Page - (000347) - in Knowledge and Networks

345 repeated ties accelerate the diffusion of information, whereas infrequent ties serve as a source of novel and nonredundant knowledge (Granovetter, 2005). Cognitive Proximity Adding to what has already been done, we unravel the multifaceted concept of cog- nitive proximity into overlap, reciprocal potential, and knowledge transfer and track their dynamics within the evolution of collaboration. Basically, the decision to form or maintain a link is continuously evaluated according to the potential gains in knowledge and in innovation (Hamel, 1991; Wuyts et al., 2005). The knowledge endowment of partners can be considered a pool of potential knowledge flows. For these flows to be take place, two conditions must be met. First, a certain minimum similarity of knowledge bases, the overlap, is necessary to provide a basis for mutual understanding. The ability to absorb external knowledge is largely a function of the relatedness of the knowledge bases of collaboration partners (Boschma, 2005; Cantner & Meder, 2007; Cohen & Levinthal, 1990). Second, the exchange of knowledge requires potential knowledge that can be acquired because it is novel for the partner and not similar to the knowledge that the partner already possesses. The implication is that the dissimilarity of knowledge bases is also fruitful for potential knowledge flows. Collaboration will be established or continued only if the expected knowledge gains are positive. From a dynamic perspective partners move along this proposed scale of cogni- tive proximity by increasing their overlap when collaborations evolve. After col- laboration has been initiated, partners who are able to learn will experience an assimilation of knowledge bases that results in both an increase in overlap and a decrease in novelty potential (Balland et al., 2015; Nooteboom, 1998; Wuyts et al., 2005). The positive effects that overlap has on mutual understanding will eventually be offset by the negative effects on novelty creation (Balland et al., 2015). These dynamic reverse effects have been found in empirical studies on the persistence of collaboration between researchers (Dahlander & McFarland, 2013) and on the performance of continuing cooperation between organizations (Beaudry & Schiffauerova, 2011; Wuyts et al., 2005). At Stanford University, too much intel- lectual similarity (overlap) of the literature cited in publications by collaborating researchers has hampered the perpetuation of their collaborative ties (Dahlander & McFarland, 2013). Lack of diversity decreases innovative performance in repeated collaborations as patent rates and the quality of patents diminish in long- term col- laborations (Beaudry & Schiffauerova, 2011), and the less variation a collaboration portfolio has, the less likely it is to result in technical novelty (Wuyts et al., 2005). We therefore assume that strategic actors who seek to maximize the benefits of col- laboration for innovation will terminate their teamwork after it has exceeded the optimal level of overlap. Hypothesis 1a The relation between the cognitive overlap of two actors and the likelihood of their continued collaboration follows an inverse-U curve. 16 Coevolution of Innovative Ties, Proximity, and Competencies