The Collected Ghost Stories of M. R. James (CGS, 1931) has formed the basis of nearly every subsequent edition or selection. James wrote to Gwendolen McBryde on 4 February 1931 that 'Proofs of the Collected Ghost Stories come in almost daily—we are at p. 448 and I think it should run to nearly 600, which seems a lot.' To the extent that James corrected and presumably approved the proofs we might suppose CGS to represent his final textual intentions. But James was not a particularly good proof-reader; as he went on to tell Gwendolen: 'It's very hard to nail the mistakes: if the words are real words, everything looks all right.' More importantly, he never looked upon the texts of his stories with a possessive or a fastidious eye, and a study of the available manuscripts suggests that he submitted happily to an imposed house style, especially in the matter of punctua tion and paragraphing.
There is thus a case to be argued against duplicating CGS uncritically where manuscripts are available for comparison. In the end, this is largely a matter of removing redundant accidentals (commas, semicolons, and so on) and reverting to the freer use of punctuation apparent in the manuscript texts. Though such changes may appear trivial in isolation their cumulative effect can be significant for the narrative energy of a particular story. This is the first edition of James's stories to draw on manuscript readings in this way, though I have refrained from reproducing an exact transcript of existing manuscripts, which would have been a rather pointless exercise in pedantry.
Substantive discrepancies between CGS and manuscript are another matter and I have assumed that, on the whole, these were instigated, or at least approved, by James at proof stage. However, in one or two cases, indicated in the Explanatory Notes, manuscript readings seem worth reinstating: for instance, the wholly M. R. Jamesian aside about the quality of Vin de Limoux in 'Canon Alberic's Scrap-book' (see p. 9).