The Widest Letters

I am preparing to update a page on my web site which talks about a pet fantasy of mine: modifying the IBM Electronic Composer so that it could also accept typewriter elements, and even the elements for the Mag Card Executive (88 character elements for the proportional-spacing styles later seen on 96 character elements for the Electronic Typewriter Model 50).

Researching the subject once again brought me face to face with one particular issue that, of course, is obsolete in today's age of the laser printer, where mechanical restrictions have largely vanished.

And that issue is?

The IBM Executive typewriter was a conventional electric typewriter with typebars for the individual letters, except for one thing, the different characters could differ in width. Called "proportional spacing", this allowed typed material to look somewhat like professionally-printed typeset material.


The characters could be from 2 to 5 units in width, so the graduations in width were fairly coarse. This tended to be visible as a shortcoming in most typestyles (as the typefaces for typewriters were called by IBM), although some, like Text and Charter, seemed to hide it quite well.

In many typefaces, the letters m, M, and W were five units wide, while most capital letters, like B or E, were four units wide, and most small letters, like a or e, were three units wide. This was at least approximately proportional to the appropriate widths for those characters.

One of the early interchangeable element typewriters, a contemporary of the Blickensderfer, was the Hammond. The Hammond typewriter design served as the basis for the VariTyper, which many offices used to prepare forms, and which also was used for the round paper labels in the center of a phonograph record on many occasions.

The VariTyper had proportional spacing. It had a three-bank keyboard, which meant that there was one shift for capital letters, and another shift for numbers and special characters. Also, its touch was not the best; typing on it was slow and laborious, so it was reserved for special-purpose use. (This is why the Blickensderfer and all the other old interchangeable element typewriters died out after the Underwood Five became the gold standard for typewriters.)

Its characters were from 2 to 4 units. The standard spacing it used was very similar to the spacing the IBM Executive typewriter used for the Documentary typeface, one of the most popular, with a few exceptions. The letter r was 2 units wide instead of 3. And the letters m, M, and W were reduced to 4 units.

This didn't seem to compromise the visual appearance of text set with it too much, and so it seems to have set a precedent.

The unit system for the IBM Selectric Composer assigned widths from 3 to 9 units to the various characters. Comparing those widths to the number of units assigned to the characters of Times Roman on the Monotype, I found that in general there was a consistent ratio, except that the letters m, M, and W ought to have been 11 units wide instead of 9 on the Composer.

A single unit could be either 1/72", 1/84", or 1/96" on the Composer, and typefaces could be from 12 points to 7 points in height.

The IBM Mag Card Executive, and later IBM 96-character electronic typewriters and daisywheel typewriters, used proportionally-spaced typestyles which seemed to be only slightly smaller, in terms of the number of units they took, than Composer typefaces. Here, characters were between 3 and 7 units in width; characters with widths from 3 to 6 units on the Composer kept their width, while those 7 and 8 units wide on the Composer were one unit less wide on the proportionally-spaced typewriter.

So perhaps m, M, and W, which "should" have been 11 units on the Composer, would have been 9 or 10 units here, but they, too, received only the maximum width of 7 units.

And, thus, this history raises the question: up to a certain point, if mechanical restrictions impose a certain coarseness on the variety of widths used for characters, is making the widest characters narrower the lesser of the two evils compared to accepting a certain additional degree of coarseness?


  • I have spent quite a while looking at these early proportional typewriter fonts.

    Perhaps *past a certain point*, making the widest characters a little narrower is preferable. But in fact, many of these devices actually make a *lot* of characters wider. Look at the most Times-like font on the IBM Executive. The “average” characters are much wider. That way the widest characters don’t suffer so much and the relative differences in character widths don’t do anything weird.

    I tried taking Times and simply rounding all the characters to the nearest width for the Selectric Composer, and the result was awful, compared to IBM Press Roman (I think that was the name of the most similar font). So it is a matter of careful and relativistic design for the constraints of the system.
  • John SavardJohn Savard Posts: 825
    edited February 2019
    I tried taking Times and simply rounding all the characters to the nearest width for the Selectric Composer, and the result was awful, compared to IBM Press Roman (I think that was the name of the most similar font). So it is a matter of careful and relativistic design for the constraints of the system.
    Yes, Press Roman was the Composer's Times Roman lookalike. This is very interesting, and it makes sense that making a character narrower is fraught with danger to a greater extent than making it wider.

    Incidentally, since 11 point Press Roman had a red arrow on the element, which meant it used the escapement with 1/72" as the size of a unit. So 11 units, not 9, was an em.
  • On the Monotype, Times Roman had a set width of 10 1/2 points rather than 11 points, so 18 unit characters like M weren't a full pica in width.

    ijl were 5 units on Monotype. 5/18 times 10 1/2 is 2.9, so 3 units on the Selectric Composer was the closest width.

    ft were 6 units on Monotype, and rsI were 7 units. Those become 3.5 and 4.08 units - and they became 4 units on the Selectric Composer.

    cezJ were 8 units on Monotype, and became 5 units on the Selectric Composer, whch is the closest to 4.67.

    bdhknopquS were 10 units on Monotype, and P was 11 units on Monotype, and both became 6 units on the Selectric composer. 5.83 and 6.4 both round to 6.

    However, agv, which were 9 units on Monotype, became 5 units on the Selectric Composer, while xy and the digits, also 9 units on Monotype, became 6 units on the Selectric Composer. As 5.25 is the equivalent scaled number of units, this is one case where the widths were adjusted upwards.

    I'm not questioning your eye as a designer - Press Roman is well-proportioned - but I am wondering about your arithmetic.
  • John SavardJohn Savard Posts: 825
    On my web page at

    I have now added an illustration of how this principle of limiting the width of the widest characters worked for ATF in the design of Quick-Set Roman, which looks much better than the typefaces of the Justowriter or the IBM Executive, despite using nearly the same unit system.

    Actually, two images are added: one showing all six point sizes of Quick-Set Roman, scaled to be the same height, and one showing that the widths in points allocated to the four width classes, A, B, C, and D, while never quite matching the 2:3:4:5 proportional typewriter ratios, hovered very close to them, and also differed in opposite directions depending on the point size.

    This discussion is in aid of my hypothesis that ATF may have used that trick - I don't yet have any print samples I'm confident of with which to check that they did - to make the 5-unit version of the ATF Typesetter produce output that looked better than that of a much less expensive Justowriter.
Sign In or Register to comment.