116 
Psyche 
[March 
independently of each other. In this case the coefficient of dis- 
persion, C.D. = 4- , (Sokal and Rohlf, 1969) will approximate 
1 (p = o 2 for the Poisson distribution). If, on the other hand, the 
observed distribution of bees per sample departs significantly 
from the expected Poisson distribution and C.D. < 1 or > 1 then 
the bees can be assumed to be evenly distributed over the flowers 
or foraging in groups (either single or multiple species groups), 
respectively. 
Temporal patterns — Counts taken at the eleven quadrats were 
tested against the expected Poisson distribution, based on the 
observed sample means, to determine the temporal patterns of the 
halictids (individual species samples taken at quadrats 8-11 are 
pooled for purposes of this analysis). The sample means are cor- 
related with the number of flowers within the same quadrats 
(Fig. 2). In fact, flower density appears to be an excellent predictor 
Fig. 2. Regression of mean number of bees per meter 2 as a function of number of 
flowers per meter 2 . Dark circles represent data from 1 1 quadrats; data from Table I. 
