Initial instruction in identifying medical disorders commonly takes the form of feature lists. Terms in these lists need to be grounded (related to perception) on two levels to function as intended. (1) General language: The initial reference of a term has to be sufficiently general to help a competent speaker. (2) Concept specific: The normal application of a term has to be concept specific to account for the person’s judgments of appropriateness of reference. The knowledge that enables concept specific grounding is memory for the various ways the feature can look knowledge of instantiated features. In decision making, the weighting of these two levels of terms, the general informational level and the more specific instantiated level, change with conditions. The learning involved in this adjustment seems to show some of the same cue interaction and blocking relations discussed in Lorraine Allen’s talk.
The metacognitive side of this development of proficiency is that students have to learn to use the “rules” appropriately. Despite the official and normative status of the stated rules, they cannot be treated as rules in a formal system. They provide the foci of attention for perceptual learning, not sufficient criteria for diagnosis. The diagnostic performance of beginners improves when instructed to “initially trust your sense of familiarity and then check for the full list of features.” Diagnostic performance of experts falls if they are initially given a list of all the features (that they subsequently admit are present) in a case. Processing as if the case were a list of individual features is a disadvantage for both experts and beginners. This relation between description and practice may be common in education.