The Alzheimer's Disease Assessment Scale's cognitive subscale (ADAS-Cog) is the most widely used instrument for screening cognitive dysfunction in Alzheimer's disease. The aim of the present study was to develop an Arabic version of this scale (A-ADAS-Cog), examine its psychometric properties (reliability and validity), and provide normative data. The A-ADAS-Cog), an Arabic version of the Mini-Mental State Examination (A-MMSE), and a Standardized Clinical Dementia Rating Scale (CDR) were administered to three Tunisian groups: 124 normal controls (NC), 33 patients with non-Alzheimer dementia (N-AD), and 25 patients with Alzheimer's disease (AD). The A-ADAS-Cog scores were significantly affected by age and education. A correction table was constructed to control these effects. The results showed that the A-ADAS-Cog has good internal consistency and reliability (α= 0.82 for AD). The test-retest reliability of the A-ADAS-Cog was stable over time (r = 0.97). An evaluation of the construct validity of the A-ADAS-Cog using principal component analysis led to a solution with three factors (memory, language and praxis), which explained 72% of the variance. The concurrent validity of the A-ADAS-Cog was established using the A-MMSE score (r = -0.86), CDR Sum of Boxes score (CDR-SB; r = 0.87), and global CDR score (CDR-Global; r = 0.74). Finally, the A-ADAS-Cog has an excellent discriminating power in the diagnosis of AD (ROC area = 0.92). A cut-off score of 10 (sensitivity = 84% and specificity = 91%) is indicated for the screening of the AD. Overall, the results indicated that the A-ADAS-Cog is psychometrically reliable and valid and provides promising results for screening of dementia in Arabic speaking patients.