You are here: Home Events Self-monitoring and control in speech production Program & Abstracts Conflict monitoring in speech production

Self-monitoring and control in speech production -

Conflict monitoring in speech production

Robert Hartsuiker - Department of Experimental Psychology, University of Ghent, Belgium

Speakers can monitor their own speech for errors and other problems by listening to their own overt speech. However, there is considerable evidence that they can also inspect an internal, pre- articulatory representation of speech. Theories differ in their accounts of such internal monitoring. According to the perceptual loop theory (e.g., Levelt, 1989), internal monitoring involves the speech perception system: Speakers would “listen” to their internal speech similarly to how they listen to their external speech. A more recent account (Nozari, Dell, & Schwartz, 2011) argues for internal monitoring based on conflict between competing representations within the language production system itself. Nozari et al. showed in simulations with their two-step activation model of word production that correct productions (e.g., saying cat when cat is intended) are characterised by less conflict than speech errors (e.g., saying dog, rat, or can when cat is intended); this was true with conflict measures based on the activation of all units (e.g., a function of the standard deviation of the activation of all nodes) or of only the two most active units (e.g., a function of the difference between these activations). Nozari et al.’s model, however, does not specify how conflict, once detected, is used. In this talk, we will first discuss the extent to which the literature supports a conflict-monitoring account vs. a perception-based account. Next, we will propose an extended version of Nozari et al.’s model that makes explicit assumptions about the processes subsequent to conflict detection. Taking our cue from the (behavioural and cognitive neuroscience) literature on conflict monitoring in action, we suggest a monitoring system that, upon detection of conflict, aselectively boosts the spreading rate for all connections, thereby steering the production system in the correct direction. We show that this steering mechanism can greatly reduce error rate and that it results in comparable performance as a mechanism that selectively boosts the input layer (i.e., semantic features). We therefore conclude that our aspecific conflict-based intervention mechanism holds promise as an internal mechanism for preventing and correcting errors.

Last checked 2016-06-28 by Zeshu

Street address
Wundtlaan 1
6525 XD Nijmegen
The Netherlands


Mailing address
P.O. Box 310
6500 AH Nijmegen
The Netherlands

Phone:   +31-24-3521911
Fax:        +31-24-3521213
E-mail:   


Public Outreach Officer
Charlotte Horn