Elsevier

Journal of Phonetics

Volume 38, Issue 4, October 2010, Pages 493-503
Journal of Phonetics

Development of voicing perception in French: Comparing adults, adolescents, and children

https://doi.org/10.1016/j.wocn.2010.06.002Get rights and content

Abstract

Previous studies suggest that the development of the perception of speech features is based on the adaptation of universal predispositions to the particular categories present in the native language. Previous studies point to a long-lasting development of the location and the precision of perceptual boundaries. However, there is no clear evidence about changes in “categorical perception”, i.e. in the degree of equivalence of identification and discrimination boundaries. The objective of the present study was to investigate the development of boundary location and precision between late childhood and adulthood, and to gather evidence for possible changes in categorical perception during this period. Voicing perception in French stops was investigated with both identification and discrimination data in children (9-year-olds), adolescents (17-year-olds), and adults. The results confirmed the effect of age on boundary precision, but did not show an effect of age on either boundary location or on categorical perception. The results also suggest that the development of categorization performance starts around the boundary and is followed by decreased sensitivity near the prototypes.

Introduction

A basic problem in the study of speech development is to understand how universal properties contribute to the perception of language-specific categories. Studies with children below one year of age point to their ability to perceive the multiple phonetic contrasts present in the world's languages (for a review: Vihman, 1996). This universal potential consists not only sound categories but also auditory boundaries between different sound categories, i.e. “basic cuts” in the acoustic space (Kuhl, 20041). These boundaries are conceptually close to the “features” of the linguistic systems, especially when the latter are defined in acoustic terms (Jakobson, Fant, & Halle, 1952). Features are units of difference: not categories but oppositions between categories (Jakobson, 1973, p. 130). Boundaries correspond to the limits between categories, have measurable properties, and can be used as the empirical correlates of features.

The adaptation of these universal boundaries to the phonological categories of the linguistic environment (i.e. to the phonetic categories specific to a given language) is a complex process. The boundaries are not only simply selected but also combined in specific ways in order to cope with the multiple possible phonological contrasts in the world's languages (Hoonhorst et al., 2009a). The development of language-specific boundaries is also a long-lasting process which starts fairly early, before one year of age, and continues later all through childhood (Burnham, Tyler, & Horlyck, 2002). Both the location and the precision of the perceptual boundaries change during these periods, revealing a progressive attunement of the universal settings to the contrasts prevailing in the ambient language (Aslin & Pisoni, 1980).

One pending question about the development of speech perception is whether it also entails changes in “Categorical Perception” (CP). CP depends on the relationship between discrimination and identification: the stronger the relationship, the higher the degree of CP (Damper & Harnad, 2000; Harnad, 1987; Liberman et al., 1957). CP is perfect when observed discrimination scores coincide with those expected from identification, and the degree of CP is assessed by the difference between the observed and expected discrimination scores. Observed discrimination scores correspond to actual discrimination responses (e.g. the correct discrimination scores or d′-transforms of stimulus pairs in an AX discrimination task) whereas the expected discrimination scores are derived from identification responses (e.g. the identification scores of the same stimuli presented individually).

CP does not make sense without some amount of boundary precision (Schouten, Gerrits, & Van Hessen, 2003). However, with the exception of extreme conditions where boundary precision is so weak that there is almost no discernible boundary, CP and boundary are two distinct categorical properties. This is illustrated by the hypothetical discrimination curves in Fig. 1. Boundary precision is assessed with the slope of the identification function—the steeper the slope, the higher the precision (Simon & Fourcin, 1978; compare Fig. 1a with Fig. 1d)—or alternatively with the magnitude of the discrimination peaks—the larger the peak, the higher the precision (compare Fig. 1b with Fig. 1c; Fig. 1c with Fig. 1d). CP depends on the match between observed discrimination peaks and those expected from identification (CP differences related to changes in peak location are illustrated in Fig. 1; compare Fig. 1b with Fig. 1c; Fig. 1e with Fig. 1f).

Previous studies have shown that both the precision and location of the boundaries change during childhood. The precision of the boundary increases with age (e.g. Krause, 1982; Simon & Fourcin, 1978; Zlatin & Koenigsknecht, 1975) and its location gets progressively closer to the adult value (Simon & Fourcin, 1978). While both the precision and the location of the boundaries improve with age, there is no available evidence about improvements in categorical perception. Most previous studies did not collect the relevant data about a possible effect of age on CP, and the only study which did so, to our knowledge, suggested a decrease of CP as a function of age (Elliott, Longinotti, Meyer, Raz, & Zucker, 1981). The principal aim of the present study was to further examine the effect of age on CP by collecting data on the development of voicing perception in French.

The voicing distinction between consonants is frequently used in developmental studies. This distinction is defined as the presence vs. absence of laryngeal vibrations during consonant production (Jakobson et al., 1952). Voicing perception is based on Voice Onset Time (VOT: for initial stops, the time interval between voicing onset and the release of the supra-glottal closure: Lisker & Abramson (1964), Lisker & Abramson (1971)) and other concomitant cues (burst loudness: Repp, 1979; F1 characteristics: Summerfield, 1982; F0 characteristics: Haggard, Summerfield, & Roberts, 1981). VOT is the most reliable voicing cue for stops in initial position, and plays a dominant role in the perception of the feature: the other cues only affect voicing identification when VOT is ambiguous (in English: Lisker, Liberman, Erickson, & Dechovitz, 1978; Abramson, & Lisker, 1985; in French: Serniclaes, 1987). Cross-linguistic differences in voicing perception are captured by changes in the locations of the VOT boundaries, i.e. by the locations of the identification boundaries between voicing categories on a VOT continuum. The VOT boundaries vary within each language as a function of the phonetic context and other stimulus factors such as speaking rate (Kessinger & Blumstein, 1998). Consequently, for the sake of simplicity, the considerations below about perceptual processes are based on mean values across different phonetic contexts, although further degrees of freedom are required for taking into account contextual variations. In languages with three voicing categories (voiced, voiceless, and voiceless aspirated), the mean VOT boundaries are located around −30 and +30 ms VOT (in Thai: Abramson & Lisker, 1970; Donald, 1978). Negative vs. positive VOT values indicate whether voice onset occurs before or after the release burst, respectively, the −30 and +30 ms VOT values mark the regions of anticipation (up to −30 ms), coincidence (between −30 and +30 ms) and delay (+30 ms and more) in voice onset relative to release. Two-category languages differ in the nature of their voicing categories. In languages with a voiceless/voiceless aspirated contrast, the VOT boundary is located around +30 ms and corresponds to the distinction between coincidence and delay in voice onset relative to release (in English: Lisker et al., 1978). Other two-category languages exhibit a voiced/voiceless contrast and their mean VOT boundary is located around 0 ms VOT, which corresponds to the distinction between anticipation and delay in voice onset relative to release (e.g. in Spanish: Williams, 1977; in French: Caramazza & Yeni-Komshian, 1974; Serniclaes, 1987).

Studies on VOT discrimination in children below one year of age have found enhanced sensitivities to both the negative and positive VOT boundaries, even when these boundaries are not phonemic in the environmental language. Infants below six months of age raised in an English environment are sensitive to both VOT boundaries, although only the positive VOT boundary is phonological in this language (Aslin, Pisoni, Hennessy, & Perey, 1981). Similarly, infants below six months of age raised in a Spanish or in a French environment also react to both the negative and positive VOT boundaries, even though the phonological boundary is located at 0 ms in these languages (in Spanish: Lasky, Syrdal-Lasky, & Klein, 1975; in French: Hoonhorst et al., 2009a). This suggests that the negative and positive VOT boundaries are either innate or arise very early in perceptual development, and that language-specific boundaries are acquired later through exposure to the environment. Children raised in a French-speaking environment react more strongly to the −30 and +30 ms VOT boundaries around four months of age, but they react more to the French/Spanish 0 ms boundary around eight months of age (Hoonhorst et al., 2009a). However, English-speaking infants also react more to the 0 ms VOT boundary around six to eight months of age and they react (again) more to the English VOT boundary (positive VOT) only at around 10–12 months of age (Burns, Yoshida, Hill, & Werker, 2007; Rivera-Gaxiola, Silva-Peyrera, & Kuhl, 2005; in production: Whalen, Levitt, & Goldstein, 20072). In summary, data collected from children raised in different languages suggest that there is a universal sensitivity to negative and positive boundaries before six months of age, and that language-specific boundaries are obtained either by combination between universal boundaries (French and Spanish) or by selection between these boundaries (English). Combination between universal boundaries means that new boundaries emerge at other locations on the VOT continuum. Selection between universal boundaries does not mean that the unselected boundaries are lost, but instead that the selected boundaries are perceived with higher precision.

The early development of VOT boundaries is fairly complex, and we can ask whether the changes in VOT categorization are completed around one year of age or whether they continue to develop later. Previous studies make it clear that the discrimination accuracy of contrasts straddling phonological boundaries continues to increase after one year of age (Elliott, Busse, Partridge, Rupert, & de Graaf, 1986; Elliott et al., 1981). Furthermore, the slope of the identification function becomes steeper between two years of age and adulthood (Burnham, Earnshaw, & Clark, 1991; Hazan & Barrett, 2000; Krause, 1982; Simon & Fourcin, 1978; Zlatin & Koenigsknecht, 1975). As both the slope of the identification function and the magnitude of the boundary discrimination peak are categorical precision parameters, the results of these studies indicate that categorical precision continues to increase up to the end of childhood. The location of the VOT boundary also continues to change after one year of age (Simon & Fourcin, 1978). However, none of these studies showed an improvement of categorical perception, i.e. of the relationship between identification and discrimination. Rather, they only showed that the precision of the VOT boundaries continues to increase well after one year of age.

In summary, previous studies show that the location and the precision of the perceptual boundaries change before one year of age and that there is a long-lasting development in boundary precision, which only ends during adolescence. However, there is no clear evidence for changes in either the locations of the boundaries or in their categorical perception (i.e. in the match between identification and discrimination boundaries) after one year of age.

The aim of the present study was to examine the development of voicing perception in French by comparing the identification and discrimination performance of 9-year-old children, 17-year-old adolescents, and adults. This study wanted to confirm the late development of boundary precision at the end of infancy, already found in English for ages ranging between 2 and 14 years (Burnham et al., 1991; Elliott, Longinotti, Meyer, Raz, & Zucker (1981), Elliott, Busse, Partridge, Rupert, & de Graaf (1986); Hazan & Barrett, 2000; Krause, 1982; Simon & Fourcin, 1978; Zlatin & Koenigsknecht, 1975). We also wanted to gather evidence about possible changes in boundary location and categorical perception during this period. We investigated the period between the age of nine and adulthood in order to avoid low categorical precision in the children's responses, which would hinder the assessment of categorical perception, and also to be able to offer evidence about possible changes in categorical perception, as previous studies in English indicate that there still are changes in categorical precision between the age of nine and adulthood.

Boundary location was measured with both identification and discrimination data (see above: location of the 50% identification score and of the discrimination peak). Boundary precision was also measured with both identification and discrimination data (see above: slope of the identification curve and magnitude of the discrimination peak). The assessment of categorical perception of the boundary was based on the match between the location and precision of the identification and discrimination boundaries. We used a /də/-/tə/ VOT continuum, as in this contrast both consonant and vowel articulation are fairly neutral and the VOT boundary was expected to be close to 0 ms, which corresponds to its mean value in French (Serniclaes, 1987).

Section snippets

Participants

Three groups of native French speakers participated in this study. One group included 17 normal-hearing children (11 boys and six girls) aged 8–11 years old (average age: 9.0 years, SD=0.7) who were attending primary schools (classes from the second to the fourth grade). The second group included 17 normal-hearing adults who were the mothers of these children. The mothers group was aged from 35 to 50 years (average age: 43.5 years, SD=5.5). The third group included 35 adolescents3

Identification

The mean identification curves of the adults, adolescents, and children are fairly similar (Fig. 3). The perceptual boundary is located around 0 ms VOT for each group and the adult and adolescent functions overlap almost perfectly. Differences in mean boundaries and slopes between the groups were tested using single-factor ANOVAs. The mean boundary was not significantly related to the groups (F<1). The mean boundary for the three groups taken together was located at +7.3 ms VOT (SD=6.3) and was

Discussion

The present results show that there is both constancy and variation in the perception of voicing in French between nine years and adulthood. The VOT boundary is quite stable during this period and is located at 7.3 ms, slightly above the expected 0 ms value. The degree of CP is limited, with discrepancies between the observed and expected discrimination scores in the 0 to +20 ms VOT region. However, these discrepancies are present for the different groups, and the degree of CP is fairly constant

Conclusion

In conclusion, there is a late effect of age on boundary precision, after nine years of age, whereas categorical perception does not change any further during this period. The late development of categorization performance is more a matter of prototype consolidation than of boundary attunement, suggesting that the boundaries between categories develop faster than category prototypes. Further investigations should make it possible to test the generality of the present findings for different

Acknowledgments

We would like to thank Marie-Pierre Geron for her help in data collection, Liliane Sprenger-Charolles for her comments on earlier versions of the manuscript, Cécile Houard and Paul Reeves for their help in the preparation of the English manuscript.

References (72)

  • W. Serniclaes et al.

    Allophonic mode of speech perception in dyslexia

    Journal of Experimental Child Psychology

    (2004)
  • D.H. Whalen et al.

    VOT in the babbling of French- and English-learning infants

    Journal of Phonetics

    (2007)
  • L. Williams

    The voicing contrast in Spanish

    Journal of Phonetics

    (1977)
  • C.G. Wolf

    The perception of stop consonants by children

    Journal of Experimental Child Psychology

    (1973)
  • A.S. Abramson et al.

    Discriminability along the voice onset time continuum: Cross-language tests

  • A.S. Abramson et al.

    Relative power of cues: F0 shift vs. voice timing

  • R.N. Aslin et al.

    Some developmental processes in speech perception

  • R.N. Aslin et al.

    Discrimination of voice onset time by human infants: New findings and implications for the effect of early experience

    Child Development

    (1981)
  • J. Bertoncini et al.

    Speech discrimination based upon temporal envelope versus fine structure cues in 5-to-7 year-old children

    Journal of Speech, Language, and Hearing Research

    (2008)
  • Bogliotti, C. (2005). Categorical perception and allophonic perception: Effect of age, reading level and coupling...
  • D.K. Burnham

    Language specific speech perception and the onset of reading

    Reading and Writing: An Interdisciplinary Journal

    (2003)
  • D.K. Burnham et al.

    Development of categorical identification of native and non-native bilabial stops: Infants, children and adults

    Journal of Child Language

    (1991)
  • D. Burnham et al.

    Periods of speech perception development and their vestiges in adulthood

  • T.C. Burns et al.

    The development of phonetic representation in bilingual and monolingual infants

    Applied Psycholinguistics

    (2007)
  • A. Caramazza et al.

    The acquisition of a new phonological contrast: The case of stop consonants in French-English bilinguals

    Journal of the Acoustical Society of America

    (1973)
  • Carré, R. (2004a). Program SyntFormVoy. Laboratoire Dynamique du Language, CNRS, Lyon, France. From René Carré Personal...
  • Carré, R. (2004b). Program Percept A. Laboratoire Dynamique du Language, CNRS, Lyon, France. From René Carré Personal...
  • Carré, R. (2004c). Program Percept AB. Laboratoire Dynamique du Language, CNRS, Lyon, France. From René Carré Personal...
  • R.I. Damper et al.

    Neural network modeling of categorical perception

    Perception and Psychophysics

    (2000)
  • Donald, S. L. (1978). The perception of voicing contrasts in Thai and English. Dissertation, University of Connecticut...
  • O. Dufor et al.

    Left pre-motor cortex and allophonic speech perception in dyslexia: A PET study

    NeuroImage

    (2009)
  • L.L. Elliott et al.

    Adult and child discrimination of CV syllables differing in Voice Onset time

    Child Development

    (1986)
  • L.L. Elliott et al.

    Developmental differences in identifying and discriminating CV syllables

    Journal of the Acoustical Society of America

    (1981)
  • D.J. Finney

    Probit analysis

    (1971)
  • F.H. Guenther et al.

    Effects of categorization and discrimination training on auditory perceptual space

    Journal of the Acoustical Society of America

    (1999)
  • F.H. Guenther et al.

    Representation of sound categories in auditory cortical maps

    Journal of Speech, Language, and Hearing Research

    (2004)
  • Cited by (37)

    • Multi-modal imaging of the auditory-larynx motor network for voicing perception

      2022, NeuroImage
      Citation Excerpt :

      For over 50 years, voicing has been a significant subject of interest in both speech production and perception studies because it is one of the most important features of speech sounds (Abramson and Whalen, 2017). Specifically, developmental changes in voicing production and perception and their language differences have often been examined to reveal the underlying mechanism of language acquisition and development (Whiteside et al., 2003; Lisker and Abramson, 1970; Aslin et al., 1981; Medina et al., 2010). In addition, the brain mechanisms involved in voicing production and perception have been explored in many neurophysiological studies to clarify the temporal representations of speech sounds in auditory and speech motor cortical activities in the order of a few tens of milliseconds (Lotte et al., 2015; Sharma and Dorman, 1999; Mesgarani et al., 2014).

    • The impact of alphabetic literacy on the perception of speech sounds

      2021, Cognition
      Citation Excerpt :

      Thus, both studies indicate that categorical precision and categorical perception are dissociable, with only the former being modulated by instruction. The same held true for the effect of age: whilst categorical perception did not change across age, categorical precision improved between kindergarten and adulthood (Hoonhorst et al., 2011; see also similar results in Medina, Hoonhorst, Bogliotti, & Serniclaes, 2010; Medina & Serniclaes, 2009). Nonetheless, the study on children was unable to dissociate the effect of literacy from the effect of age (Hoonhorst et al., 2011).

    • On the nature of the speech perception deficits in children with autism spectrum disorders

      2017, Research in Developmental Disabilities
      Citation Excerpt :

      Or do they also present an allophonic mode of speech perception, characterised by an enhanced discrimination of acoustic differences within phoneme categories? These questions are important because a CPR deficit is fairly common in a range of developmental disorders and is also found in normal-development during late childhood and even adolescence (Hoonhorst et al., 2011; Medina, Hoonhorst, Bogliotti, & Serniclaes, 2010). A persistent CPR deficit does thus probably not have serious consequences on phonological development.

    View all citing articles on Scopus
    View full text