What Children’s Pause Patterns Indicate about their Constituent Structure

In the Bock & Levelt (1994) model of language production, trees are conceptualized as ‘control hierarchies’ that coordinate the insertion of lexical material into slots once the material is retrieved and assembled. This mainstream view of production has encouraged researchers to regard the acoustic correlates of prosody, such as final lengthening and pausing, as reflective of the existence of higher order linguistic structures or units (Cooper & Paccia-Cooper, 1980; Ferreira, 1988; Selkirk, 1984; Wagner, 2005; Watson & Gibson, 2004). For example, Cooper & Paccia-Cooper (1980) used measures of final lengthening to argue that language formulation plans make use of certain syntactic constituents, as opposed to others. They also argued that pause duration positively correlates with syntactic boundary strength, and that the longer pauses found at major constituent (clause) boundaries result from “processing fatigue” brought on by the demands of producing the previous syntactic sub-constituents. Other studies have examined speech for evidence of constituent movement and gaps posited by Transformational Grammar (Chomsky, 1965). For example, Ferreira (1988) showed that the presence of a gap in surface structure affects whether a preceding word is reduced or lengthened. She argued that a trace blocks the normal process of vowel reduction in words like to in sentences like The boyi was spoken to (ti) by his teacher. Central to such theories of language production is the view that grammar is hierarchically organized, and that the evidence from speech reflects this organization. Traditional structuralist as well as Generative theories of syntactic structure (e.g. Hockett, 1958; Chomsky, 1957; 1965; Kayne, 1994) provide models of the hierarchical organization of constituents. These models allow for specific hypotheses about where to pause in an utterance (and, perhaps, for how long). As Radford (1988) puts it,