Discovering words in the continuous speech stream: the role of prosody

Finding words in sentences is made difficult by the absence of obvious acoustic markers at word boundaries, such as silent pauses. Adult speakers of a language are often assumed to rely heavily on direct identification of known lexical items (e.g. McClelland & Elman, 1986). This strategy is however not available to infants who have to first learn the lexicon of their language. Several types of word boundary cues have been experimentally studied, which may be used by infants with relatively little knowledge of their mother tongue: allophonic cues (e.g. the /t/ and /r/ in "night rates" are different from those in "nitrates", Hohne & Jusczyk, 1994), phonotactic cues (some phoneme sequences are impossible within-word, Friederici & Wessels, 1993; Jusczyk, Friederici, Wessels, Svenkerud, & Jusczyk, 1993; Jusczyk, Luce, & Charles-Luce, 1994), typical word pattern (in English most content words start with a strong syllable followed by a weak one, Jusczyk, Cutler, & Redanz, 1993), and prosodic boundary cues (Hirsh-Pasek et al., 1987; Kemler-Nelson, Hirsh-Pasek, Jusczyk, & Cassidy, 1989; Christophe, Dupoux, Bertoncini, & Mehler, 1994; Christophe, Mehler, & Sebastián-Gallés, 2001). As can be seen, Peter Jusczyk and his collaborators have contributed massively to this literature, and have studied the availability of many possible word boundary cues. The reason is that although none of these cues is sufficient to support 100% correct segmentation on its own, taken together they may allow infants to bootstrap their acquisition of a lexicon.