The yield stability of intercropping systems is important in developing cropping systems that produce economic yields over a wide range of environments. Field studies were conducted during the 1997/98 and 1998/99 growing seasons at three locations in the forest and forest–savannah transition zones of Ghana to determine yield, land use efficiency and yield stability of cassava (Manihot esculenta), maize (Zea mays), soya bean (Glycine max) and cowpea (Vigna unguiculata) intercrop systems. The cropping systems consisted of a factorial combination of two cassava varieties: ‘Gblemoduade’ (an improved, highly branched variety) and ‘Ankra’ (a local variety with fewer branches) and three spatial arrangements. Cowpea rows were planted in the minor season into cassava as a successive crop to maize and soya bean after their harvest in the major season. Intercropping significantly reduced grain or tuber yield of maize, cassava ‘Gblemoduade’ and cassava ‘Ankra’ by 23–70%, 16–49% and 24–64%, respectively. Maize yield decreased with increased number of soya bean rows. ‘Gblemoduade’ outyielded ‘Ankra’ by more than 100% under both intercrop and sole crop. The yield of soya bean increased with increased number of soya bean rows, but did not differ in response to the cassava variety. However, cowpea yield was higher when intercropped with ‘Ankra’ than with ‘Gblemoduade’. The intercrops had higher land use ratios (LER=1·27–2·83) and were more stable than the sole crops. Intercrops involving ‘Ankra’ had higher LER (2·14–2·18) than systems with ‘Gblemoduade’ (LER=1·83–1·99), but their yield stabilities were similar.