A population survey for Long-billed Curlews (Numenius americanus) was completed in the western United States and Canada in 2004 and 2005. This survey was conducted during the early breeding season, using a stratified random sample from habitat strata. The survey design was a 32-km road transect with 40 five-min point counts at 800-m intervals. Detection probabilities were estimated using the removal method in which observations in one-min intervals were removed from further consideration. Model selection based on Akaike’s Information Criterion resulted in a model where detection probability varied among observers, but was constant throughout the point count for each observer. Estimated detection probabilities for the point count duration were greater than 0.68 for all observers. Counts were adjusted for detection probability and then used to estimate the mean density within surveyed point count plots. Overall, the range-wide estimate of total population size was 161,181 individuals. The estimates were 183,231 individuals for 2004 and 139,131 for 2005, with corresponding 90% confidence intervals of 113,324 to 422,046 and 97,611 to 198,252, respectively. In addition to estimates for both the United States and Canada, population densities were estimated for geographic sub-regions: Bird Conservation Regions, Shorebird Planning Regions, administrative regions, and for each Canadian province. Issues and assumptions inherent in the study design and their implications are discussed.