The Learnability of Phonotactic Patterns in Onset and Coda Positions

Abstract

Cross-linguistically, languages allow a wider variety of phonotactic patterns in onsets than in codas. However, the variability of phonotactic patterns in coda position in different languages suggests these patterns must, at least in part, be learned. Two experiments were conducted to explore whether there is an asymmetry in English-learning infants’ ability to learn novel phonotactic patterns in word-medial onset and coda positions. In Experiment 1, English-learning 8- and 12-month-olds were familiarized with bisyllabic nonsense words with voiced fricatives restricted to either word-medial onset (Onset condition) or word-medial coda positions (Coda condition). At test, infants heard novel words that either followed or violated the familiarized patterns. Twelve-month-olds in the Onset condition showed a preference for the familiarized pattern; however, 12-month-olds in Coda condition as well as 8-month-olds in both conditions displayed no such preference. To explore whether there is a developmental pattern in these positional effects, in Experiment 2, 15-month-olds were tested in only the Coda condition. These toddlers showed a preference for the familiarized phonotactic pattern. These findings support the hypothesis that infants’ learning of positional phonotactics is asymmetrical.

Notes

¹ Though infant-directed corpora may provide more reliable approximation of infants’ input, the lack of large English IDS corpora with mothers’ speech transcribed and tagged in a way similar to CELEX prevents us from easily investigating the distribution of fricatives in IDS.

² However, we cannot be sure how infants in our study have interpreted the stimuli. It is possible that they may also have interpreted the stimuli as compound or two-word sequences, given consonant clusters are either phonotactically illegal in English or were of low frequency word-medially.

³ It should be noted that for illegal lists, we chose either CVF.CVC (for condition Onset) or CVC.FVC (for condition Coda) rather than CVC.CVC for both groups because we would like to make sure infants’ preference, if any, for legal or illegal lists in Test phase is not confounded by segmental difference. Were we use CVC.CVC for illegal lists, even if infants’ show preference for legal lists, we still do not know whether they prefer the familiarized phonotactic pattern per se, or they just prefer words with more fricatives.

⁴ Furthermore, the 15-month-olds infants (in Experiment 2) showed a novelty preference in the Coda condition, thus providing us with more evidence that 12-month-olds in Experiment 1 have not learned the pattern in coda position.

⁵ To answer the question as to whether there is any correlation between infants’ morphology and prosodic development we have been tracking 15-month-olds participants’ vocabulary development. Parents are asked to complete a MacArthur Communicative Development Inventory (CDI) – Words and Gestures when infants are 18-months-old.Twelve 15-month-olds have participated in the Coda condition; however, we were able to gather only 7 infants’ vocabulary outcome at 18-month-olds due to attrition. We conducted a bivariate correlation analysis between infant looking times and vocabulary output. The result showed that these two variables were not correlated, r = -0.367, p =.419. This data provides evidence that 15-month-olds’ learning of coda phonotactics in our study is not morphologically driven.

⁶ Segments in stressed syllables may have an acoustic advantage over those from unstressed syllables, and this is especially true in IDS given that prosodically salient units are exaggerated more than unstressed syllables (Wang, Cristià, & Seidl, in revision). Furthermore, this exaggeration applies to both vowels and consonants. For example, Cristià (2010) found that sibilants were enhanced in IDS as compared to ADS. In order to compare the acoustics of segments in onset and coda position, we segmented the target fricatives in Praat (Boersma & Weenink, 2011) and extracted three acoustic measurements, duration, amplitude, and fundamental frequency (F0), from the target fricatives across the familiarization (120 tokens) and test trials (96 tokens). We obtained the duration and amplitude measurements for all the target fricatives (108 tokens for onset and 108 tokens for coda); however, only 193 tokens (107 for coda, and 86 for onset) F0 measurements were obtained due to devoicing of some target segments. Analyses of variance (ANOVA) were performed on each of the measurements with two independent variables: syllable position and segment properties. Results indicate that word-medial coda fricatives were greater both in amplitude, F(1, 215) = 34.54, p < .0001, and F0, F (1, 192) = 20.42, p < .0001, than word-medial onset fricatives. However, there were no significant differences in duration F(1, 215) = 0.13, p = .716. Thus, analyses of acoustic measurements made on the target fricatives suggest that voiced fricatives are more salient in word-medial coda position than in onset position. This may result from hyperarticulation of fricatives in stressed syllables.