In a number of studies, various workers have used the procedure of fitting cubic splines with nodes spaced 12 months apart in order to remove seasonal effects from atmospheric CO2 observations. A number of trial examples of least squares spline fits are presented, indicating that this procedure can be subject to relatively large end effects. For variations with periods near 2 years the response is extremely sensitive to the position of the nodes and shows beating effects caused by the change in the relative positions of the signal peaks and the spline nodes. A signal whose period is exactly twice the node spacing can either be passed almost unchanged or be completely filtered out, depending on its phase relative to the nodes of the spline. These properties make such spline fitting unsuitable for analyzing the interannual variations in CO2 data.