In this paper, we employ techniques developed in spatial econometrics to analyze spatial patterns of technology diffusion, to detect clusters and to estimate theoretical models that incorporate space explicitly. These techniques correct for misspecifications resulting from the omission of spatial linkages in standard empirical models of economic growth. Our dataset consists of TFP estimates for 73 countries over the period 1960-2000, and we find that TFP growth rates and levels are positively autocorrelated over space, meaning that high or low values tend to be clustered. We also find that TFP levels are becoming more clustered over time, suggesting the possibility that technology levels are converging locally. Estimation of spatial versions of the Nelson and Phelps (1966) model shows that the impact of being located close to a country with high TFP growth rates is positive and substantial.