The availability of a large quantity of new and precise trigonometric parallaxes from the HIPPARCOS satellite raises again the need for spectral classifications. This is due essentially to ®two different reasons.
In the first place because most of the program stars for the HIPPARCOS programs were selected on the basis of spectroscopic criteria (Am, Ap, delta Del stars, giants, subgiants, etc). Poor classifications produce biased samples, and biased samples produce biased average absolute magnitudes for a group.
In the second place, when calibrating the HR diagram, one would like to know the spectroscopic luminosity class of the objects being used in the sample, to have an external check upon the stars one includes.
The present paper is written with these two purposes in mind. In the first part we shall examine how well luminosity classes derived by different classifiers agree among them, and what can be said about the general reliability of a spectroscopic luminosity class. Since our interest lies with early type stars, we shall examine this problem with samples of A-F type stars.
In the second part we shall examine one spectroscopically selected group of objects - the Am stars - to find out how well they are singled out by different classifiers, and how the classifiers agree among themselves.