During the period of study the percentage of males in most of my samples was less than 50 per cent (Fig. 2). Only once, in June, 1952, did the mean percentage of males in samples from three areas (House Field, Quarry Field, Fitch traps) exceed that level and then it was only 50.1 per cent. On several occasions, however, the percentage of males in a sample from a single area was slightly above 50 per cent. The highest percentage of males recorded was 56.69 per cent, in a sample taken from the Quarry Field population in June, 1952. In the samples taken in April, 1952, the mean percentage of males was 39.67 per cent, the lowest mean recorded. The low point for one sample was 28.02 per cent in August, 1952, from Quarry Field. The mean percentage of males in all samples taken was 45.02 ± 2.72 per cent. Percentages observed would occur in random samples taken from a population with 50 per cent males less than one per cent of the time. Exactly 50 per cent of the young in the 65 litters examined were classified as males but the sample was small and the sexing of newborn individuals was difficult.

The extent to which sex ratios in samples were affected by trapping procedure was not determined. A possibility considered was that the greater wandering tendency of males (Blair, 1940:154; Hamilton, 1937c:261; Townsend, 1935:98) impaired the formation of trap habits (Chitty and Kempson, 1949:536) on their part and thus unbalanced the sex ratios of the samples. If this were the explanation, the apparent sex ratio on larger areas would more nearly approximate the true ratio, and the frequency of capture of females would exceed that of males. The evidence is somewhat equivocal. In the populations described here the mean number of captures per individual per month was 2.31 for females, which was significantly greater (at the one per cent level) than the 2.20 captures per individual per month which was the mean number for males. This difference supports the idea that differences in habits between the sexes result in distorted sex ratios in samples obtained by live-trapping. Mean percentages of males did not, however, differ significantly between the House Field-Quarry Field samples and the samples from the Fitch trapping area, nearly five times as large.

Three age classes, juvenal, subadult and adult, were separated on the basis of condition of pelage. The percentage of adults in populations varied seasonally (Fig. 2). January, February and March were the months when the adult fraction of the population was highest and October and November were low points, with May and June showing percentages almost as low. The only marked variation in this seasonal pattern occurred in July and August, 1952, when the percentage of adults rose sharply. This was due to a depression in the reproductive rate during the dry summer of 1952, which is discussed later in this report. Juveniles made up only a small fraction of the population from December through March and a relatively large fraction in the October-November and May-June periods (Fig. 2). Again, July and August of 1952 were exceptions to the pattern as the percentages of juveniles in these months fell to midwinter levels. As expected, the curve of the percentages of subadults in the population followed that of the juveniles and preceded that of the adults. The mean percentages for the thirty month period for which data were available were: adults, 77.72 ± 4.48 per cent; subadults, 14.06 ± 3.14 per cent; and juveniles, 8.22 ± 2.62 per cent. Seasonal and yearly changes in the population structure occurred, with notable variation in the ratio of breeding females to the entire population, as discussed in this report under the heading of reproduction.

Since some of the juveniles did not move enough to be readily trapped, the real percentage of juveniles in the population was probably far greater than that shown by trapping data. I tried, therefore, to estimate the number of juveniles on the study plot each month by multiplying the number of lactating females by the mean litter size. As expected, the results were consistently higher than the estimate based on trapping data. The discrepancy was largest in April, May, June and October. During the winter there was no important difference between the two estimates. Even when the discrepancy was greatest, the estimated weight of the juveniles missed by trapping was not large enough to modify the picture of habitat utilization in any important way. I chose, therefore, to count only those juveniles actually trapped. Although probably consistently too low, such a figure seemed more reliable than an estimate made on any other basis.

A study of the age groups in each month's population revealed a differential survival based on the season of birth. Blair (1948:405) found that chances of survival in Microtus pennsylvanicus were approximately equal throughout the year. In the present populations of M. ochrogaster, however, voles born in October, November, December and January tended to live longer than those born in other months (Fig. 3). Presumably these animals, born in autumn and early winter, were more vigorous than their older competitors and were therefore better able to survive the shrinking habitat of winter. Their continued survival after large numbers of younger voles had been added to the population probably was permitted by the expanding habitat of spring and summer. The percentage of the population surviving the winter of 1951-1952 was approximately double the percentage surviving the winter of 1950-1951. This difference seemed to be due to the smaller population entering the winter of 1951-1952 rather than any major difference in the environmental resistance.

As a consequence of the differential survival, most of the breeding population in the spring was made up of animals born the previous October and November. Fig. 4 shows that in February, when the percentage of breeding females ordinarily began to rise, 51.6 per cent of the population was born in the previous October and November. Voles born in these two months continued to form a large part of the population through March (45.1 per cent), April (38.5 per cent), May (23.9 per cent), June (18.7 per cent) and July (16.2 per cent) (Fig. 4). These percentages suggest that the habitat conditions in October and November were probably important in determining the population level for at least the first half of the next year.

Fig. 4. Differential survival of voles according to month when first caught. Each column represents the percentage of the monthly sample first caught in each of the preceding months. Those voles caught first in October and November survived longer than those first caught in other months. Relatively few individuals remained in the population as long as one year.