1. Dippers were surveyed along seventy‐four acid‐sensitive streams in upland Wales in 1984 and again in 1995. At forty‐eight of the sites, and in an additional twenty‐three to twenty‐nine adjacent catchments, changes in acid‐base status and macroinvertebrate communities were assessed. River habitat surveys (RHS) and GIS (Geographical Information System) provided information on habitat features.2. pH means across all the sites were higher on average by 0.12 pH units in 1995 than 1984, accompanied by significantly increased total hardness and reduced sulphate. However, geometric mean aluminium concentrations increased significantly, while the abundances of important dipper prey either remained constant (Ephemeroptera) or declined (Trichoptera; Plecoptera).3. As in 1984, dipper distribution in 1995 was related to acid‐base status: aluminium concentrations were significantly higher, and pH significantly lower, at sites where dippers were absent. In both surveys, there were significantly more bankside broadleaves and fewer conifers where dippers were present.4. Although dippers occurred in 1995 on a similar number of streams as in 1984, there were both gains and losses, and an overall significant reduction in the number of visits on which birds were recorded (= registrations) per survey reach. Gains and losses were not related to habitat structure or acid‐base status, and might be stochastic. Mean aluminium concentrations increased more (P < 0.06) at sites where dipper registrations fell, than where they increased, but changes were large enough to explain altered occupancy at only five sites. Plecopteran abundances declined most at sites losing birds.5. We conclude that recovery from acidification has not yet been large or sustained enough to allow widescale increase in the Welsh dipper population, and continued decline cannot be excluded. There is a need for better understanding of how recovery processes will permeate foodwebs to reach top predators in acidified streams, and of the geographical scale of recovery required to increase populations in dispersed organisms such as birds. Such uncertainties, together with the mismatch in trends revealed by stream chemistry and dippers, illustrate reasons why chemical data alone are inadequate for the assessment of changing river quality.