engleski

The influence of category learning condition on metacognitive monitoring of categorization tasks

Category learning can be an explicit process when the rule that defines category membership is available, or an implicit process of slow associating members to categories when the rule is not known. The empirical question is whether the implicit process of learning categories can be facilitated by explicitly informing participants to direct their attention on relevant aspects of the category members in situation when the correct rule is not discovered. In this study, we examined whether classification accuracy and metacognitive monitoring of that accuracy reflect the category learning condition. Participants were divided into four groups that learned complex categories in which category membership was defined by a three-dimensional rule. The stimuli were arrows presented on a background. Three relevant dimensions with respect to categorization rules were arrow orientation, arrow color, or arrow numerosity. The background color and shape were irrelevant dimensions. In the course of learning, the first group was informed about the correct rule, the second group was instructed to direct the attention to the arrows and to ignore the background, the third group was informed about three relevant imensions, and the fourth group was not given any information. In the following prediction phase, participants were asked to predict the likelihood that they could correctly classify novel exemplars in the subsequent transfer phase at the category level by giving category learning judgments (CLJs). Finally, in the transfer phase, participants were asked to classify stimuli that varied in similarity to exemplars observed in the learning phase (old vs. similar vs. dissimilar) and to give confidence judgments about each classification. The analyses were performed only for participants who acquired complete accuracy during learning. The results of the prediction phase showed that CLJs were not sensitive to learning conditions (F(3, 41)= 1.94, p>.05). More precisely, participants who successfully learned correct category memberships predicted the likelihood that they could correctly classify novel exemplars in the transfer phase to be moderate (Ms< 72%). As expected, in the transfer phase the results revealed significant learning condition x item similarity interactions for both accuracy (F(6, 82)=3.33 ; p<.01) and confidence (F(6, 82)=10.31, p<.001). Informing participants about relevant dimensions in three learning conditions resulted in equally high accuracy (Ms< 82%) and confidence (Ms<86%) for three levels of item similarity. Conversely, accuracy and confidence decreased as a function of item similarity when participants were not given any information (ps<.05, Accuracy: Mold=92.92%, Ms=82.42%, Mds=74.17% ; Confidence: Mold=92.41%, Ms=86.6%, Mds=66.17%). The obtained results suggest that once attention is directed toward relevant dimensions, even when participants are not explicitly informed about the correct rule, categorization performance can be facilitated.

category learning ; learning conditions ; classification accuracy ; metacognitive monitoring ; item similarity

nije evidentirano