Detecting self-produced speech errors before and after articulation: an ERP investigation

Abstract

It has been argued that speech production errors are monitored by the same neural system involved in monitoring other types of action errors. Behavioral evidence has shown that speech errors can be detected and corrected prior to articulation, yet the neural basis for such pre-articulatory speech error monitoring is poorly understood. The current study investigated speech error monitoring using a phoneme-substitution task known to elicit speech errors. Stimulus-locked event-related potential (ERP) analyses comparing correct and incorrect utterances were used to assess pre-articulatory error monitoring and response-locked ERP analyses were used to assess post-articulatory monitoring. Our novel finding in the stimulus-locked analysis revealed that words that ultimately led to a speech error were associated with a larger P2 component at midline sites (FCz, Cz, and CPz). This early positivity may reflect the detection of an error in speech formulation, or a predictive mechanism to signal the potential for an upcoming speech error. The data also revealed that general conflict monitoring mechanisms are involved during this task as both correct and incorrect responses elicited an anterior N2 component typically associated with conflict monitoring. The response-locked analyses corroborated previous observations that self-produced speech errors led to a fronto-central error-related negativity (ERN). These results demonstrate that speech errors can be detected prior to articulation, and that speech error monitoring relies on a central error monitoring mechanism.

Keywords: conflict monitoring; error-related negativity; event-related potentials; speech errors; speech production.

Figure 1

Stimulus-locked, grand averaged waveforms for electrode sites: FCz, Cz, and CPz. Figure shows the average waveform for correct (dashed line) and incorrect (solid line) responses (panels on left side). The P2 component can be observed between 200 and 275 ms post-stimulus. The panels on the right show the averaged amplitudes within the P2 window, with standard error bars. The N2 can be observed between 275 and 375 ms post-stimulus.

Figure 2

Response-locked, grand averaged waveforms for fronto-central electrode sites: FCz and Cz. Figure shows the average waveform for correct (dashed line) and incorrect (solid line) responses (panels on the left side). The ERN can be observed after the response, peaking at ~75 ms post-error response. The panels on the right side show the averaged amplitudes within the 50–100 ms post-response interval, with standard error bars.