Abstract: The study centralized on using discriminant analysis to predict soil group membership in order to correctly classify future unknown observation into any of the five soil groups based on observed predictors (soil characteristics) in soils supporting Raphia palms of Southern Nigeria. Four functions were estimated but only one with higher eigenvalue of 1.315 which explain the variation of the soil groups by 55.6% was significant. The estimated canonical correlation, the Wilks’ Lambda and their associated Chi-square values used to measure the significant performance of the function were 0.754 and 0.178 and 82.503, respectively and was significant at p<0.05, of the 13 predictors variables used in the analysis, 7 were significant with mg ranking in discriminating among the groups. When actual grouping of the five soil types was compared to the predicted groupings generated by the discriminant functions, group 1 and 5 were predicted the best with 75% of the cases correctly classified while group 3 had a moderate performance with 64.2% success rate. Group 2 and 4 were predicted the worst with 58.3 and 56.8%, respectively of the cases were correctly classified. The overall discriminant function (hit ratio) had 67.7% success rate in classifying the samples. The estimated press’s Q statistic used to determined the predictive power of the model was 347.56 and this was greater than the calculated value (11.1), implying that the model predictive accuracy is greater than that expected by chance. Thus, the model has a good predictive power and can be generalized as a fairly good tools for classifying new cases. Finally, the result shows that discriminant analysis is a fairly good method for predicting new cases into any of the five soil types dominant in the Raphia growing zone of southern Nigeria and that mg, p, Cu, pH, Ca, Mn and K were identified as the soil properties that best discriminate among the soil types in the zone studied.