Models of the agricultural household have traditionally relied on assumptions regarding the complementarity or substitutability of family labor inputs. We show how data on time allocations, health shocks and corresponding treatment choices can be used to test these assumptions. Data from Tanzania provide evidence that complementarities exist and can explain the pattern of labor supply adjustments across household members and productive activities following acute sickness. In particular, we find that sick and healthy household members both shift labor away from self-employment and into farming when the sick recover more quickly. Infra-marginal adjustments within farming activity types provide further evidence of farm-specific complementarities.