Tania S. Zamuner, LouAnn Gerken & Michael Hammond

University of Arizona / University of Nijmegen

{zamuner, gerken, hammond}@u.arizona.ed

Previous research has shown that children’s productions are positively correlated to the frequency of sounds in the ambient language. It is also possible, however, that children’s production of the same segment differs according to the frequency of the surrounding sounds in the word. Here a deeper interpretation of the input is explored, which is one of phonotactic probability, and this allows for an examination of what input factors contribute to children’s early productions.

There is a large body of research looking at phonotactic probability in English CVC words. Although these studies looking at CVC probabilities use different tasks and span age groups, they have found that infants, young children, and adults are sensitive to phonotactic probability. Specifically all groups show a preference for non-words composed of high phonotactic probabilities. For example, Jusczyk, Luce and Charles-Luce (1994) presented infants with novel CVC words with either low or high phonotactic probabilities and found that by 9 months, infants showed a preference for the high phonotactic probability non-words. This demonstrates that even at an early age and before the onset of meaningful speech, infants are sensitive to the frequency of the sound patterns in the ambient language. It also demonstrates that they are able to encode this information at some level. Other studies, such as that by Treiman, Kessler, Knewasser, Tincoff and Bowman (2000) have found that in non-word decision tasks, young children and adults judge high phonotactic probability non-words as more "word-like" than low phonotactic probability non-words. Similar sensitivities are illustrated in a study by Storkel (1999), which found that children learn high phonotactic probability non-words with fewer exposures than non-words with low probabilistic phonotactics.

Given this body of research, it was asked whether a sensitivity to phonotactic probability would be mirrored in young children’s coda productions. To this extent, an experiment was designed that controlled for the frequency of phonemes in nonsense words and that contained the identical coda. This allowed for comparisons of children’s productions of the same coda consonant in low and high phonotactic probability environments. The rationale was that if phonotactic probability would differentially affect children’s production of the same coda consonant, this would further suggest that the frequency of sounds in the ambient language plays a substantial role in children’s acquisition of phonological structures.

The prediction was that children’s productions of the same coda consonant would differ according to the environment in which the coda occurred. That is, the same coda will be produced more often in high phonotactic probability non-words than in low phonotactic probability non-words. For example, we hypothesised that children would differentially produce the coda /l/, depending on the phonotactic probabilities of the non-word. Twenty nine children between 20 and 28 months participated in a repetition experiment of novel words. Words were controlled for phonotactic probabilities, coda type, and vowel length. Results showed that codas were more likely to be produced when they appeared in high probability environments. For example, the /l/ in the high phonotactic probability word /gEl/ was more likely to be produced than the /l/ in low phonotactic probability word /pÃl/. These results show a strong relation between frequency of sound patterns in the ambient language and children’s early productions.

In addition to testing 2-year olds production of the non-word stimuli created, 7-month-old infants were also tested on whether they could differentiate between the low and high phonotactic probability CVC non-words described above. Twenty 7-month-old infants were exposed to list of low and high phonotactic probability CVC non-words, using the Headturn Preference Procedure. Results showed that infants showed a significant preference for CVC non-words with high phonotactic probabilities (M = 9.66, SD = 2.95) than those with low phonotactic probabilities (M = 8.47, SD =1.94), t(19) 2.66, p < .05, one-tailed. This replicates the finding shown by Jusczyk et al, (1994), yet with infants who are two months younger. This illustrates that even by 7-months-of-age, infants are able to distinguish between frequency and infrequent sounds and sound combinations in the ambient language. This also suggests that children’s sensitivity to the input begins before production, such that the patterns of coda consonant development in production reflect the learner’s sensitivity to the input at a very young age.