Regression analysis applied to the time series productivity data of 114 assembly workers in an electronics firm, covering a period of 152 weeks, suggests that time-dependent individual productivity changes in the post-learning or ‘drift’ phase can be adequately represented with a quadratic time series model. Some 70% of the operators studied are shown to have a maximum shaped curve, so that the long-term drift in production level is downward and thought to be due to a decline in effort. The methodology used suggests that long-term trends of individual operator performance can be readily predicted from work records, thus highlighting the need for job redesign and other motivational aids. The results presented here also provide some support for the hypothesis that the particular form of this model is indicative of the dominant influence of antecedent changes in work attitudes.