High resolution spectra of nineteen metal deficient proper motion stars with - 3.70 less-than-or-equal-to [Fe/H] less-than-or-equal-to - 1.92 (mean [Fe/H] = - 2.9) have been analysed to determine abundances for iron and other elements. Most are based on data obtained with the UCL coude echelle spectrograph on the Anglo-Australian Telescope, operated at a resolving power of 55 000. The sample trebles the number of stars with [Fe/H] < - 3.0 for which fine abundance analyses have been performed. The comparison of results from various studies demands great caution because of systematic differences between programs. They reveal themselves as differences in the mean element ratios derived for stars in a narrow [Fe/H] interval. Since separate studies involve different metallicity ranges, systematic offsets can be misinterpreted as being intrinsic trends with [Fe/H]. Until more homogeneous data are available for stars over a wide range of [Fe/H], such apparent trends must be regarded with suspicion. There is currently little convincing evidence for variations of most relative abundance ratios of elements in stars with [Fe/H] < - 1.0, with a few exceptions. Specific results concerning the relative abundances of elements are as follows: (1) We find mild support for the claim by Tomkin et al. [ApJ, 302, 415 (1986)] and Carbon et al. [PASP, 99, 335 (1987)] of an upturn in the [C/Fe] ratio at [Fe/H] = - 2.0. This might reflect overproduction of carbon in the supernovae of 200M., zero initial metallicity stars. (2) The abundance ratios of Mg, Si, Ca, Ti, Cr, Sc, and Mn relative to Fe are all seen to continue the uniform trends previously observed for stars with - 2.5 less-than-or-equal-to [Fe/H] less-than-or-equal-to - 1.0, down to at least [Fe/H] = - 3.5. (3) Although our [Ni/Fe] abundances are higher than for some other analyses of more metal rich halo stars, the element does not appear as overabundant as Luck & Bond [ApJL, 271, L175 (1983); ApJ, 292, 559 (1985)] claimed. Our high [Ni/Fe] ratios probably reflect a systematic difference between our analysis and those with which we compare it. (4) Al is underabundant with respect to both Fe and Mg, consistent with the predictions of explosive nucleosynthesis in metal-weak supernovae. It appears to reach a constant value [Al/Mg] congruent-to - 1.0 dex at [Mg/H] less-than-or-equal-to approximately - 2.0. (5) We confirm that halo dwarfs have the same underabundance of [Mn/Fe] as field giants, - 0.33 dex. (6) The spread in relative abundances of the neutron capture element strontium in the most metal-deficient stars is greater than 1 decade. In particular, the two most metal-weak stars in the sample, with [Fe/H] = - 3.70 and - 3.40, have no measurable Sr II resonance lines. We tenatively suggest that these may be second generation stars, i.e., they formed from material which had experienced only a single episode of enrichment. They confirm the existence of large inhomogeneities in the chemical composition of the Galaxy in its earliest epochs.
