The effects were studied of four leaf development stages (LDS) and three durations of incubation (DI) on the accuracy of leaf-disc tests on eight cacao (cocoa) clones (C) for predicting field resistance to phytophthora pod rot caused by Phytophthora palmivora. The clones were known to possess different general combining abilities (GCA) for pod resistance in the field, evaluated monthly at harvest over a 9-year period. Disease severity (DS) was affected strongly by C, DI and LDS, with increasing levels of significance. Two- and three-way interaction effects were smaller than the clone effect, but still significant. Clone effects were most significant for LDS3 (i.e. leaves 50-60 days old) and for DI5 and DI7 (observations made 5 and 7 days after inoculation, respectively). Coefficients of rank correlation between DS and field results were significant for seven of the 12 treatments, with highest values obtained again for treatments LDS3/DI5 (r = 0.87) and LSD3/DI7 (r = 0.93). Pooling of data for different LDS and DI treatments did not further improve the correlation with field results. However, these correlations were improved (from an average of 0.74 to 0.88) when the GCA values for field resistance were based on weekly observations, carried out in one year, including losses of pods and cherelles. It was concluded that, when carried out in a standardized manner and under optimal conditions, the leaf-disc test may explain 75-90% of the genetic variation for field resistance of cacao genotypes to P. palmivora.