Vole grazing may be a step-function, with a critical threshold density, at which voles expand their preferred diet to lower quality forage (threshold herbivory hypothesis). Accordingly, we predicted that the establishment of unpalatable woody plants would be more strongly associated with peak herbivore abundances than with cumulative herbivory at lower numbers. We also investigated whether damage level is better explained by actual vole numbers or by numbers adjusted to the carrying capacity of the herbaceous vegetation. Our results did not support the threshold-density hypothesis. Cumulative herbivory explained the probability of sapling damage better than peak herbivory; sapling survival and growth were equally well explained by mean- and peak-vole abundances. Even at low abundances, herbivory was extended to all woody species; the damage level, however, varied according to the palatability of the woody species. Actual herbivore numbers explained sapling damage better than did abundance adjusted to carrying capacity.