Development of phonologically specified word forms

Word segmentation is qualitatively different between infants and adults, occurring in tandem with word recognition in adults, with words segmented as they are recognized from the speech stream. This requires a lexicon consisting of most commonly occurring words, which infants lack. Thus, infants are faced with the problem of needing to segment words to add them to the lexicon, but not having a lexicon large enough to segment words online as adults do. Thus, early word segmentation occurs separately from word recognition. Infants segment words primarily by attending to the predominant metrical patterns in their language as cues to likely word boundaries, and can then learn the segmental patterns within those boundaries. In adults, attention to metrical patterns is a supplementary cue, and thus word segmentation by adults is robust to suprasegmental variation that can alter metrical patterns, such as that across foreign-accented speech. Word recognition is also qualitatively different between young children and adults. Early word forms are comprised of strict phonetic patterns based on experienced pronunciations that are intolerant to previously unexperienced variation. However, mature word forms are phonologically specified, comprised of the higher-order phonological form of the word above the level of phonemic variation. This allows a robustness to considerable segmental variation, even variation that violates native phonemic boundaries, as can occur across regional accents. Adults are generally able to recognize cross-accent pronunciations via their grasp of when variation does not change a word's underlying phonological form. This thesis explores when these qualitative differences between immature and mature word segmentation and recognition begin to be resolved by young children