Probabilistic phonotactic knowledge facilitates perception, but categorical phonotactic illegality can cause misperceptions, especially of non-native phoneme combinations. If misperceptions induced by first language (L1) knowledge filter second language input, access to second language (L2) probabilistic phonotactics is potentially blocked for L2 acquisition. The facilitatory effects of L2 probabilistic phonotactics and categorical filtering effects of L1 phonotactics were compared and contrasted in a series of cross-modal priming experiments. Dutch native listeners and L1 Spanish and Japanese learners of Dutch had to perform a lexical decision task on Dutch words that started with /sC/ clusters that were of different degrees of probabilistic wellformedness in Dutch but illegal in Spanish and Japanese. Versions of target words with Spanish illegality resolving epenthesis in the clusters primed the Spanish group, showing an L1 filter; a similar effect was not found for the Japanese group. In addition, words with wellformed /sC/ clusters were recognised faster, showing a positive effect on processing of probabilistic wellformedness. However, Spanish learners with higher proficiency were facilitated to a greater extent by wellformed but epenthesised clusters, showing that although probabilistic learning occurs in spite of the L1 filter, the acquired probabilistic knowledge is still affected by L1 categorical knowledge. Categorical phonotactic and probabilistic knowledge are of a different nature and interact in acquisition.

Humans learn from statistical regularities in the environment. We tested if prediction and prediction error may play a role in such learning in the brain. We used Error-Driven Learning (EDL) to simulate participants’ trial-by-trial learning during exposure to a bimodal distribution of non-native lexical tones. We simulated incremental trial-by-trial learning to get estimates of the degree of expectation of upcoming stimuli over the course of the experiment. The expectation estimates were combined with Temporal Response Function fitting to generate a prediction of the trial-by-trial ERP waveform. EDL simulations captured the data significantly better than chance and better than models based on either stimulus characteristics or statistical distributions. The results provide tentative evidence that trial-by-trial learning as measured in neural activity is error-driven.

No abstract

This paper investigates to what extent speakers adapt to unfamiliar consonant cluster timing patterns. We exploit naturally occurring consonant overlap differences between German and Georgian speakers' productions to probe the constraints that language-specific patterns put on the flexibility of cluster articulation. We recorded articulography data from Georgian and German speakers imitating CCV clusters as produced by a German and Georgian audio model, respectively. The German participants adapted their relative overlap towards the Georgian audio model to various degrees depending on whether the cluster was phonotactically familiar to them or not. A higher degree of adaptation was observed for clusters phonotactically illegal in German. Phonotactically legal clusters showed only an intermediate degree of articulatory adaptation, even though acoustically these clusters showed a rather strong move towards the Georgian audio model in terms of the aerodynamics of the interconsonantal transition period. Georgian speakers on the other hand failed to adapt to the German audio model articulatorily and acoustically, possibly because the German cluster inventory is a subset of the Georgian inventory. This means that Georgian speakers can draw on native speaker knowledge for all clusters, which is a factor known to constrain imitation. Also language-specific cue weighting effects may partly condition the results.