In the analysis of variance (ANOVA) the F test was used at the 5% and 1% levels of probability. The trait means were compared by the Tukey test at the 5% level of probability. The ANOVA for the experiment (RCBD) was estimated according to [11] procedure. The combining ability analysis for studied traits was carried out using method 2 of model 1 of [12], where parents and F₁’s were included under the experiment excluding reciprocals.

Griffing’s Method 2 and model 1 is as follows:

X i j = u + g i + g j + s i j + ( 1 / b ) k e i j k ,

where u = the population mean; g_i = the general combining ability effect of the ith parent; g_j = the general combining ability effect of the jth parent; s_ij = the specific combining ability effect of the cross between ith and jth parents such that s_ij = s_ji; e_ijk = the environmental effect associated with ijkth observation.

Thus the experimental mccaterial comprises of n ( n + 1 ) / 2 genotypes through half diallel mating i.e. 28 genotypes (7 parents and 21 its F₁ (hybrids). The genetic components of variation were calculated according to Griffing’s numerical approach and graphical analysis (Wr-Vr graph) given by [13] [14] [15].

The GCA and SCA effects were estimated according to [16] by the following formula:

GCAeffects ( G i ) = 1 n + 2 ∑ [ ( Y i . + Y i i ) − 2 n ( Y .. ) ] Restricted to ∑ i n g i = 0

SCAeffects ( S i j ) = Y i j − 1 n + 2 [ Y i . − Y i i + Y j . + Y j j ] + 2 ( n + 1 ) ( n + 2 ) Y i i     ( i < j ) .

For estimation of heterosis in each character the mean values of the 21 F₁’s have been compared with better parent (BP) for heterobeltiosis. Percent heterosis was calculated as:

H ( B P ) = F 1 ¯ − B P ¯ B P ¯ × 100

The significance test for heterosis was done by using standard error of the value of better parent as:

S E ( B P ) = 3 / 2 × M S E / r .

Using the values of different statistics computed earlier the variance components and allied parameters as follows:

A perusal of results of combining ability analysis indicated considerable nonadditive gene action in the inheritance of majority of the attributes studied. Hence, breeding methods involving selection, intermating of selects and reselection in segregating generations followed by pedigree method of breeding may help to improve the characters [7]. The values of mean sum of square for both GCA (general combining ability) and SCA (specific combining ability) were highly significant for yield in bottle gourd, which suggests the presence of both additive and non-additive genetic variance in the population (Table 1). But the higher magnitude of GCA compared to SCA indicated predominance of additive genetic variance. The GCA component is primarily a function of the additive genetic variance. GCA of parents plays a significant role in the choice of parents. A parent with higher positive significant GCA effects is considered as a good general combiner for yield. The magnitude and direction of the significant effects for the seven parents provide meaningful comparisons and would give indications to the future breeding programme.

In the present study, parent P6 showed the greatest relative GCA effect (1.75**) followed by P1 (1.07*), P2 and P7 (1.06*) for yield. Other parents showed either insignificant negative or significant negative GCA values. So, parents P6, P1, P2 and P7 were the best general combiners in crosses for the improvement of this trait. [17] [18] [19] reported that significant positive GCA effects obtained in bottle gourd. [19] also reported KBG-16 was the best general combiner for total yield per vine. While the parents P4 (−2.74**), P3 and P5 (−1.12*) were the poor combiner for yield in this study.

Out of 21 cross combinations 14 crosses showed positive SCA effect for yield, among them 12 crosses exhibited significant positive SCA effect (Table 2). The highest positive significant SCA effects were shown by the hybrid P3 × P4 (6.90**) followed by P3 × P5 (5.12**), P4 × P5 (4.22**), P4 × P7 (3.57**), P2 × P6 (3.40**) and P2 × P7 (3.52**). Thus P3 × P4 was the best combination (poor × poor combiner) similar to other two hybrids (P3 × P5, P4 × P5) for yield in bottle gourd. The other moderately higher values of positive SCA effect may be considered as good specific combiner viz., P4 × P7 (poor × good combiner) and P2 × P6, P2 × P7 (good × good combiner) for fruit yield. [19] also found 3 hybrids viz., GH-10 × G-2, GH-9 × PSPL and GH-13 × G-2 exhibited significant

*Significant at 5% level; ** Significant at 1% level.

*Significant at 5% level; **Significant at 1% level.

specific combining ability effects for fruit yield. [17] reported best SCA effect was obtained by 7 crosses and [18] reported best combiner was IC-92362 × Pusa Naveen in bottle gourd, which was in agreement with the present findings. [19] reported that the parents having poor GCA for certain traits when crossed with parents having high GCA for the same traits usually generated high positive SCA effect. Similar trends were also observed in the present study like P4 × P7. The parents like P4 having poor GCA for yield, when crossed with P7 having good GCA, for this character, gave higher positive SCA effects (P4 × P7). So a relationship seems to exist between general and specific combining ability effects, it would safely be assumed that good or poor GCA may provide good SCA combination.

Significant difference between genotype and replication was observed in ANOVA for fruit yield (Table 1). Among 21 combinations, 17 crosses showed significant positive better parent heterosis and it was ranged from 6.27% to 49.72%. The highest significant positive better parent heterosis was exhibited for the hybrid P3 ´ P4 followed by P3 ´ P5 (44.97%), P4 ´ P5 (39.47%). While [20] reported the yield per plant of F1 hybrids L12 ´ T2 (13.80%) and L9 ´ T1 (8.08%) manifested heterosis over top parent. [21] stated the highest heterobeltiosis for yield was recorded in cross PSPL ´ NDBG-1 (106.85%). [22] got 3 best F1 hybrids (S46 ´ S54, S10 ´ S52-7 and S54 ´ S52-7) showed 84.5%, 80.5% and 80% heterosis, respectively, for yield over the best parental line, S41. [21] stated three best performing F1 hybrids LA-46 ´ LA-99, LA-12 ´ LA-76 and LA-32 ´ LA-37 also showed 51.8 and 125.9, 81.8 and 121.1 and 99.1 and 200.1% higher yield over the better parent in Kharif and summer seasons, respectively. [22] reported the best performing hybrids for yield were S36-1 ´ NC59812-1 and S39-1 ´ S1-3, with 76.4% and 58.1% heterosis over better parents, respectively. For fruit yield the hybrids ABGS11-23 ´ DBG-5, ABG-1 ´ Arka Bahar and DBG-6 ´ DBG 5 recorded the highest estimates of heterosis over standard check [6]. The yield per plant exhibited appreciably high amount of heterosis over the better parent, top parent and commercial control [24]. These were in agreement with the present findings (Table 3).

The components of variation along with the derived genetic ratios for fruit yield (Table 4) showed that the D and H components, which measure additive and dominance variation, respectively were significant. This indicated the importance of both additive and dominance components for the inheritance of all the genotypes in bottle gourd. However, the magnitude of dominance was higher than the additive component. These results agree with that reported by [25]. The H₂ representing dominance deviation due to relative frequency of positive and negative genes was significant. The proportion of positive effects as indicated by F value was non-significant for fruit yield, suggesting greater frequency of dominant alleles governing this character. The net dominance effect, obtained by

*Significant at 5% level, **Significant at 1% level.

the estimate h² expressed as the algebraic sum over all loci in heterozygous condition in all crosses, was significant. This revealed that substantial contribution of dominance effects was due to heterogeneity of loci. The environmental component E exhibited significant values, indicating influence of environmental factors in the expression. However, the magnitude of E was definitely lower due to no value than the respective value of D and H₁. This also indicated that the characters were influenced less by the environment.

The average degree of dominance as indicated by the proportion (H₁/D)^1/2 was more than unity, suggesting that over dominance was operating in the expression for fruit yield of bottle gourd. [26] [27] also found over-dominance in graphic analysis. The ratio of H₂/4H₁ provides an estimate of the average frequency of positive and negative alleles in all the parents. A value of this ratio smaller than 0.25 studied suggested asymmetrical distribution of only the negative alleles. The ratio of [(4DH₁)^1/2 + F/(4DH₁)^1/2 − F] estimates the relative proportion of dominant and recessive alleles in the parents. In the present study the ratio for the yield was greater than unity, suggesting excess of dominant alleles and minority of recessive alleles i.e., asymmetrical distribution for dominant alleles in the parents. The ratio of h²/H₂ estimates the number of gene or group of genes which control the character exhibiting the degree of dominance. The h²/H₂ values for yield per plant were 2.40, indicated the involvement of 2 to 3 genes or gene groups in the inheritance of this character as per ratio which are responsible for genetically control. Heritability in narrow sense was higher for fruit yield, indicating that this character was highly heritable. These results agree with that reported by [25]. Predominance of non-additive gene action for number of seeds per fruit and hundred seed weight is evident from the higher estimate of SCA variance component than GCA variance component [28] [29].

Graphical analysis of parent-offspring covariances (Wr) on array variances (Vr) is shown in Figure 1. It was observed from the Wr/Vr graph that the slope of the regression line for fruit yield was significantly below 1.0 (0.36 ± 0.11), suggesting

significant non-allelic interaction for this character. The regression line intersected the Wr axis below the origin, suggested over dominance in addition to the interaction.

In Figure 1, all the Wr, Vr points fell within the boundary of the limiting parabola and the parents also clustered into three distinct groups on the regression line showing diversity in the parents for this trait. [30] reported the top five high yielding crosses showed overdominance for fruit yield per plant. The relative values of Wr and Vr showed that the parent P6 had the lowest value and hence contained the most dominant alleles while the parent P4 had the highest Wr, Vr values and have the most recessive alleles. The other parents fell in between suggested that the equal frequency of dominant and recessive alleles. Similar findings were reported by [31] [32] in bean and [33] in eggplant.