Modelling How Improvements in Relational Processing can account for Developmental Trajectories in Object Recognition Performance.

Recent experimental studies have shown that development towards adult performance levels in configural processing in object recognition is delayed towards the end of adolescence. Whilst part-changes to animal and artefact stimuli are processed with similar to adult levels of accuracy between 7 years of age and adulthood, relative size changes to stimuli result in a significant decrease in relative performance for 7 and 9 year old participants. This paper presents simulation results which provide an information processing explanation for these results. One set of computational experiments were run using the JIM3 artificial neural network with adult and ‘immature’ versions by progressively decreasing the number of neurons involved in the recognition of view-independent metric relations between geons. A second set of computional experiments involved decreasing the number of neurons in JIM3’s viewdependent Surface Map. The results showed that only in computational experiments where metric-precision in JIM3 was decreased did performance on relative size stimuli decrease, with significantly higher performance on part change stimuli. These results therefore demonstrate that of the parameters investigated, only changes in metric precision in size relations between geons in JIM3 produce a similar pattern of results to recent empirical findings.