Development of reversal shift learning. An individual differences analysis

The existence of distinct modes in category learning is posited in several theories (F.G. Ashby et al. 1998), (S.Keri, 2003). In the initial phase of discrimination shift learning tasks (DSL), which is a subtype of category learning tasks, different modes of learning have been found (e.g., (K.K. Block et al., 1973), (M.E.J. Raijmakers, et al., 2001), (V.D. Schmittmann et al., 2006)). In the present paper, the existence and age-related changes of distinct modes in the reversal shift phase of a DSL task were examined, and the relation between the initial learning and the shift learning mode was analyzed. Hidden Markov models were employed to investigate the development in the cross-sectional sample of 4 to 20 year-old participants (M.E.J. Raijmakers, et al., 2001), (V.D. Schmittmann et al., 2006). The results revealed the existence of different shift learning modes in the sample, which were similar to the initial rational and slow learning modes. The probability of slow shift learning decreased with age. The response accuracy in the application of the reversed rule increased with age in slow and fast shift learners. The relation between slow and rational initial, and slow and fast shift learning modes showed age effects.