Organizations learn from their own operational experience and from the experience of their partners and competitors. Absorptive capacity – the ability to acquire and use extramural knowledge - is one possible process linking experiential and vicarious learning. The objective of this project is to develop new models to examine the organizational micro-mechanisms underlying absorptive capacity by representing both experiential and vicarious learning as network-based processes. In this perspective experiential leaning comes from deliberate investments in operational capacity, i.e., from decisions concerning the composition of the portfolio of internal organizational activities. Vicarious learning comes from contact with partners and competitors, i.e., from decisions concerning the selection of network associates. The project being proposed intends to derive, implement, and test a new model in which these decisions are linked by a co-evolutionary relation. The model posits that operational experience drives the choice of network partners, which then stimulates learning and triggers change in the internal composition of organizational activities. The addition of new organizational activities – or the abandonment of old ones - determine the conditions for the development of new trajectories of experiential learning eventually leading to the choice of new partners. A new stochastic actor-oriented model (SAOM) is proposed for representing the evolutionary dynamics of this system of coupled organizational decisions underlying the aggregate empirical regularities predicted by the absorptive capacity hypothesis: organizations with similar operational experiences are more likely to be connected by relations of mutual learning. Because the model proposed is not implemented in currently available statistical software, a parallel objective of the project involves programming, testing, documenting, and making publicly available new specialized routines for the specification and estimation of SAOM. The new model and the new estimation algorithms will be tested on a large dataset that has been collected specifically for this purpose. The new routines will be incorporated into R-SIENA (Simulation Investigation for Empirical Network Analysis) - the freely available, R-based statistical software for the estimation of models for network panel data according to the SAOM analytical framework.