Usually, the diffusion of a non-rival market knowledge externality - called a Knowledge Spillover (KS) - is related to geographical proximity. In this paper we explore the channels through which knowledge spreads. Compared with earlier work on KS measures, this study makes a step forward by calculating KS (as a balance of positive and negative absolute knowledge flows) on the basis of different proximity matrices. In particular, we focus on the relational, social, technological, and cognitive channel, along with the traditional geographical channel. In the light of previous studies on KS, we examine: (i) which types of proximity enhance or hamper the outward flow of knowledge; and (ii) whether the local endowment of absorptive capacity reduces such a flow. Our results show that KSs vary across alternative definitions of proximity. The parameter estimates of such a KS model show interesting patterns, with geographical and cognitive proximity having the highest explanatory power am ong all the types of proximity considered. Local absorptive capacity is found to be negative only when a region is surrounded by regions with similarly high levels of absorptive capacity. Furthermore, outward KSs decrease as geographical, relational, social, technological and cognitive distance increase. This points to the emergence or existence of large clusters of regions ('absorptive capacity clubs'), where relational, social, technological and cognitive proximity lock-in maximizes the returns to local investment in R&D.