Item

Abstract

In everyday communication, listeners not only deal with noisy environments but also with ambiguous linguistic inputs. Here we asked how neural oscillatory processes reflect the cascade of perceptual and decision processes that make up word comprehension. In an electroencephalography study (EEG; N=11) and a lexical decision task, the ease of lexical access was parametrically varied: Stimuli were three-syllabic German nouns (“real” words); pseudo-words with an exchanged vowel in the second syllable (“ambiguous” pseudo-words); and syllable-scrambled (“opaque”) pseudo-words. All stimuli were embedded in white noise at an individually determined signal-to-noise-ratio (based on 70% accuracy in vowel-in-noise discrimination). Participants detected well opaque pseudo-words (89% accuracy) but were significantly worse in evaluating ambiguous pseudo-words and real words (60% and 71% resp.). Nevertheless, the brain response differentiated these two perceptually close stimulus classes: First, real words were marked by an enhanced phase coherence in the alpha frequency range (500–570 ms), followed by ambiguous pseudo-words eliciting a more negative event-related potential (ERP) and enhanced alpha (8–12 Hz) power than real words. In contrast, when analysed by response (“word” vs. “not a word”), more negative ERPs, higher alpha power, and stronger alpha coherence for “not a word” responses were observed (all > 800 ms, i.e. after stimulus offset). To conclude, a left-lateralised alpha-frequency network is sensitive to uncertainties at the perception and at the decision level. The late, stimulus-independent effects highlight the interesting hypothesis that lexical decisions in speech are not entirely based on perceptual evidence but critically depend on internal brain state.