An investigation was carried out at the experimental plots of the Department of Genetics and Plant Breeding, Main Research Station, University of Agricultural Sciences, Hebbal, Bangalore, India to predict the level of heterosis based on genetic divergence through regression analysis in sunflower. One hundred and forty crosses involving 10 lines and 14 testers were affected and evaluated for ten quantitative traits. Better parent heterosis (BPH) of crosses was estimated and correlated with genetic divergence between parents as measured by Mahalanobi’s D2 statistic. The relationship between character wise parental divergence and better parent heterosis only for five important characters viz., plant height, head diameter, seed yield, oil content and oil yield, was determined using linear regression and curvilinear regression of second degree. The estimates of better parent heterosis significantly regressed towards the genetic distance of the parents. However, it was not enough for successful prediction of heterosis through either linear or curvilinear regression of second degree as indicated from highly significant chi-square values for observed and predicted estimates of heterosis for all these characters.