50
EARLY RECOGNITION
OF MORF’HOPHONOLOGICAL
ALTERNATIONS
James L. Morgan Cognitive
& Linguistic Sciences, Brown University,
Providence,
RI 02912
Word recognition is complicated by the mutability of words. Physical forms of words vary depending on speaker, speaker’s emotional state, utterance position, emphasis, etc. When words come into contact in fluent speech, segments at their edges may change, and their rhythmic patterns may be altered. Moreover, most languages possess morphophonological processes wherein segments may be modified or added to words, modulating word meaning. Across most of these changes, lexical identity remains invariant. However, modifying or adding segments to a word often changes that word’s identity. For example, suffixing /s/ to /win/ produces a distinct lexical item. But suffixing the same segment to /wit/ instead produces a morphophonemic alternate of the base lexical item. Learning which alternations are morphophonemically governed in the native language is an important step in gaining word recognition skills. Three sets of studies investigated 7- to 13-month-old infants’ abilities to recognize familiarized words in fluent speech. In these studies, infants were first conditioned to turn their heads upon hearing a target word pronounced in isolation. Standard head turning procedures were used, except that the reinforcer consisted of a few seconds of a colorful cartoon on a video monitor in lieu of the traditional mechanical toys. Infants who criterially learned the discrimination returned for a test session in which they heard a story sentence-by-sentence. Some of the sentences contained the target word; others contained phonologically or morphophonologically related foils. Infants were credited with responding to targets or foils if they initiated a head turn within 2.5 seconds from the onsets of the words. Head turns were scored from video tapes of test sessions; stimulus onsets were measured from signals written onto the videotape by computer at run-time, so that coders were blind to the nature of stimuli. In the first set of studies, targets were tokens of either /trek/ or /krIp/; foils were tokens in which one segment of the target was altered by one feature. For example, /pak/ and, /tUk/ were foils for /tack/. Twelve infants at each of three ages (7,10, and 13 months) were tested. Infants at all three ages were significantly more likely to respond to targets than to foils, although 7-month-olds were both less likely to detect targets and more likely to false alarm to foils that were 13-month-olds. In the second set of studies, foils were tokens in which an /s/ was affixed or infixed to the target. For example, /s&k/, /taask/, and /tacks/ were foils for /tack/. Only the last of these three foils is a legal English morphophonemic alternate of the target. Twelve infants at each of three ages (7,10, and 13 months) were tested. Seven-month-olds showed no significant differences in responding to targets or foils. Tenmonth-olds were more likely to respond to the target than to any of the foils. Thirteen-month-olds tended to respond to both the target and its morphophonemically legal alternate, but not to the other foils. In a third set of studies, targets were tokens ending in nasals or liquids, such as /rIn/ or /girl/. Foils were tokens to which either /s/ or /z/ had been suffixed. Both segments are phonotactically legal in these contexts, but only voiced /z/ is morphophonemically conditioned here (i.e., tens /tEnz/ is the plural of ten /tEn/, but tense /tEns/ is a different word). Twenty-four 13-month-olds were tested. Infants were significantly more likely to respond to either the base or -/z/ forms than to the -/s/ forms. These results show that infants have begun to learn about the morphophonemic patterns of English by 13 months. On-going studies are exploring knowledge of the /t/-/d/ alternation. Although there is no evidence here that infants are cognizant of the semantic or syntactic significance of these alternations, these results at least raise the possibility that knowledge of grammatical morphology is acquired much earlier that usually assumed. In addition, these results suggest that TRACE-like models of word recognition which rely on competitive activation of constituent segments must also incorporate components through which morphophonemic knowledge can be brought to bear on lexical access.