Transient attention enhances perceptual performance and FMRI response in human visual cortex

Abstract

When a visual stimulus suddenly appears, it captures attention, producing a transient improvement of performance on basic visual tasks. We investigate the effect of transient attention on stimulus representations in early visual areas using rapid event-related fMRI. Participants discriminated the orientation of one of two gratings preceded or followed by a nonpredictive peripheral cue. Compared to control conditions, precueing the target location improved performance and produced a larger fMRI response in corresponding retinotopic areas. This enhancement progressively increased from striate to extrastriate areas. Control conditions indicated that the enhanced fMRI response was not due to sensory summation of cue and target signals. Thus, an uninformative precue increases both perceptual performance and the concomitant stimulus-evoked activity in early visual areas. These results provide evidence regarding the retinotopically specific neural correlate for the effects of transient attention on early vision.

Figure 1. Experimental Design

The sequence of events in the precue and postcue trials is illustrated. Note that the onset of the Gabor stimuli within a trial is identical for the precue and postcue trials; both are 100 ms after trial onset. For the purpose of illustration, the Gabor stimuli are shown at a contrast of 50%, and the tilted Gabor stimuli are oriented ±10°.

Figure 2. Representative Results from the Localizer Scans

The diagram on the left illustrates the locations of the cue and the Gabor stimulus, which were presented in alternating blocks. Shown on the right are results from the right hemisphere of one participant, viewed on inflated surface representation of the posterior occipital cortex. Light and dark gray depict gyral and sulcal surfaces, respectively. Brain activity associated with the cue and brain activity associated with the Gabor stimulus are shown in blue and green maps, respectively. Black lines indicate the borders of early visual areas defined by the retinotopic mapping procedure (solid line, vertical meridian; dashed line, horizontal meridian). These borders were derived by using a wedge stimulus encompassing 0.25°–8.25° eccentricity, and the Gabor stimulus covered 4°–8° eccentricity. The asterisk indicates the foveal confluence where borders between areas cannot be resolved. At this statistical threshold (p < 0.001, uncorrected for multiple comparisons), the activation of the cue and the Gabor did not overlap in V1, V2, and V3. Activity started to overlap in V3a and hV4, as they contain a hemifield representation (Tootell et al., 1997; Wade et al., 2002).

Figure 3. Group-Averaged Data

(A) Behavioral results. Proportion correct (left) and reaction time (right) are shown for the four cue conditions (V-Pre, valid precue; I-Pre, invalid precue; V-Post, valid postcue; I-Post, invalid postcue). Error bars are 1 SEM. (B) Imaging results. Mean fMRI responses across participants for each cue condition and distracter are shown for each visual area. Response was obtained from the dorsal (V1, V2, V3, and V3a) and ventral (hV4) representations of the target (the green areas in Figure 2). The average standard error of all time points along a curve is shown as the error bar on the first time point.

Figure 4. Individual Data

Representative data from two individual participants—a trained psychophysical observer (FP) and a naive observer (KM). For each participant, the top row shows behavioral results, and the middle and bottom rows show imaging results for different visual areas. (For details, see the legend of Figure 3.)

Figure 5. Peak fMRI Response

Peak amplitude of the fMRI response for different trial types in all visual areas (legends identical to Figure 3). Error bars are 1 SEM.

Figure 6. Magnitude of Attentional Effect

Mean attention modulation index (AMI) across participants for different visual areas [AMI = (Peak_{valid precue} − Peak_baseline)/(Peak_{valid precue} + Peak_baseline), where Peak_baseline = average of the peak amplitude for the invalid precue, valid postcue, and invalid postcue conditions]. Error bars are 1 SEM.