High signal-to-noise (S/N approximate to 80), high-resolution (R appro
ximate to 25 000) spectra of six warm halo stars have been obtained in
the Be II lambda 3130 region. Beryllium abundances derived through sp
ectrum synthesis are generally consistent with [Be/Fe]approximate to 0
and probably do not support the simplest models of cosmic-ray nucleos
ynthesis. Alternative scenarios whereby energetic C, N, and O nuclei c
ollide with ambient He nuclei-rather than the converse-are more compat
ible with the data. However, these observations constrain net Be produ
ction only. If Galactic astration of Be is important, then Be producti
on mechanisms must be proportionally more efficient. In the case of th
e near twin stars HD 94028 and HD 194598, log(Be/H) differs by 0.3 dex
despite effectively identical Li abundances. A difference in initial
Be abundance rather than stellar depletion is proposed to account for
the discrepancy. Our Be abundances are typically similar to 0.3 dex lo
wer than those reported by Boesgaard & King [AJ, 106, 2309 (1993)] for
five stars in common. The disparity in Be abundance scales is due to
the combined influences of differences in assigned stellar parameters
and overestimated Be II lambda 3131.066 line strengths in the Boesgaar
d & King study. Systematic errors in published Be abundances may be gr
eater than has previously been suggested. A consistency check of the l
ine list used for the synthesis reveals an unidentified blending featu
re of moderate strength slightly blueward of lambda 3131.066 in the Su
n. Results of the present investigation are not significantly influenc
ed by the ambiguous identity of the contaminating feature. However, st
udies which include cooler, more metal-rich stars may, as a consequenc
e of neglecting or improperly identifying this line, report incorrect
slopes of log(Be/H) vs [Fe/H] or erroneously large scatter in log(Be/H
). (C) 1996 American Astronomical Society.