Standard models of diffusion assume spatial and temporal homogeneity.
This article develops a class of diffusion models that incorporate spa
tial and temporal heterogeneity by turning to the individual level wit
hin an event-history framework. These models permit the analyst to rep
resent social structural relations thought to channel diffusion, and t
o model decay in the influence of events over time. Heterogeneous diff
usion models are applied to a reanalysis of data reported in Coleman,
Katz, and Menzel's classic diffusion study. Network centrality and loc
al structures of influence based on cohesive relations and structural
equivalence are all shown to channel the diffusion of tetracycline.