Publications

Abstract

Listeners learn from their past experience of listening to spoken words, and use this learning to maximise the efficiency of future word recognition. This paper summarises evidence that the facilitatory effects of drawing on past experience are mediated by abstraction, enabling learning to be generalised across new words and new listening situations. Phoneme category retuning, which allows adaptation to speaker-specific articulatory characteristics, is generalised on the basis of relatively brief experience to words previously unheard from that speaker. Abstract knowledge of prosodic regularities is applied to recognition even of novel words for which these regularities were violated. Prosodic word-boundary regularities drive segmentation of speech into words independently of the membership of the lexical candidate set resulting from the segmentation operation. Each of these different cases illustrates how abstraction from past listening experience has contributed to the efficiency of lexical recognition.

Abstract

The phoneme detection task is widely used in spoken-word recognition research. Alphabetically literate participants, however, are more used to explicit representations of letters than of phonemes. The present study explored whether phoneme detection is sensitive to how target phonemes are, or may be, orthographically realized. Listeners detected the target sounds [b, m, t, f, s, k] in word-initial position in sequences of isolated English words. Response times were faster to the targets [b, m, t], which have consistent word-initial spelling, than to the targets [f, s, k], which are inconsistently spelled, but only when spelling was rendered salient by the presence in the experiment of many irregularly spelled filler words. Within the inconsistent targets [f, s, k], there was no significant difference between responses to targets in words with more usual (foam, seed, cattle) versus less usual (phone, cede, kettle) spellings. Phoneme detection is thus not necessarily sensitive to orthographic effects; knowledge of spelling stored in the lexical representations of words does not automatically become available as word candidates are activated. However, salient orthographic manipulations in experimental input can induce such sensitivity. We attribute this to listeners' experience of the value of spelling in everyday situations that encourage phonemic decisions (such as learning new names)

Abstract

Adverse listening conditions always make the perception of speech harder, but their deleterious effect is far greater if the speech we are trying to understand is in a non-native language. An imperfect signal can be coped with by recourse to the extensive knowledge one has of a native language, and imperfect knowledge of a non-native language can still support useful communication when speech signals are high-quality. But the combination of imperfect signal and imperfect knowledge leads rapidly to communication breakdown. This phenomenon is undoubtedly well known to every reader of Speech Communication from personal experience. Many readers will also have a professional interest in explaining, or remedying, the problems it produces. The journal’s readership being a decidedly interdisciplinary one, this interest will involve quite varied scientific approaches, including (but not limited to) modelling the interaction of first and second language vocabularies and phonemic repertoires, developing targeted listening training for language learners, and redesigning the acoustics of classrooms and conference halls. In other words, the phenomenon that this special issue deals with is a well-known one, that raises important scientific and practical questions across a range of speech communication disciplines, and Speech Communication is arguably the ideal vehicle for presentation of such a breadth of approaches in a single volume. The call for papers for this issue elicited a large number of submissions from across the full range of the journal’s interdisciplinary scope, requiring the guest editors to apply very strict criteria to the final selection. Perhaps unique in the history of treatments of this topic is the combination represented by the guest editors for this issue: a phonetician whose primary research interest is in second-language speech (MLGL), an engineer whose primary research field is the acoustics of masking in speech processing (MC), and a psychologist whose primary research topic is the recognition of spoken words (AC). In the opening article of the issue, these three authors together review the existing literature on listening to second-language speech under adverse conditions, bringing together these differing perspectives for the first time in a single contribution. The introductory review is followed by 13 new experimental reports of phonetic, acoustic and psychological studies of the topic. The guest editors thank Speech Communication editor Marc Swerts and the journal’s team at Elsevier, as well as all the reviewers who devoted time and expert efforts to perfecting the contributions to this issue.

Abstract

Recognising continuous speech in a second language is often unexpectedly difficult, as the operation of segmenting speech is so attuned to native-language structure. We report the initial steps in development of a novel training method for second-language listening, focusing on speech segmentation and employing a task designed for studying this: word-spotting. Listeners detect real words in sequences consisting of a word plus a minimal context. The present validation study shows that learners from varying non-English backgrounds successfully perform a version of this task in English, and display appropriate sensitivity to structural factors that also affect segmentation by native English listeners.

Abstract

If listening in adverse conditions is hard, then listening in a foreign language is doubly so: non-native listeners have to cope with both imperfect signals and imperfect knowledge. Comparison of native and non-native listener performance in speech-in-noise tasks helps to clarify the role of prior linguistic experience in speech perception, and, more directly, contributes to an understanding of the problems faced by language learners in everyday listening situations. This article reviews experimental studies on non-native listening in adverse conditions, organised around three principal contributory factors: the task facing listeners, the effect of adverse conditions on speech, and the differences among listener populations. Based on a comprehensive tabulation of key studies, we identify robust findings, research trends and gaps in current knowledge.

Abstract

Japanese listeners detected Japanese words embedded at the end of nonsense sequences (e.g., kaba 'hippopotamus' in gyachikaba). When the final portion of the preceding context together with the initial portion of the word (e.g., here, the sequence chika) was compatible with many lexical competitors, recognition of the embedded word was more difficult than when such a sequence was compatible with few competitors. This clear effect of competition, established here for preceding context in Japanese, joins similar demonstrations, in other languages and for following contexts, to underline that the functional architecture of the human spoken-word recognition system is a universal one.

Abstract

Every language manifests casual speech processes, and hence every second language too. This study examined how listeners deal with second-language casual speech processes, as a function of the processes in their native language. We compared a match case, where a second-language process t/-reduction) is also operative in native speech, with a mismatch case, where a second-language process (/r/-insertion) is absent from native speech. In each case native and non-native listeners judged stimuli in which a given phoneme (in sentence context) varied along a continuum from absent to present. Second-language listeners in general mimicked native performance in the match case, but deviated significantly from native performance in the mismatch case. Together these results make it clear that the mapping from first to second language is as important in the interpretation of casual speech processes as in other dimensions of speech perception. Unfamiliar casual speech processes are difficult to adapt to in a second language. Casual speech processes that are already familiar from native speech, however, are easy to adapt to; indeed, our results even suggest that it is possible for subtle difference in their occurrence patterns across the two languages to be detected,and to be accommodated to in second-language listening.

Abstract

Three groups of listeners identified the vowel in CV and VC syllables produced by an American English talker. The listeners were (a) native speakers of American English, (b) native speakers of Australian English (different dialect), and (c) native speakers of Dutch (different language). The syllables were embedded in multispeaker babble at three signal-to-noise ratios (0 dB, 8 dB, and 16 dB). The identification performance of native listeners was significantly better than that of listeners with another language but did not significantly differ from the performance of listeners with another dialect. Dialect differences did however affect the type of perceptual confusions which listeners made; in particular, the Australian listeners’ judgements of vowel tenseness were more variable than the American listeners’ judgements, which may be ascribed to cross-dialectal differences in this vocalic feature. Although listening difficulty can result when speech input mismatches the native dialect in terms of the precise cues for and boundaries of phonetic categories, the difficulty is very much less than that which arises when speech input mismatches the native language in terms of the repertoire of phonemic categories available.

Abstract

Children begin to talk at about age one. The vocabulary they need to do so must be built on perceptual evidence and, indeed, infants begin to recognize spoken words long before they talk. Most of the utterances infants hear, however, are continuous, without pauses between words, so constructing a vocabulary requires them to decompose continuous speech in order to extract the individual words. Here, we present electrophysiological evidence that 10-month-old infants recognize two-syllable words they have previously heard only in isolation when these words are presented anew in continuous speech. Moreover, they only need roughly the first syllable of the word to begin doing this. Thus, prelinguistic infants command a highly efficient procedure for segmentation and recognition of spoken words in the absence of an existing vocabulary, allowing them to tackle effectively the problem of bootstrapping a lexicon out of the highly variable, continuous speech signals in their environment.

Abstract

Positron emission tomography was used to investigate two competing hypotheses about the role of the left inferior frontal gyrus (IFG) in word generation. One proposes a domain-specific organization, with neural activation dependent on the type of information being processed, i.e., surface sound structure or semantic. The other proposes a process-specific organization, with activation dependent on processing demands, such as the amount of selection needed to decide between competing lexical alternatives. In a novel word retrieval task, word reconstruction (WR), subjects generated real words from heard non-words by the substitution of either a vowel or consonant. Both types of lexical retrieval, informed by sound structure alone, produced activation within anterior and posterior left IFG regions. Within these regions there was greater activity for consonant WR, which is more difficult and imposes greater processing demands. These results support a process-specific organization of the anterior left IFG.

Abstract

Three cross-modal priming experiments examined the role of suprasegmental information in the processing of spoken words. All primes consisted of truncated spoken Dutch words. Recognition of visually presented word targets was facilitated by prior auditory presentation of the first two syllables of the same words as primes, but only if they were appropriately stressed (e.g., OKTOBER preceded by okTO-); inappropriate stress, compatible with another word (e.g., OKTOBER preceded by OCto-, the beginning of octopus), produced inhibition. Monosyllabic fragments (e.g., OC-) also produced facilitation when appropriately stressed; if inappropriately stressed, they produced neither facilitation nor inhibition. The bisyllabic fragments that were compatible with only one word produced facilitation to semantically associated words, but inappropriate stress caused no inhibition of associates. The results are explained within a model of spoken-word recognition involving competition between simultaneously activated phonological representations followed by activation of separate conceptual representations for strongly supported lexical candidates; at the level of the phonological representations, activation is modulated by both segmental and suprasegmental information.

Abstract

We report detailed analyses of a very large database on timing of speech perception collected by Smits et al. (Smits, R., Warner, N., McQueen, J.M., Cutler, A., 2003. Unfolding of phonetic information over time: A database of Dutch diphone perception. J. Acoust. Soc. Am. 113, 563–574). Eighteen listeners heard all possible diphones of Dutch, gated in portions of varying size and presented without background noise. The present report analyzes listeners’ responses across gates in terms of phonological features (voicing, place, and manner for consonants; height, backness, and length for vowels). The resulting patterns for feature perception differ from patterns reported when speech is presented in noise. The data are also analyzed for effects of stress and of phonological context (neighboring vowel vs. consonant); effects of these factors are observed to be surprisingly limited. Finally, statistical effects, such as overall phoneme frequency and transitional probabilities, along with response biases, are examined; these too exercise only limited effects on response patterns. The results suggest highly accurate speech perception on the basis of acoustic information alone.

Abstract

Korean has a very complex phonology, with many interacting alternations. In a coronal-/i/ sequence, depending on the type of phonological boundary present, alternations such as palatalization, nasal insertion, nasal assimilation, coda neutralization, and intervocalic voicing can apply. This paper investigates how the phonological patterns of Korean affect processing of morphemes and words. Past research on languages such as English, German, Dutch, and Finnish has shown that listeners exploit syllable structure constraints in processing speech and segmenting it into words. The current study shows that in parsing speech, listeners also use much more complex patterns that relate the surface phonological string to various boundaries.

Abstract

Four experiments examined Dutch listeners’ use of suprasegmental information in spoken-word recognition. Isolated syllables excised from minimal stress pairs such as VOORnaam/voorNAAM could be reliably assigned to their source words. In lexical decision, no priming was observed from one member of minimal stress pairs to the other, suggesting that the pairs’ segmental ambiguity was removed by suprasegmental information.Words embedded in nonsense strings were harder to detect if the nonsense string itself formed the beginning of a competing word, but a suprasegmental mismatch to the competing word significantly reduced this inhibition. The same nonsense strings facilitated recognition of the longer words of which they constituted the beginning, butagain the facilitation was significantly reduced by suprasegmental mismatch. Together these results indicate that Dutch listeners effectively exploit suprasegmental cues in recognizing spoken words. Nonetheless, suprasegmental mismatch appears to be somewhat less effective in constraining activation than segmental mismatch.

Abstract

The results of a study on perceptual learning are reported. Dutch subjects made lexical decisions on a list of words and nonwords. Embedded in the list were either [f]- or [s]-final words in which the final fricative had been replaced by an ambiguous sound, midway between [f] and [s]. One group of listeners heard ambiguous [f]- final Dutch words like [kara?] (based on karaf, carafe) and unambiguous [s]-final words (e.g., karkas, carcase). A second group heard the reverse (e.g., ambiguous [karka?] and unambiguous karaf). After this training phase, listeners labelled ambiguous fricatives on an [f]- [s] continuum. The subjects who had heard [?] in [f]- final words categorised these fricatives as [f] reliably more often than those who had heard [?] in [s]-final words. These results suggest that speech recognition is dynamic: the system adjusts to the constraints of each particular listening situation. The lexicon can provide this adjustment process with a training signal.

Abstract

We introduce the papers in this special issue by summarising the current major issues in spoken word recognition. We argue that a full understanding of the process of lexical access during speech comprehension will depend on resolving several key representational issues: what is the form of the representations used for lexical access; how is phonological information coded in the mental lexicon; and how is the morphological and semantic information about each word stored? We then discuss a number of distinct access processes: competition between lexical hypotheses; the computation of goodness-of-fit between the signal and stored lexical knowledge; segmentation of continuous speech; whether the lexicon influences prelexical processing through feedback; and the relationship of form-based processing to the processes responsible for deriving an interpretation of a complete utterance. We conclude that further progress may well be made by swapping ideas among the different sub-domains of the discipline.

Abstract

In two word-spotting experiments, Japanese listeners detected Japanese words faster in vowel contexts (e.g., agura, to sit cross-legged, in oagura) than in consonant contexts (e.g., tagura). In the same experiments, however, listeners spotted words in vowel contexts (e.g., saru, monkey, in sarua) no faster than in moraic nasal contexts (e.g., saruN). In a third word-spotting experiment, words like uni, sea urchin, followed contexts consisting of a consonant-consonant-vowel mora (e.g., gya) plus either a moraic nasal (gyaNuni), a vowel (gyaouni) or a consonant (gyabuni). Listeners spotted words as easily in the first as in the second context (where in each case the target words were aligned with mora boundaries), but found it almost impossible to spot words in the third (where there was a single consonant, such as the [b] in gyabuni, between the beginning of the word and the nearest preceding mora boundary). Three control experiments confirmed that these effects reflected the relative ease of segmentation of the words from their contexts.We argue that the listeners showed sensitivity to the viability of sound sequences as possible Japanese words in the way that they parsed the speech into words. Since single consonants are not possible Japanese words, the listeners avoided lexical parses including single consonants and thus had difficulty recognizing words in the consonant contexts. Even though moraic nasals are also impossible words, they were not difficult segmentation contexts because, as with the vowel contexts, the mora boundaries between the contexts and the target words signaled likely word boundaries. Moraic rhythm appears to provide Japanese listeners with important segmentation cues.

Abstract

The central issues in the study of speech recognition by human listeners (HSR) and of automatic speech recognition (ASR) are clearly comparable; nevertheless the research communities that concern themselves with ASR and HSR are largely distinct. This paper compares the research objectives of the two fields, and attempts to draw informative lessons from one to the other.

Abstract

Two word-spotting experiments are reported that examine whether the Possible-Word Constraint (PWC) is a language-specific or language-universal strategy for the segmentation of continuous speech. The PWC disfavours parses which leave an impossible residue between the end of a candidate word and any likely location of a word boundary, as cued in the speech signal. The experiments examined cases where the residue was either a CVC syllable with a schwa, or a CV syllable with a lax vowel. Although neither of these syllable contexts is a possible lexical word in English, word-spotting in both contexts was easier than in a context consisting of a single consonant. Two control lexical-decision experiments showed that the word-spotting results reflected the relative segmentation difficulty of the words in different contexts. The PWC appears to be language-universal rather than language-specific.

Abstract

Current models of spoken-word recognition assume automatic activation of multiple candidate words fully or partially compatible with the speech input. We propose that listeners make use of this concurrent activation in word play such as punning. Distortion in punning should ideally involve no more than a minimal contrastive deviation between two words, namely a phoneme. Moreover, we propose that this metric of similarity does not presuppose phonemic awareness on the part of the punster. We support these claims with an analysis of modern and traditional puns in Japanese (in which phonemic awareness in language users is not encouraged by alphabetic orthography). For both data sets, the results support the predictions. Punning draws on basic processes of spokenword recognition, common across languages.

Abstract

Four cross-modal priming experiments in Spanish addressed the role of suprasegmental and segmental information in the activation of spoken words. Listeners heard neutral sentences ending with word fragments (e.g., princi-) and made lexical decisions on letter strings presented at fragment offset. Responses were compared for fragment primes that fully matched the spoken form of the initial portion of target words, versus primes that mismatched in a single element (stress pattern; one vowel; one consonant), versus control primes. Fully matching primes always facilitated lexical decision responses, in comparison to the control condition, while mismatching primes always produced inhibition. The respective strength of the contribution of stress, vowel, and consonant (one feature mismatch or more) information did not differ statistically. The results support a model of spoken-word recognition involving automatic activation of word forms and competition between activated words, in which the activation process is sensitive to all acoustic information relevant to the language’s phonology.

Abstract

In this paper, we use articulatory measures to determine whether Dutch schwa epenthesis is an abstract phonological process or a concrete phonetic process depending on articulatory timing. We examine tongue position during /l/ before underlying schwa and epenthetic schwa and in coda position. We find greater tip raising before both types of schwa, indicating light /l/ before schwa and dark /l/ in coda position. We argue that the ability of epenthetic schwa to condition the /l/ alternation shows that Dutch schwa epenthesis is an abstract phonological process involving insertion of some unit, and cannot be accounted for within Articulatory Phonology.

Abstract

Articulatory data are used to address the question of whether Dutch schwa insertion is a phonological or a phonetic process. By investigating tongue tip raising and dorsal lowering, we show that /l/ when it appears before inserted schwa is a light /l/, just as /l/ before an underlying schwa is, and unlike the dark /l/ before a consonant in non-insertion productions of the same words. The fact that inserted schwa can condition the light/dark /l/ alternation shows that schwa insertion involves the phonological insertion of a segment rather than phonetic adjustments to articulations.

Abstract

Listeners process spoken language in ways which are adapted to the phonological structure of their native language. As a consequence, non-native speakers do not listen to a language in the same way as native speakers; moreover, listeners may use their native language listening procedures inappropriately with foreign input. With sufficient experience, however, it may be possible to inhibit this latter (counter-productive) behavior.

Abstract

The recognition of speech involves the segmentation of continuous utterances into their component words. Cross-linguistic evidence is briefly reviewed which suggests that although there are language-specific solutions to this segmentation problem, they have one thing in common: they are all based on language rhythm. In English, segmentation is stress-based: strong syllables are postulated to be the onsets of words. Segmentation, however, can also be achieved by a process of competition between activated lexical hypotheses, as in the Shortlist model. A series of experiments is summarised showing that segmentation of continuous speech depends on both lexical competition and a metrically-guided procedure. In the final section, the implementation of metrical segmentation in the Shortlist model is described: the activation of lexical hypotheses matching strong syllables in the input is boosted and that of hypotheses mismatching strong syllables in the input is penalised.

Abstract

Japanese listeners detect speech sound targets which correspond precisely to a mora (a phonological unit which is the unit of rhythm in Japanese) more easily than targets which do not. English listeners detect medial vowel targets more slowly than consonants. Six phoneme detection experiments investigated these effects in both subject populations, presented with native- and foreign-language input. Japanese listeners produced faster and more accurate responses to moraic than to nonmoraic targets both in Japanese and, where possible, in English; English listeners responded differently. The detection disadvantage for medial vowels appeared with English listeners both in English and in Japanese; again, Japanese listeners responded differently. Some processing operations which listeners apply to speech input are language-specific; these language-specific procedures, appropriate for listening to input in the native language, may be applied to foreign-language input irrespective of whether they remain appropriate.

Abstract

Word blends combine fragments from two words, either in speech errors or when a new word is created. Previous work has demonstrated that in Japanese, such blends preserve moraic structure; in English they do not. A similar effect of moraic structure is observed in perceptual research on segmentation of continuous speech in Japanese; English listeners, by contrast, exploit stress units in segmentation, suggesting that a general rhythmic constraint may underlie both findings. The present study examined whether mis parallel would also hold for word blends. In spontaneous English polysyllabic blends, the source words were significantly more likely to be split before a strong than before a weak (unstressed) syllable, i.e. to be split at a stress unit boundary. In an experiment in which listeners were asked to identify the source words of blends, significantly more correct detections resulted when splits had been made before strong syllables. Word blending, like speech segmentation, appears to be constrained by language rhythm.

Abstract

In a 50,000-word corpus of spoken British English the occurrence of words embedded within other words is reported. Within-word embedding in this real speech sample is common, and analogous to the extent of embedding observed in the vocabulary. Imposition of a syllable boundary matching constraint reduces but by no means eliminates spurious embedding. Embedded words are most likely to overlap with the beginning of matrix words, and thus may pose serious problems for speech recognisers.

Abstract

Although word boundaries are rarely clearly marked, listeners can rapidly recognize the individual words of spoken sentences. Some theories explain this in terms of competition between multiply activated lexical hypotheses; others invoke sensitivity to prosodic structure. We describe a connectionist model, SHORTLIST, in which recognition by activation and competition is successful with a realistically sized lexicon. Three experiments are then reported in which listeners detected real words embedded in nonsense strings, some of which were themselves the onsets of longer words. Effects both of competition between words and of prosodic structure were observed, suggesting that activation and competition alone are not sufficient to explain word recognition in continuous speech. However, the results can be accounted for by a version of SHORTLIST that is sensitive to prosodic structure.

Abstract

This paper describes recent experimental evidence which shows that models of spoken word recognition must incorporate both inhibition between competing lexical candidates and a sensitivity to metrical cues to lexical segmentation. A new version of the Shortlist [1][2] model incorporating the Metrical Segmentation Strategy [3] provides a detailed simulation of the data.

Abstract

Infants acquire whatever language is spoken in the environment into which they are born. The mental capability of the newborn child is not biased in any way towards the acquisition of one human language rather than another. Because psychologists who attempt to model the process of language comprehension are interested in the structure of the human mind, rather than in the properties of individual languages, strategies which they incorporate in their models are presumed to be universal, not language-specific. In other words, strategies of comprehension are presumed to be characteristic of the human language processing system, rather than, say, the French, English, or Igbo language processing systems. We report here, however, on a comprehension strategy which appears to be used by native speakers of French but not by native speakers of English.

Abstract

Spontaneous self-corrections in speech pose a communication problem; the speaker must make clear to the listener not only that the original Utterance was faulty, but where it was faulty and how the fault is to be corrected. Prosodic marking of corrections - making the prosody of the repair noticeably different from that of the original utterance - offers a resource which the speaker can exploit to provide the listener with such information. A corpus of more than 400 spontaneous speech repairs was analysed, and the prosodic characteristics compared with the syntactic and semantic characteristics of each repair. Prosodic marking showed no relationship at all with the syntactic characteristics of repairs. Instead, marking was associated with certain semantic factors: repairs were marked when the original utterance had been actually erroneous, rather than simply less appropriate than the repair; and repairs tended to be marked more often when the set of items encompassing the error and the repair was small rather than when it was large. These findings lend further weight to the characterization of accent as essentially semantic in function.