Abstract
Purpose: To quantitatively relate fMRI activity in early visual areas to psychophysical responses during a yes-no threshold pattern-discrimination task. Methods: Subjects performed a rapid sequence of 2-s-duration trials each consisting of a 1-s-duration stimulus followed by a response period. A background mask pattern was presented on every trial. On 1/6th of the trials, a low-contrast target grating was added to the mask. Subjects responded by pressing one of two buttons to indicate the presence or absence of the target.Target contrast was set to each subject's threshold (d′ ∼ 1). Thus, the majority (∼70%) of the trials corresponded to correct rejections. Subjects were trained to adjust their criterion so that the remaining trials were approximately evenly divided between hits, false alarms, and misses. In separate experiments, we used two different masks, a 20%-contrast plaid or a high-contrast white-noise pattern. Functional images were recorded at a 1-s frame rate while each subject performed the task. fMRI time series were spatially averaged over voxels in several areas of early visual cortex (V1, V2, V3). Correct-rejection trials were used as a baseline. Results: In all subjects and visual areas, cortical activity was greater for hits than for the baseline correct rejects. More interesting, cortical activity was greater for false alarms than misses, contrary to the magnitude of the physical contrast in the stimulus. The amplitude of these effects was about a factor of two larger for the noise-mask stimulus than for the plaid-mask. Conclusions: Neural correlates of threshold visual perception are evident in human visual cortex. The observed activity could correspond to intrinsically noisy sensory signals, or to feedback signals from higher-level areas corresponding to the subjects' behavioral choices.
Supported by grants from the NEI and HFSF.