Thermal demagnetization of red beds (109 specimens, 42 sites) from the Maritimes and Gaspé shows that the fossil magnetization consists of a large, thermally stable component and of small, unstable components that disappear at about 400 °C. Thermal cleaning of the whole collection was therefore carried out at 450 °C to demagnetize these soft components. This procedure changes the mean declination and inclination by 6° and 7° respectively and also changes the distribution of directions about the mean. A statistical analysis shows that the site mean directions (after demagnetization) obey a Fisher's (1953) distribution whereas the individual specimen directions do not. This result is of interest in two connections: first, it shows that, in this instance, the use of Fisher's (1953) statistics to analyse the results is not permissible when unit weight is given to the specimen, but that their use is legitimate when unit weight is given to the site; second, the result is consistent with the hypothesis that the earth's field, when sampled in this way (paleomagneticalty), can be represented by an axial geocentric dipole field perturbed by randomly directed magnetic components at the earth's surface. Comparison with other results available for this geological period shows a certain divergence among the paleomagnetic poles for North America. It is suggested that a partial demagnetization of these earlier collections might reduce the observed divergence.