OBJECTIVES: This study aimed to translate the Michigan Diabetes Knowledge Test (MDKT) into Pakistani (Urdu) language, and to examine the psychometric properties of the Urdu version.  METHODS: A standard procedure of “forward-backward” translation procedure was used to create the Urdu version of the MDKT from the original English version. A convenience sample of 325 outpatients with type 2 diabetes was approached between June and November 2010. All data were collected from Bolan Medical College Hospital, Quetta, Pakistan. In addition to MDKT, socio-demographic data of the patients was also collected. Patient medical records were explored to get clinical and hemoglobin A1c data. For the purpose of test-rest analysis, Spearman’s rho coefficient was used and data was available from 51 patients. Internal consistency was used as a measurement of reliability using Cronbach’s alpha. Known group validity was also measured to ensure the consistency of the MDKT.  RESULTS: By using the recommended scoring methods of MDKT, the mean ± SD of MDKT scores was reported as 7.56±2.98. Cronbach’s alpha value was 0.702 showing good internal consistency. Test-retest reliability value was 0.812 (p <0.001). Significant relationship between MDKT categories and HbA1c categories (chi-square = 20.555; p <0.001) was found for known group validity. CONCLUSIONS: The findings of this validation study reveal that that the Pakistani (Urdu) version of the MDKT is a reliable and valid measure of diabetes knowledge that can be used in clinical and research practice of Pakistani health care system.