Investigation of time trends in contaminant levels in fish stocks has been of primary interest to the International Council for the Exploration of the Sea (ICES) since about 1974. Conventional analysis of covariance (ANCOVA) has been widely used for analysing data on contaminant levels and lengths (both transformed to logarithms), measured on individuals from a number of years of collection. It is an inadequate procedure when slopes (coefficients of regression of contaminant concentration on length) for separate years and residual variances about these separate regressions are not equal for all years. By employing a weighted procedure this methodology is extended to analyse such data sets where homogeneity of regression coefficients and residual variances is not a requirement. The procedure is illustrated on Belgian cod and flounder data.