G.C. Williams's 1957 hypothesis famously argues that higher age-independent, or ‘extrinsic,’ mortality should select for faster rates of senescence. Long-lived species should therefore show relatively few deaths from extrinsic causes such as predation and starvation. Theoretical explorations and empirical tests of Williams's hypothesis have flourished in the past decade but it has not yet been tested empirically among humans. We test Williams's hypothesis using mortality data from subsistence populations and from historical cohorts from Sweden and England/Wales, and examine whether rates of actuarial aging declined over the past two centuries. We employ three aging measures: mortality rate doubling time (MRDT), Ricklefs's ω;, and the slope of mortality hazard from ages 60–70, m′50–70, and model mortality using both Weibull and Gompertz-Makeham hazard models. We find that (1) actuarial aging in subsistence societies is similar to that of early Europe, (2) actuarial senescence has slowed in later European cohorts, (3) reductions in extrinsic mortality associate with slower actuarial aging in longitudinal samples, and (4) men senesce more rapidly than women, especially in later cohorts. To interpret these results, we attempt to bridge population-based evolutionary analysis with individual-level proximate mechanisms.