Yield efficiency of cocoa, defined as yield over a period of time divided by the increment in trunk cross-sectional area over that period, was estimated in four factorial matings in Sabah, Malaysia, evaluated at a single planting density. The parents were 18 seedlings of Upper Amazon Forastero type, broadly representative of Peruvian Amazon material, and 15 clones of Trinitario type, selected from cultivated cocoa in Ecuador, Trinidad and Papua New Guinea. Both the Upper Amazon and Trinitario parents varied in general combining ability (gca) for yield efficiency, with strong evidence of additive inheritance of the trait. Some of the Upper Amazon parents showed higher yield efficiency than many of the Trinitarios. Correlations between gcas for yield efficiency and yield were inconsistent. This is thought to be because the parents differed markedly in optimal planting density, which is higher for many of the Upper Amazons with high gca for yield efficiency than it is for some of the Trinitarios. At the current stage of cocoa breeding, selection for adaptation to planting density is a higher priority than selection for yield efficiency.