Studies of the "testing effect" have shown that retrieval significantly improves learning. However, most of these studies have been restricted to simple types of declarative verbal knowledge. Five experiments were designed to explore whether testing improves acquisition of route knowledge, which has a procedural component consisting of actions to be performed at decision points (Golledge, 1991). Participants learned a route through a series of connected rooms in a virtual building. Each room contained multiple doors, only one of which led to the next room. During encoding, participants were shown the correct sequence of doors in a manner similar to global positioning system (GPS) navigation guidance. During subsequent exposures to the route, participants were either shown the correct sequence again or had to recall the sequence from memory. Participants later completed a final test in which they traversed the route without guidance or feedback. Testing improved route memory compared to studying, but only when participants were given feedback about the correct door prior to moving through the room. When feedback occurred after moving to an incorrect door, testing resulted in worse performance compared to studying. These findings parallel work on errorless learning, in which procedural skills are acquired more quickly when errors are minimized during learning.