Rolling element bearings are an essential component in rotating machinery and their failure is one of the most common reasons for machine breakdown. To minimize both the risk of breakdown and maintenance costs, predictive maintenance based on condition monitoring is a useful tool. Current methods for bearing condition monitoring typically require human interaction or prior knowledge of the system as well as specialized measurement equipment. In addition, their performance at low speeds is often relatively poor. A new patented method based on adaptive filtering has been developed by LeBlanc and Pääjärvi. This method works with standard measurement equipment and requires no prior knowledge of the system. Preliminary results have been good, and this thesis is a more thorough investigation of the low speed performance. The performance was evaluated using extensive measurements from a test rig as well as field measurements from a number of real machines operating at varying speeds. Using the test rig, outer race defects were studied at speeds from 73 to 8.6 rpm and inner and combined race defects were studied from 73 to 23.2 rpm. The results show that detectability is very good and that, contrary to the general direction in the industry, lower sampling frequencies should be used to improve detectability at low speeds with this method. Results from the field measurements prove that the method works not only for simple structures such as the test rig, but also for real, more complex systems and extremely high noise environments.