Management scholars and practitioners emphasize the importance of the sizeand diversity of a knowledge worker's social network. Constraints onknowledge workers' time and energy suggest that more is not always better.Further, why and how larger networks contribute to valuable outcomesdeserves further understanding. In this study, we offer hypotheses to shedinsight on the question of the diminishing returns of large networks and thespecific form of network diversity that may contribute to innovativeperformance among knowledge workers. We tested our hypotheses using datacollected from 93 R&D engineers in a Sino-German automobile electronicscompany located in China. Study findings identified an inflection point,confirming our hypothesis that the size of the knowledge worker's egocentricnetwork has an inverted U-shaped effect on job performance. We furtherdemonstrate that network dispersion richness (the number of cohorts that thefocal employee has connections to) rather than network dispersion evenness(equal distribution of ties across the cohorts) has more influence on theknowledge worker's job performance. Additionally, we found that thecurvilinear effect of network size is fully mediated by network dispersionrichness. Implications for future research on social networks in China andWestern contexts are discussed.