Purchase this article with an account.
George Sperling, Ian Scofield, Arvin Hsu; Computational model of the spatial resolution of visual attention. Journal of Vision 2008;8(6):396. doi: 10.1167/8.6.396.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
The resolution limits of visual attention are conceptualized as being determined by an attention spread function. The parameters of the attention spread function are derived from a search experiment. Subjects search attended areas of a 12×12 array for a single target disk (a large disk) among 133 distracters (small disks) and 12 false targets (large disks in unattended areas) that force subjects to confine attention to the to-be-attended locations. To derive the parameters of the attention spread function, to-be-attended locations are defined as rows or columns interleaved with to-be-unattended rows or columns, analogous to spatial frequency gratings. Two model parameters characterize the width and shape of the attention spread function. One parameter characterizes task difficulty (how discriminable targets are from distracters), three parameters describe the asymmetric decline of spatial acuity with eccentricity. Once these 6 parameters have been determined from experiments with attention confined to rows or columns, the model makes predictions (with no new measurements) for all of the 1.5*1042 possible ways of requesting subjects to attend to 72 of 144 locations. To test the model, the 12×12 stimulus was conceptually divided into 16 3x3 blocks. Eight blocks were arbitrarily chosen to define to-be-attended areas. In a still more complicated task, subjects were required to attend to an arbitrarily chosen 18/36 blocks. The model made 288 good a priori predictions for one subject (not merely fits to data) when attending both 8/16 blocks and 18/36 blocks. However, the other subject, despite extensive practice, could attend to only slightly more than half of the to-be-attended blocks in these complex arrays; predictions within the attended subset were in accordance with the model. Conclusion: A simple spread function adequately describes the resolution of spatial attention but, when extremely complex distributions of attention are requested, complexity itself becomes a limiting factor.
This PDF is available to Subscribers Only