University of Virginia Library

I

In recent years we have all come to recognize the need for what might be called 'scientific' investigation in bibliography, a phrase which at its best implies, as Professor Bowers has succinctly put it, a strict regard for certain fixed bounds of physical fact and logical probability.[1] The achievements resulting from such a concern are clear and important. In descriptive bibliography we have gained a new, accurate and rational vocabulary, and formulae that are both economic and unambiguous. In analytical bibliography — with which I am principally concerned — we have been taught to use the critical tools of comparison and analysis in a new way; and the importance of establishing press variants by collation, of detecting setting by formes, of distinguishing between compositors by spellings or impressions by press-figures is no longer questioned. Scientific bibliography, complete with its laboratory aids, has become a new orthodoxy.

Yet, as T. S. Eliot puts it, All our knowledge brings us nearer to our ignorance. In the very act of opening our eyes to new and exciting

possibilities, our discoveries raise fresh problems of understanding and breed an awareness of our limitations. In particular, the detailed relation of analytical bibliography to the editing of seventeenth- and eighteenth-century texts depends upon a number of assumptions about printing-house conditions that are only now beginning to be tested. Indeed, if I were to give this paper an epigraph, it might well be that quoted by Sir Karl Popper from Black's Lectures on the Elements of Chemistry published in 1803: "A nice adaptation of conditions will make almost any hypothesis agree with the phenomena. This will please the imagination, but does not advance our knowledge."[2] Our ignorance about printing-house conditions in the 17th and 18th centuries has left us disastrously free to devise them according to need; and we have at times compounded our errors by giving a spurious air of 'scientific' definitiveness to our conclusions.

There has of course been nothing morally reprehensible in this, for 'scientific' in such a use has meant little more than an honesty of method in respect to the physical phenomena available for study.[3] As in the physical and natural sciences, at least in their more elementary stages of observation and classification, it has, as I have indicated, simply meant placing the stress in the first instance on a finite number of particulars and drawing a conclusion from them. And since bibliographers very rarely have anything to work on but the physical evidence of the books themselves it has seemed only natural that any methodology should work within these limitations and seek its exactitude by describing and relating only the observable facts of the paper and the marks it bears.

Some distinguished bibliographers, it is true, have had their doubts. R. C. Bald remarked that "whatever bibliography may or may not be, it is not an exact science, if one understands by an 'exact' science a branch of study which arrives at its conclusions through experiment and observation and can reproduce the conditions of an experiment so that the results can be repeated and checked at any time ("Evidence and Inference," pp. 2-3). And R. B. McKerrow, late in life and writing

more particularly of textual criticism, made much the same point and added the comment that "Nothing can be gained, and much may be lost, by a pretence of deriving results of scientific accuracy from data which are admittedly uncertain and incomplete" (Prolegomena, p. vii). It is this last point which I wish to take up.

The effect of Bald's suggestion that bibliography "cannot claim for its conclusions the same universal validity as belongs to those of the exact sciences" is simply that we should rest content with a very different order of certainty but take the precaution of scrutinizing more frequently than we do the procedures of bibliographical scholarship. McKerrow's comment, however, runs far beyond his intention and offers a radical criticism of the very bases of all knowledge inductively derived. It is not that bibliographical inquiry differs in any essential respect from 'scientific' inquiry as described, but that the method common to both is itself logically unsound. Bibliography, as it happens, is a convenient area in which to demonstrate its unsoundness.

For whatever the short-term advantages, to assume, as we have been asked to do, that analytical bibliography must be empirically based, and to limit our knowledge to that which may be derived by inductive inference from direct observations, is to invite the obvious objection that no finite number of observations can ever justify a generalization. Bertrand Russell remarked that so far as he could see induction was a mere method of making plausible guesses. "It is far from obvious, from a logical point of view," writes Sir Karl Popper, "that we are justified in inferring universal statements from singular ones, no matter how numerous; for any conclusion drawn in this way may always turn out to be false: no matter how many instances of white swans we may have observed, this does not justify the conclusion that all swans are white" (Scientific Discovery, p. 27). David Hume had made essentially the same point: "Even after the observation of the frequent conjunction of objects, we have no reason to draw any inference of any object beyond those of which we have had experience."[4] And more graphically the Lilliputians' ignorance is ironically exposed: "Besides, our histories of six thousand moons make no mention of any other regions than the two great empires of Lilliput and Blefescu."

Nor is it simply that there is no logical way of arriving at general truths from the examination of sampled cases. To observe at all is to bestow meaning of some kind on the thing observed; to gather particular pieces of evidence is to seek those relevant to some preconceived

notion of their utility.[5] But the main point is that any general laws derived by the inductive method remain highly vulnerable to fresh evidence. Where the known particulars are few, this risk will be greatest.

The inductivist in all subjects may of course freely admit that incompleteness is almost invariably a characteristic of his evidence and that his conclusions may therefore be subject to modification when new evidence comes to hand. Looked at in this way the inductive process carries a burden of assumed truth waiting to be converted into proven error: knowledge, that is, comes with the act of disproof. And until that moment arrives we may be offered conclusions that lay claim only to some degree of 'reliability' or 'probability,' based on reasonable assumptions about the comprehensiveness of the evidence used and about the predictability of past or 'normal' examples into the future.[6]

Even such a qualified attitude, however, is still inadequate to the demands of bibliography. For one thing, I doubt that 'normality,' in any serious and extended sense, is a meaningful concept.[7] For another,

even granting the inductivists' case, I doubt whether bibliography has yet either the body of documentary evidence or the fund of experimental proof necessary to bring most of its conclusions within a usefully narrow range of probabilities. Under present conditions, therefore, we are tempted either to over-stress the few 'proofs' achieved (by restricting further inquiry to a demonstration of their 'normality' and so asking that new conclusions be consistent with it), or conversely (in the greater number of cases where we have little developed knowledge of the causative conditions) to tolerate any number of 'probable' explanations for the same limited range of phenomena. Neither alternative is a particularly happy one. The first protects us from open-ended speculation by a crippling limitation of the subject; the second, although it characterizes the subject as practised, is rarely capable of conclusive demonstrations.[8]

But there is another way of looking at the whole problem; nor does it involve any question-begging distinctions between bibliography and 'science,' for it applies equally to both. It simply consists in recognizing the present situation of multiple 'probabilities' as the desirable one and regarding them as hypotheses to be tested deductively. I am naturally tempted to it because the productive conditions in early printing houses display an incredible variety which, if it is to be re-conceived at all, demands an imaginative facility in devising hypotheses;[9] and also because bibliographers, as a matter of fact, are becoming increasingly concerned to trace processes involving complex relationships less susceptible of conclusive demonstration. More seriously, deductive reasoning (by which a general hypothesis dictates particular possibilities or 'predictions' and rules out others) does offer a sound way to knowledge. Like induction, it is open to logical objection, for no amount of positive evidence can ever conclusively confirm an hypothesis;

but one piece of negative evidence, one contradictory occurrence, will conclusively falsify it.

These comments are philosophic commonplaces and stated so baldly must seem slightly naive. Yet they do serve to point quite sharply to the two main directions open to bibliographical inquiry. If bibliographers wish to persist as inductivists then they must diligently search out the historical facts which will alone provide a fairly accurate definition of 'normality' and offer these as a corrective to the logical defects inherent in the method. Alternatively they may confess outright the partial and theoretic nature of bibliographical knowledge, proceed deductively, and at the same time practise a new and rigorous scepticism.

In fact the nature of 'normality' as so far revealed by historical evidence suggests that the 'norm' comprised conditions of such an irrecoverable complexity that we must in any case adopt the latter course. If the 'scientific' proofs offered in some recent bibliographical analyses of older books were seen philosophically for the conjectures they are, we should I think be nearer the true spirit of scientific inquiry and the humility that always accompanies an awareness of the possibility of fresh evidence and therefore of falsification. The subject would not then be circumscribed by the demand for demonstrable proofs; rather it would be expanded in its hospitality to new ideas and in its search for fresh historical evidence in the service of disproof. Such a method would be, in the best sense, scientific.

In the following section I wish to offer some varieties of fresh evidence. Its main implication is that the very fixity of the physical bounds within which we are asked to work is inimical to the development of a sound methodology — first, because if the stress is laid on 'proof' then the small number of paradigms available to us unreasonably restricts the subject; second, because in the present state of our knowledge the finite particulars with which we must work are too few and therefore permit too many alternative generalizations to be induced from them; third, because the conception of 'normality' as a corrective to the undisciplined proliferation of generalizations misrepresents the nature of the printing process; fourth, because induction is necessarily an inconclusive method of inquiry. The evidence is consistent with my belief that we should normally proceed in our inquiries by the hypothetico-deductive method which welcomes conjectures in the positive knowledge that productive conditions were extraordinarily complex and unpredictable, but which also insists that such conjectures be scrutinized with the greatest rigour and, if refuted, rejected.