INTRODUCTION Pearl millet [Pennisetum glaucum (L.) R. Br.] is a diploid (2n = 14) species in the poaceae family. It is a fascinating dual-purpose crop that can be cultivated as a cereal or as fodder and possess a wide range of agronomically important traits that are adaptable to various agroecological conditions. It is commonly produced in marginal lands with low rainfall during the Kharif season (Vidyadhar et al., 2007; Bhoite et al., 2008) because of its ability to withstand drought and adverse agro-climatic conditions. The current production potential of pearl millet green fodder is relatively poor (Shashikala et al., 2013). Multi cut pearl millet varieties with excellent fodder yields must be developed to boost animal production (Harinarayana et al., 2005). Genetic improvement for quantitative traits in fodder pearl millet is dependent on the degree of variability for the desired traits in the base material, as well as the degree to which the desirable traits are heritable, and plays a significant role in selecting superior cultivars. GFY is a complex trait that is the result of activities and interactions of various traits; thus, understanding the relationship between yield and its components is all the more important. Understanding the pattern of variability, the inheritance of target traits, and the nature of character association is required for effective breeding strategies and successful selection for developing superior cultivars. In keeping view of above, the present study aimed to assess the magnitude of variability, inheritance pattern, and character association among contributing traits in 121 pearl millet accessions. MATERIALS AND METHODS The experiment was conducted during the year of 2019 at new area farm of Department of Forage Crops, Centre for Plant Breeding and Genetics, Tamil Nadu Agricultural University, Coimbatore. The experimental material included 121 pearl millet genotypes which were evaluated in Randomized Block Design (RBD) with two replications. Each pearl millet accession was grown in two rows of four metres each with a 45 cm x 15 cm spacing, which was maintained by thinning/transplanting at the 3 leaf stage. To raise a healthy crop, the entire suggested agronomic package of procedures was followed. In each replication, five competitive plants were chosen at random from each genotype, avoiding border plants. Plants to be selected were tagged before initiation of ear emergence for recording the observations at the time of 50 per cent flowering on 20 quantitative and qualitative traits viz., plant height, number of leaves per tiller, number of tillers per plant, number of nodes in main tiller, leaf width, leaf length, stem girth, stem weight, leaf weight, leaf to stem ratio, crude fibre, neutral detergent fibre, acid detergent fibre, crude protein, ash, crude fat, dry matter content, dry matter yield and green fodder yield per plant. The Australian Fodder Industry Association (AFIS) laboratory procedures manual (http://www.afia.org.au) was used to estimate crude fibre, acid detergent fibre, neutral detergent fibre, crude protein, crude fat, and ash. Plant samples were taken at the time of panicle initiation, chopped, air dried, and then oven dried for two days at 60°C. The samples were crushed and sieved using various sieve sizes after being dried in the oven. Sieve size of 0.5 mm was used for estimation of neutral detergent fibre, acid detergent fibre, crude fibre, ash and 0.1 mm sieve size was used for crude protein and crude fat estimation. Crude fat by soxhlet method using petroleum ether and crude protein was estimated by Kjeldahl’s method while crude fibre, acid detergent fibre and neutral detergent fibre was estimated by using crude fibre apparatus by following the method made through in AFIS laboratory methods manual. The genetic parameters were computed using mean values, and data was statistically analysed for each character as reported by Panse and Sukhatme (1967). Burton (1952) suggested estimating genotypic and phenotypic coefficients of variation, while Falconer (1964) and Dewey & Lu (1959) suggested calculating correlation coefficients and path coefficient analyses, respectively. RESULTS AND DISCUSSION A. Genetic Variability The development of an effective breeding programme is dependent on the presence of genetic variability, which provides basic information about the genetic properties of the population, the nature and extent of variability, the heritability of the characters, and genetic advance, all of which are used to formulate breeding methods for crop improvement. The analysis of variance revealed extremely significant differences among genotypes for all of the characters evaluated in this study, indicating that the pearl millet experimental materials have sufficient variability. The PCV was greater than GCV indicating little influence of environment on the expression of characters (Table 1) (Fig. 1, 2). The evaluated pearl millet genotypes revealed that the greater PCV and GCV were obtained for 14 characters viz., PHT (28.520, 28.046), number of tillers per plant (28.015, 22.811), number of leaves per tiller (22.365, 21.900), number of nodes on main tiller (29.356, 24.186), leaf width (37.733, 32.874), stem girth (40.879, 36.010), leaf weight (48.382, 47.911), stem weight (26.485, 26.310), leaf to stem ratio (20.841, 20.117), dry matter yield (33.007, 30.853), green fodder yield per plant (32.173, 32.005), crude fat (43.006, 38.208), crude protein (23.957, 20.867) and ash content (39.003, 35.948) except for days to 50% flowering (6.906, 4.042), leaf length (23.119, 18.960), dry matter content (20.308, 16.734), crude fibre (15.451, 13.819), acid detergent fibre (16.300, 14.999) and neutral detergent fibre (14.045, 13.179). It revealed the presence of higher genetic variability and it suggests that the selection may be effective. Concomitant results for high GCV and PCV for plant height, total number of leaves, leaf weight, stem weight and green fodder yield per plant was also reported by Suthamathi and Stephen Dorairaj (1995); Vidyadhar et al. (2007); Bhoite et al. (2008); Kumar et al. (2017); Thomas et al. (2018) in pearl millet. Moderate PCV and moderate GCV were observed for leaf length (23.119, 18.960), crude fibre (15.451, 13.819), acid detergent fibre (16.300, 14.999), neutral detergent fibre (14.045, 13.179) and dry matter content (20.308, 16.734) while low PCV and low GCV was found for days to 50% flowering (6.906, 4.042 respectively). Similar results for dry matter percent and crude protein, acid detergent fibre content was reported by Suthamati and Dorairaj (1997); Rahul Kapoor (2017) in Napier grass and Govindaraj et al., (2011) in pearl millet respectively. Heritable variation cannot be estimated with the help of GCV alone. Yield and its characteristics are significantly influenced by the environment, making it difficult to determine whether the observed variability is heritable or not. As a result, it's critical to separate the observed variability into heritable and non-heritable components. The precision with which a genotype can be determined by its phenotypic performance is referred to as heritability. Burton (1952) suggested that a combination of genotypic coefficient of variation and heritability estimations would be more efficient for selection. The kind of gene action influencing a given character can be determined by comparing heritability estimates and predicted genetic advance as a percentage of the mean. High heritability with high genetic advance as per cent of mean was observed for plant height (96.700, 56.816), number of tillers per plant (66.300, 38.262), number of leaves per tiller (61.700, 27.404), number of nodes in main tiller (67.900, 41.050), leaf length (67.300, 32.003), leaf width (75.900, 59.002), stem girth (77.600, 65.347), leaf weight (98.100, 97.736), stem weight (98.700, 53.842), leaf to stem ratio (93.200, 40.001), crude fibre (80.000, 25.460), acid detergent fibre (84.700, 28.431), neutral detergent fibre (88.100, 25.477), crude protein (60.600, 27.431), crude fat (78.900, 69.926), ash (84.900, 68.254), dry matter content (67.900, 28.404), dry matter yield (87.400, 59.409) and green fodder yield (99.000, 65.586). This indicated the lesser influence of environment in expression of these characters and prevalence of additive gene action in their inheritance and hence selection of these traits in breeding programme would facilitate the improvement of both fodder yield and quality. Similarly high heritability for these characters studied viz., number of leaves per plant, green fodder yield per plant was reports by Suthamathi and Dorairaj (1997); Babu et al. (2009) in Napier grass, Vedansh et al. (2010) in fodder maize and Sharma et al. (2003); Vidyadhar et al. (2007); Vinodhana et al. (2013) in pearl millet, Jain and Patel (2012) in fodder sorghum. Among the traits evaluated in pearl millet accessions, days to 50% flowering alone registered moderate heritability and low genetic advance as per cent of mean. High values for desirable variability, heritability, genetic advance as percent of mean were observed in the traits such as, plant height, number of tillers per plant, number of leaves per tiller, number of nodes on main tiller, leaf width, stem girth, leaf weight, stem weight, leaf to stem ratio, crude fat, crude protein, ash, dry matter yield and green fodder yield. It indicates that character-based selection might be quite effective and influenced by additive gene action. As a result, selecting these traits could help improve the yield and quality of green fodder in pearl millet. B. Correlation The green fodder yield, which is finally harvested in a forage crop, is influenced by a variety of vegetative plant features. Knowing the relationship between biomass yield and other biometrical features, as well as the relationship between the component qualities themselves, would substantially aid indirect selection for high green fodder yield. In the present investigation, the traits, plant height (0.9399 0.9220), number of tillers per plant (0.9458, 0.7703), number of leaves per tiller (0.9218, 0.5800), number of nodes on main tiller (0.9582, 0.7948), leaf length (0.9309, 0.7711), leaf width (0.7733, 0.6702), stem girth (0.8456, 0.7486), leaf weight (0.9450, 0.9309), stem weight (0.9517, 0.9401), leaf to stem ratio (0.6902, 0.6631), crude protein (0.7979, 0.6163), crude fat (0.2745, 0.2356) and dry matter yield (0.8582, 0.8032) were found to exhibit highly significant and positive correlation at genotypic and phenotypic level while crude fibre (-0.3483, -0.3171), dry matter content (-0.3384, -0.2800) and acid detergent fibre (-0.1963, -0.1811) have registered negative and significant correlation at genotypic and phenotypic level with green fodder yield (Table 2). The obtained result coincided with the findings of number of tillers with green fodder yield by Mangat and Satija (1991) in fodder pearl millet and Khan et al. (2002) in Napier grass. Vijendra Das and Ratnam Nadar (1991) remarked positive association of green fodder yield with plant height, number of tillers and leaf length. Plant height inter-correlated positively with number of tillers per plant, number of leaves per tiller, number of nodes on main tiller, leaf length, leaf width, stem girth, leaf weight, stem weight, leaf to stem ratio, crude protein, crude fat and dry matter yield and negatively with crude fibre, dry matter content and acid detergent fibre. The positive inter-correlation between the number of tillers per plant, the number of leaves per tiller, and the leaf stem ratio suggested that indirect selection for these traits could assist in increasing green fodder yield besides providing more leaves, which would improve palatability and intake. Similar result was reported by Kumar et al. (2017) in fodder pearl millet. It is concluded that all the biometric traits and quality traits viz., crude protein, crude fat were studied in the paper are highly associated with green fodder yield. Hence, these traits are important during selection of high biomass plant. C. Path analysis Path analysis is a statistical method for dividing the total effect into direct and indirect effects (Wright, 1921). Because the component traits are interdependent, they frequently influence their direct relationship with yield, limiting the reliability of selection indices based solely on correlation coefficient. As a result, the association between the characters was examined in order to determine the direct and indirect effects. The stem weight (0.4276) on green fodder yield has registered highest positive direct effect on green fodder yield, followed by leaf weight (0.3169) (Table 3) (Fig. 3). The result was in accordance with the results of Iyanar et al. (2010); Jain and Patel (2012) in fodder sorghum traits viz., number of tillers per plant, stem weight on green fodder yield had high positive direct effect. Therefore, these traits expected to have a true correlation and could be taken as component traits while selection for the improvement of green fodder yield. On the other hand, the traits viz., number of nodes on main tiller (-0.0520), neutral detergent fibre (-0.0240), ash (-0.0330) and dry matter content (-0.0482) showed negative direct effects. Stem weight had high positive indirect effect through plant height (0.3744), number of tillers per plant (0.3973), number of nodes on main tiller (0.3809), number of leaves per tiller (0.4801), leaf length (0.3627), leaf breath (0.3185), stem girth (0.3407), leaf weight (0.3466) and leaf weight. Stem weight had low negative indirect effect through crude fibre (-0.1303, -0.1230) and dry matter content (-0.1006, -0.1422) on green fodder yield was observed. Leaf weight had moderate positive indirect effect for plant height (0.2937), number of tillers per plant (0.2695), number of nodes on main tiller (0.2988), leaf length (0.2938), leaf width (0.2375), stem girth (0.2597), stem weight (0.2569), leaf to stem ratio (0.2760), crude protein (0.2263) and dry matter yield (0.2547). The dry matter content can be increased through the selection of high tillering, leaf and stem weight. The remaining traits had negligible inter-correlation, while green fodder yield had none. Plant height, number of tillers per plant, number of leaves per tiller, number of nodes in main stem, leaf length, leaf width, stem girth, leaf weight, stem weight, leaf to stem ratio, crude protein, and dry matter yield may be responsible for increasing green fodder yield per plant, according to correlation and path analysis studies. These traits could be used as selection criteria in future breeding programmes to improve fodder yield and quality.