The North American Breeding Bird Survey was started in 1966, and provides information on population change for >400 species of birds. it covers the continental United States, Canada, and Alaska, and is conducted once each year, in June, by volunteer observers. A 39.4 kIn roadside survey route is driven starting 30 min before sunrise, and a 3 min point count is conducted at each of 50 stops spaced every 0.8 kIn. Existing analyses of the data are internet-based (http://www.mbr-pwrc.usgs.govlbbslbbs.html), and include maps of relative abundance, estimates of population change including trends (%/yr), composite annual indices (pattern in time), and maps of population trend (pattern in space). At least 36 species of marsh birds are encountered on the BBS, and the survey provides estimates with greatly varying levels of efficiency for the species. It is often difficult to understand how well the BBS surveys a species. Often, efficiency is judged by estimating trend and its variance for a species, then by calculating power and needed samples to detect a prespecified trend over some time period (e.g., a 2%/yr trend over 31 yr). Unfortunately, this approach is not always valid, as estimated trends and variances can be of little use if the population is poorly sampled. Lurking concerns with BBS data include (1) incomplete coverage of species range; (2) undersampling of habitats; and (3) low and variable visibility of birds during point counts. It is difficult to evaluate these concerns, because known populations do not exist for comparison with counts, and detection rates are time-consuming and costly to estimate. I evaluated the efficiency of the BBS for selected rails (Rallidae) and snipes (Scolopacidae), presenting estimates of population trend over 1966-1996 (T), power to detect 2%/yr trend over 31 yr, needed samples to achieve power of 0.75 with alpha= 0.1, number of survey routes with data for the species (N), average abundance on survey routes (RA), and maps of relative abundance. Examples include Yellow Rail (Coturnicops noveboracensis) (T=12 %/yr; P= 0.0085; N =28; routes; RA=0.05; Power=0.37; Needed samples=85), Black Rail (Laterallus jamaicensis) (No trend data or power information available, N =8), Clapper Rail (Rallus longirostris) (T=1.9%/yr; P=0.55; N =64; RA=0.31; Power=0.35; Needed samples=590), King Rail (Rallus elegans) (T=-4.2 %/yr; P= 0.03; N =76; Power=0.41; Needed samples=159), Sora (Porzana carolina) (T=0.98 %/yr; P= 0.24; N =720; RA= 0.92; Power=0.69; Needed samples= 377), and Common Snipe (Gallinago gallinago) (T=-0.24 %/yr; P= 0.54; N =1412; RA= 2.19; Power=0.98; Needed samples=205). With regard to quality of BBS data, marsh birds fall into 3 categories: (1) almost never encountered on BBS routes; (2) encountered at extremely low abundances on BBS routes; and (3) probably fairly well sampled by BBS roadside counts. BBS data can provide useful information for many marsh bird species, but users should be aware of the limitations of the BBS sample for monitoring species that have low visibility from point counts and prefer habitats not often encountered on roadsides.