Spatial scale is a critical consideration for understanding ecological patterns and controls of ecological processes, yet very little is known about how rates of fundamental ecosystem processes vary across spatial scales. We assessed litter decomposition in stream networks whose inherent hierarchical nature makes them a suitable model system to evaluate variation in decay rates across multiple spatial scales. Our hypotheses were (1) that increasing spatial extent adds significant variability at each hierarchical level, and (2) that stream size is an important source of variability among streams. To test these hypotheses we let litter decompose in four riffles in each of twelve 3rd-order streams evenly distributed across four 4th-order watersheds, and in a second experiment determined variation in decomposition rate along a stream-size gradient ranging from orders 1 to 4. Differences in decay rates between coarse-mesh and fine-mesh litter bags accounted for much of the overall variability in the data sets, and were remarkably consistent across spatial scales and stream sizes. In particular, variation across watersheds was minor. Differences among streams and among riffles were statistically significant, though relatively small, leaving most of the total variance (51%) statistically unexplained. This result suggests that variability was generated mainly within riffles, decreasing successively with increasing scale. A broad range of physical and chemical attributes measured at the study sites explained little of the variance in decomposition rate. This, together with the strong mesh-size effect and greater variability among coarse-mesh bags, suggests that detritivores account, at least partly, for the unexplained variance. These findings contrast with the widespread perception that variability of ecosystem characteristics, including process rates, invariably increases (1) with spatial extent and (2), in stream networks, when analyses encompass headwaters of various size. An important practical implication is that natural variability need not compromise litter decomposition assays as a means of assessing functional ecosystem integrity.