We examine the effect of the matching mechanism on learning in 2 x 2 games. Six games are played repeatedly under either fixed pairs or random matching. Unlike most economics experiments, the games are played under limited information: subjects are never shown the games payoff matrices nor given information about opponent payoffs. We find that behaviour, while initially similar between treatments, diverges over time. In most but not all games, fixed-pairs matching is associated with increased coordination on pure-strategy Nash equilibria, higher-payoff equilibria being reached and faster convergence toward pure-strategy play.