- I used program CAPTURE to analyze 1,535 data sets for 33 North American species of small mammals for sources of variation in capture probabilities and to characterize species-specific responses to markrecapture trapping. CAPTURE incorporates seven models to account for all possible combinations of heterogeneity (h), behavioral response (b), and time effects (t) on capture probabilities, in addition to the null model (M₀) which requires equal capture-probabilities. The following information was tabulated for each data set: 1) results of the closure test; 2) probabilities of fit of the data to the models; 3) model-selection criteria; 4) abundance estimates, confidence intervals, and estimated probabilities of capture and recapture; and 5) mean maximum-distance moved. Population closure was not rejected 70-100% of the time for 30 species when models M₀ or M[subscript h] were chosen by CAPTURE. Differences within families were significant only for Heteromyidae. Frequency of closure decreased significantly as length of trapping session increased for 4 of 5 species for which I performed the test. Model M₀ was selected most often by CAPTURE as the most appropriate model for most species, which merely reflected the inability of the program to detect unequal probabilities of capture from small data sets. However, heterogeneity was evident in Sylvilagus floridanus, Tamias townsendii, Spermophilus richardsonii, Glaucomys sabrinus, Peroqnathus parvus, Chaetodipus intermedins, Dipodomvs merriami, Dipodomvs ordii, Reithrodontomys spp., Peromvscus maniculatus, Sigmodon hispidus, Clethrionomys spp., Microtus californicus, and M. pennsvlvanicus. Strong behavioral responses (trap proneness or trap shyness) were detected in T. townsendii, S. richardsonii, R. megalotis, P. maniculatus, and M. pennsvlvanicus. Time effects, particularly with heterogeneity, were apparent for Sylvilagus nuttallii, Tamias minimus, D. ordii, C. gapperi, M. californicus, and M. pennsvlvanicus. Sylvilagus spp., Tamias amoenus, G. sabrinus, Perognathus longimembris, and Microtus ochrogaster were trap shy. Tamias striatus, J_. townsendii, S. richardsonii, P. parvus, R. megalotis, P. maniculatus, Peromvscus truei, S. hispidus, M. californicus, and M. pennsvlvanicus were trap prone. Murid species had the highest estimated probabilities of capture, followed in decreasing order by heteromyids, sciurids, and leporids. Estimates of abundance from model M[subscript h] were greatest and often significantly greater than those from, in decreasing order, M[subscript bh], M[subscript b], M₀, and M[subscript t]. Coefficients of variation of abundance estimates from models M[subscript b] and M[subscript bh] were usually twice as great as those from models M₀, M[subscript h], and M[subscript t], and the coefficients of variation from M[subscript bh] were significantly greater than those from M[subscript t] and M₀ Coefficient of variation is not an appropriate measure of the reliability of abundance estimates from CAPTURE models. Abundance estimates from the Jolly-Seber model were almost always less than those from CAPTURE models, regardless of which CAPTURE model was the most appropriate, and they rarely exceeded the total number of animals captured during the trapping session. The negative biases of these estimates were so great that they more likely were due to unequal probabilities of capture rather than to emigration.