The Ghosts of Bindings Past:Micro-Computed X-Ray Tomography for the Study of Bookbinding
This essay describes the results of a new application of micro-computed X-ray tomography (µCT) to conduct nondestructive investigations of the binding structures of premodern books. This application addresses a twofold challenge in the study of historic bindings and their construction. Few premodern books survive in their original bindings. Moreover, until recently, when books were rebound, the original structures were rarely documented, and the remains were usually discarded. Where original bindings do remain in situ, much of their structure is, by design, hidden. Particulars of construction may be surmised; but without destructive disbinding, little can be proven. µCT enables an exploratory, multilinear approach to codicological investigations that makes bindings accessible in the form of tractable volumetric data.
[End Page 142]
Judging Books by the Outsides
"Why, it's one o' the books I bought at Partridge's sale. They was all bound alike,—it's a good binding, you see,—and I thought they'd be all good books …—sermons mostly, I think,—but they've all got the same covers, and I thought they were all o' one sample, as you may say. But it seems one mustn't judge by th' outside. This is a puzzlin' world." So opines Mr. Tulliver in George Eliot's Mill on the Floss, whose more literate daughter, Maggie, has been discovered reading one of these series-bound books, which is not a sermon at all but Daniel Defoe's 1726 Political History of the Devil. "Have you no prettier books?" asks their worldly visitor. He means something more suitable in content for a young girl, but the metaphor of "th' outside" hangs over the exchange, and over Maggie in particular, whose bookish desires and cleverness, whose very inner life, will here and elsewhere always be subordinated to the expectations of her gender. "'It's the wonderful'st thing'—here [Mr. Tulliver] lowered his voice—'as I picked the mother because she wasn't o'er 'cute—bein' a good-looking woman.'"1
This scene in Eliot's novel is an early literary record of the proverbial commonplace "don't judge a book by its cover." The binding is the part of a codex that transforms the leaves or bifolia into the durable structure of a book. But the binding is the part of the book that it is easiest to forget because so many of its elements are concealed by design and because reading is, by definition, a mostly textual pursuit. Those who fail to look beyond "th' outside" miss the logocentric point. Yet Eliot also seems to recognize in this passage that "th' outside" has its own ineluctable history.
Our purpose in this essay is to contribute to that history, although not through any further literary investigation. Eliot suggests that bindings matter to the history of books. We concur, and we report here on work we have recently undertaken to reveal evidence concealed by the covers on books, using the technology of micro-computed tomography—abbreviated as micro-CT or µCT.
µCT creates a three-dimensional representation of an object's interior by projecting a narrow beam of X ray through the object to a high-resolution detector behind. The three-dimensional effect is created by rotating either the object or the X-ray source and detector, and the CT gathering and compiling many narrow slices, giving a researcher many possible ways of looking through the object. µCT has a wide variety of applications in modern science, including clinical medicine, where [End Page 143] it is used for tissue analysis. It is also used in phenotyping and morphology in small-animal studies; for nondestructive precision engineering of jet engines; and in archaeology, where it is one of the methods available for visualizing the interior forms of ancient artifacts, from hominin jawbones to mummies.2 Our research presumes that old books are also objects worth looking into—that under the surfaces on which they bear texts and art, books offer valuable but often-overlooked information about their own and their makers' and users' histories.
Three major trends in the study of historical bookbinding provide context for our experiments with µCT technology. The first is often remarked: a shift, in recent decades, away from an interest in bindings as an example of decorative arts, toward an interest in the materials, craft practices, technologies, and histories contained by a book's structure. J. A. Szirmai's Archaeology of Medieval Bookbinding (1999) was an important milestone as the field developed in this new direction.3
The second trend encompasses the expansion of bookbinding research to new domains. This includes, for example, Georgios Boudalis's development of Szirmai's and others' work on the codex in late antiquity, including in Coptic Egypt. It also comprehends new work on nonwestern traditions—for example, by Jinah Kim (Buddhist book production), Mina Song (early Korean books), and Karin Scheper (bindings on books from the premodern Islamic world).4 In the discussion that follows we focus on the results of our µCT experiments with a Book of Hours from the special collections of Western University in Ontario, Canada: Western University, BX2080.C365 1500, MS Canon Grandel's Prayer Book. The book is a fairly typical late medieval French manuscript, but we embarked on our project hoping that, as we learned more about µCT's affordances, we would discover its applications for lessstudied, more ancient, and more fragile binding structures.
The third trend involves digital enhancement of the processes of systemizing and visualizing data about books' structures. W3C's Simple Knowledge Organisation System (SKOS) underpins the vocabulary and descriptive principles laid out by the Ligatus "Language of Bindings" project, led by Nicholas Pickwoad with Aurelie Martin, Alberto Campagnolo, Georgios Boudalis, and Athanasios Velios.5 Dot Porter's data model, VisColl, can be used to model the way in which pages are gathered for binding into books.6 A team at the University of Toronto led by Rachel Di Cresce worked with Dot Porter and Alberto Campagnolo to develop this data model into a web application, VisCodex, which produces interactive visualizations of codicological models.7 The University [End Page 144] of Pennsylvania's Schoenberg Institute has recently released VCEditor, an updated version of this co-conceived web app.7
Digital tools such as VisColl offer scholars new ways to organize and share the knowledge they individually gather from books. However, those scholars are limited by what they are able to observe, and the older the book under study, the harder the limits. Most medieval manuscripts no longer retain their original bindings. Szirmai estimates that no more than 1 to 5 percent of medieval manuscripts survive in an original or early binding.8 Under heavy use, a "permanent" medieval binding with wooden boards made in the European-style might last for roughly a generation.9 Many books were therefore rebound multiple times during the medieval and early modern periods. Books invariably contain evidence of their earlier structures: traces of old adhesives, original spine linings, reused boards, and unused sewing and lacing holes. But books' concealed structures further restrict an already limited corpus of evidence, even in the case of those that have survived from early periods with their bindings intact. Aside from some singularly paired back structures, most styles of European bookbinding (for example) obscure the binding's mechanical workings almost entirely. Leather covers, which wrap around the front, back, and spine of the book, conceal the stitching and supports that attach groups of pages together. Only in damaged books—for example, a twelfth-century copy of Bede's Seven Catholic Epistles, Oxford, Jesus College, MS 70, on which the spine is coming away, exposing the lining, supports, and sewing pattern—are the inner workings of bindings structures directly observable (fig. 1).
In 1975, Graham Pollard sought to bypass limits like these by using single-shot plain film X-rays to investigate a group of early English books in which boards from before 1066 had been reused to bind the same manuscripts after the Norman conquest. Pollard's work was elaborated by Christopher Clarkson in 1996.10 The X-ray images they and their collaborators produced focused on attachment methods—that is, the mechanisms by which the binder laced the bands onto which the gatherings of pages were sewn into the boards. Their results were somewhat mixed. Both Pollard and Clarkson confirmed the presence of modes of attachment that had not been identified in these books before. However, the limitations of the static images, particularly the superposition of structures that is inherent to plain film radiography, led to "inevitable difficulties" in interpretation.11 Without the ability to view the inner and outer extent of the boards, and trace the path of the band through them, much analysis had to remain deductive. [End Page 145]
High-resolution data from µCT alleviate some of these problems by producing a volumetric visualization of the book at ~40–100 µm voxel size.12 In October 2013, a British Library team including Christina Duffy, Claire Breay, Paul Garside, Flavio Marzo, Kristine Rose-Beers, Shaun Thompson, and Nicholas Pickwoad transported the early eighth-century Saint Cuthbert Gospel, the earliest book from medieval Europe to survive intact in its original binding, from the British Library to the Natural History Museum, where it was imaged in a Nikon Metrology HMX ST 225 µCT scanner. The team conducted three scans: two of the manuscript at 85 µm and 65 µm voxel size respectively, and one of a modern facsimile created by Jim Bloxam and Rose-Beers. Using the results, they were able to answer a number of longstanding questions about the book's unique historic binding. They found, for example, that the book's binder created the raised decorative pattern on the leather cover by attaching it over a clay-like core laid on the board.13 [End Page 146]
Scholars of early books have made other use of µCT. In recent years, teams led by the computer scientist William Brent Seales achieved headline-grabbing results using µCT to investigate carbonized rolls—a Pentateuchal scroll found at En-Gedi and a fragment of papyrus from Herculaneum—and a Coptic codex. In each case, the focus of the µCT research was on the recovery of the text, which had been lost to history inside an object too damaged and friable to open and read.14 Projects exploring other advanced X-ray technologies have likewise been focused on the recovery of text. In 2016, a group led by Erik Kwakkel and Joris Dik reported on their use of macro X-ray fluorescence spectrometry (MA-XRF) to "see through" the leather on a medieval book's spine and identify the text on the fragment of a medieval manuscript used as a liner.15 And, as recently as 2021, a research team led by Jana Dambrogio used a dental µCT scanner to unfold and read undelivered seventeenth-century letters without breaking their wax seals.16
Our group's work is not primarily on lost texts (though we will have some more to say about these later in the essay) but on the further development of µCT as a tool for the investigation of old books. Our experiments with Canon Grandel's Prayer Book suggest, as did work on Saint Cuthbert's Gospel, that µCT can bring to light otherwise inaccessible evidence of books' collation and bindings in the form of tractable digital data that allow multidirectional examination and dimensional and volumetric measurement of a book's internal structure. We report some specific results here: for example, empty sewing holes in Canon Grandel's Prayer Book—remnants of an earlier structure, a "ghost binding." We also report our development of µCT as a nonlinear method of a book historical inquiry. As it opens views into and around the book's three-dimensional form, µCT prompts further questions that can then be followed up through reinterrogation of data from the same scan. µCT, we conclude, encourages a dynamic mode of codicological research—a new way of puzzling at books, that has the potential to change our understanding of their history.
A Book and Its Binding, Examined through µCT
Western's Book of Hours—an abbreviated form of the Divine Office intended for personal use—was copied in the late fifteenth century (fig. 2)17 It is modest in scale (about 130 x 90 mm), and its decoration consists primarily of alternating red and blue minor initials alongside gold two-line and four-line capitals on red, blue, and occasionally gold grounds. However, as the most recent sales catalogue suggests, a series [End Page 147]
of stubs may be all that remains of a program of six miniatures since removed.18 The liturgical calendar is also missing, but the book contains an unusually full litany, including some saints whose veneration was limited to Flanders: Saint Bavo (Ghent), Saint Gaugericus (Cambrai), and Saint Ghislain (Tournai). Inscriptions on the paper flyleaf at the rear of the manuscript provide the earliest surviving provenance information. A partially erased ex libris places the book in the library of a religious institution in Lille.19 A number appears beneath the note in an apparently similar eighteenth-century hand. The number, 1734, which is replicated in the upper margin of the manuscript's opening leaf, could be a year—perhaps of acquisition, donation, or rebinding—a pressmark, or both. Another inscription, "Aurait appartenu au chanoine Grandel (would belong to Canon Grandel)," was added below it in pencil later. It is this note that gives the manuscript its current moniker, MS Canon Grandel's Prayer Book. [End Page 148]
François-Joseph Grandel, whose portrait remains in the collections of the Cathédrale Notre-Dame-de-la-Treille in Lille, was installed as a canon of the Collégiale Saint-Pierre de Lille on 11 September 1766. It is not clear at what stage the manuscript transferred out of the institutional library and into the possession of an individual canon. Before his installation Grandel was serving as a priest in the diocese of Tournai (an area that includes Lille).20 Grandel's presence at Saint-Pierre is registered in 1777, when he appears listed as canon in the general calendar for Flanders, at which time he lived in a residence on Rue du Glen.21 On 3 March 1783, he witnessed the church's annual statement of account for the years 1779–80.22 In 1781 a court case was brought against Grandel by the neighboring Priory of Fives after he failed to present a contract of sale for land he had purchased in the area in 1768 to the justice du seigneur within the requisite amount of time.23 The Collégiale Church at Lille was destroyed during the French Revolution. The canons celebrated their last office at Saint-Pierre on 5 November 1790 and were expelled from the premises the following day.24 Despite signing the official register in submission and obedience to the Republic in 1792—the same year that the church buildings were dismantled—Grandel did not return to the community after the Concord (1802) but resided in the neighboring town of Haubourdin until he died on 15 June 1815.25
One possibility is that 1734 marks a date contemporary with the manuscript's current binding. The binding is covered in dark-brown tanned leather, rough and cracked and patched in places where it was most worn. Under raking light, it is possible to see faint impressions on this leather cover left by metal clasps. There are no other traces of these clasps, however, and no evidence they were ever attached to the book's existing boards. The impressions suggest instead that the leather cover is earlier than the current binding. Due to this discrepancy between leather and boards, we suggest that the cover was removed from the book's previous binding and recycled for use on this one. That might explain another idiosyncratic feature of the book: the sales catalogue from the manuscript's 2011 acquisition by Western describes the spine title as appearing "upside down," suggesting that the words upon it were inverted when compared to the text block. The spine has since been rebacked, but perhaps the recycled leather was attached the wrong way up.26
The boards on Grandel's Prayer Book are lightweight. At several points, where the leather is worn through, it is apparent that they are made from an early precursor to cardboard rather than wood. However, the exposed areas were pasted, stained, and waxed during the binding repairs commissioned when the manuscript was acquired, obscuring [End Page 149] a more detailed assessment of the material.27 Three raised sewing supports are visible from external observation. In all visible aspects, then, the binding conforms to Nicholas Pickwoad's description of inexpensive commercial bookbindings from the eighteenth century.28 Abigail Bainbridge produced a similar binding by following the instructions given by Diderot in his 1751–72 Encyclopédie, ou dictionnaire raisonné des sciences, des arts et des métiers. Other examples provided in Bainbridge's article are comparable, though the covers on these books are made of new, rather than recycled, leather.29
Several considerations prompted our choice of this manuscript as a candidate for our pilot study. Because it is not a manuscript of particular note for historians, we were permitted to work with it in a more intensive way than the British Library team could with Saint Cuthbert's Gospel. Because of its relatively small size, it fit easily inside the µCT scanner, which allowed us to obtain fairly high-resolution scans. (The largest object that could be comfortably scanned in the machine we used is an object about 180 mm wide and/or tall.). The binding on the Western Book of Hours is of a type familiar to students of European binding, so we were relatively confident about what we might expect to see. As a result, we felt able to assess the capabilities of the technique to produce clear and accurate data, even as we tested some hypotheses about what might be going on beneath this book's covers and used µCT's potential to reveal new information about its history. Finally, and most significantly for our purposes, the eighteenth-century binding of this Book of Hours apparently replaced one or more lost earlier structures. We hoped it would serve to answer one of our first questions: is µCT an effective, non-, or minimally destructive method for investigating a book's earlier bindings?
Canon Grandel's Prayer Book was imaged using a Nikon XT H 225 ST µCT scanner housed at the Museum of Ontario Archaeology in London, Ontario. The book was mounted in a foam carrier with the longest side held vertically, as it might be on a modern bookshelf. Two scans were taken, one of the entire book at a resolution of 80 µm voxel size, and the other of just the text block at a resolution of 40 µm voxel size, with smaller voxel size indicating higher-resolution images. Molybdenum is used as a source (target) in radiography because it emits X rays at lower energies than other alternatives do. The lower-energy photons emitted by molybdenum are more likely to interact with less dense materials (for [End Page 150] instance, soft tissue, leading to its use in mammography), so the resultant images exhibit a greater contrast than higher-energy alternatives do (for instance, tungsten, which is used for most osteographic applications in which the priority is a clear and unobscured image of a dense material, bone). We therefore employed a molybdenum target to maximize the contrast between the chemically similar materials that make up the book. Technique specifications were the same for the two scans: 110 kVp (kilovoltage peak), 97 µA, one-second exposure, 1 mm Al filter and 3141 projections for a fifty-three-minute scan. The capture software was X-Tek Inspect-X v 4.4, and the projections were reconstructed using X-Tek CT Pro v 4.4. The volumes were exported to sixteen-bit grayscale TIF slices for viewing in Image-J, or they were viewed in native VGI/VOL format in ORS VisualSI and Dragonfly v 2020.1.30
In preparation for scanning medieval boxwood prayer beads intended for display in the Art Gallery of Ontario's "Small Wonders" exhibition in 2016, Madalena Kozachuk and colleagues demonstrated that the X ray energies encountered during typical laboratory analysis, including µCT imaging, did not damage wooden artifacts. Their results confirm that the analysis undertaken here is a safe, nondestructive approach to the study of organic heritage materials.31 Further work is in process at Western University and the University of Toronto to examine the effects of other X-ray and laser-based analytical techniques on parchment and collagen.
Look-up tables (LUTs) were applied to several of the µCT images presented here in order to produce pseudo-color images. The human eye can only differentiate approximately thirty shades of gray; so when µCT imaging is rendered in a sixteen-bit grayscale format, some features may appear indistinct. However, we can recognize a much broader range of other colors, so LUTs map a predetermined color spectrum onto the gray values assigned in the images to enhance the visibility of features where color variation is subtle.32
Results and Discussion
CT and µCT scans capture data from an object by taking a series of projection images (essentially digital X-rays) that are then algorithmically reconstructed into a volumetric digital model. The model can be viewed in three dimensions or sliced and resliced in any plane. Medical CT scans are generally sliced in three orthogonal planes: the axial (cutting the body into upper and lower portions), the coronal (cutting the body into front and back portions), and the sagittal (cutting the body into [End Page 151] right and left portions). Because a book does not share its anatomy with the human body, we needed to redefine the planes. For this set of experiments, we still refer to the axial plane, which bisects the book horizontally through the spine, cutting it into top and bottom portions. We refer to the plane running parallel to the spine as the spinal plane, and the plane that is parallel to the pages and covers as the paginal plane (fig. 3).
µCT confirmed what visual examination had already indicated. The book is sewn on three supports cognate with the ridges across the outside of the spine. Volumetric images showed us sewing supports made of single lengths of cord, raised up from, rather than recessed into,
[End Page 152]
the book block. Pickwoad associates the use of single sewing supports with the economizing of binding structures after the first quarter of the sixteenth century. They are not commonly found before that point.33 Our imaging from the middle and ends of the cord clarifies its composition: it is roughly 2 mm in diameter and comprises a two-ply strand that has been doubled up and twisted together (fig. 4).
[End Page 153]
Scanning also confirmed that the manuscript's protective boards are made from a precursor to modern cardboard (fig. 5). The lamination visible in axial segments is indicative of the layering of materials used in the production of pasteboard (paste-laminate board). Internal fissures across the board in both axial and spinal planes show points at which the adhesion between layers is beginning to degrade, resulting in separation and air pockets. Such fissures support our identification of the material: breaks are clean, occur in a single plane (between two flat surfaces), and do not produce the "crumbling" effect typical of uneven or pulped materials. At the edges, where the board is vulnerable, wear has caused further separation, creating a frayed or fanned-out appearance. In European bookbinding, pasteboard was largely replaced by pulpboard by the late eighteenth century; its production was less labor intensive as the paper was pulped and dried in a single operation rather than layered by hand.34 So the use of pasteboard in this binding supports the dating evidence laid out above.
From the fifteenth century onward, pasteboard was most commonly made out of paper, which was by then widespread and relatively cheap. When we compared the thickness of the layers within the binding to that of the manuscript's parchment pages, we found the material in the pasteboard to be considerably thinner. Additionally, whereas parchment tends to stiffen and curl with age, paper remains softer and more pliable. The ripples visible along the edges of fissures, then, are more typical of paper. Until the late eighteenth century, paper was made from rags, collected
[End Page 154]
from the surrounding community, soaked, pulped, and dried into sheets.35 This genesis might neatly explain several threads seen between or within the layers in paginal slices through the board, although these could equally have entered the material during processing.
The fabrication of pasteboard was one way to recycle wastepaper, from lists and documents to unwanted books. Indeed, much important information about premodern texts and their contexts is preserved only in the fragments used to make pasteboard or otherwise stiffen later bindings.36 Given this context, it would not be surprising to find written pages among the layers of material that make up the board; in fact, we might expect it. It is notable, then, that such evidence is not immediately apparent. European printing ink was almost always carbon-based, so it is possible that there is writing on this material that contains little to none of the metallic compounds found in writing inks of the same period and easily picked up by X-rays.37 However, the pointlike inclusions dispersed evenly throughout the boards in the images indicate that another substance is present alongside the paper. Another explanation, then, is that the lack of legible writing comes as a side effect of the production process. Eighteenth-century board manufacturing processes are detailed by Joseph Jérôme Lefrançais de Lelande in his Art du cartonnier (1762), and they help us to interpret our results. Unlike in previous centuries, when layers of wastepaper were stuck together using flour-based adhesive (or simply "packed" into bindings to stiffen them), eighteenth-century rag paper was repulped to produce new sheets for use in boards. Pulping proceeded in several stages, which may have varied depending on the scale of the operation. First, the paper was heaped up, soaked, and left to ferment for about a week. Once the material was soft enough, it could be broken up into small pieces by hand with a wooden shovel or iron scraper. If the new paper that made up the board was manufactured at this point, we would expect to find small fragments of written material embedded throughout, resulting in inclusions clustered in specific areas of the scan. But (at least by the 1760s) many French board makers employed horse-driven mills to grind the material into a finer pulp. In this case any remaining iron-gall ink on reused waste would be deposited evenly throughout the resulting board, as we see in our scan. Molds were used to form the boards, which were then couched onto felts to dry. The order of drying, pressing, and pasting material together varied depending on the desired quality of the final product. To make cartons rédoublés (couched laminates or multi-couched boards), wet layers of pulp were couched directly on top of each other and either pressed or dried under their own weight into light and durable boards. [End Page 155] No adhesive was used in this construction method, which explains why we cannot see the same characteristic layers of paste within the boards that we find attaching the leather to their outsides.38
Nevertheless, we know that writing sometimes did survive inside pasteboard and that, when metallic compounds are present in ink, writing can be identified using µCT. One of the most compelling images rendered by our µCT of Canon Grendal's Prayer Book emphasizes the contrast between dense writing and the less dense, parchment substrate. The gold-painted decorative capitals show up with dazzling clarity; but, between these, faint impressions of the rest of the iron-gall ink text are visible, too (fig. 6). The fact that we can identify the separate layers of paper used to produce the pasteboard in our scans suggests that in the future the same techniques used to virtually unwrap the carbonized En-Gedi scroll or unfold sealed premodern letters may facilitate the recovery of lost fragments of texts within bindings without damaging their structures, at least where these were copied in metal-based inks.39
[End Page 156]
In several key places fissures also run vertically through multiple layers within the pasteboard. In the frontal plane some of these breaks are shown to be the holes of bookworms (fig. 7). However, none of the cracks found are easily detectable from external examination without involving some process of disbinding. µCT imaging, we suggest, presents a nondestructive method of diagnostic assessment without the potential for further damage caused by the removal of covers and other binding structures prior to the development of a treatment plan.
In most western European books, pages are produced in conjoined pairs (sometimes as part of larger leaves), which are piled into groups, folded down the middle, and then sewn through the center channel to produce small gatherings. In the axial plane the threads that attach the leaves of the book together are visible where they run up the center of the gatherings (fig. 8). Digital measurements show that such threads have a diameter
[End Page 157]
[End Page 158]
of roughly a third of a millimeter. Given their visibility, in the long term it would be possible to trace the path of a single thread through the volume in order to reconstruct sewing patterns. For the present, orthogonal views provide a rough impression.
When a book is made up of more than one gathering, as most bound books are, one method of connecting the gatherings together is to sew them one after another onto the fixed sewing supports identified above. In the paginal plane the paths of the threads may be traced as they wrap around the sewing supports (fig. 9). Here, the threads run up the center of gatherings, exit to the outside, and pass over the top of the supports and back in, either through the same hole or sometimes through a second hole up to 4 mm beneath the first (two variations of all-along sewing). The same view in the spinal plane seems to suggest that the gatherings are sewn with straight, evenly spaced (as opposed to packed) stitches. The sewing is anchored by a typical row of linked
[End Page 159]
changeover stitches—in which each stitch loops around the thread of the previous gathering and then enters the following one—at either end of the spine (fig. 10). Given that the analysis of sewing patterns was possible in fewer than a third of the Gothic bindings surveyed by Szirmai, leaving him with a sample "too small to allow any definitive conclusion," the ability to examine sewing patterns through µCT could dramatically enhance the field of codicology.40
In books like this one, the sewing supports also provide the primary way of attaching the binding to the book. The ends of cord or leather
[End Page 160]
supports are laced into holes in the boards of the binding to secure them around the book block. Different configurations of holes were used to lace in the sewing supports in different times and places. The manner of lacing in, then, is important evidence for dating a binding and determining its point of origin. However, in European bindings the lacing in is usually partially or wholly obscured by the addition of covers, pastedowns, and flyleaves (as is the case here).
When discussing the lacing in of endbands, Pickwoad records two styles used in abbreviated bindings like this one. The cord might be pushed through holes from the outside to the inside of the board and then back, or the cord might be frayed out and stuck to the outside of the board with adhesive.41 In this case, the book has no endbands. But moving slice by slice through the paginal plane allows us to trace the path of the other sewing supports (fig. 11). Here, supports start by wrapping around the outside of the pasteboard beneath the leather cover. They are then pushed through holes roughly 5 mm from the spine edge. The cords run along the inside of the board, beneath the pastedowns, for approximately 13 mm, before exiting through holes roughly in line with their point of entry. After this the binder has taken the cords off at an angle (between approximately 55 and 75 degrees) where they reenter the boards roughly 8 mm later. The end of the cord is knotted as close to the board as possible and then frayed out on the inside of the board beneath the pastedown. This three-hole lacing-in
[End Page 161]
method is typical of French bookbindings produced in the mid- to late eighteenth century.42 We determined, from the regular deformation of the material around the edge of holes from the outside toward the inside of the board, regardless of the direction of lacing, that all of the holes were prepared prior to lacing in.
Not all of the evidence visible from our scan was immediately meaningful to us. In other words, the application of µCT does not relieve the book historian of the burden of interpretation, nor of its challenges and risks (from confirmation bias to plain error). What µCT offers, consistently, is more evidence of a book's history. Among the data produced by our µCT scans of Western's Book of Hours, approximately two-thirds of the way down the axial stack, was a bright point of dense material. The material produced an X-ray artifact in the fold at the center of a gathering. It remains in shot for 251 slices before disappearing again. When we located the same position in the paginal plane, the dense area turned out to be a thin curved object, possibly a narrow piece of metal (fig. 6). During the first team examination of the scan, we speculated on what this metallic object might be. Our best guess, from its narrow dimensions and the metal's curving shape, is a thin bristle from a wire brush or a metallic thread. The object is not visible in external examinations of the book, which implies either that it is caught deep in the gutter, between two leaves, or that it is lodged outside the book block, between the outer leaves of two different quires, held in by the sewing and then the cover over the spine. Due to the curvature of the pages and the object, our µCT images did not allow us to decide between these two possibilities.
µCT imaging allowed us to vastly expand our observations about the existing binding on Western's prayer book. It also gave us some insight into past interventions in the book's structure. The µCT images, for example, revealed aspects of the manuscript's collation to us that were invisible or puzzling. We could see all of the stubs of the leaves that are thought to have housed the book's miniatures (see detail in fig. 8) in a single image, even when those stubs were hidden from visual observation within the spine fold.43
More importantly for our purposes, µCT exposed four parallel rows of holes, evenly spaced along the spine between the current sewing supports. The holes pierce the central fold of each gathering, marking the now unused sewing stations from a previous binding. We take these to [End Page 162] be impressions of a lost earlier binding structure, a ghost binding visible to us only from the X-ray images. By using a curved plane reformat to flatten the spine along the sewing path—a similar manipulation to those used by dentists when taking panoramic X-rays—we could observe a full set of holes in a single frame (fig. 12). Doing so revealed that each sewing point is made up of a set of two sewing holes, with holes for a linked kettle stitch at either end of the spine. This pattern can be explained if the sewing attached the gatherings to double supports using a two-point stitch. That was a common-enough method of sewing in fifteenth- and sixteenth-century books, but it was more labor intensive than other available methods. Sewing on four double supports also
[End Page 163]
means marginally more investment (of time and perhaps cost) than the current binding was afforded, given its three single supports. The prayer book's original binding, then, was more elaborate, if only in the details of its construction, than was the binding that replaced it.
When viewed in the paginal plane, the direction of the book's earlier sewing is sometimes identifiable. Some of the holes appear to show deformation of up to 1 mm as if punctured by a sharp object. Such deformation was likely caused by the binding needle, either as the gatherings were sewn onto the supports or perhaps before sewing if the gatherings were pierced preemptively to make the process easier. To adjudicate between these scenarios, we would need to undertake a more detailed investigation of all of the empty sewing holes in order to identify patterns. A regular pattern of inward and outward deformation might allow us not only to reconstruct the style of the preexisting supports but also to replicate the process of attaching pages to them. We did not embark on a study of that sort here, but we note it as a potential further use of µCT imaging because sewing styles tell bookbinding experts much about a book's date and locale. Such a study would present a host of new ways to expand the corpus of evidence available for book history more broadly.
As we have already suggested, new evidence will, of course, present new interpretive challenges. Most premodern books have had many bindings, both temporary and (semi)permanent over the course of their history. Canon Grandel's Prayer Book is a case in point. We argue here that the impressions of clasps left on its leather cover mean that the cover was reused. But we were not able to determine whether that cover, or its now-missing clasps, came from the book's earliest binding or an intermediate structure, between the original binding and the present one. Binders often reused pieces of previous bindings to expedite repairs and upgrades: covers, boards, and sewing supports might all be saved and repurposed during rebinding efforts.44 Parchment is a durable material; and if stitch holes were in good repair, then a book could easily be resewn through the same holes (not only incrementally reducing the labor involved but also avoiding further weakening of the fold). When we reconstruct early bindings from the traces revealed by µCT, we must be aware of the multilayered, multitemporal nature of binding history. The evidence revealed by a µCT scan might represent a current binding, a previous one, or separate elements of several bindings. The materials we identify may have been reused in many consecutive structures. While the sewing holes we found in this book match recognizable sewing patterns from the period in which the book was copied, that is not a given. µCT [End Page 164] images of other books might reveal multiple ghost bindings that will need to be interpreted carefully. Yet the layered traces themselves promise a compelling new way for scholars to recall, detect, and describe the diachronic life of a single book—that is, to describe its bibliobiography.
We selected Canon Grandel's Prayer Book, a medieval manuscript in an eighteenth-century rather than a premodern binding, because we were confident that we would understand its structure and recognize its features, even through the unfamiliar medium of µCT. µCT confirmed, extended, and nuanced our knowledge by allowing us to describe the structure and date of the pasteboards, the French style of the lacing in, and the various sewings of the book block. We uncovered evidence of various interventions in this manuscript's history—one or perhaps two rebindings, the excision of several illustrations, a piece of wire—that helped to fill in this modest prayer book's bibliobiography further. µCT revealed a more elaborate structure, predating the current, commercial binding on the book. Our findings add this ghost binding to the limited corpus available for scholarly investigation. Our experiment, therefore, amply demonstrated the potential of µCT to reveal aspects of the diachronic life of a single premodern book. However, µCT did not relieve us, and cannot relieve any other book historian, of the need for interpretation of bibliographic evidence. What µCT offers, consistently, is more of that evidence, and with it a deeper and richer account of a book's history.
Based on our success at imaging this manuscript, we propose several other applications for µCT. As suggested above, we think it is worth investigating whether µCT could uncover text on the leaves layered into early pasteboard bindings. In the past, the only method of recovering such material involved disbinding the book and soaking the boards to separate the leaves within.45 Such an application would build on other scientists' and scholars' work to use X-ray technology to recover text concealed within unopenable books.
Now that we are confident in parsing the structures of a well-documented binding type we can also apply our knowledge to more fragile books as well as those with more complex or more significant histories and those that, because they do not fit the western, Eurocentric bias of much book history, tend not to be subject to much codicological investigation. Some of the authors of this article are pursuing a collaboration with Giovanni Grasselli's Geomechanics Group and the Thomas Fisher Rare Book Library at the University of Toronto, which seeks to use µCT to uncover the history of a premodern Kashmiri paper manuscript [End Page 165] with a Persicate textile and pasteboard binding. With Williams College and Harvard University, we are working on µCT imaging another early modern Kashmiri manuscript—in this case, copied on birchbark. These new µCT experiments will direct attention to neglected, regionally specific binding techniques, offering new evidence of the international contexts for premodern cultural inventiveness and exchange.
Finally, we think that µCT is a valuable tool for guiding collections care in libraries and museums and adding to records of the history of that care. Here, we are guided by Christopher Clarkson's principles of "minimum intervention in [the] treatment of books." As he writes, "for minimum intervention to be practised correctly a damaged item must be very carefully analysed, understood, and assessed before any action is taken."46 Scanning of books at the outset of conservation projects would allow conservators to assess the condition of objects and prepare a treatment plan before taking more invasive steps. Scanning them again at the end would offer a valuable record of the changes made to a book's structure at a particular moment in time.
µCT is already in use to find out what is written on the surfaces of textual objects too fragile to open. Our work on the bindings of old books is, for us, a salient reminder that any heritage object is in some sense unopened, concealing the story of its past within its materials and structure. µCT can offer important new insights into these elements of a book, just as it can reveal an old book's textual secrets. It seems to Mr. Tulliver in The Mill on the Floss that "one mustn't judge by th' outside," but this conclusion does not help him much, for it is still "a puzzlin' world." New techniques for historical study will not answer all the questions we have about the past, but µCT does give us some new pieces for the puzzle.
Contributors: J. D. Sargan drafted the article and (unless otherwise cited) produced the figures. A. Gillespie, Andrew J. Nelson, and Jessica J. Lockhart contributed sections to the draft. Andrew J. Nelson imaged the manuscript. J. D. Sargan and Andrew J. Nelson analyzed the data. D. L. Meert-Williston photographed and examined the manuscript. All of the authors revised the essay critically for important intellectual content, gave final approval to the version to be published, and agreed to be accountable for all aspects of the work.
Funding: This research was funded by the Andrew W. Mellon Foundation and the Leverhulme Trust.
2. Stock, MicroComputed Tomography; Ritman, "Micro-Computed Tomography," 185–208; Singhal et al., "Micro/Nano CT for Visualization," 16–22; du Plessis et al., "Laboratory X-Ray Micro-Computed Tomography," 1–11; Conlogue et al., "Chapter 7. Computed Tomography (CT)," 111–78.
3. Among many other milestones in this transition are Pickwoad, "Interpretation of Bookbinding Structure," 209–49; Hadgraft, "English Fifteenth Century Book Structures"; Jennifer M. Sheppard's privately printed pamphlet, A Guide to the Census of Western Medieval Bookbinding Structures to 1500 in British Libraries, described in Sheppard, "British Medieval Binding Structures," 25–30.
12. Analogous to a pixel in two-dimensional digital images, a voxel is a discrete element of volume or data point in three-dimensional digital space. The voxel size is defined as the size of one side of a pixel on any one plane (slice). In µCT, the depth between slices falls within the range of 1–100 µms; in three-dimensional clinical radiography, minimum slice thickness is between 500–650 µms.
14. Seales et al., "Damage to Discovery," e1601247; Parker et al., "From Invisibility to Readability," e0215775; Fredericks, "Inside Story," https://www.themorgan.org/blog/inside-story-using-x-ray-microtomography-see-hidden-features-manuscript-codex; Wade, "Scanning an Ancient Biblical Text," https://www.nytimes.com/2018/01/05/science/biblical-codes-morgan-library.html. For an overview of projects, see Rosin et al., "Virtual Recovery," art. no. 11901.
19. London ON, Western University, MS Canon Grandel's Prayer Book, flyleaf ii v. King Alfred's Notebook LLC, Enchiridion 6, lot 2, p. 4 suggests that the inscription is legible under ultraviolet light and confirms a connection with a Capuchin house in Lille. However, digital manipulation may indicate a relationship with the Collégiale Church of Saint-Pierre.
21. Calendrier general, 72.
23. Mathon, Blog Fives, http://jean-delaphalecque.simplesite.com/438546767; Lille, Archives Départementales du Nord, 14H 97.
25. Ibid., 3:387, 406.
27. Bookbinder Dan Mezza's assessment and treatment notes from May 2013 are in Western University Special Collections and Archives.
36. For instance, an unknown play attributed to Skelton: Frost and Nash, "Good Order," 483–91. See also Upper, "Red Frisket Sheets," 477–522; Dane, "Pseudo-Donatus," 275–82; and Watson, "Medieval Manuscript Fragments," esp. 65, 69–70. On the practice of reusing bibliographical fragments within bindings, see especially Hellinga, "Fragments Found in Bindings," 13–34; and Ker, Fragments of Medieval Manuscripts.
39. The problem of how to read carbon-based inks, with which most western books are printed and most other books from around the world are handwritten, is one our group plans to investigate further in the future, developing the work of Parker et al., "Invisibility to Readability."
43. de Hamel, Cutting up Manuscripts. Excising valuable images was not an unusual practice among nineteenth-century collectors and remains acceptable among some sellers today. The diaries of the famous Victorian art critic, John Ruskin, offer an insight into his use of medieval manuscript illuminations: he and his children cut them out to make into scrapbooks (Clements and Graham, Introduction to Manuscript Studies, 67). For an account of modern (illicit) examples of the practice, see Hunt, "Missing Miniatures," 97–100.
45. This was the technique used, for example, by Schullian, "Here the Frailest Leaves," 201–17, to retrieve fragments from bindings in the collections of the Armed Forces Medical Library in Washington, DC. The fragments are now known as the Bathtub Collection.