University of Virginia Library

Mechanized Textual Collation and Recent Designs
by
Gordon Lindstrand

Since Charlton Hinman's perfection of the first mechanical-optical instrument for comparing texts within a single edition-typesetting, the approving response of analytical bibliographers and textual editors has led to two recent developments: the invention of collators with far fewer pretentions than the Hinman machine, thus creating a virtual textual center in the bibliographer's bedroom; and, second, the recognition in the area of textual studies of the need for even more ambitious mechanical, optical, and electronic instruments to assist with the labor of establishing a text. The operation of the Hinman Collator is probably sufficiently well known to exempt it from description here. Its purpose is simply to compare any two copies, presumed to be identical, of a printed, written, or drawn document. For the textual bibliographer the primary use of the machine is for comparison of two copies of a text printed from the same type arrangement — whether by the impressing of standing type, by plate duplication of type,

or by planographically produced type configuration. The Hinman uses the original text, and thereby possesses a distinctly important advantage over more recently developed collators that must employ either microfilm or photocopy-facsimile. This advantage of the Hinman is, significantly, the opportunity to scrutinize the condition of the original document, to determine the actual deterioration of printed symbols, not just the photocopy-facsimile which, however well produced, often gratuitously provides distortions of the type images as well as particles of matter not on the original. Moreover, I believe that many users of collation devices are not sufficiently mindful of the telescopic zoom-in capacity of the Hinman and, as well, the fact that there is little point in zooming in, for example, for an examination of broken type showing up on a photocopy.

Two of the recent additions to the growing list of collators have been devices that compare text prepared on microfilm or by photocopy-facsimile; a third apparatus makes use only of photocopy-facsimile, not film images.[1] In spite of the obvious drawbacks of copy by film and by photocopy-facsimile, these three new collators are valuable scholarly aids for the reason that availability of the many texts often required either for a study of textual corruption or for a critical edition depends in so many cases entirely on reprographical processes.[2] Since my purpose here is to describe still more collating instruments I will assume general knowledge of how the devices already mentioned operate and bring attention only to those features that relate importantly to the objectives of comparing texts.

The disadvantages reported by the developers and users of existing collating machines seem universal to the process. The alignment of the two texts is always difficult, and this problem is certainly aggravated, if not actually created by, the lack of uniformity inherent in the microfilm process — often the source of photocopy-facsimile. It is not surprising that alignment, of actual texts or their photocopies, on the Hinman Collator is both crucial and difficult to achieve. Indeed, because the Hinman machine is a professionally engineered and constructed instrument that must do its job with polished thoroughness and fair reliability, it is necessary to accomplish all adjustments for alignment, focus and for subsequent operation

via the potential of the mechanics of the instrument, thus involving movement vertically and laterally of texts that have been "strapped in" for the journey. An advantage of latecomer rivals to the Hinman is that the copies fed to the device and the parts of the device are always quickly accessible to the operator for easily-made adjustments that are so necessary for constant alignment. But however encouraging the prospects seem for these collators that have recently been brought to our attention, the perspective and suggestion offered at the conclusion of one scholar's description of his machine should be kept in mind (although I must disagree with his statement that individual sight collation is the most accurate method for comparing texts): when using a substitute for the Hinman, and even perhaps when using the Hinman itself, one ought to accept the results with some hesitation and be ready to emend the data with those got from other methods of collation. Further, since each textual chore has its distinct challenges, an exchange of problems and solutions would profit the many researchers now so occupied.[3]

The ultimate development in collators will undoubtedly be electronic and will presumably permit for the first time rapid and accurate cross-edition or multiple-edition (my terms for collations involving comparison of texts not printed from the same type or type-facsimile reproduction source) collation of texts. There now exist prototypes of electronic instruments that scan texts, "recognize" the symbols examined, and compare them if required with other symbols previously stored in a computer memory.[4] But the optical scanner-computer does not yet possess the versatility necessary to accommodate the many type faces and sizes likely to be encountered in books, nor is its cost low enough to make it accessible to the university textual center. This electronic miracle can also compare texts within the transmission of a single edition-typesetting, but the optical scanner will then of course be keyed to a single type fount and the computer memory bank will contain only symbols of that particular type face and size.

There are projects underway now in the application of the computer alone to cross-edition collation, but published results so far do not prove this method, even when available, a satisfactory substitute for visual collation by a team of trained readers. The computer is first supplied — by convenient units of information, such as a line length of verse — with the anchor text (the copy-text, when known at this stage) or text against which all others are to be compared. Each successive text is prepared in a manner after the original on cards (or, conceivably, tape) before being

fed into the computer for comparison. For either purpose, the collation across editions or within editions, the tedious and expensive preparation of computer "messages" (units of text) and the number of stages at which accidental alteration of the text could occur (each stage requiring a proofreading check) make the computer sans optical scanner at the present time a questionable tool for textual editing, except for the very largest and best financed editorial projects. While the computer comparison itself is lightning fast, all is still dependent on the speed and fallibility of the keyboard console operators.[5]

Not only have we lacked an effective and generally available mechanical or electronic aid for cross-edition collation, the cost and accessibility of a thoroughly reliable collating machine, such as the Hinman Collator, for texts from the same parent edition-typesetting have placed this mechanical aid out of reach for the already large population of scholars working on textual projects. A recently published figure indicates that there are only twenty-eight Hinman machines in existence.[6] But there are alternative solutions, as the recent "poor man's" collators demonstrate. When one is faced with individual sight collation the urgency of finding ways of both cutting the labor and increasing the reliability of the collating grows in proportion to the frustrations endured. In desperate need of some form of assistance for the collation of a large number of texts of novels by Joseph Conrad, I developed in 1960 two mechanical-optical instruments for the performance of the two different kinds of collating chores, the reprint comparisons and the cross-edition collation, which I describe below, respectively, as the Mark I and the Mark II Comparators.

My early and elementary experiments with an optical collating apparatus for comparing texts from a single parent edition-typesetting revealed that if each eye viewed one and only one of a pair of texts, presumed to be nearly identical, any differences would be apprehended as the visual sensory apparatus received and mentally compared the two images. The superimposition of images is effected by mental process, and not by the apparatus prior to visual inspection of the texts by an observer. It then remained only to devise a framework and appropriate optical system by which to present the images of the text to an observer. The arrangement of the mechanical and optical components of the instrument is not itself a very delicate process (except to insure correspondence between the two planes of the matter viewed and the viewing optical assembly), and it should be noted that the principle does not require so elaborate and rigid

an optical system as to prohibit quick and easy adjustment for alignment and focus of the texts and their images.

Figure 1 shows the front and end views of one model of the Mark I Comparator. The supporting framework for the optics and the texts is a completely open cabinet to provide access, from both front and rear, for the rapid alignment of texts. The end view shows how the instrument when in use is tilted toward the seated operator to permit comfortable viewing without the introduction of a very elaborate and costly optical system that would "bend" the images of the texts toward the observer. As a consequence of the tilt, it is necessary to introduce movable book mounting platforms that hold the texts or copies of texts being compared. One proceeds through the text first on one side of the leaf and then, after turning the book mount around, on the other; special attachments, illustrated in Figure 1, keep the stock of leaves collated held away from the opened book. The addition of two high intensity lamps for adequate illumination for the texts completes the basic design. While the eyepiece prisms are adjustable only for the short distance of the range of interpupillary separation, the large 45° angle mirrors must be adjustable laterally in order to accommodate variations in the focal lengths of the two texts. In other words, if one text when placed on its supporting platform is closer to the eye than the other text, as occurring most frequently when one text is book copy and the other photocopy-facsimile, the instrument must have the capacity to adjust precisely for the differing distances of the two lines of sight. (The eyes, of course, cannot focus together on objects that are not the same distance — focal length — from the observer.) The instrument should possess a fine adjustment of this kind for the reason that the crude alternative is to move the text having the longer focal length, by piling up supports underneath it, closer to the eye. Finally, the size of the texts to be collated determines the size of the instrument one will need. For the collation of most modern,

machine-printed books, an instrument measuring approximately three feet long and no more than fifteen inches for each of the other two dimensions should be sufficient.

The optical principle by which variants are detected through the Mark I Comparator is crucial in more than one sense. First, the principle must be understood completely in order to appreciate its potential; second, the principle requires, as with most optical instruments, precise positioning of the eyepiece prisms; third, exact correspondence is necessary in the alignment of all reflecting surfaces. The "mechanism" that detects variance between the two images viewed is not the blinking lights system used in the Hinman, but the capacity of the human nervous system to distinguish the characteristics of the separate input of the visual nerves and so determine (1) if the objects in the field are identical, and (2) the amount of physical separation of the objects in the field, thereby giving us our dimensional apprehension or depth perception. Thus, the design, operation, and theory of this instrument differ entirely from their counterparts in the Hinman or any other existing apparatus. The fact that the eyes are receiving different sources of stimuli means automatically that the eyes are not converging. One is actually, by the muscular action of the eyes, looking into infinity instead of converging for focus as one is doing for almost every other visual stimulus-response event.[7] In the actual operation, then, the process amounts to this: if a discrepancy exists between the two texts, however slight the difference, the observer receives the two images and "mentally" perceives that there is a difference. In the case of a textual variant that is of sufficient size to have necessitated a change in the location in one text of several lines of material, until one text had been "evened up" with the other, the effect is one of total confusion for the area on the page following the beginning of the variant: in short, the effect and the variant are inescapable. When the variant is slight or if so much as a very small portion of a single type symbol is missing in one text, the mode of detection is three-dimensional perception. That is, whatever material that is in one text but not in the other is seen as distinctly raised off the surface of the page. Since experimental depth perception is not a familiar phenomenon, the effects described here are difficult to appreciate unless one has experienced

them. If, however, one has ever used a stereopticon or stereoscopic viewer or seen the short-lived "3-D" motion picture innovation, then the effect can at once be understood.

The Mark I Comparator is particularly useful for locating reset lines in a text, thanks to the depth perception our visual sensory apparatus possesses. Whenever a line or part of a line has been reset, whether to accommodate a new textual version or from whatever cause that could lead to an alteration in standing type or plates, the observer witnesses a "hill-valley" effect; that is, the reset material seems to stand out in full three-dimensional effect on a surface containing otherwise flat detail. It is apparent that for the hand-printed book the detection of a reprint, set word for word from an earlier edition, is a relatively easy matter. I am of the opinion that the Mark I Comparator may also have the capacity of distinguishing between two texts identical in their symbols but printed in one case from plates or standing type — that is, by the letterpress process — and in the other from a planographic process. The distinction between the two, and the basis for their differentiation by the instrument, is in the degree to which the printed page has been impressed or "embossed": in the planographic process the inked symbols are merely rolled onto the paper, while in letter-press they are impressed on by actual characters mounted on the printing press bed or cylinder. A fourth general capacity of the Comparator permits discerning between a text printed from standing type and one printed from plates.

There are additional features and characteristics, chiefly advantages, of the Mark I Comparator and its operation. Since the detection of textual variance is dependent upon mental perception, it is obviously necessary that the two sensory stimuli, fed independently to the brain through each eye, be transmitted in equal strength. In other words, one needs nearly perfect, whether natural or corrected, vision or one must install a simple lens system to correct, for example, for one weak eye. The entire instrument while capable of accommodating through quick adjustment texts in a wide range of page sizes is at the same time portable. I took an early model, for instance, to do collating at a distant special collection, a model that more resembled an egg crate than a respectable institutional apparatus. The Mark I Comparator is, for the reasons given above, easier to use and the operation considerably faster than the Hinman machine because it is an easy matter to align pages and to change leaves. The speed and simplicity achieved depend on the ingenuity in the design of the book mounts or cradles. As to the instrument's additional potential, if it becomes necessary to compare microfilm copy an adaptor can be devised that will permit one eye to view a positive microfilm image. The special requirements for this particular operation are a lens system (for the microfilm image only) plus care and uniformity in the microfilming. A shutter arrangement can also be fitted to the instrument should it ever be found advantageous to replace

the original mode of variant detection with the shutter or alternating image principle. Finally, with the reflecting surfaces mounted at appropriate angles, one can arrange the apparatus so as to view the images from directly overhead, or at any angle desired.

Given the correct arrangement of the components I have identified, and the optimum in environmental conditions, a practiced operator can collate printed matter of any historical period (during tests the earliest text used was photocopied pages from the works of Sir Thomas Elyot) more rapidly than is ordinarily possible on the Hinman and, as tests have confirmed, with as high a degree of accuracy as is now possible by any collation method. The Mark I Comparator has been subjected to ambitious testing, including speed-accuracy competition with the Hinman.[8] One of the subjects tested happened to have been at work on the texts of Sir Thomas Elyot for about a year on the Hinman Collator, and reported that the Mark I Comparator was easier to operate, speedier in collation, and less tiring on the eyes. I had always assumed that persons with less than perfect vision and who needed to wear corrective lenses could not discern the effects that inform an observer of the presence of either a variant or a reset line. But the tests of four subjects who wore glasses or contact lenses were especially encouraging. A subject who normally wore glasses for correction of a weakness in one eye did not use his glasses for the test, performed beyond expectations, amused himself picking out flyspecks existing only on one text, and finally volunteered the guess that having one weaker eye may actually have aided in perceiving a textual variant. Other subjects who wore their glasses while being tested reported that the prisms of the eyepiece component had sufficient "eye relief" (distance from lens to eye) to permit an observer to wear glasses in comfort. I save for the last the knowledge that for a particular model of the Mark I Comparator I have used extensively the expense did not exceed one hundred dollars for the optics and the professionally formed aluminum parts. For a professionally designed and manufactured model, one may expect to add from two to three hundred dollars for the labor and additional refinements. I believe that "comparison shopping" among comparable instruments will reveal this sum a surprisingly modest investment for a reliable mechanical collating device.

The second "comparator" I wish to describe is to be used in a cross or multiple-edition collation — that is, when collating texts that are not from the same edition-typesetting. While it is true that only with the perfection of an electronic computer-scanner device will multiple-edition collation emerge from the limbo of tedium and error of sight collation, in the meantime there are certain conveniences that can assist in this most formidable chore of textual studies. I am convinced through experience that

the head-wagging, finger-pointing method of individual sight collation in addition to being the most exasperating kind of ordeal is subject to much error. As an alternative, one can, after devising a system of shorthand designations to accommodate everything appearing on the printed page and to separate homonyms and comparable similarities arising from verbalized symbols, read a text into a tape recorder and then listen to himself as he collates one text after another. I have resorted to this method for a very long work of fiction and then abandoned the idea after discovering that the only truly predictable result was to put the operator to sleep.

A much better approach was to design an instrument that would virtually — that is, optically — juxtapose a segment from one text with the corresponding segment from another. In theory, the idea is promising: the image of a small portion of text matter — preferably a single line length — from one text together with the image of the corresponding passage from a second text are through a system of right-angle prisms placed side by side, or juxtaposed with as little white space between them as possible. The observer will then see the texts through his peripheral vision if he looks steadily at an imaginary line drawn between the two lines of text. A shift in the juxtaposed matter to bring new text into view is accomplished not by optics but by moving the text pages around manually. In practice, however, it is necessary to locate corresponding words from the two texts by eye movement, not by peripheral vision, a routine that often requires the observer to compare a word found in one text at the left margin but at the right margin in the other text. While the comparison of most prose passages may be somewhat tedious, this instrument permits, on the other hand, very rapid and accurate comparison of text containing completed lines between page margins, namely, the text of verse and portions of plays.

Figure 2 illustrates the arrangement of prisms and texts for the Mark II Comparator. The housing for the various components can be conceived from any of a number of design viewpoints. The portions of texts viewed and the juxtaposition they achieve cannot be made apparent except through a depiction such as that in Figure 3. Here the "picture window" that is seen by the operator corresponds to the top flat surface of Prism B in Figure 2, and the thin line drawn through the center is actually the righthand edge of Prism C. This line separates the images from the two texts and corresponds to the imaginary line mentioned above in my theoretical explanation of this instrument. (In the illustration only two juxtaposed lines of sample text are shown: additional text matter in the area would also be visible if the window were not masked to reveal only the center area.) The exact matter viewed through the window can be shifted by moving the texts along their platforms by introducing rails for this purpose. A second set of rails, but for just one of the texts, extending the other direction at a 90° angle will permit the sideways movement of one text so that its words can be matched up to word position in the other text. This refinement effectively minimizes the amount of eye movement necessary in comparing a word located at one margin in one text but at the opposite margin in the other text. Finally, since once again there are two focal lengths to contend with, adjustment for this factor is made by moving Prism A along its plane, closer to or farther away from the observer. As with the Mark I Comparator, a brief study here of the illustrations will reveal the origin and problem of focal length discrepancies arising from the differing distances from the eye of the two portions of text matter viewed.

It is at once obvious from Figure 2 that it is better to work with photocopies since they are easier to position and move about than the actual books. Also, if the books themselves are used it is necessary to shift frequently the prisms (or front surface mirrors, if preferred) to keep both texts in identical focus. Mounting the platforms for the texts on guides or rails permits the operator, after he acquires lost ambidexterity, to move the texts along at the necessarily differing intervals as he drops down

the page of text, line by line. Because an extra set of prisms is introduced into the optical system to allow viewing at a comfortable angle (an innovation that increases the focal length), it is desirable to introduce compensatory magnification into the apparatus, thus making the perceived images larger and bringing the whole performance closer to the phenomenon described above as that experienced when one is using his peripheral vision. Since the laws of optics forbid a magnifier exceeding a four inch or so diameter, and the average line length of a printed page exceeds that measurement, one of the large flat magnifying plastic sheets constructed on the Fresnel lens principle can be used, although there is some attending distortion — produced from the concentric rings of a Fresnel lens — to become accustomed to. Finally, masking or housing the prism system wherever appropriate will keep out all unwanted ambient light from the area of the prisms.

The two Comparators I have described, if constructed with an appreciation for optical precision and the rather disturbing effects experienced when images are not in focus, will provide the user with rapid, accurate, portable, and inexpensive collating instruments. In using these instruments, as well as others recently described to the profession, one risks only the temporary notoriety that is the bane of the innovator.

The Mark I Comparator has recently been redesigned for the purpose of production in quantity, and three separate models are currently proposed, each conforming to the distinct purpose for which the instrument is intended. The redesign and manufacture of the instrument was in response to the specific request by colleagues in textual studies after they had witnessed the performance of the Comparator. The result of this public exposure of the instrument has been the complete reaffirmation of previous test results and the unqualified performance as described in this paper. Additional information on the availability of the Mark I Comparator can be obtained from its designer and developer.

Notes