Authors:

SHTYROV, Y., GARAGNANI, M., Kujala, T., Wennekers, T. & PULVERMULLER, F.

Reference:

Year of publication:

2007

CBU number:

6623

Abstract:

When subjects are presented with words and matched meaningless pseudowords while they are distracted in an oddball paradigm, their early mismatch negativity (MMN) brain response is greater for the words. In contrast, greater late N400 brain responses to pseudowords than words emerged in tasks where subjects have to attend to the language stimuli. A series of neurophysiological, neuroimaging and neurocomputational studies investigated this apparent contradiction. Using EEG and MEG and a variety of languages, we found a robust lexical enhancement (words > pseudowords) in passively acquired MMN responses at approximately 150 ms after critical auditory information is available. This is considered as neural signature of automatic lexico-semantic processes (activation of word memory traces) in the brain. MEG results suggested auditory cortex in planum temporale vicinity as a structural seat for this enhancement. More specifically, our fMRI investigation indicated a role of left-hemisphere's superior-temporal gyrus and sulcus, and possibly middle-temporal gyrus, in sustaining such word-specific memory circuits. To assess the differential MMN and N400 pattern, we modelled the neuronal processes involved in word recognition computationally. The results indicated dissociating effects of attention, modelled as non-specific inhibition and feedback control, to auditory input on neural processing. Simulation of passive non-attend conditions led to an early word advantage, whereas simulation of attended-conditions explained relatively enhanced late responses to pseudowords. This explains the divergence between MMN and N400 results on the basis of attention level mechanistically realised as neuronal feedback regulation with different gains. To further substantiate this model of interactions between cerebral processes of attention and lexical access, we recorded neurophysiological brain responses to words when the subjects were asked to attend to the spoken input or to ignore it. The results confirmed further predictions of the model, especially a robust automatic activation of word-specific memory traces which, during approximately the first 140 ms of language processing, was not significantly influenced by attention level. Within the same time window, pseudoword responses were significantly modulated by attention level. The data confirm earlier suggestions that initial stages of lexical processing are not affected by attentional demands and may thus be automatic. Attention effects on lexical processing emerged at latencies >200ms.