The CTIO 4 m telescope, RC spectrograph, and the "2D-FRUTTI" detector have been used to obtain spectra for 20 47 Tuc (= NGC 104 = C0021 - 723) stars. The stars observed were 10 main-sequence (MS), four luminous red-giants, and six subgiants, giving a total observed range of 12.1 < V < 18.5 (- 1.2 < M(v) < 5.2). Variations in CN and CH band strength were detected across the entire range. The CN and CH band strengths have been analyzed using the S(3839) and S(CH) spectral indices, which have been compared to indices from corresponding synthetic spectra to obtain abundances. An Mg-sensitive index, Mg2, has also been measured, with the result that there is no observed correlation between Mg2 and CN-band strength. The observed C and N abundance variations are impossible to reproduce with the assumptions of mixing involving only C --> N-processed material and initial C and N abundances appropriate to either the CN-normal stars or the solar ratio. Adding the possibility of O --> N-cycle mixing will allow total C + N + O to be conserved. However, abandoning the assumption of CNO in the solar ratio will allow CN-process mixing to result in the observed C and N abundances while maintaining constant C + N, but mixing of the CN-normal stars is then also required. The distribution of CN band strengths appears to be consistent with the existence on the upper MS of the 3883 angstrom CN bimodality, previously observed throughout the more evolved regions of the CM-diagram. Any mixing episode(s) hypothesized to account for the CN and CH variations must therefore have taken place before the upper MS. The C and N abundance differences are very similar to the differences observed among the more luminous cluster stars. Any discussion of mechanisms for altering the surface C, N, and O abundances (as well as isotopic ratios) of evolved stars must allow for a significant population of upper MS stars with strong CN bands, which would appear to represent a greater challenge at present for mixing, rather than for primordial, mechanisms.