"If Big Red contains specimens of Electra (not just the supplements) then it dates from at least 1935. If it contains Caledonia, then it would be after 1939. But I seem to recall that it doesn't have the latter."

Kent

The issue I have does not contain Caledonia. I have found some Linotype information that would seem to indicate that it was released at the turn of 1938. References indicate it to be available in very early 1939 (January).

I'll be going over these last few very interesting posts later - for now let me just put up my scans. I'm not even going to comment anything on them yet.

This is 11 (not 12) point Electra from that issue of The Colophon:

scaled 12/11 to the same Em as Kent's "Emblems ..." sample, here again for reference:


And here's a composited cascade from the supplement to that big Linotype specimen book:

And that cascade with each size scaled to 14 Em.


Fire away.

hhp

There's a problem with my second cascade image. I've now left-aligned the lc snippet in each line:


hhp

These are very interesting, Hrant. I can certainly see the great disparity in gain between the heavy inking on antique paper of the Emblems and the finer impression on calendared paper from Big Red. All I can say at this point is that I would also not specify Linotype metal Electra for a book to be printed on very smooth paper -- it was obviously possible for the original Electra to be too dazzling also.

Slight tangent: Here is an interesting early statement from Dwiggins, from Layout in Advertising (1928):

"But as a matter of fact, printing type on antique-surfaced paper seems to be something of a stumbling-block to the modern craft. The fine points of the time-honored process have been forgotten in learning (and learning very thoroughly) an entirely different trade -- viz., the trade of printing half-tones. . . . If [the designer] is a stickler for good workmanship, he will demand more impression than is compatible with the ideals of most pressmen; and an inking -- and an ink -- that will leave crisp, solid inlays with sharp edges printed from the face of the type, not from its shoulder."

I have no personal experience printing letterpress, so I am not sure what to make of this declaration. Gerald may find it provocative.

He goes on in this section to say that Antique and English Finish (what we would probably call Vellum and Wove today) are "essentially type surfaces," while Machine Finish and Sized & Super-calendered (Smooth and Super-Smooth or Plate) are "general utility half-tone papers." And then: "It is interesting to note how the picture, exclusively, has set the pace in all this career after smoother surfaces. Type has had to trail along as best it could." [all WAD's italics]

He also mentions briefly the development of coated papers as the current evolution in this search for ever greater polish and smoothness. He concludes the section with "The problem of assembling type and pictures on half-tone papers has not yet begun to be solved. The solution waits, first, for new styles of type letters; and after that for a sharp break with the traditions of the past."

I can't say for certain what he might have meant exactly, but Hrant, I thought you'd find this last statement fascinating.

I suppose it would be reasonable to assume that Electra (begun a few years later) represented some of Dwiggins's thoughts along the lines of a new style of type letters for printing on half-tone papers, which lends credence to the thrust of Gerald's arguments, as I understand them. If this was part of Dwiggins's goal, then I confess that I am ambivalent about his success toward this particular end. Sharpness? certainly; Lightness? maybe; High contrast? no.

It is interesting to note, however, that Dwiggins himself (as far as I know) did not specify any jobs to be printed with Electra on machine finish or s&sc papers -- continuing to prefer antique and english finished stocks.

-- K.

Kent

I suspect this may very well be a key to Electra. With the standardization of photoengraving in the very late nineteenth century the problem of the photograph was of great concern. With the letterpress process, halftones cannot be printed the same as type. They need a very smooth and unresilient paper. Any kind of overinking and the tonal quality is lost. Impression has to be exactingly uniform and quite light compared to type. Beatrice Warde's study concerning the differences between the letterpress process and the offset process on various kinds of substrates would also be revealing here. Many of the old standby typefaces did not perform well on the kind of paper required for halftones.

Kent

One other thing here that may or may not shed some light... It would probably have been quite rare for bookwork to have been printed directly from the Linotype castings. I doubt Big Red was printed from the castings. Linotype metal and Monotype metal were far too soft to withstand such repeated impression. Monotype, in fact, was not initially intended for printing. At the time, these casting processes were used for the initial composition. Once proofed and okayed, a paper mould of the form would be made and from this a stereotype was cast. The printing would be done from the stereotype. At a later point (when stereotype fell out of practice), or in less moneyied shops, they would probably have just taken a proof and made a copper photoengraving and would have printed from this. I would suspect you would see more of a bolder character from any printing done directly from the Linotype metal, which would grow bolder (or conversely, weaker, with heavier inking to compensate) as the metal face wore down during the printing process.

This is a great thread.

> the digital looks a whole lot better when you scan it in and blow it up

Yes, that's a classic indication of the need for optical scaling, as well as the trickiness of evaluating a text face: you can't just do it "directly" - you have to account for subvisibles. For example, you can look at a large lc "i" and say: "That dot looks slightly large. Good."

> I would suspect you would see more of a bolder character
> from any printing done directly from the Linotype metal

This is interesting.

--

Let me try to put out some thoughts:
1) The "Emblems ..." looks to me like bad printing, I'm sorry to say.
2) The digital Electra is very close to the 14 as it prints in the specimens. I can't be sure what Dwiggins intended, but I think chances are he would have approved of the digital for 14 point usage.
3) The difference between the 14 and the even the 12 is large. The digital should not be used for smaller sizes, unless there's a lot of gain (and it's tracked looser). Even then the x-height might be too small. (BTW, Linotype had two versions for each size: long and short descenders. Although I didn't see a long for the 14 in the Supplement.)

--

Also, here are some numbers (derived from the alphabet-length specs in the Supplement) that show how much wider each cut was compared to the 14:
7 : 16.25%
8 : 13.75%
9 : 13.75%
10 : 9.37%
11 : 7.39%
12 : 3.54%

BTW, coupled to my second cascade image, I think one can conclude that they messed up the 8 and/or the 9.

hhp

>With the letterpress process, halftones cannot be printed the same as type. They need a very smooth and unresilient paper. Any kind of overinking and the tonal quality is lost. Impression has to be exactingly uniform and quite light compared to type.

This is as I suspected. Presumably Dwiggins felt that when printed alongside half-tones, type looked too light and lacked proper substance. But given that, don't you think that the type designer's challenge would be to create a face that would project and demonstrate "substance" when subjected to a light impression? Electra seems to perform in exactly the opposite way (to my taste): it appears too fragile and spindly with too light a touch. But, again, perhaps I am not thinking about this properly, not being a printer.

You mean that BW article from the Bowater Papers, right? I'm still trying to get my hands on a copy.

>It would probably have been quite rare for bookwork to have been printed directly from the Linotype castings.

This is a new insight for me. Thank you. What would you say is the average run size where the Linotype metal would have been used directly? For instance, the Typophiles' Chapbooks 35 & 36 (the Postscripts on Dwiggins) were printed in an edition of 1050 (at the Anthoensen Press, on Curtis Colophon Text); this edition would have been printed directly from the castings, right?

-- K.

Kent

I think Electra probably is a bit fragile and spindly as it was possibly intended this way. I can't say why that would be. You would know more than I. But compared to other faces at the time I would say yes, but also compared to other faces that were quite "light" in color, such as Centaur, it seems to have a bit of strength built in to it. Centaur, printed direct from the Monotype, self destructs very quickly. I doubt Electra would. I noticed something else about its quirkiness. That lowercase g, for instance, seems to have ink "supports" built in to it.

I only have a photocopy of the Bowater. Someone on TYPO-L a while back had a couple of extra copies but I didn't get a reply when I posted about it.

I can't say whether or not the Anthoensen Press would have been printed directly from the metal. It would certainly have been cheaper to do so and those are not long enough runs to have damaged the metal much. I noticed in the issue of the Colophon that Hrant showed me and that was put up here, that there was metal flashing between some of the characters. Flashing is a leakage between the matrices as the result of dirt or misalignment. I'd "suspect" then that this printing, in particular, was printed from the metal.

>2) The digital Electra is very close to the 14 as it prints in the specimens. I can't be sure what Dwiggins intended, but I think chances are he would have approved of the digital for 14 point usage.

As a match for stroke weight and contrast, yes, so it would seem. But you didn't address the differences in character shape. Maybe I'm imagining things -- you have access to the original: check the width of the counters in 'h n m u'. Also that digital 'a' -- isn't it just way too wide? (Or am I way too sensitive or delirious or something?) And the rounds . . .?

>(BTW, Linotype had two versions for each size: long and short descenders. Although I didn't see a long for the 14 in the Supplement.)

The original drawings that Dwiggins did were on the standard 12-point alignment, but with long descenders -- meant to be cast on a body one point larger. Presumably Linotype also fashioned the short descenders for those who wanted to set solid.

BTW, an undated broadsheet specimen I have (a photocopy) lists the specifications for l.c. alphabet lengths as follows:

7 pt, 93 points l.c. alphabet
8 pt, 104
9 pt, 117
10 pt, 125
11 pt, 135
12 pt, 142
14 pt, 159

How does that compare with your figures? I don't know what might be going on in the supplement.

-- K.

"But given that, don't you think that the type designer's challenge would be to create a face that would project and demonstrate "substance" when subjected to a light impression? Electra seems to perform in exactly the opposite way (to my taste): it appears too fragile and spindly with too light a touch."

Actually, I've always assumed just the opposite. That a face needed to be thinner to work well with halftones. Photographs have a great deal of tonal range! My photo students like Optima and Univers (in the thinner versions) best for captions etc because they are much less interfering in terms of weight (among their other qualities). The opposite would be printing with wood engravings which carry a tremendous amount of weight. There you would need a fairly heavy face. Poliphilus, Plantin, etc.

So maybe Electra is really something of a Trojan horse. I'd like to think so, and give a tip of the hat to Dwiggins.

Gerald --

Okay, what you say makes sense. Dwiggins argued throughout his career in favor of engraved illustrations (rather than photos or halftoned illustrations) as the best complement in book texts because they are of the same nature as type; halftones are too gray in character. I suppose, then, that you are right that an appropriate type to match up with halftones would need to be lighter in overall color.

Lighter perhaps, but so contrasty? It still just doesn't jibe for me with his comment about being drawn to avoid the extremes of thick and thin usually seen in modern style typefaces. In this way, I would think Gill's Joanna is perhaps a better answer to this matter.

-- K.

Kent

I don't see Electra as all that contrasty. Compared to Joanna sure, but the latter is more of the oldstyle approach to letterpress design. Electra certainly was not a Modern typeface design in the historical sense. There is certainly something else going on there. Moderns also required a fairly smooth paper and there was initially a reaction in opposition to them because of this.

I believe that the contrasts built into Electra serve a technical purpose. If Hrant would put up the 96 point, the Linotype specimen, (hint, hint) along with the similar sized digital, you will seem some interesting characteristics to the original that would more have served the face as a letterpress face. I suspect these were "knowingly" altered by the digital designers.

>> "The solution waits, first, for new styles of type letters; and after that for a sharp break with the traditions of the past."

It's hard to say how concrete versus visionary this was. But I know I love the guy.

> I'm still trying to get my hands on a copy.

Who have you asked? :-)
"Lo-Fi" is my middle name.

BTW, Warde's text I found disappointing.

> Someone on TYPO-L a while back had a couple of extra copies

Shelley Gruendler: shelleygruendler@HOTMAIL.COM
I can't compete with her originals (assuming they make it to you).

> metal flashing between some of the characters.

What's interesting is that it's consistent on almost all the lc "o"s, and what's even more interesting is that it's also on the italic lc "o"s, but very few others in either style... What does all that mean?

> check the width of ....

OK, give me some time.

> Presumably Linotype also fashioned the short descenders for those who wanted to set solid.

Yeah, those who realized what a real text face needs to be... What's weird is that the 14 (which could be seen as a display face) is the only one missing the long descenders! Also, the Cursive seems to have not been cut in a 14. But this is all based just on the Supplement.

> 7 pt, 93 points l.c. alphabet
> 8 pt, 104
> ....

Those are all identical except for the 14 being shown as both 159 and 160 in different places in the Supplement.

> If Hrant would put up the 96 point

OK, OK, even though it's basically just the 14.
I feel like a lowly scan-monkey, what with my inability to fathom all this interesting halftone business you guys are talking about... I think I need pictures.

hhp

"Who have you asked? :-)
"Lo-Fi" is my middle name."

From a previous post, I thought your middle name was anal. That H must be silent.


"BTW, Warde's text I found disappointing."

At the time, you were probably looking for something else in it. I didn't find what I was looking for in it either. But what I did find was quite educational.


"What's interesting is that it's consistent on almost all the lc "o"s, and what's even more interesting is that it's also on the italic lc "o"s, but very few others in either style... What does all that mean?"

The o matrix isn't fitting properly. Molten metal is leaking to the printing surface, where it appears as a feathery crack-like line.


"I feel like a lowly scan-monkey..."

Yes, but you're our "lowly scan-monkey" and we love you for it.

None of my letters are silent. But the ones to the left of spaces are very much anal. BTW, what's the head of a word, the first letter or the last? Maybe it's the last, since that's the direction the word is going. Unless you're a luger.

> The o matrix isn't fitting properly.

Is the italic "o" having the same problem just an amazing coincidence? There are very very few other flashes.

--

Me Tarzan, Warde Jane.

Pop quiz: What was Beatrice's mother's maiden name?

hhp

You guys owe me like a Double-Kingburger. With egg.

--

To be honest, I'm confused. I don't know what's coming from where...
I'll just show you my composite, maybe you can figure it out.



L12: Metal Electra 12 (long descenders) from the Supplement.
L14: Metal Electra 14 (short descenders) from the Supplement.
L60: "Enlarged" Electra letterforms from a 1954 Linotype
publication called "Distinction in Type Design".
DT: Digital Electra - Text cut.
DD: Digital Electra - Display cut.

I tried to normalize them all vertically, but in some cases
it's not clear what's going on. For example, the Digital
Display seems to be smaller on the body than the Text...

--

Kent, I'll start working on the Digital-versus-14 widths stuff soon.

hhp

BTW, the red guides are all based on the DT.

hhp

I find it very interesting that Dwiggins carefully adjusted the weight and width with each point size. I think this might partly explain Kent's and my nostalgia for the pre-optical and digital typeface printing.

If the weight adjustment was dropped in the optical era, as I believe Gerald said, it might explain why many of the old books looked more 'right' to our eyes.

As someone noted, with modern computer programs, the adjustment of weights should be less onerous, and could be a selling point. You type designers tell me: how much of a headache would this be, and how much do you think the benefit to the usability of the type?

This URL:
http://www.printinghistory.org/htm/journal/articles/31-32-Cost-Benton.pdf
includes an account of how different sizes were adjusted with the pantograph machines the Benton's invented, and I guess were the basis for the production of 20th century typefaces until the optical & digital eras.

It says that Zapf dealt with the old system, in which three sizes would be designed, and adjustments made in the rest. If TYPOFILE could ask Zapf about this issue of adjusting type weights and widths point size by point size, I bet the answer would be fascinating.

In Open Type could the shifts in weight with point size be built in to the typesetting program?

Kent, I think your instincts on the digital Electra being too light are completely right (meaning I agree!) Dwiggins may have wanted a very clean and new look, but he carefully adjusted the weights. The 12 point is noticably heavier than the 14 point basis (according to Hrant) for the digital and the 10 point more so, in Hrant's scans.

Hrant

The L60 and the DT were the ones I compared. The enlarged L60 read out at 84 or 96pt, if I remember correctly, and DT was adjusted to 119 points to match. DD, the digital display, is actually slightly smaller in height than DT when set to the same size.

The actual L60 is a bit finer in some areas than the DT as it shows here. Since L60 was well printed, I believe the swelling in the junctures is a part of the design that does differ from the digital, and is not an indication of ink gather.

William

There are a number of optimized fonts out there. I believe this has been a concern for well over a decade. The dfTYPE font Rialto is optimized. In fact, a version of it, Piccolo Pressa, is the only digital font that has been configured for letterpress. I've put up an attachment here of it as part of this thread. The first instance is Piccolo (normal), the second is the Pressa. Note the ink traps on the Pressa. Piccolo, of course, is optimized for text sizes. There is also a small text size and a display size. Interestingly, the metal Electra has almost the opposite effect (swelled junctures) compared to its digital version. Perplexed by this.

df TYPE's Rialto Piccolo/Pressa Piccolo inktraps.pdf (6.7 k)

In Open Type could the shifts in weight with point size be built in to the typesetting program?

The OpenType Layout feature tag registry includes a <size> feature, registered by Adobe. This feature does not apply any transformations to fonts that contain it, but it enables a font to be linked to a specific point size or range of point sizes. To date, there are no applications that support this, but the idea is that when the user selects a font that includes this feature and sets the type size to e.g. 7pt, the appropriate font from the family will automatically be used.

I met a very interesting man from Berlin at the TypoTechnica event in Heidelberg last month. He has made a study of the ways in which types were adapted to different sizes in different periods and technologies, and has identified ten principles for optimising type for different sizes. He showed me samples of a custom version of Aldus, optimised for 6-12pt. The 6 and 12 point sizes were manually designed, and the intermediate sizes were interpolated. Adobe, of course, tried in some of their Multiple Master fonts to interpolate between a much greater range of sizes (between 6pt and 72pt). I tend to agree with my German colleague that interpolation for optimised sizes only works across a smaller range of sizes.

>I find it very interesting that Dwiggins carefully adjusted the weight and width with each point size.

William -- Credit where credit is due: Dwiggins would not have been responsible for any of the adjustments you notice. At that time, these optical adjustments were all performed as a matter of course by Linotype. The man most responsible would have been the unsung hero of Linotype's drawing office, Nils Larson. There were separate drawings made by hand for each character of each size of each design. When Larson retired in 1959, a commemorative monograph published by Mergenthaler gave the example that for just the Caledonia series alone, Larson made and/or supervised a total of 6,850 production drawings. That's just one design. During his career, Larson was involved with probably on the order of one hundred different designs.

The subject of optical scaling for different point sizes is an involved and interesting topic and is a perennial favorite on lists and forums, worthy of its own thread.

-- Kent.

Okay, here's the report I promised. There are still gaps in some of the details, but I think this provides the broad strokes of the provenance of the Linotype outlines.

The primary sources for this research are conversations with Mike Parker and Matthew Carter, augmented by the timeline given on the Linotype Library website. For those who don't know: Mike was Director of Typographic Development at Mergenthaler Linotype Co. (MLCo) from 1959 to 1981. He succeeded Jackson Burke, who in turn succeeded Chauncey H. Griffith. Matthew worked with Mike at Linotype from 1965 until 1981, when they both left to co-found Bitstream (with Cherie Cone and Rob Friedman).

* * * * *

In 1958, Mergenthaler Linotype Co. introduced the Linofilm phototypesetting system. The original Linotype production drawings were used to convert the designs to film. The Linofilm system, however, was limited to an 18-unit em, so the original drawings had to be adapted to this system. In many cases, this may have been a simple matter of marking new side-bearings on existing drawings, but some faces may have had characters redrawn, as well. Also, new drawings would usually have been made for characters that would benefit from the absence of kerning and duplexing constraints -- characters like the 'f', as well as many italic designs.

In a process which is somewhat analogous to the metal process of Drawings -> Brasses -> Punches -> Matrices, the drawings were used as patterns to cut friskets from rubylith; these friskets were photographically reduced to create glass "plaques;" the plaques were assembled into an oversized version of the Linofilm grid; and this array was used to produce the Linofilm film master.

The Linofilm optical system was only capable of enlarging to 200%, so three grids were needed to cover the size range: one for 6-12, one for 12-24, and one for 18-36. The master drawings used were usually 8, 12, and 18 point, but there may have been exceptions. Later the V-I-P system was capable of covering a wider range with a single master and the middle, 12-point size became the default master.

At some point in the late 1960s or early 1970s, MLCo began converting the fonts into digital data. The friskets were scanned on a drum scanner and the data was "quantized" on a main-frame and converted into fonts for the Linotron machines. Linotrons set encoded data as vertical stripes, but it is unclear whether this data was initially stored as such or as outlines. The Linotron 202, introduced in 1978, is mentioned as the first model to store the font data as outline vectors. I could not determine what format these vectors were encoded in, but they may very well not have been Bezier splines.

In 1987, Linotype licensed its typeface library to the fledgling Adobe and the fonts were converted into the now-familiar Type 1 Postscript format. Type 1 was a proprietary format until 1989 or 90, so I would think that Adobe did the conversion themselves, but a recent Linotype catalog indicates that the fonts were digitized by Linotype. (This could, of course, simply refer to the original digitization from the friskets back in the 70s.)

* * * * *

When the American Linotype offices were closed down for good in the late 1980s, Mike Parker managed to save some of the materials from destruction. There was a complete set of Linotype matrices which had been used by the MLCo. in-house typesetting department. He got these donated to RIT. Later these went to the Smithsonian Institute, where they are stored off-site to this day. I don't know what happened to the brasses. The production drawings were initially put into storage and later offered to the Smithsonian. The Smithsonian was reluctant to take them because of concerns about mold contamination. So Parker found a home for these with the Museum of Printing in North Andover, MA.

* * * * *

Matthew said that in the early days of photocomp and offset printing, the results were really terrible. Over time there was some improvement. Although apparently there was some speculation that the font companies went back and added weight to the characters, Matthew said this is baloney. Instead, plate-makers and printers probably just learned how to use the photo output better.

Faced with the daunting task of converting the metal library to photocomp back in the late 50s, the MLCo Drawing and Matrix Manufacturing departments must have felt that it would be impossible to redraw every single character to compensate for the difference in gain -- not and get the Linofilm system to market in a timely fashion. Matthew said, "Having just done that to one family, I cannot see how it could ever have been accomplished at the time for the entire library." He is referring here to his recent reworking of Linotype Monticello for the Princeton University Press.

Also in reference to that project, he told me, "As I discovered with Monticello, the work of improving the photocomp data until it looks reasonable is really daunting. The digital font will never look exactly like the hot-metal face printed by letterpress, but at least it will not be as anaemic as digital data made straight from the hot-metal drawings."

* * * * *

Respectfully submitted, ;-)

-- Kent.

Type 1 was a proprietary format until 1989 or 90, so I would think that Adobe did the conversion themselves, but a recent Linotype catalog indicates that the fonts were digitized by Linotype. (This could, of course, simply refer to the original digitization from the friskets back in the 70s.)

Some of the T1 fonts may have been digitised by Adobe, but I believe Linotype did a lot of this work themselves using the Unix-based tools provided by Adobe.

> how much of a headache would this be, and how much do you think the benefit to the usability of the type?

There's the manual headache of making more masters, but the real trick is deciding what the numbers are (assuming that the compensation is continuous - some people don't think it is - but I do, although it's not linear). Exactly how much wider/darker/looser/larger_on_the_body/more_open/etc. does the 6 have to be compared to the 12?

As for the benefit, the sad answer is that if people were willing to pay for this stuff, it would never have stopped being done... But I think there is a niche market. And there's always one's priceless love for quality.

> sizes were adjusted with the pantograph

Actually, people soon stopped using that feature of the pantograph: too much work, nobody cared - not even back then. You see, the deprecation of optical compensation started way before the photo era.

> The enlarged L60 read out at 84 or 96pt, if I remember correctly, and DT was adjusted to 119 points to match.

I actually normalized everything to 59.5 points (based on your 119), so I don't know where that 84/96 is coming from...

> Nils Larson

Is there a way we could dig up his numbers? That would be golden.

> worthy of its own thread

Totally. What do you guys think if we moved all the posts since Paul Schliesser's of March 8 to a new thread? If there's no complaints, let's ask Jared/Joe to do it, OK?

> here's the report

Wow. Is this good stuff or what. Thanks a bunch.
Are you vegetarian?

> original drawings ... had to be adapted

Now my "Rg" scans make more sense.

> only capable of enlarging to 200%

Funny. Technological limitation helping to preserve typographic quality... :-/

hhp

BTW, were you guys asking about specimen dates?
The Supplement is from 1948, and it cites 1939 for the main volume.

hhp

I still think the Digital-Text was based on the L60, but I now think that the L60 in turn was based on the metal 12, as opposed to the 14. Also, there were progressive adjustments, like making the tail of the "R" protrude less (in the digital). One clue is the long descenders, but the proportions -and even the color- seem to be closer to the 12 than the 14.

As for the Digital-Display, it seem to be an interpretation of the metal 14, but with some effects from the Digital-Text, like the more reserved tail and the longer descenders.

--

My upcoming closer comparisons between the Digital(s) and the 12/14 should help...

hhp

OK, it's not the best, but...



One -significant- thing you can't see there is that the digital's x-height is closer to the 12's than the 14's.

hhp

>Is there a way we could dig up his numbers? That would be golden.

What numbers do you mean?

>Are you vegetarian?

Not for many years. What makes you ask?

>One clue is the long descenders,

I forgot to mention earlier, the specimens I have copies of all mention the 14 available with long descenders also. I don't know why the supplement doesn't mention it.

-- K.

> What numbers do you mean?

Larson must have used numbers to scale the original drawings, at least in terms of width, probably x-height and sidebearings, and maybe even weight and other variables. There's no way he would have eyeballed everything. What was his background?

One clue would be the relative compensations of regular versus irregular letters: if groups of letters like "hnu" are scaled very similarly, but something like the "g" occasionally deviates, especially in its internal dimensions. The less regular a feature, the more likely eyeballing would take precedence over measured compensation.

Another clue would come from measuring the relative scaling in different fonts, to see if they match.

This type of stuff cries out for formulas. And maybe that's one reason it's done less these days: too much of an artsy attitude. Even the best succumb: in "Letters of Credit" Tracy discovers that Falcon's smaller size (which he considers ideally compensated) have a numerical basis to their scaling (p 53), but he can't help reverting to the comforting but consistently over-emphasized "The eye is the final aribter" in the end...

> What makes you ask?

'Cause I owe you a meal now. :-)

> the 14 available with long descenders

Huh.

hhp

BTW, one thing that got lost in the commotion: none of the printing of Electra outside of "Emblems ...." seems to have that baseline shifting! I really think that was a defect. Remember that the type wasn't released yet.

One more thing:
I just compared the color of the Colophon printing against that of the Supplement (which frankly seems as light as the digital to me): the former is slightly darker. The paper stock of the Colophon seems less coated though - if that makes sense.

hhp

One more clue that Larson used formulas: if you look at the alphabet lengths, and their gradual [relative] increase from the 14, you see that the percentages are suspiciously round. If you take the 14 to be 160, you get: 3.54, 7.39, 9.37, 13.75*, 16.25. If you take the -more probable- 159, you still get: 4.19, 8.06, 9.09, 14.47*, 16.98.

* Two of those - again, must be a mistake. Hey, that's even more proof! What are the chances the ratios would be identical?

As for the small variations (like 14.47, from 14.5), they come from rounding the lengths to integers.

hhp

Hrant

Yeah, I'd say paper stock has a profound effect on the look of Electra. At the end of the specimen showing in Big Red are two pages from the Limited Editions Club publication of Ravelais' _Garganatua and Pantogruel_. In twelve point Electra. Quite nice. Any baseline shifting is not readily noticeable. Includes decorations by Dwiggins. The reproductions are extremely clean and precise, as must be the original, which I doubt was printed on slick stock. Probably a bunch of copies of this available through places like abe.com.

Again though, I'd suggest baseline shifting is common for most of the small text sizes in the faces presented by Linotype. Even Dwiggins' Metro. I believe this is done on purpose, as an aid to legibility, reversing my earlier judgement that this was a technical problem (?) In many cases, for instance, the lowercase v will drop severely in the 7 point and begin to slowly correct itself as per the point sizes increase. The ce combination and lu combination begin to adjust to each other likewise as the size increases.

So, how about scanning up a bunch of those for us, bud. Ha! Just kidding.

> baseline shifting is common for most of the small text sizes

OK, I haven't looked there very closely (yet). But the huge shifts in Kent's scan from the 12 point in "Emblems ...." are not present in either The Colophon (11 point), nor anything above about 10 point in the Supplement.

hhp

Hrant

Without going back and checking... the thread is a monster... maybe "Emblems..." was not printed in 12 point but in a smaller size?

I think Kent said 12.

BTW, here's a book about Larson: "Letters and Nils Larson: reflections on his contributions to typographic development, 1922-1959". Mergenthaler Linotype Company, 1959. Typophile monograph 58.

It's at the Clark Library, which is both good and bad...
If it's any good, I'll buy it.

hhp

Hrant --

Letters and Nils Larson is the monograph I referred to above. It is quite small -- 4.5x7", 36 pages. Nicely set in 10 pt Monticello. There are seven short tributes from Walter Tracy, R. Ruzicka, P.J. Conkwright, Hermann Zapf, Paul Bennett, Jackson Burke, and James Darley. All, save Burke's, are average about 500 words. There are also five pages listing the typefaces Larson had a hand in.

You won't find any magic formulas.

That "eye is the final arbiter" statement is Griffith's. Tracy was just quoting CHG. This was a common Griffithism.

Larson's first job was as an apprentice in a brass foundry and machine shop in Sweden. He started at Mergenthaler as a clerk in the matrix department. Later he started training in the letter-drawing department. He took night classes in art and architectural design. Larson advanced through the drawing office to eventually become the supervisor. He basically came up under Griffith and would have absorbed a lot of his knowledge from CHG's methods.

Only after the pilot size of a typeface was completed and approved for all technical details of shape and proportion and fitting, then the remaining sizes of the series would be graded for correct optical proportion. From Burke's tribute:

"In order that a type design may pervade a series, optical illusions need to be effected. Six point type looks like its related twelve point only because it is carefully drawn to be different. It is proportionately wider, heavier, squarer and more open. It is Larson's trained eye which establishes how much wider, heavier, squarer and more open. There is no formula which serves to shepherd a design through all these permutations."

There wasn't a formula, but that doesn't mean that everything was done arbitrarily or willy-nilly. I would guess that Larson (and Griff) had some rules of thumb in their heads. Once the grading of the essential dimensions of a design was determined, the figures were recorded on a graph. This became the master blueprint for that particular typeface. But the graph would vary by design. The point is that the graph recorded what the eye determined, not the other way around. That was Griff's stance on fitting as well. "The eye is the final arbiter. . ." I know, it frustrates the hell out of you. (It frustrated Dwiggins somewhat too, as he kept looking for a formula for fitting. I believe Tracy's method may have been based on Dwiggins's search.)

One could probably deduce much of those rules of thumb by analyzing different designs in the Linotype line. When I'm able to get over to the Museum of Printing I hope to take a closer look at this matter.

Is that Colophon of yours the new series, Vol I, Number 1, Summer 1935? In a letter from June 20, 1935, Dwig writes, "Emblems and Electra is going to be pretty swank. It will be interesting to see how the printing trade takes it. My bet is that it will be good merchandising. The Colophon use of Electra will be a good christening, I think. The type looks fine in the pages we have got so far." The two publications were contemporary.

The Emblems is set in twelve point, the pilot size. Electra was not released yet, but the pilot was completed and the series was being graded. Obviously, the eleven was also done, since The Colophon used it.

I'm at a loss to explain the degree of ragged baseline in Emblems related to the other examples. But in general, the movement of the bottom overshoot on the rounds and pointeds as one traverses the different sizes was very likely another dimension that was graded for optical adjustment by size. Somewhere I recall reading Harry Carter discussing some general figures for overshoot -- maybe in the Fournier; I'll have to try to dig that up some day.

-- Kent.

> You won't find any magic formulas.

:-(
But it seems nice anyway.
BTW, I'm not looking for magic.

> Only after the pilot size of a typeface was completed and approved for all technical details of shape and proportion and fitting, then the remaining sizes of the series would be graded for correct optical proportion.

BTW, wouldn't you say this limits the effectiveness of the design across the whole range of sizes? Isn't it better to design with more than one size in mind (like some masters do actually do)? Or maybe Larson took enough liberties with the original to make it work well at all sizes. They must have trusted him. JvK would have killed him.

>> "It is Larson's trained eye which establishes how much wider, heavier, squarer and more open. There is no formula which serves to shepherd a design through all these permutations."

I don't buy this for a second. It's just romanticism. The mysterious eye (which it is) is much sexier than plain thought.

But let me be clear: I'm all for the eye deciding what the numbers are - no doubt about that. And I don't pretend it's worth devising a Magic Formula that accomodates everything (even if that's possible). What I'm saying is that math was -and is- used a lot, even by those who seem to see benefit in denying it. And the clues are everywhere. The clean interpolation in Falcon for example. I will find more.

> "The eye is the final arbiter. . ."

No, and even Griffith couldn't really have believed that - the evidence is there. The eye and math work in parallel. In fact, the straight interpolation in Falcon is an indication of too much reliance on math at the expense of the eye (which is known to be non-linear) on the part of Linotype. ATF had better math.

> it frustrates the hell out of you.

I can see how it would have frustrated Dwiggins, but I have no boss to tell me what fonts not to make, so in my case it's just disappointment: that people I otherwise admire have a coping mechanism preventing candor.

--

My Colophon is Spring '38, V3#2.
I really don't understand how Dwiggins could have approved of both that and Emblems... Very disconcerting.

hhp

"I really don't understand how Dwiggins could have approved of both that and Emblems... Very disconcerting."

Hrant

I'd agree that the Emblems... looked like poor presswork but in all probability Dwiggins would have okayed, at most, the setting. The presswork would have been out of his hands. And, a problem with letterpress is that there will not be consistency throughout the run. The goal of a fine printer is to maintain consistency but there will be some variation. In some cases, with bad presswork, a lot of it. And that particular showing of the Emblems... might have been an errant copy that just didn't get weeded out, or most likely, there was no weeding out.

A good army buries its dead. Some printers don't.

Pop quiz: What was Beatrice's mother's maiden name?

Lamberton.

You got it!

You win an all-expenses paid night on the town with Ms Warde! You'll just need our patented time travel machine ($9,999.99) - just make sure you aim for a date before her marriage (or at least before her swap).

hhp

Kent Lew's wonderful research seems to confirm my suspicions that commercial pressures resulted in poor initial digitizations of old typefaces.

John Hudson's information that Open Type has the potential to seamlessly switch between weights designed for different point sizes makes me wonder: Is Adobe planning to provide this in an upgrade of InDesign?

This new capacity would serve their current Open Type 'opticals', and my impression from seeing the huge variety of weights in faces intended for newspapers is that newspapers would gladly use point by point adjustments.

Since Adobe are the leaders on this, I would think that they would want to keep their 'edge' on this issue...

As someone noted, optical sizes really deserve a thread of their own, but briefly:

Adobe are implementing the <size> feature in all their own families with 'opticals', and are encouraging other foundries to do the same. At the moment, it is a little difficult to add <size> information to a font, as the feature utilises the GPOS table in an unorthodox way (basically, it uses the table to store a string of bits that provide information to an application but do not constitute a normal GPOS lookup). I have not experimented with the Adobe FDK or FontLab in this regard, but since I used VOLT for all my OT development I would probably need to post-process fonts with TTX in order to add <size> information to the GPOS table.

InDesign currently has an 'Automatically use correct optical size' preference, but this currently only works with MM fonts with an optical axis. My guess is that this function will be extended to the OT <size> feature, but I don't know when.

Hrant --

>Or maybe Larson took enough liberties with the original to make it work well at all sizes.

I believe he did. The evidence is in the specimens.

>They must have trusted him.

I believe they did. Witness the tributes.

>But let me be clear: I'm all for the eye deciding what the numbers are - no doubt about that. And I don't pretend it's worth devising a Magic Formula that accomodates everything (even if that's possible). What I'm saying is that math was -and is- used a lot . . .

I can't argue with you there. And I don't think Griffith would either. They clearly used plenty of math. But they didn't have a formula. Just because a certain proportion or interpolation worked in one instance, doesn't mean that it would work in the next instance. Which is why a trained eye is needed to judge the results. Hence Griffith's statement. I think you take read too much into it. "The eye is the final arbiter" doesn't mean you don't measure, multiply, and cogitate. It just means that when all is said and done, you *look* at it and decide. Use the math, but trust your eyes.

In that Falcon memo, by the way, the very next sentence following that "arbiter" statement, and which Tracy did not include in his quotation, is "Just how accurate is the eye and whose eye is a safe and acceptable standard, is an open question that can be answered -- by whom?"

The Falcon grading is by and large mostly linear, but it is not *strictly* linear. The advance proof (the final proof made of each size of each design, which becomes the official record of the font) lists several key measurements for a given font. Those that bear most on this discussion are the face height (from top of ascenders down to the bottom of descenders), the height of the capital H, and the length of the lowercase alphabet.

I've put together a graph of these three features over the range from 6- to 12-point Falcon. In creating my graph, I used a convenient scale for the beginning and end points of the lc alphabet line; then I placed the end points of the other two in the same places and located the in-between points proportionally. So there are three completely different scales. My purpose was to be able to overlay the lines so that we could compare their degrees of linearity. I'm sure someone more mathematically inclined than I will tell us if my method is flawed. The figures are included in my graphic for you to do your own analysis, if you wish.



I don't have an explanation for the divergences in the upper range. You can see by the number of decimal places they bothered to measure that the processes were geared for precision machining, so I don't think you can explain them away as some sort of manufacturing errors.

Make of this what you will.

>that people I otherwise admire have a coping mechanism preventing candor.

Do you mean Griffith? I don't think he was trying to hide anything. That "romanticism" about Larson was Burke, remember, not Griffith. I think the thing preventing candor is he's dead -- he can't join the forum and explain it all for you. You're going to have to do some digging for this stuff.

Maybe this will help. In a footnote appended to one of WAD's letters regarding Falcon, Griffith offers this explanation of something Dwiggins makes reference to:

"* By 'graph' is meant a method devised by me for charting and recording the predetermined visual proportions and basic dimensional data in relation to each point size of a series, in this case 6- to 11- point, inclusive, above mentioned, predicated on the norms established for the pilot 12-point size. The proportions of height and width and the thickness of stems and hairlines in each size are thus determined in advance of and as a basis for subsequent development of the indicated point sizes required for the series. Dimensional data on the graph is a recording of the visual conception of relative proportion. Visual conception is given a graphic test by cutting experimental characters of each size. If found correct, the standard list of trial characters, 'E H M O R T W e f h j m n o p' are cut in each size as a further test. After final approval of these, the letter draughtsmen proceed with letter drawings for each of the point sizes in the series. N.B., for perfect visual gradation of sizes, each is drawn separately, and none is cut from the master patterns or drawings of any other in geometrical proportion by the pantograph process."

I'm still digging.

-- Kent.

> I've put together a graph of these three features

Wow, nice.
The only problem I see with your method is that you can't appreciate the scale of divergence.

> the number of decimal places they bothered to measure

Since the Blue and Red are floating-point, the divergence is unlikely to be an error, I agree. On the other hand, I don't think they had precision down to about 2.5 microns (like in the Red 7) back then. In fact, look how round those numbers are: they're a result of some sort of integer math ratio applied to some [whole] source number. Beyond the mil range there's a uniformity that cannot come from eyeballing.

Here's some thoughts:
- The deviation in Black (from the straight line) must be due to rounding to integer.
- I could understand the deviation in Blue: you want caps smaller in reading sizes than either in really small sizes or sizes verging on display setting.
- I have no ideas about the Red divergence.

>> "The proportions of height and width and the thickness of stems and hairlines in each size are thus determined in advance of and as a basis for subsequent development of the indicated point sizes required for the series.

There you go!
Makes total sense. You take a 12 point master (the "intended" design*), generate benchmark characters in the 6 that the eye likes using predetermined numbers (that were arrived at through an iterative process with the eye as "arbiter"), refine the numbers some more (using the eye), then generate everything with those numbers, interpolating all the middle sizes (with tweaks as your graph shows). The only problem is that the eye is not linear... so I would have actually established an eye-determined interpolation scheme in addition. BTW, I once measured InDesign's optical spacing compensation for different sizes, and found that it was bi-linear**: no change below 4 point, a straight line between 4 and 12, a separate straight line between 12 and 72, and then no change above that.

* Like I said, this is problematic.

** It probably wasn't some non-linear formula for performance reasons.

Anyway, please let me/us know if you find more good stuff like this!

hhp