Using numerosity judgments to determine what is learned during automatization

An important question in learning is the nature of the information required to support skilled, or automatic, performance. In 2 experiments, participants counted patterns of 8-10 objects displayed on a computer screen for 4 sessions of 240 trials each before being transferred to a different set of patterns for a final session of 240 trials. The patterns in the final session differed from those seen in the 4 training sessions in either overall configuration (Experiment 1) or identity of constituents (Experiment 2). Results indicated that both types of information are important in learning and automatization of a counting task and support the idea that what is attended to during training will be necessary to support subsequent automatic performance.