The Early Perception of Length Contrasts

Infants’ perception of consonantal length contrasts has been studied far less than, for example, the early perception of laryngeal or PoA contrasts. The scarce data available focus mainly on single age groups and not much is known about how the perception of quantity contrasts in consonants devel-ops during the first and second year of life.

A study by Richardson et al.(2003) examines Finnish infants’ ability to categorise stops varying in CD. Finnish is similar to Swiss German in that it also uses a quantity contrast to distinguish two phoneme categories, as the example mato [t] ‘worm’ versus matto [t:] ‘carpet’ illustrates (Richardson et al.2003:387). In an experiment using the CHTP, Richardson and her col-leagues investigate whether Finnish 6-month-old infants and adults are able to categorise posttonic coronal stops with varying CD values ranging from 95 ms to 255 ms on a CD continuum.²⁴ The authors report successful cate-gorisation by the adults as well as by the infants, with a phoneme boundary location at around 135 ms of CD. Yet, a direct comparison of adults’ and infants’ responses reveals that the infants’ discrimination results are less distinct than the adults’, as can be seen in figure 4.2.

24The principal aim of the study was to discover whether dyslexic adults and their infants differ in their perception of phonemic length contrasts as compared to adults and infants not suffering from dyslexia. Since language impairment is not an issue in the present thesis, only the results of the healthy parents and infants are reported here. The data of the dyslexic adults and their children will be disregarded.

4.4 The Early Perception of Length Contrasts 91

Figure 4.2: Categorisation curves of Finnish 6-month-olds and their parents for posttonic coronal stops differing in CD as estimated from the graphs in Richardsonet al. (2003:392,394).

The pooled adult results form a neat s-curve with 0% [t:]-perception for the stimulus with the lowest CD value and 100% [t:]-perception for the highest CD values. The infants’ curve, by contrast, starts at around 30%

[t:]-perception for the lowest CD value and rises rather gradually than in an s-shaped form to approximately 70% [t:]-perception for the highest CD value. Although there is a noticeable leap in the infant curve at a value near the adults’ cross-over point (between 135 ms and 155 ms), it is obvious that the Finnish 6-month-olds fall far behind their parents in categorising stops varying in CD only.

Infants’ perception of quantity contrasts in stops is also studied by Co-henet al. (1992). In a VF task they test whether American English infants aged 7.5 months are better at discriminating cross-category length contrasts in posttonic labial stops than within-category length contrasts. English does not employ a pure quantity distinction in its stop system. Rather, CD – as also closure voicing and the length of a preceding vowel – can be an addi-tional cue to a laryngeal tense/lax opposition marked by ACT differences.

Tense stops tend to have longer closures than lax stops. Basing themselves on adult data (cf. Parkeret al.1986), they assume a category boundary at 60 ms of CD. In the within-category conditions, stops with a CD value of

20 ms are opposed to stops with a CD of 50 ms. Within the ‘long’ category, stops with CDs of 70 ms and 100 ms, respectively, are contrasted. The cross-category contrasts oppose CD values of 70 ms to 40 ms and of 50 ms to 80 ms.

Since the authors are interested also in potential trading relations between the CD cue and closure voicing, the stimuli are presented with and without glottal pulsing during closure. As a result, Cohen et al. (1992) report cat-egorical discrimination for the unpulsed stimuli; that is, when the stimuli are all voiceless, the dishabituation effect is stronger when the contrast to be discriminated crosses the 60 ms-boundary than when it remains within one category. At first sight, this finding suggests that American English 7.5-month-olds are not only able to discriminate small contrasts in CD but also that their perception has already adapted to the native categories in terms of consonantal length distinctions. However, a closer look at the data reveals that the effect is present only in one of the two conditions, namely when a short stop is presented during habituation and a long stop follows in the test phase. In the reverse condition, infants’ attention remains low during the test phase.

In a second experiment, the stimuli are presented crosswise in terms of voicing. Infants exposed to voiceless stops during the habituation phase are presented with prevoiced stops in the test phase and vice versa. The ‘most natural’ combinations of the voicing cue with length provoke the strongest dishabituation effects. A voiceless stop with a long closure phase is reliably discriminated from a prevoiced stop with a short closure phase, a finding that is reported for both directions of testing. All other contrasts, including the aforementioned conditions with voiceless stops during both habituation and test phase, are less well discriminated. These findings suggest that CD alone is not a strong cue to phoneme distinction for infants acquiring American English.²⁵

Summarising so far, both the study by Richardson et al.(2003) as well as Cohen et al.’s (1992) investigation seem to portend that infants aged 6 and 7.5 months might be sensitive to CD contrasts in stops, but at the same time the data suggest that consonantal length might be more difficult to discriminate than some other consonant contrasts (e.g., PoA contrasts or the short lag/long lag opposition in ACT).

Support for that assumption comes from two further studies which ex-plore older children’s perception of length contrasts. Kuijpers (1996) finds

25Also striking in the study by Cohenet al.(1992) is the unusually high attrition rate (68.4% in the first experiment, 57.1% in the second experiment), which suggests that the results may be generalised only with caution.

4.4 The Early Perception of Length Contrasts 93 that Dutch 4- and 6-year-old children differ significantly from 12-year-olds and adults when they have to categorise labial and coronal posttonic stops varying in 20 ms-steps along a CD-range from 10 ms to 130 ms. The younger children show a markedly higher percentage of ambiguous responses and the cross-over point from one category to the other is extended to a broader boundary region. Dutch does not have a pure quantity contrast in order to distinguish two stop categories. Rather, CD can be a cue in addition to the laryngeal cue provided by the presence and absence of closure voicing.²⁶ As indicated above, the situation is different in Finnish and also in Japanese where consonantal quantity is phonemically distinctive.

A study by Aoyama (2002, also cf. Aoyama 2001) focuses on Japanese and Finnish 3-, 4- and 5-year-olds’ abilities to perceptually discriminate and to produce a length contrast in a coronal nasal. Her results suggest that the acquisition of length contrasts is highly language-specific. The Finnish chil-dren clearly discriminate /hana/ from /han:a/ in a picture-pointing task, whereas the Japanese children are less confident in their judgments. The 3- and the 5-year-olds’ responses are at chance level. With 78.6% correct responses the 4-year-olds are slightly better than the other Japanese sub-ject groups but still they do not reach the same level of performance as the Finnish children. Aoyama (2002) finds a similar difference in produc-tion. The Finnish children produced more systematic quantity distinctions than the Japanese children. Pertaining to children’s differentiation skills in production, Kunnari et al. (2001) also report that Finnish children be-tween 13 and 17.5 months distinguished singleton from geminate stop con-sonants more reliably and earlier than Japanese infants of the same age range. Richardson (1998) finds that in an imitation task at least some of the 18-month-old Finnish infants participating in the study produced reli-able length distinctions between [t] and[t:].

The studies presented in the current section are not sufficient to provide a holistic picture of the early perception of consonantal length. They can offer only a rough impression of how the perception of CD contrasts might develop. The data are hard to compare to each other and do not allow to draw a coherent line of development. The subjects of the studies are from different language backgrounds and of different ages. Moreover, the stimuli differ in the consonants chosen for examination and in the values selected from the CD continuum (presumably due to language-specific conventions

26Note that Kuijpers (1996) remarks that other cues such as formant transitions, burst amplitude and burst duration may have had an influence on the discrimination results in her study.

how CD is used). Furthermore, on the basis of the studies at hand it is not possible to assess the role of the phonological status of length in language acquisition. In other words, no conclusions can be drawn as to whether perceptual development is different for children learning a language with a phonemic length contrast, like Finnish, Japanese or Swiss German, as compared to children acquiring a language in which CD is phonemically irrelevant or just an additional cue to a phoneme contrast primarily marked by laryngeal configurations as in English, Dutch or German.

So far, it can be recorded that infants undeniably possess at least some sensitivity to consonantal length. This is also supported by an ERP-study by Lepp¨anenet al.(2002), who find that electrical activation in the brains of Finnish 6-month-olds is different for short (95 ms) and long (255 ms) coronal stops. Nonetheless, consonantal length contrasts are apparently not at all easy to discriminate as the discrepancies between infants/children on the one hand and adults on the other hand suggest. The available data seem to portend that the perception of the length contrast might develop in a similar way as the laryngeal contrast between prevoiced and voiceless unaspirated stops, namely being weakly present at birth and becoming more reliably discriminable with a sufficient amount of linguistic experience (‘facilitation’).

The present thesis will provide more insights into infants’ perception of a length contrast and its development during the first 16 months of life, both in infants learning a language in which quantity contrasts are represented phonologically in the prosodic structure of the language as well as in infants for whom CD is at most a secondary cue to discriminate tense/lax contrasts.