Most of the paleomagnetic data available from oceanic plates consist of measurements of inclination but not declination from azimuthally unoriented vertical cores. Similarly, the basic paleomagnetic data to be analyzed from crustal blocks that have rotated by varying amounts about vertical axes are a set of paleoinclinations. A new method is proposed for determining the paleocolatitude of a sampling site and the associated confidence limits. Paleomagnetic inclinations are first converted to paleocolatitudes and averaged to produce a mean colatitude. After showing that this mean is biased, we derive a paleocolatitude correction to remove the bias, using a model for geomagnetic secular variation based on the latitude dependence of secular variation during the past 5 million years. After testing the validity of the model by applying it to paleomagnetic data from Quaternary lavas for which both inclination and declination are known, we apply the method to inclination‐only paleomagnetic data from Suiko Seamount, the most complete data set from a drilling site. The method is then applied to paleomagnetic data from all Deep Sea Drilling Project sites for the Pacific plate to obtain for each site an unbiased mean paleocolatitude and its confidence limits. These paleocolatitudes and confidence limits are then used as input for a least squares analysis to obtain Early and Late Cretaceous poles for the Pacific plate and 95% confidence limits for these poles. It is shown that the scatter of individual site paleocolatitudes about the best fit pole is consistent with the confidence limits at each site as determined by the new method of analysi