We are studying a sample of Maryland (2 %) and New Hampshire (4 %) Atlas blocks and a small sample in Maine. These three States used different sampling methods and block sizes. We compare sampling techniques, roadside with off-road coverage, our coverage with that of the volunteers, and different methods of quantifying Atlas results. The 7 1/2' (12-km) blocks used in the Maine Atlas are satisfactory for coarse mapping, but are too large to enable changes to be detected in the future. Most states are subdividing the standard 7 1/2' maps into six 5-km blocks. The random 1/6 sample of 5-km blocks used in New Hampshire, Vermont (published 1985), and many other states has the advantage of permitting detection of some changes in the future, but the disadvantage of leaving important habitats unsampled. The Maryland system of atlasing all 1,200 5-km blocks and covering one out of each six by quarterblocks (2 1/2-km) is far superior if enough observers can be found. A good compromise, not yet attempted, would be to Atlas a 1/6 random sample of 5-km blocks and also one other carefully selected (non-random) block on the same 7 1/2' map--the block that would include the best sample of habitats or elevations not in the random block. In our sample the second block raised the percentage of birds found from 86% of the birds recorded in the 7 1/2' quadrangle to 93%. It was helpful to list the expected species in each block and to revise this list annually. We estimate that 90-100 species could be found with intensive effort in most Maryland blocks; perhaps 95-105 in New Hampshire. It was also helpful to know which species were under-sampled so we could make a special effort to search for these. A total of 75 species per block (or 75% of the expected species in blocks with very restricted habitat diversity) is considered a practical and adequate goal in these States. When fewer than 60 species are found per block, a high proportion of the rarer species are missed, as well as some of the common ones. Similarity indices based on fewer than 60 species per block reflect coverage rather than habitat affinities. Atlas blocks that are ecologically similar should have similarity indices (S) of at least 0.80 to be considered adequately covered. S = 2C/(A + B), where C is the number of species in common and A and B are species totals for each of the two blocks being compared. A series of 15 13-minutes roadside stops yielded more species than 15 off-road stops, but off-road stops always had some species not detected at the roadside stops. A series of timed stops is an excellent way to map relative abundance if the stops are standardized with respect to time of day and weather, and the counts are made by observers of comparable ability. Efforts to estimate Atlas block populations in powers of 10 (as in the French Atlas) have not gained acceptance in U.S.A. Most observers feel unqualified to make estimates. An efficient way to Atlas a block is to make at least 3 early morning visits to 15 or more specific stops. Arrive in the block early enough to check for nocturnal species on at least two days; and after completing the specific stops, search the block for other species and for confirmations.