An experiment was conducted in Winnipeg, Manitoba, Canada to : (1) ascertain the efficiency of gridded mass selection ; (2) to determine if additional selection efficiency can be achieved through refinement of the gridded mass selection procedure ; and (3) to determine if the degree of homozygosity has an effect on response to single-plant selection . Three selection methods [simple mass selection (SMS) ; gridded mass selection (GMS) ; and a new rhombic grid selection (RGS)] were applied to four populations [a conventional F2 (CON-F 2); an intermated F2 (INT-F2) ; and 2 partial backcrossed populations] developed from a biparental cross involving the Canadian spring wheat cultivars, 'Glenlea' and 'Sinton' . Averaged over the 4 populations, a single cycle of RGS (10% selection intensity) increased yield by about 9 .5%, whereas GMS and SMS increased yield by approximately 4 .5% and 3 .1%, respectively . The degree of homozygosity of individual plants that make up the population appeared to have an effect on the efficiency of selection . In general, selection was more efficient in the two partial backcrossed populations, and least efficient in the CON-F2 and INT-F2 populations . The largest response to selection (15 .9%) was attained when RGS was applied in the backcross to Glenlea population, and the lowest response (1 .2%) when SMS was enforced in the CON-F2 population . This paper represents the first report on the efficiency of rhombic grid selection for yield improvement in cereal crops .