The results of generation mean analysis further exhibited that absolute totals (ignoring signs) of second order interactions [(w) + (x) + (y) + (z)] were much higher than the main effects and first order interactions [(i) + (j) + (l)] in all three crosses under both the sowing dates except in the cross Raj 911 x DWL 5002 in normal planting (Table 2). In this cross first order interactions contributed maximum than the main effects. These results of this study confirmed the greater role of non-allelic interactions in the genetic control of the inheritance of grain filling period in durum wheat. Experimental results showed that parameters (h), (l) and (z) were significant in the cross HI 8062 x JNK-4W-128 in both the sowing dates. The relative signs of (h) and (z) were in same direction but opposite to (l) in normal sowing whereas in late sown the relative signs (h) and (l) were in same direction but opposite to (z), which indicated that all the three genes were showing duplicate type of epistasis. The parameter (h) was observed significant negative and (l) was significant positive in the cross Raj 911 x DWL 5002 in normal planting indicated duplicate type of epistasis at digenic level (Table 1). Conclusion regarding any type of epistasis could not be drawn in other cases because among (h), (l) and (z) one or the other parameter was found non-significant.

A perusal of absolute totals of fixable [(d) + (i) + (w)] and non-fixable [(h) + (j) + (l) + (x) + (y) + (z)] gene effects revealed that in all the crosses non-fixable (non-additive) gene effects were many times higher than the fixable (additive) gene effects indicated greater role of non-additive gene effects and hence, the requirement of complicated procedure of breeding for their further exploitation is needed (Table 2). Other studies (Prabhu and Sharma 1984; Singh and Rana 1987; Raghuvanshi et al. 1988; Srivastava et al. 1992 and Singh 2002) also led to similar conclusion.