In the cases of large kills, or limited manpower, or both, it may be impossible to process all birds, however desirable this might be. If possible, however, all should be collected, identified, the numbers and species recorded, and rarities saved. Further, partial analysis, or more properly, complete analysis of a partial sample, can be made. Analyses which for any reason require randomness of sample pose a special problem. We think that in very large kills the best way to solve this problem is probably to make one or more transects across the area where dead birds are found. These transects should cross both the areas of greatest and least density (to allow for fast and slow flying species). Their width may be adjusted to give the desired number of birds, that is, the number that can be adequately processed. Another alternative would be to decide to study certain abundant species and pick up all of these. There are other possibilities, but in any event the method of sampling should be thoroughly described wherever all birds have not been processed. Summary The foregoing paper discusses accidents in which large numbers of night-migrating birds are killed. A brief historical review of ornithological interest in such occurrences is given, and the types of data provided by the accidents are listed and discussed. In particular, recent accidents occurring in early October, 1954, through much of eastern United States are cited, and detailed analysis is presented of a sample of 1090 birds killed one mile west of Topeka, Shawnee County, Kansas, between September 25 and October 23, 1954. At Topeka during the period mentioned, 1090 birds representing 61 species were collected and were processed at the University of Kansas. For all specimens, weight, sex, age, and fat condition were recorded, and certain species were measured as well. Some notes on molt were taken. A total of 193 birds was preserved as study skins, and 49 as skeletons. Comments on weight, size, sex, age, subspecific identity, and status in Kansas are presented in an annotated list. Randomness of this and other similar samples is discussed. A theoretical computation is given for several nights of the numbers of migrants passing through a plane one mile in width, from 450 to 950 feet above ground level, and oriented to face the arriving migrants. The computed totals give some idea of the tremendous volume of nocturnal migration under some conditions. Potentialities of further study of such events are discussed and a comparison is made with lunar observations.