Pedestrian delay at signalized intersections must often be considered when describing the quality of pedestrian flow in downtown and other busy locations. Pedestrian delay at signals is usually modeled with the assumption that (a) pedestrian arrivals are random and (b) pedestrians proceed only when a Walk signal provides the right-of-way. To reduce their own delay, some pedestrians will at times proceed without a Walk indication. The range of delay reductions that pedestrians achieve by violating the pedestrian signal was examined through a field study of 18 crosswalks. Delays experienced by pedestrians were found to be 22 percent lower than those that would be predicted with complete signal compliance. Although relatively small reductions in delay were due to crossings while the signal was effectively red, the majority of delay reduction was due to people who began their crossing during the flashing Don't Walk phase. The number of pedestrians who entered crosswalks during the flashing Don't Walk phases was 69 percent of the number who would be expected to arrive at the curb during these periods if arrivals were random. It is inferred that some pedestrians increase their speeds to enter during the flashing Don't Walk phase, rather than wait for the next Walk interval. A modification to a delay equation was developed to provide an improved estimate of pedestrian delay at signalized intersections.