A number of shortcomings associated with the use of smoothing and least squares splines for data analysis in general, and also in some specific past CO2 analyses are considered. A potentially useful alternative, through generalized smoothing (GeS) splines, which do not require the data and spline knots to coincide, is offered. This may be computationally more cost effective and, with appropriate care, produces equivalent results. The presented examples examine fits to unselected Cape Grim CO2 data and a number of preprocessed cases are also compared. In all these cases the GeS splines produce a stable output, unlike the smoothing spline, which does not possess the capability of directly handling very large data sets. A range of diagnostics may be applied to a problem to evaluate the suitability of the GeS spline as a replacement for the conventional smoothing spline in data smoothing. A proof of asymptotic equivalence between the smoothing and generalized smoothing splines is also given as an appendix. ¿ American Geophysical Union 1995