A tentative theory of age-discrimination is indirectly tested through labor statistics concerning the duration of unemployment. After first reviewing the inadequacies and inapplicability of sex and race discrimination hypotheses derived from the conventional economic models, this analysis synthesizes statistical models of discrimination to explain the negative correlation between age and the duration of unemployment. Under a quasi-experimental model for the varying levels of overall economic activity over the period studied, a statistical explanation is provided for the relative stability in age-related unemployment duration differences and a paradoxical relationship to the general unemployment rate. (Author)