A mobile recording system, with integrated laser speed gun, video from CCD-cameras and auxiliary battery system, was used to observe driving behavior at an urban intersection and a suburban analog. After removal of instances of interference, 1538 driving behaviors were recorded. Multiple regression was then utilized to examine the factors affecting approaching speed. Speed limit violation was considered a dichotomous variable with two categories, violation and compliance. Binary logistic regression was also used to examine the risk of speeding as a function of covariates and interaction terms. The results of analysis revealed that the major contributing factors for approaching speed were site, rush-hour-status, traffic light condition, vehicle type and driver gender. In particular, light status was the highest contributor to speed. In addition, the results of logistic regression showed significant sites and rush-hour effects on speeding, with the risk of limit violation in the suburbs nearly six-fold that in urban areas. The relative risk of speeding for travelling in non-rush hours is three times higher than that for rush-hour. In terms of driver characteristics, male drivers under 55 years of age had the greatest speeding propensity in our sample. The results of the present study may provide meaningful information applicable to the design and operation of signalized intersections.