speech requires detailed knowledge of the native language. In the first year of life, infants acquire considerable word segmentation abilities. Infants at this early stage in word segmentation rely to a large extent on the metrical pattern of their native language, at least in stress-based languages. In Dutch and English (both languages with a preferred trochaic stress pattern), segmentation of strong-weak words develops rapidly between 7 and 10 months of age. Nevertheless, trochaic languages contain not only strong-weak words but also words with a weak-strong stress pattern. In this article, we present electrophysiological evidence of the beginnings of weak-strong word segmentation in Dutch 10-month-olds. At this age, the ability to combine different cues for efficient word segmentation does not yet seem to be completely developed. We provide evidence that Dutch infants still largely rely on strong syllables, even for the segmentation of weak-strong words.

Abstract

Two artificial-language learning experiments directly compared English, French, and Dutch listeners’ use of suprasegmental cues for continuous-speech segmentation. In both experiments, listeners heard unbroken sequences of consonant-vowel syllables, composed of recurring three- and four-syllable “words.” These words were demarcated by(a) no cue other than transitional probabilities induced by their recurrence, (b) a consistent left-edge cue, or (c) a consistent right-edge cue. Experiment 1 examined a vowel lengthening cue. All three listener groups benefited from this cue in right-edge position; none benefited from it in left-edge position. Experiment 2 examined a pitch-movement cue. English listeners used this cue in left-edge position, French listeners used it in right-edge position, and Dutch listeners used it in both positions. These findings are interpreted as evidence of both language-universal and language-specific effects. Final lengthening is a language-universal effect expressing a more general (non-linguistic) mechanism. Pitch movement expresses prominence which has characteristically different placements across languages: typically at right edges in French, but at left edges in English and Dutch. Finally, stress realization in English versus Dutch encourages greater attention to suprasegmental variation by Dutch than by English listeners, allowing Dutch listeners to benefit from an informative pitch-movement cue even in an uncharacteristic position.

Abstract

Dutch listeners outperform native listeners in identifying syllable stress in English. This is because lexical stress is more useful in recognition of spoken words of Dutch than of English, so that Dutch listeners pay greater attention to stress in general. We examined Dutch listeners’ use of the acoustic correlates of English stress. Primary- and secondary-stressed syllables differ significantly on acoustic measures, and some differences, in F0 especially, correlate with data of earlier listening experiments. The correlations found in the Dutch responses were not paralleled in data from native listeners. Thus the acoustic cues which distinguish English primary versus secondary stress are better exploited by Dutch than by native listeners.

Abstract

In contrast to initial L1 vocabularies, which of necessity depend largely on heard exemplars, L2 vocabulary construction can draw on a variety of knowledge sources. This can lead to richer stored knowledge about the phonology of the L2 than the listener's prelexical phonetic processing capacity can support, and thus to mismatch between the level of detail required for accurate lexical mapping and the level of detail delivered by the prelexical processor. Experiments on spoken word recognition in L2 have shown that phonetic contrasts which are not reliably perceived are represented in the lexicon nonetheless. This lexical representation of contrast must be based on abstract knowledge, not on veridical representation of heard exemplars. New experiments confirm that provision of abstract knowledge (in the form of spelling) can induce lexical representation of a contrast which is not reliably perceived; but also that experience (in the form of frequency of occurrence) modulates the mismatch of phonetic and lexical processing. We conclude that a correct account of word recognition in L2 (as indeed in L1) requires consideration of both abstract and episodic information.

Abstract

The difficulty of listening to speech in noise is exacerbated when the speech is in the listener’s L2 rather than L1. In this study, Spanish and Dutch users of English as an L2 identified American English consonants in a constant intervocalic context. Their performance was compared with that of L1 (British English) listeners, under quiet conditions and when the speech was masked by speech from another talker or by noise. Masking affected performance more for the Spanish listeners than for the L1 listeners, but not for the Dutch listeners, whose performance was worse than the L1 case to about the same degree in all conditions. There were, however,large differences in the pattern of results across individual consonants, which were consistent with differences in how consonants are identified in the respective L1s.

Abstract

Listeners rely on native-language rhythm in segmenting speech; in different languages, stress-, syllable- or mora-based rhythm is exploited. The rhythmic similarity hypothesis holds that where two languages have similar rhythm, listeners of each language should segment their own and the other language similarly. Such similarity in listening was previously observed only for related languages (English-Dutch; French-Spanish). We now report three experiments in which speakers of Telugu, a Dravidian language unrelated to Japanese but similar to it in crucial aspects of rhythmic structure, heard speech in Japanese and in their own language, and Japanese listeners heard Telugu. For the Telugu listeners, detection of target sequences in Japanese speech was harder when target boundaries mismatched mora boundaries, exactly the pattern that Japanese listeners earlier exhibited with Japanese and other languages. The same results appeared when Japanese listeners heard Telugu speech containing only codas permissible in Japanese. Telugu listeners' results with Telugu speech were mixed, but the overall pattern revealed correspondences between the response patterns of the two listener groups, as predicted by the rhythmic similarity hypothesis. Telugu and Japanese listeners appear to command similar procedures for speech segmentation, further bolstering the proposal that aspects of language phonological structure affect listeners' speech segmentation.

Abstract

Previous studies have shown that segmentation skills are language-specific, making it difficult to segment continuous speech in an unfamiliar language into its component words. Here we present the first study capturing the delay in segmentation and recognition in the foreign listener using ERPs. We compared the ability of Dutch adults and of English adults without knowledge of Dutch (‘foreign listeners’) to segment familiarized words from continuous Dutch speech. We used the known effect of repetition on the event-related potential (ERP) as an index of recognition of words in continuous speech. Our results show that word repetitions in isolation are recognized with equivalent facility by native and foreign listeners, but word repetitions in continuous speech are not. First, words familiarized in isolation are recognized faster by native than by foreign listeners when they are repeated in continuous speech. Second, when words that have previously been heard only in a continuous-speech context re-occur in continuous speech, the repetition is detected by native listeners, but is not detected by foreign listeners. A preceding speech context facilitates word recognition for native listeners, but delays or even inhibits word recognition for foreign listeners. We propose that the apparent difference in segmentation rate between native and foreign listeners is grounded in the difference in language-specific skills available to the listeners.

Abstract

We investigated whether non-native listeners can exploit phonetic detail in recognizing potentially ambiguous utterances, as native listeners can [6, 7, 8, 9, 10]. Due to the phenomenon of intrusive /r/, the English phrase extra ice may sound like extra rice. A production study indicates that the intrusive /r/ can be distinguished from the onset /r/ in rice, as it is phonetically weaker. In two cross-modal identity priming studies, however, we found no conclusive evidence that Dutch learners of English are able to make use of this difference. Instead, auditory primes such as extra rice and extra ice with onset and intrusive /r/s activate both types of targets such as ice and rice. This supports the notion of spurious lexical activation in L2 perception.

Abstract

SPEECH, in any language, is continuous; speakers provide few reliable cues to the boundaries of words, phrases, or other meaningful units. To understand speech, listeners must divide the continuous speech stream into portions that correspond to such units. This segmentation process is so basic to human language comprehension that psycholinguists long assumed that all speakers would do it in the same way. In previous research1,2, however, we reported that segmentation routines can be language-specific: speakers of French process spoken words syllable by syllable, but speakers of English do not. French has relatively clear syllable boundaries and syllable-based timing patterns, whereas English has relatively unclear syllable boundaries and stress-based timing; thus syllabic segmentation would work more efficiently in the comprehension of French than in the comprehension of English. Our present study suggests that at this level of language processing, there are limits to bilingualism: a bilingual speaker has one and only one basic language.

Abstract

A model of speech segmentation in a stress language is proposed, according to which the occurrence of a strong syllable triggers segmentation of the speech signal, whereas occurrence of a weak syllable does not trigger segmentation. We report experiments in which listeners detected words embedded in nonsense bisyllables more slowly when the bisyllable had two strong syllables than when it had a strong and a weak syllable; mint was detected more slowly in mintayve than in mintesh. According to our proposed model, this result is an effect of segmentation: When the second syllable is strong, it is segmented from the first syllable, and successful detection of the embedded word therefore requires assembly of speech material across a segmentation position. Speech recognition models involving phonemic or syllabic recoding, or based on strictly left-to-right processes, do not predict this result. It is argued that segmentation at strong syllables in continuous speech recognition serves the purpose of detecting the most efficient locations at which to initiate lexical access. (C) 1988 by the American Psychological Association

Abstract

Although spontaneous speech occurs more frequently in most listeners’ experience than read speech, laboratory studies of human speech recognition typically use carefully controlled materials read from a script. The phonological and prosodic characteristics of spontaneous and read speech differ considerably, however, which suggests that laboratory results may not generalize to the recognition of spontaneous and read speech materials, and their response time to detect word-initial target phonemes was measured. Response were, overall, equally fast in each speech mode. However analysis of effects previously reported in phoneme detection studies revealed significant differences between speech modes. In read speech but not in spontaneous speech, later targets were detected more rapidly than earlier targets, and targets preceded by long words were detected more rapidly than targets preceded by short words. In contrast, in spontaneous speech but not in read speech, targets were detected more rapidly in accented than unaccented words and in strong than in weak syllables. An explanation for this pattern is offered in terms of characteristic prosodic differences between spontaneous and read speech. The results support claim from previous work that listeners pay great attention to prosodic information in the process of recognizing speech.

Abstract

Previous research comparing detection times for syllables and for phonemes has consistently found that syllables are responded to faster than phonemes. This finding poses theoretical problems for strictly hierarchical models of speech recognition, in which smaller units should be able to be identified faster than larger units. However, inspection of the characteristics of previous experiments’stimuli reveals that subjects have been able to respond to syllables on the basis of only a partial analysis of the stimulus. In the present experiment, five groups of subjects listened to identical stimulus material. Phoneme and syllable monitoring under standard conditions was compared with monitoring under conditions in which near matches of target and stimulus occurred on no-response trials. In the latter case, when subjects were forced to analyze each stimulus fully, phonemes were detected faster than syllables.

Abstract

The order of some conjoined words is rigidly fixed (e.g. dribs and drabs/*drabs and dribs). Both phonetic and semantic factors can play a role in determining the fixed order. An experiment was conducted to test whether listerners’ reaction times for monitoring a predetermined phoneme are influenced by phonetic constraints on ordering. Two such constraints were investigated: monosyllable-bissyllable and high-low vowel sequences. In English, conjoined words occur in such sequences with much greater frequency than their converses, other factors being equal. Reaction times were significantly shorter for phoneme monitoring in monosyllable-bisyllable sequences than in bisyllable- monosyllable sequences. However, reaction times were not significantly different for high-low vs. low-high vowel sequences.