Faithful revivals from early printed pages

George Horton's picture

I hope to spend next academic year producing a faithful revival of Griffo's last roman in its 1499 state - the font, in adulterated form, of the Hypnerotomachia - and his 1503 italic for Soncino. 'Faithful' here means avoiding the potentially banalising effects of human interpretation; with incunables, it can't really mean the copying of selected printed letter-forms. That worked better for Justin Howes with the relatively limited variation of Caslon's printed letters than it would for the mess of an Aldine page.

So I've written a little program which takes a large number of copies of a letter, normalises their colour, finds the right offset at which to superimpose each on a master-image, blends each with the master, repeats the process with the completed master from the first round as seed-image, and then thresholds the result to give a monochrome image around which to draw an outline. The degree of boldness from inkpress one wants determines the threshold. But I had previously had a very helpful exchange of emails with Sergei Egorov on his own method, by which tracings from a smaller number of characters were superimposed and an outline for the font drawn through the resulting blur. Still another method would be to try to model in software the Aldine printing process from punch to paper, which, run in reverse, could get to the punch - a thing at least of interest, even if an averaged character better represents Griffo's expectation of how the letter would look. The question: which method is best, and why? Would another method be superior?

A 'p' from the 1501 italic for Aldus is shown here, with Sergei's method shown in the second and his end result in the third image. My blended composite is fourth, and final image for tracing is fifth. This p shows, I think, something which I suspect Griffo practiced pretty consistently in his later work: rotation of the counter. I don't have enough data to be sure yet, but it's finding that sort of thing which makes painstakingly literalist revival interesting.


Nick Shinn's picture

Would another method be superior?

I think the best way to do a revival is by eye, with a magnifying glass.
Whatever method of revival you choose, you are making an interpretation, by the mere act of choice. IMO, it is not "human" interpretation that is banalizing, but attempts to automate the process of interpretation.

By using any system of averaging, you are opting for a composite which is softened and blurred, which is counter to the visually arresting effect of a letterpress impression. A flat two-dimensional photo-reproduction can never convey the bite of letterpress into paper, so the digital revivalist has to find some way of compensating, of adding an ingredient. But averaging does the opposite. The end result looks crude and ugly, whereas Griffo's intent and sense of style is clearly apparent in most impressions -- the calligraphic ductus underlying the form -- which averaging obscures.

...even if an averaged character better represents Griffo’s expectation of how the letter would look.

It doesn't. Griffo was well aware that the individual impressions of a particular character would vary, and would have designed his glyphs accordingly, to account for that variation of the final image. Had he been assured of a uniform reproduction of letter shapes, he would have designed them differently. More like Bodoni.

The variety of serif treatment in Incunabula type accomodates the fact that had they been made the same, they would have ended up different, due to the vagaries of the printing process. So the reader's attention would have been drawn to this deviation from the ideal. However, to make them varied to begin with is logical, it disguises the depradations of the printing process.

There is something disconcerting about a page of Poliphilus, especially when it is digitally set. Perhaps not if one only reads it once. But if one is to savor the text by rereading, and if one is reading the typography as well as the content, it palls. Hence Monotype Bembo, the antithesis of banality.

So if you want to use digital technology to reinterpret old letterpress printing, I think you would be better off exploring the possibility of contextual alternates in an OpenType font, to capture the variety of impression, which is an integral part of the original design. I suspect that Incunabula printers used textual context to inform their selection of a particular sort of a letter, and may even have physically "fine tuned" some of them, especially the lower case "r".

The long left side of the main stroke has acquired a texture (of paper?) in the original, but in both the derived versions shown it is straight. This texture is part of the design, like the resonance of the room in which a piece of music is performed. You cannot produce a "faithful" copy without finding some way of paraphrasing this quality of the printing.

Perhaps you could start with an idealized glyph -- clean, sharp and simple outline -- then apply a "letterpress" filter to it, similar to the way that the texture brushes in Illustrator can be applied to a glyph outline. This, and my comment about contextual alternates, is along the lines of your thought "to model in software the Aldine printing process from punch to paper".

George Horton's picture

Thanks Nick, all very interesting.

A flat two-dimensional photo-reproduction can never convey the bite of letterpress into paper, so the digital revivalist has to find some way of compensating

I think this is true: averaged forms should be toughened up by removing the glutinous deposite from inside the crannies and hardening up angles.

Griffo’s intent and sense of style is clearly apparent in most impressions — the calligraphic ductus underlying the form — which averaging obscures.

I think this begs the question rather. The calligraphic ductus is partly your imposition on the actual form, though it's probably the interpretation Griffo wanted his readers to impose: in 1500 departure from calligraphic ductus was thought of as simply erroneous. But Griffo was clever, and made small (in this p's case not so small) deviations from calligraphic ductus, presumably in order to better unite the word-image. My averaging makes this this very clear - perhaps too clear, because the illusion of calligraphy necessary to the total effect is weakened.

Must sleep now, more later.

raph's picture

Very interesting. I think I've written approximately the same program. I've wondered what would happen when applied to very old samples - I know I've been pleased with the results I've gotten from originals in the 100-year old range.

My own approach is simply not to threshold. In my curve editor, I use the grayscale averaged image to trace from. In my opinion, there is valuable information present in the grayscale which is lost in thresholding.

Look closely at two places you have grayness - the top join of the bowl of the 'p' to the vertical stem, and the spur that protrudes to the left. In the first case, it's reasonable to conclude that the grayness is caused by varying amounts of "glutinous deposit". Arguably, the truer shape of the metal can be recovered by tracing towards the light side.

On the spur, the grayness represents the probability of damage to the metal type - much, much higher for delicate projections such as this spur than, say, along the edge of a stem. With this interpretation, it should be clear that the truer shape of the metal is recovered from a trace of the dark side.

So when I'm tracing, I leave the image in grayscale, and make these judgements by eye.

A couple of other notes about my "recipe".

1. Start with a really high resolution scan. I almost invariably use 2400 dpi.

1a. Use as many impressions as humanly possible. I'm astonished by the difference in clarity between a handful and dozens, and again in the jump to hundreds.

2. I do my contrast adjustment after the averaging. There's no need to get the paper white pre-averaging, because the contrast of the paper texture will be dramatically reduced. In my opinion, this increases the chance of being able to see subtle gray variations.

3. After averaging, I sharpen the image pretty significantly. I believe this compensates for the blurring caused by the (inevitable) errors in aligning the individual impressions to the master template.

4. After sharpening, I tend to lighten the image so I can see the subtleties in the darker gray shadings more accurately.

Nick has a point about variation - you're likely to get results similar to what printers would have produced had they combined Griffo's punches with the type founding, paper, and printing technologies of, say, the 1912 ATF specimen book. If Griffo lived 400 years later, he may well have designed his letters differently to suit, but then again, perhaps not.

I disagree that averaging obscures the calligraphic ductus of the original letters. I think it lets you visualize an image much closer to the actual shape of the metal than can be seen in any of the individual impressions. In some ways, I think it's even better to work from the averaged image than if you knew the shape of the metal, because the latter was probably overthinned to compensate for the ink gain on the press.

Note, though, that in saying this I'm assuming that there was one punch, so each individual letter is a fair copy of the same original shape. Averaging together impressions from different punches would, I agree, be a travesty. If I recall Sergei Egorov's research correctly, part of what he found is that, for many letters, there may have been two or three similar, but subtly different, punches. If you carefully classify the impressions so that the vast majority come from the same punch, that should take care of that issue.

A "letterpress filter" for type is a very, very interesting idea. I've been toying with the idea of building such a thing, but the research, both into characterizing the letterpress process, and into getting usable digital outlines of the results, is nontrivial. How strong would be the demand for such a thing?

Anyway, George, very best of luck with your project. It sounds very much worth doing.

John Hudson's picture

If your aim is to capture something of he 'mess of an Aldine page', then I think Nick's suggestion of making multiple versions of each letter and using contextual substitution to vary their use makes a lot of sense. I think the peculiar sensation that I sometimes get when looking at a page of Aldine italic type comes from the tension of variation constrained by the typographic identity of each sort. The italic text seems to hover somewhere between the chirographic and the typographic, in way that the roman doesn't and which later improvements in printing technology gradually diminished and erased.

George Horton's picture

Further replies to Nick:
To make [glyphs] varied to begin with is logical, it disguises the depradations of the printing process.

Even the dullest reader could hardly have been fooled. I doubt that this was the primary motivation, when variety of seriffing has so many other virtues. I think the 1499 caps are the best proof of this: the serifs are designed to scoop white into a convincing counter (L) or eject it (C), thus making perceived counter-size sufficiently similar for the caps to make a readable text face - something not repeated by Garamond and the later tradition, and certainly not by Bodoni!

Texture is part of the design, like the resonance of the room in which a piece of music is performed. You cannot produce a “faithful” copy without finding some way of paraphrasing this quality of the printing.

I could take the averaged letterform, toughen it, and then consistently waver the outline, paraphrasing the texture without simulating the actual and extreme variations of the printed letters.

As for Poliphilus, in 13+ point letterpress on suitable paper the only problems I see are in the consistently inconsistent weights of, for instance, h compared to m, resulting from Monotype's own attempts to paraphrase this texture. In fact I think letterpress Poliphilus is nearly one of the very finest typefaces - letterpress Bembo is the finest of all, but some of its letters are too wide and so too white (above all y) and its caps are a little too Garamondian.

George Horton's picture

Hello Raph,
Thanks for your help, and there goes my patent. Tracing from a sharpened version of the blended master sounds like a good idea. To clarify what you say about tracing the light side and dark side, presumably you're urging that I trace hard against the black at the top join of bowl and stem, but trace around the paler grey of the spur, on the understanding that it has been often blunted by damage. That would make sense - though it would be nice to investigate just how this damage characteristically operated, because I want to avoid unfounded interpretation. And Aldine type-metal was unusually good for the time, though the press-work was of course generally far below the level of a Ratdolt performance.

Failing to normalise the colour in advance means each letter contributing very different amounts to the blend, and also increases errors in superposition. It's a compromise: with these scans I think careful normalising is a good thing. I've tried without, incidentally. I'm careful not to mix punches - and that's something software is very good for. Sergei's conclusion was that mixing of punches was not done in a form consistent enough to be aesthetically motivated - more likely an all hands on deck response to a big job, or large number of concurrent jobs. And the compositors were, according to Aldus himself, a pack of dunces, while nothing but good came from the 'daedaleis Francisci manibus Bononiensis'.

Finally, is there a way to automate the selection from scans of all the letters from each punch? Cutting them out one by one is a bore.

George Horton's picture

As you say, John, the degree of mess varies: in the 1499 Scriptores Astronomici Veteres, which I expect will be my roman source, inkpress is less noisy than in most pages from a Hypnerotomachia, and much less noisy, simply because of font size, than in a page of Aldine italic. Meanwhile the Soncino italic, though approximately the same size as its predecessor, received better treatment than the Aldine: UPenn page. So in neither of my sources will there be the sheer randomness of the Aldine italic, such that using alternative glyphs to represent the same punch might look like affectation, if only because of the obvious artificiality of the simulation. On the other hand some Beowulf-style postscript coding to provide a letterpress filter would be valuable - as it would be, in carefully regulated doses, for almost any text face. Minute variation is good for orientation on a page, and the sense that the reader is making progress through the word-stream.

raph's picture

there goes my patent

Ah well, and the untold fame and fortune that goes along with.

I finally remembered the thread in which I first posted real results from this method: Christian Gothic. In that thread, you'll find a bunch of sample images, a heated discussion about the merits of the averaging technique and the process of designing revivals, and of course the inevitable flamewar in which Hrant labels others as "fascist-apologist".

Finally, is there a way to automate the selection from scans of all the letters from each punch? Cutting them out one by one is a bore.

Yes, I wrote some code to segment the image by bounding boxes, then to do a pairwise correlation between all samples of approximately matching size. I can post my code if you're interested, but it's quite unpolished and often requires manual touch-up of the results.

That Soncino sample is indeed gorgeous. Any way to acquire a higher resolution scan?

Nick Shinn's picture

...attempts to paraphrase this texture
...faithful revival

George, I'm not critical of your project per se (technique), only of your mission statement, and the word "faithful" in particular (philosophy).

Those who keep the faith are still printing letterpress -- fundamentalists. What you are planning is a convincing deception. Nothing wrong with that, as it one of the destinies of new technology to perpetuate the virtues of the old, and mimicry is the sincerest form of respect. But the idea that new tools remove the element of interpretation is mistaken. can’t really mean the copying of selected printed letter-forms.

It's a pity Justin Howes isn't around to explain his method. Copying in this instance is not like duplicating a digital file. He was digitizing an analogue 3D image, for offset reproduction, which requires not just selection, but interpretation. For instance, every digitiser has to decide "how long to make the coastline of Britain" -- and you too are doing that by picking a certain number of specimens to average. The more you choose, the shorter the coastline.

I just had a look at "The Well Made Book", Updike essays, published by Mark Batty, which is set in Founder's Caslon. Read for content, nice. Looked at for typography, one thing that stikes me is the bounce, i.e. the deviation from baseline. Howes has incorporated this quality --- I first noticed that the "o" rides high. That's not mere copying, it's a design specification. En masse, it adds to the personality of the page, but damn my typographic reading, (which must come into play when passages are re-read, certainly in a book for typophiles!), I'm thinking, "why is the 'o' always up? -- contextual alternates could fix that".

It's all about the mise-en-page.

There are two approaches to "redoing Griffo's last roman", engineering or design. The engineer would perhaps provide a tool that gives the font-user sliders to control simulations of:
- alternate character shape (corresponding to variable wear and sorts from different punches)
- paper texture
- ink spread
- bounce

The designer would work from an exemplary specimen (or edit several), choosing specifics for the above variables.

George Horton's picture

OK Nick: the mission statement is essentially an expression of piety. I am not as wholly naive about revival as I may, for brevity's sake, have given you to understand; and I know the complexity of Justin Howes' superb work, about which I recently spoke with a mutual friend.

To be more explicit, and perhaps more honest: I want to produce a set of letters as close as possible to the printed images Griffo while working can most reasonably be conjectured to have predicted, from type freshly cast from the first matrices struck from his new punches, evenly inked, and pressed into paper consistent in texture and thickness with zero bounce (not the same thing as a consistent apparent baseline/x-height - indeed I think Griffo, perhaps following his experience of making Greek type, cut variety in vertical alignment into his latin punches). What I am not trying to do is simulate Aldus' or Soncino's page, which I agree is not remotely possible without (as would be very pleasant) making a separate version for matrix engraving (and even then...). So: I've got to interpret, but I want to avoid selecting specimen letters. Now perhaps Griffo's expectations for his letters were so irreducably plural that modelling them in their most characteristic forms (ie averaged, then drawn rather than traced to take into account damage, clean out inkpress, and harden angles) would be pointless. But would he really, with his graver, have excitedly anticipated all the splodges and blurs given his work by second rate press-work? If so, I will misinterpret him fatally, but I would still like to see what comes out of the other end of this project.

George Horton's picture

Hi Raph, I've just read the thread containing your prior art, and both your method for automating glyph selection and this 'deconvolution' stuff sound interesting; I'm not much of a programmer, so pseudo-code in words of one syllable or Python would be nice, unless you have a windows or linux binary you'd be prepared to share.

I may find my vocation in this forum as a lightning conductor for political attacks - I'm a genuine Scrutonian conservative, something pretty rare in these circles, and for want of which all sorts of mild liberals have had to dress up as fascist apologists.

As for high-res Soncino: if only I had some myself! I might need to take a trip to Pennsylvania: can't find anything in the Bodleian, though All Souls has a Soncino omnilingual thing which might have some italic. (If I'd known sooner, I wouldn't have walked out of the Fellowship exams there in September: they maul with tigers any strangers who get too near their books). One thing I do have is a scan of the not-very-good repro in Mardersteig's Griffo book, provided along with many other goodies by the kindly Sergei, which I'll gladly email you if you'd like.

hrant's picture

Although human interpretation is indeed unavoidable, wanting to tame it is a great and often useful act of humility - very Craft, as opposed to Art. As a result, even though you never -or at least very rarely- want to slavishly trace, good Method (like your technique, and Raph's) is central. Anybody who tells you otherwise simply wants to feel more... relevant than one could ever really be.

> A flat two-dimensional photo-reproduction can
> never convey the bite of letterpress into paper

The two are unrelated. This fact hits home especially when you consider
that digital fonts can be printed lettepress too! In the end, averaging is
not necessarily soft.

> he would have designed them differently. More like Bodoni.

I personally can't see Aldus/Griffo going for that junk at all.
And it's quite untenable to claim to read another designer's
mind, especially when he's so dead.

> contextual alternates in an OpenType font

That cannot substitute for true irregularity, it can only be a caricature of it.
Remember again the words of Von Neumann. The only way to really emulate
the original is through letterpress (although not necessarily metal type).

> Those who keep the faith are still printing letterpress — fundamentalists.

This is untrue. I print letterpress (again, not
necessarily from metal) because there's really
something there to revive. Although I don't
know exactly what it is yet.

> What you are planning is a convincing deception.

Like using OpenType to mimic true variance...

> There are two approaches to “redoing Griffo’s last roman”

1) There are always more ways.
2) Everything exists not in any pure state, but within extremes.


George, I'm certainly not a fan of revivals, especially not literal ones. But when I see somebody applying good craft -as opposed to merely his emotions coupled to a loupe*- to an effort, I dearly want to buttress that against discouragement.

* And why the loupe, really? :-/


hrant's picture

Oh, and Raph, yet again: "fascism-apologist".
Huge difference. But it's OK, I won't ask for
an apology.


Nick Shinn's picture

I dearly want to buttress that against discouragement. :-/

I wasn't critical of the project, only that expression of piety which was the mission statement.

And why the loupe...?

(A magnifying glass actually, which may be used with two eyes)

For the same reason one would draw from life, rather than a photograph.
It was in answer to the question "would another method be superior?"
The context was about being faithful.
If one is attempting to make a two dimensional version of a three dimensional artefact, why not work directly from the three-dimensional impression?
It seems to me that the more intermediate elements that are removed, the more opportunity for the process to be faithful to the intentions of the original type designer. opposed to merely his emotions coupled to a loupe*

Why do you consider the more direct approach, without tracing over a simulacrum, more emotional?
I would say the opposite is true. The conscious effort required to work from life is more difficult than working from a tracing; one becomes immediately aware of how incorrect one's initial, spontaneous drawing is, and works to modify that, in order to get one's glyphs to look right. So the repetititve process of drawing, comparing, and redrawing is one of suppressing emotion. It gives a better understanding of the original, of why the forms were made the way they were, and of one's preconceptions. This enables one to better address the grey areas that are open to interpretation (caused by loss of resolution between metal and impression, and variablitity of impression). It's here that interpretation is unavoidable, and the draughtsman's personality may be expressed decisively, because he has acquired a knowledge of the original designer's intentions by going through a similar working process of trying to get all the design elements to come together as a balanced whole.

hrant's picture

> I wasn’t critical of the project, only that
> expression of piety which was the mission statement.

It seems to me -through both this thread and Raph's- that you're critical of the method - in fact I think Method in general. And this project being centered on Method, I think you're critical of a lot more than a supposedly tangential "expression". I don't want you to hold back the criticality - that would be unnatural; but I in turn need to counter it when it shows up.

> For the same reason one would draw from life, rather than a photograph.

But I'm saying why bother with magnification.
Scale affects the essence of something, and it
seems to me that shunning George's (and Raph's)
methods requires shunning the magnifier as well.

> the more intermediate elements that are removed, the
> more opportunity for the process to be faithful to
> the intentions of the original type designer.

If that's true, then the magnifier's gotta go.

> Why do you consider the more direct approach,
> without tracing over a simulacrum, more emotional?

Because it assumes the reviver is infinitely gifted.

I would say that shunning Method is emotional. Don't get me wrong, I think emotion is indispensable in anything a human does, but wallowing in it, for example to the point of shunning Method, is anti-Craft.

> It gives a better understanding of the original

I don't think so. I think mostly it causes the reviver to express himself more, to tap what's in him more than the source; this confuses the -intended- results. Unless one's intent isn't to do a literal revival - which is great by me, but not the same thing, and not what George wants. I find that the best crits are the ones that take into account the creator's highly personal desires - not those of the person giving the crit. For example, I don't have to like humanist sans fonts to be able to make Ricardo feel like I helped him improve Lisboa. In giving a crit, I have to separate emotion and method, and focus on the latter. Trying to get people to make fonts I would make just makes everybody miserable. Like when I once showed a "progressive" design to a Famous Designer, and I told him the intent is such-and-such, and he come (sic) back to me with a ton of details of how to make it all homogenous like him (sic) own fonts, that's just a waste of everybody's time (not to mention arrogant).

Basically I guess I don't think anybody (short of a medium :-) is capable of subverting his own preconceptions to allow what was in the mind of another person to surface - especially not centuries apart.


hrant's picture

What can [re]surface though is the outward forms.
Which is exactly where Method helps.


Nick Shinn's picture

It seems to me -through both this thread and Raph’s- that you’re critical of the method - in fact I think Method in general.

George wondered if there was a better way to be faithful to Griffo's last roman than averaging. I suggested a magnifying glass, and contextual alternates, which to me are better methods than averaging. But as it transpired, I was thinking more of the effect of the Aldine page, and he of the type. He clarified his position as an interpreter, rather than a simulator, which is fine by me. End of that part of the thread.

But I’m saying why bother with magnification.

My eyes need a little help.
The magnifying glass is far less of a mediation than a scan.

Because it assumes the reviver is infinitely gifted.

Not at all. The reviver can never be photo-exact, but that's not the point, as such exactitude can't cross the barrier between the imperfect, varying impression, and the ideal piece of type. The reviver can only be relatively gifted, compared to the designer.

I think emotion is indispensable in anything a human does, but wallowing in it, for example to the point of shunning method, is anti-Craft.

I don't think you got my point about drawing from life being a process of un-learning emotion. The "magnifying glass" method is this: Look at the original specimen(s) through the lens, draw part of the glyph (with beziers), look at the original again and refine the glyph, as many times as necessary. Then do a print-out of the same text as the specimen at the same size, and compare them. Then go back to the glass and the beziers. And so on.

Trying to get people to make fonts I would make just makes everybody miserable.

I would hope that by answering my criticisms, George could be a little clearer in his own mind of what he's doing and why. Especially as this is an MA project, and he can expect some harsh crits there. I also made a number of what I consider to be constructive contributions, rather than criticisms, such as the contextual alternates for varying wobble thing, and the outline texture filter idea. I've used the outline-filter technique, but not the contextual.

Basically I guess I don’t think anybody (short of a medium :-) is capable of subverting his own preconceptions to allow what was in the mind of another person to surface - especially not centuries apart.

But don't you feel that you have a different understanding and greater insight into why Baskerville's type is the way it is, now that you have been working on a revival? And why wouldn't such an understanding correspond to what JB had in mind?

raph's picture

Arggh. I wrote a detailed response, but the flaky internet in the United lounge ate it. In brief:

Hrant: sorry for the innaccurate characterization.

George: I'll be happy to send you the code. I'm on the road at the moment, and would like to package it up with instructions and so on as soon as I get home. Bug me if I let it slip my mind.

Thread: the averaging technique is a tool. Having more and better tools avaiable is a Good Thing. Whether typographers prefer reproductions faithful in detail or modern interpretations inspired by historical models is up to them. In either case, the averaging technique can be very useful for visualizing much more clearly what the original type designers did.

hrant's picture

> My eyes need a little help.

And those of Griffo's readers didn't?

Nick, you claim your methods allow you to get inside Griffo's head.
I guess I can't believe you.

> ... why wouldn’t such an understanding correspond to what JB had in mind?

It might. But the chances are slim enough that my own ideas are sure to overpower what I think JB might have been thinking if I look too far beyond the forms themselves. Thought patterns are not some mathematically pure, eternal things; they're entirely individual, highly dynamic and contextual to boot; there's no way a 21st century Armo is going to get inside an 18th century Brit's head (at least not enough to make type like him).

I think that in a revival it's important to distill the essence of a design, which comes from the forms, not what happened to be in the original designer's head. For one thing, stuff in people's heads changes. Should I revive the JB on the morning he decided to get into typography? Or the JB the night he finished printing his Virgil? Or maybe the JB on the verge of giving up on typography in disgust? It's hopeless to approach it that way. So for example when I'm making the British Pound character, I can look at what JB made and say "that was wrong", and make one that fits the design, as opposed to JB's perceived mindframe, better. Also, a key difference here is that my revival is not literal.

Raph, I know how it feels to lose a long post - sorry to hear it. More than the lost content I think it's the combination of improvidence (for not copy-pasting it periodically) and guilt (for rambling on so much in the first place) that makes it so frustrating, and the ensuing decision of how far to take the replacement post so unnerving.

> the averaging technique is a tool.

One problem is that powerful tools can be overly challenging, even scary.
Like how I'm much less inclined to learn now software these days.


John Hudson's picture

I'm sort of with Nick on this one, at least insofar as I consider the human eye and brain to be a powerful analytical tool. I find it bizarre that 'emotional' and 'expressive' are instantly assumed to be the predominant characteristics of human interpretation, as if we are somehow unable to actually see what we see and perform a rational analysis, but must rely on machines to get past our overriding emotional and expressive faculties. This is bollocks. I also think Hrant is wrong to attempt to try to frame this in terms of his favourite 'art vs craft' dichotomy, as if craft cannot come from the eye and hand: where else has it come from for the millennia of human making? We have an analytical faculty and the ability to make rational judgements regarding what we observe. Among other things, we have an obligation to employ this faculty critically in considering the value of methods and tools. Looking at the images in the first post of this thread, the first thing I immediately see is that all of the derived forms based on averaging have a large black bump in the lower left, while the one photographic image shows something quite different; this seems to me to be a problem of method, because the averaging does not take into account relative lightness and darkness of the type image. Similarly, in the photographic image, there is a clear but very faint indication of the full extent of the terminal of the upper bowl where it crosses the stem, and this is completely lost in the derived shapes, presumably because it was too faint to be preserved in the threshold conversion. So these are just two places in a single letter where the human eye tells us more than the proposed methodology and tools. This is not to say that the methodology and tools are without merit or promise but, ironically, that they must be subject to the same critical and interpretative analysis that we can also apply to the subject.

hrant's picture

> as if craft cannot come from the eye and hand

It's certainly not that it can't come from there, it's that it can come
from elsewhere as well; shunning other sources is what's anti-Craft.

> Looking at the images in the first post of this thread ...

Looking at them with magnification.
With a method not unlike George and Raph's averaging.

> they must be subject to the same critical and interpretative
> analysis that we can also apply to the subject.

Totally. And maybe it's this extra burden of thought that
causes some people to avoid new/elaborate methods outright.


Nick Shinn's picture

Right on, John. And good eye!

George, I seem to recall that in the original Raph thread, I suggested that a good (and scientific) way to test the veracity of the averaging technique would be to recreate it now. ie take a typeface where you actually have a clean master, then produce a letterpress page, average the impressions, and see what divergence there is between the result and the original. Are you up for it?

hrant's picture

The problem of course is that evaluating the veracity of something requires much objectivity and humility. People will see what they want. Some people more than others. Ergo: it's not a test that can really work (at least not in this case).


John Hudson's picture

Hrant, you seem to have seen what you want to see even before you have seen anything :)

I think Nick's proposed test has a lot of value, even though I consider 'veracity' in this context to be a value judgement rather than an objective proof. But that's okay because I'm not expecting a proof result. Nick's test would provide insight into what the averaging methodology produces relative to a known form. This is valuable information, regardless of the conclusion that one draws from it or the relevance it might have to a particular goal. Obviously if your interest is in reproducing the letterpress printed appearance of a typeface, rather than the forms of the type itself, then what Nick's test would show is not directly relevant. On the other hand, if that is your goal, perhaps you should be making a full colour, high res Photofont, which would capture the paler artifacts of the letterforms in the printing.

hrant's picture

> I’m not expecting a proof result.

Except of course when the person making a claim is me.

(More on this soon.)


John Hudson's picture

No, Hrant, I have not asked for a proof result from you, I've asked for some kind of result, some kind of evidence, some quantifiable indicator, and only when you have made specific claims about reading performance improvements based on your theories. If I have ever used the word 'proof' in that context, I apologise, because I've spent enough time around philosophers to know better. What I'm interested in is evidence to support ideas, results that indicate whether a theory or a practice does what is claimed for it. I'm interested in information that extends our knowledge. And yes, in the absence of such information, I may favour my own prejudices, ideas or theories against your prejudices, ideas or theories. The expression of the latter has not convinced me, so I'm left wanting evidence. Sorry.

hrant's picture

John, I would get into a detailed rebuttal of what you wrote,
but I think the problem is bigger than you, so doing so would
simply distract from real, broad improvement.


There's entirely too much personality-conflict-determined discourse on
Typophile (and, to be fair, most elsewhere). This drains much useful
energy from high-quality typographic discussion. For example, why do
I have to be the only person to counter Nick when he makes a claim
about the spacing of italics that's false in so many ways? Guys, if you
want me to lay off Nick (although, like I've said, it's not him, it's his
ideas), or whoever, a great way is to get off your finials and do some
of the countering instead. I think in the past (and on other lists) we've
had a lot more ego-sidelining passion for objective exploration of type.
Isn't it a beautiful thing, to strongly counter something you know is
wrong no matter who happens to have said it? Well, I think it is.

You're being civil? Civility is one thing, but too many of you guys seem
to be taking the "if you don't have anything nice to say then don't say
anything at all" proverb to a whole new level. Except of course when it
comes to Hrant, but that's OK because he can handle it, because "what
goes around comes around" or whatnot. Well, the result is just a bunch
of people yammering away, droning on about what they happen to believe,
throwing their weight around in an insular sandbox, not really interested
in changing (neither themselves nor others). This kills theoretical progress,
cripples cultural relevance, and really does start turning Typophile into
a cocktail party. Largely useless; and boring as hell.

People, come on - if you love type, fight for it a bit!


enne_son's picture

George, I think there is real value in mechanizing back-propogation from letterpress impressions. And even in deciding on the best method of back-propogation Nick's iterative process of adjusting the thresholding (or whatever) protocol in dialogue with interpretive inspection is not only indispensible, it is intrinsic. But considering ductus and writing-book precedants must have been at least a point of reference for Griffo, wouldn't it also make sense to approach things from the other side (ductus) and see where the two meet.

Approaching it from the other side would be "calculating systematic norm-violation" to swept-object routines.

Finding where mechanized back-propogation and swept-object norm-violation meet might tell us something of what Griffo intended.

hrant's picture

It's one thing to evaluate a method in order to improve it, it's quite another to suggest evaluation with the intent of eventually rejecting the method outright. The decisions have already been made - and that's part of the problem I mention above.


Nick Shinn's picture

There’s entirely too much personality-conflict-determined discourse on

Thank-you for toning things down recently.

For example, why do I have to be the only person to counter Nick when he makes a claim about the spacing of italics that’s false in so many ways?

They probably agree with my observation. Karsten said "Yes!" to it.

hrant's picture

> Thank-you for toning things down recently.

I can't know if you actually mean that.
Not that it matters much if I know I guess.

> They probably agree with my observation.

Some of them I'm sure do. But the problem is that I've been paying attention; and I know that some other people don't. If you can figure out the intentions of a 300-year-dead guy by looking at some of his glyphs, then I can certainly do the same with people who have explicitly expressed their ideas in writing in front of me for years.

The human complaint/praise parameters are strange: when it comes to people they don't know, people only write to complain; but when it comes to people they feel even a slight friendship with, they can only get themselves to be nice. To me, that's not real friendship.


Nick Shinn's picture

I can’t know if you actually mean that.

I think so. At least, my nipples were erect when I typed it.

If you can figure out the intentions of a 300-year-dead guy by looking at some of his glyphs,

Not by looking, but by going through a similar design process to what he did.
That would be making the glyph shapes from scratch, rather than tracing.
Pre-photography and pantographic punch-cutter, those guys didn't trace over scans.
Was that Morris' epiphany, on seeing Walker's slide-show of Jenson type?
Was Morris the first to use the "tracing over photo" technique? -- One would think so, given the crude nature of the result, and in fact many of the faces of the years immediately after LB Benton's punchcutter were a bit clumsy, as would be expected of a new technology.


Also, perhaps people are reluctant to disagree with me because of my scary avatar.

hrant's picture

> Not by looking, but by going through a similar design process to what he did.

Well, this is all tangential to my real point above, but:

> making the glyph shapes from scratch, rather than tracing.

The mistake you're making is that averaging necessarily leads
to tracing. I never trace - not even my own drawings [slavishly].

> those guys didn’t trace over scans.

Simply because they didn't have the technology!

Hmmm, what do you think of Bruce Rogers and Centaur?
Rogers doesn't have a scary avatar so feel free to be straight.


John Hudson's picture

Hrant, what I understood Nick to be saying over in the italic thread was simply that italics can be better fitted in that they may require far fewer kerning pairs than romans. This seems to be to be self-fulfilling: if an italic has fewer kerning pairs than the associated roman, it meets the criteria for 'better fitted' that Nick is using. So what is to counter? He didn't say that this was good for readability, in fact he explicitly stated that it might not be a benefit. It is simply an observation about the way italic letters fit together as compared to the way roman letters fit together. If you consider even one factor affecting the fitting of italics vs romans, that the typical italic a is a cognate form with d and q, then you have identified a significant reason for the need for fewer kerning pairs. If you consider also that italic v and w and sometimes y often have convex sides, you have another why fewer kerning pairs would be needed. So I really don't see anything to counter in Nick's observation.

If you want an example of me spending a long time trying to change Nick's mind about something, go read the threads related to Greek casing and tonos angle.

If I've seemed to spend more time hammering at you over the years, it is because I've had so little success getting you to change your mind even about obvious things like the distinction between visual rhythm and temporal rhythem, even when I provide detailed illustrations and demonstrations. It has nothing to do with your being 'able to handle it' -- there are stronger souls than both of us in this forum --, but everything to do with your being stubborn in your mistakes. With most people, if I criticise or point out flaws in their reasoning, they respond directly to the points I raise and the conversation progresses, whether they accept my criticisms or disagree with them. You tend to just keep repeating the original statement or claim, and ignore the points that have been raised against it. So you come in for more and repeated persuasion, while conversations with other people proceed with more fluidity and less repetition.

hrant's picture

John, I wasn't necessarily talking [just] about you.
Although I certainly do think you go out of your way
to counter me more than most/most. It all started in
Boston in '99, when I pointed out that a certain nose
is too big...

And I think you're not getting my point about italics.
Please check the other thread for an elaboration.

> it is because I’ve had so little success getting you to change your mind

On the contrary.
Which of course doesn't mean you should expect me to change my mind about everything you say; like the rhythm business is just hopeless - no matter how many illustrations or whatnot you provide (although it's flattering that you put in the effort - thank you for that). And I'm not the only person who thinks it's contrived. I guess it's just hard to give up a term so dearly used for so long - and you are fond of terminology [for its own sake] aren't you?

And John, what have I ever convinced you of?
I do manage to convince all kinds of other people about
various things, so I have to think the problem isn't me.

Also, I think you're mistaking civility in discourse with discoursive progress, and giving yourself too much credit for changing minds [other than mine]. You know when I know for sure that I've changed a person's mind? When they tell me I did. Like when Colin Banks did so in Leipzig, about blackletter. But don't try that with those who think they've arrived (especially when they don't admit that [as a] limitation) because all you'll get is deflections embalmed in tact.


Nick Shinn's picture

If you want an example of me spending a long time trying to change Nick’s mind about something, go read the threads related to Greek casing and tonos angle.

And thanks for that. I thought those threads had massive "discoursive progress", otherwise I don't think you would have been able to convince me to have a tonos angle system different than Adobe.

hrant's picture

Ripples on the surface.


Nick Shinn's picture

Ripples on the surface.

To me, the accents in traditional Greek seem more like a flock of birds in a forest.

John Hudson's picture

But I wasn't just talking about changing minds, I was talking about progressive discourse through direct engagement with criticism. So even if someone disagrees with my criticisms, he or she can further the conversation by responding to them directly. But you tend just to repeat your original assertion. You have not once even tried to explain why what everyone else calls rhythm is not rhythm, and your rejection of that term is based in the very mistake of associating it with flow, so the rejection is begging the question. You keep saying that rhythm implies flow and therefore there is no rhythm. But I've been demonstrating precisely that rhythm does not imply flow -- indeed, flow implies lack of rhythm --, but refers to the spacial relationship of letters within words. I come back to this again and again because I think your rejection of the term is unfortunate, because this spacial relationship is crucial to bouma construction and good notan. At its simplest level, the distinction between the rhythm of letters within words and the rhythm of word spacing (which you have elsewhere insisted should be even, i.e. rhythmic) is key to identifying what constitutes a word. But if word recognition does indeed involve bouma recognition, then regular, i.e. rhythmic, spacial relationships are also important at the sub-word level because it is only in these relationships that role architecture can define recognisable shapes larger than a single letter.

It seems to me, as I write this, that we have spent a long time talking at cross purposes, because you've jumped on the word rhythm with the statement that 'There is no rhythm in reading because there is no flow'. We don't disagree about their being no flow (although, discounting regressions, saccades have an average length, and when they skip words they do so in a predictable fashion, and periodic pulse is the very stuff of rhythm; in fact, if there were smooth flow then that would really be a lack of rhythm). But has anyone actually been talking about rhythm in terms of reading physiology? No, we've been talking about rhythm in terms of spacial relationships of letters in word formation. And my 'contrived' illustrations, have shown how word formation falls apart if you mess with the rhythm of the letter relationships. I've never understood why you have taken so strongly against this commonly understandable and accurate terminology. As soon as you recognise that there needs to be a relationship between the white space inside a letter and the white space between the letters, you are moving in the area of rhythm. This is what rhythm is: a proportional relationship. If you are not going to call that rhythm, then you'll have to come up with another term, because the phenomenon is real, demonstrable, and important to the matters that concern us both.

Give some serious thought to the nature of proportional spacial relationships in the way letters form words, and tell me why it doesn't make sense to talk about rhythm in letter spacing. I've adopted 'notan' and even 'bouma', and I think you should reconsider a term that accurately describes a phenomenon that we need to be able to talk about. And that's not terminology for its own sake.

hrant's picture

> I’ve been demonstrating precisely that rhythm does not imply flow

Why is it so hard to accept that I think you haven't?

I've said this before: to me we're talking about Pattern, and
there's no reason to encumber our terminology, especially not
with such a misleadingly overloaded term. Just because I repeat
that every time doesn't make it false.


John Hudson's picture

Why is it so hard to accept that I think you haven’t?

Because you have never once addressed the specific points of the argument. This is what I'm saying: you don't engage with the criticism, you don't make counter demonstrations, you just keep repeating your original assertion.

We agree on pattern, and I have explained why 'rhythmic' is an accurate description of the kind of pattern. It isn't an 'overloaded term', it is just a common and accurate term. By itself, the word pattern, doesn't tell us what we need to know, since patterns can take many different forms, requiring only repetition to qualify as patterns. What matters is the spacial relationship within the pattern, which we have a need to describe.

Just because I repeat that every time doesn’t make it false.

The repetition doesn't make it false, but it does nothing to demonstrate why it might be true. You just don't make any counter demonstration at all.

enne_son's picture

One could also argue that 'pattern' is a misleadingly overloaded term. When I talk of 'a pattern of salient whites' in a bounded map, I don't mean pattern in the sense that a tartan has or is a pattern. If I did, you would have to say an orthographically meaningful cluster of letters doesn't have a pattern.

A literal notion of flow is not intrinsic to the uses of the term 'rhythm' that we invoke when applying it in the context of descriptions of visual material. In it's deployment of salient whites and vertical blacks a normatively spaced cluster of letters has a phasal structure of discrete pulses (like the discrete pulses of consonantal sound in speach) that hovers around a phasal mean. When we talk of rhythmn in optical-grammatical contexts, it is conventionally that characteristic, or something like it, we wish to evoke. When I hear talk of rhythmn in optical-grammatical contexts, it is typically that characteristic, or something like it, that comes to mind.

It seems to me that Hrant's misgivings about rhythm have to do with the fact that in visual wordform resolution role-architectural subcomponents of a bounded map are not processed sequentially but parallelistically, and sequential processing associations are hard to supress, because of how we learn about rhythm.

Please pardon the intrusion of my lingo in this thread, but what I'm trying to convey with it seems to need to be gotten out of the way or cleared up.

George Horton's picture

Hi Nick,
Thanks again. Looking with a magnifying glass would allow access, as you imply, to information that scans cannot provide - the shape of the metal's indentation into paper, as opposed to ink absorption. Unfortunately there is no way to measure these indentations accurately enough to inform, except in very general terms (like "ink tends to overextend indentation at this point on that letter"), the positioning of digital outlines, once one accepts the aspiration to preserve intentions on Griffo's part of which one is not aware. As Hrant says of Baskerville, Griffo's intentions will have altered during his work - that's consciousness for you - making any theory of his intent reductive.

I accept that I would learn more about type design from your method than from mine (in whatever its final form is) but that's not the primary goal here. My final result should be something that I can learn from - and if I've done my work properly then its interpretation should be interpretation of Griffo's letter design, and so should remain arguable.

A further note on "bite": the swerves in direction of interior and exterior outlines will in my version be much more similar to those of the punches than in most instances from the original 3D page; since a good deal of bite comes from these outlines working against one another, as they do in Griffo's but not in, say, Garamond's work, some of the bite that comes from the 3D impression will be made up for naturally (though I wouldn't begin to pretend that it's a tradeoff any sensible typographer would welcome).

I'm very interested in your outline filtering - how did you do it? It'd be best to apply it to the PDF rather than the beziers, of course.

As for a trial of averaging from known punches, one would have to ensure that the process of printing was very similar to Aldus' for the results to be useful. That means matrix engraving, type cast from historical alloy, repeated impression at slightly varying angles and depths onto uneven rag paper, a small amount of ink of totally different consistency and composition to the modern stuff... Hard, as well as time-consuming and expensive, but not quite impossible: I think that we (not I!) know a surprising amount about Aldus' methods. I agree that it would be worthwhile. Let me ponder - but probably No, and certainly Not yet. Good doctorate though!

George Horton's picture

Thanks for your buttressing, Hrant!

John, thanks for your comments. The averaging method does take into account the relative lightness and darkness of parts of the image. Thresholding was done only at the end, to give a potentially traceable outline; but, following Raph, I have abandoned it even there. I think you and Nick might underestimate the variation of these printed characters, and so not see that one cannot, as Nick has it, "Look at the original specimen(s)" and see any outline to draw at all. The averaged p, the fourth in my initial post, contains more useful information than any single original character image (eg concerning damage to the spur), and is a more trusty basis for interpretation than any selected subset of printed originals. Here are four; and they're by no means unrepresentatively varied.

What the averaged master does lose is the coexistence of features in single original images; that's why some detailed analysis through software modelling of what's going in terms of terms of inkpress, off-centre impressions and so on would be desirable. I just don't know whether it's possible.

Nick Shinn's picture

George, couldn't you run a "control" test of averaging just using ordinary letterpress type -- made from a digital master, even? It wouldn't be a perfect simulation, but you might uncover some useful information/principles nonetheless.

Outline filtering: I've only ever done it in Illustrator, using that app's "pre-built" filters.
I haven't used the Brush filter, which i suspect could be the most malleable and useful.

George Horton's picture

Sorry about your internet connection, Raph; I look forward to the code.

John might be reassured by the output from my software tweaked such that there is no thresholding for any reason at any stage. It's rather beautiful:

John Hudson's picture

Peter, thank you very much for giving fresh form to the discussion about rhythm, and saving Hrant and I from going around in circles again.

It seems to me that Hrant’s misgivings about rhythm have to do with the fact that in visual wordform resolution role-architectural subcomponents of a bounded map are not processed sequentially but parallelistically, and sequential processing associations are hard to supress, because of how we learn about rhythm.

Yes, but supress them we must if we want to talk meaningfully about 'rhythmn in optical-grammatical contexts'. This is why in the various threads that touch on this subject I have insisted on the all-at-onceness of perception of visual rhythm, stressing that it is not sequential. It seems to me that simply surrendering to associations of rhythm with sequence or flow is to accept a kind of phenomenological blindness. It impoverishes both discourse and experience, not only in understanding of typography and reading, but in a broad range of visual experience.

Everything you need to know about the non-sequential nature of visual rhythm is embodied in this painting, and by the same token this painting cannot be understood without suppressing 'sequential processing associations' of the word rhythm even though, brilliantly, the painting alludes to several different kinds of sequential rhythm: jazz music, traffic, street noise.

George Horton's picture

Peter, I think the strategy you're talking about is the most attractive kind of reductionism I can think of. Are you recommending in practical terms that I get to the punch outline if possible, and if not that I take an outline derived with consideration for inkpress in corners and blunting at spurs from an averaged image like the one two posts up, write a fixed-front ductus over it, see where the ductus doesn't match and how the metal departs from it, assess the consistency of those modifications, and of second-order variation in manipulation (going on potentially ad infinitum but really it wouldn't be, not with humans), and then redraw the font appropriately, making further alterations until the result is intuitively satisfying and perceptually authentic?

It might help if I say that, though the primary purpose of this project is merely scholarly, a secondary purpose is in the production of fine digital books. I will Multiple Master for optical size, probably from 10 to 15 point, the maximum range for which I suspect MM is going to work effectively and manageably. Titling and photopolymer/matrix engraving versions might eventually follow, as might a set of sans titling caps (the 1499 caps would make an absolutely gorgeous very lightly modelled sans).

John Hudson's picture

Thanks for the new images, George, especially the last which is, as you say, rather beautiful.

My concern, looking at the variation is as stated much earlier in this thread: it is the variation itself that contributes to the look of the letterpress Aldine page, so averaging the variation to produce single letterforms seems to me not to be a very good way to produce a 'faithful revival' of the type. To a very large extent, what you are averaging is not letterforms per se but printing artifacts affecting letterforms. Consider, for example the scan of the third p from the left in the four examples a few messages back. In this specimen, the spur is very clearly present as a part of the letterform: it is really there in a very clean and structural way. Looking at this occurence of the letter, one knows for certain that this feature of the letter is Griffo's design. But in other examples, perhaps even the majority of examples, this feature may not be so clear. But lack of clarity does not imply lack of presence and certainly doesn't imply lack of design. It just means that the impression was very often bad or that individual sorts were worn or broken, such that the design cannot be ascertained from those examples. This means that you have to apply judgement about which samples display the design and which obscure the design through printing artifacts. Looking at that beautiful combined image, we can see that what is a clear design element in one example is, through averaging of printing artifacts, made to disappear: the averaged form embodies the lack of clarity.

To me, this suggests that the averaging methodology has merit relative to the careful selection of input. There needs to be an analytical process to determine which specimens show forth the design, and which are simply poor specimens.

George Horton's picture

I sort of agree. But that third p is very unusual in the length and solidity of its spur, and I think it shows some perhaps unknowable degree of merely random ink-press. We need something to sway us besides aesthetics - Griffo might chosen to make only a little spur. So we turn to the averaged master and see that, as one might anticipate knowing the properties of softish metal, beyond the spur and also to the right of the descender serif there are darker pale-greys - ie more than background-random cases of ink impression in these regions, implying that yes, these projections have been blunted and should be restored, according to judgement. On the other hand if one chose to believe that the splurge on the bottom of the second p of the four was a ravishing example of Griffo's real intent, lost in too many instances by blunting, then one could check this aesthetic choice against the average and see that the progression from darkest to lightest is much quicker moving vertically down from the serif than to the right of it, so that the descender of the second p is simply botchy.

Almost all the impressions are visibly poor in some part or other, and when errors coincide it can be hard to see them as such - in other words two 'p's can both look convincingly perfect but be wholly unlike. Thus in selecting good material one must be careful not to exclude too much, though in moderation I'm sure it's sensible, and I have done so in previous runs of the software.

Syndicate content Syndicate content