Confidence and gradedness in semantic categorization: definitely somewhat artifactual, maybe absolutely natural.

Artifacts tend to be categorized in a graded (i.e., continuous) manner, whereas natural categorization tends to be absolute (i.e., discrete). This domain-specific categorization is assumed to reflect a domain difference in representation. However, another tenable but untested explanation is that graded categorization arises from uncertainty, which is greater in artifact categories than in natural categories. Confidence ratings were used as an index of certainty in two experiments that tested whether confidence in category judgments can explain the apparent gradedness of those categories. Both experiments revealed that artifact categories were more graded and were judged with greater confidence than were natural categories. Confidence and gradedness were negatively correlated within both domains. Thus, confidence did indeed predict gradedness within the artifact and natural domains but failed to predict the difference in gradedness between those domains. There is more to gradedness than just uncertainty.