Abstract Objective: This study examines the utility, both broadly and at the subtest level, of the cortical-subcortical index (CSI) of the RBANS in classifying cortical and subcortical cognitive profiles. Methods: Outpatients in a southeast hospital were grouped by diagnosis: Alzheimer’s Disease (AD, n=33; age=77.03[6.04], ed=13.55[2.58], 88% female, 91% Caucasian), Subcortical (n=36, age=72.17[11.51], ed=12.63[3.70], 50% female, 75% Caucasian) and No Diagnosis (n=30, age=71.97[5.59], ed=13.24[3.04], 80% female, 70% Caucasian). The CSI formula ([visuospatial construction + attention]/2]–[delayed memory + language]/2)) was calculated for patients. Groups were compared on CSI score using one-way analysis of variance (ANOVA) and Tukey’s HSD. ROC analysis was used to identify CSI cutoff that optimizes classification. Independent t-tests for subtests comprising the CSI were run comparing Cortical/Subcortical groups, with stepwise regression to evaluate significant subtests. Results: CSI differed significantly between groups (F[2, 96]=30.58, p<.001]). Tukey’s HSD found CSI was significantly different between Cortical/Subcortical (p<.001, 95% CI=16.38, 31.27) and Cortical/No Diagnosis (p<.001, 95% CI=9.62, 25.20), but not Subcortical/No Diagnosis (p=.118, 95% CI=-14.04, 1.22). ROC yielded a CSI cutoff of <-1 for optimal classification (AUC=.888, SE .039, p<.001, 95%CI .813-.964, Sensitivity=.750, Specificity=.909). Independent t-tests resulted in significance in 6 subtests. Stepwise regression using significant subtests revealed the best CSI-prediction model explained 28% of the variance (F[1,111]=44.06, p<.001) and included only List Recall (B=-2.96, p<.001). Conclusions: CSI’s accuracy in classifying Cortical from Subcortical profiles was 81.6% using a cut score of -1. List Recall accounted for most of the variance in the CSI. Limitations included small sample size and using RBANS scores to determine original diagnostic category.