Statistics and inference
One aspect of Donald Foster's 1989 monograph which evidently impressed many readers was the quantity of statistics it assembled. Every possible issue was quantified, percentages or ratios were calculated for an amazingly wide range of linguistic markers, and there are twentyfive statistical tables. A typical example of this quantification process is Foster's computation that the Funerall Elegye contains 'twenty-two compound words, for a net frequency of 5.1 / 1000 words, as opposed to 4.4 in The Winter's Tale and 5.5 in The Tempest …' (1989, p. 97). That example is one of many in which Foster placed a statistic derived from Shakespeare alongside one from the Funerall Elegye, silently claiming that stylometric data established Shakespeare's authorship of that poem. Foster's statistical procedures, however, are flawed in many respects. He himself showed some awareness of the criticism often voiced, that even 'striking verbal parallels' would not amount to 'proof' of authorship:
What is wanted rather is the closest possible scrutiny of the available evidence for Shakespeare's authorship, together with all possible contrary evidence. Nor is it enough simply to demonstrate that the poem has certain Shakespearean qualities. It has to be shown that such qualities, at least in this particular combination, are found nowhere but in Shakespeare, a formidable task. (p. 80)
Those are admirable principles, but we can now see how little Foster did to realize them. In particular, he gave very short shrift to the 'possible contrary evidence' that I have set out here – all of it freely available to anyone with the curiosity and scholarship needed to explore the wider tradition. As for showing that the distinctive qualities of the Funerall Elegye are found 'nowhere but in Shakespeare', the gap between desire and fulfilment is immense.
The first deficiency in Foster's use of statistics concerns the scope of comparison. His statistics for the frequency of compound words would be meaningful only if he had produced comparable figures for a very