Abstract
Recent studies show that speech errors involve the co-production of the phonetic properties of both targets and error outcomes. Based on the spatial and temporal properties of these co-productions, Pouplier and Goldstein argued that they are influenced by speech production mechanisms that detect and suppress errorful articulations. In this commentary, we provide simulation data supporting an alternative account based on cascading activation. Novel simulation results in the Gradient Symbol Processing framework show that in speech errors target and error phonological representations are gradiently co-activated. Using the TADA system, we show how these types of co-activation patterns could give rise to the articulatory patterns observed by Pouplier and Goldstein. These results show that cascading activation could provide a plausible account not only for the presence of co-productions in speech errors but also for their temporal and spatial properties.
Acknowledgements
This research was supported by the National Science Foundation (NSF) Grant BCS0846147 to Goldrick. Any opinions, findings and conclusions or recommendations expressed in this material are those of the authors and do not necessarily reflect the views of the NSF. Thanks to Honsung Nam and Louis Goldstein for assistance with the TADA system and Janet Pierrehumbert and the SoundLab for helpful discussion and comments.
Notes
1. As activation values in the simulation are designed to fall within (0, 1), this is given by 0.5 * [a[+voiced] – a[Øvoiced] + 1].
2. Specifically, relative to a prototypical voiceless stop we shorten the activation interval of the glottal gesture.
3. Specifically, relative to a full constriction we shorten the activation interval of the closure gesture and reduce the coordination angle between the onsets of the closure and release gestures.