Point counts are a controversial sampling method for bird populations because the counts are not censuses, and the proportion of birds missed during counting generally is not estimated. We applied a double-observer approach to estimate detection rates of birds from point counts in Maryland, USA, and test whether detection rates differed between point counts conducted in field habitats as opposed to wooded habitats. We conducted 2 analyses. The first analysis was based on 4 clusters of counts (routes) surveyed by a single pair of observers. A series of models was developed with differing assumptions about sources of variation in detection probabilities and fit using program SURVIV. The most appropriate model was selected using Akaike's Information Criterion. The second analysis was based on 13 routes (7 woods and 6 field routes) surveyed by various observers in which average detection rates were estimated by route and compared using a t-test. In both analyses, little evidence existed for variation in detection probabilities in relation to habitat. Double-observer methods provide a reasonable means of estimating detection probabilities and testing critical assumptions needed for analysis of point counts.