Brief articleThe role of exposure to isolated words in early vocabulary development
Introduction
Between the ages of 9 and 21 months, children typically progress from speaking at most a handful of words to speaking over 200 (Fenson, Dale, Reznick, Bates, & Thal, 1994). To learn a word, a child must store its sound pattern, its meaning, and an association between the two. Since fluent speech contains no known acoustic analog of the blank spaces between words of printed English, children must segment the speech signal in order to learn words from multi-word utterances. Early in the study of language acquisition, the segmentation problem received little attention; it was tacitly assumed that children learned words primarily from isolated occurrences. In the 1980s, it was suggested (Peters, 1983, Pinker, 1984) that words learned in isolation could help children segment multi-word utterances containing novel words. However, there was little systematic empirical evidence for this hypothesis. In fact, there was some evidence that infant-directed speech does not reliably provide isolated words (Aslin, Woodward, LaMendola, & Bever, 1996).
In the last 10 years there has been tremendous progress in understanding infants' ability to segment fluent speech. By 7.5 months, infants can recognize words that they have heard in fluent speech when those words are later presented in isolation (Jusczyk & Aslin, 1995) and can even do so after a 2-week delay (Jusczyk & Hohne, 1997). Further, 8-month-old infants are able to exploit patterns in sequences of nonsense syllables to help isolate word-like units from synthesized speech that lacks any other segmentation cue (Saffran, Aslin, & Newport, 1996). This research has been accompanied by a shift away from the view that early vocabularies are learned primarily from isolated words.
In this paper, we revisit the potential role of isolated words in early word learning. It has long been known that infant-directed speech tends to consist of short utterances (e.g. Snow, 1977) separated by relatively long silent pauses (Fernald et al., 1989). Recently, the question of isolated-word frequency has been addressed in a study of a single subject (van de Weijer, 1998). This paper reports a multi-subject investigation of four empirical questions:
- 1.
Are isolated words a normal and reliable feature of spontaneous infant-directed speech?
- 2.
Do children hear repeated instances of a variety of distinct words in isolation?
- 3.
Does a significant proportion of children's earliest vocabularies consist of words that they have heard in isolation?
- 4.
Does exposure to isolated instances of a word predict later knowledge of that word, above and beyond total frequency of exposure?
Section snippets
Subjects
Thirteen English-speaking mothers with first-born infants under 9 months old were recruited by advertising in Baltimore's Child, a free newspaper distributed in and around Baltimore, Maryland. Three mothers who missed four or more recording sessions each were dropped from the study. Of the remaining ten mothers, two were eliminated due to consistent recording difficulties attributable to experimenter error. The remaining eight were selected for transcription and further study. Three of the
Frequency and reliability of isolated words
The first question we asked was how frequent isolated words are. To investigate this question, we analyzed three early transcripts per child. In order to separate the frequency of isolated words from the overall number of utterances produced by each mother, we used only the first 600 maternal utterances from each transcript, approximately the smallest number of maternal utterances in any transcript. On average, 9.0% of the maternal utterances consisted of isolated words. There was no
Discussion
The data and analyses described above provide positive answers to the four empirical questions posed in Section 1. First, isolated words are a regular occurrence in the experience of almost every infant in the population from which this sample was drawn. Second, the isolated words to which infants are exposed comprise a variety of distinct word types, a number of which are repeated in close temporal proximity. Third, a substantial proportion of the first 30–50 words produced are words typically
Acknowledgements
We wish to thank the undergraduates at Johns Hopkins University who recorded and transcribed these sessions: Sarah Carricaburu, Anya Kanevsky, Jeannie Park, Matt Shomphe, and Marni Soupcoff, and the research assistants who worked on this project at Washington University: Angela Pelch and Bridget Gaertner. Evan Keibler provided essential programming support. Thanks also to Richard Aslin, Sven Mattys, Elissa Newport, and Jenny Saffran for helpful comments on earlier drafts of this manuscript.
References (24)
- et al.
Distributional regularity and phonotactics are useful for segmentation
Cognition
(1996) - et al.
Acoustic determinants of infant preference for motherese speech
Infant Behavior and Development
(1987) - et al.
Infants' detection of sound patterns of words in fluent speech
Cognitive Psychology
(1995) A computational study of cross-situational techniques for learning word-to-meaning mappings
Cognition
(1996)- et al.
A cross-language investigation of infant preference for infant-directed communication
Infant Behavior and Development
(1994) - et al.
Models of word segmentation in fluent maternal speech to infants
An efficient, probabilistically sound algorithm for segmentation and word discovery
Machine Learning
(1999)- et al.
Preference for infant-directed speech in the 1st month after birth
Child Development
(1990) - et al.
Issues of complexity in English and Inuktitut child-directed speech
- Fenson, L., Dale, P. S., Reznick, J. S., Bates, E., & Thal, D. (1994). Variability in early communicative development....