Traditional urban drainage degrades receiving waters. Alternative approaches have potential to protect downstream waters, but widespread adoption requires robust demonstration of their feasibility and effectiveness. We conducted a catchment-scale experiment over 19 years to assess the effect of dispersed stormwater control measures (SCMs), measured as a reduction in effective imperviousness (EI) on stream water quality in 6 sites on 2 streams. We compared changes in those sites over 7 years as EI decreased, to changes in the 12 preceding years, and in 3 reference and 2 control streams. SCMs reduced phosphorus concentrations and summer temperature to reference levels in dry weather where EI was sufficiently reduced, but effects were smaller with increased antecedent rain. SCMs also reduced nitrogen concentrations which were influenced by septic tank seepage in all sites. SCMs had no effect on suspended solids concentrations, which were lower in urban than in reference streams. SCMs increased electrical conductivity: along with reduced temperature this is evidence of increased contribution of groundwater to baseflows. This experiment strengthens inference that urban stormwater drainage increases contaminant concentrations in streams, and demonstrates that such impacts are reversible and likely preventable. Variation in degree of water quality improvement among experimental sites suggests that achieving reference water quality would require SCMs with large retention capacity intercepting runoff from nearly all impervious surfaces, thus requiring more downslope space and water demand. EI is a useful metric for predicting stream water quality responses to SCMs, allowing better catchment prioritization and SCM design standards for stream protection.