The majority of detailed analyses of OB stars have been subject to the
assumption of hydrostatic equilibrium in a plane-parallel atmosphere (e.g.
Herrero et al. 1992; Puls et al. 1996). However, should stellar winds
be strong, absorption lines will be contaminated by wind effects
(Schaerer & Schmutz 1994; Crowther & Bohannan 1997).
Optical HeI-II lines generally provide temperature diagnostics for
O stars (e.g. Herrero et al. 1992) while SiII-IV lines
are used for B stars (e.g. Becker & Butler 1990), though
Kudritzki (1992) discussed the complicating role of velocity fields for investigating supergiants. Kilian et al. (1991)
employed both He and Si diagnostics for early B dwarfs and giants
which agreed relatively well, though with differences of up to 1800K.
Kilian et al. (1991) and Grigsby et al. (1992) used line blanketed
atmospheres to obtain effective temperatures systematically higher
(2000K) than those
from 
--
photometric calibrations (e.g. Lester et al.
1986). The 
scale for B supergiants has been
improved upon recently by McErlean et al. (1997) who used SiIII-IV
diagnostics for early-types, with non-LTE corrections to results
from the Lester et al. (1986) calibration applied to mid/late B supergiants.
Recent ionization equilibrium techniques for O stars represent considerable
advances relative to earlier (insensitive) continuum methods.
For example, Bohannan et al. (1986) obtained 
=42kK from a non-LTE
wind blanketed analysis of 
Pup
(O4I(n)f), with 
=32--50kK obtained from continuum techniques.
However, widely varying results have also been
found in recent non-LTE spectroscopic analyses of late O/early B dwarfs.
For instance, for HD214680 (10Lac, O9V)
Grigsby et al. (1992) derived 
=30.0kK using a line-blanketed
model atmosphere, in sharp contrast to
=37.5kK from the unblanketed
analysis of Herrero et al. (1992). Despite these conflicting results,
HD214680 remains widely used as a standard in
--photometric calibrations following
Malagnini et al. (1986) who used
LTE Kurucz continuum fits to obtain 
=31.75kK.
Studies based on similar non-LTE
models can result in discrepant temperatures when based on optical or
UV diagnostics (e.g. Melnick 42, Heap et al. 1991, Pauldrach et al. 1994).
Indeed, within a single analysis different diagnostic
line ratios often yield considerably different temperatures -- for HD46150
(O5V(f)), Herrero et al. (1992) obtained 
=42.5kK from
HeI 
4922/HeII 
4542 compared with
=47kK from HeI 
4471/HeII 
4542.
Detacted OB binaries provide useful tests of spectroscopic analyses since their masses, radii, gravities can be independently and accurately determined, though temperature determinations generally rely on Strömgren photometric calibrations (Schönberner & Harmanec 1995; Hilditch et al. 1996) which are imprecise for O and early B stars.
Figure 3: Comparison of our revised temperature scale for OB stars
(PAC, thick solid line), available in tabular form on request,
with the standard calibrations of Böhm-Vitense (1981) and
Schmidt-Kaler (1982), plus the recent linear O star scale of
Vacca et al. (1996)
Various MK spectral type--
correlations have been attempted
for OB stars over the past couple of decades, including
Böhm-Vitense (1981), Schmidt-Kaler (1982) and Humphreys & McElroy (1984).
While these remain widely used, in ionizing
flux calibrations for instance (necessary for photoionization studies
of HII regions and starbursts), more recent
scales have been obtained by Howarth & Prinja (1989),
Theodossiou & Danezis (1991) and Vacca et al. (1996)
using more recent observational and theoretical results.
Due to the plethora of new results in this field
we present a revised 
scale of OB
dwarfs and supergiants in Fig. 3, based on results from
various direct, continuum and ionization
equilibrium techniques. Previous standard scales are also included --
note that we find considerable differences relative
to the recent Vacca et al. (1996) calibration for O dwarfs since we
incorporate
results from blanketed studies (e.g. Voels et al. 1989; Grigsby et al. 1992)
and do not impose a linear 
scale.