The overall kinetics of retting, a spontaneous fermentation of cassava roots performed in central Africa, was investigated in terms of microbial-population evolution and biochemical and physicochemical parameters. During the traditional process, endogenous cyanogens were almost totally degraded, plant cell walls were lysed by the simultaneous action of pectin methylesterase and pectate lyase, and organic acids (C 2 to C 4) were produced. Most microorganisms identified were found to be facultative anaerobes which used the sugars (sucrose, glucose, and fructose) present in the roots as carbon sources. After 24 h of retting, the fermentation reached an equilibrium that was reproducible in all the spontaneous fermentations studied. Lactic acid bacteria were largely predominant (over 99% of the total flora after 48 h) and governed the fermentation. The epiphytic flora was first replaced by Lactococcus lactis, then by Leuconostoc mesenteroides, and finally, at the end of the process, by Lactobacillus plantarum. These organisms produced ethanol and high concentrations of lactate, which strongly acidified the retting juice. In addition, the rapid decrease in partial oxygen pressure rendered the process anaerobic. Strict anaerobes, such as Clostridium spp., developed and produced the volatile fatty acids (mainly butyrate) responsible, together with lactate, for the typical flavor of retted cassava. Yeasts (mostly Candida spp.) did not seem to play a significant role in the process, but their increasing numbers in the last stage of the process might influence the flavor and the preservation of the end products. MATERIALS AND METHODS Origin of the plant material. Cassava roots (Manihot esculenta var. MM86) were harvested near Brazzaville, Congo, 15 months after planting. Retting procedures. Approximatively 100 kg of roots was washed, peeled, and placed in a 200-liter barrel filled with rainwater, which was left at ambient temperature. This retting procedure was repeated four times, always yielding similar results; the data presented are from a single representative retting. Sample preparation for bacterial enumeration. Sampling was performed every 12 h for the first 2 days and then daily until retting was completed. Six randomly selected root sections were cut into 0.5-cm-diameter cubes and mixed under sterile conditions. A 60-g sample was diluted in 540 ml of sterile peptonized water reduced by boiling and addition of cysteine-HCl (0.1% [wt/vol]); this corresponded to a 10 Ϫ1 dilution. The solution was then mixed in a blender (Turnmix ME 88; SOFRACA,