Using high-resolution, moderate signal-to-noise ratio spectroscopy obt
ained with the 10 m Keck I Telescope and efficient HIRES echelle spect
rograph, we derive abundances of several elements in subgiants near th
e M92 turnoff. As a consistency check, we also analyze the metal-poor
field star HD 140283 and find an Fe abundance in fine agreement with m
any previous determinations. However, our M92 value ([Fe/H] = -2.52) i
s a factor of 2 lower than the abundance derived from its red giant me
mbers. Differences in model atmospheres, gf-values, and instrumental e
ffects might account for this difference, but whether they in fact do
so is unclear. We note possible evidence for [Fe/H] differences within
M92. Our spectroscopic analysis suggests that the M92 reddening, E(B-
V), may be 0.04-0.05 mag greater than canonical values, but various un
certainties mean that this conclusion is not definitive; the significa
nt difference in interstellar Na I line strengths in the M92 and HD 14
0283 spectra may be consistent with an increased reddening. Regardless
, the conclusion that either the [Fe/H] of M92 has been significantly
overestimated from red giants or current reddening/photometry estimate
s are too small/red is not easily escaped. If the reddening/photometry
were in error by this amount, turnoff color-based ages for M92 could
be reduced by similar to 4 Gyr. The adjustment to the M92 distance mod
ulus required for a similarly reduced turnoff age that is luminosity-b
ased can be accommodated by increases in extinction and alterations to
the metal-poor field star distance scale recently inferred from Hippa
rcos Cepheid and subdwarf data. Our M92 subgiants demonstrate [Cr/Fe],
[Ca/Fe], and [Ti/Fe] ratios that are unremarkable and essentially ide
ntical to the values for HD 140283. [Ba/Fe] is 0.45 dex larger for the
M92 subgiants than for HD 140283. Surprisingly, we find [Mg/Fe] to be
0.55 dex lower in our M92 subgiants than in HD 140283, and [Na/Fe] to
be 0.76 dex larger in our M92 subgiants than in HD 140283. These diff
erences (and indeed nearly all our abundance ratios) seem immune to va
rious data, analysis, and parameter errors. If real, this striking abu
ndance pattern is suggestive of material in our M92 stars' photosphere
s that has undergone Ne --> Na and ME --> Al cycling like that inferre
d for red giants in M92 and other clusters. While this is generally be
lieved to be an in situ process in cluster giants, the presence of abu
ndant Li in our M92 objects suggests a polluting source acting either
primordially or via accretion after cluster star formation. This may b
e consistent with CN and Na variations on the 47 Tucanae main sequence
, recently reported Ba and Eu variations in M15 red giants, possible c
luster-to-cluster n-capture abundance differences, and very low [O/Fe]
ratios observed near the base of the M13 giant branch. We thus sugges
t that a polluting source of light-element alteration, in addition to
the in situ source for more evolved stars, may be required for M92. Co
mparison of our M92 subgiant abundance ratios with those of M92 red gi
ants may indicate that pollution occurred after the present generation
of cluster stars formed, but until the cause or causes of the subgian
t versus giant Fe abundance discrepancy are definitively identified, t
his conclusion is uncertain. A polluting source of our Na and Mg anoma
lies produced via processing in a previous stellar generation also has
complications; namely, how the Mg and Na anomalies arise without appa
rently any net influence on our subgiants' Li abundances and on the C
abundances of other M92 subgiants. A similar quandary may exist in som
e 47 Tuc turnoff stars. An understanding of cluster abundance variatio
ns (by whatever mechanisms) and their behavior with evolutionary state
may be needed for a complete understanding of absolute and relative g
lobular clusters ages, and for derivation of the primordial Li abundan
ce.