Two experiments, each involving a set of six sweet potato clones, the first set developed in sites differing in altitude, and the second in sites differing in soil type, were done at three locations in 4 years in Cameroon. Data obtained were subjected to analyses of variance to determine the presence of genotype × environment (G × E) interactions, and to joint regression analyses to measure the performance of clones across environments. The first experiment produced higher yields and contained more stable clones than the second. In both experiments, mean yields were almost twice as high in 1984 (21·1 t/ha) as in each of the other years (c. 11·0 t/ha), and highest at Nyombe (18·0 t/ha). In Expt 1, the G × E interaction mainly concerned interaction with location, whereas in Expt 2 it concerned interaction with years.
Clones 1611 (Expt 1) and 048 (Expt 2) yielded above average and gave linear regressions significantly above unity (b > 1·0) for most traits, indicating specific adaptation to high-yielding environments and hence below average stability. Clones 1112, 1639 and TIbl (Expt I) yielded above average and had regression slopes equal to unity (b = 1·0), indicating average stability and thus general adaptability. Clones TIb2 (Expt 1) and 1487 (Expt 2) produced below average yields (b < 1·0), indicating specific adaptation to low-yielding environments. Since sweet potato is grown mainly for human consumption in Cameroon, a preferred clone must have stable marketable yields. Only clones 1112, TIbl and 1639 could be considered desirable for release to growers.