Archives for category: Typesetting

Originally typecases would all be locally designed. Each printer would bang up a typecase out of bits of wood to suit his own preferences. A different case would be made for different type sizes and different typefaces. Their layout would of course be the same. As time went by layouts would doubtless tend to converge: here’s an 18th century English layout from John Smith: The Printer’s Grammar (1755).

The two cases would be arranged one above and behind the other. The upper case contained Caps and Small Caps and other less frequently used sorts, and this arrangement gave rise of course to our common parlance distinction between upper and lower case letters. Whether the layout of this double frame of typecases from Oxford is the same as Mr Smith’s example or not, we cannot tell.

A composing frame from Oxford

The point of the two part arrangement was to cut down on arm movement in order to increase the speed of typesetting. The size of the space allowed for each sort was determined by the letter’s frequency of use. The most commonly used letters would be placed together in the middle, for the same reason. The upper case tended to be left in alphabetical order. But, lovers of tradition as printers must ever be, Cap J and Cap U followed Z, because in the earliest days of printing J and U were not used by English printers, and were only added on later!

The California job case, which originated in San Francisco, swept America in the mid nineteenth century and got rid of the upper/lower arrangement, putting caps on the right. It was alleged that it cut down ½ mile’s worth of arm travel a day as compared to its predecessor the Italic case layout. I guess experienced compositors had gotten used to having their Js and Us next to Z by then.

Note the spaces, Quad, Em quad and En quad almost next to the Caps, and the 3-em space nearer the middle of the lower case sorts, as it would be what was most frequently used as word space. A 3-em space is, confusingly, not 3 ems wide: three of them together will make up one em.

A visit to the Mackenzie & Harris typefoundery, from which this layout is a keepsake is highly recommended. It is housed in the same building in The Presidio in San Francisco as Arion Press, whose future, consequent upon the retirement of Andrew Hoyem, is we all hope in the safe hands of the Grabhorn Institute.

The concept of typesetting is pretty easily understandable to us, born into an alphabetic culture. There’s such fit between the two that one might almost imagine that the alphabet was invented with hot metal hand setting in mind. But Chinese doesn’t have an alphabet, so you kind of need a bit of type for every word in the dictionary, or at least every word in the work in hand. So you end up with a whole lot of sorts and a terrible navigation task to find the one you need next. As Joseph Needham’s Science and Civilisation in China tells us (Vol. 5, Pt.1) “Because a set of metal type might comprise two to four hundred thousand characters, we can readily comprehend the magnitude of the work and cost involved in printing with movable type. Very often two sets of type, large and small, had to be made to print, for example, text and commentaries . . . The tin types of Foushan even included three sets.”  “The principle of assembling individual characters to compose a piece of text can be traced back many centuries before Christ, as inscriptions on bronze vessels, pottery objects and cast metal seals have made evident, but the use of movable type for printing was not begun until the middle of the +11th century.” [NB: Gutenberg “invented” all this in the middle of the +15th century.]

Although typesetting from individual pieces of type was invented in China, woodblock printing remained the commonest method of print reproduction until the introduction of offset lithography in the late 19th century. Chinese script has always been an artistic as well as a communication medium, and the ability simply to photograph a page of handsome brush-drawn characters and print from that, rather than dizzyingly trying to reproduce their sense in individual types, must have seemed like a liberation to Chinese printers. Wikipedia tells us that to typeset his agricultural gazetteers around the turn of the 13th/14th centuries, Wang Zhen “used revolving tables about 2m in diameter in which the [individual wooden] characters were divided according to the five tones and the rhyme sections according to the official book of rhymes. The characters were all numbered and one man holding the list called out the number to another who would fetch the type.”

An improvement on simple chaos perhaps but not exactly a model of efficiency. Another strike against movable Chinese type was that if a reprint was called for the types (often wooden) would have to be reassembled with the same lengthy process, whereas a book printed by woodblocks could be reprinted just by taking out the blocks and mounting them on the press. There’s a Chinese Blockprinting Museum in Yangzhou, where they run demonstrations of woodblock carving and printing.

Here is a video of members of the Wang clan who still use movable wooden types for their genealogy printing. They still use a rhyming chant to select the individual types. This print survival seems in the process of being moved into a museum setting where it will be preserved in some form. Luckily the commentary is supplemented by subtitles.

If you don’t see a video here, please click on the title of this post in order to view it in your browser.

Here’s a Flickr post with lots of pictures of Rixing Type Foundry in Taipei. (I am, I admit, making an assumption that the pictures all come from the same foundry: I can’t read Chinese.) Atlas ObsuraTyperoom and NeoCha also show some good pictures.

See also Chinese typewriter



To make a typewriter for Chinese along the same lines as the typewriters we all are familiar with would require a keyboard so big that nobody could reach the middle of it. Because each character in Chinese is (by and large) a syllable or a word*, such a keyboard, thinking in terms of an English comparison, would need to have the same number of keys (plus key shift options) as there are entries in the dictionary and then some, as you’d need to conjugate and decline too. The Remington Typewriter Company, founded in 1816, was the early brand leader in “English-language” typewriters. They claimed their machines were universal, but this required them to turn a blind eye to the Chinese language.

In the late 19th and early 20th centuries, many Chinese were determined to overcome the problem and they set about solving the conundrum along three avenues: firstly, restrict the number of characters to the most frequently used 2,000 to 3,000 — surprisingly this actually takes care of a good deal of what anyone might normally need to communicate. (By the 20th century there were over 85,000 characters in the Chinese lexicon.) Then secondly, look at dividing the characters into constituent parts and combining these separate parts to create a variety of characters. This proved almost impossible to work out because of aesthetic concerns about fit and size variation. Thirdly, simplify Chinese script — not something anyone other than the odd enthusiast was prepared to consider.

The solution used in the first working Chinese typewriter was to abandon the idea of a keyboard, and put in its place a rectangular tray carrying little metal characters. The tray, about 18″ x 9″, carried about 2,500 individual characters and offered the chance of substituting some special characters to customize the machine to particular subject matter. The typist moves the tray about to locate the required character, then pushes a lever to cause the mechanism to pick up the sort and bang it against the paper. Chinese typing was always a bit slow. By rearranging the characters in the tray so that characters frequently occurring together were placed adjacent to each other, individuals were able to get their speed up to 3,000 an hour in the revolutionary period. Attempts to standardize this sort of thematic arrangement always failed and by the 1980s manufacturers would supply the tray bed empty, so that each purchaser could load the characters in their own personal layout.

This slightly hazy video shows the system in operation.

If you don’t see a video here, please click on the title of this post in order to view it in your browser.

The adaptation of telegraphy to Chinese took a different tack. There all the characters were assigned a four-digit numerical code, arranged in the sequence of number of strokes, a method often used in China for ranking in the absence of an alphabet. This involved the printing of directories showing the numerical code for each character, which the operator would have to look up. An efficiency break-through came when someone reprinted the book with all the characters beginning with the same two digits printed on one page. This meant that rather than flicking through the pages of the directory the operator could go straight to, say, page 32 to find the character 3261 which he’d just received. Cumbersome, but better than not having access to the telegraph at all. The work was slowed down by the fact that the digits in Morse code are all five clicks long, so it took longer to type a message than if they had had fewer clicks — though of course no character was longer than four digits, twenty clicks.

In 1947 Lin Yutang came up with a working prototype of a typewriter which contained over 8,000 characters, plus a system of partial characters which would enable the typing of every known Chinese character. He named it MingKwai, meaning Clear & Fast. It worked on the basis of a series of 8-sided metal bars, each face engraved with 29 characters. Six of these bars were mounted on a rotating mechanism. There were five more such 6-bar mechanisms which would rotate around one another as well as rotating themselves. The keyboard didn’t activate a character; it directed the machine to the character to be printed. A group of eight characters would be summoned up by two keystrokes; a third stroke using the number keys at the bottom of the keyboard would select the character in that location and print it onto the paper. All was going well when Lin ran out of money just in time for the whole project to be engulfed in the revolution and subsequently the Korean War, making foreign investment impossible to come by. The single prototype machine was apparently chucked out by Mergenthaler Linotype later on. Future archaeologists, please be on the look out. It may be rusting away in the Freshkills landfill.

This information is gleaned from The Chinese Typewriter: A History by Thomas Mullaney (MIT Press, 2017), a strangely fascinating volume which tells the totally unsuspected story of an object one had always thought of as utterly mundane. The book was reviewed in the Times Literary Supplement of 1 June, 2018. You need, I fear, to be a subscriber to read it in full. Professor Mullaney promises us a second volume, continuing the story into the world of computers. One can perhaps sense that the telegraphy trick might become significant here.

The website ozTypewriter gives an extensive well-illustrated account of some Chinese typewriters, including Dr Lin Yutang’s MingKwai machine.


* Wikipedia tells us “Chinese characters represent words of the language using several strategies. A few characters, including some of the most commonly used, were originally pictograms, which depicted the objects denoted, or ideograms, in which meaning was expressed iconically. The vast majority were written using the rebus principle, in which a character for a similarly sounding word was either simply borrowed or (more commonly) extended with a disambiguating semantic marker to form a phono-semantic compound character.”


Nice of Tottenham Hotspur to offer to personalize a shirt for me, but I’d prefer a bit more letterspacing before I’d consider plonking down my £80.

In an ideal world Caps would always be letterspaced — have a little extra space added between the letters, varying depending on the fit of the characters to make the line look even in its “color”. Just look at that IC. You need quite a bit of space between these two to prevent it looking like they are crashing together. Clearly the Ls create their own space, and the counter of the C makes it look OK next to the K. I’d probably want a little letterspacing between the O and the first L, and a bit more between the H and the O. I’m sure this isn’t an option at the Spurs shop — besides I wouldn’t have the side to append my name to Harry Kane’s number.

In hot metal setting, Monotype or hand setting especially, where each letter is a separate piece of metal, letterspacing seems to me to be easy to understand. You just stick a little bit of metal in between the characters so that that H is faced a lilt distance away from the O following. Remember that in order to print a page of metal types every line has to be exactly the same length so that the type can be locked up in a forme so as not to move under the pressure of the press, and this means calculation has constantly to be made. Add space here, take it out there, or break a word, hyphenate it and once more adjust the space. Letterspacing is usually 1/9 of an em, but adjustments can be made by the compositor. Here from Hugh Williamson’s Methods of Book Design is a description of the universe of available spaces: “The set of fixed spaces in metal is related to that of the em of the fount. A hair-space is 1/6 em or a fraction over 16 percent: a thin, 1/5 or 20 percent; a middle or mid, ¼ or 25 percent; a thick, ⅓ or just over 33 percent; and an en or nut, ½ or 50 percent. The em itself, 100 percent, is often called a mutton in conversation, to distinguish it from the en. In most composition, the space between words should not be narrower than thin or wider than nut, and the thick space should be wide enough for the majority of lines.” So for letterspacing that shirt, you might reach for a mixture of hair, thin, and maybe even a mid for that IC crash leaving the 1/9 space between the two Ls and the C and K.

Via Twitter Erik Kwakkel sends these pictures together with the information that before anyone thought of creating a title page at the front of the book, they would give the “publication” details in a colophon at the back. This practice was taken over from the manuscript tradition.

Illustrated is a Bible Commentary which, according to Professor Kwakkel, was printed in Nuremberg in 1487. Printer’s errors seem to have a long and noble tradition: isn’t there a typo in the date? Or is this actually a manuscript created in 1387? Hard to tell without seeing it up close. I couldn’t track it down at the University of Leiden’s library using the call number quoted, Groenh. 014. The rubrication would be done by hand even if the black was printed.

The European Book in the Twelfth Century, edited by Professor Kwakkel and Rodney Thomson, has just been published by Cambridge University Press.


Drop initials always look nice. Well, I like the look at least. Magazine Designing tells us “Drop caps and initials are an effective way of grabbing readers attention because they add personality and visual strength to the page.” To me, they have a sort of old fashioned, quality appearance. We can see an origin in those illuminated and historiated initials in manuscripts.

The Missal of Cardinal Angelo Acciaiuoli. Fitzwilliam Museum


Magazine Designing also tells us that drops dropped out of favor in the early 20th century under the influence of Bauhaus typographical rigor. That may have had something to do with it, but I’d bet that the main reason was economics. Drop initials add cost, and as labor costs went up publishers found themselves less and less willing to pay for “frills” like decent paper, generous margins, good book cloth, footnotes, drop initials etc.. Therefore if you are going to pay for drop initials you probably ought to do them right. Here’s The New Yorker doing it wrong:

Took me a moment or two to figure out that “live” isn’t being used here as an adjective. Here’s Hart’s Rules showing us how it ought to be done.

As you may see, Hart (the Bible of Oxford bookmaking) also disagrees with The New Yorker‘s handling of the open quotation mark.

I would also argue that good book composition manners demand that the rest of the word be set in letterspaced small caps or at least caps. That alone would have helped a little in the “live” confusion.

Adding negative space in hot metal days used to involve getting a saw and cutting out part of the type to allow the rest of the word to tuck in next to the top of the “A”. In modern computer setting it’s much easier — you just have to have your system programmed to apply a rule which you need to define in code. But “Hey — it’s not worth the (tiny) hassle — nobody’ll notice.”



Tables are usually taken for granted. (In this grant we can include those bits of wood on which we rest our books while examining tables within them.) The Oxford English Dictionary gives as its first example of the use of the word in the sense of “a systematic arrangement of words, numbers, symbols etc.” the 11th century (Old) English of Byrhtferð: “Þæra geara getæl hæfð seo tabule þe we amearkian willað”. So the table has been around for a long time. However the scribes may have dealt with tabular material*, it has long been a topic of debate for book compositors, and each printing house would establish house rules for the layout of tables, all with the aim of making the information contained therein as clear and accessible as possible.

Naturally Oxford and Cambridge University Presses have evolved different ways of dealing with the same material. One occasionally imagines them saying “So they do it that way over there. OK, we’ll do it this way here.” The main difference comes down to the head and foot rules where Oxford favors bold or semi-bold rules, while Cambridge goes for a double rule. To my (obviously utterly unprejudiced) eye, the color of the Cambridge version makes it superior. The bold rules clunk a bit as you flip through a book.

Oxford style

Cambridge style

The Chicago Manual of Style rather wanly opts for a single rule at top and bottom, losing any distinction from internal rules.

The parts of a table, all of which will be identified at least in the early going in a full manuscript mark-up, include the stub, which is the list of the elements you’d look up in the table, table number, table head, column heads, spanner rules etc. This picture from Cambridge University Press’ excellent Copy-editing handbook by Judith Butcher, shows some of this.

The use of leader lines (rows of dots) is usually frowned upon in bookwork. Newspapers may routinely use them, but book compositors always tried to work out any problems of the eye jumping from one line to another by the use of spacing, both vertical, between lines, and horizontal, between the  columns.


* Here’s a manuscript page showing a rather fancy table from a manuscript of Ptolemy’s Almagest. The table lists values of arcs and chords of angles. The manuscript’s creation date is uncertain, but majority opinion inclines to the 9th century, with one or two preferring the 7th or 8th centuries.

Photo: Paris, Bibliothèque Nationale, ms. grec 2389, folio 17 recto.

Here’s a typeface recognition game from Better web type.


A five-minute film which makes it all pretty straightforward.

If you see no video here, please click on the title of this post in order to view it in your browser.

We used to spend a lot of time focussing on the quality of the output we’d get from various typesetting systems. You’d get out your loupe and scrutinize raster lines in Pacesetter or Linotron 202 repros, as if something you could only detect under high magnification could ever matter at all. But in the early days of phototypesetting this was the focus of concern: the image produced by a bit of metal type was something we knew. You knew you could rely on that good old analog reality; you could always see it; you could observe by eye and touch how smooth those curves were; you could always feel it as it indented the paper.

From Hugh Williamson: Methods of Book Design, 3rd edn. © Yale University Press 1983

Because a digital image can only be 1 or 0, on or off, black or white, the image maker has to make a decision around those curved edges: is more of my little dot black, or is it mostly white? On a curve this will obviously lead to a jagged edge as the decision goes one way or the other as you move along the slope. But it’s all on such a tiny scale that you really can’t see it happening.

Nobody would waste their time like this any more.  When a book is processed through a modern-day text processing system it remains a digital entity until it reaches the printing press, or just before that when it is used to create a printing plate. There’s not really any output to look at until you see a set of printed sheets, so we’ve just stopped worrying about it. I think this illustrates two things:

  1. If people don’t have enough to do they will worry about needless stuff, and
  2. It is difficult for people used to a tactile process, literally a hands-on-workflow, to repress the urge to touch their work, just to make sure it’s still there.