The sensitivity of transient, pattern-reversal visual evoked potentials in the detection of early compressive lesions of the chiasm is controversial in the literature. There have been claims that the technique is capable of detecting an abnormality in the absence of any demonstrable visual field loss, and conversely that VEPs are not reliable for the detection of chiasmal lesions even when a bitemporal hemianopsia is clearly recordable. Using nine patients with pituitary adenoma we attempted to quantify the extent of visual field loss and correlate the diagnostic capabilities of topographically recorded potentials following full- and half-field stimulus presentations of various field and check sizes. Differential light thresholds were measured and quantified according to one investigator's graticule for the neural representation of visual space. Results show a strong correlation between the degree of information loss and the diagnostic value of the visual evoked potential. The technique was, however, capable of detecting abnormality in the absence of recordable field loss when large field and check sizes were used.