Among digenic interactions additive x additive (i) and dominance x dominance (l) were contributed maximum towards in controlling the inheritance of this trait under normal and late sown environments, respectively. Among trigenic interactions most of the interactions were found significant in the cross of Cocorit71 x A-9-30-1, whereas only additive x dominance 3K dominance (y) found significant in the cross Raj911 x DWL5002 under normal sown condition (Table 1). Absolute totals of epistatic effects revealed that epistatic effects were much higher than the main effects in all the cases except where additive-dominance model was fitted to data. These results further confirmed that the non-allelic interactions, such as first order and second order had somehow more important though the additive (d) and dominance (h) were also contributed significantly in controlling the inheritance of trait studied (Table 2). Duplicate type of epistasis was observed in the cross Cocorit71 x A-9-30-1 and HI8062 x JNK-4W-128 under late sown conditions only.

Results of the study further revealed that the absolute totals of the non-fixable effects were higher than the fixables in all the cases except in the cross Raj 911 x DWL 5002 in late sown condition signifying the greater importance of non-additive gene effects, a major portion of which was shared by epistatic effects (Table 2). Earlier, Widner and Lebsock (1973), Ram et al. (1977), Bhatia et al. (1979) and Srivastava et al. (1982) in durum Wheat and Singh and Rana (1989) and Joshi, (1997) in bread wheat reported that both additive and non-additive gene effects were important in the inheritance of grain yield per spike.

Analysis of components of heterosis revealed that epistatic interactions had important role to cause significant and positive heterosis in the cross Cocorit71 x A-9-30-1 whereas non-additive gene effect causes heterosis in the Cross H18062 x JNK-4W-128 under normal sown condition. Most of the trigenic interactions [(x), (y) and (z)] and only additive x additive (i) digenic interaction in the former cross and dominance (h) component in latter cross, contributed maximum towards significant heterosis. Absence of significant heterosis in remaining cases could be explained due to the internal cancellation of heterosis components (Table 3). Significant inbreeding depression in all the three crosses in normal sown condition, showed reduction in F₂ over F₁ due to the dissipation of non-additive dominance effects or epistatic effects involving dominance in F₂. Thus, it is confirmed that non-additive genes had a significant importance in durum wheat. Earlier, Sharma et al. (1986) and Dasgupta and Mondal (1988) reported that non-additive gene actions were the responsible for significant heterosis and inbreeding depression in grain yield per spike.

This study further revealed that varied results were recorded in different crosses and environments regarding the relative importance of gene action in controlling the inheritance of grain yield per spike in durum wheat. Such variations may arise due to the differences in the parental material of investigation and the differences caused by sowing environments.

However, in view of high magnitude of inter-allelic interactions of digenic and trigenic level for most of the cases, it was suggested that wheat breeder should follow such methods, which can mop-up the genes, to form superior gene constellations interacting in a favorable manner. The breeding methods suggested to achieve this objective are restricted recurrent selection (Joshi 1979) and / or diallel selective matting (Jensen 1978). The presence of considerable additive variation with or without fixable gene effects for this trait (HI8062 x JNK-4W-128 and Raj911 x DWL5002 under normal and late sown condition, respectively) indicated that simple procedures like single seed descent, bulk method of breeding, and their various modifications should be rewarding in obtaining superior recombinant pure breeding lines in durum wheat. Thus, it is clear from the study that suitable environment should be selected for rewarding superior recombinant in a favorable environment with suitable material for desired improvement in this trait for further tangible advancement of grain yield in macaroni wheat in future.