All available paleomagnetic data for the last 130 m.y. meeting certain selection criteria are drawn from standard compilations and returned to precontinental drift site locations and orientations using spreading poles based on sea floor magnetic anomalies. The resulting data set is divided into six time periods, and global patterns of inclination and declination anomalies are used to identify important nondipole components. Spherical harmonic analysis is then performed to evaluate the corresponding coefficients. An axially symmetry quadrupole component (g20) is found to be important for at least the last 100 m.y., and an axially symmetric octupole component (g30) for the last 50 m.y. Possible correlations are noted between the variation with time of the reversal frequency of the geomagnetic field, compiled by A. Cox, the polarity bias published by Irving and Pullaiah, and our nondipole field magnetude. Both current explanations for the nondipole components seen in paleomagnetic data, the offset dipole model and Cox's zonal nondipole model, are incompatible with the large values of g30 found in this study. However, as a possible alternative, Cox's model can be generalized to accomodate the new information.