The abundance of Be in the lowest-metallicity stars is a probe of Big Bang Nucleosynthesis, and its abundance in halo and disk stars is a probe of galactic evolution and stellar structure. We present observations of the Be II resonance lines in 14 halo stars and 27 (mostly old) disk stars with [Fe/H] from - 2.7 to + 0.13. The spectra were obtained at the Canada-France-Hawaii 3.6 m telescope and have a measured resolution of 0.13 angstrom and a median signal-to-noise ratio of approximately 50. For 18 of the 41 stars we have also made observations of the O I triplet at the Palomar 5 m telescope, the UH 2.2 m telescope, and the CFH telescope. Stellar parameters of T(eff), log g, and [Fe/H] were carefully determined from several independent estimates. Abundances are determined for log N(Be/H) and [O/H] from measured equivalent widths, model parameters, and Kurucz (1991) model atmospheres with the RAI10 model atmosphere abundance program. The agreement with previously published Be detections is very good (a mean difference of 0.05 dex) for five of six determinations in four halo stars and in four of five disk stars. Our typical internal error from uncertainities in the stellar parameters and the S/N ratio of the observed spectra is 0.10 dex. The agreement with very recently published O abundances is 0.07(5) dex. It is plausible, but far from conclusive, that there is a plateau in the amount of Be present in the lowest metallicity stars: log N(Be/H) approximately - 12.8 for [Fe/H] < - 2.2. As [Fe/H] increases from - 2.2 to - 1.0, log N(Be/H) increases and the slope is 1.2-1.3, indicating a faster increase in Be than in Fe. This is consistent with the production of Be by spallation reactions between cosmic rays and 0 atoms from massive stars and the production of Fe from intermediate mass stars. The Be and Fe data can also be represented by N(Be) proportional to N(Fe)2.0 as [Fe/H] goes from - 3.0 to - 2.0 and N(Be) proportional to N(Fe)1.0 as [Fe/H] goes from - 2.0 to - 1.0. The disk stars ([Fe/H] = - 1.0 to 0.0) show a spread in Be of nearly an order of magnitude; part of this must be real since the typical uncertainity is 0.10 dex. Evidence for stellar processing of Be exists in the disk stars and in at least two of the halo stars. A plot of Be abundance vs 0 abundances shows that Be increases as O1.12, indicating that Be is produced primarily in the vicinity of supernovae envelopes, but a small and interesting fraction is produced in the general interstellar gas in the halo.
