Variation for incomplete resistance to coffee leaf rust was studied in Coffea arabica. Disease level in the field was scored by a 0 to 9 scale. Components of resistance observed in laboratory and greenhouse tests were latency period (LP), lesion density (LD) and leaf retention period (LRP). LRP determines the duration of sporulation. Lines of ‘Mundo Novo’ and ‘Catuai’ showed relatively small but significant differences for disease level in the field. Heritibility was low (0.31) and part of the variance (34%) was explainable by a significant correlation between disease level and yielding capacity of the lines. The high susceptibility of ‘Ibaare’, in comparison to other cultivars, was best explained by longer LRP values for ‘Ibaare’. Also its LP was generally shorter than that of other cultivars. Among coffee accessions from Ethiopia great variation was observed for disease level in the field. A significant part of the variance (35%) was explained by differences in yielding capacity. Variation for resistance components was observed between accessions, tested in the greenhouse and laboratory. However, results of the two tests were inconsistent. Transgressive segragation for incomplete resistance was observed in F2 populations of the cross between Agaro C1164–19 and ‘Catuai’, tested in the laboratory and greenhouse. Resistance was expressed by a longer LP, a lower LD, a certain percentage of non-sporulating lesions and, in some populations, by early necrosis of lesions. Prospects for breeding for incomplete resistance to H. vastatrix in C. arabica are discussed. Factors which may hamper selection progress are: a) the positive correlation between yield and disease level in the field, b) the relative small genetic variation for incomplete resistance among C. arabica cultivars and productive breeding lines, and c) inconsistency of results between resistance tests.