[Next] [Up] [Previous] [Home] [Other]

Printing Terminals and Proportional Spacing

The IBM Executive typewriter allowed typed documents to somewhat resemble typeset text by varying the amount of space allocated to different letters. For example, with the Documentary font, a character could be from 2 to 5 units in width, where each unit was 1/32 of an inch. Other fonts, even when still typed at 6 lines to the inch, had a different basic unit; the Mid-Century font, for example, used a unit of 1/36 of an inch.

Using a crude dot-matrix font, the diagram below illustrates a unit system of this type:

and this keyboard chart includes typical character widths for such a unit system:

| ! | @ | # | $ | % | ¢ | & | * | ( | ) | _ | + |
|  2|  5|  3|  3|  5|  3|  4|  3|  2|  2|  5|  3|
| 1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 | 9 | 0 | - | = |
|  3|  3|  3|  3|  3|  3|  3|  3|  3|  3|  3|  3|
  | Q | W | E | R | T | Y | U | I | O | P | ¼ |
  |  4|  5|  4|  4|  4|  4|  4|  2|  4|  4|  5|
  | q | w | e | r | t | y | u | i | o | p | ½ |
  |  3|  4|  3|  3|  2|  3|  3|  2|  3|  3|  5|
   | A | S | D | F | G | H | J | K | L | : | " |
   |  4|  3|  4|  4|  4|  4|  3|  4|  4|  2|  3|
   | a | s | d | f | g | h | j | k | l | ; | ' |
   |  3|  3|  3|  2|  3|  3|  2|  3|  2|  2|  2|
     | Z | X | C | V | B | N | M |   |   | ? |
     |  4|  4|  4|  4|  4|  4|  5|   |   |  3|
     | z | x | c | v | b | n | m | , | . | / |
     |  3|  3|  3|  3|  3|  3|  5|  2|  2|  3|

space = 2

2 fijlt I .,:;'!()
3 abcdeghknopqrsuvxyz JS 0123456789 "?#+-*/= 
5 m WM @%_½¼

This is based on the unit system for Documentary at a unit size of 1/32", but may deviate from it for some of the special characters, my actual primary source being a font with the same spacing as Documentary for the letters which was used on a Friden Justowriter.

There were typewriters that offered proportional spacing long before the IBM Executive, but they were not successful; the IBM Executive could type as rapidly as a normal electric typewriter, but those early attempts apparently were encumbered by the added complexity of proportional spacing.

The Olivetti Graphika typewriter, which came out after the IBM Executive, and offered proportional spacing of a similar kind, with charcters from two to five units in width, spaced with a unit of 0.8 or 0.785 mm, and was manual instead of electric.

It came out in the 1950s, after the IBM Executive, which was the first commercially successful typewriter with proportional spacing. Uniform spacing, however, was recognized as a deficiency of the typewriter from the outset, and thus earlier attempts at a proportionally-spaced typewriter were made.

One web site gives the first such typewriter as the Crandall 1, from 1883, and also notes the Columbia 1 as another early typewriter with proportional spacing. One that was particularly well known was the Maskelyne, from 1889, which continued to be sold into the 1920s, from a firm which was indeed founded by the famous stage magician of that name. There was even a typewriter called the Automatic that went to the trouble of offering proportional spacing despite only typing capital letters.

While we are on the topic of IBM trivia, and if we have wandered from punched card codes to line printers, another important input-output device, the IBM 2741 printing terminal, may also deserve some attention.

In the top two rows of this diagram, the two major families of 2741 arrangements are illustrated, first the arrangement of characters on the keyboard, and then the arrangement of characters on the typing element.

The arrangement at the top, the Correspondence arrangement, uses exactly the same elements as a standard office Selectric typewriter. The one in the middle is the PTTC/EBCD arrangement, which is designed so that the position of characters on the element leads to a code for the letters which has some relation to punched card codes and the EBCDIC internal code.

On the bottom is the keyboard arrangement, and arrangement of characters on the element, for another major family of devices using the IBM Selectric "golfball" element, but which was not made into a form of the 2741 terminal.

The charts of 2741 codes on the right shows the codes for the printing characters applicable to the Correspondence arrangement and then for the PTTC/EBCD arrangement; the relation between these is determined by matching characters in corresponding positions on the elements. The space and control characters always have the same codes.

Now that we have discussed the IBM Selectric Composer, here is a chart of its keyboard, including the widths of each of its characters in units, a unit being 1/72", 1/84", or 1/96" depending on the size of type being set, these sizes being indicated by red, yellow, and blue triangles upon the element respectively:

| ! | † | + | $ | % | / | & | * | ( | ) | _ | @ |
|  4|  6|  6|  6|  8|  4|  8|  6|  4|  4|  8|  8|
| 1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 | 9 | 0 | - | = |
|  6|  6|  6|  6|  6|  6|  6|  6|  6|  6|  3|  6|
  | Q | W | E | R | T | Y | U | I | O | P | ¼ |
  |  8|  9|  7|  8|  7|  8|  8|  4|  8|  6|  8|
  | q | w | e | r | t | y | u | i | o | p | ] |
  |  6|  8|  5|  4|  4|  6|  6|  3|  6|  6|  6|
   | A | S | D | F | G | H | J | K | L | ¾ | ½ |
   |  8|  6|  8|  7|  8|  8|  5|  8|  7|  8|  8|
   | a | s | d | f | g | h | j | k | l | ? | [ |
   |  5|  4|  6|  4|  5|  6|  3|  6|  3|  5|  5|
     | Z | X | C | V | B | N | M | ` | ' | : |
     |  7|  8|  7|  8|  7|  8|  9|  3|  3|  4|
     | z | x | c | v | b | n | m | , | . | ; |
     |  5|  6|  5|  6|  6|  6|  9|  3|  3|  3|

space = 3

3 ijl  .,;`'-
4 ftrs  I  :!()/
5 acegz  J  [
6 bdhknpquvxy  PS  0123456789  ]+*=$†
8 w  ADGHKNOQRUVXY &@%½¼¾ <em dash>
9 m  WM

where the underscore is filling the space that should contain a dash (I've seen pages that try to use &emdash; but that doesn't work), and I have cheated somewhat by using † or &#134; to represent a dagger, which assumes you are viewing the page from within Microsoft Windows. The difference in width between colon and semicolon, and open and close square bracket, is intentional, and has to do with the desired amount of space after each character.

And, while we're on this topic, here is my attempt at a reconstruction of the unit system of the Mag Card Executive:

| ! | @ | # | $ | % | ¢ | & | * | ( | ) | _ | + |
|  5|  5|  5|  5|  5|  5|  5|  5|  5|  5|  5|  5|
| 1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 | 9 | 0 | - | = |
|  5|  5|  5|  5|  5|  5|  5|  5|  5|  5|  5|  5|
  | Q | W | E | R | T | Y | U | I | O | P | ¼ |
  |  7|  7|  6|  7|  7|  7|  7|  4|  7|  6|  5|
  | q | w | e | r | t | y | u | i | o | p | ½ |
  |  6|  7|  5|  5|  4|  6|  5|  3|  5|  6|  5|
   | A | S | D | F | G | H | J | K | L | : | " |
   |  7|  6|  7|  6|  7|  7|  5|  7|  6|  5|  5|
   | a | s | d | f | g | h | j | k | l | ; | ' |
   |  5|  5|  6|  4|  6|  5|  3|  5|  3|  5|  4|
     | Z | X | C | V | B | N | M |   |   | ? |
     |  6|  7|  7|  7|  7|  7|  7|   |   |  5|
     | z | x | c | v | b | n | m | , | . | / |
     |  5|  5|  5|  5|  6|  5|  7|  5|  5|  5|

space = 5

3 ijl
4 ft  I  '
5 acehknorsuvxz J 0123456789 .,:;!?"@#$&¢()+-*/=_½¼
6 bdgpqy EFLPSZ

Although the Mag Card Executive is a rare machine, the same typefaces were used with the 96-character elements for the Model 50 Electronic Typewriter from IBM. This typewriter also typed in pica (1/10" monospaced) and elite (1/12" monospaced), and so I am confident that the unit size for this typewriter, and the Mag Card Executive, was 1/60", although at least one paper published in the IBM Journal of Research and Development claims a 1/72" unit for the Mag Card Executive. Comparing the spacing of this font to that of others, it can be seen that the spacing was designed for a font with a large x-height, the widths of the lower-case letters being larger in relation to those of the upper-case letters than in the other examples. Characters like @, %, and & were specifically designed to fit into the same relatively narrow space as digits to make it easier for typists to still put tables of numbers in uniform columns, even when some special symbols are used.

There may well be serious inaccuracies in the chart above, and someone out there with a Model 50 Electronic Typewriter, or a Model 65 or Model 85 may well be able to supply corrections. The forwards-pointing triangle on a Selectric Composer element indicated its spacing with its color; on a typewriter, a solid white triangle indicated pica, and a hollow triangle elite; proportional spacing was indicated by a circle, both on Model 50 elements and on those for the Mag Card Executive.

The 1/60" unit for proportional spacing of typed copy was, of course, also used on virtually all daisywheel printers that provided proportional spacing. A few even provided the option of printing small monospaced typing with letters 1/15" in width. Rather than being similar to the Mag Card Executive font, however, the proportionally-spaced fonts for some daisywheel printers tended to have small x-heights, a clear distinction between small and capital letters emphasizing that the font in use is proportionally-spaced.

In addition to the Mag Card Executive, and machines intended for typesetting, such as the Selectric Composer and the VariTyper, another modern single-element typewriter before daisywheels offered proportional spacing.

From the 1970s, many of you might remember the Olivetti Lexikon 82 and 83 portable typewriters, and the Brother Correct-O-Ball, which used plastic elements which resembled the Selectric typewriter "golfball" element.

Before those typewriters, Olivetti made a series of office typewriters, the Lexikon 90, 92, 93, and 94, of which the 92 had proportional spacing; there were also the later 90C, 92C, 93C and 94C models, which had a correction ribbon added. These also used a single element. However, that element was sometimes called a "pineapple", as it looked very different from a Selectric element.

A Selectric element had 88 raised characters in four circles of 22 characters which stood side-by-side next to each other, the circle in the middle of the element being the largest one, with those above and below spaced more closely together. Teeth on the bottom of the element were used to align the element exactly as it struck the page. A lever at the top was used to mount and dismount the element.

The elements for the Olivetti Lexikon 90 and related typewriters had 96 raised characters in six circles of 16 characters which stood vertically one on top of another.

So the two circles in the middle had the characters most widely spaced, but the spacing was vertical, and so there would be a preference for putting characters with descenders in the middle instead of wider ones there.

Between each circle of characters were columns of V-shaped indentations so that a pair was immediately beside each character; these were what was used to align the element as it struck the paper. A rod through the center of the element was pulled out to dismount the element and pushed in to mount it.

One model in the series, the Lexikon 92 (and the Lexikon 92C as well) could be switched between Pica, Elite, and proportional; the proportional spacing was comparable to that of a normal Executive typewriter, not a Mag Card Executive.

One had to lift the cover on the top of the typewriter to change this setting, which was controlled by a pair of levers covered by sliders, but that was reasonable as normally it would only be changed when changing the element.

It may also be noted, though, that the Crandall 1 typewriter also used a single element (like many other early typewriters) as well as having proportional spacing (unlike all but a very few).

A Note on Typewriter Innovation

The book Century of the Typewriter notes that many typewriters were invented before the one by Sholes and Glidden, who are often regarded as the inventors of the typewriter.

That is true, but it is also true that their typewriter succeeded, and became the ancestor of the manual typewriters that continued to be made and sold by many companies for many years, because its design permitted relatively fast and efficient typing.

Similarly, there were several typewriters that permitted the typist to see what was typed on the paper as soon as it was typed before Underwood came up with the "visible" typewriter that became such a success as to drive the older kind of typewriter off the market.

So, while it definitely is mistaken to refer to the IBM Executive as the first typewriter with proportional spacing, as it dates from 1944, and other typewriters with that feature date back to 1883, typing on an IBM Executive was as fast and efficient as typing on any other ordinary electric typewriter. Of course, there was still the additional difficulty involved in backspacing; IBM did have a memory backspace achieved by mechanical means for the original Selectric Composer some years later.

Having typed on a VariTyper, I know that one had to press quite forcefully on the keys to type with it, and thus typing on a VariTyper or a Hammond would be slower than typing on a conventional manual typewriter.

The IBM Selectric could not keep up with a few of the very fastest typists, but for almost everyone else, it was at least the equal of a typebar electric typewriter.

Thus, as with the Sholes and Glidden typewriter, while IBM didn't invent proportional spacing for the Executive, and it didn't invent single element typewriters with the Selectric, in both cases they were indeed the first to achieve a level of speed and ease in typing that made these innovations, at long last, acceptable for use in the normal office environment.

The Blickensderfer, as well as the Hammond, had elements resembling those used in some toy typewriters, in that all the symbols associated with one shift state of the keyboard were in a single circle; thus, those elements had to move further to print a letter than the typeball of the Selectric, which moved in two directions. However, some of the older single-element typewriters, such as the Crandall, did not have that limitation. Although the Crandall had an unusual keyboard, and had a figures shift, the fact that depressing a key would have had to have moved the element in two directions means that the most critical element of an efficient single-element typewriter design would have been available without infringing on IBM's newer patents.

In any case, the Selectric didn't inspire any competitors of IBM to revive any of the older single-element typewriter designs. Some of the designs would have required a three-bank keyboard, which could be argued was only a problem due to fashion, but the absence of any attempt to produce a competing single-element typewriter based on the earlier designs would seem to indicate that IBM had, genuinely, through innovative technology to which it held patents, made single-element typing fast and efficient enough to be acceptable in the contemporary office environment, and the earlier designs had not reached that target.

Of course, it's possible that this alternative was only rejected because typewriters based on it would "look" old-fashioned, and thus not be saleable, and they would have worked as well, but at this point I doubt it.

Instead, Olivetti came out with the Lexikon 90 series, which used an element that, although it looked very different from an IBM Selectric "golfball" element, still moved in two directions, and flipped over to the other side when shifted.

Then, as IBM's patents expired, the Olivetti Lexikon 82 and 83 portable typewriters, and the Brother Correct-O-Ball came out, along with the Facit 1850 and the Triumph-Adler SE1000, in addition to typewriters from Remington and Silver-Reed that used actual IBM Selectric elements.

None the less, while IBM does deserve credit, therefore, for making the single-element typewriter with interchangeable elements, thus providing the potential of typing in many different languages, and using a wide repertoire of symbols for special purposes, practical, the Blickensderfer and the Hammond typewriter do, of course, show that the basic idea of such a typewriter itself was around long before the IBM Selectric, and to the extent that IBM advertising created a contrary impression, it would be subject to criticism.

A Vari-Typer could both type proportionally in multiple spacings, like a Selectric Composer, and it could also use monospaced typewriter fonts as well. Only the Mag Card Executive and the IBM Electronic Typewriter models 50, 65, and 85 could mix proportional spacing and normal typewriter monospaced type, and those typewriters could only produce proportionally spaced type in one size.

But the unit system of the Vari-Typer was much coarser than that of the Selectric Composer, which limited its usefulness as a device for cold type composition; it was a genuine piece of typesetting equipment, and was used as such, quite commonly, in fact, for purposes such as designing forms and producing the text on the central label portion of phonograph records. But in general, its limited unit system meant that the type it produced was unattractive in appearance compared to more conventional typography.

Chart of Unit Systems

Devoting all this space to unit systems and character widths, I might as well go "whole hog", and present this little cross-reference of some historical font metrics:

Characters Font, Device
12-point Oldstyle type, Foundry (1/4 point) Times Roman, Monotype Graphotype (improved) IBM Selectric Composer (1/72" for 11 point) IBM Mag Card Executive (1/60") Diablo 630 (1/60") Self-Spacing Type Quick-Set Roman Typotabular Gothic No. 4 (1/72") Graphotype (original) Documentary, IBM Executive; Friden Justowriter (1/32") Charter, IBM Executive (1/45") Mid-Century, IBM Executive (1/45") Varityper Varityper, Coder 2 No. 17 with Condensed Title No. 5, Linotype (1/36") Raphael, Underwood
i j l 15 5 5 3 3 (add I)
2 A (I: 3 units)
2 2 (add z, J)
2 2 2 1 1
f 6 6 4 4 (remove I)
t 17 1/2 3 3
I 7 7 (add Z)
r (remove g, y)
(add o)
s (J: 3 units)
3 (add R, X, Y)
c e z 21 8 (g, v, x, y: 9 units)
5 5 (remove z, J)
J 4
a g v (add o)
(remove g, x)
(0, 1, 2, 3, 4, 5, 6, 7, 8, 9: 24 units)
9 (add Z)
(remove o)
x y 0 1 2 3 4 5 6 7 8 9 6
h k n o u (add g, x)
(remove o)
10 (b, d, o, p, q: 9 units)
b d p q (add g, y)
S (add Z)
6 3
P 33 11 6 (remove w, X, Z)
5 4 4 (remove Z)
B E F L Z 12 (remove E, T, Z)
T (remove w)
(add A,U)
7 4
C 13 7 D 4 (remove R, X, Y)
w V (remove w, G, R, X)
(add D,E,T)
A G O Q R X Y (add w, K)
(remove A, U)
D N U 15 (add w, G, R, X)
(remove D, H)
H K (remove K)
(add w, X)
m 9 8 5 5
M 18 (add H)
7 5
W 50 16

For the oldstyle foundry type shown in the first column of the table, the digits are oldstyle or non-ranging digits, and this may be a reason that they are somewhat narrower.

Note that the lowercase letter g is narrower than the digits on the Selectric Composer, and the lowercase letters g and y are both narrower than the lowercase letters h, k, n, o, and u in Monotype Times Roman, while on the Mag Card Executive, lowercase g and y are wider than the digits and the lowercase letters h, k, n, o, and u, and, thus, there is a note to that effect in its column, as these letters cannot merely be displaced to an adjacent row in the diagram.

Similarly, other letters are moved around in Self-Spacing Type, a special kind of foundry type originally created by Linn Boyd Benton, the father of Morris Fuller Benton, which was designed to a unit system as an aid to the setting of tabular matter.

In addition to having a unit system, it had one other thing in common with the IBM Selectric Composer. The IBM Selectric Composer used only three sizes of unit, 1/72", 1/84", and 1/96". It allowed margins to be set, and tab stops to be placed, at locations spaced one-sixth of an inch apart. (Note that the unit sizes actually had 1/12" as a common multiple.)

Although it had only three unit sizes, more than three point sizes were offered of several of its more popular typefaces; this led to the problem (as noted, for example, in Production for the Graphic Designer) that changing the point size of text might only change the size of the text in the vertical direction, producing less impact than expected for copyfitting purposes. This also meant that some sizes of a typeface would be slightly more or less condensed than others.

The Varityper had four different unit sizes, designated A, B, C, and D. The A scale had the largest unit size, the D scale the smallest.

The number of units to the inch for these sizes, the size of each unit in points, and the size of each unit in millimetres, was approximately:

A: 39.1   1.8447   0.650    2.42    52/43
B: 42.8   1.6832   0.593    2.67    32/29
C: 47.6   1.5254   0.537    2.95    1
D: 52.8   1.3649   0.481    3.27    17/19

thus, while the Selectric Composer covered a 4:3 range of sizes with three sizes, the Varityper covered a range only slightly greater (28.4 to 21, about 1.35) with four sizes. Incidentally, the Varityper appears to have used the exact printer's point of 0.013837", rather than the approximation of 1/72" used by the Selectric Composer.

The fourth column contains the company's own description of the size of each scale in terms of characters per pica, and the fifth column integer ratios between the sizes that appear to be close to the truth, but which don't point either to the four sizes being common multiples of a small unit, or aliquot parts of a larger unit.

In patent documents, on the other hand, the four spacings, A, B, C, and D are referred to as providing 10, 12, 14 and 16 characters to the inch respectively. However, since that gives a ratio of 1.6 rather than 1.35 between A and D, it clearly is merely illustrative. Yet, in the earlier catalogue of non-differential spacing elements for the Varityper, elements are given with horizontal spacings of 12, 14, and 16.

The manual for a more recent VariTyper, with proportional spacing, notes that when the lever is pulled out to select Standard instead of Differential spacing, the approximate number of characters per inch for each size is 13, 14, 16, and 18 characters per inch, which indicates that the Standard setting causes each character to be three units wide.

Since proportionally-spaced fonts include characters that are up to four units wide, and the shape of a VariTyper element was identical to that of a Hammond typewriter element, that the newer units with Differential Spacing might not have supported typing with the largest sizes of fixed-spacing printing styles, and examination of a catalogue of fixed-spacing printing styles to compare their relative sizes seems to indicate that this is indeed the case.

The apparatus illustrated in U. S. Patent 1,918,299, however, would seem to indicate that the four spacings should be aliquot parts of a common large unit.

If there were slight inaccuracies in the illustration of a spacing rule that I've been studying, perhaps 2 11/16" might be exactly 105 A units, 115 B units, 127 C units, and 142 D units; the gears required for that would not be unreasonable in size.

At a late date, a second version of the English-language coder was offered for the newer models of Varityper in which the coder could be removed and changed. This allowed some type fonts, such as their version of Univers, to provide a narrower width, 3 units instead of 4, for a few capital letters. While the regular coder was very similar to the spacing for the Documentary typestyle on the IBM Executive typewriter, Coder 2 was more like the spacing used for Charter or Mid-Century.

The units for Self-Spacing Type were, like those of the Selectric Composer, also subdivisions of the Pica em. While Selectric Composer units were 12, 14, and 16 to the Pica em, Self-Spacing Type units were 7, 8, 9, 10, 11, and 12 to the Pica em.

Later, another form of type based on this principle, Quick-Set Roman, was made and sold by American Type Founders. Instead of using units which were different fractions of the Pica em, the fundamental unit was 1/2 point, or 1/144", although the effort was made to minimize the use of half points. Characters were made in only four different widths, and which character was assigned to which width did not vary with the size of the type, but the ratios between the various widths were allowed to change. The four widths for 6, 8, 10, and 12-point type, as given in the 1923 ATF catalog, and the widths for 14 and 18-point type, of which specimens are also given, are noted in the table below:

Point Size: 6 8 10 12 14 18
A 2 3 4 5 6
B 4 5 6 7 9
C 4 5 6 7 9 12
D 8 9 12 16

Despite the greater limitation in the variation of the widths of the characters, this later form of type seems to me to have been more attractive in appearance than Self-Spacing Type, its predecessor.

In addition to also making monospaced Mailing List Type, there was yet another unitized font for rapid setting available from ATF; this was their line of Typotabular Gothics. Most of these faces resembled Copperplate Gothic, but there was also a condensed gothic as well. These were all-caps fonts, with I (and sometimes J and 1) being one unit wide, and all other characters two units wide. There was only one exception to this, Typotabular Gothic No. 4, which did have lowercase, and four different sizes of characters; this one is shown in the table above.

A table in the famed reference work Typographical Printing-Surfaces, by L. A. Legros and J. C. Grant, giving the widths of the characters in two fonts of foundry type, one a modern typeface and one an oldstyle typeface, appears to indicate that a unit system was in use for foundry type as well:

Units     Points    Inches    Modern           Oldstyle
 16 2/3    12 1/2   0.17296                    W
 16        12       0.16604   W+=@             +=
 13 1/3    10       0.13837                    HMm@
 12 2/3     9 1/2   0.13145   KMm              DGKNOQRXw&
 12         9       0.12453   HGNUX$
 11 2/3     8 3/4   0.12107                    ACTUVY
 11 1/3     8 1/2   0.11761   ADEOQRVY
 11         8 1/4   0.11416                    BEFLPZ$
 10 2/3     8       0.11070   BCFLTw%&
 10         7 1/2   0.10378   PZ
  9         6 3/4   0.09340                    Sbdghknopqux
  8 2/3     6 1/2   0.08994   SJbdghknpqu
  8 1/3     6 1/4   0.08648                    aovy
  8         6       0.08302   vxy1234567890*   1234567890*%
  7 1/3     5 1/2   0.07610   ao
  7         5 1/4   0.07264   ce               Jcez
  6 1/3     4 3/4   0.06573   Irs?
  5 5/6     4 3/8   0.06054                    Irst-/?
  5 1/3     4       0.05535   fjt
  5         3 3/4   0.05189                    fijl[]()
  4 2/3     3 1/2   0.04843   il-/[]()
  4         3       0.04151   .,:;'!           .,:;'!

The fundamental unit is 1/6 of the unit used, but note that only once is the unit split into sixths, and there does seem to be a preference for whole units as opposed to thirds of units. As 13 1/3 units, or 40/3 units, equals 10 points, 1/3 unit appears to be 1/4 of a point, and 1/6 unit appears to be 1/8 of a point. There is reason to believe that the practice was not to scale the unit size with the size of the type, but rather to have the widths of all type slugs as multiples of 1/4 point or 1/8 point, since several ATF catalogues note that spaces which are nominally 3-to-em, 4-to-em, or 5-to-em in fact have their sizes altered so as to be multiples of 1/4 point, as shown below:

Point size:    6      7      8      9     10     11     12     14
3 to em        2      2 1/2  3      3      3 1/2  3 1/2  4      5
4 to em        1 1/2  1 3/4  2      2 1/4  2 1/2  2 3/4  3      4
5 to em        1 1/4  1 1/2  1 1/2  1 3/4  2      2 1/4  2 1/2  3

Optical Scaling

One characteristic of the IBM Selectric Composer is that it had three possible unit sizes, to permit a range of horizontal scalings; a unit could be 1/72", 1/84", or 1/96". Tab stops and margins could be placed at 1/6" intervals. (Because printers usually think of horizontal distances in ems, while points and picas apply to vertical distances, a horizontal distance of 1/6" is normally called a pica em in printing rather than just a pica.) However, 72, 84, and 96 are all multiples of 12.

Because there were only these three sizes, sometimes changing to a different point size of a typeface did not change its horizontal spacing, and some sizes appeared visibly more condensed than others. Let us imagine that someone were to bring out a daisywheel typewriter based on the ideas behind the Selectric Composer, but removing its limitations. Thus, for example, M and W would be increased in size from 9 units to 11 units.

Let us suppose the available unit sizes on such a typewriter are:

1/54", 1/60", 1/66", 1/72", 1/78", 1/84", 1/90", and 1/96".

It might also have a number of alternate assignments of units to the individual letters, so as to permit the printing of smaller type with a coarser unit system, or larger type with a finer unit system.

In addition, 3 units of 1/96" correspond to 1/32" in width, and 2 units of 1/72" correspond to 1/36" in width: with units of 1/90" added, 2 units correspond to 1/45", allowing daisywheels with the typestyles of all the various IBM Exectuive typewriters.

Even with eight different unit sizes instead of three, the sizes of the units are aliquot parts of one-sixth of an inch, instead of multiples of a finer unit. So the relation of sizes between them won't correspond exactly to the point sizes of type.

This, however, is not as much of a problem as it might seem. In general, as a typeface is reduced to smaller point sizes, it is also widened (and, as well, made bolder). As an example of this, Times New Roman on the Monotype followed this pattern, in part:

Point     Set       Stretch
-----     ---       -------
11        10 1/2     -4.5%
10         9 3/4     -2.5%
 9         9          0
 8         8 1/4      3.1%
 7         7 3/4     10.7%
 6 1/2     7 1/4     11.5%
 6         6 3/4     12.5%

where "Set" means that the Monotype machine's system of 18 horizontal units to the em is applied as if the point size of the type were the figure given in that column, rather than the vertical size of the type.

Another example of optical scaling would be the different sizes of Century Expanded as provided by American Type Founders:

Point      Width      Stretch
-----      --------   -------
11         13 1/2      0
 8         14 1/2      7.4%
 6         15 1/2     14.8%
 5 1/2     17         25.9%

where "Width" is the width of the lower-case alphabet in ems.

While optical scaling was a part of type design even in the days when type designs were realized through punch-cutting, when this firm used the pantograph to make matrices for new type designs more quickly and with a more polished finish, it also made the degree of optical scaling required explicit as part of the new process.

On a Selectric Composer, an 11-point type might have a unit size of 1/72". If we allocated other sizes of type to unit sizes from the expanded scale of units noted above, the results could be:

Point     Unit       Stretch
-----     ----       -------
15        1/54"      -2.2%
12        1/66"       0
11        1/72"       0
10        1/78"       1.5%
 9        1/84"       4.8%
 8        1/90"      10%
 7 1/2    1/96"      10%

and thus all of these choices, at least, would be quite reasonable.

As noted, one of the limitations of the Selectric Composer is that the capital M and W were not as wide as they should have been, in proportion to the other letters, in order to allow reasonably large sizes of type within the limitations of the size of the Selectric typewriter element, originally designed for use for normal monospaced typing.

While the layout of characters on the regular Selectric type element for typewriters did put the wider characters on the "equator" of the typeball, and narrower ones on the top row, the different arrangement on Selectric Composer elements was adopted because further optimization was needed to avoid "shadow printing", impressions made by adjacent characters on the element.

In my fantasy of making an improved version of the Selectric Composer based on a daisywheel element, in addition to having more alternatives of spacing, I would also seek to support a larger character set.

While daisywheel printers typically offered 88 or 96 characters on a printing element, there were several different approaches to making more characters available.

Ricoh daisywheel printers used a small daisywheel with 64 petals, each having two characters, one above the other, for a set of 128 characters.

The NEC Spinwriter changed the daisywheel to a thimble; it also had 64 petals with two characters, but in this way, one could move the thimble vertically on a rotating axis without having to raise or lower the motor that spun it. Although it had the disadvantage that the thimble-shaped elements, not being flat, would take more space to store, at least one reviewer of the Spinwriter noted that those elements seemed more durable than conventional daisywheels.

Also, NEC offered a significantly wider selection of print elements for the Spinwriter than were offered by most daisywheel makers.

There was also at least one manufacturer that made a daisywheel printer that could handle elements with 192 characters; it could also handle elements for their previous conventional daisywheel printers with up to 96 characters. But the extra character, added to the outer portion of the petal, was angled to the original character on the element.

This meant that instead of raising and lowering the element and the motor that spun it, the element was moved horizontally relative to the hammer that struck the correct character against the ribbon and the paper. So when the outer characters passed in front of the hammer, they were aligned vertically.

While the Diablo 630 ECS used printwheels with 192 characters, they had the two characters in the same vertical alignment, like those on the Ricoh daisywheels, as can be seen by a photograph of one which appeared on page 184 of the May 29, 1984 issue of PC Magazine, so I seem to be remembering a daisywheel from a third major manufacturer, other than Diablo and Qume, and so far, searching has not turned up the daisywheel printer advertisement which I so vividly remember.

The Qume TwinTrack printer allowed 192 characters in a more flexible manner; two printhead assemblies were present, each of which could have a daisywheel chosen by the user placed on it. Qume also made printwheels with up to 130 characters, but those simply went from 96 petals on the daisywheel to 130 petals.

For my purposes, I would want to combine both approaches, and have three, or even five, characters on each petal of a printwheel in addition to having two of them present.

Of course, at this point, the superiority of the laser printer becomes even more obvious.

Flexible Spacing

Another limitation is that all the typefaces had exactly the same spacing for each character. Later electronic versions of the Selectric Composer allowed a two-digit type code to be entered, permitting a different arrangement of spacing to be used with Cyrillic elements and for elements which contained two sizes of fonts like Copperplate Gothic, where a smaller size of capital letters replaced the lower case.

In general, this is not normally a problem for most typefaces in normal use. However, the proportions of the capital letters of some oldstyle typefaces are patterned after classic Roman carved letters, such as those on the famous Trajan column, in which case the capital letters B, E, F, L, and P, among others, are more noticeably narrower than most other capital letters than is the case in more ordinary typefaces.

Also, there are typefaces in which the size of the small letters relative to that of the capital letters is unusually small.

The chart above shows how a unit system derived from that of the Selectric Composer could be augmented with two alternative spacings for these cases. Note that capital J has become narrower in both additional settings; that is because, in the fonts concerned, it descends below the line, and thus is kerned: that is, part of the letter overprints the space alloted to the preceding letter.

It can also be noted that in these diagrams, the typefaces shown are in different point sizes, so that the width of the lower-case alphabet is very nearly the same for each one. Since the relative widths of the lower-case letters do not change as noticeably as the widths of some upper-case letters between typefaces, keeping their size fixed reduces the amount of change needed to match the different typefaces.

Also, in preparing this diagram, I examined three typefaces as examples of an old style typeface in which the classical proportions of the capital letters were used. There was a noticeable difference in the relative sizes of small letters and capital letters between those three faces, and the one in the middle was chosen as the example.

While still restricting the number of alternative spacings, there are a few other possibilities not shown here. The width of the capital J is something of a compromise value; a number of typefaces call for it to be made one unit wider rather than one unit narrower. Also, there are the newspaper legibility faces with a very large x-height; reducing the width of the capital letters 8 or more units in width by one unit may be appropriate for them.

The diagram above illustrates how these additional cases might be handled.

Another thing to consider, which involves relatively subtle changes, is how the widths of the characters may vary somewhat when going to italics and boldface:

Also depicted is the appropriate spacing for small capitals, which, of course, is not a subtle variation, but an entirely new series.

However, there are further variations in some fonts that might be worth accomodating, as illustrated above. Given, though, that the capital letters are not used as frequently as the lower-case letters, which have much less variation in widths from one typeface to another, it is understandable how it was possible to get away with one unit system for all typefaces.

Small Capitals

The appropriate proportions for small capitals are a question worthy of some examination. The Selectric Composer had available features that would allow it to vary the spacing of some letters, so that it could use Copperplate Gothic elements which had the capital letters in a smaller size as the lower case. The spacing involved was:

    Upper           Lower         Small
    Case            Case          Capitals

3                   ijl           I
4   I               frst          J
5   J               acefvz        LPSTZ
6   PS              bdhknopquxy   ABCDEFGHKNOQRUVXY
7   BCEFLTZ                       MW
9   MW              m

Here, the ratio of the two sizes appears to be about 3:4.

But Copperplate Gothic might not be a good example to use. To get a trustworthy example for the proportions of small capitals in a normal text face, and for that matter to address the question of italics and boldface, it makes sense to resort to the Monotype matrix-case; one example being that for Times New Roman, series 327:

   Normal                         Italics                Bold

   Upper   Lower       Small      Upper     Lower        Upper   Lower
   Case    Case        Capitals   Case      Case         Case    Case

 5          ijl        I                    ijl                  ijl
 6          ft         J                    ft                   ft
 7 I        rs         S          I         rs           I       rs
 8 J        cez        EF                   cevz                 cevz
 9          agvxy      BLPT       J         abdghnopuy           abdghnopquy
10 S        bdhknopqu  ACOQRVXYZ            kx           J       kx
11 P                   GDHKNU     FS                     FS
12 BEFLTZ              M          BELPT     w            BELPT   w
13 CV       w                     AGKRVXYZ               ACKRXZ
14 AOQRXY                         DNOQU     m            DGNU    m
15 DGHKUN   m          W          H                      HOQ
18 MW                             MW                     MW

Uniform Proportions

On this page, I note that, prior to the adoption of a uniform point system for printing, one James Fergusson adopted a scheme which approximated the old named type sizes by dividing a unit of fourteen lines of Nonpareil type, or seven picas, into different parts, just as the Selectric Composer divided the pica em into different numbers of parts.

In order to allow small type sizes to be handled, the actual unit that would have to be divided would be one of 42 points, of which the circumference of the roller would have to be a multiple. Since dividing it into 6 parts yields 7 points, and into 7 parts yields 6 points, those two of the six possible sizes would not require additional gears, so only four additional gears would be left:

(7 points)  Minion    (7 points)  x2 English       x4 Two-line English
(6 points)  Nonpareil (6 points)  x2 Pica          x4 Double Pica
 8 parts    Agate     (5.25)      x2 Small Pica    x4 Two-line Pica
 9 parts    Pearl     (4.66667)   x2 Long Primer   x4 Paragon
10 parts    Diamond   (4.2)       x2 Bourgeois     x4 Great Primer
11 parts    Brilliant (3.81818)   x2 Brevier       x4 Columbian

In order to match this up with the horizontal divisions, so that no rescaling of typefaces would be required between sizes, one could use the 1/72" unit for English, the 1/84" unit for Pica, the 1/96" unit for Small Pica. Since 11 point type, rather than 14 point type, was used with the 1/72" unit, this would mean a unit system that was more finely grained for those sizes of types; for smaller sizes, though, the number of units to the em would need to be divided in half, so instead of 11 units to the em, either 14 or 7 units would be used.

The fact that these scales would only coincide every 42 points would create a limitation in their use. One way this difficulty could be reduced is this: if the normal gear for advancing the platen according to the point system operates in steps of 1/144", then it would not be unreasonable for the four additional gears to divide the basic 42 point space not into 8, 9, 10, and 11 parts respectively, but into 56, 63, 70 and 77 parts. One would still need 11 lines of Brevier to add up to exactly 84 points, but now those 11 lines could start at any position that was a multiple of 6 points. As well, the diameter of the roller would now only have to be a multiple of 6 points instead of a multiple of 42 points, so that a standard typewriter roller could be used without a change in its diameter.

Also, while the system of James Ferguson appealed instinctively to me as something that might have been likely to be used as the basis for type heights before the point system was used, further study of the old sizes of type has led me to the conclusion that, instead, it seems like the pica was simply divided into a greater number of equal parts than the twelve points we use now. If the pica is divided instead into thirty parts, the sizes from Pearl to Pica seem to be approximated closely.

[Next] [Up] [Previous] [Home] [Other]