Are Word-Final Sounds Perceptually Salient for Infants?

Using the HPP, this study extended Jusczyk and Aslin (1995) by familiarizing 7.5month-olds with nonwords (e.g., [beyp]) and then presenting test passages containing words that differ from the nonwords in their final consonant (e.g., 'bike', [beyk]) along with new unfamiliar words. The infants did not false alarm to the test passages containing the similar sounding words. Our results suggest that infants have detailed phonetic representations of familiar sound patterns. Introduction A critical task for young children acquiring their native language is building a vocabulary that allows them to refer to objects in their environment. In order to develop this vocabulary, young children must first be able to isolate individual words from the continuous stream of speech that they hear. A recent series of experiments by Jusczyk and Aslin (1995) suggests that the ability to segment individual words from fluent speech may emerge as early as 7.5 months of age. For example, infants, familiarized with the words 'bike' and 'feet', listened longer to passages containing these words than to passages containing new unfamiliar words such as 'cup' and 'dog' (see Figure 1). This finding suggests that 7.5-month-old infants are able to form representations of sound patterns with relatively minimal exposure. However, the question remained as to how detailed these representations were. Jusczyk and Aslin (1995) explored this issue in a second experiment. They familiarized 7.5-month-olds with nonwords such as [geyk] and [zit] and then presented the infants with four passages containing the real words: 'cup', 'dog', 'bike', and 'feet'. These nonwords differ from the real words in their initial consonants. In this case, they did not find a significant difference between the infants' responses to the passages containing words similar to the nonwords heard during familiarization (e.g., [geyk] repetitions and 'bike' passage) and the passages containing new words (see Figure 2). These results suggest that the initial consonant of a familiar sound pattern is very salient for infants. Jusczyk, Goodman, and Bauman (1999) provide further support for the perceptual salience of initial consonants. They found that 9-month-old infants preferred listening to isolated word lists that shared the same first consonant as compared to unmatched lists. In addition, Newport, Gleitman, and Gleitman (1977) hypothesized that a possible learning strategy for the child is to "listen to familiar beginnings (p. 138)". Given that infants are more focused on the beginnings rather than on the ends of words, it is possible that they may not detect changes to the endings of targets in the word detection paradigm. However, it is also possible that for infants to segment the speech stream into isolated words they need to identify the end of a word as well. Results from Saffran, Newport, and Aslin (in press) show that adults learning the "words" of a nonsense language appear to learn the ends * Poster presented at the Fifth Conference on Laboratory Phonology at Northwestern University, Evanston, Illinois July 1996. of words first. Slobin's (1973) operating principle, "pay attention to the ends of words (p. 412)", also stresses the importance of language learners focusing on word final sounds. Figure 1. Study 3 mean looking times (error bars = SE), Jusczyk & Aslin (1995). The present experiment extends Jusczyk and Aslin (1995) by including nonwords, such as [beyp] and [fik], that differ from the real words 'bike' and 'feet' in their final consonant. Infants are first familiarized with repetitions of the nonsense words and then listen to the test passages containing the similar sounding words or new unfamiliar words. If the infants have detailed representations of the familiarized nonwords, then they should not false alarm to the similar sounding words in the test passages. Stimuli We constructed the target nonwords by changing the place of articulation of the final consonant in the original target words used by Jusczyk & Aslin (1995). The original target words were 'cup', 'dog', 'bike', and 'feet'. Our manipulation yielded the target nonwords used in the present experiment: [kut], [dçb], [beyp], and [fik]. The test passages in the present experiment were the same passages used by Jusczyk & Aslin (1995). The position of the target word in the sentences varied such that the target word was spoken in the beginning of two sentences, in the middle of two sentences, and at the end of two sentences. Each target word was preceded by one of six words that were used in all four sets of passages. (e.g., bike passage :His bike had big black wheels. The girl rode her big bike. Her bike could go very fast. The bell on the bike was really loud. The boy had a new red bike. Your bike always stays in the garage.) The same female talker (native speaker of American English) who had recorded the test passages also recorded the isolated nonwords that the infants would listen to in the familiarization phase. For each target nonword, the talker repeated the item with some variation fifteen times in a row, in a lively voice. Participants We tested twenty-four infants (10 females, 14 males) approximately 7.5 months of age (mean age: 231 days, range: 213-244 days). All infants were from monolingual English-speaking homes. An additional twelve infants failed to complete the entire testing session.