The mating system of the tropical rainforest tree Carapa procera was studied for a population located within a sylvicultural trial in French Guiana. We used the mixed-mating model with the genotypes of 47 openpollinated progenies at ten polymorphic loci, obtained through isozyme analysis. Seeds were collected on both logged and unlogged (control) plots, which were interspersed. A high multilocus outcrossing rate (0,78) was found for the global population, and there was indirect evidence for nonrandom mating for outcrosses. It is argued that this may be due to some kind of positive assortative mating, probably biparental inbreeding. Correlation of selfing between sibs was not significantly different from zero, and the proportion of full-sibs among outcrossed sib-pairs was significant but low (0.16). We found significant differences between outcrossing rates when computing separate estimates for the 14 trees located on logged plots (tm = 0.63) and the 33 trees located on undisturbed plots (tm = 0.85). It is argued that the decrease in density is probably not the only cause of the decrease in outcrossing rate on logged plots. Lastly, several points concerning the reliability of the estimation are outlined. In particular, the downward bias due to ignoring null alleles during genetic interpretations of zymograms is quantified in a few cases by simulating artificial populations.