V . L E I A N D M . J A K O B S E N . 2004.Aims: To identify and examine the diversity of predominant lactic acid bacteria (LAB) in koko and koko sour water (KSW) from different Ghanaian production sites with regard to pattern of fermentation (API 50 CHL), genotype, antimicrobial activity, and resistance to low pH and bile salts. Methods and Results: In total 215 LAB were isolated from koko and KSW. The isolates were identified using intergenic transcribed spacers (ITS)-PCR restriction fragment length polymorphism (RFLP), API 50 CHL, restriction enzyme analysis with pulsed-field gel electrophoresis (REA-PFGE) and sequencing of the 16S rRNA gene. The dominating micro-organisms in koko was found to be Weisella confusa and Lactobacillus fermentum, followed by Lact. salivarius and Pediococcus spp. Chemometric data analysis were used to link the LAB species to the different production stages and production sites. At intra-species level the isolates were found to have a great diversity. The isolates were investigated for antimicrobial activity using agar diffusion assays, and acid and bile tolerance. Most isolates showed low levels of antimicrobial activity towards the indicator strain Listeria innocua, but not towards the bacteriocin-sensitive Lact. sakei. Growth of all LAB isolates was unaffected by the presence of 0AE3% (v/v) oxgall bile. The isolates were able to survive, but were not able to grow in growth medium adjusted to pH 2AE5. Conclusions: The dominating LAB of koko and KSW were W. confusa and Lact. fermentum showing a pronounced taxonomic biodiversity at sub-species level between stages within the production as well as between production sites. Other species observed in KSW were Lact. salivarius, Ped. pentosaceus, Ped. acidilactici and Lact. paraplantarum. They occurred in levels of 10 8 CFU ml )1 in fresh KSW and showed uniform antimicrobial activity, and acid and bile tolerance. Significance and Impact of the Study: The present study gives a detailed picture of the taxonomy and diversity of LAB in an African-fermented millet product that may have potential as a probiotic product for the local population. The chemometric tools Principal Component Analysis and ANOVA ANOVA Partial Least Squares Regression were proven to be useful in the analysis of microbial groupings and associations with specific sites and stages in the production of koko and KSW.