Instead of a "basket" of pivoting typebars the Selectric had a pivoting type element (frequently called a "typeball") that could be changed so as to display different fonts in the same document, resurrecting a capacity that had been pioneered by the unsuccessful Blickensderfer typewriter sixty years before. The Selectric also replaced the traditional typewriter's moving carriage with a paper roller ("platen") that stayed stationary while the typeball and ribbon mechanism moved from side to side.
Selectrics and their descendants eventually captured 75 percent of the United States market for electric typewriters used in business.
The ability to change fonts, combined with the neat regular appearance of the typed page, was revolutionary, and marked the beginning of desktop publishing. Later models with dual pitch (10/12) and built-in correcting tape carried the trend even further. Any typist could produce a polished manuscript. By 1966, a full typesetting version with justification and proportional spacing was released.
The possibility to intersperse text in Latin letters with Greek letters and mathematical symbols made the machine especially useful for scientists writing manuscripts that include mathematical formulas. The typical look of Selectric typed documents is hence still familiar to any scientist who reads conference proceedings, monographies, theses and the like from these times. (Proper mathematical typesetting was very laborious before the advent of TeX and done only for much-sold textbooks and very prestigious journals.)
The machine had a feature called "Stroke Storage" that prevented two keys from being depressed simultaneously. When a key was depressed, an interposer, beneath the keylever, was pushed down into a slotted tube full of small metal balls (called the "compensator tube") and spring latched. These balls were adjusted to have enough horizontal space for only one interposer to enter at a time. If a typist pressed two keys simultaneously both interposers were blocked from entering the tube. Pressing two keys several milliseconds apart allows the first interposer to enter the tube, tripping a clutch which rotated a fluted shaft driving the interposer horizontally and out of the tube, making way for the second interposer to enter the tube some milliseconds later. While a full print cycle was 65 milliseconds this filtering and storage feature allowed the typist to depress keys in a more random fashion and still print the characters in the sequence entered. The powered horizontal motion of the interposer selected the appropriate rotate and tilt of the printhead for character selection.
The spacebar, dash/underscore, index, backspace and line feed repeated when continually held down. This feature was referred to as "Typamatic."
The Selectric typewriter was first introduced on 23 July 1961. Its industrial design is credited to influential American designer Eliot Noyes. Noyes had worked on a number of design projects for IBM; prior to his work on the Selectric, he had been commissioned in 1956 by Thomas J. Watson, Jr to create IBM's first house style — these influential efforts, in which Noyes collaborated with Paul Rand, Marcel Breuer, and Charles Eames, have been referred to as the first "house style" program in American business.
Both Selectric and the later Selectric II were available in standard, medium, and wide-carriage models and in various colors, including red and blue as well as traditional neutral colors.
Mechanically, the Selectric borrowed some design elements from a toy typewriter produced earlier by Marx Toys. IBM bought the rights to the design. The typeball and carriage mechanism was similar to the design of the Teletype Model 26 and later, which used a rotating cylinder that moved along a fixed platen.
The mechanism that positions the typing element ("ball") is partly binary, and includes two mechanical digital-to-analog converters, which are basically "whiffletree" linkages of the type used for adding and subtracting in linkage-type mechanical analog computers. Every character has its own binary codes, one for tilt and one for rotate.
When the typist presses a key, it unlatches a metal bar for that key. The bar is parallel to the side of the mechanism. This bar has several short projections ("fingers"). Only some of the fingers are present on any given code bar, those present corresponding to the binary code for the desired character.
When the key's bar moves, its projections push against a second set of bars that extend all the way across the keyboard mechanism; each bar corresponds to one bit. All bars for the keys contact some of these crosswise bars. Those bars that move, of course, define the binary code.
The bars that have been moved cause cams on the driveshaft (which is rotating) to move the ends of the links in the whiffletree linkage, which sums (adds together) the amounts ("weights") of movement corresponding to the selected bits. The sum of the weighted inputs is the required movement of the typing element. There are two sets of similar mechanisms, one for tilt, one for rotate. The reason for this is the type element has four rows of 22 characters. By tilting and rotating the element to the location of a character, the element could be thrust against the platen, leaving an imprint of the chosen character.
The motor at the back of the machine drove a belt connected to a two-part shaft located roughly halfway through the machine. The Cycle Shaft on the left side provided the energy that was used to tilt and rotate the type element. The Operational Shaft on the right side provided functions such as spacing, back spacing and case shifting. Additionally, the Op Shaft was used as a governor; limiting the left-to-right speed with which the carrier moved. A series of spring clutches were used to power the cams which provided the motion needed to perform functions such as backspacing. The Cycle Shaft was rotated when a spring clutch was released, driving a set of cams whose rotational motion was then converted into left-and-right motion by the whiffle tree. The system was highly dependent upon lubrication and adjustment and much of IBM's revenue stream came from the sale of Service Contracts on the machines. Repair was fairly expensive, so maintenance contracts were an easy sell.
The locations of the characters on the element was not random. Punctuation marks and the underscore were deliberately placed so the maximum amount of energy was used to position the element, thus reducing the impact made by them and lessening the chance that the underscore would cut through the paper. Later on, a deliberate mechanism was added that reduced the force of the impact made by punctuation.
Tilt and rotate movements are transferred to the ball carrier, which moves across the page, by two taut metal tapes, one for tilt and one for rotate. The tilt and rotate tapes are both anchored to the right side of the carrier (the metal contraption upon which the type element is located). They both wrap around separate pulleys at the right side of the frame. They then wrap behind the carrier the are wrapped around two separate pulleys at the left side of the frame. The tilt tape is then anchored to a small, quarter-circle pulley which, through a gear, tips the tilt ring two one of four possible locations (The tilt ring is the device to which the type element is connected). The rotate tape is wrapped around a spring-loaded pulley located in the middle of the carrier. The rotate pulley under the tilt ring is connected through a universal joint (called a "dog bone"; it looked like a small bone) to the center part of the tilt ring. The type element is sprint-latched onto that central post. The type element rotates counter-clockwise when the rotate tape is tightened. The clock spring underneath the rotate pulley rotates the element in the clockwise direction. As the carrier moves across the page (such as when it returns), the tapes travel over their pulleys, but the spring-loaded pulleys on the ball carrier do not pivot or rotate.
To position the ball, both of the pulleys on the left side of the frame are moved by the whiffletree linkage. When the rotate pulley is moved to the right or left, the rotate tape spins the type element to the appropriate location. When the tilt pulley is moved, it tips the tilt ring to the appropriate location. When it moves, the tape rotates the spring-loaded pulley on the ball carrier independent of the carrier's location on the page.
Case was shifted between caps and lower case by rotating element by exactly half a turn. This was accomplished by moving the right-hand rotate pulley using a cam mounted on the end of the operation shaft.
There was a proportional-spacing Selectric called a Composer that would backspace proportionally for perhaps 40 characters. The spacing code for the last characters typed was stored by small sliding plates in a carrier wheel.
After the Selectric II was introduced a few years later, the original design was designated the Selectric I. These machines used the same 88-character typing elements. However they differed from each other in many respects:
In addition to the "typeball" technology, Selectrics were also associated with a series of innovations in ribbon design. The original Selectric had to be ordered to use either cloth reusable ribbon or one-time carbon film ribbon; the same machine could not use both. The same was true of the original, non-correcting Selectric II. IBM had used a similar carbon film ribbon on their earlier "Executive" series of typewriters. As with these older machines, the carbon film ribbon presented a security issue in some environments: It was possible to read the text that had been typed from the ribbon, seen as light characters against the darker ribbon background.
The "Correctable" nature of the Correcting Selectric II's carbon film ribbons had an additional issue in that the carbon pigment could easily be removed from a typed document, thus facilitating unauthorized changes.
The Correcting Selectric II used a new ribbon cartridge mechanism. The ribbons were wider than had been used previously, giving more typed characters per inch of ribbon. Successive characters were staggered vertically on the ribbon, which incremented less than a full character position each time. Any Correcting Selectric II could use any of three types of ribbon, which all came in similar-looking cartridges: Reusable cloth ribbon with associated Cover-Up tape; Correctable (carbon) Film ribbon with associated Lift-Off tape; and the Tech-3 permanent ribbon, introduced later, which used the same Cover-Up tape as the earlier cloth ribbon. The Tech-3 ribbon essentially replaced the cloth ribbon, as they offered similar typing quality to the film ribbon but at a cost comparable to the reusable cloth.
Tech-3 ribbons provided much higher security and longer life than the Correctable Film ribbon. Like the cloth ribbon, Tech-3 ribbons incremented only a fraction of the character width after being struck. Unlike the cloth ribbon, the Tech-3 ribbon provided high quality impressions for several characters from each spot on the one-time-use ribbon. Because characters overstrike each other on a Tech-3 ribbon several times it could not be easily read to discover what had been typed.
In addition, where the Correctable Film ribbon was unsuitable for documents such as checks due to the ease of lifting the ink from the document, the Tech-3 ribbon's impressions were permanent as soon as they were struck. Some colored ribbons (e.g. brown) were also available.
There were four classes of carbon film ribbons available for the Selectric II series. The thumb wheel on the ribbon and the correction tape spools were color coded so they could be easily identified and matched with the appropriate correction tapes. There were two lift-off correctable ribbons/correction tapes, one color coded yellow and the other orange. Yellow meant the ribbon was a higher quality and would produce a better quality type image. Orange was a general purpose ribbon for everyday typing. The yellow and orange coded lift-off tapes would work with either ribbon type because they were both sticky (similar to adhesive tape) and would pull the ink off the paper. Later there was a less "sticky" version of these lift of tapes that wouldn't damage more delicate paper surfaces but some people believed it didn't remove the ink as well. As a side note, if you ran out of lift-off tape, you could use a piece of adhesive tape (such as Scotch tape) to correct a mistake.
The Tech-3 (Tech III) ribbons described above were color coded blue and the high quality carbon film ribbon was color coded pink. The pink coded ribbons could be used for the more sensitve documents because the ink was not easily removable from the paper and it gave a clearer/crisper image than the Tech-3 ribbons. The correction tapes for these covered up the typewritten characters with white ink. This complicated corrections on paper colors other than white.
In 1964 IBM introduced the "Magnetic Tape Selectric Typewriter" and in 1969, a "Magnetic Card Selectric Typewriter." These were sometimes referred to as the "MT/ST" and "MC/ST", respectively. The MC/ST was also available in a "communicating" version that emulated an IBM 2741 terminal. These featured electronically-interfaced typing mechanisms and keyboards and a magnetic storage device (either tape in a cartridge, or a magnetic-coated card the same size as an 80-column punch card) for recording, editing, and replaying typed material at c. 12-15 characters per second.
These machines were among the first to provide word processing capability in any form. They used the same elements as ordinary office Selectrics.
In 1966, IBM released the Selectric Composer. This highly modified Selectric produced camera-ready justified copy using proportional fonts in a number of font sizes and styles. Like the Varityper with which it competed, the machine required that material be typed twice if the type was to be justified. The first time was to measure the length of the line and count the spaces, recording special measurements on the right margin. The second time it was typed, the operator used the measurements to set justification for each line. The process was tedious and lengthy but provided a way to get camera-ready, proportionally spaced, justified copy from a desktop machine. The elements for the Selectric Composer would fit on a Selectric, and vice versa, but they could not actually be used on each other's machines: the characters were arranged differently around the element and were also positioned differently within each character area. Selectric Composer elements can be identified by a colored index arrow (the color is used to set a median character width on the machine) and a cryptic series of letters and numbers identifying the font, size, and variation, for example "UN-11-B" for Univers 11 point bold (Adrian Frutiger had adapted his Univers font specifically for the Selectric Composer).
In 1967, a "Magnetic Tape Selectric Composer" appeared, and in 1978, a "Magnetic Card Selectric Composer." The "Electronic Composer" (with c. 5000 characters internal memory and similar to the later Magnetic Card model but without external storage) was marketed from 1975. All these models used the same elements and measurement mechanism as the previous Selectric Composer; due to the magnetic/internal storage, they avoided the need to type justified text twice or to set the mechanism for the justification needs of each line. Furthermore, tapes or cards originally recorded on the much less expensive and easier to operate "Selectric" versions, the MT/ST or MC/ST, could be read by the "Composer" equivalents.
In the 1980s IBM introduced a Selectric III and several other Selectric models, some of them word processors or typesetters instead of typewriters, but by then the rest of the industry had caught up with the trend, and IBM's new models did not dominate the market the way the first Selectric had. This was to be expected, as by the late 1970s the Selectric typewriter's dominance was under assault from both 35-45 character per second proportional-spacing electronic typewriters with inbuilt memory (e.g. the 800 from Xerox based on Diablo's 'daisywheels' and from OEMs of Qume who had a similar 'printwheel' technology) and CRT-based systems from AES, Lexitron, Vydek, Wang and Xerox (see the Word Processor article for further details of these brands). In addition, IBM had already (c. 1977) brought to market the CRT-based Office System/6 (from Office Products Division) and 5520 (from IBM GSD) both of which used the new 6640 inkjet printer capable of 96 characters per second with two paper trays and sophisticated envelope handling, and was about to introduce Qume-based printers for the existing System/6 range and the new Displaywriter launched in June 1980 and described by IBM as "not your father's Selectric."
Nevertheless, IBM had a large installed base of Selectric typewriters and to retain customer loyalty it made sense to introduce updated models.
The Selectric III featured a 96 character element vs. the previous 88 character element. IBM's series of "Electronic Typewriters" used this same 96 character element. The 96 character elements can be identified by yellow printing on the top plastic surface and the legend "96," which always appears along with the font name and pitch. The 96 and 88 character elements are mechanically incompatible with each other (they won't fit on each others' machines) and 96 character elements were not available in as many fonts as the older 88 character types.
Most Selectric IIIs and Electronic Typewriters only had keys for 92 printable characters; the 96 character keyboard was an optional feature. Fitting the additional keys onto the keyboard required shrinking of the Return key and this was annoying to many typists, so it was not the default configuration. The keytops on the Selectric III and Electronic Typewriters were larger and more square than those on earlier Selectrics.
The Selectric I, Selectric II, and all of the "Magnetic Card" and "Magnetic Tape" variations except for the Composers, used the same typing elements. These were available in many fonts, including symbols for science and mathematics, OCR faces for scanning by computers, cursive script, "Old English" (fraktur), and more than a dozen ordinary alphabets. The Selectric III and "Electronic Typewriters" used a new 96-character element.
There were two visibly different styles of mechanical design for the elements. The original models had a metal spring clip with two wire wings that were squeezed together to release the element from the typewriter. Later models had a fragile flip-up black plastic lever that could break off. This was later redesigned to have a substantial plastic lever that did not break.
Some of the interchangeable font elements available for the Selectric models included:
Starred fonts were 96-character elements made for the Selectric III.
Many of the fonts listed here came in several sub-varieties. For example, in the early years of the Selectric, typists were used to using the lower-case L for the numeral 1. The Selectric had a dedicated key for 1/!, but this was also marked [/], and many of the early elements had square brackets in these positions, necessitating that the typist continue the old convention. Later elements tended to have the dedicated numeral 1 and exclamation point characters instead. Some moved the square brackets to the positions formerly occupied by the 1/4 and 1/2 fractions, while others lost them completely. Some put a degree symbol in place of the exclamation point. IBM would furthermore customize any element for a fee, so literally endless variations were possible. Such customized elements were identified by a gray plastic flip-up clip instead of a black one.
Many specialized elements were not listed in IBM's regular brochure, but were available from IBM provided the right part number was known. For example, the element for the APL programming language was available. This element was really intended for use with the IBM 2741 printing terminal.
Despite appearances, these machines were not simply Selectric typewriters with an RS-232 connector added. A Selectric is a marvel of mechanical and production—but not electronic—engineering. As with other electric typewriters, and electric adding machines of the era, Selectrics are best thought of as electromechanical devices: The only electric components are the power cord, power switch, and electric motor. The electric motor runs continuously. The keys are not electrical pushbuttons, as they are on a computer keyboard. Pressing a key does not produce an electrical signal, but rather engages a series of clutches which couple the motor power to the mechanism to turn and tilt the element. A Selectric would work equally well if hand-cranked at sufficient speed.
Adapting this mechanism to the needs of computer input/output was nontrivial. The keyboard and printing mechanism were mechanically separated (so that keystrokes do not necessarily result in immediate printing), microswitches were added to the keyboard, solenoids were added to allow the computer to trigger the typing mechanism, and interface electronics were needed. Several mechanical components, in particular the motor and the main clutch, had to be upgraded from the typewriter versions to reliably support continuous operation. Additional microswitches had to be added to sense the state of various parts of the mechanism, such as case (upper vs. lower).
Even after adding all those solenoids and switches, getting a Selectric to talk to a computer was a large project. The Selectric mechanism, as documented in its service manual, had many peculiar requirements. If commanded to shift to upper case when it was already in upper-case, the mechanism locked up and never signaled "done". Same thing for shifting the ribbon direction or initiating a carriage-return. These commands could only be issued at particular times, with the Selectric in a particular state, and then not again until the terminal signaled the operation was complete.
In addition the Selectric spoke neither ASCII nor EBCDIC, but a unique code based on the tilt/rotate commands to the golf ball. That and the bit-parallel interface and peculiar timing requirements meant the Selectric could not be directly hooked up to a modem. Indeed it needed a relatively large amount of logic to reconcile the two devices.
Particularly vexing was the Selectric's lack of a full ASCII character set. The late Bob Bemer wrote that while working for IBM he lobbied unsuccessfully to expand the typing element to 64 characters from 44. The Selectric actually provided 44 characters per case, but the point remains that with 88 printable characters it could not quite produce the full printable ASCII character set.
Nevertheless, between 1968 and about 1980, a Selectric-based printer was a relatively inexpensive and fairly popular way to get high-quality output from a computer.
The 96-character element introduced with the Selectric III and Electronic Typewriter series could (with some customizations) handle the full ASCII character set, but by that time the computer industry had moved on to the much faster and simpler daisy wheel mechanisms such as the Diablo 630. The typewriter industry followed this trend shortly afterward, even IBM replacing their Selectric lineup with the daisy wheel-based "Wheelwriter" series.