This work re-considers current descriptions of pre-colonial Swahili stone architecture by applying formal analysis to structures from the 'golden age' of the fourteenth and fifteenth centuries and the Swahili heartland regions of mainland Tanzania and Kenya. The results of this analysis are compared to mainstream descriptions of Swahili architecture. They are then used to make comparisons between the remains of pre-colonial Swahili stone structures and those in built environments that are popularly identified as essentially 'Islamic.' This project seeks to generate an updated description of early Swahili architecture while providing a useful case study of Islamic architecture at the margins of the field. The data for this work was attained through a combination of archival and onsite research. The latter was carried out through visits to Kenya and Tanzania over the course of two summers, 2006 and 2009. The analysis is separated into sections focusing on religious, domestic and public spaces that were chosen from a cross section of stone town sites. The findings support that pre-colonial Swahili spaces were designed by and for the local Swahili community. The analysis of religious structures suggests that nine-bay mosques may have had a commemorative function. For larger mosques it is proposed that longitudinal orientation and the visual obstruction of the mihrab by central pillars are deliberate (rather than accidental) features that were intended to emphasize the direction of prayer. The analysis of housing reveals that pre-colonial Swahili stone houses exhibit only minimal degrees of visual restriction: the stringently gendered, intimacy gradient that is thought to circumscribe both social and physical Swahili space was a relatively late development that can be linked to the period of Omani colonization in the eighteenth and nineteenth centuries. The late colonial period also marks the earliest evidence of a formal relationship between Swahili stone houses and coastal South Arabian houses. The analysis of public space suggests that pre-sixteenth century parallels between Mijikenda and Swahili urban form are tenuous at best. When these parallels exist they can be ascribed to a shared cultural climate without the necessity of identifying direct links between one or another local groups.