Using data collected from an experimental double auction market, we study the dynamics of interaction among traders. Our focus is on the effect the trading network has on price dynamics and price-fundamental convergence. At the aggregate level, the network of empirical exchanges reveals properties that are dissimilar from random graphs and, in particular, high centrality and high clustering. Precisely, these properties are identifiable as the cause of price volatility and divergence from the fundamental value. At the microscopic level, we find out how the topological properties of the network derive from the behavior of traders. In fact, our findings show that it is the unbridled trading action of very centralized players who implement a minority game, to give rise to volatility clustering and arbitrage opportunities.JEL codes: D03 • D53 • D85 • G12