We have been concerned here with the interpretation of data to give
stellar temperatures that are as accurate as we can manage, and with
an attempt at assessing their relative and absolute accuracies as
reliably as possible. No account has been taken of inaccuracies in the
original data, for we judge that the largest contributions to errors
in the final temperatures originate in the theoretical interpretion of data.
The derivation of monochromatic fluxes from observations made using wide
wavelength bands is difficult and uncertain, and may introduce unacceptably
large errors which depend on uncertain line strengths. Our judgement of
the relative accuracies of the final temperatures is reasonably well based
and may be regarded as reliable. However, the more difficult task is the
assessment of final absolute probable errors. Some contributions to these
errors are clear. For example, [Fe/H] values are not available for many
stars in the observed list, and even the solar iron abundance is now
regarded as uncertain (Blackwell et al. 1995);
where no [Fe/H] value is
available, we have given an accuracy of 0.9 per cent. A theoretically
derived atmosphere is needed, and although these are now more reliable
than formerly, some uncertainty about their use remains. An important and
extensive investigation of the accuracy of model predictions has been
made by van der Bliek et al. (1996),
following on the earlier work
by Mégessier (1994).
In the present paper we suggest that the wholly
empirical use of plots of temperature against intrinsic photometric indices,
e.g. 
and 
, provide useful evidence for
relative accuracies, whilst comparisons with measures made by a
variety of methods are useful for detecting gross errors.
Acknowledgements
We are indebted to the ISO Ground Based
Preparatory Programme which
has entrusted us with the work reported here. Rapid progrsss would not
have been possible without the generosity of Drs. van der Bliek
and P.L. Hammersley, who kindly made their observations available in
advance of publication. Hipparcos parallaxes were provided by
Dr. Perryman, also in advance of publication. Dr. Marie Jourdain de
Muizon gave constant and patient help during the investigation and
the authors are especially grateful to her. Obtaining infrared
monochromatic flux densities would have been more difficult
without Dr. Hammersley's advice. The authors have also benefitted
from correspondence with Professor G.P. di Benedetto.
Comments and suggestions by an anonymous referee have contributed
to the present paper.