| Abstract: | It was proposed that prediction hypotheses do not emerge full blown in the repertoire of the young child. Instead, the component units out of which predictions are consolidated (win-stay, lose-shift) are acquired independently and at different rates. The young child's behavior is at first dominated by response sets in which the win and lose components involve the same action (win-stay, lose-stay or win-shift, lose-shift). The components of response sets must be separated and coordinated with task-relevant information before appropriate components are consolidated to yield prediction hypotheses. A combination of learning-set and blank-trial procedures was used to evaluate this conception with preschoolers (mean CA = 4:1) and second graders (mean CA = 7:6). Results revealed that children of both age levels exhibited response patterns that corresponded to hypotheses in most of their probes (.89 to. 98). The striking finding was the relative weakness of win-stay on object cues at the outset of acquisition, particularly among the younger children. In contrast. lose-shift object was strong from the outset among children of both age levels. Once the two components (win-stay object, lose-shift object) were consolidated, the resulting scheme was durable, in that transfer to more complex problems was nearly perfect. Children who did not achieve criterion (in 70 problems) failed because they exhibited mostly position hypotheses and/or because the strength of win-stay object remained weak throughout all sessions. |
