Arterial tonometry provides for the continuous and noninvasive recording of the arterial pressure waveform. However, tonometers are affected by motion artifact that degrades the signal. An arterial tonometer was constructed using two piezoelectric transducers centered within a solid base. In two subjects, one transducer was positioned over the radial pulse (p) and the other was positioned on the wrist not overlying the pulse (n). The presence of induced motion artifact and any noise was removed after signal digitization by noise cancellation. Besides fixed weighting, two adaptive algorithms were used for cancellation-LMS and differential steepest descent (DSD). Criteria were developed for comparison of the adaptive techniques. The best fixed weighting for noise cancellation was w=0.6. For fixed-weighting, LMS, and DSD, the mean peak-to-peak errors were 1.22+/-0.54, 1.18+/-0.30, and 1.16+/-0.23 V, respectively, and the mean point-to-point errors were 15.86+/-3.15, 11.40+/-1.96, and 10.13+/-1.25 V, respectively. Noise cancellation using a common-mode reference input substantially reduces motion artifact and other noise from the acquired tonometric arterial pulse signal. Adaptive weighting provides better cancellation than fixed weighting, likely because the mechanical gain at the transducer-skin interface is time-varying.