Pacific salmon provide resource subsidies to thousands of nutrient-poor streams along the North Pacific Rim where they are believed to increase stream productivity.  However, salmon disturb sediments during migration and spawning, potentially reducing productivity.  We hypothesized that the net outcome of these contrasting effects varies among stream habitats depending on their characteristics.  We measured functional (NCP, CR) and structural (chla, AFDM) biofilm metrics in three habitats (stream edges, riffles, pools) every eight days, spanning the salmon run in a Southeast Alaska stream.  Before the run, GPP increased 1.75-fold from mid-July to highest levels (mean±SE, 4.88±0.29 mgC·m^-2·h^-1) in early-August, likely due to increases in temperature and light.  While GPP generally did not differ among habitats, pools leveled out at highest GPP by end-July.  With the arrival of salmon, production declined in all habitats (0.89±0.36 mgC·m^-2·h^-1), but least rapidly in pools, likely because of less salmon disturbance.  Structural measures did not always track functional changes, suggesting such measurements are not reliable surrogates for production responses to salmon in all habitats.  Although salmon effects were detected, to understand the net impact of salmon on ecosystem productivity, both structural and functional measures should be considered as well as relative abundances and characteristics of habitats.