We have reached in this paper three conclusions, namely
a) that the average precision of a MK luminosity class given in the literature
is 
L.C., which is equivalent to 
magnitudes in the
spectral range of A to F 2 stars.
b) That the probability that two observers agree to call an object
a "metallic line star" is of 60%.
c) A detailed comparison of three different series of classifications
carried out in the framework of the MK system shows that in the three cases
the precision is very similar, a fact which is in line with the
expectations.
Acknowledgements
We thank the referee, Prof. B. Hauck for comments leading to the improvement of the manuscript. One of us (C.J.) is indebted to M. Golay for many useful discussions concerning Am stars.