For 50 years, ‘hybrid vigour’ in crosses involving genotypes collected from the centre of diversity of the crop in the Peruvian Amazon has been the guiding principle of cocoa breeding. In the work described here, six Amazonian genotypes and one Trinitario clone were used to produce four bi-parental crosses. Twenty-two plants of each cross were propagated from plagiotropic buds as ‘cloned seedlings’ and evaluated for growth, yield and yield efficiency (the ratio of yield to continuing vegetative growth) in replicated trials in comparison with the parental clones and the reproduced seedling family. The reproduced seedling families grew faster than and outyielded all the clones in all four crosses, although they showed no gain in yield efficiency. The vigour and yield of the ‘cloned seedlings’ of the three inter-Amazon crosses exceeded mid-parent value (MPV), but not the single Trinitario cross. The pattern of departures from MPVs was inconsistent with current understanding of the genetic distance between the parents. The superiority of the seedlings over the ‘cloned seedlings’ is interpreted as an epigenetic effect. The results suggest that simple recurrent selection is an appropriate breeding strategy in cocoa, with different optimum planting densities for the seedling and clonal phases. It is concluded that if the concept of very large ‘hybrid vigour’ did not already exist in cocoa, it would not be adopted on the evidence of this work, the first in which clones were compared to their progeny grown as clones.