Data obtained from functional magnetic resonance imaging are often limited by a low signal-to-noise ratio. The time-course data obtained from activated regions contain both system noise and physiological noise, primarily linked to the heart and respiratory rates, that are superimposed on task induced signals. Time averaging of a practical number of data sets is not very effective in improving the signal-to-noise ratio because neither system nor physiological noise is truly random. In this paper, a method is presented for filtering unwanted physiological fluctuations, including aliased signals that are formed as a result of long repetition time (TR) values. A pulse oximeter was used to obtain cardiac and respiratory information during the scanning period. Finite impulse response band-reject digital filters were designed to remove the physiological fluctuations. For comparison, cross-correlation analyses were performed at the same level of statistical significance on both filtered and unfiltered data. We demonstrate that this method can improve the detection of weak signals without increasing the probability of false positives.