Cortical sequence for processing and learning spoken word forms

Spoken words have a special status in human auditory processing. Spoken language is special also in that new words and a foreign language sound system can be acquired quickly and effortlessly, especially during childhood. This thesis focuses on the cortical substrates of processing and learning spoken words. Cortical activation is measured with magnetoencephalography (MEG), which enables tracking the sequence of activation with high temporal precision and separating different processing stages reflected in this activation. Study I investigated how spoken word processing is different from other everyday sounds with corresponding meanings. We used machine learning methods to examine how different features of speech sounds (acoustic, phonemic, semantic) are encoded in cortical activation. Decoding of spoken words based on their cortical evoked responses improved markedly when the MEG signals were modeled as time-locked to the spectrogram, compared to non-time-sensitive models; this was not true for environmental sounds. The results indicate that the spectrotemporal content of speech is processed in the auditory cortex in a special time-locked manner. Successful prediction of the phoneme sequences of the words at the same time delay as the acoustic content indicated that the function of this time-locking could be the transformation of the continuous acoustic signal to discrete linguistic representations. Studies II-IV examined how incidental learning of new native and foreign word forms in an overt repetition task is reflected in cortical activation. The results indicate that in adults learning is confined to individual word forms, reflected in left-lateralized areas that are related to phonological input and output processing (superior temporal and premotor cortex, respectively). Manipulations of noise level and task after learning highlighted the role of the superior temporal responses in memory retrieval and the premotor responses in articulatory planning. Comparison of school-aged children and adults revealed that children show a functionally similar cortical effect of word form learning in superior temporal cortex as adults, but this effect is located to the right hemisphere, outside of the traditionally left-lateralized core of the language network. Additionally, children showed generalized changes in the processing of new phonological input overnight. The results point to developmental differences in the degree to which processing of new phonological input is confined by native-language phoneme segmentation in left-hemispheric regions. The results of this thesis show that features of speech are more accurately tracked in time in auditory cortical activation than other sounds, and that phonological processing and learning is different in school-aged children than adults. These findings highlight the special nature of speech and the abilities of the human brain to specialize and adapt to the linguistic environment.