Fish ecologists often use species-discharge relationships (SDRs) to understand how species richness varies with aquatic habitat availability, but few SDR studies have considered whether the reported SDRs are scale-dependent, or attributed the SDR to a specific causal mechanism. Here, we assessed whether the SDR is scale-dependent by using individual river reaches, rather than complete river basins, as sampling units in a SDR analysis. We also determined whether the SDR is a function of among-reach habitat diversity. To do so, we first tested for longitudinal zonation along three major Pacific Northwest (U.S.A.) rivers. Our zonation tests consistently detected ‘lower,’ ‘middle,’ and ‘upper’ river fish assemblages, each of which was characterized by common patterns in adult habitat use, feeding guild structure, and reproductive behavior, and was associated with predictable habitat conditions. When these longitudinal zones were used as sampling units in a SDR analysis (i.e., total discharge and species richness within each zone), we detected strong linear relationships between discharge and species richness (log₁₀ data). Because individual zones predicted species richness more effectively than complete basins, we conclude that the SDR is scale-dependent. And we infer that among-zone habitat shifts are an important determinant of the SDR, as the slope of the SDR is a function of the differential richness found in each zone.