Abstract Variation in offspring size with female size and other aspects of the maternal phenotype is commonly observed and taxonomically widespread. However, life history theory predicts that optimal offspring size should not depend on maternal size or total reproductive effort. This incongruity persists despite various modifications to theory, that nonetheless, either are limited in their applicability or fail to alter the prediction of fixed offspring size. I demonstrate that the persistence of this theoretical outcome stems from an ideal assumption that reproductive effort relates only to direct material costs, and therefore, equal or proportional to clutch mass or the product of offspring size and number. A major innovation in my study is to explicitly distinguish between direct and overhead components of the costs of reproduction. When overhead energetic costs of reproduction are explicitly incorporated, I readily obtain variation in optimal offspring size with maternal phenotype. This consequence of overhead costs of reproduction has not been demonstrated before. I identify functional forms of such overhead costs that facilitate variation in optimal offspring size. In particular, costs that are more sensitive to offspring size than to offspring number are most effective in causing variation in offspring size. The novelty of the model lies in succeeding to resolve the above incongruity both within the framework of traditional models of optimal offspring size and within more dynamic description of the lifecycle (addressing simultaneously both offspring and maternal performance), including stochastic effects, difference between reserves and structural components of size, and distinction between starvation and extrinsic mortality. My predictions explain several patterns of variation in size and body composition of offspring, with respect to both environmental conditions and maternal phenotype.