Computational modelling reveals the influence of object similarity and proximity on visually guided movements.

The paper aims to understand how humans reach for a single target object in multi-object scenes. In a previous empirical study, human subjects were asked to execute reaches to a single target among non-targets (choice reaching task). In the current work, we re-analysed the human data and implemented a neurobiologically-plausible cognitive robotics model (choice reaching with a LEGO arm robot, CoRLEGO) that mimics human reaches in the choice reaching task. The results from the experiment confirmed the commonly made assumption that proximity and similarity between objects (also termed perceptual grouping) affect the quality of the reaches. However, what was novel here was that modelling the reaches also allowed to temporally separate these factors, as the start of the movement was affected by both factors while the reach trajectory was affected only by proximity between target and distractor objects indicating that human information processing of visual stimuli applies these factors in a serial fashion. In particular, our model architecture and the optimised parameter settings suggest that object proximity directly influences the movement onset. In addition, our computational model confirmed this interpretation but also revealed that the relationship between the two factors may be affected by how the participants balanced speed (starting time of the movement) and accuracy of reaching (straightness of reaches). Future research will need to test whether this plausible prediction is correct.