The annual rate r of tag loss may be estimated from observations on the proportion of tag retentions among recoveries of fish which were both tagged and permanently marked at the time of release. In samples recovered t years after release, the proportion of marked fish still bearing tags is assumed to be an estimate of rt. Homogeneity of these t-year loss rates for different year classes may be tested by Chi-square, and the information from all samples may then be combined in the form of a maximum-likelihood estimator of r. Finally, a goodness-of-fit Chi-square may be calculated to test the assumption of a constant annual rate of tag loss. When applied to recapture data on lake whitefish in Georgian Bay, these tests failed to detect any errors in this simple model.