ISCA Workshop on Plasticity in Speech Perception (PSP2005)

Senate House, London, UK
June 15-17, 2005

Segmental Details in Children's Early Lexical Representations

Suzanne van der Feest, Paula Fikkert

Radboud University, Nijmegen, The Netherlands

Studies on Dutch child language production have claimed that children's phonological representations are underspecified: Fikkert & Levelt (2004) provided evidence for the underspecified nature of coronal place of articulation and Kager et al. (2004) provided evidence for the underspecification of voiceless stops. In both studies the evidence came from asymmetries in production. If phonological representations are underspecified, we expect to find evidence for this in perception as well. Evidence from previous word perception studies suggests that children store word with phonetic detail. Fennell & Werker (2003) used the switch procedure to investigate whether Canadian children perceive the difference between well-known minimal pairs like ‘ball’-‘doll’. They showed that while 14- month-old children did not perceive the difference between the nonce minimal pairs ‘bin’-‘din’ in a word learning task, they did perceive the contrast between the known words ‘ball’-‘doll’. Swingley & Aslin (2000) and Swingley (2003) have shown that 18- to 24-month-old American children are able to detect mispronunciations in well-known words. Using a preferential looking paradigm, they established that children look slower or shorter to the target ‘ball’ when hearing ‘dall’ than when hearing ‘ball’. A variety of mispronunciations were used in these experiments, including changes in vowel height, consonant place, manner and voice, but it was not systematically tested whether asymmetries in perception exist.

Using the same split-screen preferential looking paradigm as Swingley (2003), we tested the perception of mispronunciations of well-known Dutch words by forty-eight 24-month-old Dutch children. Targets consisted of CVC words starting with one of the following stops: /p, b, t, d/. Each target word was used in three conditions: correctly pronounced (‘ball’), with a voice-mispronunciation (‘pall’) and with a place-mispronunciation (‘dall’).

Mispronounced words resulted in non-words or low frequency words unknown to the children. Children were sensitive to both place and voice mispronunciations. Furthermore, the results clearly show perceptual asymmetries in the predicted directions. Perceiving a voice-mispronunciation resulted in longer looking latencies for voiceless targets than for voiced targets. However, when a voiceless stop was perceived, there were no differences in looking latencies for voiced and voiceless targets. Place-mispronunciations led to longer looking latencies for labial-initial than for coronal-initial targets.

These results indicate that children have stored underspecified phonological representations of words. The results cannot be accounted for by assuming that children merely perceive changes in the phonetic realization of targets. Moreover, the data suggest that there is a tight link between perception and production.


Bibliographic reference.  Feest, Suzanne van der / Fikkert, Paula (2005): "Segmental details in children's early lexical representations", In PSP2005, 96.