Current GIS data models have problems representing fuzzy, interconnected data sets, due to problems inherent with the underlying abstractions used for storing and manipulating geographical information. These limitations are becoming problematic when processing traditional knowledge of spatio-temporal patterns. The authors present a new approach to the representation of fuzzy, spatio-temporal knowledge, as found in traditional knowledge systems. The approach used centres on the motif, the folk thesaurus, narratives, and a data model that connects them. The authors suggest modelling spatio-temporal knowledge in a way that maintains some of the strengths of oral depiction, interconnectedness, and precise ambiguity. We discuss how spatio temporal motifs interact with other motifs via shared elements, via narratives and via processes. Finally; we explore how a working data model centred on motifs may be conceptually and logically represented.