The experimental material generated from six diverse parents, comprised three crosses namely, Cocorit 71 x A-9- 30-1, HI 8062 x JNK-4W-128 and Raj 911 x DWL 5002. One of the parents (A-9-30-1, JNK-4W-128 and Raj 911) in each cross had higher number of spikelets per ear. Twelve basic generations viz, two parents, F₁ and F₂, first backcross generations with both parents (BC₁ and BC₂), where BC₁ was the cross between F₁ x female parent and BC₂ was F₁ x male parent, their selfed progenies (BC₁ F₂, BC₂ F₂) and second backcross generations ie the BC₁ and BC₂ plants again crossed with both original parents (BC₁ x female parent; BC₁ x male parent and BC₂ x female parent; BC₂ x male parent). All these populations were raised together in randomized block design with three replications at 30 cm x 15 cm spacing under normal and late sown environments in the same cropping seasons at research farm of Rajasthan Agricultural University, Agricultural Research Station, Durgapura, Jaipur. Each parent and F₁ generations were sown in 2 rows, each backcross generation in 4 rows and F₂ and the second cycle of backcrosses in 6 rows of 5 m length. Number of spikelets per ear was recorded on 15 random plants in each parent and F₁, 30 plants in each backcross generations and 60 plants in each F₂ and second backcross generations in both environments.

The data of each population in both environments were analyzed separately by joint scaling test of Cavalli (1952) to determine the nature of gene action. Components of heterosis in the presence of trigenic interactions were calculated as suggested by Hill (1966).

Results and discussion

Significant differences were observed among generation means for spikelets per ear in all the three crosses in both the sowing environments, which revealed the presence of genetic diversity for this attribute in the material. On the basis of different models fitted to data, the joint scaling tests revealed the presence of epistatic interaction in both the environments except in the cross Cocorit 71 x A-9-30-1 under normal sown, where additive-dominance model was fitted to the data. In all other cases, 10-parameter model was adequate to explain genetic variation in different generations in both the sowing environments except in the cross Raj 911 x DWL 5002 under normal sowing where even this model was not adequate, indicating the more complex genetic systems involved in controlling this trait. However, looking to the low value of chi-square and very low probability of 10- parameter model, the different gene effects were estimated according to this model (Table 1).

The analysis of gene effects revealed that the additive (d) effect was only significant in the first cross (Cocorit 71 x A-9-30-1) under normal sowing conditions, indicating that this character was simply inherited and can be improved by simple breeding method such as progeny selection. Additive (d) effects were also found significant in the second cross (HI 8062 x JNK-4W-128) in both the sowing environments only. However, dominance (h) was only found significant in the third cross (Raj 911 x DWL 5002) under normal sowing. Both digenic and trigenic interactions had important role in controlling the inheritance of this trait in the cross HI 8062 x JNK-4W- 128. Whereas, in the first and third cross under late sowing only trigenic interaction additive x additive x additive (w) and additive x dominance x dominance (y) control the inheritance of trait studied, respectively. However, none of the digenic and trigenic interactions were found significant in the cross Raj 911 x DWL5002 under normal sowing, confirmed that more complex type of non-allelic interactions or linkages are involved to control spikelets per ear (Table 1). Magnitude and directions (signs) of epistatic interactions were depended upon the material and sowing environments.

Results of the absolute totals of epistatic effects (Table 2) clearly indicated that second order interactions had higher value than the first order interactions and the main effects in all the cases in both the sowing environments except in the first cross where additive dominance model was fitted to the data. Thus, it is clear that epistatic effects particularly trigenic effects contributed maximum to control the inheritance of this trait. Earlier Nanda et al. (1981) reported that additive-dominance model was sufficient to explain genetic variance among the generation means for this trait in aestivum wheat whereas additive x additive (i) and dominance x dominance (l) variance along with additive variance were reported by Singh and Anand (1971), which had significant role to control this trait. Type of the epistasis could not be ascertained as either (h) or (l) or both were non-significant in each case.

Furthermore, results indicated that the absolute totals of the non-fixable gene effects were higher than fixable in almost all cases except in the cross Cocorit 71 x A-9-30-1 (under normal sown), where fixable gene effect (additive) had important role. Dasgupta and Mondal (1988), Raghuvanshi et al. (1988) and Patil et al. (1997) also reported the predominant role of additive gene effects in the expression of this trait. Epistatic interactions shared a major portion of the non-fixable effects in both the environments. Earlier Maloo (1978), Reddy (unpub), Verma (1981), Saini (1987), Kathiria (1991), Mann et al. (1992) and Khedar (1998) also observed the significant role of non-additive gene effects to control this trait.