P300 and uncertainty reduction in a concept-identification task.

The relationship between the amplitude of P300, the mean amplitude of the Slow Wave, and uncertainty reduction after (dis)confirmation of hypotheses was studied in a Concept-Identification task. The subjects had to categorize stimuli according to a conceptual mIe (joint denial or exclusion) and to rate the confidence that their classification was correct. Three types of feedback were distinguished: confirming (subject's cate-gorization was correct), disconfirming (subject's categorization was incorrect), and non-informative feedback. The EEG was averaged separately according to the three types of feedback and the two confidence ratings (low, high). The data showed the predicted interaction between type of feedback and confidence level. A larger P300 amplitude tumed up after confirming feedback when the subject was less confident, than when he was more confident. The reverse was found after disconfirming feedback. The P300 amplitude after non-informative feedback was not influenced by confidence. The mean amplitude of the Slow Wave showed approximately the same interaction pattem. The results were interpreted in terms of changes in the probability of hypotheses which subjects use to categorize stimuli in a Concept-Identification task. This thesis implies that a sub-model ofthe orienting reflex to evaluate the quanti-ject selects a hypo thesis and classifies stimuli ac-tative change in the probabilities of the neural cording to this hypothesis. If the classification tums modeis. This change resulted from a combination of out to be incorrect (disconfirmation) the subject re-information from the outside world and the prior jects the hypothesis and selects another Doe. In case probabilities of the modeis. De Swart and Das-of correct classifications (confirmations) the subject Smaal (1976, 1979a, 1979b) argued that Sokoiov's stays with his hypothesis. Many studies have shown model is a neural analogue of the cognitive process that subjects perform according to this strategy 're-of hypothesis testing. The authors (1979b) found ject your hypothesis only after disconfirmation' that the amplitude of the Skin Conductance Re-firmed a low-confidence classification than a high-Bayes' theorem offers a model to describe the confidence classification. Disconfirmation revealed change in subjective probabilities after confirma-the opposite result: SCR was higher after a high-tion and disconfirmation in a deterministic Concept-confidence classification than after a low-confidence Identification task, given the strategy 'reject your classification. hypothesis only after disconfirmation.' A vast number of studies demonstrated that the Several Concept-Identification studies revealed amplitude ofthe late positive component (P300) of data which are in agreement with the Bayesian pre-the cortiçal event-related brain potential (ERP) is dictions …