v2 test was performed with each training task to compare the predicted and observed accuracy distributions. For participants trained with direct classification, the predicted and observed accuracies were similar [v2(2) = 0.36, n.s.],suggesting that participants could use the knowledge acquired during a direct classification task to perform the same–different categorization task. This is in line with the decreased performance when the participants switched to the same–different categorization task. In contrast, participants trained with the same–different categorization task
did not seem to use their category knowledge when transferred to the direct classification task [v2(2) = 17.05,p.001]. In all three conditions, the participant performance should have improved substantially when transferred to the direct classification task (because only one categorization judgment is required). The previous analyses showed that their performances did not statistically
change when transferred to the direct classification task. A third measure of transfer is to ask whether accuracy in the first transfer block (Block 15) is the same as the firstblock performance of other participants who were originally trained with the same task and the same category structure (e.g., compare the first transfer block of Cat-II
with the first training block of SD-II). If there was transfer, the first block of the transfer task should have been better than the first block of the corresponding training task. These accuracy differences are shown in Table 2 (transfer gain). Independent sample t tests were performed in each condition. Only the group performing direct classification with the conjunction rule showed evidence of transfer
[t(28) = 4.69, p.001]. The same–different categorization task was easier after previous training in direct classification with the conjunction rule. All other conditions showed no sign of positive or negative transfer [all |ts(28)|1.51, n.s.].3