Abstract
The results of Grossman et al. (1982), which suggest that the stream fish association described by Whitaker (1976) is not discernably different from random, is an artifact of mixing together species of different parts of the stream. Analysis of covarying species as the basis of identification of community components suggests that the riffle-inhabiting fish and pool-dwelling fish comprise 2 independent associations. Riffle-inhabiting species are closely correlated with the total number of other fish collected and with May rainfall. Population flux of the species within this group is significantly concordant and each species feeds in a different manner than the rest. Populations of pool-dwelling species show no significant concordance in their variation from year to year, show no similarity in their response to rainfall patterns and are all feeders on the benthos. Only the niche diversification hypothesis (Connell, 1978) is in accord with these observations for the riffle-inhabiting species and only the intermediate disturbance hypothesis (Connell, 1978) conforms to the patterns observed for pool dwellers.