We have analyzed spectra of 83 subdwarfs obtained with the Lick Observatory 3.1-m Shane telescope for C, N, and Fe abundances. Our stellar sample is composed principally of stars in the lowest metallicity domain, -1.5 > [M/H] > -3.2. Applying the model-atmosphere technique of spectrum synthesis to our intermediate-resolution spectra, we deduced [C/Fe] from the CH bands (X ~ 4300 À), [N/Fe] from the NH bands (X ~ 3360 À), and [Fe/H] from strong mostly Fe I features (3400 A < X < 3550 A). We find excellent agreement between our [Fe/H] values and the results obtained by other investigators based on high-resolution spectra. We find [C/Fe] to be essentially constant, with a scatter of cr ~ 0.18 dex, over the range -0.75 > [Fe/H] > -2.5, confirming that C is a primary nucleosynthetic species in this metallicity domain. We find a distinct upturn in the [C/Fe] values at the very lowest metallicities, i.e., [Fe/H] ^ -2.0. This behavior suggests a contribution to [C/Fe] from comparatively high-mass stars at early epochs. We find that the behavior of <[N/Fe]> as [Fe/H] declines is compatible with the assumption of a primary, as opposed to a secondary, origin of N, but the limiting value of <[N/Fe]> is found to be several tenths dex smaller than in previous investigations. We confirm that HD 74000 has [N/Fe] distinctively outside the range of typical subdwarfs, as does HD 25329. However, we found very few stars in our comparatively large sample with very high [N/Fe] indicating that such objects, whatever their origin, are quite atypical. Finally, we find strong evidence in our study and in the studies by Laird and by Tomkin and Lambert for a T eír dependence in the deduced subdwarf [N/Fe] values. This dependence, if not due to an analysis error common to the three investigations, may be evidence for the mixing of C -» N processed material to the stellar surface on the subdwarf main sequence. Key words: abundances-metal-poor stars-subdwarf stars such analyses have been based exclusively on high-resolution spectra and have been limited, perforce, to quite small samples of stars. Nevertheless, there has been evidence for some time that carbon and iron abundances keep close pace with one another (Peterson and Sneden 1978; Clegg, Lambert, and Tomkin 1981; Barbuy 1981) and that oxygen becomes somewhat overabundant relative to carbon and iron as [Fe/H] decreases (Sneden, Lambert, and Whitaker 1979). These results strongly suggest a primary origin for C and O in the halo and disk.The problem of small sample sizes has been most serious in the case of nitrogen. From a study of CN band