SIGNAL TO SYNTAX : Bootstrapping From Speech to Grammar in Early Acquisition

The hypothesis that young infants rely on prosodic cues in speech 10 bootstrap their way into syntax has received considerable attention in recent discussions of early language development (e.g., Gleitman, Gleitman, Landau, & Wanner, 1988; Hirsh-Pasek, Kemler Nelson, Jusczyk, Cassidy, Druss, & Kennedy, 1987). The appeal of the prosodic bootstrapping hypothesis is easy to understand. If the boundaries be­ tween syntactic constituents in speech were indeed reliably marked by constellations of prosodic features such as pauses, pitch contours, and vowel lengthening, this acoustic punctuation could potentially be useful to the child beginning to learn language. And if this syntax-to-prosody mapping were more distinctive and reliable in infant-directed speech (IDS) than in adult-directed speech (ADS), the prosodic structure of IDS could provide even greater support for the infant's initial efforts at pars­ ing the speech stream. The prosodic bootstrapping hypothesis has cap­ tured the imagination of many researchers in the field on the strength of its apparent plausibility and explanatory promise. We argue here, however, that the popularity of the prosodic boot­ strapping notion has proceeded far in advance of the data necessary 10 support it. Although a few critical voices have been raised (e.g., Pinker, 1987), there has been insufficient attention either to the logic of the argument or to the limitations of the data. Support for the prosodic bootstrapping hypothesis rests on a selective use of indirect evidence, and some of the central findings cited in its favor need further replica­ tion. Because this hypothesis is about how the child uses prosodic cues to induce grammatical rules, it can be directly tested only by manipulat­ ing the relation of syntactic and prosodic units in speech to the child and assessing the effects of these manipulations on language acquisition. In the absence of direct evidence, prosodic bootstrapping advocates rely on indirect evidence to argue that prosodic features in language input are operative in the child's induction of language structure. The force of the argument lies not in any particular finding, but in the broad sweep of what appears to be convergent evidence from diverse sources. Four major categories of indirect evidence are typically cited: