The essays presented in this dissertation strive to narrow the gap between Operations models and practice. They describe three models with seemingly paradoxical or counter-intuitive predictions and then test them in the controlled environment of the laboratory.Essay 1 studies the departure time decisions of commuters traversing a Y-shaped network with two bottlenecks, who wish to arrive at their common destination at a desired time. Imposed on the network are costs associated with arriving either too early or too late with respect to an exogenously determined arrival time as well as to the delay experienced due to the bottlenecks. The equilibrium solution implies that, for certain parameter values, expanding the capacity of the upstream bottlenecks while keeping the capacity of the other fixed may induce a shift in the endogenously-determined departure times so as to increase total travel costs. We report the results of a large-group experiment designed to test this counterintuitive hypothesis. Our experimental results are strongly supportive of this prediction.Essay 2 examines the Braess Paradox which is a counterintuitive discovery that removing a link from a network that is subject to congestion may decrease the equilibrium travel cost for each of its users. We demonstrate this phenomenon in a complex network and test it experimentally with large groups of players. Our main purpose is to compare two information conditions. In the PUBLIC condition every user is informed of the route choices and payoffs of all the users. In the PRIVATE condition, each user is only informed of her own payoff. We show that under both information conditions, aggregate route choices converge to equilibrium.Essay 3 examines the impact of information on the routing decisions that drivers make in a congestible two route traffic network. We present a model and theoretical predictions of driver choices in such a network and compare outcomes under conditions of full and no-information regarding the capacities of each route. Under certain circumstances, the model predicts a paradox: aggregate travel delays increase with the provision of a priori information regarding stochastic travel conditions. We report evidence supporting this paradox in a laboratory experiment.