Abstract: In this article, we compare classification learning with inference learning. In the inference task, participants predict the value of a missing feature of an item given its category label and other feature values. In the classification task, participants predict the category label of an item given its feature values. Yamauchi and Markman (1998, 2000, 2002) showed that these two types of learning did not result in the learning of equivalent knowledge. Categories defined by a family resemblance structure were more easily learned by inference learning than by classification learning, whereas categories defined by a nonlinearly separable structure were more easily learned by classification learning than by inference learning. Chin-Parker and Ross (2002, 2004) found that Classification learners were highly sensitive to diagnostic features but not sensitive to nondiagnostic, prototypical features. Inference learners were less sensitive to the diagnostic features than were classification learners and were also sensitive to the nondiagnostic, prototypical features. In the current experiments, we examined the sensitivity of classification and inference learners to another critical type of category information- feature correlation information. We further systematically explored the two learning tasks that might lead to differential learning efficiency, strategy and outcome.
Method
As Chin-Parker & Ross (2002，2004) pointed out, Yamauchi & Markman ignored irrelative transfer effects in their 1988 study and used less learning exemplars in their 2002 study. We improved Yamauchi & Markman’s experiment (2002) by designing 10 learning exemplars and 3/5 prototypical typicality category structure, using learning-transfer task paradigm and feature category detecting method. We explored the category feature correlation processing in different way in classification and inference learning. One hundred and forty-four volunteers, who took an introductory psychology course for partial credit at South China University of Technology, took part in the experiments (ninety-six for other two complementary experiments).
Results
For participants who reached the 90% learning criterion, we found that those in the inference learning condition (M = 34.33) were required significantly more blocks than those in the classification learning condition (M = 17.84). For participants given classification learning, we found that they classified transfer stimuli A6~A7 more accurately than they did the transfer stimuli B6~B7. In contrast, participants given inference learning did not show this trend.
The study also showed that classification learning was better than inference learning to catch the category prototype. At inference transfer phase, classification learners were more likely to infer the absent feature as the prototypical feature than inference learner did. At single feature classifying, classification learners showed more accurately than inference learner did.
Conclusions
Inference learning was easier than classification learning with accepting feature correlation information; classification learning was more efficient than inference learning in reaching the learning criterion and switching strategies. Classification learning was better than inference learning in integrating prototype although classification learning was opt to exemplars memorizing comparing inference learning to prototype memorizing

Key words: category learning, classification, inference, prototype memorizing, exemplars memorizing