Within the set of all vehicle crashes that occur annually, of intersection-related crashes are over-represented. The research conducted here uses an empirical approach to study driver behavior at intersections, in a naturalistic paradigm. A data-mining algorithm was used to aggregate the data from two different naturalistic databases to obtain instances of unprotected turns at the same intersection. Several dependent variables were analyzed which included visual entropy, mean-duration of glances to locations in the driver's view, gap-acceptance/rejection time. Kinematic dependent variables include peak/average speed, and peak longitudinal and lateral acceleration. Results indicated that visual entropy and peak speed differs amongst drivers of the three age-groups (older, middle-age, teens) in the presence of traffic in the intersecting streams while negotiating a left turn. Although not significant, but approaching significance, were differences in gap acceptance times, with the older driver accepting larger gaps compared to the younger teen drivers. Significant differences were observed for peak speed and average speed during a left turn, with younger drivers exhibiting higher values for both. Overall, this research has resulted in contribution towards two types of engineering application. Firstly, the analyses of traffic levels, gap acceptance, and gap non-acceptance represented exploratory efforts, ones that ventured into new areas of technical content, using newly available naturalistic driving data. Secondly, the findings from this thesis are among the few that can be used to inform the further development, refinement, and testing of technology (and training) solutions intended to assist drivers in making successful turns and avoiding crashes at intersections. / Master of Science