Two populations, one non-improved of Dactylis glomerata and one improved of Agropyron cristatum were selected in an isolated spaced plant environment (80 cm) by three selection procedures: (1) Mass honeycomb selection (MHS), (2) pedigree honeycomb selection (PHS) and (3) pedigree honeycomb selection-combined criterion (PHS-CC). The third procedure is a between-families selection based on increased family mean and reduced family coefficient of variation (CV). Response to selection assessed in sward trials in the first and second harvest year was effective for the three procedures, which ranked PHS-CC > PHS > MHS. Evaluation was based on consistent efficiency as measured by both cloned and half-sib families. The precision of assessing progeny performance under sward conditions on the basis of single plant evaluation in the isolation environment, increased with a reduction in competition among plants, the use of moving-ring selection, and the application of high selection pressures (1·6%). The efficiency of selection between unimproved families increased when families were selected concurrently for high family mean and reduced family coefficient of variation.