What if you are not sure? Electroencephalographic correlates of subjective confidence level about a decision


Creative Commons License

Selimbeyoglu A., Keskin-Ergen Y., Demiralp T.

CLINICAL NEUROPHYSIOLOGY, cilt.123, ss.1158-1167, 2012 (SCI İndekslerine Giren Dergi) identifier identifier identifier

  • Cilt numarası: 123 Konu: 6
  • Basım Tarihi: 2012
  • Doi Numarası: 10.1016/j.clinph.2011.10.037
  • Dergi Adı: CLINICAL NEUROPHYSIOLOGY
  • Sayfa Sayıları: ss.1158-1167

Özet

Objective: Action monitoring has been studied in terms of EEG correlates of accuracy/inaccuracy. EEG signals related to the confidence level (uncertainty) of the subject about given responses remain to be clarified.

Objective: Action monitoring has been studied in terms of EEG correlates of accuracy/inaccuracy. EEG signals related to the confidence level (uncertainty) of the subject about given responses remain to be clarified. 

Methods: Two discrimination tasks that create difficulty in stimulus processing and response selection stages of decision making were used. Conditions were certain-hit (hit trials where subjects were certain), certain-error (trials where subject was aware of the error), uncertain (trials where subject was uncertain). Analyzes of event-related potentials (ERP) (ERN, Pe, P3) in the time domain were extended by time–frequency analysis of both phase-locked and non-phase-locked oscillatory activities.

Results: Errors caused by difficulty in stimulus processing led more uncertainty than the ones related to the response selection, and tended to elicit ERN with shorter latency. Each condition was discriminated from the others by at least one parameter. Uncertainty was discriminated by total theta activity following response, total delta activity following stimulus and mean P3 amplitude.

Conclusions: Subjective confidence levels of decision making can be discriminated by time and time–frequency analyses of ERPs and EEG.

Significance: Time–frequency parameters of EEG can be useful in the detection of subjective confidence in single-trials, which might be efficiently used for faster human–computer interaction.