Abstract
Analyses of regressions of atlas bone width against fish length (mixed regressions) of Atlantic salmon Salmo salar can be used to infer the parr:smolt ratio of the source populations. Analyses of mixed regressions have been used to infer that, contrary to previous reports, sawbill ducks Mergus spp. eat only a few smolts. Here a Monte Carlo simulation model is presented which quantifies the effect of variation in the smolt:parr ratio on the form of the mixed regression. Based on published mixed regressions, the model predicts 95% CI of 1–36% and 1–42% smolts in diets of sawbill ducks, which are broadly consistent with results of earlier studies. However, other large inconsistencies still exist between results of previous studies and these are briefly discussed.