Evaluating models of phonotactic constraints on the basis of sC cluster acquisition data

This paper aims to test the model of phonotactics called NAD (Net Auditory Distance) with the acquisition data to see whether it makes better predictions than sonority-based models. NAD assumes that the phonotactic preferability of a cluster depends on perceptual distance between particular sounds in this cluster. The index of perceptual distance within NAD is based on phonetic features: manner of articulation (MOA), place of articulation (POA) and voicing. To test the predictions of our model we used data gathered from 50 Polish kindergarten children, who produced for us six initial double clusters: /st/, /sp/, /sk/, /sx/, /sm/ and /sw/. This sC set ensured that clusters were similar, yet differing on NAD scales. To check for the preferences, we calculated the number of cluster reductions applied for each cluster group. In accordance with the newest version of NAD principle, we predicted that the /st/ would be the most dispreferred (reduced) cluster, followed by /sp/, /sk/ and /sx/, /sm/ and finally /sw/. The results only partially confirmed this hypothesis. For the /sx/, /sm/ and /sw/ clusters, our prediction turned out to be accurate, but the theory did not predict correctly the results for the s+stop clusters. In fact, the results pointed out to /st/ (the most dispreferred cluster according to NAD), as the most preferred sequence. Overall, our results point to a special status of the s+stop clusters among all the other sound sequences. The ambiguous nature of these clusters (they are dispreferred in many phonological theories but relatively salient acoustically) may render them more vulnerable to frequency effects. This would explain the preference for /st/, which is the third most frequent initial double cluster in Polish. Our study thus shows that in ambiguous cases language-specific idiosyncrasies can override the universal tendencies in language acquisition.