Predicting rhesus monkey eye movements during natural-image search

Journal of vision

Emory Kuo¹  Sara Caddigan¹  Emily A. Berthiaume¹  Mark A. Segraves¹  Konrad P. Kording² 

[1] Department of Neurobiology, Weinberg College of Arts and Sciences, Northwestern University, Evanston, IL, USA;Departments of Physical Medicine and Rehabilitation and Physiology, Feinberg School of Medicine, Northwestern University, Chicago, IL, USA

关键词: eye movement;    cercopithecidae;    macaca mulatta;    saccades;    pedestrian;    natural scenes;   

DOI  :  10.1167/17.3.12

学科分类：眼科学

来源: Association for Research in Vision and Ophthalmology

【 摘 要 】

There are three prominent factors that can predict human visual-search behavior in natural scenes: the distinctiveness of a location (salience), similarity to the target (relevance), and features of the environment that predict where the object might be (context). We do not currently know how well these factors are able to predict macaque visual search, which matters because it is arguably the most popular model for asking how the brain controls eye movements. Here we trained monkeys to perform the pedestrian search task previously used for human subjects. Salience, relevance, and context models were all predictive of monkey eye fixations and jointly about as precise as for humans. We attempted to disrupt the influence of scene context on search by testing the monkeys with an inverted set of the same images. Surprisingly, the monkeys were able to locate the pedestrian at a rate similar to that for upright images. The best predictions of monkey fixations in searching inverted images were obtained by rotating the results of the model predictions for the original image. The fact that the same models can predict human and monkey search behavior suggests that the monkey can be used as a good model for understanding how the human brain enables natural-scene search.

【 授权许可】

CC BY