GCA vs SCA in Diallel
General Combining Ability (GCA) and Specific Combining Ability (SCA) are the two key concepts that come out of a diallel mating analysis. GCA is the average performance of a parent in crosses with all other parents. SCA is the deviation of a particular cross from what its two parents' GCAs would predict.
GCA, SCA and reciprocals
Sprague and Tatum (1942) introduced the partition. GCA captures additive genetic variance; it indicates whether a parent is generally good in crosses. SCA captures non-additive variance (dominance and epistasis); it indicates whether a specific combination is better or worse than expected. Reciprocal effects, where the cross P_i x P_j differs from P_j x P_i, capture maternal and cytoplasmic contributions and are estimable in Griffing methods 1 and 3 (which include reciprocals). Methods 2 and 4 omit reciprocals.
Variance components
Sigma squared GCA estimates one quarter of the additive genetic variance. Sigma squared SCA estimates the dominance variance. The narrow-sense heritability of yield in a population of related lines can be approximated from the GCA / SCA ratio. A high GCA-to-SCA variance ratio means selection on parents (general performance) will work; a low ratio means specific cross combinations matter and pedigree breeding or hybrids are more efficient.
Reading combining ability tables
Diallel output gives a GCA estimate per parent and an SCA estimate per cross. Both come with standard errors. Test each effect against zero with a t-test. A parent with a positive significant GCA for grain yield is a good general combiner. A cross with a positive significant SCA for grain yield is a specific combination worth advancing as a hybrid. Negative significant SCAs are also informative; they mark combinations to avoid.