The results of generation mean analysis and standard errors for grain filling period in the three inter-varietal crosses of durum wheat under two dates of plantings revealed significant differences among all the twelve generation means indicated presence of genetic variability for this trait in the materials studied. The joint scaling test exhibited that subsequent fitting of different models viz., non-epistatic model (3parameter model), epistatic models assuming presence of digenic interactions (6-parametr model) and trigenic interactions (10-parameter model), 10-parameter model showed non-significant values of chi-square with maximum significant estimates of parameters in all the crosses under both the sowing dates except in the crosses HI 8062 x JNK-4W-128 and Raj 911 x DWL 5002 in normal sown environment (Table1). This indicated adequacy of this model to account the genetic variance among the generation means existing in the crosses. In the cross HI 8062 x JNK-4W-128, 3-parameter and 6-parameter models gave very high chi-squareof values in normal sowing. Subsequent fitting of 10-parameter model did not give probability higher than 0.001, revealed that even 10-parameter model did not fit to the data indicated involvement of more complex interactions or linkage in the inheritance of this trait. The various gene effects in this cross were, however, estimated following the trigenic interaction model, in view of the fact that the chi-square value for this model was the lowest. In the cross Raj 911 x DWL 5002 (normal sown) the chi-square value for 3-parameter was significant. Subsequent fitting of 6-parameter and 10-parameter models showed non-significant chi-square values, however, the probability of fitness of the 6-parameter model was the highest along with maximum significant parameters. Therefore the 6-parameter model was considered more appropriate to explain the genetic variation among the generation means existing in this cross in normal sown condition (Table 1). These results clearly indicated that epistatic interactions had a greater role in controlling the inheritance of grain filling period in all the crosses under optimum and sub-optimum sowing environments. The results of the cross Raj 911 x DWL 5002 in late sown environment further exhibited that environment played a greater role to the expression of different non-allelic interactions in such a way that a significant contribution for trait changed drastically in changing the environments. As a matter of fact that when sowing is delayed, the genotypic expression is affected hence, the possibility is that the true phenotypic difference are not resolved leading to observation of non-significant differences between the different genetic parameters estimated in normal and delayed sowing. In such situations the estimates made under late sowing are found to be less or non-significant in comparison to the normal environments unless the parents involved are specially selected for the targeted environment. It was also noted in the cross HI 8062 x JNK-4W-128 that changes in planting dates (normal to late) somehow reduce/break the complex genetic barrier of different gene effects in late sown condition.

The analysis of gene effects revealed that both additive (d) and dominance (h) gene effects were significant in the crosses HI 8062 x JNK-4W-128 and Raj 911 x DWL 5002 under both the sowing dates. However, in the cross Cocorit 71 x A-9-30-1 only additive (d) gene effect was significant in late planting. In the present study relative magnitude and direction (signs) of the main effects [(d) and (h)] changed in change in cross as well as sowing environment. Results further revealed that dominance (h) had a higher magnitude in most of the cases than additive (d) gene effect indicated preponderance of dominance (h) gene effect in the expression of this trait. A perusal of the results of digenic epistatic interaction [(i), (j) and (l)] revealed that additive x additive (i) and dominance x dominance (1) were equally and significantly contributed towards the inheritance of the trait in most of the crosses under both the planting dates, however, magnitude of the dominance x dominance (1) was observed higher. Additive x dominance (j) digenic interaction was not significantly contributed so frequently towards the genetic control of this trait( Table 1). The results of trigenic epistatic interactions [(w), (x), y) and (z)] showed that all the interactions played a significant role in controlling the inheritance of this trait in the cross HI 8062 x JNK-4W-128 in both sowing conditions except additive x dominance x dominance (y) in late sowing. In the cross Cocorit 71 x A-9-30-1 additive x additive x dominance (x) and additive x dominance x dominance (y) in normal planting, whereas in late planting additive x additive x additive (w) and additive x dominance x dominance (y) were responsible in governing the inheritance of this trait. In the cross Raj 911 x DWL 5002 additive x dominance x dominance (y) and dominance x dominance x dominance (z) were significantly contributed in the inheritance under late sowing. These results further confirmed that non-allelic interactions played a significantly greater role than the additive (d) and dominance (h) gene effects in the genetic control of inheritance of this trait in durum wheat. Earlier studies (Singh et al. 1977; Edward et al. 1976; Dasgupta and Mondal 1989; Nayeem 1994; Sood and Dawa 1999) also confirmed the role of epistatic interactions in the inheritance of grain filling period in wheat.