This paper compares two different approaches for indexing archived text documents. The data used for testing the different systems are said to be drawn from commercial and technical material. The statistical parameters needed for text access are defined in terms of number of (non) (common and/or distinct) words (i.e., number of words; number of non-common words; number of common words; number of non-common distinct words; and number of common distinct words), ratios, etc. STAIRS and two closely related methods for text indexing based on signature files are briefly described. These were designed by Christodoulakis and Faloutsos [1], and Tchichritzis and Christodoulakis [2]. Performances were estimated using analytical models; several environment parameters were tested (mainframe vs. microcomputer, different processing loads, buffer sizes, etc). The conclusion of the tests is that “the ‘best’ system for retrieving text documents in an office environment seems to be the SC system, i.e., a system based on signatures generated by superimposed coding.” A few graphs are included. The main interest of this paper lies in its precise description and clarity.