Though collisions in wireless networks can be caused either by known stations within the transmission range or by unknown hidden stations, rate adaptation algorithms that are designed to adapt to varying channel quality deal with both cases in the same manner so that they do not effectively respond to the different channel conditions. In this paper, after examining the core issues regarding rate adaptation, we propose a novel rate adaptation algorithm, called RACD (Rate Adaptation with Collision Differentiation), with the ability to differentiate the sources of collision by utilizing a new feature of IEEE 802.11, CTS-to-self frame, and adjusting the size of the contention window accordingly. Our simulation results show that RACD outperforms previous rate adaptation schemes, in terms of data throughput, by effectively distinguishing the cause for adverse channel conditions.