Although some of the components of egg-size variation in birds have been studied, there is a lack of approaches in which phenotypic variation is both partitioned and its causes are analyzed. We partitioned and analyzed the phenotypic variation in egg size in 1588 eggs from 572 clutches of the Imperial Cormorant (Phalacrocorax atriceps) over three breeding seasons (2004–2006) at Punta León, Argentina. We estimated repeatability and phenotypic plasticity of egg size, investigating the effects of year, date of laying, body size, and body condition on egg-size variation within and between clutches. Egg size varied widely, the largest egg being >2× as heavy as the smallest. The repeatabilities of both egg size and mean egg size (0.761 and 0.894, respectively) are among the highest reported for any bird. Most variation among clutches was due to differences among individuals, being weakly related to date of laying and unrelated to year, body size, or body condition. Egg size decreased with the egg's order. This general pattern was not related to year, date of laying, body size, or condition. Proximate constraints did not explain variation either within or among clutches. There is no obvious adaptive benefit of intraclutch variation because the effect in brood reduction of intra-clutch variation in egg size was negligible. However, egg size was positively related with the survival time of the last (third) chick. Therefore, investing in a large third egg should benefit females of the Imperial Cormorant, a brood reducer, by keeping the last chick alive longer.