Macrofossils from woodrat (Neotoma) middens serve as an important proxy for reconstructing past vegetation in arid and semiarid regions of North America. The presence/absence of plant macrofossils in middens can provide valuable information on temporal and spatial patterns of plant migration and range boundaries. The primary aim of this study was to determine how local plant abundance, distance of plant populations from midden sites, and species population density on the landscape affect the probability of occurrence of macrofossils in middens. The study was designed with the primary intent of determining the reliability of middens in detecting scattered populations of Pinus ponderosa. We analyzed macrofossil assemblages from 42 modern woodrat middens from West Carrizo Canyon in southeastern Colorado, near the current eastern range margin of Pinus ponderosa. We compared midden contents with composition of the surrounding vegetation, measuring distance from the midden to the nearest individual of selected plant species, and the percent cover of each species within 30 m of the midden. We used this information to model the probability of species presence in a midden across a range of population densities on the landscape. Macrofossils of Juniperus spp., Quercus gambelii, and Opuntia spp. were consistently found in middens regardless of their local abundance in vegetation, although populations occurred within 30 m of all middens. Pinus edulis and P. ponderosa occurred in nearly all middens within 20–30 m of individual trees. P. ponderosa was rare in middens >20–30 m away from individual trees. Results of a simple simulation model suggest that middens become absolutely reliable indicators of P. ponderosa presence on the landscape only when average tree density exceeds 50 stems ha−1. Woodrats reliably collected macrofossils of Pinus edulis, P. ponderosa, Juniperus spp., Quercus gambelii, and Opuntia spp. when populations of these taxa occur within 20–30 m of a midden site. Woodrats did not collect P. ponderosa when the nearest individuals were more than 30 m away. Low-density populations of these and other species may be difficult to detect in fossil woodrat-midden series owing to reduced probability that individuals grow within foraging distance of the middens. Data from this and similar studies can be used to construct and parameterize a forward model of macrofossil representation in woodrat middens.