In the right atrium (RA) we globally investigated: (1) the properties of noncontact electrograms measured by multielectrode cavitary probes, (2) the features of endocardial electrograms computed from the noncontact probe electrograms, and (3) the impact of the probe size on both the noncontact and the computed electrograms. We deployed a custom catheter in the dog RA, which consisted of a cylindrical probe with 64 electrodes on its surface, for measuring noncontact cavitary electrograms, and a concentric endocardial basket carrying an additional array of 64 electrodes, for measuring contact endocardial electrograms (the "gold standard"). Both a 5-mm- and a 10-mm-diam probe (P5 and P10, respectively) were sequentially tested in the same RA of one dog. Unipolar electrograms from both the probe and the basket were simultaneously acquired during normal as well as during paced rhythms (n > or =24 protocols per probe). Boundary element method and numeric regularization were applied to compute endocardial electrograms at the basket electrode locations. We found that noncontact electrograms were attenuated and smoothed, and this effect was exaggerated with the small probe. Computed endocardial electrograms more accurately reconstructed important amplitude distribution and morphological features; peak-to-peak amplitude error, 35% for P5 and 34% for P10. Activation and spatial errors of computed endocardial electrograms were 8.8+/-6.8 ms and 5.1+/-6.1 mm for P5, respectively, and 6.0+/-5.5 ms and 3.2+/-4.4 mm for P10, respectively. In conclusion, global RA activation may be delineated directly from noncontact cavitary electrograms alone, but may be affected by volume attenuation, smoothing, and probe size. Accurate endocardial electrograms, however, can be successfully computed from noncontact electrograms acquired with small probes and be used to reconstruct both electrogram amplitude and detailed morphology.