Sorghum is ranked the fifth most produced food crop in the world, and is a dietary staple for over 500 million people in over 30 countries. It is the second most produced food crop in Africa, where cultivation of local varieties (landraces) of sorghum is the predominant form of agriculture.This study investigated the genetic diversity and structure of 161 sorghum accessions, which included landraces and wild/weedy sorghum, using 17 microsatellites. The material represented three geographical scales. For the continental and country scale studies, landrace accessions from throughout Africa and throughout Tanzania were obtained from gene banks (ICRISAT and NPGRC). For the local scale study, eight landraces and wild/weedy sorghum were collected from five households in Hombolo, Tanzania.The genetic diversity of sorghum at all three geographical scales was found to be mainly structured according to geography and less structured according to race, temperature and precipitation. At the continental scale, the accessions were (based on STRUCTURE analysis) largely divided into an eastern, western, northeastern and southern group. However, accessions from Sudan were found in all four groups, a result which supports the suggested origin of domesticated sorghum in northeastern Africa. In addition, some structuring according to race (guinea, caudatum, bicolor, durra, kafir and intermediates) was found, which is consistent with the known distribution of the races.The cultivated accessions from Hombolo were genetically structured according to landrace and for the most part differently named landraces were genetically distinct. In addition, there was some geographical structuring of genetic diversity for the cultivated accessions (even though fields were only from 150m to 1.6 km apart ), but not for wild/weedy sorghum growing alongside the sorghum crop fields. This could be explained by higher outcrossing rates in wild/weedy sorghum compared with cultivated sorghum. Considerable gene flow was detected between wild/weedy and landraces based on genetic overlap, no significant differences in genetic diversity and the number of migrating individuals. Gene flow was higher between cultivated and wild/weedy sorghum than between different landraces of cultivated sorghum.The mainly geographical structuring of sorghum diversity can be explained by traditional cultivation practices based on indigenous landraces and a self-fertilizing mating strategy. This reflects the wealth of diversity found amongst indigenous landraces, which is important to conserve for present and future food security needs.Keywords: sorghum, landrace, genetic diversity, wild/weedy sorghum, microsatellites.