Abstract.Simple body measurements were used to predict the fat content of Atlantic salmon parr. For three different sizes (wet weights 18–98g, 44–207g and 137‐151·8g) at two times of year (November/December and July) the percentage body fat (as measured by lipid extraction) could be predicted accurately using multiple regressions based on a combination of the following measurements: wet weight, fork length, opercular height, dorsal height, opercular width, dorsal width, anal width, adipose fin length and condition factor. In every case predictions based on multiple regression equations were more accurate than condition factor alone. This biometric method is easily employed and can potentially be used both in biological studies and on fish farms to assess body condition of individual fish. Since it is non‐destructive it has the additional advantage of allowing repeated monitoring of fat levels in the s