Mechanisms of interaction in speech production

Abstract

Many theories predict the presence of interactive effects involving information represented by distinct cognitive processes in speech production. There is considerably less agreement regarding the precise cognitive mechanisms that underlie these interactive effects. For example, are they driven by purely production-internal mechanisms (e.g., Dell, 1986) or do they reflect the influence of perceptual monitoring mechanisms on production processes (e.g., Roelofs, 2004)? Acoustic analyses reveal the phonetic realisation of words is influenced by their word-specific properties – supporting the presence of interaction between lexical-level and phonetic information in speech production. A second experiment examines what mechanisms are responsible for this interactive effect. The results suggest the effect occurs on-line and is not purely driven by listener modelling. These findings are consistent with the presence of an interactive mechanism that is online and internal to the production system.

Keywords:

• Speech production
• Interactive processes
• Lexical-phonetic variation

Acknowledgements

This research was supported by National Institutes of Health grant DC 0079772. Portions of this work were presented at the 10^th Conference on Laboratory Phonology (Paris, France, 2006), the 3^rd International Workshop on Language Production (Chicago, IL, 2006), and the 48^th Annual Meeting of the Psychonomic Society (Long Beach, CA, 2007). We thank Arim Choi, Yaron McNabb, and Jaro Pylypczak for help in running the experiments, Eva Tomczyk and Kristin Van Engen for assistance in acoustic measurements and Ann Bradlow for extremely helpful discussions of this research.

Notes

¹The studies reviewed below assume a categorical definition of phonologically related – specifically, all words related by the addition, substitution, or deletion of a single phoneme (see Vitevitch, 2002, for additional discussion of this measure in the context of speech production). In contrast, spreading activation theories tend to assume a more graded definition of neighbourhood structure – where various non-target words are partially activated based on their degree of phonological overlap with the target (see Goldrick and Rapp, 2001, for further discussion).

²Note that Goldrick and Rapp (2007) failed to find any influence of lexical variables on categorical measures of phonetic processing (accuracy and categorical properties of error outcomes). It is possible that interaction between these processes is limited such that interactive effects are limited to subcategorical modulations of fine-grained phonetic aspects of production.

³As noted by a reviewer, the phonetic variation reported in this study (5% of VOT) is substantially smaller than that typically examined in the perceptual studies cited above. Although no study has examined the perceptual consequences of VOT variation of the specific type and magnitude we report here, we believe that the substantial body of results cited above make the perceptual monitoring account a plausible hypothesis. (Although, as shown by our results, ultimately incorrect; see below.)

⁴Three participants were run with two fillers in the /t/-initial list that rhymed with targets in the /k/-initial list; these were replaced for subsequent participants.

⁵It should be noted that this analysis is slightly unconventional as it combines across subjects from two different experiments.

⁶As noted with the previous analysis, this analysis is unconventional as it combines subjects across two experiments. Additionally, the mean VOTs for participants in Experiment 1b collapse across both the /t/- and /k/-initial lists, whereas the mean VOTs for participants in Experiment 1a are from the /p/-initial list alone.

⁷It should be noted that, although we only compared items used in both Experiments 1 and 2 in our analysis, the two experiments were not strictly comparable as the experimental context differed. Therefore, the comparison should not be considered conclusive.

⁸Other instances of phonetic variation seem to be hybrids of listener modelling and more automatic processes. For example, Fowler and Housum (1987) examined second mention reduction. When speakers produce a word for the second time in a discourse it is shorter and less intelligible than the first mention. While this is sensitive to the communicative context, it is clear that it is a more automatic process than clear speech – which appears to be more under talker control. Similarly, the more general effects of articulatory reduction in predictable speech contexts (see Aylett and Turk, 2004, for a recent review) also appear to be due to more automatic processes that may or may not reflect explicit listener modelling.