Semantic activation and verbal working memory maintenance in schizophrenic thought disorder: insights from electrophysiology and lexical ambiguity

Abstract

We have examined language processing using ambiguous words (homographs like panel or toast) and rapid or slow presentation rates while recording event-related brain potentials (ERPs). Homographs allow for tracking the train of thought at points of lexical ambiguity and detecting modulation of associative threads by previous context. Rapid presentation rates stress automatic semantic activation, and slow rates stress controlled verbal working memory contextual modulation. In conjunction with reaction times and performance, ERPs allow for objective measurement of activity related to language processing from word presentation through overt behavioral response. Smaller N400 to related and unrelated items at short stimulus onset asynchronies (SOAs), the presence of a semantic bias, and large N400 to related and unrelated items at long SOAs are present in schizophrenia. We describe a model of initial semantic memory hyper-priming and subsequent decay of information in verbal working memory stores, the activation-maintenance model of schizophrenic thought disorder hypothesized to underlie the thought disorder in schizophrenia.

Figure 1

Comprehension errors on a homograph disambiguation task in chronic and first hospitalized psychosis. Congruent: Semantically congruent endings for unambiguous nouns. Dominant: Semantically congruent endings for dominant homograph noun meanings. Subordinate: Semantically congruent endings for subordinate homograph noun meanings.

Figure 2

Panel A. N400 ERPs on a passive reading homograph disambiguation task in chronic schizophrenia. Note the largest N400 to subordinate homograph endings, but large N400 to all sentences in schizophrenia. (After Salisbury et al. 2002.) Panel B. N400 ERPs on an active comprehension of homograph disambiguation task in chronic schizophrenia. Note the large N400 to subordinate homograph endings that made sense in schizophrenia (Subordinate) is as large as their N400 to subordinate endings they did not comprehend (Subordinate Errors), whereas controls showed modulation by comprehension. (After Salisbury et al. 2004.)

Figure 3

Panel A. Graded N400 responses in controls is absent in schizophrenia. (After Salisbury et al., In Press). Panel B. The lack of graded N400 response to different word types in schizophrenia at slow presentation rates confirmed in a new sample. Panel C. Note that with fast presentation rates, the graded response disappears in controls, with as large N400 to subordinate meaning endings as to incongruent endings. The N400 is smaller here in schizophrenia.