James Hogg's Confessions of a Justified Sinner is one of the more remarkable studies of the uncertainties of evidence, and one of its more remarkable moments is the 'editor's' account of the court-room scene, where Bessy Gillies refuses to admit to the possibility of certainty about a matter of fact. Under cross-examination, she refuses to acknowledge that a gown presented in evidence is one that has been stolen by Bell Calvert from Bessy's mistress Mrs Logan:
'Did you ever see this gown before, think you?'
'I hae seen ane very like it.'
'Could you not swear that gown was your mistress's once?'
'No, unless I saw her hae't on [. . .]'
'But you say that gown is very like one your mistress used to wear.'
'I never said sic a thing [. . .] it is very like ane I hae seen Mrs. Butler in the Grass Market wearing too; I rather think it is the same. Bless you, sir, I wadna swear to my ain fore finger, if it had been as lang out o' my sight'.
There is more than one way to interpret this teasing passage in this prismatic text. Bessy's scepticism may be viewed as the ultimate honesty. It may also be viewed as a mode of prevarication. The Confessions are a fiction at best, and there are uncertainties in the narrative of the editor (not least because of his trust in 'tradition' rather than in written testimonies). None the less, it seems that Bessy's scepticism is not, or not only, a principled position, but a device to deny the probable: Bessy serves Mrs Logan in her anxiety to save Bell Calvert from the gallows, in the hope that Calvert may reveal to them how young George Colwan died. Bessy confronts the cross-examining lawyer by denying the possibility of truth. The master of evidence is nonplussed by Bessy's effective rejection of the possibility that any evidence could ever be of value; that anything could ever be known with certainty. This is pleasingly subversive, as Bessy invents a new and female character for the boy bishop in this feast of narrative misrule, but it is also an act both of bad and of little faith. Bessy's ruse is the stock in trade of the sceptic: to assert (what is clearly true) that knowledge can never be certain, and thence to conclude (what is much less clearly true) that no knowledge can therefore be had. In this Bessy is a prototype of a central method of modern editorial scepticism, and has much of its charm.
While happily agreeing with Bessy Gillies and her followers in the rejection of certainty, I shall attempt to argue here for probabilistic knowledge in a discrete but highly significant area of textual work, the exercise of choice amongst, and the explication of, local verbal meanings. I shall here be concerned with the interpretative rather than the bibliographical aspect of editorial analysis. I shall draw on my own experience as an annotator and explicator of eighteenth-century poetry, and I shall argue for a position similar to that adopted by Josephine M. Guy and Ian Small in their recent study of Politics and Value in English Studies: that is, that textual editing is at least in part an interpretative procedure, and that interpretative determinacy is a necessary condition for textual editing and explanation as it has been practised by literary scholars since the eighteenth century. My argument will make against, implicitly and at some points explicitly, a powerful sceptical position which has been persuasively and variously argued by Randall McLeod, D. C. Greetham, and Leah S. Marcus amongst others: that texts and textual meanings are indeterminate, and that the familiar procedures of interpretative textual editing, which normally produce a single eclectic text in printed form, and commonly attempt to elucidate that text, had better be replaced by a project of 'unediting', whose preferred forms are multiple, photographic, or electronic, and un-'policed' by annotation. In the course of this essay I shall briefly rehearse some of the no doubt familiar arguments which bear on the question of whether textual editing and commentary might or might not be considered 'scientific' in its methods; explore some of the principles upon which knowledge in textual editing might be validated; provide some practical instances of such validation; and suggest some of the implications for interpretation if we are prepared to accept not small scale certainty, but the reasonable assurance of small-scale knowledge.
The relation between the physical sciences and the discipline of bibliography has been widely discussed in a number of well-known essays. The case for a relation between the physical sciences and the procedures of textual criticism and interpretation has been rather less considered. Karl Popper has argued for an essential similarity in the methods of the sciences and the humanities, and his statement of that similarity bears particularly explicitly upon textual editing:
Labouring the differences between science and the humanities has long been a fashion, and has become a bore. The method of problem solving, the method of conjecture and refutation, is practised by both. It is practised in reconstructing a damaged text as well as in constructing a theory of radioactivity.
The analogy between the methods of science and the humanities has been resisted on a number of grounds: that, whereas the physical sciences are characteristically predictive, aiming to generate reliable rules of wide application on the basis of empirical observation, the humanities are almost invariably concerned with the individual case; that the subjects of the humanities are human creations and activities rather than natural phenomena; that the humanities are inevitably concerned with the historical, which is uncertainly knowable; and that, despite Popper's assertion, the hypothetical method of problem solving in the humanities can never be as decisively based on a process of falsification as in the physical sciences.
For many of the sciences, for much of the time, it is possible to test understanding by the construction of experiments which will serve to falsify certain hypotheses and leave one or more other hypotheses in place. Science proceeds, like Sherlock Holmes, by a process of elimination amongst alternative possibilities. The possibility of falsifying experiment has seemed much less available in the humanities. A doctor, as A. E. Housman pointed out, may devise an experiment to test the efficacy of a drug, but 'our conclusions regarding the truth or falsehood of a MS. reading can never be confirmed or corrected by an equally decisive test; for the only equally decisive test would be the production of the author's autograph'. More recently, with reference to interpretation, and with a more philosophic attention to the doctrine of inductive asymmetry, E. D. Hirsch has made essentially the same point as Housman:
Since we can never prove a theory to be true simply by accumulating favorable evidence, the only certain method of choosing between two hypotheses is to prove that one of them is false. In the predictive sciences this can be accomplished by devising an experiment [. . .]. In the historical sciences such a result can seldom be achieved because decisive, falsifying data cannot be generated at will, and if such data had already been known, the two hypotheses would not have been in serious competition. Sometimes, of course, decisive data does by good fortune turn up, but usually neither competing hypothesis can be falsified, and both continue after their separate fashions to account for the evidence.
Because 'the direct path of falsification is closed', Hirsch concludes, 'we have to make our way through a thicket of probability judgments on the basis of the evidence that we have'.
'Probability' as a concept has a long history, of course, in many fields, including law, mathematics, and the physical sciences. Its value has been much questioned by many philosophers of science, and rejected by some. Certain kinds of problem in bibliography might be addressed by a strictly scientific probability theory. Laplace's 'law of succession', for instance, addresses the question, 'If n instances of a all agree in being b, what is the probability that the next instance of a is b?'. Laplace's answer is given in the formula n+1/n+2, which could be helpful, for instance, in indicating how many individual copies of a given impression of a hand-printed book might have to be examined to reach a reasonable level of probability that all variants had been found. Probability judgements in science and in life, however, are commonly non-numerical, and textual criticism and interpretation, it is possible to argue, follow a more general process by which degrees of qualitative rather than quantitative probability might be established.
This text-critical or interpretative process may be 'scientific', but, as Housman remarked in relation to the classical literatures, it is not an exact science, and can scarcely be reduced to a 'cut and dried method'. None the less, some of the broad outlines of the process …