✍ Scribed by Wolfgang Manz
No coin nor oath required. For personal study only.
Probability-matching in two-choice experiments has stimulated discussions about strategic behaviour of Ss'. Neimark and Shuford found that Ss in a two-choice situation estimated correctly the proportion of events to be predicted and did not use the suboptimal matching strategy. Does this finding generalize to situations with more than two choices ?
A probability learning experiment with six events was conducted with 20 Ss in each of four groups. The usual prediction method had to be changed into a successive guessing method to secure the same information feedback on each trial as in the two-choice situation. A 2 • 2 design was used to vary the probability of the most frequent event at two levels (p --0.33 and p = 0.50) and the distribution of the remaining events (equal proportions vs. decreasing proportions).
Ss' estimates of subjective probabilities did not differ from objective probabilities. Analysis of Ss' choices in the final block of 30 trials and an analysis of variance on the number of guesses indicated that no probability-matching did occur. Ss used the optimal strategy instead but failed to do so consequently enough to reach optimum performance scores. The results show that even with more than two events Ss are able to learn the probability distribution and effectively try to adopt the optimal strategy if the situation is a well defined one.
📜 SIMILAR VOLUMES