Repair or Violation Detection? Pre-Attentive Processing Strategies of Phonotactic Illegality Demonstrated on the Constraint of g-Deletion in German.

PURPOSE: Effects of categorical phonotactic knowledge on pre-attentive speech processing were investigated by presenting illegal speech input that violated a phonotactic constraint in German called "g-deletion." The present study aimed to extend previous findings of automatic processing of phonotactic violations and to investigate the role of stimulus context in triggering either an automatic phonotactic repair or a detection of the violation. METHOD: The mismatch negativity event-related potential component was obtained in 2 identical cross-sectional experiments with speaker variation and 16 healthy adult participants each. Four pseudowords were used as stimuli, 3 of them phonotactically legal and 1 illegal. Stimuli were contrasted pairwise in passive oddball conditions and presented binaurally via headphones. Results were analyzed by means of mixed design analyses of variance. RESULTS: Phonotactically illegal stimuli were found to be processed differently compared to legal ones. Results indicate evidence for both automatic repair and detection of the phonotactic violation depending on the linguistic context the illegal stimulus was embedded in. CONCLUSIONS: These findings corroborate notions that categorical phonotactic knowledge is activated and applied even in the absence of attention. Thus, our findings contribute to the general understanding of sublexical phonological processing and may be of use for further developing speech recognition models.

Mismatch negativity (MMN) to successive deviants within one hierarchically structured auditory object.

The Mismatch Negativity (MMN), a component of the event-related potential (ERP), is elicited by a deviant following a series of standard stimuli. The present study addressed the question whether two deviants occurring successively in one hierarchically organized auditory object would elicit separate MMN responses. An example of such an object is a VC syllable (vowel+consonant). In a passive oddball protocol, the syllables [ɔx] and [ɛʃ] were presented both as standards and deviants so that both phonemes, the vowel and the consonant, changed in the respective deviant. Two negative responses were found in the deviant-minus-standard difference waves. Due to the latencies these effects could not be interpreted as separate MMN responses to the phonemic changes. Instead, the first effect (108ms) was taken as an N1 modulation whereas the second negativity (168ms) was interpreted as an MMN mainly reflecting the change of the initial vowel. Furthermore, the present data were statistically compared with related results from Steinberg et al. (2010a, b) obtained partly from the same stimuli presented in oddball blocks with simple mono-phonemic deviances due to changing vowels. Higher MMN amplitudes were found for the syllable [ɛʃ] in the present data compared to those previous findings. Although this difference was only present for one of the stimulus syllables, it is discussed as a possible indicator of an anticipatory MMN enhancement due to the predictability of the second change in the deviant of the present experiment.