An empirical study was conducted of driver behavioral theory of multi-lane freeway traffic flow and capacity drop. The study focused on a freeway segment with a horizontal curve located in a stretch of the O-1 highway approaching the European side of the Bosporus Bridge in Istanbul, Turkey. The stretch behaves as a recurrent bottleneck in the afternoon peak. Flow data were extracted manually from video recordings taken from stations along the highway stretch and then converted into cumulative vehicle count curves constructed for each lane at the count stations. Radar detector data taken from a point within the horizontal curve were used to construct an occupancy-flow diagram. The empirical results show that drivers behave in some respects as outlined in the theory. For example, drivers in the median lane do not redistribute themselves across lanes although their speeds are below those in the shoulder lane, as opposed to the theory, but redistribution occurs later. This implies that more than speed differential across lanes affects driver behavior in the median lane. Flow differences and then equalization were observed in the median and shoulder lanes in free-flow and congested-flow states, respectively. There was a fast-moving queue in the median lane in a semicongested state while free-flow conditions in the shoulder lane were prevailing, in harmony with the theory. A relation between capacity drop due to bottleneck activation and bottleneck segment density is also presented.