I've never heard of the smallest unit in font editors being called an em. As I've seen it, it's always the body size, however it's spelled. The Font lab manual spells it eM. The wikipedia spells it em. The typowiki spells it Em and has the same definition. The Oxford Concise spells it em.

If you are a real glutton for punishment you can read this thread about the em.

The history of units is covered in Walter Tracy’s book Letters of Credit. Unfortunately it was written too long ago to cover digital units, but it gets the old stuff covered.

Yes, before digital type entered the stage things were quite clear.

A friend writes me off list:

The fontlab manual does say eM, as William says.
and calls the smaller unit UPM Unit Per eM. Maybe
fontographer used different terminology?

It looks to me like they only capitalized it that way to illustrate the acronym "UPM", a la "eXtensible Markup Language" or "Get Rid Of Slimy girlS".

Yeah, I don't think there's any significance to the capitalization of the word. An em is an Em is an eM.

An em is a width dimension equal to the height of the type being used. Thus 1 em of any font gives a square (as opposed to rectangular) area.

When an em is divided by 1000, the only term I have heard is unit. Unit is not a consistent unit -- in older systems units were 18 or 54 to the em. The closest one could come to a standard definition is that the unit is the smallest amount that an em could be split into when providing different widths for glyphs.

If I am not mistaken, one of the sales points of the Monotype system early in the last century was that it used 54 units per em, while Linotype machines were at 18 units per em.

>Linotype machines were at 18 units per em.

IIRC I heard that Linotype only introduced units in the photo type era. In in the hot metal era, I don't believe it had a unit system, in contrast with Monotype. Their big restrictions were no kerning of the f's, and the duplexing of the italics, so the italic and roman characters had to be the same width.

As Bill points out, Don, unitized widths were a characteristic of Monotype and not of Linotype.

The Monotype system employed a matrix-case which was typically 15 x 15. All characters in a given row needed to have the same set width. So, designs were restricted to a system which divided the em into 18 units (not all increments were used for each font). This unitized system was also integral to the calculations of the Monotype justification method.

The Linotype, on the other hand, had no unitized width restrictions. However, it was not possible to have any kerns (character elements projecting beyond the metal body), which was possible in Monotype. Also, most Linotype families had two styles on a single matrix, which meant that the two designs had to have the same widths -- i.e., be "duplexed". There were usually options for Roman duplexed with Italic or Roman duplexed with Bold. This had a distinct advantage for certain types of typesetting, but a distinct challenge to the type designer.

As Bill notes, early Linotype photo-typesetters did have unitized systems, because of certain mechanical restrictions in the advance drive motor. Also, in the newspaper industry, TTS was a pervasive standard which required unitized widths, and this was applied to Linotype newspaper types.

Regarding the original question, I've never heard that capitalization made any difference. An em is as has been described. I've always heard the units used in digital font applications referred to as "em units" -- which for Postscript fonts is 1/1000th of an em.

-- K.

Having a limited number of character widths in a font does provide a certain rigor and discipline to the type design.
And the reductive duplexing of roman and italic (eg Sabon, Optima, Univers) gives a distinctly modern character to typefaces designed with that in mind.

With so many options for width now, and global kerning, it's easy to be sloppy, and the aesthetic departs from one which is "scalar" -- i.e. having an inbuilt, mandated predisposition towards harmonic relationships between character widths.

Foundry types had scalar widths (and this goes at least as far back as Manutius, see Types of Space, a very practical reason being that this facilitated justification. The system incorporated a variety of space characters, so that "s+hair space=n" (or something like that, I'm not familiar with the exact maths).

> Foundry types had scalar widths

If you mean that traditional typefounders had a kind of unit system for letters, of the kind that Monotype used in order to justify its lines, I'm doubtful if this was ever the case: the width of hand-cast type depends on the setting of the registers of the mould, and is beyond the control of the punchcutter or matrix maker. Given the idiosyncracies of individual casters, it can be different every time a new fount is made. To be sure modular ‘self-spacing' type was introduced at the end of the 19th century – but only when long-haul hand setting of type was nearly over.

As for the exactly regular modular spaces of Aldus, which Peter Burnhill thought he had detected, judging from my limited experience as a compositor I think these would be very unhelpful. The classic set is thin, mid and thick – or 5, 4 and 3 spaces to the em quad. Their odd combinations produce subtle increments, or reductions (or whatever the opposite of increment is), which are a godsend when a line is just too loose or just too tight, as often happens. But judging what just too loose, or just too tight, or indeed just right, all mean is a matter of feedback through your fingers. The number of words to be spaced varies unpredictably from line to line, and you may want bigger spaces between two l's than two o's. Setting type well by hand is not a completely rational and mathematical business, thank goodness.

Yes, that makes much more sense. The word spacing in old typography often varies wildly within lines.
Nonetheless, I still suspect the typefounders of yore of harbouring secret mystical numerical systems of letter proportion!
But as you say, whether this had any practical effect would be moot.
I suppose it would be possible to compare the drawings of the Romain du Roi with printed samples of the face and see how much deviation there was.

"....widths were a characteristic of Monotype and not of Linotype."

Widths were, I believe a characteristic of both. When I was at Linotype I found many drawings on an 18 unit system and all the new faces we made were on 54 units. There are unit systems hiding in other places as well. E.G. Highway Gothic, we found, was made on a 10 unit system, most likely we surmised, to make the composition of road signs much easier for USDOT employees in a the precomputer age.

Regarding the original question, some of us are waiting with great anticipation for the all-encompassing definition of "Em or em" as promised by T. Phinney on Wikipedia a year ago. Since this promise chased out all the other definitions of "em", it should be interesting to see what he comes up with, no?

Cheers!

I admit my memories of the 18-unit Linotypes is based on the Linotron 202 photosetter. I assumed (incorrectly, I guess) that this was based on the old metal machines.

Thinking back over how those wonderful Linotype and Intertype machines worked, with sliding space bands, it is clear that a unit system would not be needed, as the bands would slide to the exact thickness needed (as long as there were enough of them in the line--else ... spurt).

David --

I don't doubt that you encountered many unitized drawings. You were at Linotype toward the end, right -- late 70s? As I alluded to in my post, the Linotype photo typesetters were unitized. The early machines were constrained to 18 units. The V-I-P was able to handle a finer system of 54 units. I would expect this is what you were drawing for, right?

Also, the TTS system was 18 units, IIRC. (I remember an anecdote from Mike Parker: something to the effect that Jackson Burke lamented that much of his reign at MLCo consisted of overseeing the continual revision of Corona to the TTS system as newspapers desired to increase their body sizes.)

I also wouldn't be surprised if faces that were originally Monotype faces maintained their original 18-unit widths when done up for Lino (even in metal) -- e.g., TNR. And, of course, Sabon complied with 18 units to be "cross-platform."

My original point was that unitized widths were not an inherent restriction of metal Linotype, in order to counter Don's earlier recollection.

-- K.

"You were at Linotype toward the end, right — late 70s?"
Uh, we're both still here. So... the end of what? ;) Yes we drew for VIP fonts and they were on 54ths. My impression was that unitized widths were a feature of all mechanized type setting systems. All pre-digital systems needed to count and move in increments coarser than those of the grid until PS where the increment met the grid at 1:1, (finally).

Cheers!

I meant toward the end of "American" Linotype. My recollection of the history was that MLCo was sold off to Allied Chemical in 1979, and then moved offshore not too long thereafter (what I call "the end").

By "unitized widths" I mean a system of discreet and relatively coarse divisions of the em. Is this what you mean as well?

Obviously, there was a minimum discreet increment for the drafting and machining of Linotype characters. Off the top of my head, I think it was 0.00025 inches. So, I suppose you could say that the Linotype used a 333-unit system (for 12-point type; more like 4000 units for 72 pt). But I don't think that's what we're talking about.

> All pre-digital systems needed to count and move in increments coarser than those of the grid

Almost all; but not quite.

Yes: from what I know, the Monotype composition system (in metal) employs a mechanical calculator -- justifying by basically adding up character widths in the line and then calculating the wordspace by distributing the leftover space when the compositor indicates an end-of-line. Different arrangements of the matrix-case (i.e., which characters are in which unit row) require different cylinders for the composing machine, in order to properly calculate the line. (That is, the machine needs to "know" whether the n character is 9 or 10 units in the particular face at hand.)

In distinct contrast, however, the Linotype justifies by a purely mechanical means. The space bands are wedge shaped. As the compositor reachs the end of a line and decides where to break, the wedges are then physically forced upward to push out the line before casting the slug. No discreet increments (smooth, evenly tapered wedge) = no math = no calculation necessary. So, no organized system of widths needed.

On the other hand: yes, the early Linofilm photosetters rely upon a writing prism that travels at a constant speed, and I believe justification is achieved through some kind of calculation. (My understanding of this system is not as thorough.) Factors such as the speed of the drive and the minimum amount of time required to open and close the shutter have some bearing on the constraints of advance widths.

The V-I-P, developed later, employs a stepping motor and can move discontinuously (even standing still!), which allows more flexibility (thus, 54 instead of 18 units).

-- K.

Thanks for the summary Kent. I think "No discreet increments (smooth, evenly tapered wedge) = no math = no calculation necessary. So, no organized system of widths needed." this may be theoretically true, but practically everything was on unit systems because of the plethora of mechanisms and the eternal shortage of steady hands.

Cheers, and thanks!

Well, David, you may know better than I. I'm more the historian; you were actually there. ;-)

I haven't run across any discernible systems (or references to same) in the materials from the period that I've been studying -- Dwiggins, Ruzicka, et al., circa 1925–50.

But perhaps after Griffith left more was put on unit systems than I'm aware of.

-- K.

Starting with the book Typographical Printing Surfaces by Legros and Grant, and then reading between the lines of the ATF catalog - there's a table where it gives the actual widths of the different spaces that are billed as "3 to em" and "5 to em" in different point sizes - it appears that their foundry types were indeed made to a unit system, but unlike Monotype's, it did not scale with the size of the type.

Instead, the width of every type slug, whatever the point size, was an integral multiple of 1/4 of a point. So foundry type was unitized, and the evidence of that can be found.