Perception mirrors production in 14- and 18-month-olds: The case of coda consonants

Abstract

In a word learning experiment, 14- and 18-month-old infants are tested on their perceptual sensitivity to coda-consonant omissions. The results indicate that 14-month-olds are not sensitive to coda consonant omissions, showing a parallel with the omission of target coda consonants in early child language productions. At 18 months, infants are sensitive to coda-omission. The study strengthens the hypothesis that phonological wellformedness constraints influence infants’ speech processing in general, and might restrict what is stored in their initial lexical representations. A lexical representation lacking information on the target coda consonant is, in turn, a likely source for coda-omissions in production.

Introduction

When babies produce their first meaningful words, for some time they omit target coda consonants (e.g. Demuth, 1995, Fikkert, 1994, Ingram, 1978, Menn, 1976, Moskowitz, 1970, Smith, 1973). Examples from English (Menn, 1976) and Dutch (CLPF database: Fikkert, 1994, Levelt, 1994) are in (1):

In the literature, this type of omission is accounted for in terms of phonological wellformedness constraints on young children’s output forms: a syllable template that only allows C(onsonant)V(owel) sequences (Demuth, 1995, Macken, 1979, Menn, 1978), a coda-parameter with a default setting “no codas allowed” (Dresher and Kaye, 1990, Fikkert, 1994) or a no-coda constraint that is initially high-ranked in an optimality theoretic grammar (Levelt, Schiller, & Levelt, 2000).

In the above accounts it is, implicitly or explicitly, assumed that children’s underlying forms, i.e. the stored lexical representations, are segmentally detailed. The argument for this assumption is that infants are able to perceive phonetic detail from very early on (Smith, 1973, Smolensky, 1996). Indeed, in several studies infants have demonstrated sensitivity to mispronunciations (e.g. Bailey and Plunkett, 2002, Ballem and Plunkett, 2005, Mani and Plunkett, 2007, Swingley, 2003, Swingley, 2005, White and Morgan, 2008). However, since infants failed to do so in other studies (e.g. Pater et al., 2004, Stager and Werker, 1997, Van der Feest, 2007), in the present study the lexical representation of target coda consonants that infants construct during word-learning is investigated.

Jusczyk (1998) hypothesized that principles of phonological wellformedness may play a role in speech processing from the start, predisposing infants to attend to certain aspects of the speech signal. Jusczyk, Smolensky, and Allocco (2002) indeed found evidence for a perceptual equivalent for the phonological output constraint on nasal place assimilation, in 4,5- and 10-month-old infants.

In Jusczyk (1998) preliminary results are reported showing that 6-month-old infants have a perceptual bias for syllable structures that satisfy both the coda constraint and a constraint requiring onsets, i.e. CV-syllables. If this bias would also lead infants to parse CV-syllables from the acoustic signal during word-learning, it is conceivable that target coda consonants initially remain un(der)represented in the lexicon. In the present study the hypothesis is, thus, that infants’ lexical representations for target coda consonants remain heavily underspecified. Such incomplete representations could form the initial source for coda-deletions in early word productions.

Up until now, only a few mispronunciation studies have looked at coda consonants. Swingley (2005) tested mispronunciations in both onsets and codas of known words in a head-turn experiment with 11-month-olds. While he found a clear mispronunciation effect for onsets, the results for codas were less obvious; Infants preferred listening to familiar words over nonwords, but this preference went away when the familiar words’ codas were mispronounced. This suggests that infants knew the codas’ phonological features. On the other hand, infants did not prefer familiar words over altered-coda variants of those words, suggesting that they did not distinguish between correct and incorrect codas. Zamuner (2006) found similar negative results for coda-mispronunciations with 11-month-olds and mixed results with 16-month-olds. In Swingley (2009) onset and coda mispronunciations were tested again using a referential visual fixation task with 17-month-olds (mean age). For this age group he did find a mispronunciation effect for both onsets and codas. In a similar experiment with 14-month-olds by Langeslag, Altvater-Mackensen, and Fikkert (2008), however, no mispronunciation effect was found for stop-fricative mispronunciations in codas, while the effect was present for onsets. In general, then, younger infants, i.e. the 11- and 14-month-olds, seem to pay less attention to codas.

All of these studies on coda consonants tested the infant’s sensitivity to a segmental change. In the present study the perception and representation of coda consonants is considered from the perspective of the earliest word productions, and here target coda consonants are omitted, rather than changed. In the experiments described below, the coda in mispronounced test items is therefore omitted. In the control condition the target word has a CV syllable structure, and a coda is added in the mispronounced test item. Coda-addition never happens in early word-productions.

We tested 14-month-olds, since this age group often omits codas from target words in production. The prediction is straightforward: if 14-month-olds indeed build incomplete representations during word-learning, parsing CV structures from the acoustic input, they should be tolerant of coda-omissions in mispronounced items. We do expect sensitivity to coda-addition. One experiment was repeated with 18-month-olds, who do usually produce the codas of target words.¹

If this entails that the coda-constraint on speech processing is lifted, they should be sensitive to coda-omission in the mispronounced test-item.