This experiment attempted to identify the features of air traffic geometry that would influence the difficulty and biases of pilots' conflict detection using a cockpit display of traffic information (CDTI). There was previously no systematic study on effects of such features on conflict detection using the CDTI. Twenty-four pilots viewed dynamic encounters between their own aircraft and an intruder aircraft on a simulated CDTI; difficulty was varied by the intruder aircraft's distance and time to closest point of approach (CPA), relative speed, miss distance at CPA, approach side, and conflict angle. Participants estimated the intruder's location at, and its time to, CPA. Effects on three estimation error measures were explored: intruder's miss distance at CPA, orientation at CPA, and time to CPA. (a) Estimation errors increased with slower speeds, longer times to CPA, and longer distances to CPA and with longer miss distances at CPA; (b) the best performance occurred at a conflict angle of 90 degrees; (c) there was a bias to judge conflicts to be more risky than was actually the case; and (d) there was a "distance-over-speed" bias, such that two aircraft farther apart and converging rapidly were perceived as less risky than when they were closer to each other and converging at a slower rate, despite identical time to CPA. Pilots' conflict detection with CDTI was subject to various errors and biases, which has important safety implications. The design of procedures, displays, and decision support tools for the free flight environment needs to take these human performance limitations into account.