Typos, something that every reader will notice and correct for at one time or another, are among the simplest examples with which to introduce textual criticism. One of the pleasures of textual scholarship is how it reveals simple typos to be not so simple after all. “Typo” is, of course, short for “typographical error.” Taken literally, it is now a largely outdated term, since type (whether as movable pieces of metal or as part of a typewriter) is now far from the dominant technology for producing and reproducing texts. It hasn’t been dominant for decades. Type presses long ago gave way to other technologies and are now primarily used in hobbies or for books as works of art. Some years ago, I was a member of an English department that still had a $400 budget line for typewriter repair, though no one in the department still had a typewriter. In an effort to get the money transferred to another line so that we could actually use it, I tried to find out if there were any typewriters anywhere on campus and surprisingly found some in storage. None of them looked like they had been used in fifteen years. But it certainly wasn’t the case that no one at that college had created a typo in all that time. In short, “typo” has become a metonym for all manner of textual errors. We talk about “typos” in email messages and web pages that are never printed. We could just as logically call such errors “slips of the pen.”
Textual scholars are interested in how we use particular technologies — the quill pen, the lead type, the word processor — because it can help explain how a particular text came to have certain features. Sometimes, by studying the methods of medieval scribes we can show how a scribe might be reasonably confused about what he (or she) was copying and produce a particular error. And some kinds of errors are common enough to have their own names, like homoioteleuton, or “eye-skip,” which occurs when a scribe moves his eyes too far ahead in the exemplar as he is copying, unintentionally deleting some material that should appear in his copy.
One of the most common ways typos get introduced now is through optical character recognition of scanned documents. I suspect that an OCR error is what is behind one troubling mistake in a textbook I taught from recently: “human actions, while certainly hard facts, are not hard facts as distinguished flora language.” The sentence becomes much clearer when we correct the erroneous “flora” to “from.” It seems unlikely that “flora” comes from someone mis-“typing” at a keyboard, especially given that “flora” is one letter longer. Then there is the fact that “l” and “r” are not near one another on the standard keyboard; they’re not even struck by the same hand, much less the same finger, and so they are unlikely to be switched accidentally by a keyboardist. It seems much more likely that an OCR program mistakenly found “flora” to be the closest match to the forms it was “seeing,” the “r” being mistaken for an “l” (which is easier to do than one might think), and the “m” being misread as an “ra.”
While OCR is very good at what it does, it is nowhere near perfect. (Peter Shillingsburg has a wonderful discussion of what OCR error means for readers of digital texts in his book, From Gutenberg to Google. The most remarkable data for the relative reliability of OCR is available on Google books, whose enormous database of scanned books allows us to search for text that we can check against the scanned images of those books. Searching Google books for “be distinguished flora” provides a handful of examples where the OCR reads “flora” for “from.” (I chose the phrase “be distinguished flora” because I thought that it would result in fewer false hits. “As distinct flora” produces many false hits, pointing to texts about unusual plant life.) While there aren’t as many, the errors do confirm that OCR sometimes reads “from” incorrectly as “flora.” I learned to use Google books to look for OCR typos in this way from another blogger, whose example is much, much funnier than mine, and shows how prevalent and how significant OCR errors can be. As Wesley Raabe writes, “Romance writers, be forewarned, the child who leaps into her arms or the heroine who leaps into his has almost a five percent chance of ending up elsewhere, when GoogleBooks OCRs your text.” Gothic scripts made it difficult for medieval scribes (and modern readers, now) to determine which minims, the short vertical strokes that constitute the bodies of letters like “r” (one minim) and “m” (three minims), should be grouped with which minims. Something very similar happens with these OCR errors.
One typo I explored in Textual Editing and Criticism appears in several electronic editions of James Fenimore Cooper’s The Last of the Mohicans:
“It is extraordinary!” said Heyward, taking his pistols from the place where he had laid them on entering, “be it a sign of peach or a signal of war, it must be looked to. Lead the way, my friend; I follow.”
In the textbook, I wrote that the error, “peach” for “peace,” was more likely an OCR error than an error produced at the keyboard, as the “e” and “h” are keyed by separate hands. But “peach” might also have been an actual typographical error, that is one produced in the age of type. The earliest editions of Cooper’s novel would have been set in type by hand, each letter, punctuation mark, and space created by a compositor placing lead type and non-printing spaces in the proper order. “Peach” and “peace” might have been easily mis-set for one another, because in the cases that stored the letters in a printer’s shop, the bin for “h” was immediately below the bin for the “e.” (Click here for many examples of the “English lay,” or the general pattern of letters in cases used by English language printers.)
Here’s a video of two men setting type, and while it’s low in production values, it does show how each letter, each punctuation mark, and each space needed to be added one at a time. At the very end of the video, you can hear one of the typesetters (or compositors) explain how errors can creep in to type when someone else put away the letters incorrectly. (In technical terms, the compositor “distributed” the type incorrectly, resulting in a “foul case.”)
I am fairly certain that “peach” is a genuine typo, not a computer keyboarding error, nor an OCR error, because the little typesetting I have done tells me that this kind of error is easy to make. But I have no proof, and other explanations are quite plausible. I looked at many nineteenth century editions and didn’t find any with the error.
When I read Wesley Raabe’s blog post, the one about Google books and OCR errors, I thought that this might be a great way to find the “peach” error. But my results surprised me. First of all, I did not find the source of the error. While it means that I have again struck out in finding an edition that might have served as the exemplar for the electronic text, it also suggests that “peach” might be an unlikely OCR error for this particular edition. Interestingly, “peach” does show up as an OCR error for “peace” in other scans in Google books, but most often when the words are in all caps. Capital “E” looks much more like capital “H” than lower-case “e” looks like lower-case “h.” Second, there were only four hits returned by Google books for “‘sign of peach’ Cooper,” fewer than I might have expected, and one of them was my own book. But third, and most surprisingly, the other three hits are from books published in the last ten years. Neither of the two publishers gives much information about how they prepared their texts of the novel, one saying only that the text was downloaded from the internet and “extensively edited and typeset,” and another saying nothing at all about the text’s provenance, so far as I can tell. I think it fairly likely that each of these editions was likely printed from the text originally released by Project Gutenberg, which many other sites have re-distributed, sometimes without acknowledging their sources. To their credit, Project Gutenberg works to reduce errors, and this error no longer appears in either of the two texts of the novel that they offer for download.
Part of me is scandalized that two publishers simply downloaded a free version of a long-out-of-copyright novel, printed it (and that probably very cheaply), and now sell it through the various online bookstores, charging more than $20 a copy for their labor — if you can call it that. But only a little time spent studying the history of book publishing and of copyright will show that such behavior has never been all that unusual. The first printed version of Shakespeare’s Hamlet, after all, what we know as the “Bad Quarto,” was very likely an early seventeenth century version of the same sort of publisher’s game. So, another way of looking at my peach is that it might be good proof that textual scholarship has an actual value in the marketplace. Typographical errors are usually only annoying. “Peach” is easily corrected by most readers. “Flora” may be a little harder, but not impossible. But in the aggregate, as the errors add up, proliferate, and circulate, the value of textual scholarship becomes clearer.
David Foster Wallace’s novel, The Pale King, unfinished when the author died, has just been published. The fact that an editor has been wrestling the very unfinished documents into shape makes for interesting news. For my money, it’s what Wallace’s editor, Michael Pietsch, is quoted as saying at the end of the article that is most interesting. It begins as an anecdote about editing Ernest Hemingway’s The Dangerous Summer, which was also unfinished when Hemingway died:
the original text was more than 100,000 words. Pietsch helped cut it to around 75,000 and "The Dangerous Summer" was released in 1985.
"It was fascinating," Pietsch said. "You saw his great, mature writing style, these sentences that were long and flowing and muscular and beautiful, and he was writing about himself as a public figure, with a sense of humor, which was just delicious.
"It was also long and baggy in parts, and there were passages I felt you could cut out very cleanly. But you don't change the author's words. Not David Foster Wallace's and not Ernest Hemingway's. You either cut out a passage or you let it stand."
It wasn’t that many weeks ago that a similar debate over the powers of an editor was staged in the media, namely concerning Mark Twain’s The Adventures of Tom Sawyer and Huckleberry Finn, edited by Alan Gribben. That edition got the notice of the national news media for Gribben’s decision to eliminate certain racial epithets from the text. (Interestingly, what very few people seemed to have noticed is that this “censorship,” as others have called it, is probably not as big a change as is the fact that this edition treats two works as one continuous piece.) I think the Twain edition is the unspoken context for the Wallace article’s concluding remark, “you don’t change the author’s words. Not David Foster Wallace’s and not Ernest Hemingway’s. You either cut out a passage or you let it stand.”
Pietsch’s logic is shared by many, I suspect. Changing an author’s words feels like an ethical violation to many, and Pietsch seems to express it as such: “you don’t change the author’s words.” Except that he and many, many others are willing to cut them altogether. My question is, how is cutting passages amounting to 25,000 words less of a change to an author’s words than substituting one word for another, like “Indian Joe” for “Injun Joe?” Does authorship only function at the level of individual words? Now, Pietsch is not producing a scholarly edition, in which the requirement for theoretical and methodological consistency is different. But it does make me wonder whether the ethical complaint we often voice about changing an author’s words isn’t covering for something else?
The value of teaching textual criticism to undergraduates in literature classes is the way in which it requires a kind of close reading. Among the essential and disciplined activities that a textual editor engages in — including what many people think of as almost mindless, like transcription and proofreading — collation is perhaps most valuable in the literature classroom. I would guess that collation is talked about most commonly when discussing Sir Thomas Wyatt’s sonnet, “They Flee From Me.” The version available at the link is edited from a manuscript in the British Library copied under Wyatt’s supervision. But the first printed version of the poem, appearing in an anthology of poems we now call “Tottel’s Miscellany,” revises the poem in significant ways. Tottel revised the poem, trying to make the rhythms of the poem suit his sense of meter, but also changing the tone in places. When students encounter this poem in a British Literature class, they’ll usually encounter both versions, printed one after the other. Having students collate these versions leads to a lively classroom discussion every time.
It’s this principle of collation as close reading that animates W. D. Snodgrass’s book, De/Compositions: 101 Good Poems Gone Wrong.
Snodgrass includes both versions of Wyatt’s poem, and he prints a draft and revision of Donald Hall’s poem, “The Man in the Dead Machine.” For the most part, Snodgrass doesn’t count on finding competing versions of poems but instead rewrites the poems himself. Snodgrass isolates some aspect of the craft of poetry, like meter, and compares the authorial version to his own bad version, throwing that aspect of the craft into relief. In order to show how the trochees of Blake’s “The Tyger” matter, Snodgrass rewrites the first stanza in iambics:
O tyger, beast that burns so bright
In darkling forests of the night,
What godlike hand, what deathless eye
Dare frame thy fearful symmetry?
Then he re-rewrites the first stanza in anapests:
O tyger, you creature that’s burning so bright
In the threatening darkening forests of night,
What hand of immortal, what diety’s eye
Dare hope it could fashion thy feared symmetry?
Snodgrass’s plays on his nickname, “De,” in the title of the book, an attitude in keeping with his playfulness throughout — he’s often trying to be funny. In the end, Snodgrass’s aims are traditional: he wants to illustrate formal qualities of poetry and how form functions. But he’s acutely aware of how difference creates meaning, too. As he writes in his preface, “The better the poem, the more likely it is . . . To accept a multiplicity of readings. . . . Often, my de/compositions are easier to understand than the originals — usually because there is less to understand in them.” While being able to talk about formal matters is generally useful, for me the value of the book lies in the reading method that the book insists upon, one in which collation has some prominence. More than one editor has said to me that they do not feel they know a work at all until they have begun to edit it. I think that it’s the unheralded, supposedly lower-order tasks, like transcription, proofreading, and collation, that bring the familiarity and understanding that these editors feel they have. Snodgrass’s book certainly isn’t for everyone, but I wonder if we wouldn’t do well to have more textbooks like Snodgrass’s, textbooks that emphasize versions and genetic editing — editing that recovers the process of a text’s creation or circulation.
Here's a talk given a few months ago on a topic I find fascinating — error.
Kathryn Schulz speaks entertainingly about being wrong. She argues that our culture teaches us that
getting something wrong means there's something wrong with us. So we just insist that we're right, because it makes us feel smart and responsible and virtuous and safe.
There's a great deal in this snippet to unpack and apply to textual criticism and textual theory, including the ethical dimension of error — why "virtuous," after all? (I have long noticed that my students approach textual editing assignments with some trepidation, not so much because they fear they'll get the wrong answer, but because they feel they owe something to the author to get it right.)
Perhaps I will get a chance to discuss this in a future post. For now, I'd like to say that I find her discussion interesting because textual criticism was for so long consumed with thinking about error. A. E. Housman famously defined textual criticism as "The science of discovering error in texts and the art of removing it." But Housman was also wrote in the same essay that
we should neglect no safeguard lying within our reach; that we should look sharp after ourselves; that we should narrowly scrutinise our own proceedings and rigorously analyse our springs of action.
Similarly, speaking about a surgery gone horribly wrong, Schulz says,
trusting too much in the feeling of being on the correct side of anything can be very dangerous.
In the last few decades, textual theory has begun replacing thinking about error with thinking about difference. There are many reasons for this, not the least of which is the influence of post-structuralist theory' critiques of binary thinking. I like to think that the lessons of those critiques are already present in the processes of textual criticism, that textual critics have begun to learn the lessons of textual criticism and so have begun to develop a language that frees us from binary right/wrong thinking.
Before we abandon the binary completely, I'd like to say that becoming attuned to the ways in which we might be wrong and the ways in which the texts we read might be wrong — what Housman writes so eloquently about — is among the chief benefits of studying textual criticism. For Housman, one needed to learn to think in order to do textual criticism. I would argue that doing textual criticism teaches us to think more carefully, that is, to be skeptical and self-aware as we read, as we weigh evidence, and as we make decisions. Textual criticism is, in short, remarkable for providing an opportunity for students to grow intellectually. It may be that we need to pass through the binary thinking of "the discovery of error" before we can begin thinking about textual difference without error.
I've only just become aware of an event, a "Karma Chain," that happened at the end of April during the annual PEN World Voices Festival of International Literature. There are so many delicious ironies packed into the thing that I have to comment.
This is what I heard at one time: Like a shimmering star, or a flickering lamp, a fleeting autumn cloud or a shining drop of dew, a phantom, a dream, a bubble, so is all the existence to be seen.
and ended as:
Follow the glass stone. Follow the glass stone. The droid from hell. If anything exists, it changes.
The droid from hell? That's fabulous! But even the interpolation of Star Wars (Use the force, Luke!) into a Buddhist sutra is not the most delicious part. It was who waited for the message at the other end of the line:
Salman Rushdie, of course.
Rushdie's brilliant novel, The Satanic Verses raised the ire of many in the Muslim world, including Ayatollah Khomeini of Iran, who issued a fatwa calling for Rushdie's death. The Satanic Verses is itself very interested in textual history- it draws its title from a probable apocryphal story of a textual problem in the Koran. A large segment of the novel depicts in a dream (of a man who is losing his mind, no less) the passing of the text of the Koran from the archangel to Mohammed to a scribe, who willfully miscopies some verses, and who is not caught by Mohammed when the scribe reads the verses back to him.
Who better than Rushdie to receive the message of the Karma Chain? And what better message in this age of Christian and Islamic fundamentalism (both of which tend to reject not only evolutionary theory but also the idea that their holy texts might be subject to the instability inherent in all communication) than a Buddhist message about the fragility and transitoriness of all existence, including texts. (And about how beautiful that existence is, perhaps because it is fragile and transitory!) And what better way to end than for Rushdie to tweet the result when the game was over. Odd, though: I haven't been able to track down those tweets.
Finally, where better than on the High Line? Every city is a palimpsest, but no part of New York is more visibly a palimpsest than the High Line, a great park emerging from an obsolete train line. More photos here.
With the end of the term, I've been better able to finish off the old New Yorkers I wasn't able to read while grading composition essays. In the same issue as the story about Salman Rushdie and Lama Pema Wangdak's Karma Chain is a Talk of the Town piece about the revival of Anything Goes currently running on Broadway.
In it we have the confluence of theater-, film- and television history, musicology and more. But for me the takeaway with this story is that some of the research that artists do to prepare looks a great deal like academic research, particularly textual criticism.
It turns out that the director and choreographer, Kathleen Marshall, visited the Paley Center for Media to research previous versions of the musical, and the article lists in bullet points some of the interesting differences she discovered in the archives. As Michael Schulman, the author of the piece, puts it:
the title is more apt than you think. Half a dozen versions have existed since its première, in 1934, and many of the Cole Porter songs now associated with it— "Friendship," "It's De-Lovely"— were originally written for other shows and spliced in later. A Paramount film, released in 1956, starring Bing Crosby and Mitzi Gaynor, revamped the plot completely. (Two actors haggle over two leading ladies — high jinks ensue.) Just about the only thing that has remained constant about "Anything Goes" is that it takes place on a boat.
The bulleted list of unusual aspects of the 1954 television version and of the 1936 film version is entertaining. It is interesting that the list begins with what was being advertised in the commercials, a paratextual feature that would normally not have been included in scholarly discussions not so long ago but that now is understood to be enormously important. But maybe the best bits are the lyrics you thought you knew: "You're the top./ You're the Swanee River./ You're the top./ You're a V-8 flivver."
I had to look up "flivver." I was guessing it would be a kind of alcoholic drink. I was wrong.
What are these roots that clutch, what branches grow
Out of this stony rubble?
Here’s one answer: this very cool app — a digital edition of T. S. Eliot’s The Waste Land. It’s got the text of the poem, clickable shoulder notes taken from a student edition, recordings of the poem, and a gallery of images meant to put the poem in context. Yes, the notes and contexts are rather weak for my tastes. No, of course, it’s not the first digital book for the iPad. Nor is it the most visually arresting. (For a sense of how programmers have taken advantage of the iPad for other books, watch this TED video about Al Gore’s Our Choice.) But the app is getting rave reviews, like this one from The Telegraph. And those rave reviews are probably deserved.
As a teacher interested in bringing textual criticism to my students, there are two aspects of the edition that strike me.
But the images included are old ones, black and white, relics from a time when printing in color was very expensive. The iPad does colors beautifully, and I know that there are different colors of ink in the manuscript that would help readers distinguish different stages of revision. Moreover, this is an app given to making use of the latest media for splashy effect. For instance, it includes a filmed recital of the poem by Fiona Shaw (a brilliant actress unfortunately most known for her role as Petunia Dursley) that is synchronized with the text. That’s right: if you hold the iPad in portrait (rather than landscape), the top half displays the movie, while the bottom half of the screen displays the text, scrolling with Shaw’s recital. Good color scans of the manuscript were not out of reach for this app, so, that the old facsimile images have been used is an unfortunate shortcoming. Nor would it have been prohibitive to include Valerie Eliot's very valuable transcription, which is what makes the facsimile edition so worthwhile. It’s still brilliant, mind you. But how much more brilliant it could have been with high-quality color images of the manuscript. I should also add that there are no credits for the images that I could find; I’m basing my claim that these are the old photographs on my experience with the printed facsimile. As a scholar and teacher, I find the lack of credits disturbing.
The second aspect is related to Shaw’s performance. One of the tricks of the app is the ability to listen to several recorded performances, including two by Eliot himself and one by Sir Alec Guinness. Like Shaw’s, the readings are synchronized with the text, so that you can read along with Guinness, switching back and forth between readings almost seamlessly, starting a stanza with Eliot’s voice but ending with Viggo Mortensen’s. That’s some cool collation. I noticed a couple of times that the readings did not exactly match up with the text on the screen, and I wanted some way of annotating or bookmarking the moment to explore it in detail, but couldn’t find out how to do so. Maybe that function is there, somewhere, but I gave it a pretty thorough checking.
Ultimately, the edition raises questions for our digital publishing future: what features do we need in our electronic textbooks and electronic scholarly editions? What will our apparatuses look like? Does anyone out there know of digital scholarly editions being made specifically for eBook readers or being made as apps for tablets like the iPad? If you do, tell me about them. I’d like to hear.
Many years ago, I argued with my father about how much computer programming would become a part of our high school curricula. I’d been told that the inexpensive TRS-80 and Apple II meant that computers would soon be everywhere, so I reckoned that everyone would need to know how to program. But my dad predicted the industry would turn computers into gadgets. You wouldn’t program it yourself. You’d only need to push this button or that to get it to do what you wanted. He didn’t live to see my iPad, but it’s exactly what he was talking about. (If you’re feeling nostalgic for those early machines, by the way, there’s an app for that!)
I think a similar argument is playing out now in the world of digital texts.
It is an oversimplification to state that all digitized texts can be classified on a continuum describing how “structured” the data is. On one end of this continuum is the text tagged thoroughly with XML (extensible markup language) following the latest guidelines from TEI (the Text Encoding Initiative). The most pleasurable way to see what TEI encoding looks like is to watch this iconic Bob Dylan music video with a scrolling structured transcription of the text on the signs Dylan reveals, ideal for the classroom!
On the other end of the continuum is the text file with little or no markup at all.
The argument divides along the lines of those who want (and spend much time and money creating) highly structured digital texts, versus those who figure that we need smarter and smarter search engines to address unstructured data.
The best way to understand what’s at stake in the continuum of structuredness in text files is to imagine the kinds of searches possible with them. Book titles are the classic example. Let’s say you’re searching through novels from the second decade of the nineteenth century for the titles of any books named in those novels, as The Mysteries of Udolpho is named in Austen’s Northanger Abbey, but without knowing in advance what those titles are. If you already know the titles, it would be easy. But if you don’t have strings of text to search for, to perform this sort of search with Google Books is not much different from searching through the books by hand, looking page by page for the display codes that signal book title to our reading brains: capital letters, italics, or quotation marks.
Files that preserve those display codes and search engines that can find those display codes without specifying the text strings would speed things up — this is possible with most word processing programs, for instance. But you’d still need to sift through your results to remove false hits. We use italics, after all, for other purposes — headings, emphasis, and foreign words, to name a few.
At the most structured end of things, performing this search with data that has been encoded according to the careful and complex rules of the TEI, would be easier still, as TEI tags more than display codes. (Click here for an example of a highly structured encoding of Shakespeare’a sonnet 17, tagged so that each metrical foot of each line is indicated.) You’d need only search for items tagged as book titles — just so long as the encoders thought to tag all the book titles.
Structured data means that a human has begun to interpret the texts for readers, and the argument is that doing so adds value to the digitized text by making it more user-friendly. The other side would argue that what we’re doing is making the digitized text more friendly for weak search engines. We don’t need to expend the time and money it takes to structure data, my dad would say. Rather, we need to build smarter search engines that can sift through unstructured data intelligently. Years of teaching students how to write researched essays makes me wonder if we have thought enough about how search engines are modes of interpretation, too.
In an earlier post, I pointed to some reasons not to trust wholeheartedly the digitized texts on Google Books. The argument in favor of structured data aligns well with those worries — fewer false positives, fewer outright errors. Just as human proofreaders catch typographical errors, human taggers can make an unruly text more usable. But the tradeoff is that interpretation is, in part, an act of exclusion, limiting (perhaps unconsciously) the ways in which we can understand ambiguities. Perhaps unstructured data and a good search algorithms will empower readers in ways that taggers can’t yet imagine.
