the Oxford Fell Types in contemporary typography - Lin Yangchen
©Lin Yangchen

I was never completely happy with most digital fonts for flowing text, even those like Caslon and Garamond. They are too uniform and disciplined. The letters are sized and aligned to nanometer accuracy, the serifs are exactly the same, it's like laser, like a robot. They are missing one of the endearing characteristics of the old letterpress typefaces: the handcut, manually typeset look. Faithful digital reproductions of original letterpress typefaces are hard to come by; one of them is Founder's Caslon (sample pdf) by Justin Howes. Sadly it is no longer commercially available after he passed away from a heart attack at the age of 41 in 2005.

William Caslon typeface specimen sheet 1741
One of the original specimen sheets of William Caslon, printed in London in 1741.

One day, I decided to visit my university library to see the original specimen sheet printed by William Caslon in 1741, in the library's rare book collection. It was a timeless experience seeing and handling the specimen sheet itself, not a digital scan of it. It was like a live performance of the Vienna Philharmonic at the Musikverein, versus an mp3 on the iPod. Sentiment was awakened again, and I began to look for a way to emulate letterpress printing in my PhD dissertation.

The so-called 'Dutch Fell types' from c. 1670 quickly captured my attention, as they are the primeval typeface from which Caslon was derived and they predate Caslon by about half a century. It turns out that type engineer Igino Marini has embarked on a labour of love in the last decade to faithfully digitize the Fell types, empower it with all the advanced typographic features of OpenType, and make it freely available for the benefit of humanity. More recently it became available in Google Fonts. In fact, you are reading it right now.

Fell Types Specimen 1693 Oxford University Press
The very first specimen book of the Fell types, printed at Oxford University Press in 1693, when J. S. Bach was only 8 years old. Only four copies are known to exist. Furthermore, many of the original punches and matrices, some of which were made of wood, have been lost (Oxford University Press 1900), so these specimens are the only record.

Fell is quite a badass typeface, even by letterpress standards. It has an inconsistent x-height and ragged baseline, and no two serifs are the same. But the person who cast these types was no reckless driver. Inspect the characters and you'll see that underlying all the noise is an exquisite sense of balance not only within any given character, but also in the `flow' from one character to the next, making it quite comfortable to read despite the noise. Indeed, these qualities characterized similar typefaces widely used in northwestern continental Europe during the time, from the Netherlands (see Enschedé 1978) to Denmark (see Nielsen 1934), and would later go on to take the world by storm in Caslon, and they are still celebrated today. They have the spontaneity of a live performance of a great symphony, in which musical instruments played by humans are never perfectly in tune and occasionally quite far out of tune, and with all the harmonic transients at the beginning and end of every note that make the music breathe like a living organism.

If you compare Caslon and the Fell types glyph by glyph, there are mainly superficial differences. But I realized from viewing the blocks of text as a whole that there is a more significant, though subtle, difference at the level of words, lines and paragraphs. At reading-text optical sizes, the Fell types are slightly heavier in the vertical strokes than Caslon, giving the former not only a slightly darker colour on the page but also a slightly more `picket-fence' look. In extreme cases, the `picket-fence' effect can make a typeface harder to read at least to the modern eye (Gothic script comes to mind). The `picket fence' can make the reader feel a little `held back', especially in long passages of text. The abnormally large line spacing required by universities for dissertations only exacerbates the imbalance between vertical and horizontal density.

Fell types English Roman and English Italic 1693 specimen Oxford
The English Roman and English Italic of the Fell types in the original 1693 specimen. Note the slight `picket-fence' effect of the English Roman text as a whole, caused by relatively thick vertical strokes, which I tried to alleviate by increasing the letterspacing (see main text below).

William Caslon 1741 Specimen English Roman
The corresponding English Roman from the Caslon specimen, showing slightly less `picket fencing' because the strokes within individual letters are more uniform in weight.

Igino Marini very kindly contributed his views in 2015: The Fell Types project premise was that of reproducing as much as possible the original types without any other active interpretation. Applied both to the digitized shapes and the global spacing tightness. So I can't say anything about the latter because I didn't decide it. The punchcutter could. Eventually the original samples from Carter and Morison's books were set even tighter. It's obvious that a digital typeface, like Adobe Caslon, conceived in a different century, intended to be used at a different size on different media for a different audience may be somewhat different. The Fell Types were meant to be faithful not pleasant nor modern. — Igino Marini

After careful consideration, I decided to break one of the cardinal rules of typesetting: that one should never change the letterspacing of the lowercase letters, because it has already been perfected by the original creator for optimal appearance and readability. I am perhaps more liberal with this rule, as I believe that typography, like music, is an art that that might be subject to some nuances of interpretation. One should of course be wary of overdoing it. In this case I spaced out all the letters just enough that the vertical and horizontal 'resistances' of the text felt just about equal, using the fontspec package in LaTeX. The difference between the end result and the original is almost unnoticeable at the microscopic level, but at the macroscopic level there is in my eyes a very subtle and positive release of 'pressure' that allows the text to breathe and read more freely. It is certainly not as far-fetched as Leonard Bernstein's controversially slow reading (rehearsal video) of Nimrod from Elgar's Enigma Variations with the BBC Symphony Orchestra.

IM Fell English Pro default letterspacing
Fell English Roman at default letterspacing, exhibiting 'picket fencing' (but less so compared to many other typefaces).

IM Fell English Pro increased letterspacing
The same paragraph with increased letterspacing in LaTeX to reduce `picket fencing'. The difference is hardly noticeable but may reduce eye strain over long periods of continuous reading. Also notice the margin kerning (protrusion) in both cases, achieved using the microtype package in LaTeX.

But it turned out that that wasn't the end of the story. After spacing out the lowercase letters it was noticed that the italic type now looked too spaced out, because italic letters resembling handwriting are predisposed to be somewhat more connected together than roman letters. I therefore compressed the italic words back to their default letterspacing, which looked and read better. Then I noticed that the italics in the page headers, which were rendered at a smaller font size, looked too cramped. A re-examination of the 1693 specimen revealed that italics at smaller font sizes were actually set with a larger letterspacing, improving their readability. So I re-adjusted my headers to space out the italics there until they looked perfect, which required an increase in spacing larger than that applied previously in the main roman text. All in a day's work for typesetting—they had an even harder time positioning every single character individually back then. When I awoke the next morning, the normal-sized italics in the main text looked too tight to my rejuvenated eyes,  so I re-adjusted them again. Now they look right. See my dissertation for the final result.

A word about LaTeX. The Fell font file exploits such advanced configurations of digital font standards that one encounters a series of challenges getting it to typeset in LaTeX and getting the pdf to display and print properly; see this thread on Stack Exchange where fellow users and I discuss some of our problems and solutions.

Adjustments to individual characters

Finally, I manually enhanced some of glyph shapes and kerning in the font, the final step in making my thesis typographically unique and achieving my personal typographic ideal. Changes to the glyph shapes and kerning tables of a font cannot be made in a word processor or Adobe Illustrator; a specialized font editing programme is needed. My top recommendation is FontForge, a free and very powerful full-featured cross-platform font editor. Installing FontForge on OS X is a Unix-style operation (human-understandable instructions), but it pays. With it, you can take a font apart, explore its gears, springs and nuts and sculpt it to perfection like an artisan with a hammer and chisel. If you wish to modify a font, first check that the license allows you to do that for your intended use. The IM Fell fonts came with the SIL Open Font License at the time I made the changes.

FontForge Mac screen
FontForge with IM Fell English Pro.

The IM Fell font is actually not exactly identical to the original Fell types, because it was digitized from a later reproduction of the originals using letters newly cast in the late 19th century, which I found to be slightly different by comparing with the original specimens from 1693 in the library. However, the quest for originality wasn't my main motivation for making changes; it was that a few of the letters in IM Fell had (faithfully reproduced) obvious kinks, presumably introduced during the 19th-century recasting process, that detracted from its readability. The digitized font is actually slightly overweight compared to the original (see illustrations above), but I decided not to try reducing its weight, because there is a real risk of distortion when trying to reduce the thicker parts more than the thinner parts, and the current colour of the text block as a whole does not come across as too dark for reading comfort, somewhat thanks to my increased letterspacing.


1. Removed the overlap between the solidus and the dot of 'i', which can occur occasionally e.g. 'out/in'. Another option is to use a dotless i in a sort of '/i'-ligature—in LaTeX just use {\i} to get the dotless i. But the solidus should ideally not be used in prose.

Before: solidus overlapping with dot of i.

After: overlap removed.

2. Fixed the upwards-protruding upper right serif of letter u. In the original specimen of the Fell types there is no protrusion (see specimen above). I also made the serif horizontally shorter so that it doesn't appear joined to the left-hand stroke, making it more readable.

Letter u before, with protruding right serif.

Letter u after.

3. Made the small-cap C look less like G. In this case the specimen itself looks like G, but it is most likely an unintended aberration caused by imperfect cutting or printing, because all the other C's don't look like that.

Small-cap C before, resembling G.

Small-cap C after.

4. Lengthened the en dash and em dash, which previously looked like the hyphen and en dash respectively. Also made them sleeker so that they blend well into the 'colour' of the text block as a whole.

5. Not modified: overlap of consecutive characters gg. I leave it as it is, although this is physically impossible in letterpress unless cut together, because separating the bottoms would result in too much space between the two characters. Unfortunately there is no example in the specimen book to indicate how this would have been typeset back then. Short of drawing a custom gg ligature, it doesn't look too bad as it is; a little intimacy won't hurt.

Overlapping italic gg (not modified).

6. Finally, it seems that there is no way around this other than avoiding words that did not exist at the time:

In these close-up illustrations of the glyphs one can see how badass the font is with its kinks and nicks. The wood and metal type itself was unlikely to have had so many defects; the roughness of papers made at the time also had a role to play. When printed at normal size on modern-day smooth papers, these kinks and nicks achieve the intended effect of making it look like handmade paper pressed onto handcut letters.


Alas, looking back at it all, it might seem that I have taken things a little too far. There is veritable doubt as to whether a dissertation involving 21st-century computer modeling should resemble a book printed in the Renaissance. Have I broken another cardinal rule in typography, that of using an inappropriate typeface? Here are three possible justifications: (1) no harm echoing the earliest scientific accounts in the Philosophical Transactions of the Royal Society, in which Antonie van Leeuwenhoek in 1676 expounded upon the curiosity of animalcules that has fired the imagination of natural philosophers to this day; (2) an atmosphere of nostalgia for bygone centuries would not be wholly inappropriate in a thesis from a university that is more than 800 years old; (3) an even more extreme precedent for time-juxtaposition can be found in Stanley Kubrick's timeless and profound motion picture 2001: A Space Odyssey. At the beginning of the film, a prehistoric landscape inhabited by ancestral apes is punctuated by the appearance of an unidentified object from a future so distant that it is undecipherable even to the spacefaring civilization of 2001 A.D.

2001: A Space Odyssey dawn of man prehistoric ape
A scene from 2001: A Space Odyssey, a juxtaposition of distant past and distant future even more extreme than pairing a 17th-century typeface and 21st-century computer graphics.

Well, I may end up falling back on good old Adobe Caslon Pro if all this 17th-century letterpress printing ends up freaking people out. The degree committee might even ask me to change the whole thing to Times New Roman…

[Update: I stuck with the Fell types for my dissertation, and the examiners complimented it. See it here.]

Literature cited

Enschedé, C. 1978. Typefoundries in the Netherlands from the Fifteenth to the Nineteenth Century. Stichting Museum Enschedé, Haarlem.

Nielsen, L. 1934. Dansk Typografisk Atlas 1482–1600. J. Jørgensen & Co., København.

Oxford University Press 1900. Notes on a Century of Typography at the University Press, Oxford 1693–1794.
Powered by SmugMug Owner Log In