Error awareness as evidence accumulation: effects of speed-accuracy trade-off on error signaling

Abstract

Errors in choice tasks have been shown to elicit a cascade of characteristic components in the human event-related potential (ERPs)-the error-related negativity (Ne/ERN) and the error positivity (Pe). Despite the large number of studies concerned with these components, it is still unclear how they relate to error awareness as measured by overt error signaling responses. In the present study, we considered error awareness as a decision process in which evidence for an error is accumulated until a decision criterion is reached, and hypothesized that the Pe is a correlate of the accumulated decision evidence. To test the prediction that the amplitude of the Pe varies as a function of the strength and latency of the accumulated evidence for an error, we manipulated the speed-accuracy trade-off (SAT) in a brightness discrimination task while participants signaled the occurrence of errors. Based on a previous modeling study, we predicted that lower speed pressure should be associated with weaker evidence for an error and, thus, with smaller Pe amplitudes. As predicted, average Pe amplitude was decreased and error signaling was impaired in a low speed pressure condition compared to a high speed pressure condition. In further analyses, we derived single-trial Pe amplitudes using a logistic regression approach. Single-trial amplitudes robustly predicted the occurrence of signaling responses on a trial-by-trial basis. These results confirm the predictions of the evidence accumulation account, supporting the notion that the Pe reflects accumulated evidence for an error and that this evidence drives the emergence of error awareness.

Keywords: error awareness; error positivity; error-related negativity; event-related potentials; performance monitoring; single-trial analysis.

Figure 1

Sequence of stimulus events in a trial. Participants were first required to indicate which of two boxes in the stimulus was brighter. Following the error prompt, they pressed a signaling key if they judged that their primary task response was an error.

Figure 2

Response-locked ERPs separately for the highSP and lowSP conditions. (A,C) Mean ERP waveforms at electrodes FCz and POz for all errors and correct responses. (B,D) Mean ERP waveforms at electrodes FCz and POz for signaled errors and unsignaled correct responses. Zero indicates the time of the response. HighSP = high speed pressure. LowSP = low speed pressure.

Figure 3

Spatial distribution of ERPs for the difference between errors in the lowSP condition and errors in the highSP condition. Upper row: Time period of the Ne/ERN (−50 – 50 ms). Lower row: Time period of the Pe (150–400 ms). Left column: Data from all errors. Right column: Data from signaled errors. HighSP = high speed pressure. LowSP = low speed pressure.

Figure 4

Response-locked ERPs for RT matched data separately for the highSP and lowSP conditions. (A,C) Mean ERP waveforms at electrodes FCz and POz for all errors and correct responses. (B,D) Mean ERP waveforms at electrodes FCz and POz for signaled errors and unsignaled correct responses. Zero indicates the time of the response. HighSP = high speed pressure. LowSP = low speed pressure.

Figure 5

Spatial distribution of ERPs in RT matched data for the difference between errors in the lowSP condition and errors in the highSP condition. Upper row: Time period of the Ne/ERN (−50 – 50 ms). Lower row: Time period of the Pe (150–400 ms). Left column: Data from all errors. Right column: Data from signaled errors. HighSP = high speed pressure. LowSP = low speed pressure.

Figure 6

Error signal extracted by single-trial analysis. (A) Spatial distribution of the error signal as illustrated by normalized coupling coefficients. (B) Mean error signal for signaled (Sig) and unsignaled (NoSig) errors.