The CIE has compiled color matching functions (CMFs) at one nm sampling to be used for computational color. Since most instruments produce sampled data at resolutions considerably coarser than one nm, the recorded data must be interpolated to finer resolution before the multiply and sum required to compute the tristimulus values. This paper points out that the interpolation which should be used depends on the characteristics of the CIE interpolations which produced the CMFs, the characteristics of the reflectance signals and the noise associated with the recording instrument.