Are there interactive processes in speech perception?

Abstract

Lexical information facilitates speech perception, especially when sounds are ambiguous or degraded. The interactive approach to understanding this effect posits that this facilitation is accomplished through bi-directional flow of information, allowing lexical knowledge to influence pre-lexical processes. Alternative autonomous theories posit feed-forward processing with lexical influence restricted to post-perceptual decision processes. We review evidence supporting the prediction of interactive models that lexical influences can affect pre-lexical mechanisms, triggering compensation, adaptation and retuning of phonological processes generally taken to be pre-lexical. We argue that these and other findings point to interactive processing as a fundamental principle for perception of speech and other modalities.

Figure I

Architecture of the TRACE model. Bi-directional excitatory connections are shown in red: mutually consistent elements at adjacent levels support each other through excitation. Units within a layer compete through inhibitory connections (blue; the full set is shown for the lexical layer, for clarity only a schematic connection is shown at the phoneme level).

Figure 1

Schematic representations of information flow in interactive and autonomous models of speech perception. (a) An interactive model posits bi-directional excitatory connections between processing levels with phoneme-level responses produced at the pre-lexical processing level. Units within each layer compete through mutually inhibitory connections. (See Box 1 for architectural details of the interactive TRACE model.) (b) An autonomous model of the sort advocated in [4] posits strictly feedforward excitatory connections from pre-lexical processing to word processing and a separate phoneme identification layer that integrates inputs from phoneme and word processing layers. Units in the word processing and phoneme identification layers compete through mutually inhibitory connections. In both panels red arrows indicate excitatory connections and blue curves indicate inhibitory connections.

Figure 2

Acoustically and lexically mediated compensation for coarticulation. (a) and (b) show data from simulations of the TRACE model, and (c) and (d) show behavioral data from human listeners. Panels (a) and (c) show that an ambiguous stop will be identified as /k/ more often when it is preceded by a clear /s/ sound than when it is preceded by a /∫/ sound (acoustically mediated compensation for coarticulation). Panels (b) and (d) show that this effect persists (albeit more weakly) when the fricative is replaced by an ambiguous one and lexical context is manipulated. That is, a lexically defined fricative causes a shift in identification of neighboring sounds (lexically mediated compensation for coarticulation). This lexically mediated compensation for coarticulation effect requires that lexical influences act directly on phoneme processing rather than on a post-perceptual decision stage. Data replotted with permission from Ref. [18].