Previous research has shown that a range of physical and biological drivers can influence the composition of faunal assemblages occupying localities within streams. There is much debate in the literature about which of these is more important. Descriptive and experimental field studies were conducted in two relatively undisturbed, second order rainforest streams in southeast Queensland, Australia. The principal objectives were to describe spatial and temporal patterns in pool fauna and explore relationships between these patterns and physical attributes of habitat, disturbance and biotic interactions. The macroinvertebrate and vertebrate fauna of 12 small stream pools were sampled approximately monthly over a period of 15 months. Samples were collected from all major within-pool habitat types and concurrent measurements of potentially important environmental parameters were made at landscape scales of stream, pool and habitat patch. Faunal assemblages were consistently different between the two streams and between the various within-pool habitat types, although the latter may partially be explained by differences in sampling protocols applied in the different habitat types. However, spatial and temporal variation in faunal assemblages within habitat types was large at the scales of whole pools and within-pool habitats, and this variation occurred apparently independently of variation in physical habitat attributes. These results indicated that very little of the local scale faunal variation could be explained by abiotic drivers and that some other factors must be responsible for the observed faunal patterns. Previous research had indicated that atyid shrimps can play a significant ecological role in rainforest streams, where they act as 'ecosystem engineers' by removing fine sediment from hard surfaces. This subsequently alters algal dynamics and faunal composition in streams. A pool-scale manipulative experiment was conducted to investigate the role of the atyid Paratya australiensis, which is an abundant and conspicuous component of the fauna. Removal of shrimp from pools had no effect on sediment accrual on hard surfaces and consequently did not affect algal biomass or faunal assemblages. The lack of effect on sediment accumulation was attributed to the low rate of deposition in these streams, which was an order of magnitude lower than in streams where atyids have been demonstrated to play a keystone role. The fish Mogurnda adspersa was found to be the primary predator of pool fauna in the study streams, where it preyed on a wide variety of taxa. Dietary analyses revealed that an ontogenetic shift occurred in both diet and the within-pool habitat where fish fed. Within this general framework, individual fish had strong individual prey preferences. Significant correlations were found between the natural abundance of Mogurnda in pools and faunal assemblage patterns in both gravel habitat and pools in general, indicating that predation had an effect on pool fauna. The nature of this effect varied between habitats. A direct density dependent response was observed in gravel habitat. In contrast, the response in pools varied considerably between individual pools, perhaps reflecting the differing prey preferences of individual fish. Despite these correlations, an experimental manipulation of the density of Mogurnda at a whole-pool scale did not conclusively identify a predation effect. This may have been due to problems with fish moving between treatments, despite attempts to constrain them, and low experimental power due to the inherent high variability of pool fauna. Overall, the results of the study indicated that there was considerable spatial and temporal variation in pool fauna despite similarities in the physical attributes of pools and their close proximity. This variation appeared to occur at random and could not be explained by abiotic or biotic factors. Predation had a small effect, but could not explain the overall patterns, whereas disturbance by spates had very little effect at all. Stochastic processes associated with low level random recruitment were identified as a possible and plausible explanation for observed patterns. These conclusions are discussed in terms of their implications for our understanding of the ecology and management of streams.