University of Virginia Library

5. Some Problems and Suggestions

Defoe represents a unique challenge. A tiny number of personal letters survive, and those are late-life and probably not typical; the writings we can attribute with confidence tell us much about his causes and commitments but not a lot about him as a man. Other major eighteenth-century writers published anonymously, but he is unusual in having put his name on so few works. From the end of the eighteenth century to the third quarter of the twentieth, a gigantically inflated canon grew up around him, most of it based on irresponsible attributions. At this late date we now need to confront two separate but related problems. The first is that many of our deeply ingrained ideas about Defoe have their basis in Moore’s canon, which has been (for most of us) effectively demolished. The second is that the Furbank and Owens canon is nowhere near as solid as we would like to believe. If we are going to tackle Defoe with confidence on a factual basis, then we need to reconsider just about everything.

The Critical Bibliography represents a huge advance on Moore’s Checklist, and we owe Furbank and Owens a profound debt of gratitude—but (as they themselves would say) we need to use the new canon responsibly if we are to avoid creating a further mess. The blunt truth is that we cannot afford to pretend that Furbank and Owens have settled the canon for us—which they never claimed to have done. In Starr’s favorable review of the Bibliography, he rightly praises Furbank and Owens for not writing with "an air of finality, of having fixed the canon once and for all" (585). Works entered with no cautionary "P" are not necessarily proven attributions, and those with a "P" are anything but close-to- certain. Furbank and Owens understand this—but any item-by-item survey at book length inevitably creates a close focus on individual items. For details of particular cases, it is a tremendous resource; for comprehension of the basis of the canon as a whole, it necessarily has limitations. The table constituting the appendix to this essay was conceived as a kind of second step, an enhanced view of the same material for different ends.


Page 147

The bottom line is that we cannot assume that the Critical Bibliography represents a firm foundation. In some ways, it seems to be more conservative than it actually is. Most of us feel quite certain that Defoe wrote a lot of the items not labeled "claimed by Defoe" in the table, and we are possibly right. But our sense that he wrote some and even many of these works does not justify our behaving as though he wrote all of them or any particular one of them. Instinct and plausibility and even some degree of probability we have—but that is not grounds for attribution. I wholly concur with Starr’s judgment: "My own view is that there are really rather few works that we can say were Defoe’s with anything approaching ‘certainty,’ using that term in the everyday rather than the Humean sense, and that the sum should be smaller than Furbank and Owens’s figure of 160" (587–588). By my reckoning, at present our solid canon consists of 67 works (not counting the thirteen items reprinted from the Review).

Might quantitative stylistic analysis allow us to expand that canon with confidence? The answer unfortunately seems to be No. Quantitative methods of attribution have not worked very well to date, and there have been some problems—as for example Martin C. Battestin’s New Essays by Henry Fielding (1989), which claimed Fielding’s authorship of forty-one Craftsman essays almost entirely on the basis of stylistic parallels.30 Reviews of the volume were few and largely uncritical, though some critics did complain about illegitimate stylistic parallels and generally inadequate evidence.31 The Craftsman attributions were not disproven until 2008,32 by which point they had done serious damage to the field; they will no doubt continue to do so for many years to come. Battestin has since transferred his allegiance to the Cusum method, a different kind of stylometric analysis,33 whose legitimacy has also been called into question. In a devastating rejection of quantitatively-based theories of attribution, and of Cusum (or QSUM) analysis in particular, Stephen Karian offers this blunt verdict:

QSUM uses vague definitions for its terms, misuses a valid statistical technique, and relies on visual inspection without employing a standardized method for calculating scale. Finally, it does not, in any sense of the word, ‘work.’ QSUM has no validity. ... This method is so faulty that one can manipulate it to claim any position on any particular attribution.34

Karian’s sober analysis and critique bears out his conclusion that the Cusum method "cannot be rescued."


Page 148

Supposing someone actually did develop a more reliable method of stylistic analysis, how effective would it be if applied to Defoe?35 Defoe turns out to be highly problematic as a subject for quantitative analysis. The lack of a sufficient number of works irrefutably attributed to him makes stylistic comparison dangerous at best. We just do not have enough to constitute a satisfactory sample.36 A writer like Fielding, whose output is mostly signed, is a different proposition. The majority of the works we can feel reasonably confident that Defoe wrote, moreover, date from quite early in his career. Would stylistic markers from The True-Born Englishman or The Shortest-Way or even the Review help us with Roxana? This seems highly unlikely: such works differ generically as well as modally, and never mind the broad chronological spread of composition involved.37 Including a work in the canon on the basis of internal parallels to other supposed "Defoe" pieces is unwise: as Furbank and Owens warn, we cannot count a work as Defoe’s because it sounds like other things he might have written. A related danger is that of over-emphasizing stylistic similarities and not considering significant stylistic discrepancies, a point Rodney M. Baine made in his de-attribution of Madagascar: or, Robert Drury’s Journal38

What about paragraph lengths and common phrases? Maximillian E. Novak in particular weights these very heavily, maintaining that "wide paragraph length


Page 149
variations" are uniquely characteristic of Defoe, as are "[p]aragraph beginnings with generalized phrases such as "Tis’."39 So if we see a pamphlet whose paragraphs often begin with "It is said," "It is," "I know," and "As to the," should we assume that is Defoe revealing himself to us? The utility of such a test seems far from clear. "Defoeisms" such as "I say" and "viz." and "In a word" seem equally unhelpful: such generic language patterns seem unlikely to provide the answer to our problems. Stylistic analysis may some day be more useful in particular cases than it has hitherto proved, but the uncertainties surrounding Defoe and his canon make application of such analysis to him especially difficult. If future work in this realm can convincingly settle the authorship of individual items, then that would of course be a major improvement, but at present there seems little possibility of using style to substantiate uncertain attributions.

What, then, are we to do? The crux of the answer is that we cannot afford to ignore the uncertainties that are probably an irresoluble feature of the Defoe bibliography. We need to learn to live with a much-reduced canon. In the post- Furbank-and-Owens era, more scholars will naturally want to re-attribute questionable titles than de-attribute yet more of them, but the temptation to rescue favorite items is best resisted. Erring too far in the direction of conservative attribution is at worst a shame; erring too far in the opposite direction is a disaster. We therefore need to come to each item in the Critical Bibliography—no matter how often taught or how heavily written upon—without commitment to its retention. In time we can return to the De-Attributions and reconsider them. But we need first to make sure we are using the Bibliography in ways that are intellectually honest and methodologically defensible—in ways worthy of the careful work it represents. Ideally we would find evidence either to bolster current attributions or to disprove them, and (in principle) we ought to be equally glad of either.40

At this point, we can decide simply to ignore the problems and carry on, building anew atop woefully shaky foundations. Alternatively, we can decide not to base interpretations or biographical constructions on the Furbank and Owens canon without considerable caution and admission of its limitations. What does "caution" look like in practical terms? The 67 works unambiguously claimed by Defoe are safe. Beyond that, how a particular text gets treated should depend not on whether Furbank and Owens include it but on what sort of evidence we have for it—and even where the nature of the evidence is similar, the probability can vary drastically.

This is an obvious conclusion, but it needs to be made explicit and acted upon, especially since even Furbank and Owens at times seem too casual about


Page 150
their own distinction. Take the example of Atalantis Major (122(P)). In a 26 December 1710 letter to Harley, Defoe explains that he had come into possession of "Two Vile Ill Natur’d Pamphlets," one of which (he says) is Atalantis Major. "It is Certainly Written by Some English man, and I have Some Guess at the Man, but dare not be positive," he tells Harley, but also that he will insure that the pamphlet is suppressed in Edinburgh. Furbank and Owens conclude that, Defoe’s comments notwithstanding, "It is ... generally agreed that this ‘Vile’ tract was Defoe’s own work: above all because it corresponds so closely to what he had written earlier to Harley."41 Whether Defoe is being partially or wholly disingenuous we will never know, which is presumably why Furbank and Owens regard Atalantis Major as merely probable. When they discuss the pamphlet in their Political Biography of Daniel Defoe (2006), however, they say simply that "the author, undoubtedly, was Defoe himself."42 If there is in fact no doubt, why brand the item "P" in the Bibliography? Conceivably their position changed between 1998 and 2006, but if so the reason for the shift should have been explained. The discussion of The Commentator, one of the "P" journals, likewise reflects misleading certitude about the attribution. Furbank and Owens explain that Defoe "launched a new periodical," that he "began to lose interest in his Commentator," and so on, with no indication of dubiety. Making distinctions between probable and certain is of little use if we are not going to apply those distinctions to our critical treatments of Defoe.

One of the implications of the canon problem is that our comprehension of "Defoe" is based principally on the body of works attributed to him: who he was depends upon what he wrote.43 And if a significant percentage of the items included in the Furbank and Owens canon are uncertain, then so too are our characterizations, both critical and biographical. The two most substantial modern biographers of Defoe, Paula R. Backscheider (1989) and Novak (2001), both decided to stick with Moore’s canon, with the exception of a small number of works.44 Backscheider was finishing her Daniel Defoe: His Life just as Furbank and Owens were beginning to publish their misgivings about the canon. Novak had more opportunity to revise his account, though by the time of Furbank and Owens’s advent on the Defoe scene, he too had been working for many years on his Daniel Defoe, Master of Fictions. In his introduction, he objects to Furbank and Owens’s dismissal of stylistic evidence, explaining that he is "draw[ing] upon [his] years of reading [Defoe’s] texts and those of his contemporaries." Novak’s commitment to his own sense of Defoe’s corpus (which is not identical to Moore’s


Page 151
canon) reflects his conviction that he can recognize his subject’s "style and thought."45 From the conservative standpoint espoused by Furbank and Owens, this is problematic. In any case, we are left with two much-quoted accounts of the life that are based on a huge canon for which we have no solid proof—and in a position of having to rethink from the ground up almost everything we thought we knew about what sort of person and writer Defoe really was.

We need to be prepared to test every bit of inherited wisdom about Defoe. What kind of writer was he, really? Does the fact that he was a master fabricator, able to mimic voices not his own, strengthen the case for his authorship of Moll Flanders and Roxana? Perhaps, but first we need to ask whether he was in fact a master fabricator. The letters to Harley are ample proof of his willingness and ability to lie, but how many of the 67 works we can be sure about involve mimicry and impersonation? Item no. 20 is ostensibly by a member of the "Yeomandry, and poor Freeholders of England" (p. 3); no. 144 is in the voice of "an Englishman at the court of Hannover"; in no. 148 (like no. 144, one of the 1713 succession pamphlets), Defoe assumes the guise of a Jacobite pressing for the return of the Pretender; and no. 37 is The Shortest-Way, his most brilliant piece of satiric impersonation. No. 52 is "in the person of a Dissenter" (CB, 48), so not Defoe-speaking-as-Defoe but hardly a feat of mimicry. In these five works, Defoe assumes a voice not his own, or at least not only his. The Shortest-Way alone proves his capacity as a mimic, but his inclination toward mimicry is not necessarily on the scale critics have long imagined. If we add Robinson Crusoe, we have another example of a persona not identical to Defoe. If Defoe did write (say) Moll Flanders, then we have a dissenting Protestant male assuming the persona of a female pickpocket cum prostitute. This does not seem to be impersonation of the sort we find either in The Shortest-Way or Crusoe. Could Defoe have done a different kind of mimicry? Yes. Does the rest of his output lend credence to the argument that he actually did so? No. What is crucial here is that we need to look again, and skeptically, at every established verity. Some of them will withstand further scrutiny—but they cannot be relied upon until they are tested against a much smaller canon.

This is a frustrating state of affairs, all the more so because it could have been avoided. Half a century ago, Moore produced a bibliography which was de facto unreviewable; he did not explain the evidentiary basis for particular attributions, making judging plausibility functionally impossible. That reviewers did not assess his Checklist in nitty-gritty detail is not astonishing. Lacking the evidence, how could they have done so? What is dismaying is that the contents of an unverified and essentially unverifiable list were so widely accepted. One of the lessons we should learn from the history of the Defoe canon is that we need to be much tougherminded about how we respond to new attributions. R. S. Crane long ago warned against letting one’s "guiding principle" in interpretation be "a will to believe."46 In the case of attribution, we would do well to be driven by a will to doubt. For the


Page 152
present problem, Defoe is himself to some extent to blame. Had he left behind a signed or validated canon, we would have one. Instead, we are left with a famously protean writer about whom we have painfully little definite personal information and who evidently relished his ability to "Disguise" his "Stile."

The future of Defoe studies will depend heavily upon how we proceed from here. Some scholars have resisted abandoning Moore’s Checklist, though many Defoe scholars have cheerfully restricted themselves to Furbank and Owens’s canon. The idea of shrinking further is probably a hard sell for most scholars, especially when several of the major works are not exempt from skepticism. I do not see my conclusions as negative or gloomy. We are probably always going to be left with a significant degree of uncertainty about quite a lot of even the reduced "Defoe" canon: the problem is not lack of certitude in attribution but in refusing to admit that lack. Historical study is at best difficult, and the troubling question of how we can know the past at all is always with us. G. Thomas Tanselle has observed that "The essence of any inductive process, such as ... the pursuit of the past, is uncertainty," and he has a point.47 The record that has come down to us is incomplete—as the increasing concern with attribution and (more broadly) with the realities, frustrations, and implications of anonymous publication makes clear. A considerable part of what we think we know is the result of potentially fallible inductive reasoning, not just something read or discovered. Refusing to go beyond the bounds of what critical judgment can justify is crucial—which brings us back to the practicalities of weighing such evidence as we have and living with its limitations.

We need to remember that the "Defoe" we now study and teach is a relatively recent construction. Half a century ago, Defoe was very differently judged and valued. Let me propose a counterfactual. Imagine that we were transported back to the 1930s—before Ian Watt and others turned Defoe into the father of the English novel; before Moore inflated the canon; before a flock of New Critics set about demonstrating Defoe’s artistry and the cohesion of his novels. Suppose that in those circumstances scholars had started from scratch, applying the sober attribution methods and standards that specialist readers for leading learned journals would insist upon today. What canon would be constructed for Daniel Defoe? What picture of the life-and-works? Doggedly maintaining faith in a canon we cannot defend does not make sense: if we are going to set the bibliographic situation aright, then this is the time to do so. The question now before us is whether we will finally face facts or continue to play the ostrich.48


Battestin, New Essays by Henry Fielding: His Contributions to the Craftsman (1734 –1739) and Other Early Journalism, "With a Stylometric Analysis by Michael G. Farringdon" (Charlottesville: University Press of Virginia, 1989).


The only vehemently negative review was by Claude Rawson, "Fielding in the Dock," London Review of Books, 5 April 1990, 20–23. Milder reservations were expressed by Hugh Amory in ‘"It Is Very Probable I Am Lord Bke’: Reflections on Fielding’s Canon," Eighteenth-Century Fiction 8 (1996): 529–533; and by Simon Varey in The Scriblerian 25 (1992): 56–59.


Thomas Lockwood, "Did Fielding Write for The Craftsman?" The Review of English Studies, n.s., 59 (2008): 86–117.


For Battestin’s defense of the Cusum (cumulative sum) technique, see "The Cusum Method: Escaping the Bog of Subjectivism," Eighteenth-Century Fiction 8 (19g6): 533–538.


Karian, "Authors of the Mind: Some Notes on the QSUM Attribution Theory," Studies in Bibliography 57 (2005–2006): 263–286, at 281. The following quotation is at p. 285.


Furbank and Owens raise the issue of stylometry, and even produced computergenerated concordances of Moll and Robinson Crusoe. They point out that while stylometry might conceivably be useful if one is trying to decide between two candidates for authorship, where the field is virtually limitless much of the utility vanishes: "it was illusory to suppose, as some stylometrists claim, that an author has a unique statistical ‘fingerprint,’ distinguishing him or her from all others." They also warn that the "temptation to get the answer you want ... is simply too compelling." See "The Defoe That Never Was: A Tale of De-Attribution," The American Scholar 66 (1997): 276–284, at 283.


In his 1974 stylometric study of parts of Defoe’s canon, Stieg Hargevik identified words, phrases, syntactical structures, and sentence patterns that he found to be indicative of Defoe’s style, measuring their appearance in Defoe’s writings against their appearance in the work of his contemporaries. He demonstrated that the stylistic tendencies he regarded as Defovian appeared only very rarely in select texts by Abel Boyer, Anthony Collins, Benjamin Hoadly, Swift and others. See The Disputed Assignment of Memoirs of an English Officer to Daniel Defoe, in two parts (Stockholm: Almqvist and Wiskell, 1974), Part I. The problem with this methodology is that for it to be seriously effective we would need more definitive attributions and then we would need to test them against a very large sample of different kinds of writings. Hargevik himself admits that statistical methods can only "indicate likelihood and not proof" (26). For a cautious defense of Hargevik’s methods and findings, see Irving N. Rothman, "Defoe De-Attributions Scrutinized under Hargevik Criteria: Applying Stylometrics to the Canon," PBSA 94 (2000): 375–398. I concur with Furbank and Owens that, especially in the case of Defoe, stylistic analysis is insufficient to solve attributional puzzles. As they remind us, "stylo-metricians are unlikely to present their results as ‘unambiguous’ and self-sufficient, requiring no support from other and non-stylometric types of reasoning" (Canonisation, 177).


I do not share Moore’s faith that "the Defoe who wrote the boyish religious verse of 1681 is astonishingly like the elderly man who published social and economic tracts forty-nine years later. We can trace him through his Protean changes of appearance. If we hold him tightly, he remains essentially himself" ("The Canon of Defoe’s Writings," 164).


Baine, "Daniel Defoe and Robert Drury’s Journal," Texas Studies in Literature and Language 16 (1974): 479–491, especially pp. 484–486. "The wider the Defoe canon becomes, the easier it is to establish other attributions of the basis of style and idiom. We have a widening gyre whose center will not hold" (484).


Novak, "A Vindication of the Press and the Defoe Canon," Studies in English Literature 27 (1987): 399–411, at 406–407. Stephen Bernard has demonstrated conclusively, on the basis of internal and external evidence, that A Vindication of the Press was written not by Defoe but by Giles Jacob. See "After Defoe, Before the Dunciad: Giles Jacob and A Vindication of the Press," The Review of English Studies, n.s., 59 (2008): 487–507.


Schoenbaum’s counsel is wise: "It is good, I believe, that now and then we pay tribute to the virtue of recognizing our limitations," he concludes, and the scholar "must have his own kind of negative capability. He must know and accept the often frustrating limitations of the methods available to him if, in his quest to dispel illusions and errors, he is not to create new ones in their place" (Internal Evidence, 219).


CB, 111. For Defoe’s quotations, see Letters, 306–307.


A Political Biography of Daniel Defoe (London: Pickering & Chatto, 2006), 113, 177, 182.


For a discussion of the various ways in which "Defoe" has been constructed by biographers and critics, see my "Fabricating Defoes: From Anonymous Hack to Master of Fictions," Eighteenth-Century Life 36.2 (Spring 2012): i–35.


John Richetti’s more recent biography of Defoe is based on the Furbank and Owens canon (with only a few exceptions). Richetti makes a good point: "even Furbank and Owens cannot resolve the uncertainty surrounding some of what we think is Defoe’s massive output, and in their Critical Bibliography ... they list works that are ‘probably’ by Defoe, and in their ongoing collected edition of large numbers of Defoe’s works they ... reprint and annotate some of those pieces that they mark as merely probable in their Critical Bibliography." The Life of Daniel Defoe (Malden, MA, and Oxford: Blackwell, 2005), vii.


Novak, Daniel Defoe, Master of Fictions: His Life and Ideas (Oxford: Oxford University Press, 2001), 5.


Crane, "On Hypotheses in 'Historical Criticism’: Apropos of Certain Contemporary Medievalists," in The Idea of the Humanities and Other Essays Critical and Historical, 2 vols. (Chicago: University of Chicago Press, 1967), 2:241.


Tanselle, "Printing History and Other History," Studies in Bibliography 48 (1995): 269–289, at 282.


An oral version of this essay was presented at the second biennial meeting of the Defoe Society, July 2011. For advice on drafts, I am grateful to Julian Fung, Patricia Gael, Robert D. Hume, Leah Orr, and David Wallace Spielman. The anonymous readers for Studies in Bibliography provided extensive and extremely useful commentary for which I am grateful. For technological assistance I am much obliged to Tom Minsker. Obviously I am deeply indebted to Furbank and Owens, who meticulously assembled the complex mass of evidence I have recast and analyzed here. This enterprise depends on their pioneering work. I did not consult them about this piece, and any blame attaching to it must be mine. If Defoe studies has a good future it is largely because of Furbank and Owens’s efforts.