The aim of this investigation was to study the nature of genetic constraints on life-history evolution in a natural population. To identify possible constraints, estimates of variance and covariance components were obtained in the laboratory for a range of life-history characters in a population of 'Tribolium castaneum' three generations after collection from the wild. Estimates were obtained by means of a diallel analysis. To test predictions about genetic constraints based upon the variance-covariance matrix, selection was conducted for duration of development in both directions for six generations followed by a phenotypic assay of all life-history traits in all lines. After six generations of selection, there were significant differences between selected lines for both males and females for duration of development and growth rate, but no significant correlated responses in other characters. Some correlated responses, such as a decrease in late life egg production in slow developing lines approached significance and, if the selection program had been continued, may have become significant. ... All fecundity indices were positively correlated but genetic correlations of less than unity and the decrease in late life egg production in slow developing lines suggest that the reproductive schedule is amenable to modification by selection. Depression of late life fecundity in slow developing lines was not expected as late life fecundity had no heritable variation, and had low and insignificant positive correlations with duration of development. The depression may have been due to disruption of developmental processes as a result of selection for duration of development. It is possible that patterns of development characteristic of a taxon may restrict the range of future adaptations in that taxon. However, further investigation is needed before conclusions about this depression of fecundity can be drawn. Experimental results do not support theories about the evolution and/or maintenance of senescence by the mechanisms of antagonistic pleiotropy between early and late life history characters or the accumulation of deleterious mutations. Results do support the hypothesis r, that senescence is due to a running out of genetic program for internal repair and maintenance capabilities in organisms. Whilst it is recognised that genetic variation and covariation are dependent upon environment, and thus genetic constraints may differ in different environments, studies which combine a quantitative genetic analysis of a population with a selection program carried out in a single environment are still useful. Firstly, they identify how the genetic variance-covariance matrix affects responses to selection in a specific environment. Secondly, they can help to identify possibly fundamental and universal constraints. Negative genetic correlations reflecting physiological trade-offs may be operational in all "realistic" environments. Studies would need to be repeated in a number of different environments to confirm the identity of these universal genetic constraints.