Possible changes in streamflow associated with logging were analyzed for 23 western Washington catchments with drainage areas from 14 to 1600 km2. Statistically significant trends in annual streamflow minima, uncorrected for climatic influences, are all decreasing and are apparently dominated by a regional climate signal associated with the Pacific Decadal Oscillation, rather than land cover change. Using paired catchment analysis, the number of statistically significant trends detected for the peak flow series is largely within the range of statistical noise. Only in the case of the annual minima were more trends detected than could be attributed to change, owing in part to the lower relative variability, hence greater detectability of trends in low flows. Investigation of the effect of return period on peak flow changes shows an apparent increase in flood peaks for treatment relative to control catchments, the mean magnitude of which decreases with increasing return interval up to about the 10-year return period. In large part, owing to the small number of catchment pairs available, this analysis cannot be considered conclusive. An alternative approach to evaluating trends in peak flows based on time series residuals of observed flows from hydrology model predictions detected increasing trends in peak flow series, which were largely absent in the paired catchment analysis. This is attributed both to the ability of the model, which acts as the control, to filter out natural variability and to a larger trend signal in the residuals analysis resulting from the ability of the method to fix the vegetation condition in the model control. ¿ 2000 American Geophysical Union