Decoupling Language Processing from Time

Speech understanding requires listeners to map temporally unfolding speech to lexical items. For decades there has been consensus around the principles governing this process: lexical items are activated immediately as the input arrives, acoustic information flows to word recognition in a continuous cascade, perceptual and lexical representations rapidly decay to make room for new information, and lexical entries are temporally ordered. In this view speech processing is tightly coupled to the input (following “left-to-right” processing). However, recent work challenges this view: listeners often revise earlier decisions, they maintain low-level information for some time; lexical access may be delayed in some circumstances; and lexical representations are not fully ordered. These findings argue for a deep revision of models of language processing.