Constraints on the perception of synthetic speech generated by rule

Within the next few years, there will be an extensive proliferation of various types of voice response devices in human-machine communication systems. Unfortunately, at present, relatively little basic or applied research has been carried out on the intelligibility, comprehension, and perceptual processing of synthetic speech produced by these devices. On the basis of our research, we identify five factors that must be considered in studying the perception of synthetic speech: (1) the specific demands imposed by a particular task, (2) the inherent limitations of the human information processing system, (3) the experience and training of the human listener, (4) the linguistic structure of the message set, and (5) the structure and quality of the speech signal.     

Perception of synthetic speech produced automatically by rule: Intelligibility of eight text-to-speech systems

We present the results of studies designed to measure the segmental intelligibility of eight text-to-speech systems and a natural speech control, using the Modified Rhyme Test (MRT). Results indicated that the voices tested could be grouped into four categories: natural speech, high-quality synthetic speech, moderate-quality synthetic speech, and low-quality synthetic speech. The overall performance of the best synthesis system, DECtalk-Paul, was equivalent to natural speech only in terms of performance on initial consonants. The findings are discussed in terms of recent work investigating the perception of synthetic speech under more severe conditions. Suggestions for future research on improving the quality of synthetic speech are also considered.     

Effects of phonological similarity on priming in auditory lexical decision

Two auditory lexical decision experiments were conducted to determine whether facilitation can be obtained when a prime and a target share word-initial phonological information. Subjects responded “word” or “nonword” to monosyllabic words and nonwords controlled for frequency. Each target was preceded by the presentation of either a word or nonword prime that was identical to the target or shared three, two, or one phonemes from the beginning. The results showed that lexical decision times decreased when the prime and target were identical for both word and nonword targets. However, no facilitation was observed when the prime and target shared three, two, or one initial phonemes, These results were found when the interstimulus interval between the prime and target was 500 msec or 50 msec. In a second experiment, no differences were found between primes and targets that shared three, one, or zero phonemes, although facilitation was observed for identical prime-target pairs. The results are compared to recent findings obtained using a perceptual identification paradigm. Taken together, the findings suggest several important differences in the way lexical decision and perceptual identification tasks tap into the information-processing system during auditory word recognition.     

Adaptation of the relative onset time of two-component tones

The results of three selective adaptation experiments employing nonspeech signals that differed in temporal onset are reported. In one experiment, adaptation effects were observed when both the adapting and test stimuli were selected from the same nonspeech test continuum. This result was interpreted as evidence for selective processing of temporal order information in nonspeech signals. Two additional experiments tested for the presence of cross-series adaptation effects from speech to nonspeech and then from nonspeech to speech. Both experiments failed to show any evidence of cross-series adaptation effects, implying a possible dissociation between perceptual classes of speech and nonspeech signals in processing temporal order information. Despite the absence of cross-series effects, it is argued that the ability of the auditory system to process temporal order information may still provide a possible basis for explaining the perception of voicing in stops that differ in VOT. The results of the present experiments, taken together with earlier findings on the perception of temporal onset in nonspeech signals, were viewed as an example of the way spoken language has exploited the basic sensory capabilities of the auditory system to signal phonetic differences.     

Training Japanese listeners to identify English /r/ and /l/: long-term retention of learning in perception and production.

Previous work from our laboratories has shown that monolingual Japanese adults who were given intensive high-variability perceptual training improved in both perception and production of English /r/-/l/ minimal pairs. In this study, we extended those findings by investigating the long-term retention of learning in both perception and production of this difficult non-native contrast. Results showed that 3 months after completion of the perceptual training procedure, the Japanese trainees maintained their improved levels of performance of the perceptual identification task. Furthermore, perceptual evaluations by native American English listeners of the Japanese trainees' pretest, posttest, and 3-month follow-up speech productions showed that the trainees retained their long-term improvements in the general quality, identifiability, and overall intelligibility of their English/r/-/l/ word productions. Taken together, the results provide further support for the efficacy of high-variability laboratory speech sound training procedures, and suggest an optimistic outlook for the application of such procedures for a wide range of "special populations."     

Phonotactics, neighborhood activation, and lexical access for spoken words.

Probabilistic phonotactics refers to the relative frequencies of segments and sequences of segments in spoken words. Neighborhood density refers to the number of words that are phonologically similar to a given word. Despite a positive correlation between phonotactic probability and neighborhood density, nonsense words with high probability segments and sequences are responded to more quickly than nonsense words with low probability segments and sequences, whereas real words occurring in dense similarity neighborhoods are responded to more slowly than real words occurring in sparse similarity neighborhoods. This contradiction may be resolved by hypothesizing that effects of probabilistic phonotactics have a sublexical focus and that effects of similarity neighborhood density have a lexical focus. The implications of this hypothesis for models of spoken word recognition are discussed.     

Simple and contingent adaptation effects for place of articulation in stop consonants

Experiments on selective adaptation have shown that the locus of the phonetic category boundary between two segments shifts after repetitive listening to an adapting stimulus. Theoretical interpretations of these results have proposed that adaptation occurs either entirely at an auditory level of processing or at both auditory and more abstract phonetic levels. The present experiment employed two alternating stimuli as adaptors in an attempt to distinguish between these two possible explanations. Two alternating stimuli were used as adaptors in order to test for the presence of contingent effects and to compare these results to simple adaptation using only a single adaptor. Two synthetic CV series with different vowels that varied the place of articulation of the consonant were employed. When two alternating adaptors were used, contingent adaptation effects were observed for the two stimulus series. The direction of the shifts in each series was governed by the vowel context of the adapting syllables. Using the single adaptor data, a comparison was made between the additive effects of the single adaptors and their combined effects when presented in alternating pairs. With voiced adaptors, only within-series adaptation effects were found, and these data were consistent with a on,level model of selective adaptation. However, for the voiceless adaptors, both within- and cross-series adaptation effects were found, suggesting the possible presence of two levels of adaptation to place of articulation. Further, the contingent adaptation effects with the voiceless adaptors seemed to be the result of the additive effects of the two alternating adaptors. This result indicates that previously reported contingent adaptation results may also reflect the net vowel specific adaptation effects after cancellation of other, nonvowel dependent effects and that caution is needed in interpreting such results.     

Implicit statistical learning in language processing: Word predictability is the key

Fundamental learning abilities related to the implicit encoding of sequential structure have been postulated to underlie language acquisition and processing. However, there is very little direct evidence to date supporting such a link between implicit statistical learning and language. In three experiments using novel methods of assessing implicit learning and language abilities, we show that sensitivity to sequential structure – as measured by improvements to immediate memory span for structurally-consistent input sequences – is significantly correlated with the ability to use knowledge of word predictability to aid speech perception under degraded listening conditions. Importantly, the association remained even after controlling for participant performance on other cognitive tasks, including short-term and working memory, intelligence, attention and inhibition, and vocabulary knowledge. Thus, the evidence suggests that implicit learning abilities are essential for acquiring long-term knowledge of the sequential structure of language – i.e., knowledge of word predictability – and that individual differences on such abilities impact speech perception in everyday situations. These findings provide a new theoretical rationale linking basic learning phenomena to specific aspects of spoken language processing in adults, and may furthermore indicate new fruitful directions for investigating both typical and atypical language development.