Ungulate browsing in predator depleted North American landscapes is believed to be causing widespread tree recruitment failures. However, canopy disturbances and variations in ungulate densities are sources of heterogeneity that can buffer ecosystems against herbivory. Relatively little is known about the functional response (the rate of consumption in relation to food availability) of ungulates in eastern temperate forests, and therefore how “top down” control of vegetation may vary with disturbance type, intensity, and timing. This knowledge gap is relevant in the Northeastern United States today with the recent arrival of hemlock woolly adelgid (HWA; Adelges tsugae) that is killing eastern hemlocks (Tsuga canadensis) and initiating salvage logging as a management response. We used an existing experiment in central New England begun in 2005, which simulated severe adelgid infestation and intensive logging of intact hemlock forest, to examine the functional response of combined moose (Alces americanus) and white-tailed deer (Odocoileus virginianus) foraging in two different time periods after disturbance (3 and 7 years). We predicted that browsing impacts would be linear or accelerating (Type I or Type III response) in year 3 when regenerating stem densities were relatively low and decelerating (Type II response) in year 7 when stem densities increased. We sampled and compared woody regeneration and browsing among logged and simulated insect attack treatments and two intact controls (hemlock and hardwood forest) in 2008 and again in 2012. We then used AIC model selection to compare the three major functional response models (Types I, II, and III) of ungulate browsing in relation to forage density. We also examined relative use of the different stand types by comparing pellet group density and remote camera images. In 2008, total and proportional browse consumption increased with stem density, and peaked in logged plots, revealing a Type I response. In 2012, stem densities were greatest in girdled plots, but proportional browse consumption was highest at intermediate stem densities in logged plots, exhibiting a Type III (rather than a Type II) functional response. Our results revealed shifting top–down control by herbivores at different stages of stand recovery after disturbance and in different understory conditions resulting from logging vs. simulated adelgid attack. If forest managers wish to promote tree regeneration in hemlock stands that is more resistant to ungulate browsers, leaving HWA-infested stands unmanaged may be a better option than preemptively logging them.