Information, as published knowledge, is vulnerable to error at any point in the extended “communications circuit” that connects authors and readers (as Robert Darnton describes it, in the case of print), by virtue of having undergone the error-prone processes of composition, editing, proofreading, printing, binding, selling, and so on. More generally, it is liable to error throughout the even more diverse and open-ended circuits by which it is acquired, analyzed, disseminated in any media, and made use of (to use a four-stage sequence conveniently identified by Peter Burke in the entry “knowledge” in this volume). Some errors along those circuits may be easy to identify as errors, but others raise difficult questions about how to understand past and foreign contexts. The former case includes errors that can be identified through their clear violation of a particular correctness condition in a past context; the latter may involve errors where the very criteria for correctness were disputed in the past or are unclear to us now. To distinguish between errors in the local performance of knowledge and errors in criteria for knowledge requires careful attention and sometimes subtlety, as does identifying them as errors in the first place, since we are at risk of anachronistically imposing onto other contexts our own standards for correctness and our own disputes about criteria for error. Normally we can understand error through the range of practices that have aimed to prevent, correct, or eliminate it; through the diversity of vices and dispositions believed to lead to it; and through its envisaged consequences. Although error criteria often remain implicit in these practices, and can be inferred only with caution, at times they are explicitly identified as such, for example in the literary *genres that exaggerate errors in order to expose them. Satire and comedy, caricature and character types, invective and polemic, and parody constitute rich sources for the explicit representation of error, not of course because they are meant to be plainly believed, but because they offer evidence about what was believable about error and makers of error.
Error may be located in a variety of places in information circuits. It may be discovered, for instance, not in the knowledge itself but in the producer of the knowledge, in some deficiency of character or other personal disqualification. Biography, in some periods, is invoked as a test for the validity of certain kinds of knowledge, the defects of a life implying a failure of that person’s knowledge, or at least grounds for suspecting it of error. In courtroom speeches and in political deliberations, for instance, moral and *epistemological error are commonly presented as one. Since the nineteenth century, biography has generally been held irrelevant to the truth of philosophical or scientific statements, but in antiquity the actual course of a philosophical life was a prime arena for testing philosophical knowledge. Errors in other domains of knowledge have also been traced to errors of the knowing subject rather than to the nature of the knowledge itself. The history of the biographies of knowing subjects may illuminate the connections that different ages drew between life errors and knowledge errors and clarify the degree to which legitimate personae of knowers were explicitly established.
When biography is found relevant to error, it is often understood to involve specific epistemological vices. The hardened hearts of sinners blind them to a knowledge of their sin (though some may be cured through repentance and *conversion); the curiositas of inquirers leads them in self-trivializing directions where their preference for vulgar knowledge would disqualify them from higher knowledge. Bias is one of the most commonly cited causes of error. Classical historians and orators, for instance, frequently rebut error that they trace to the bias of their rival authors and speakers, whose defects of character, they claim, are manifest in their tendentious and unreliable information; as a result, ancient authors developed a wide range of techniques and rhetoric to distinguish their reliable words from those of their rivals (e.g., by admitting conflicting evidence, insisting on their own independence and impartiality, interrogating sources). However, it is not necessary for a character to be flawed in order to be a source of error in information. For instance, with the emergence in the nineteenth century of mechanical objectivity as a method of truth, any trace of the knowing self in knowledge was regarded as a contaminant; subjectivity itself, the mere presence of a self, was taken to be a ground for error. In addition, error may be understood neither as a specific defect of a character nor as a defect intrinsic to human subjectivity, but rather as the result of the exclusion of some people from the discovery of knowledge: such a democratized knowledge procedure would root error in the inadequate participation of potential knowers in the production of knowledge. It would seek wisdom in crowds or identify coming to consensus as the means to discover valid knowledge.
The converse situation is perhaps more common, when the aspect of character that leads to an error in knowledge is taken to be the result of ignorance, a lack of training, a failure of expertise. Error, in this case, is created because those unqualified for knowledge have participated in producing it. These different ways of locating error within the character of the knower involve the widest range of prophylactics against error, and include diverse examples of spiritual training, disciplined habits of inquiry, repression of self-will, the steady management of a reputation for expertise, the use of torture to elicit truthful accounts from unreliable witnesses (such as slaves or suspects), and the extension or the reduction of the number of qualified knowers.
Alternatively, error may be located not in a flaw within the knowing subject but in knowledge itself. In these cases, the tests that establish the reliability or trustworthiness of knowledge claims are not directed to the character of the person making the claim. Instead, the knowledge has to meet other kinds of criteria: it must be consistent, either internally, or with a body of trustworthy knowledge that is prior to and outside of it; or it must be deducible from principles or axioms (less strongly, justifiable as knowledge by reference to principles); or it must be an appropriate generalization, neither excluding too many cases nor including too many exceptions; or it must be verified by some empirical observation (a prediction or postdiction that may encompass a wide range of possible circumstances); or it must be of the correct age, as both young information and old information may be challenged on the basis, respectively, of insufficient and excessive longevity. Sometimes errors are identified by means of aesthetic criteria: the structure of knowledge may be required to be beautiful, and flaws in information (published knowledge) may be due to aesthetic mistakes, such as a lack of clarity, vividness, narrative skill, or proportion. It is sometimes claimed that knowledge must have potential implications for action, in order to be either genuine knowledge or a higher knowledge.
Knowledge, in such cases, may be subject to additional criteria of wisdom, meaning, or value; to count as knowledge, it must be knowledge of something worth knowing (not, as the title of Lajos Hatvany’s early twentieth-century satire puts it, Die Wissenschaft des Nicht Wissenswerten [The science of what’s not worth knowing]). Ancient denunciations of the idleness or vanity of mere knowledge, separated from the conduct of life or the possibility of redemption, have been regularly renewed. William James (1842–1910), for instance, insisted that a valid difference in truth must correspond to a difference that we can trace elsewhere, one that is concretely imposed on somebody at some time and place. In the twentieth century, Gregory Bateson sought to define information as “any difference that makes a difference.” Knowledge, finally, must not be of the sort that is forbidden. Some true things, on penalty of divine sanction, social taboos, or legal liabilities, are to be excluded from the knowledge circuit; that is, they are not permitted to introduce a difference into it: Adam and Eve were forbidden to eat of the tree of the knowledge of good and evil, medical data from Nazi experiments or from unauthorized exhumations have been barred from use, information thought to be pornographic may be banned for that reason, and legal proceedings exclude evidence deemed irrelevant. Other matters may be forbidden on logical or epistemological grounds, as things about which knowledge cannot properly be had.
Error may also be introduced in the reception of information, for instance when it is received by the wrong kind of learner or in the wrong manner. Knowledge correct in itself may nonetheless produce error by being communicated to unfit learners; such communication, it has sometimes been claimed, not only lacks prudence but may even unfit some people, or whole populations, for what was considered to be the proper conduct of their lives, duties, or social roles. (These are cases of epistemological truths yielding moral errors, or what were taken to be moral errors; in the ordinary usage of European languages, “error” encompasses mistakes of both knowledge and action.) The fear that knowledge may be misused or misunderstood, or pose a danger to potential knowers, locates error in people who learn, rather than in the knowledge itself. This fear is often a fear of the risk posed by the unregulated distribution and consumption of information and is typically managed by restricting access to it. Censorship need not be limited to political and legal means; pricing out particular populations may be just as effective in restricting access, as likewise the use of written language to exclude the illiterate, or learned languages or technical jargons to exclude *vernacular speakers. Some groups have been disqualified from both the consumption and the production of knowledge; these disqualifications, however, have divergent histories, and different practices and rationales have been associated with them. For instance, in colonial America, teaching slaves to read (as E. Jennifer Monaghan observed) was legally permitted everywhere, while instruction in writing was sometimes forbidden; after the revolution, a number of Southern states made instruction in reading illegal, thereby barring slaves from the reception as well as the production of written information. (This was often done on the grounds that, as Frederick Douglass’s master feared, teaching a slave to read “would forever unfit him to be a slave.”)
Forbidding the wide communication of knowledge does not necessarily presuppose defective knowers; it may be a means to establish the value of arcane or secret knowledge. Such an “economy of secrets” (Daniel Jütte’s phrase) in the premodern and *early modern periods understands knowledge and publication to be at odds with each other, in contrast to more recent times when they are generally understood as mutually dependent, and with publication as the ground for the self-correction of knowledge. More commonly, however, error is located not in the presence or absence of publication itself but in some defect within potential consumers of knowledge, in the incapacity for truth on the part of some audiences: children, women, enslaved, colonized, and dependent populations, the disabled, the laity or non-coreligionists, for instance. Pedants, bluestockings, opsimaths (who came late to learning), and autodidacts may have accumulated information, but at the cost of becoming spoiled or denatured, and unfit for their assigned roles; as they are not gentlemen, their class or gender (and hence exclusion, in some periods, from formal education) has disqualified them from truly knowing as gentlemen do, that is, from integrating knowledge into their lives without error. Children in particular may be represented as uniquely vulnerable to the ill effects of adult knowledge.
Even capable learners may become flawed as a result of their knowing, if they learn in the wrong way. In the ancient world, Thales fell into a ditch as he was studying the stars: knowledge incapacitated him from everyday life, a theme and an example that, along with Socrates in cloud-cuckoo-land, long persisted in representations of the theorist. The Preacher found that “much study is a weariness of flesh” (Ecclesiastes 12:12), and the Roman procurator of Judea, Festus, complained of St. Paul that too much learning had made him mad (Acts 26:24). Seneca, in the first century CE, in one of the epistles to Lucilius identifies not only immoderate but also hasty or varied reading as the cause of distraction in learners. The eighteenth-century American physician Benjamin Rush believed that booksellers were particularly prone to insanity because their profession led them into the habit of “the frequent and rapid transition of the mind from one subject to another.” Enduring stereotypes of melancholy, abstracted, desiccated, emotionally stunted, and asexual or uxorious scholars (Dryasdust, Dr. Syntax, Edward Casaubon, and many others) present the excessive consumption of knowledge as the prime cause of error, while stereotypes of haughty, charismatic, or dangerously predatory scholars tend to place the emphasis on the self-love of the scholar or the forbidden nature of the knowledge.
Wherever it is located, error elicits different kinds of reactions and different estimates of the danger it poses. When a member of an accredited or expert elite makes a mistake, the error may put into question the individual’s expert status, which is what allowed the opinion to have authority in the first place; shame or defensive rage may be an appropriate response, allowing for differences in individual temperament and the gravity of the error. In contrast, in more democratic structures of knowledge, the threat of error may not be a fundamental discrediting of a self but rather an ordinary aspect of life: an error may be corrected in the next newspaper issue, installment of a novel, vote by the citizenry, and so on. The gravity of an error, the interpretation of its significance to the larger society, is also likely to be influenced by social contexts. The cultural theory of risk offers a sociological interpretation of varying responses to the significance of error, taking individualistic societies generally as discounting the danger posed by error, collectivist egalitarian ones as tending to fear catastrophic scenarios, and hierarchical ones as interpreting error as manageable so long as it falls within an expertly defined margin.
Some of the errors to which information is prone are specific to the medium by which it is disseminated. Compared to literate cultures, oral ones have a relatively limited ability to fix utterances and so cannot impose analogous kinds of tests for coherence and consistency on large corpora of written statements. Tests for errors of consistency can be facilitated by manuscript and print instruments for making comparisons (e.g., cross-references, *concordances, annotation, tables); these nonsequential reading practices assisted in making errors visible in new ways. In contrast, written works introduce new problems of error that arise from the difficulties of establishing the trustworthiness or authority of text. Speakers or reporters of speech in an oral culture are generally known, so the question of whether to credit them does not raise to the same degree the problems of anonymity, pseudonymity, and forgery associated with written texts, which are more easily alienated from the circumstances of their production. (The difference is one of degree, not kind, of course; the oral transmissions of the Jewish Mishna and Islamic hadith both placed a strong emphasis on citing original sources.) Readers, for their part, without an oral informant to guide them or a witness they might cross-examine, may use a text in ways that are remote from its original purpose and thereby introduce error; moreover, the use of writing itself may atrophy the memory of readers who rely on it (cf. the much debated passage in Plato, Phaedrus 275a).
The introduction and growth of print amplified anxieties about written errors. It was feared that readers might succumb to diametrically opposed temptations: their attention may be dissipated by the sheer variety of print matter available to them, or they may be corrupted or radicalized by their absorption in newly available dangerous texts (such as vernacular translations of the Bible in the sixteenth century, or novels in the nineteenth). By virtue of its mechanical *reproducibility, print standardized the information disseminated in any print run (of course there was some variation even within the run, due to stop-press corrections, the removal or addition of material, or other irregularities); on the other hand, there was no guarantee that the disseminated information was itself free from error, as the *publishers may have chosen a defective source to print, or carelessly converted it into print. Moreover, printed matter now often existed in competition with alternative texts supplied by other printers. Works that were designed to be definitive sources could be subject to proliferating ripostes and polemics; learned reference works sometimes inspired the production of counterworks of reference; and a whole genre developed to correct common or popular error (erreurs populaires, pseudodoxia epidemica).
The problem of trust that ensued with the proliferation of competing texts was handled in a number of ways: works might obtain an imprimatur or the approval of censors, or the support of prestigious patrons or subscribers; they might be brought out by a printing house, academic society, or other body with a reputation for the quality of its authors and members, and which perhaps also employed correctors or, from about the middle of the sixteenth century, professional proofreaders; royal academies secured the agreement of peer reviewers prior to the publication of articles from the eighteenth century, as they did with book publication in the seventeenth century, as Mario Biagioli has shown. More informally, works might advertise their authority through scholarly apparatuses, dedications, frontispieces, blurbs, or other displays; they might be published, or claim to be published, in certain locations with a reputation for publishing works trusted by particular readers (Amsterdam, Paris); or they might be published by coreligionists (the number of theological works published in the early modern period is far greater than other kinds of learned books because, for the most part, they appeared in three different confessional versions, as Ian Maclean observed in Scholarship, Commerce, and Religion). In the long history of print, the many changes to practice and technology created many new possibilities for errors and correction. Even a narrow subject such as proofreading shows much historical diversity: in the use of correctors, authors, or both, to proofread; in the capacity to produce proofs of an entire work at one time, rather than in installments, which changed the ability to check for internal inconsistencies; in the method of proofreading, solitary and silent or joint and aloud; and so on. In general, we may see print as both a stabilizing and a disruptive force in the relationships of information and error, amplifying both. The same is true of *new media, where analogous problems of the trustworthiness of information and its capacity to deform provoke anxiety.
The quantity of information that is mechanically or *digitally reproduced creates new sources of error in the retrieval of information. There is too much to know. Ann Blair documented the major techniques of information management in the early modern period (note taking, indexing, compiling, and finding devices). One technique, the index card or equivalents (scheda, fiche, Zettel), has been used since antiquity. As important as the cards themselves were systems for the storage and retrieval of cards. One of the most elaborate of these was the large “literary closet” (scrinium literatum) devised in the seventeenth century. Scholars would improvise a number of makeshift solutions over the centuries (as Ian Jackson documented): cards would be stored in bags placed on shelves or hung from clotheslines, in baskets suspended by hooks, inside biscuit tins on shelves, in cubbyholes, or in slip boxes (Zettelkästen). With the widespread use of index cards, the kind of person who made use of the information atomized on them (the fichiste) became an explicit object of scorn (for example, by Joris-Karl Huysmans, Anatole France, or Charles Péguy, in the decades around 1900).
In the case of digital information, the crucial difference from print is the availability of keyword search. This permits the navigation of information on a vastly enlarged scale. The enormous and unwieldy number of possible results in a keyword search is reduced to manageable size through page rankings, which has the effect of concentrating knowledge in a few places, whose high rank then tends to be perpetuated. Moreover, page rankings may be customized to particular users, and even when they are not, a particular phrasing of the keyword may dramatically alter the results; as a consequence, users may be directed to their preferred version of information in cases where there are competing accounts. Digital information with keyword searching is even more efficient than print at creating distinct epistemological communities, who employ different error criteria, trust different authorities, and know different *facts.
Error may be a result not only of the scale of the information but also of its system of organization. Alphabetical order has had many discontents: in a letter of 1530 Erasmus (Ep. 2260) complained that a pirated version of his unpublished epitome of Valla’s Elegantiae spoiled its usefulness by imposing alphabetical order onto it. Joseph de Maimieux sought to replace the “alphabetical chaos” of dictionaries with his new pasigraphic order (a proposal from the 1790s that stands in a long history of attempts to construct symbols for a universal language corresponding to ideas rather than sounds). Coleridge (1772–1834) believed that the alphabetical order of *encyclopedias had brought about the “dangerous habit of desultory and unconnected reading.” Conceptual orderings are generally contested: the spatialized representation and dichotomized brackets of Ramist texts encountered considerable resistance, speculative dictionaries and encyclopedias were rejected by positivists, and the order of Roget’s Thesaurus was mocked by Nabokov and others. Critics of historicism may object to chronological arrangement. The absence of order is likewise subject to criticism: Coleridge elsewhere worried over the harm incurred by “the general taste for unconnected writing,” which he traced to the Spectator (no. 46, 1711); a generation before, James Beattie complained about the new fashion for “uncemented composition,” blaming it on Montesquieu; and sometimes even madness was feared to be a possible outcome of reading disorganized information.
Since error is usually traced to multiple sources in any period, many of the changes to how error is understood come about through the reevaluation of the threat posed by existing sources of error, rather than by the discovery of new ones. Christianity, for instance, defined itself in opposition to two kinds of error: the literalism of the Jews and the allegorism of the Greeks. The tension between letter and spirit as sources of error runs through its history, but at different moments the errors due to lawless enthusiasm and those due to the dead weight of the past have been judged differently. Another example is the tension between learning from books and learning from nature or experience, a conflict frequently articulated by historical actors themselves, but which varies in time and among different populations. When one criterion of error is dominant, there is an incentive to forge statements that pass as true by that criterion, a situation that—depending on the quantity, significance, or exposure of the forgeries—may lead to a readjustment of criteria or to the establishment of new ones, which may in their turn invite new kinds of forgeries. The accumulation of knowledge among polyhistors in the seventeenth century routinely incurred the accusation of charlatanry in the early eighteenth century (Mencke’s De charlataneria eruditorum [1715] was the most famous of these satires); the good manners of the *saloniste in the *Republic of Letters were routinely denounced as hypocrisy by the end of the century. Positivist knowledge (or “mechanical objectivity,” as Lorraine Daston and Peter Galison describe it) in the nineteenth century sought to correct errors attributable to the vagaries of the observing subject by removing traces of the self altogether; in turn this criterion for separating knowledge from error was followed, in the early decades of the twentieth century, by one overseen by the trained judgment of experts competent to recognize patterns.
The understanding of error may also change when the authority guaranteeing the validity of information is challenged or alters. One influential argument from the last century (still controversial in its scope) ties the emergence of open-ended inquiry and rhetorics of persuasion in sixth- and fifth-century Greece to the hoplite reform by which members of the warrior class were included in some institutions. Christianity is marked by a central tension as to whether the authority of truth is to be located in the body of believers or a hierarchy within the church, or rather the degree and the domains of these respective authorities to define error. The inclusion of religious minorities in scientific pursuits (Dissenters in nineteenth-century England, Jews in twentieth-century America) changed, and reflected a change in, scientific practices. How error changes as a result of challenges to the trustworthiness of traditional authority is unpredictable; it may either challenge or reinforce the tradition in question. The debate over miracles led not only to Hume’s skepticism about whether to credit previously authoritative witnesses of unusual events but also to Bayes’s mathematical interpretation of their probability, Babbage’s defense of miracles in the Ninth Bridgewater Treatise (1837), and Whateley’s incisive satire on historical criticism (Historic Doubts Relative to Napoleon Bonaparte, 1819).
New sources of error are sometimes discovered. One of the most significant was the protracted realization of the role of randomness in observation and its ability to be quantified. It was a major shift for early modern experimenters to record all the results of their experiments, rather than just the result of the best experiment, as it was contrary to their practice as craftsmen to record good and bad results equally (as Jeb Z. Buchwald and Mordechai Feingold have argued). Including tables of all the discrepant data was a practice slow to develop in the history of the publication of scientific reports (Buchwald and Feingold date it “towards the end of the eighteenth, or the beginning of the nineteenth, century”). The realization that randomness exhibited statistical regularities was even slower: the statistical regularity of nature was controversial even in the nineteenth century, when fundamental laws of thermodynamics took a statistical rather than Newtonian form, and the statistics of human behavior are occasionally controversial even today (it is difficult to assign blame for random errors and thereby prevent or correct them). That error might paradoxically become a source of truth was also an early discovery of *philology: rather than identifying a best text, or combining the best readings from several texts, philological criticism took errors (or rather a special category of error, the Leitfehler, or significant variants) to constitute true evidence for the history of the text, permitting the reconstruction of earlier versions that were less error prone than surviving witnesses, at least in theory. Psychoanalysis likewise converted errors (such as verbal slips and other parapraxes) into a previously unsuspected source of truth.
New technologies have changed how error is made visible, as we have seen in the case of the technologies of writing and print, but it is difficult to know whether to attribute to them a larger role as agents of change. The display of information so that error is not removed or corrected but made prominent—for instance, the presence, since the sixteenth century, of a critical apparatus to accompany texts, or the use of the scatterplot in the natural sciences since John Herschel in the nineteenth—did not require technical innovation but a new practice, relatively slow to emerge, of systematically including rather than excluding error. It is sometimes argued that the competition between epistemological communities that have been organized in the wake of massive new disseminations of information, and the consciousness of different error criteria active in different communities, may spur innovation to attempt to bridge them via new techniques of truth finding and error management, but such awareness may instead entrench communities further and increase their remoteness from each other.
This discussion has been limited to inadvertent errors, but mention should also be made of carefully deliberated ones: lies, flattery (regularly depicted in antiquity as a major cause of political and ethical misjudgment), willful mistranslations, typographical errors inserted in order to escape censorship, fictitious entries in reference works to establish priority or *copyright, vandalism (a Wikipedia article about a fake Akkadian demon with striking similarities to Jesus Christ lasted a dozen years, as have three other hoaxes identified so far), and errors in the form of parodies and spoofs designed to escape detection.
Kenneth Haynes
See also censorship; excerpting/commonplacing; information, disinformation, misinformation; knowledge; observing; proofreaders; storage and search