The methods developed from Brunswik's lens model were applied to a learning task in vector composition in which only an indication of the direction of the vectors was given. The feedback consisted of the direction of the resultant (criterion) vector. The relative lengths of the three stimulus vectors determined the direction of the resultant vector and thus could be considered as the relative weights to be attached to the stimulus dimensions. The task consisted of learning these three relative weights. There were three levels of task predictability. Half of the subjects got 140 learning trials followed by 80 transfer trials where the weights of the most and least important dimensions were reversed. The other half received 140 learning trials followed by 80 trials without any feedback. There was an indication for a two-stage learning process in which first the general structure of the task was learned after which the subjects stabilized their response strategy. Also there was a trend towards optimizing behavior; probability matching could not account for the results attained. The authors concluded that the judgmental processes studied are of an automatic, not clearly verbalizable, character. Whatever the quality of performance, verbal reports were only of a very global type. © 1972.
van Bolhuis-Bouma, A., & Oostlander, A. M. (1972). Multiple cue probability learning: An experiment with circular dimensions. Organizational Behavior and Human Performance, 7(3), 383–396. https://doi.org/10.1016/0030-5073(72)90024-4