Resistance to witches' broom disease (WB) in cocoa, caused by Moniliophthora perniciosa, was investigated in 56 cross families. Twenty-four micro-grafted clones (16 parental and 8 control or reference clones) were also inoculated, including 16 parental clones representing various genetic groups. The clones and seedlings were tested for resistance to the disease using an optimised agar-droplet inoculation method in screen house experiments with plants of one-year-old, and by inoculating one bud per plant. For clones, three replicates of three plants each were used and for seedlings two replicates of 25 seedlings). Resistance components recorded on an individual plant basis were infection frequency (IF = % of swellings or brooms), broom frequency (BF, %), incubation period (IP, days to express first symptoms) and symptom severity (expressed as broombase diameter or BBD in mm). As expected from earlier work with the same type of agar-droplet inoculation method, average IF and BF were high. This allowed for better distinction among other post-penetration resistance parameters. Seedling BF was approximately 100%, whereas this parameter varied widely for clones (22 to 100%). Significant differences were observed among the clones and seedling progenies both for IP and BBD. For the clones, BBD was negatively correlated with IP (r = - 0.68) and positively correlated with IF (r = 0.62). IF and IP were not significantly correlated (r = - 0.22). For 14 clones tested in the screen houses, it was possible to compare the resistance components with field results (% vegetative infection). BBD was correlated with field levels of resistance (r = 0.71), whereas the correlation between IF and field resistance was lower (r = 0.42), though still significant. IP was not correlated with field results (r = - 0.16). These results clearly show that BBD is an important trait to be used in selection for WB resistance. There was a strong linear relationship between mid parental values and progeny