The aim of the present paper was to compare the development of perceptual categorization of voicing, colors and facial expressions in French-speaking children (aged from six to eight years) and adults. Differences in both Categorical Perception, i.e. the correspondence between identification and discrimination performances, and in Boundary Precision, indexed by the steepness of the identification slope, were investigated. Whereas there was no significant effect of age on Categorical Perception, Boundary Precision increased with age, both for voicing and facial expressions though not for colors. Further, the precision of the voicing boundary was correlated with reading abilities. These results are discussed in the light of two developmental theories: the reading hypothesis and the general cognitive hypothesis.
1 Hoonhorst I, Medina V, Colin C, Markessis E, Radeau M, Deltenre P, Serniclaes W. Development of categorical perception: comparisons between voicing, colors and facial expressions (in revision). Journal of Experimental Child Psychology.
Introduction
CATEGORICAL PERCEPTION
Categorical perception is commonly defined by the mismatch between the monotonic variation of a set of stimuli regularly varying along a physical continuum and its resulting non-monotonic perception. Harnad (1987) emphasized two characteristics associated with categorical perception: “1/ a set of stimuli ranging along a physical continuum is given one label on one side of a category boundary and another label on the other side and 2/ the subject can discriminate smaller physical differences between pairs of stimuli that straddle that boundary than between pairs that are entirely within one category or the other”.
This definition was shown to suit both non linguistic and linguistic continua. Cutting and Rosner (1974) evidenced that auditory rise-time variations are perceived categorically either as presented in a linguistic or in a non-linguistic continuum. Bornstein, Kessen and Weiskopf (1976) evidenced the same categorical mode of perception with color hues and Ekman (1992) with facial expressions. Concerning linguistic continua, Liberman, Harris, Hoffman and Griffith (1957) demonstrated that stimuli varying along a /b–d–g/ continuum, i.e. stimuli varying in place of articulation, are perceived categorically. The same conclusion was reached with continua where Voice Onset Time (VOT), i.e. the delay between voicing onset and closure release (Lisker & Abramson, 1967), was modified. Specifically, the categorical perception of VOT was evidenced in human adults (Abramson & Lisker, 1970), human infants as young as one month of age (Eimas, Siqueland, Jusczyk & Vigorito, 1971) and even in non-human animals such as chinchillas (Kuhl & Miller, 1975; 1978). It is now accepted that categorical perception is neither speech- nor human-specific.
THE DEVELOPMENT OF CATEGORICAL PERCEPTION
The issue of the development of categorical perception during infancy has been largely documented. As far as VOT is concerned, it is now accepted that during the first year of life, infants move from a language-general to a language-specific mode of perception, i.e. whereas all babies discriminate voicing contrasts according to the three voicing categories separated by two universal boundaries located at -30 and +30 ms VOT (Aslin, Pisoni, Hennessy &
Perey, 1981) during the first months of life, they soon become specialists in their mother tongue by adopting the phonological boundary/ies relevant in the language spoken in their
environment (0 ms VOT in French: Serniclaes, 1987). Far less is known about the following steps of maturation in voicing perception. Zlatin and Koeningsknecht (1975) tested two- and six-year-old English-speaking children with words beginning with a voiced or a voiceless phoneme and showed that from two to six years of age the identification function undergoes a development leading to greater Boundary Precision. Hazan and Barrett (2000) spanned these results to different phonetic contrasts and to English-speaking children from six to 12 years of age. Concerning discrimination scores, Elliott, Busse, Partridge, Rupert and de Graaf (1986) evidenced a negative correlation between age and the size of the just noticeable difference discriminated by English-speaking children from eight to 11 years when tested on a /ba-pa/
continuum.
This study was designed as a continuation of these researches. We aimed at documenting 1/
the development of categorical perception and 2/ the determinants involved in this development, the difficulty being the interacting influences between the different factors of maturation. This study is composed of three experiments, one pertaining to the development of voicing perception and the last two pertaining to the development of colour and facial expressions perception. Specifically, the comparison between these three experiments enabled us to test two specific hypotheses: The reading hypothesis (Burnham, Earnshaw & Clark, 1991; Burnham, 2003) according to which “the intensification of language speech perception between two and six years is related to the onset of reading instruction” (pp. 573) and the general cognitive hypothesis (Karmiloff Smith, 1991; Lalonde & Werker, 1995) according to which categorical perception of speech sounds evolves in synchronicity with other abilities through the influence of cognitive maturation. According to the reading hypothesis, there should be no difference between age groups for either color or facial expression perception.
Since most of the studies in the field have used either identification or discrimination tasks, it seemed important for the purpose of the present study to use both identification and
function is used to determine the Boundary Precision, i.e. the steeper the identification function, the greater the precision (Simon & Fourcin, 1978) and to predict the discrimination scores thanks to a probabilistic formula (Pollack & Pisoni, 1971). The comparison between the discrimination scores expected from identification scores and the observed scores obtained through the discrimination task represents the classical Categorical Perception test.
The stronger the correspondence between the expected and the observed discrimination scores, the stronger the Categorical Perception (Damper & Harnad, 2000). Throughout the paper, we will distinguish categorical perception as the general phenomenon by which a monotonic variation leads to a non–monotonic perception and Categorical Perception (written with capital first letters) as the test that compares expected and observed discrimination scores.
Although either the Boundary Precision or the Categorical Perception tests have been extensively used to assess categorical perception, few studies have integrated them in a single study. Yet, these tests are complementary rather than redundant. Figure 1 presents the possible combinations between Boundary Precision and Categorical Perception measures. In example a/ both Boundary Precision and Categorical Perception are strong, a situation that corresponds to perfect categorical perception. In example b/ Boundary Precision is strong but Categorical Perception is weak indicating a mismatch between expected and observed discrimination scores (described in Damper & Harnad, 2000 ; Medina, Hoonhorst, Bogliotti, Sprenger-Charolles & Serniclaes, submitted). In example c/ Boundary Precision is weak but Categorical Perception is strong reflecting the pattern of results obtained by illiterate adults (Serniclaes, Ventura, Morais & Kolinsky, 2005). In example d/ Boundary Precision and Categorical Perception are weak reflecting both poor labeling abilities and a mismatch between expected and observed discrimination scores.
Figure 1 : Categorical model of perception integrating both Boundary Precision and Categorical Perception indexes as co-factors. Boundary Precision is assessed by the magnitude of the discrimination peak, as illustrated by the differences between the right-hand and left-right-hand columns of the figure. Degree of Categorical Perception is inversely related to the difference between observed discrimination scores (continuous lines) and those expected from identification (dotted lines), as illustrated by the difference between upper and lower lines of the figure.