Benthic diatoms are important indicators of human impacts on streams. Epilithic diatoms are collected most often for bioassessments, but potentially important ecological information from other habitats could be missed. Within our study region, substrata ranged from 100% rock to 100% sandy silt at 61 sites, leading us to question the appropriateness of sampling solely epilithic surfaces in riffles. We compared two protocols for collecting diatoms:(1) epilithic habitat samples (EHS) within riffles and (2) multi-habitat samples (MHS), which proportionately included different habitats (e.g., riffles and pools) and substrata (e.g., sand, silt, and rock). Three streams were not compared because rocks were absent. Nonmetric multidimensional scaling indicated that diatom communities from EHS and MHS responded similarly to alkalinity and agricultural gradients, and Procrustes analysis showed that ordinations were similar (P \ 0.01). Percent motile diatoms and metrics indicating high or low P and N from MHS showed stronger relationships with land use variables than those metrics from EHS. Bray-Curtis (BC) similarity between the protocols increased as diversity and richness decreased in MHS, as agricultural impacts increased, and as motile diatoms became more abundant in EHS, which likely indicated greater habitat homogeneity or an overriding effect of water chemistry. The two protocols classified 78% of sites concordantly as minimally impacted or impacted. Multi-habitat sampling is preferred where stream habitats vary greatly because it assessed land use impacts more effectively than EHS, while indicating water chemistry impacts as effectively. Multiple habitats also contained greater species diversity, which is important to documenting species distributions, and were always available for sampling, whereas epilithic habitats could be sparse or absent.