The reliability of certain of the conclusions which might be drawn from data of the kind presented herein depends largely on the randomness of the sample. To what degree does this sample provide a true cross-section of the nocturnal migrants present over the area on a given night or succession of nights? As far as the relative abundance of species in the sample is concerned, there is little doubt that it is not at all random. The absence of such species as the Gray-cheeked Thrush (Hylocichla minima), among the passerines, and many of the shorebirds known to be migrating through the area at the time is evidence for this statement. Quite possibly many seminocturnal species did not strike the tower at all for the simple reason that they could see it, and certain large-eyed diurnal species (such as thrushes and shorebirds) may avoid collision to some extent, thus not appearing in the sample in proportion to their actual numbers. Finally, some or all of the species concerned probably migrate partly by day. The sample may to some degree reflect the true relative abundance of closely related species. For example, there is little doubt that, as shown by the sample, Nashville Warblers are more numerous locally at this season than Tennessee Warblers, a fact that can readily be corroborated by ordinary field observation. Also, the sample is useful in suggesting the actual abundance of species which are furtive and/or difficult to identify under normal field conditions, for example, the Mourning Warbler and Philadelphia Vireo. It is obvious that the sample should reflect the true relative abundance at one place and time of any two species with equal tendency to migrate by night and equal tendency to strike the tower. Since the facts in regard to both tendencies are at present unknown for most species, we think that interspecific comparisons should be avoided or approached with extreme caution.

In respect to the relative abundance of the various sex- and age-classes within a given species, the sample is, we think, as close to random as is possible to obtain. Certainly it is greatly superior to samples obtained by field collecting, where possible differences in habits, wariness, and experience of the birds, and subconscious (if not conscious) selection by collectors can all bias the results. Dwight (1900:128-9) believed that the greater wariness of adult birds was almost entirely responsible for the seemingly disproportionate number of immatures in autumn and gave some observational evidence in favor of his views. The large percentage of adults in some of the samples here treated tends to reinforce Dwight's position. To a somewhat lesser extent, this advantage in randomness of accidental kills over routine collecting may be supposed to apply also in demonstrating the composition by subspecies of a single migrant species.

So far as particulars already mentioned are concerned, the present sample or other samples of tower-killed birds would seem to be in no way superior (that is, more nearly random) to samples obtained in connection with lighthouses and other lighted objects, and ceilometers. In one important respect, however, it is probably somewhat superior to these as the dimly red-lighted structure has not been shown to have any important collecting or attracting influence. Thus, in computations intended to estimate the over-all abundance of migrants, the sample should be more reliable than samples involving bright light with its possible attracting effect.