1 Introduction

Published values of [Fe/H] are now available for more than 1000 evolved G and K stars. If those data are to be used successfully, two basic problems must be solved: 1) the data must be accessible, and 2) their quality must be well understood. The first of these problems has been addressed by a well-known series of catalogs by Cayrel de Strobel and her associates (see, for example, Cayrel de Strobel et al. 1997). For the second problem, however, there is no solution that is generally accepted.

One part of an acceptable solution would be a reliable zero point. As Griffin & Holweger (1989) have noted, it is not safe to assume that published values of [Fe/H] are already on a uniform zero point. Another part of an acceptable solution would be a reliable set of accidental errors. Such errors must be known for any data set if the data are to be used correctly. It is worth noting that values of [Fe/H] are almost always quoted without accidental errors, and that there appears to be no consensus about how large those errors should be.

This paper describes procedures for a) establishing zero-point data, b) using those data to correct the zero points of other results, and c) determining accidental errors. These procedures are steps in the production of a catalog containing averaged values of [Fe/H] and rms errors. The stars listed in the catalog have spectral types of G and K and luminosity classes ranging from II through IV. An initial version of the catalog has been published by Taylor (1991, hereafter T91). That version includes zero-point and error analyses, but both must be revised because many more data are now being included.

In Sect. 2 below, there is a review of the way the new analysis is prepared. Section 3 is concerned with the zero point and the assembly of a final data base. In Sect. 4, accidental errors are considered. In Sect. 5, a comparison is made between the new results and counterparts in T91. Section 6 concludes the paper with a brief review.