Hollander's insulin test was evaluated by performing successively partial denervation of the parietal cell area, highly selective, selective, and truncal vagotomy operations on three dogs. The rise in acidity, the timing of the highest rise in acidity, and the rise in acid output were examined. The mean rise in acidity (m-equiv/l) before vagotomy (106 +/- SE11, 120 +/- 23, 60 +/- 16) did not differ significantly from those obtained after partial denervation (67 +/- SE14, 125 +/- 12, 45 +/- 12). After selective vagotomy, a presumed complete denervation, the rises in acidity were significantly lower (0 +/- 0.1, 2.5 +/- 4, 0 +/- 9). The Hollander test, however, was occasionally negative before any nerves had been divided, and was positive in 20% of tests after presumed complete parietal cell denervation. The highest rise in acidity before vagotomy most commonly occurred 30 minutes after insulin, and following partial denervation this rise occurred significantly later, most commonly at 60 minutes. After the other operations there was no consistent timing. The rise in insulin-stimulated acid output after presumed complete denervation provided near complete discrimination between innervated and denervated stomachs, and its repeatability as assessed by the coefficient of variation was significantly better than the repeatability of rise in acidity. These findings provide support for the contention that the results of the insulin test should be expressed quantitatively in terms of the rise in acid output rather than using Hollander-type criteria of changes in acidity.