In a recent article in CJL/RCL 13, 1, H. M. Logan discussed some of the problems involved in as well as the kind of results obtainable by the use ot computers in dialect studies of Middle English. While many of Logan’s points were interesting and suggested ways in which some of the statistical burdens could be raised from human shoulders, in my opinion much ot what he says tends both to oversimplify the problem of computer usefulness in this field and to underestimate the achievements of precomputer scholarship. Although it is the narrow scope of the linguistic criteria utilized in such projects as the MEDC on which Logan focuses his main objections, he himself, it would seem, fails to expand them in any significant way or to suggest, for example, that they may be more profitably derived from one descriptive model rather than another. For him “it is precisely [the] central problem of exhaustiveness which the computer solves.” But exhaustiveness in what sense? The number of occurrences of and in a text in a full or abbreviated form is surely in itself of peripheral importance (as Logan admits); yet it is the very tendency to stress the value of data collection supported by a minimum of relational classification which is prominent in his arguments. While one must admit the usefulness of the computer as a concordance builder, its ability to effect “a classification of all the grammatical forms and the significant phonological features” is only as good as the descriptive adequacy of the theory which can account for such forms and features.