The external errors on the mean abundances are obtained by propagating
the uncertainties due to the measured 
, the atomic
parameters, and stellar 
(which affects the ionization balance),
assuming they are independent of one other. Following the SD93
convention, we have adopted estimates of the errors on each parameter
as follows: 
0.25dex in 
, 
250K in 
, 
50
percent in the Stark damping (
),
0.5kms
in
the microturbulence (
), and 
5 percent in
. Propagating these errors through the curve-of-growth
analysis for the Mn II 
4478
line, using a model atmosphere
based around 
= 13000K, 
= 4.0, 
=1, with a
+2dex Mn abundance enhancement over the solar value, leads to the
following representative errors in the derived Mn abundances:
0.01dex (
); 
0.02dex (
); 
0.05dex
(
); 
0.06dex (
); 
0.06dex (
).
To obtain the internal error on the mean value of abundance for each star
we calculate the standard deviation of the set of measured lines
(Table 2). The mean standard deviation over the entire
sample is 
0.09 dex. Remarkably,
there is no significant overall
difference between the mean abundances derived from the Mn I and Mn II
lines, despite the fact that the 
-sensitivity of MnI is
0.15dex for variations of 
250K. There is no evidence for
systematic differences between abundances derived from the different lines
in Table 2; the average deviations, 
, are only a few hundredths of a dex.
Finally, we can estimate a purely `experimental' error on the abundance
determinations for individual lines by varying the abundance values used
in our synthetic fits (see Section 3.4) and seeing how rapidly the
synthetic profiles depart from the observed profiles. For visual fitting
of synthetic profiles to the observed data, changes of approximately
0.02-0.05dex produce a significantly degraded quality of fit,
depending upon the S/N of the spectrum.
These comparisons give us confidence that the abundances we have
derived are internally consistent with expected errors of
observation. It is remarkable that there appears to be little
significant extra scatter due to errors in 
. Throughout this
analysis we assume that Mn is homogeneously distributed with
depth. Stratification of elements by diffusion and gravitational
settling, and non-LTE, are both possible complications which must be
considered in such analyses. However, we do not maintain that the
full agreement of Mn I and Mn II `proves' that there are no
stratification or non-LTE effects on the Mn I abundances compared to
Mn II, only that if any such effects are present, either they are
small or the factors involved apparently cancel out over a wide range
of 
. Detailed considerations of such possibilities are beyond the
scope of this paper.