DETECTION AND EFFECTS ON MORTALITY ESTIMATES OF CHANGES IN TAG LOSS

When estimates of mortality are obtained from tagging experiments the rate of tag loss is usually assumed to be constant. Our analysis shows that if the rate of tag loss increases over the time after tagging an d is measured over a short period, it will be seriously underestimated when extrapolated to a longer period, with a corresponding overestima te of mortality. We describe a method for testing the assumption of co nstant tag loss on single-tagged animals, using a multiple release of tagged animals combined with recaptures or ''searches'' at a number of later times. An experiment of this type using polyethylene shellfish tags attached with cyanoacrylate (''superglue'') to the limpet Cellana tramoserica (Sowerby) suggested that the rate of tag loss increased w ith increasing time after release. We recommend that the assumption of constant tag loss should always be tested using the method described, or all animals should be double tagged so that mortality can be estim ated free of bias even if tag loss is not constant.