258 TELFORD: MULTIMAMMATE RAT POPULATION BIOLOGY IN TANZANIA

 The first year of the study (Fig. 3) was notably warmer than the previous
 10-year average from January through March, with maxima for those months
 exceeding 330, about two degrees higher than average.
 Other years more closely resembled the 1971-80 average (Figs. 3, 4).
 The curve of mean monthly precipitation for the 10 years preceding the
 study, 1971-80 (Fig. 2), shows two peaks: the maximum, an average of 184 mm
 in April, and a second, somewhat lower peak of 126 mm in January. Minimum
 levels occur from June to October, with the least rain in August, 6.2 mm. As
 will be seen later, the critical period for rainfall for the rodent population is
 from November through March; this is described in detail below, for each year
 of the study showed significant differences from the pattern formed by the 10-
 year average.
 In 1981-82 (Fig. 5), rainfall was normal from September through
 November, slightly below normal in December, less than half that expected for
 January, and virtually absent in February, when only 4.4 mm were recorded.
 The rainfall peak for the year, in April, was only 97.2 mm, slightly over half
 that expected, and only 2 mm more than had been received in December.
 In 1982-83 (Fig. 6), precipitation was double or greater than the 1971-80
 average from September through December, and then decreased sharply to
 produce a very dry January. Rainfall remained below normal until May, then
 exceeded the expected in May, becoming normal thereafter. The heavy rain
 from October through December had important consequences for the rodent
 population that year.
 In 1983-84 (Fig. 7), rainfall was below normal from September through
November but then rose sharply to levels about one-third above normal for
December and January. February and March were similar to the average, but
precipitation in April was again one-third more than expected, reaching nearly
300 mm in that month. The remainder of the year was normal. Heavy rain in
January, followed by a normal February and March appears to be the critical
environmental influence for the rodent population of that year.
 In 1984-85 (Fig. 8), rainfall was normal in September and October, but in
November more than doubled the expected amount. Precipitation decreased
sharply in December to about half the average and continued to decline in
January, when only 10.8 mm was received, in comparison to the 126 mm
expected from the 10-year average. The study ended in February, which had a
greater than normal amount of rain.
 The hypothesis advanced in this paper is that the events observed in the
rodent population studied can be explained as the result of annual variation in
the timing and level of precipitation during the critical period from November
through February of each year. This variation is summarized thus:

 1981-82: January and February had extremely low
 rainfall.