Three cowpea (vigna unguiculata (L. ) walp) lines were selected based on their per cent crude protein and used as parents in crosses in the following combinations: 2/180-4-12 (High) x 11/48-2 (Low); 2/180-4-12 (High) x L-25 (Low) and 11/48-2 (Low) x L-25 (Low). In the first cross the -following generations were obtained in addition to the parentals, reciprocal F1's, reciprocal backcrosses and F2. In the remaining two crosses only the reciprocal F1's and F2 populations were obtained in addition to the parentals, The F1 means per cent crude protein content were either closer to or lower than the low protein parent, thus indicating partial dominance to overdominance of low protein over high, No significant difference was observed between reciprocal F1's, indicating the absence of maternal effects. The F2 distributions for per cent protein were normal suggesting the involvement of multiple genes, thus supporting the estimates of three major genes with possible modifiers. Broadsense heritabiiity estimates ranged from 40.7% in the cross low x low to 73.4% in the crosses high x low. Narrowsense heritabiiity estimates ranged from 26.8% to 40.3%. These estimates indicate considerable? genetic influence some of which were additive, In all crosses, negative phenotypic and genotypic correlations predominated for yield, number of seeds/pod and number of days to flowering with per cent protein, Significant negative phenotypic correlations were obtained between number of pods/plant and per cent protein, while 100-seed weight showed mostly positive phenotypic and genotypic correlations with per cent protein. Positive correlations predominated between days to maturity and per cent protein. Some high yielding segregant had high protein percentage, suggesting that high yield and high protein percentage can be combined into a suitable genotype most especially with the weak negative correlation between these characters