The fact that the pupil constricts not only to increases in stimulus luminance, but also to a range of stimuli even when there is no change in luminance, makes it a potentially valuable method of assessing visual function. Thus, it responds to achromatic gratings depending on their spatial frequency and contrast, to colored stimuli, depending on wavelength, and to movement (Barbur & Forsyth, 1986;Barbur & Thomson, 1987). These properties make it especially useful for assessing visual function in the blind fields of human and animal hemianopes and, indeed, in any situation where the verbal demands of psychophysical judgments may be fraught, as in blindsight, where subjects are asked to ''guess'' about stimuli they do not ''see,'' or where verbal instructions are impossible, as with animals or human infants.
It is now known that the pupil constricts to the presentation of achromatic gratings and to colored stimuli in the blind hemifields of human and monkey hemianopic subjects (Weiskrantz, Cowey, & Le Mare, 1998). With gratings, the response is relatively narrowly tuned to spatial frequencies between 0.5 and 7 cycles/degree, with a peak sensitivity at 1 cycle/degree. This spectrum is narrower than is the case for the normal hemifield, which can respond to spatial frequencies up to approximately 30 cycles/degree-i.e., there is a reduction in acuity of about 2 octaves and the sensitivity is also decreased. These properties correlate well with the results obtained with two-alternative forced-choice psychophysics for a human blindsight subject (Weiskrantz et al., ibid.). With colored stimuli in the blind hemifield, the pupil shows a robust constriction to a red stimulus on a gray ground (matched in luminance), but not to an equiluminant green stimulus on gray, again, with reduced sensitivity. Here, too, the pupillometric responses correlate with 2AFC psychophysical judgments in blindsight subjects, with good discrimination between red and achromatic stimuli and poor discrimination between green and achromatic stimuli (Barbur, Harlow, Sahraie, Stoerig, & Weiskrantz, 1994). They also correlate with activation levels measured by FMRI (Barbur, Sahraie, Simmons, Weiskrantz, & Williams, in press).
All of these results, however, are ambiguous with respect to the strict definition of ''blindsight''-residual discriminative capacity in the absence of acknowledged awareness. The reason is that the functions just described were all obtained with stimuli with sharp temporal onset and offsets. Some hemianopic subjects, including G.Y., report having a ''feeling'' or a ''knowing'' that something has occurred with