4 



NORTH AMERICAN FAUNA 65 



METHODS 



Since differences in coloration, especially of females, are often 

 subtle, some of the races can generally be separated most accurately 

 on the basis of body size and dimensions. Differences in ecology, time 

 of breeding, and extent of migration also help in characterizing some 

 subspecies. 



When field observations and study of laboratory skins suggested 

 differences in a particular population, a statistical analysis was made 

 to determine the possible significance of morphological differences. 

 Mayr considers the conventional level of subspecies difference to be 

 90 percent or more (Mayr, Linsley, and Usinger, 1953) . The cri- 

 terion used here in determining the validity of races was whether 95 

 percent or more of the specimens of a population were separable 

 from 95 percent or more of the specimens of the adjacent race or 

 races. To determine the degree of difference and the percentage of 

 joint nonoverlap between characters of races, the standard error was 

 calculated for each mean. Confidence limits were determined for the 

 .05 probability level as tlie mean ± "t" times the standard error. 

 The means, confidence limits, and ranges are given in the table and 

 in figures 2 to 4. When the confidence limits do not overlap, 

 statistically significant differences are indicated. 



In this review, 463 adult specimens and several juveniles were 

 studied. They represent most of the known populations from Arizona 

 and Texas south to Ecuador and Peru. The largest series were those 

 of the Fish and Wildlife Service collection in the U. S. National 

 Museum that I had obtained in Mexico and Central America. Speci- 

 mens from the West Indies were not included. Birds from Jamaica, 

 Cuba, Haiti, Dominican Republic, the Bahamas, and Old Providence 

 Island were examined, but additional breeding specimens are needed 

 before some puzzling questions can be answered and an adequate 

 appraisal made of t^ieir taxonomy. The specimens of asiatica in- 

 cluded in the present tabulation of measurements were from breeding 

 grounds in southern Texas and northeastern Mexico. 



Measurements used in this study include lengtli of wing (chord) , 

 tail, culmen, and tarsus, although the tarsus is not included in most 

 of the comparisons as it is not of significant diagnostic value. Statistical 

 differences in dimensions are shown in the table and in figures 2 to 

 4 (see pages 22-27) . Plumage colors were studied under natural light 

 in most instances. A few were determined under special lights which 

 approximate daylight in the Bird Division, U.S. Museum of Natural 

 History. Color names are from Ridgway (1912) . 



Since migrant subspecies may mingle with resident birds during 

 the winter, it is essential that full information, particularly condition 

 of gonads and amount of fat, be recorded in the field. In Mexico and 

 Central America it was usually possible, from February to April, to 

 separate the resident and migrant forms by the greater development 



