Abstract
Spatial attention can target specific locations in the visual field to enhance processing of stimuli within that location. The top-down control of spatial attention is thought to be mediated via signals projecting from prefrontal cortex (PFC) to areas of early visual cortex. However, PFC neurons typically have relatively large spatial receptive fields and thus it is not clear how they guide highly precise attentional modulations in early visual areas. Here, we used fMRI to test the hypothesis that broad top-down spatial attentional signals only achieve high precision after being combined with precise bottom-up stimulus drive (N=5, two sessions per subject). On each trial, subjects were briefly shown (500ms) either a "focused" or a "diffuse” attention cue indicating the possible location of a subsequently presented target. The target was an orientated Gabor rendered at one of 5 contrasts between 0-50% that was flashed briefly (150ms) at one of 12 equidistant locations in the periphery. The cue and the target were separated by dynamic filtered noise presented across all 12 target locations, over a variable delay period (2 or 6-8s), and the contrast of the filtered noise matched that of the target to provide variable amounts of bottom-up sensory drive. Subjects reported whether the orientation of the Gabor was closer to horizontal or vertical. The fMRI results show that the average univariate responses increased with higher contrast, but that the fidelity of multivariate representations for spatial positions stayed largely constant across contrast levels. These results place constraints on possible interactions between top-down and stimulus driven signals in determining the size and specificity of attentional modulations.