A two-phase study of driver behavior at intersections was conducted using a naturalistic paradigm. In Phase 1, the behavior of teen, middle-aged, and older drivers was compared for left turns at three types of left turns across path (LTAP) intersections. Phase II was a follow-on effort focused more narrowly on a single T-intersection that included both left and right turns. A data-mining algorithm was used to aggregate data from two different naturalistic databases to obtain instances of unprotected turns at the same intersection and instances from a comparison protected turn at a signalized intersection. Several dependent variables were analyzed, including visual scanning measures, head-turning measures, speed approaching and driving through the intersection, and gap acceptance and rejection times. Results from Phase 1 show that driver behavior differed between the unprotected turns and the comparison protected turn. The two types of unprotected turns also had different effects on scan patterns and glance durations through the initiation, conflict, and completion zones of the turn. Definitive age-group effects were seen with head turning. The Phase 1 results suggest that older drivers scanned more narrowly and that they strategically oriented their scans in the direction of greatest threat for certain types of unprotected left turns. Results from Phase 2 replicated those from Phase 1. For left turns, most driver groups had a similar distribution of glances by location for this T-intersection, with the forward-looking glances constituting almost half of all glances, followed by glances through the window areas. However, older drivers showed a higher proportion of glances to the right, indicating a strategic shift of attention that was consistent with the head-turn analysis in Phase 1. Link analyses done on the visual scan patterns of drivers indicated that older drivers and middle-aged drivers had more-organized and more-strategic visual scans of the intersection compared with young drivers. While middle-aged drivers and younger drivers scanned broadly, older drivers had a narrower scan that was oriented slightly toward the right for a left turn and to the left for a right turn, perhaps indicating attention to gap selection and threats in the stream of traffic into which they were merging. Visual entropy analyses showed that active visual scanning to a larger number of areas increased for all age groups when traffic was present. This increase was most pronounced for younger drivers. Significant differences were observed for peak speed and average speed during a left turn, with older drivers using lower speeds throughout intersections and completing their turns at lower speeds than other age groups. These findings suggest that the behavior of older drivers is at least partly strategic in nature. The changes in scanning exhibited by older drivers appear to be more focused “versions” of the visual scanning typical of middle-aged drivers, tuned specifically to areas of highest threat during turns. However, this may lead to some neglect of other areas, and it is possible that age-related changes in perception and cognition may be contributing to some of the differences.