J. agric. Sci., Camb. (1978), 91, 505-508 Printed in Great Britain 505 Triple test cross analysis in first backcross populations of four wheat crosses BY S. SINGH Department of Agricultural Botany, J.V. College, Baraut, India AND R. B. SINGH Department of Genetics and Plant Breeding, B.H.V., Varanasi, India (Received 24 April 1978) SUMMARY The triple test cross analysis (Kearsey & Jinks, 1968; Jinks & Perkins, 1970) was used to detect and estimate the additive, dominance and epistatic components of genetic variation for four metric traits, namely, final plant height, number of spikelets per spike, 100-kernel weight and yield per plant, in the first backcross populations of four wheat crosses (Norteno 67 x Moti, Sonalika x Moti, Kalyan Sona x Sonalika and Kalyan SonaxNP 876). Epistasis was a more important component of variation for final plant height and yield per plant than for number of spikelets per spike and 100- kernel weight. On the other hand, the additive component was highly significant for all four characters in all eight backcrosses. INTRODUCTION Of all the genetic procedures based on second degree statistics which are used to estimate the components of continuous variation, the triple test cross analysis has the least assumptions (restric- tions) and therefore is more widely applicable for investigating materials of various kinds. In addi- tion to allowing an unambiguous detection of epi- stasis and unbiased estimation of the additive and dominance components in the absence of epistasis, the analysis is also unaffected by differences in allelic frequencies, degree of inbreeding and gene correlation. Furthermore, the method can be used to investigate inbred lines as well as the F 2 genera- tion (Kearsey & Jinks, 1968; Jinks, Perkins & Breese, 1969; Jinks & Virk, 1977) and backcross populations (Jinks & Perkins, 1970). MATERIALS AND METHODS Experimental design The F X 8 of four spring wheat crosses Norteno 67 x Moti, Sonalika x Moti, Kalyan Sona x Sonalika and Kalyan Sona x NP 876, were back-crossed to their P t (larger parent) and P a (smaller parent) to obtain B ^ x P J and Bjfi^xPu) generations. From each of these eight backcrosses, 18 plants were randomly chosen and crossed, as male parent, to their corresponding P v P 2 and F 1 to produce 54 progeny families. All 432 families (produced from all eight backcrosses) were planted in three com- pletely randomized replicate blocks. Observations for four characters were taken on five plants randomly chosen from each progeny family in each replication. The four characters measured were final plant height, number of spikelets per spike, 100-kernel weight and yield per plant. The data obtained from each backcross were analysed separately (individual analysis) for each character except where both the Bj^ and B 2 of a cross did not differ significantly against a full error component. In the latter case, a combined analysis (B 1 +B 2 ) was carried out. Detection of epistasis The test for epistasis (L u + L 2i — 2L 3i , where L u is the mean of B u or B 2i x P v L 2( , the mean of B lt or B 2 i x P 2 and L 3( , the mean of B 1( or B 2i x F v cross) was carried out according to Kearsey & Jinks (1968). The sums of squares due to epistasis and replicate error, in each of the eight backcrosses, were obtained for 18 D.F. and 36 D.F., respectively. A within-families error (1296 D.tf.) was used to test the significance of these two items, as ^ 2 s (the sum of squares of the item divided by the average mean square within families), except where the replicate error was significant. In the latter case, the epistasis item was tested, as a variance ratio, against the replicate error. Whereas the items