Keys and Keyboards

Keys and Keyboards

Handbook of Human-Computer Interaction M. Helander (ed.) © Elsevier Science Publishers B.V. (North-Holland), 1988 Chapter 21 Keys and Keyboards 21.1...

3MB Sizes 0 Downloads 16 Views

Handbook of Human-Computer Interaction M. Helander (ed.) © Elsevier Science Publishers B.V. (North-Holland), 1988

Chapter 21

Keys and Keyboards 21.1

Kathleen M. Potosnak The Koffler Group Santa Monica, California

Introduction

Keyboards have been around for over 100 years and are in widespread use both on typewriters and as in­ put devices to computers. Although market forces and convention have kept the basic design of keyboards fairly constant over the years, the keyboard as we know it today is still undergoing changes. Refinements of the typewriter keyboard initially were aimed at improving its mechanical action so that it would operate more smoothly with fewer malfunc­ tions. Technology solved those mechanical problems decades ago. Later work was aimed at improving typ­ ing speed and accuracy. Most recently, the widespread use of computer terminals raised new interest in key­ board design. Scientists shifted their concerns to typist fatigue, comfort and muscular strain. This chapter focuses on design factors which af­ fect skilled typing and data entry. The information presented should apply equally well to typewriter and computer keyboards. Some data also apply to tele­ phones and other specialized keypads that are used for data entry tasks. Physical characteristics of the key­ board as well as details of alphabetic and numeric typ­ ing functions are discussed. Specifically, this chapter covers:

21.1 Introduction Design Criteria

475 476

21.2 Keyboard Layouts The QWERTY Layout The Dvorak Simplified Keyboard Layout Conclusions on Dvorak Simplified Keyboard Alphabetical Keyboards Other Keyboard Layouts

476 476 476

21.3 Data-Entry Keypads Layout of Numbers and Letters Multifunction Keypads

479 480 480

2L4 Physical Features of Keys and Key­ boards Keyboard Height and Slope Size of the Keyboard Detachable Keyboards Keyboard Profile Key Size and Shape Key Force, Travel and Tactile Feedback Auditory Feedback Visual Feedback Error-Avoidance Features Color and Labeling

481 481 482 482 483 483 484 485 486 486 487

21.5 Innovations in Keyboard Design Split and Tilt Keyboards New Methods of Typing

487 487 490

Data-Entry Keypads

21.6 Summary

491

• multifunction keypads

21.7 Acknowledgements

491

Physical Features of Keys and Keyboards

21.8 References

491

• keyboard height and slope

478 479 479

Keyboard Layouts • QWERTY • Dvorak • alphabetical • other layouts

• layout of numbers and letters

• keyboard size 475

CHAPTER 21. KEYS AND KEYBOARDS

476 • detachable keyboards • keyboard profile • key size and shape • key force, travel, and tactile feedback • auditory feedback • visual feedback • error-avoidance features (rollover, hysteresis, in­ terlocks, buffer length, repeat features) • color and labeling Innovations in Keyboard Design • split and tilt keyboards (K-keyboard, STR key­ board, Maltron keyboard) • new methods of typing (one- and two-handed chord keyboards, wipe-activated keyboard) Topics that are not covered include word processing functions, programmable keys, and virtual keyboards (i.e., those whose functions vary according to the lay­ out presented on a screen display) due to the lack of research in these areas. Design Criteria Typing speed, error rate, fatigue, muscular strain, and preferences are all reasons for selecting one keyboard design over another. Because these criteria for defin­ ing keyboard quality do not necessarily result in con­ sistent recommendations, there will always be trade­ offs in keyboard design. For example, a keyboard de­ signed for high performance typing might eventually be preferred by typists, but might also induce more fatigue or muscular strain, so that more frequent rest breaks would be required. Thus, the first step in key­ board development or selection should be a determi­ nation of which criteria are most important. Preferential factors often outweigh performance dif­ ferences between keyboards. Experienced typists strongly resist changing their methods of typing, even if they are convinced that it could improve their out­ put. So, for example, the best keyboard layout is one that meets typist expectations and permits the use of well- learned motor skills. As long as they are using a QWERTY layout (or whatever is most familiar), typists are very adaptable to other design modifications. This ability to adapt to a design makes it difficult to find short-term per­ formance differences between, for example, key top

shapes or keyboard slopes and profiles. Even though typists can cope with a wide variety of keyboards so that their typing performance does not suffer, they do have strong preferences for particular designs. More­ over, adaptation usually comes at a cost, so there may be long-term effects. For instance, maintaining high performance may be more fatiguing on one keyboard design than another. However, differences in fatigue may not become apparent until the typist has been working for more than a few hours.

21.2 Keyboard Layouts The locations of letters and numbers on keys has been a matter of research, theory, debate, contests and patent applications since the appearance of the first conventional typewriter keyboard. Although other typewriters are known to have existed previously, the design patented in 1868 by Sholes, Glidden and Soule is generally accepted as the first to include many of the characteristics of modern typewriters (Yamada, 1980). The letters were originally arranged alphabetically. The QWERTY Layout Fast typists ran into trouble with the early design of the Sholes keyboard because the typebars of successive keystrokes would interfere with each other. The cur­ rent QWERTY layout (named for the top left-most row of letters) was developed to increase the spacing be­ tween common pairs of letters so that the sequentially struck typebars would not jam. The first patent show­ ing the QWERTY layout appeared in 1878 (Cooper, 1983; Noyes, 1983b)1. The Dvorak Simplified Keyboard Layout While the QWERTY arrangement was developed for mechanical reasons, rapid advances in keyboard tech­ nology and typing methods made it possible to im­ prove on the original design (Yamada, 1980). Among the suggested modifications were those that sought to rearrange the letters of the alphabet to increase typing speed and accuracy. The most well-known of these attempts was that made by August Dvorak, who de­ veloped the Simplified Keyboard (known as DSK) and received a U.S. patent for his design in 1936. The Dvorak layout was designed along principles of time-and-motion study and scientific measurement 1

There are other theories about how QWERTY came into exis­ tence. For a summary, see Noyes (1983b). Alleviation of typebar jamming problems was the most frequently cited explanation in the literature.

212. KEYBOARD LAYOUTS of efficiency (Dvorak, 1943). Dvorak also based his analyses on touch-typing performance. QWERTY was designed for two-finger typing. Touch typing was not generally accepted until around 1900 (Yamada, 1980). The principles underlying Dvorak's layout included such concepts as: simple motions are easier to learn and perform rapidly than more complex motions, and rhythmic motions are less fatiguing that erratic ones. With the DSK layout, the right hand is used more than the left, and the fingers are assigned proportionate amounts of work. Almost (70%) of the typing is done on the home row, which reduces hurdling by as much as 90% 2 . The placement of vowels and frequentlyused consonants on opposite halves of the keyboard also allows for more two-handed typing sequences. All of these figures are based on the letter frequencies in English text. Many experiments, field trials and analytical studies have been carried out to compare the Dvorak Simpli­ fied Keyboard with the QWERTY arrangement. Dvo­ rak conducted some of his own tests with reportedly positive results. Four studies comparing DSK and QWERTY keyboards are discussed below. The Navy Department study. In the 1940s, the Navy Department compared two groups of typists who were given on-the-job retraining (U.S. Navy Depart­ ment, 1944a, 1944b). The first group consisted of QWERTY-trained typists who were retrained on the DSK layout. The second group of QWERTY typ­ ists was given additional training on the standard key­ board. Increases in typing speeds and decreases in error rates were higher for the DSK group. However, the gain in net words per minute (n.w.p.m.) was not sta­ tistically significant. Also, there were pre-existing dif­ ferences between the two groups: the DSK typists ini­ tially were faster on QWERTY than the other group. The Navy Department report does not focus on the final n.w.p.m., but instead describes differences between the groups in terms of percentage gain in n.w.p.m. as a function of the number of hours of addi­ tional training. In measuring this learning rate, the re­ searchers used zero as the baseline for the DSK group because they had never used DSK before. The QW­ ERTY group baseline was their typing rate before ad­ ditional training. The use of different baselines affects 2

This number was calculated as the percentage of miles that a typist's fingers travel in a day. Dvorak figured that on the DSK *A typist's finger tips in a eight hour day travel only a little over a mile in motions from row to row as compared to 12 to 20 miles a day for a typist on the Standard [QWERTY] Keyboard' (Dvorak, 1943). There is currently a debate over the correctness of these calculations of finger travel {Dvorak Developments, 1986)

All

the calculation of learning rate. Also, the initial learn­ ing rate for the DSK group is expected to be quite high because some previously learned typing skills (such as finger movements) would be relevant for learning DSK. Later performance might not show such a rapid rate of learning. Navy Department researchers also calculated the costs and benefits of retraining compared to addi­ tional QWERTY training. However, the cost figure was "corrected" by subtracting the value of increased production during the latter part of the retraining pe­ riod for the DSK group only. That is, once typists in the DSK group exceeded their original QWERTY typing rates, the increase in production was given a dollar value. The correction factor was the number of hours each typist worked at greater than 100% of QWERTY typing speed multiplied by the individual's hourly wage. Without this correction factor, the av­ erage cost of retraining was actually cheaper per hour for additional QWERTY training (L27/hour DSK and 0.90/hour QWERTY). The Navy Department report concluded with highly favorable statements about DSK retraining and recom­ mendations for implementing such retraining. The fol­ lowing facts invalidate this conclusion: a) differences in final typing performance were not statistically sig­ nificant; b) measures of learning rates were biased in favor of the DSK group; and c) calculations of costs and benefits were also biased in favor of DSK. The Strong study. A study was conducted by Strong (1956) for the U.S. General Services Administration. Strong trained QWERTY typists on DSK keyboards until they had reached their previous QWERTY typ­ ing performance levels. This took an average of about 28 days. In the second part of the experiment, the DSK group was given additional instruction time to increase their speed and accuracy on DSK. A compa­ rable group of QWERTY typists began the experiment in the second half and was given only this additional training (but on QWERTY). After training, the QW­ ERTY group performed better on typing tests than the DSK group (Alden et al., 1972; Noyés, 1983b). Strong concluded that there were no advantages to the DSK and that "brush up" training on QWERTY was more effective (Yamada, 1980). Other researchers have questioned the Strong report. Some tried to obtain the original experimental data for re-evaluation, but were told that all data had been de­ stroyed. There has been speculation that the study was unfairly biased in favor of QWERTY (Noyes, 1983b) and that Strong himself was "hardly an unbiased inves­ tigator" (Yamada, 1980, p. 188). Regardless of what-

478 ever motivated Strong to write such a report, it clearly was a major blow to public acceptance of DSK and the adoption of DSK by the U.S. government (Cassingham, 1986; Yamada, 1980). A non-experimental comparison. It is difficult to conduct a fair experiment to compare DSK and QW­ ERTY because QWERTY is so widely used and this previous training can affect both DSK retraining and additional QWERTY training in unknown ways. To circumvent the methodological difficulties of train­ ing and retraining typists on each keyboard, Kinkead (1975) collected data on the times required for the fingers to type each possible sequence of two letters (called a "digram" or "digraph") on the QWERTY key­ board. The second part of the analysis was to obtain the frequency at which each digraph occurs in English. Kinkead assumed that the time to make a particular finger motion ("keystroke time") would be the same for both DSK and QWERTY. That is, the keys and rows are arranged the same on both keyboards, so the same finger motions are required. The only difference between layouts was the assignment of letters to the key locations and thus the frequency with which each finger motion would be used. The sum of all "digraph frequency x keystroke time" values was used as the estimate of typing speed for each keyboard layout. Kinkead concluded that at best DSK is 2.3% faster than QWERTY. (This value of 2.3% is shown below his table of calculations; in the text, the number is given as 2.6% faster.) There are some discrepan­ cies between these values and calculations based on the numbers in Kinkead's report. For example, us­ ing the same numbers as Kinkead, the advantage of DSK over QWERTY can be calculated at either 3.1% or 3.2%, depending on whether Kinkead's first (155 msec/keystroke) or second (151 msec/keystroke) esti­ mate of average keystroke time is used. Another difficulty in interpreting the Kinkead data stems from the effect of context on typing speed and "the levelling effect." Context and typing speed. The speed of typing a character depends on the context surrounding that character. This context is larger than the digraph. In general, the typing speed for one letter is influenced by the two characters before it and one character af­ ter it (Gentner, 1983). The levelling effect: Calcula­ tions for trigraphs (three- letter sequences) have shown that when a particularly slow keystroke is made, other keying sequences surrounding it are slowed down. Fast keying sequences tend to speed up surrounding keystrokes (Hiraga, Ono, and Yamada, 1980). This tendency toward maintaining a constant typing speed

CHAPTER 21. KEYS AND KEYBOARDS from one keystroke to the next is known as the level­ ling effect. It is believed to be a means by which the typist maintains typing rhythm. The context and levelling effects could explain why Kinkead obtained such a low estimate for DSK's ad­ vantage over QWERTY, since the analysis was based on digraphs only. A contributing factor may also have been the use of an electronic typewriter. The earlier data from Dvorak and others were most probably de­ rived from typing on manual typewriters, which are estimated to be about 9 words per minute slower than electronic typewriters (Alden et al., 1972). A computer simulation. Another comparison of DSK and QWERTY was performed by Norman and Fisher (1982) using "a computer simulation of the hand and finger movements of a skilled typist" (p. 514). Their model accounts for the context effect because it "allows for simultaneous movement of the fingers and hands toward different letters of the word being typed, thus capturing the parallel, overlapping move­ ments seen in high-speed films of expert typists" (p. 515). They calculated that DSK provides about a 5.4% advantage in typing speed over QWERTY. The model resulted in a typing rate of about 58 w.p.m. for DSK compared to about 56 w.p.m. for QWERTY. This study addresses some of the criticisms of the Kinkead report. The computer model did take into ac­ count more than just digraph frequencies in determin­ ing speed of finger motions. The conclusions, how­ ever, are limited only to speed. Although their model also simulates normal error rates for skilled typists, they did not report any differences in errors between DSK and QWERTY. Accuracy and fatigue are also important factors to consider in the evaluation of typ­ ing on different keyboards. Conclusions on Dvorak Simplified Keyboard Most studies confirm that DSK is faster than QW­ ERTY. However, there is disagreement on the size of the difference between the two keyboard layouts. Ear­ lier accounts claimed that DSK was anywhere from 15% to 50% faster than QWERTY (Yamada, 1980). More recent calculations give much smaller numbers ranging from 2.3% to 17% (Kinkead, 1975; Norman and Fisher, 1982; Yamada, 1980). Because there are so many unknowns (such as how long it will take a particular typist to retrain) a decision to switch to DSK must be based on estimated risks. With a conservative estimate of 5% to 10% increase in output over QWERTY, the switch may be cost ef­ fective for some typists, but essentially worthless for

479

213. DATA-ENTRY KEYPADS others. As for market trends, use of DSK is on the rise. One source (Cassingham, 1986) reports that the num­ ber of DSK typists has grown from 2,000 in 1974 to over 100,000 by 1985. Whether this trend continues may depend on the success of current changeovers and the results of research, such as that being conducted by the Dvorak International Federation (Dvorak De­ velopments, 1986). Alphabetical Keyboards Another method of designing a keyboard is to place letters on the keys in alphabetical order. Such a lay­ out is used on some children's toys and on a popular stockbroker's quotation terminal. A moment's consid­ eration will reveal that there are quite a few ways to create an alphabetical arrangement (e.g., ordered hor­ izontally versus vertically). Research comparing QWERTY and alphabetical keyboards. Hirsch (1970) tested one group of non-typists on QWERTY and another group on an alphabetically-arranged keyboard. After seven hours of practice, the QWERTY group improved their typing speed from 1.47 to 1.99 keystrokes per second. The Alphabetical group, however, did not even reach their pre-experiment QWERTY typing rates (1.47 compared to 1.11 keystrokes per second). Michaels (1971) expanded on the work of Hirsch by including people with a broader range of typing skills, ages and backgrounds. Results showed that both the high- and medium-skill groups were signifi­ cantly faster on QWERTY, while the low-skill group showed no significant difference in typing speed on the two keyboards. It was also found that skilled typists were faster at keying numerical sequences on QW­ ERTY than on the Alphabetical keyboard even though the number keys were exactly the same on both type­ writers. The levelling effect, as described in the above discussion of Kinkead's comparison of DSK and QW­ ERTY, is the tendency of fast keystrokes to speed up slow ones and of slow keystrokes to slow down fast ones. Since the speed of typing letters on the Al­ phabetical keyboard was so slow, the letter keystrokes might have slowed typing of numbers. Norman and Fisher (1982) tested non-typists on four different keyboards: QWERTY, Alphabetical (letters A through Z arranged from left to right starting at up­ per left of keyboard), Diagonal (also alphabetical, but with letters arranged from top to bottom and then from left to right starting at upper left of keyboard), and a Random keyboard (letters assigned to keys at random).

Typing was more that 65% faster on QWERTY than on any of the other layouts. Statistical tests revealed that the first three keyboards were all significantly bet­ ter than the Random arrangement and that QWERTY was better than the second two keyboards (the alpha­ betical arrangements were not significantly different from each other). A study of small keypads for an enhanced-telephone application compared QWERTY and Alphabetical lay­ outs (Francas, Brown, and Goodman, 1983). The size of the keypads limited typing to a one- or two-finger strategy. Average time for entering sentences was 54.4 seconds for QWERTY and 97.5 seconds for the Alpha­ betical layout. Only one person preferred the Alpha­ betical keypad over QWERTY. Conclusions. Alphabetically arranged keyboards provide no advantages over QWERTY. It is easier to resort to simple visual search of an alphabetical key­ board than it is to mentally figure out the position of the letter in the alphabet and then translate this into a keyboard location. Performance on QWERTY may be better than alphabetical or random keyboards even with a visual search strategy because many of the more commonly-used letters are centrally located, which re­ duces the area of search. Another possible explanation is that QWERTY keyboards are so common that most people have had at least some experience with one.

Other Keyboard Layouts Given the structure of the standard keyboard (three rows of letters with an upper row of numbers), there are many ways to arrange the alphabetic keys. To date, no other keyboard has received more recogni­ tion than DSK as an alternative to QWERTY. Most proposed keyboards have never been tested and quite a few never made it past the prototype stage. Other than as an academic exercise, the redesign of the QW­ ERTY layout appears to be a fruitless effort.

21.3

Data-Entry Keypads

In addition to the main alphanumeric section, most computer keyboards incorporate a separate, numeric keypad for data entry. Also, use of push-button tele­ phones as remote terminals to computers is on the rise (see Bayer and Thompson, 1983; Hagelbarger and Thompson, 1983). This section considers the design of keypads for telephones and other applications.

480 Layout of Numbers and Letters Lutz and Chapanis (1955) tested six key configurations to determine where people expected each letter and number to appear on ten-button keysets that were to be used by long-distance telephone operators. Keys were arranged in either two horizontal rows of five keys, two vertical rows of five keys, or three rows of three keys with a single key placed at the top, bottom, left or right of the block of nine keys. In general, people placed letters and numbers on keys in the same order as they read text (that is, from left to right and from top to bottom) regardless of the key configuration. When numbers were already on the keys: a) people placed letters on the keys from left to right and from top to bottom if the numbers were arranged that way; and b) if the numbers were not arranged from left to right and top to bottom about half of the people placed the letters to be consistent with the ordering of the numbers, while the other half persisted in arranging the letters from left to right and from top to bottom even though this conflicted with the number arrangement. The most frequent number arrangement was that found on present-day push button telephones. Later studies have compared performance with the so-called telephone layout (1, 2, 3 across the top with zero at the bottom) and the common calculator lay­ out (7, 8, 9 across the top). Conrad and Hull (1968) asked housewives to enter numeric codes and found the telephone layout was superior in both speed and accuracy. Paul, Sarlanis and Buckley (1985) tested air traffic controllers and found that the telephone layout was better for entry of letters and mixed (letter and number) data, but that performance was the same for entry of numbers only. A related study (Goodman, Dickinson, and Francas, 1983) used a simulation methodology to deter­ mine the best layout of keys for Telidon (Canada's Videotex system) keypads. They tested both a horizon­ tal (1 through zero in a single row) and a telephone arrangement. Speed and accuracy were slightly better on the telephone layout in a reaction-time task. How­ ever, differences in performance were not statistically significant. Preferences were strongly in favor of the telephone arrangement for a problem- solving task that simulated expected use of the Telidon system, but were slightly in favor of the horizontal arrangement for the reaction-time task. Conclusions. The design and use of data-entry key­ pads depends to a great extent on the keying task re­ quired. For numeric input, the telephone layout is slightly superior to the calculator layout, especially for people who are not familiar with calculators or adding

CHAPTER 21. KEYS AND KEYBOARDS machines. It is expected that there will be increased use of the telephone layout on computer keyboards be­ cause of proposed additions of telephone-related func­ tions to computer terminals. Anyone who has tried to dial a telephone number on a non-telephone keypad (such as an auto-dialer or computer keyboard) will ap­ preciate the inclusion of a telephone layout for this and similar functions. Multifunction Keypads The data-entry procedure that is used with multifunc­ tion keypads (such as telephones) is also of ergonomie importance. Because telephones and other limited key­ sets contain more than one letter on each key, a strat­ egy is required for designating which letter is meant for each keypress. Davidson (1966) suggested using the two extra keys on push-button telephones as "control" keys. The left (*) button would indicate the first letter, the right (#) button would indicate the third letter, and a keypress without a control key would indicate the middle letter. One experiment (Francas, Brown, and Good­ man, 1983) compared Davidson's suggested method with both a miniature QWERTY keypad and an alphabetically-arranged one. For entering alphabetic data, the multifunction keypad with left and right con­ trol keys was significantly slower than either the QW­ ERTY or alphabetical keysets, but accuracy was the same for all entry methods. The authors concluded that the multifunction keypad was not suitable for en­ tering letters in their telephone application. However, for tasks that primarily require accuracy and have se­ vere space limitations, a multifunction keypad with control keys might be acceptable. A study that is relevant to aircraft cockpits compared three keypads which required one-handed keying (Butterbaugh and Rockwell, 1982). Task performance was fastest and most accurate with a keyboard having sep­ arate keys for each letter and number. However, fur­ ther analyses showed that this difference was mainly due to the fact that the two other (multifunction) key­ pads required more keystrokes to enter each letter. Keypressing speed was actually fastest with a tele­ phone layout that had letters assigned horizontally to the number keys. These investigators recommend that the full key­ board should be used if at all possible. If space re­ quirements necessitate the use of smaller keypads, they suggest a telephone layout with letters assigned from left to right and from top to bottom. They additionally recommend using three control keys (for left, middle

21 A. PHYSICAL FEATURES OF KEYS AND KEYBOARDS and right characters) located in the top row of keys. Conclusions. For the entry of letters or mixed al­ phanumeric data, small multifunction keypads are no match for full-sized QWERTY or alphabetical key­ boards. Even for telephone applications, separate number and letter keys are an advantage. If speed is not important, but accuracy and space are critical, then a combined function keypad might be acceptable. In such cases, the layout of the keys and the proce­ dures for entering data should be carefully designed and tested for compatibility with the requirements of the task.

21.4 Physical Features of Keys and Key­ boards Experimental investigations of physical aspects of keys and keyboards are presented in this section. Because the research has not provided data on how the features might interact with each other in affecting typing per­ formance, each aspect is discussed separately. Keyboard Height and Slope An experiment was conducted to determine the best slope for keysets used by long-distance telephone op­ erators (Scales and Chapanis, 1954). People who had no previous experience with this keyset entered se­ quences of letters and numbers with the keyset sloped at either zero, 5, 10, 15, 20, 25, 30 or 40 degrees for each session. There were no differences in speed or errors among the eight slope conditions. All partic­ ipants preferred some slope over a flat (zero degree) keyset; half of them preferred an angle between 15 and 25 degrees. A computer keyboard was tested at 9, 21 and 33 degrees (Galitz, 1965). While there were no perfor­ mance differences due to keyboard slope, typists pre­ ferred the 21-degree angle. This slope was closest to the 16- to 17-degree angle on equipment the typists normally used. It was recommended that computer keyboards have a slope adjustable between 10 and 35 degrees to satisfy individual preferences. Emmons and Hirsch (1982) compared three slopes for an IBM 30- mm keyboard. Since a change in an­ gle also resulted in a different home-row height, their height (angle) settings were: 30 mm (5 degrees), 38 mm (12 degrees), and 45 mm (18 degrees). They also used a non-IBM 30 mm (5 degrees) keyboard. Tests of 12 experienced typists revealed no differences in error rates among the three angles/heights. With re­ gard to typing speed, the 38-mm and 45-mm keyboards

481

resulted in faster rates than either of the 30-mm key­ boards. Typists preferred the 45-mm keyboard most, and the 38-mm keyboard second-most. When asked about discomfort, five participants found "everything uncomfortable", five said the 30-mm keyboards caused discomfort, one reported the 45-mm keyboard as un­ comfortable, and one person had no discomfort with any of the keyboards. Miller and Suther (1981; 1983) examined pre­ ferences for keyboard slope and height. U.S. and Japanese participants from the 5th, 50th and 95th percentiles in height (compared to their respective pop­ ulations) were asked to adjust a workstation to their preferred settings and then transcribe some text from a written document into a computer terminal. Keyboard slope settings ranged from 14 to 25 de­ grees with an average of 18 degrees. Preferred slope was significantly correlated with seat height (r=-.71) and with individual stature (r=-.43). Short people or people who preferred lower seat heights also liked to have a keyboard with a steeper slope. Because stature is known to be correlated with hand length, a steeper slope makes it easier for short-handed people to reach all of the keys. They recommended that keyboard slope be adjustable up to at least 20 degrees (25 de­ grees is even better) to suit individual preferences. The keyboard used in the study was 77 mm high. Preferred home-row height was 637 to 802 mm above the floor, with an average of 707 mm. Keyboard height was significantly correlated with: stature (r=.71), pre­ ferred seat height (r=.74), and preferred CRT height (r=.57). They recommended that keyboards be as thin as possible to satisfy table-height requirements and that the keyboard-support surface be independently raised and lowered relative to the CRT height. Experienced typists used a thin-profile (30 mm) key­ board at 5, 10 and 25 degrees and a thick-profile keyboard at 15 degrees (Suther and Mciyre, 1982). There were no differences in typing performance for the four angles. None of the typists preferred the 5degree keyboard. One person liked 25 degrees best. Everyone else rated the 10- and 15-degree keyboards as preferable. This study also found that preferences were related to stature and hand length. Taller people and those with long hands tended to like the lower slope, while short people and those with short hands liked the steeper slope. Recommendations were given for keyboards with a slope that is adjustable between 10 and 25 degrees. Abernethy (1984; Abernethy and Akagi, 1984) com­ pared a 30-mm (8 degrees) keyboard with a 66-mm (12 degrees) keyboard. Typists' hands tended to "curl"

482 more with the 30-mm keyboard. That is, "the fingers curved around more as though they were forming a [loose] fist ... for the lower keyboard" (C.N. Abernethy, personal communication, September 12, 1985). A comparison of the same 30-mm keyboard with a 44.5-mm (8- degree slope) keyboard showed that the wrist angle "pronated" more with the lower keyboard. Pronaüon describes the motion of the hand turning inward about the axis of the wrist, such that "the wrist angle flattened out, becoming more parallel to the floor, from the higher to the lower keyboard height" (C.N. Abernethy, personal communication, September 12, 1985). Pronation was more pronounced when the keyboards were placed on a lower, typing stand than when they were at desk height (8-degrees pronation compared to 5-degrees pronation). In another test, the 30-mm keyboard was modified to allow adjustment of the slope from 8 degrees to over 30 degrees. The av­ erage angle chosen was 16.1 degrees at desk height and 14.4 degrees at typing stand height. In an experiment designed to simulate dual-task con­ ditions in aircraft cockpits (Hansen, 1983), three slopes of a small keypad were tested. There were no perfor­ mance differences between zero-, 15- and 35-degree slopes, but seventy-five percent of the pilots tested preferred the 15-degree slope. A keyboard with a fixed, 11-degree slope was tested at three heights: 35, 64, 84, and 104 mm (Burke, Muto, and Gutmann, 1984). Again, there were no significant differences in either speed or accuracy of performance. The lowest keyboard was the least pre­ ferred. The 84-mm keyboard was most preferred. The 64-mm keyboard was also rated highly. Conclusions. The results of research clearly show that performance is unaffected across a wide range of keyboard heights and slopes. It is also clear that people prefer some slope in a keyboard and that the angle should be adjustable to at least 15 degrees and perhaps even steeper to accommodate individual preferences. The finding that preferred slope is related to stature and to hand length implies that preferences should be predictable. The height and slope of keyboards have become a matter of debate since the West Germans announced their requirement for low- profile (30 mm) keyboards having a slope of no more than 15 degrees. This regulation has been enforced since January 1, 1985. Long-term comfort and avoidance of muscular strain appear to be the primary considerations behind the law. Specifically, a slim keyboard is an advantage for short people who cannot adjust their furniture to achieve a comfortable typing height. Taller people

CHAPTER 21. KEYS AND KEYBOARDS can always make use of a thicker keyboard or they can raise the height of a low one by placing a book or other object under it. Although the 30-mm require­ ment caused quite a stir when first announced, lowprofile keyboards have come to be accepted by key­ board designers, manufacturers, researchers and users (e.g., Paci and Gabbrielli, 1984). Size of the Keyboard Very little research has been done comparing key­ boards of different sizes. In general, the size of a keyboard is dependent on the task and the number of keys required for functionality. Whether the aim is to reduce space requirements or to increase the num­ ber of functions of a system, tests should be conducted to determine whether operators can achieve acceptable performance levels with the proposed design. Small keyboards. Most of the work with "small" keyboards has concentrated on those containing less than a full-alphanumeric set of keys. The main in­ terest in miniaturized keysets has been for telephone entry (Francas, Brown, and Goodman, 1983), mili­ tary and aerospace environments (Hufford and Coburn, 1961),and special applications (Goodman, Dickinson, and Francas, 1983). Smaller keyboards are generally not meant to be operated by touch typing, so keying rates can be expected to be lower. There is some ev­ idence that the use of a probe or stylus to actuate the keys on a small keyboard results in better performance than using the fingers (Hufford and Coburn, 1961). Large keyboards. There is no data available on the use of oversized keyboards. It is expected that the more keys a keyboard contains, the more difficult it will be to find individual keys and the longer the learn­ ing time to reach peak performance levels. Detachable Keyboards There is no research available on the need for key­ boards that are detachable from the screen housing. The purpose of detachable keyboards is to satisfy in­ dividual sizes, preferences and task needs for locating the keyboard on the work surface. Tradeoffs with furniture. The advantage of a de­ tachable keyboard might be limited by the selection of work surfaces. Although a separate, keyboard-support surface increases flexibility for height of the keyboard, it can reduce flexibility for placing the keyboard off to one side of the workstation. Locating the keyboard support at the center of a workstation might interfere with tasks that do not require the use of a keyboard.

21 A. PHYSICAL FEATURES OF KEYS AND KEYBOARDS Tradeoffs with task. Another important point is that it is not always necessary to have a detachable key­ board. For office tasks that are brief or infrequent, a detachable keyboard is not required. In some cases, a detachable keyboard is impractical. Laptop computers would be more difficult to use if the keyboards were separate. For other applications, such as public ac­ cess terminals, a detachable keyboard may even be a liability. Keyboard Profile The relative angles at which different rows of keys are arranged on the keyboard is called the keyboard profile. Only one study was found that tested different key­ board profiles (Paci and Gabbrielli, 1984). Three typ­ ists were filmed while using either a stepped or a dished profile keyboard. The angle at which the fin­ gers touched the keys ranged from 2 to 13 degrees on the stepped keyboard and from 8 to 11 degrees on the dished keyboard. Performance was reportedly better with the dished profile and the typists expressed a preference for this keyboard. These effects may have been influenced by a difference in slopes: the stepped keyboard had a 9-degree slope, while the dished keyboard had a 12degree slope. They recommended the dished profile for the alphanumeric keys, the stepped profile for the numeric keypad (because this is common for calcula­ tors), and the sloped profile for function keys because visual requirements (e.g., labeling and readability) are more stringent for these keys. Conclusions. Although there may be performance differences between stepped, sloped, dished, and flat keyboards, recommendations generally agree that each of these is acceptable. Individuals may have strong preferences for particular keyboard profiles, but these probably are highly dependent on what keyboards are familiar to the typist. More research is needed to make firm recommendations. Key Size and Shape Because the keytops are the most immediate contact between a user and a keyboard, it might be supposed that the size and shape of the keys have a lot to do with typing performance. However, little research has been conducted to test this hypothesis. As late as 1972, the design of individual keys depended more on "design conventions rather than empirical data" (Alden et al., 1972, p. 280).

483

Deininger (1960) tested different key sizes and shapes for a ten-key numeric keypad. Keying times and accuracy improved when key size was increased from 0.375 to 0.50 inches (0.95 to 1.27 cm). How­ ever, these results cannot be generalized to full-size keyboards. Clare (1976) presents four goals for the design of key shapes: 1. the operator should be able to see the label of the key, 2. the finger should be able to locate the key without hitting other keys or fingers, 3. the distribution of pressure on the finger should indicate the location of the finger on the key, and 4. the force of pressing the key should be distributed to the proper portion of the finger. Clare recommends that key tops should be 0.5 inches (1.27 cm) square and have a distance of 0.75 inches (1.91 cm) between keytop centers. Smaller keytops (i.e., 0.375 inches or 0.95 cm) were found to be "less satisfactory" (p. 102). The top of the key should slope away from the horizontal position at 21.5 degrees to distribute the pressure on the finger, but should slope forward about 45 degrees to satisfy viewing require­ ments. The front of the key should be raised to provide for adequate finger positioning. By placing the labels on the front edge of the key and sloping the front sur­ face backwards for visibility, the visual requirement can be met. The back part of the top surface can be slightly raised to aid finger positioning. Another study (Hansen, 1983) compared two key sizes for use by pilots with and without gloves. One keyboard had small (5.16- mm diameter square) keys and the other had large (12.4-mm diameter square) keys. Both keyboards used a 3 x 3 layout. Key­ ing speed was fastest with the large keyboard without gloves. There were no differences in error rates. PC Magazine (Rosch, 1984) reported the results of their typing test to evaluate various computer key­ boards. Taping performance was much poorer with keyboards having small keys. When these same key mechanisms were used with larger keys, performance was among the best for the eleven keyboards tested. Monty, Snyder and Birdwell (1983; Texas Instru­ ments, 1983) compared six keyboards for both perfor­ mance and preferences. Since the keyboards differed in many ways, it is difficult to determine the specific causes of any effects on typing speed, errors or user ratings. In general, speed and error differences were

484 inconsequential. One keyboard that did not require use of the shift key was found to be faster and more errorprone than the others. Two other keyboards elicited faster, but less-accurate performance. The inverse re­ lation between speed and accuracy makes it difficult to interpret these data because it indicates a tradeoff in the typists' criteria for performance. (A more ex­ pected result would be that a "better" design would result in faster typing with fewer errors.) Typists preferred keycaps that "resemble the some­ what rounded, dished keycaps of earlier model Selectric typewriters" (Texas Instruments, 1983, p. 26). The next most preferred keys were those with large, square touch surfaces that were cylindrically indented from front to back. The worst ranking was given to the one keyboard having round keytops. Conclusions. Although the research is sparse, there are some commonalities in recommendations for key size. For finger operation (as opposed to using a stylus or probe), 0.5-inches (1.27 cm) square is a lower limit for keytop strike area. There is no data available on upper limits for key size. Values for the center-tocenter spacing of keys are generally around 0.75 inches (1.91 cm). With regard to the shape of keys, there is some ev­ idence that they should be indented to fit the shape of the finger tip for ease of location and finger place­ ment. Preferences appear to lean toward keys that are spherically indented as opposed to those having only a cylindrical indentation. The latter are more preva­ lent on current keyboards in the United States and have not been found to adversely affect performance. Some guidelines cite round keys as acceptable, but square keys may be better because they provide more surface area within the same amount of space between keytop centers (Cakir, Hart and Stewart, 1980). Choice of round versus square keys may be a cultural variable: round or oval keys are more popular in West Germany than in the United States. Key Force, Travel and Tactile Feedback The force with which a finger must press a key to actuate it and the distance the key travels before, dur­ ing and after actuation are described by the so-called "force/travel function." A generalized force/travel function is presented in Figure 1, which shows: • the distance the key travels on the horizontal axis; • the force with which the key must be pressed shown on the vertical axis;

CHAPTER 21. KEYS AND KEYBOARDS

switch closed

switch open KEY TRAVEL DISTANCE Figure 1: A generalized force/travel function. • one curve for the downstroke and one for the up­ stroke of the key (the direction of travel is shown by arrows); • the point at which the key is actuated ("switch closed"); • the "switch open" point at which the key can again be pressed to create a character; • changes in the slope of the function corresponding to the "feel" of the key at different points along its travel; • "tactile feedback" caused by a drop in force be­ fore the "switch closed" point and an increase in force beyond this point Studies on the effects of different force/travel func­ tions have not been systematic, so it is difficult to create a model of how a particular aspect of the func­ tion will affect keying performance or preferences. Research has uncovered ranges within which perfor­ mance is not affected by either the amount of force or the distance of travel. Other investigations have compared keyboards for which the entire force/travel function varied.

21 A. PHYSICAL FEATURES OF KEYS AND KEYBOARDS A study of telephone usage and occasional data en­ try (Deininger, 1960) found no performance differ­ ences due to either a decrease in maximum force from 14.1 to 3.5 ounces (400 to 100 grams) or a decrease in maximum travel from 0.19 to 0.03 inches (0.48 to 0.08 cm). Key pressing performance was found to be best at low levels of force and travel (Kinkead and Gonzales, 1969). Recommended values were between 5.3 and 0.9 ounces (150.3 and 25.5 grams) for force and be­ tween 0.25 and 0.05 inches (0.64 and 0.13 cm) for key travel. Brunner and Richardson (1984) compared three switch technologies: a snap-spring keyboard which had a very slight drop- off in force before the point of actuation and a gradual increase in force after this point; a linear-spring keyboard that had no change in the force/travel function to indicate tactile feedback; and an elastomer switch that provided distinct tactile feedback. The elastomer keyboard was about 2% to 6% bet­ ter than the other keyboards counting both speed and errors. Characters were incorrectly inserted more of­ ten on the linear-spring keyboard. Ratings by typists indicated that the elastomer keyboard was acceptable relative to the other keyboards tested. It is not pos­ sible to determine exactly which features of the key­ boards caused the performance differences reported in this study. It is, in general, faster to type on a keyboard that requires less force and shorter travel because the fingers must do less work. Tactile feedback also could have contributed to the differences in performance. In another study typists rated "smooth key move­ ment" as a highly important facet of keyboard qual­ ity (Monty, Snyder, and Birdwell, 1983; Texas Instru­ ments, 1983). Comparing six keyboards, the factors found to be most important to users were: 1. key switches that do not have noisy key bottom­ ing; 2. tactile-snap feedback caused by an abrupt change in the force required to actuate the key; 3. a force/travel curve shaped like a "roller coaster"; 4. a smooth force/travel curve that is not disturbed by "jitter"; 5. minimal lateral wobble of the keycap. Because the data showed a speed-accuracy tradeoff, it was impossible to determine the effects of different force/travel curves on performance.

485

Touch keys that lack key travel such as capacitance and membrane technologies have been shown to re­ sult in slower keying performance than conventional mechanical keys (Cohen, 1982; Pollard and Cooper, 1979). Although the disadvantage of these switches decreases as users become accustomed to the absence of tactile feedback, the negative effects can be signif­ icantly reduced by the addition of cues such as em­ bossed edges, metal domes, and tones or clicks on actuation (Roe, Muto, and Blake, 1984). Clare (1976) recommends that force/travel curves should be different for different fingers and key lo­ cations. Upper keys should have shorter travel and lower keys should have longer travel to indicate the same "feel" for the fingers. Conclusions. The literature on actuation force and travel indicates that performance is not affected within a wide range of these parameters. Recommended val­ ues range from about 1 to 5 ounces (approximately 28 to 142 grams) of force and about 0.05 to 0.25 inches (approximately 0.13 to 0.64 cm) of travel. The exact values may depend on preierences for the "feel" of the keys. More important than the amount of force or travel is tactile feedback, which is caused by a gradual in­ crease in force followed by a sharp decrease in force required to actuate the key (called "breakaway" force) and a subsequent increase in force beyond this point for cushioning. The result is a curve shaped like a roller coaster. From the data available, it is recom­ mended that keys provide tactile feedback because it improves keying performance and is preferred by typ­ ists. Capacitive and membrane keys that require only a minimal touch and no travel are inferior to conven­ tional keys in terms of typing performance. Because many factors could have influenced the results of the relevant studies, more research in this area would be useful. Auditory Feedback Auditory clicks, beeps and tones have been found to be unnecessary for skilled typists in high speed data entry tasks (Alden et al., 1972). For telephone use, however, there is some evidence that a single tone allows faster keying than a click or a visual signal (a light) (Pol­ lard and Cooper, 1979). When the user cannot see the keys for dialing telephone numbers, speech feedback has been found to reduce keying errors (Nakatani and O'Connor,1980). Newer key devices, such as elastomer and capaci­ tive switches have allowed designers to create truly

486 silent key action. Auditory feedback can be incorpo­ rated into these keyboards as an add-on feature that can be turned on and off. Some also allow adjustment of the volume of the mechanical click. Performance data with such keyboards indicate that typing is somewhat faster and more accurate with auditory feedback on than with it off (Monty, Snyder, and Birdwell, 1983; Roe, Muto, and Blake, 1984). However, tactile feed­ back is more important for keying speed than is audi­ tory feedback. Most typists prefer auditory feedback, but they also want the ability to turn it off. Another aspect of auditory feedback is the time lag between key depression and the tone or click. If the lag is too long, it may actually interfere with typing performance (Clare, 1976; Texas Instruments, 1983). Visual Feedback Common sense suggests that it might be helpful to have a visual display of a telephone number to re­ duce errors when dialing the phone. In fact, such a display has been found to be "of no value for ordi­ nary dialing of telephone numbers" (E.T. Klemmer, personal communication, December 10, 1981). Visual displays for telephones are not an advantage because "Telephone users can easily operate with an accept­ able low error rate without the display and the effort involved in checking the display is not worthwhile. Moreover, if the user suspects that an error was made (more than half all errors are self-detected) it is more efficient to simply re-key the number than to read the display, check for accuracy, and then re-key the num­ ber" (E.T. Klemmer, personal communication, Decem­ ber 10, 1981). For touch typing on regular keyboards, visual feed­ back has not been found to provide any advantage for speed of typing, but it does affect the typist's abil­ ity to catch and correct errors (Alden et al., 1972; Rosinski, Chiesi, and Debons, 1980). With less than 9 characters displayed at a time, typists were less likely to correct their own errors (Rosinski et al., 1980) than when a larger number of characters was displayed. Vi­ sual feedback also may be useful when first learning to type (Alden et al., 1972). As with auditory feedback, timing of visual feed­ back is important. When the print mechanism of a teletype was delayed and irregular in relation to typing on the keyboard, speed of typing was slowed for both unskilled and experienced typists (Long, 1976). The effect was eliminated with practice, but only for skilled typists. Delay of visual feedback on a computer dis­ play was also found to have important effects on typist

CHAPTER 21. KEYS AND KEYBOARDS behavior and satisfaction (Williges and Williges, 1981, 1982). Error-Avoidance Features Variables such as rollover, buffer length, hysteresis and repeat functions have effects on the prevalence of typ­ ing errors, but have not been investigated experimen­ tally. Rollover. The ability of the keyboard to store each keystroke in proper sequence is known as rollover. Without rollover, each key must be released before the next key is depressed. During high-speed typing without rollover, some characters will be lost or will generate erroneous codes. Two-key rollover will gen­ erate two keystrokes accurately if the second key is depressed before the first key is released. With n-key rollover, any number of keystrokes can overlap and will result in the proper sequence of characters. Another aspect of rollover is shadow rolling, which refers to the sequence of keystrokes in which the sec­ ond key is depressed and released before the first key is released. If only two-key rollover is provided, shadow rolling does not generate the second character. Only with n-key rollover will the sequence be correctly recorded. Thus, n-key rollover is preferred over twokey rollover (Cakir, Hart, and Stewart, 1980; Davis, 1973). Hysteresis. A key is actuated at the "switch closed" point on the downstroke. Upon release, the switch remains closed until the key travels past the "switch open" point. If the closed and open points occur at the same place, extra keystrokes can be inserted when the typing finger hesitates or is not smooth on the upstroke (called key "bounce"). This problem is eliminated by hysteresis, which is the travel distance between the "switch close" and "switch open" points. That is, the switch remains closed on the upstroke even past the point at which the key was actuated on the downstroke (see Figure 1). Interlocks. If the key does not have mechanical hysteresis, an electronic interlock can be used to pro­ vide the same benefit. On most modern keyboards, an electronic polling device imposes a minimum time be­ tween keystrokes. Short inter-key times are assumed to be due to key bounce or unintended keystrokes. The effect of the interlock is to transmit keystrokes at a controlled rate so that these short inter-key times are ignored. The optimal transmission rate depends on maximum keying rates, which usually occur in fast "bursts" of typing. If the interlock period is too short, it will not

21.5. INNOVATIONS IN KEYBOARD DESIGN be effective at screening out unintended keystrokes. Recommendations for interlock systems vary, but in general should account for a typing rate of at least 100 gross words per minute. This value represents an interlock period of about 100 msec. Since typing burst rates can be much faster than 100 msec per keystroke (inter- keystroke intervals have been observed as low as 4 msec for skilled keying of certain digrams, but these occur infrequently), shorter interlock intervals may be necessary. A lower limit of 50 msec has also been proposed (Alden et al., 1972). Further research is necessary to make a firm recommendation. Buffer length. Many computer keyboards allow the user to "type ahead" of the display. This feature is particularly useful when the display is dependent on the response of the system. If the user knows ahead of time what input should be entered next, the delay of the system can be avoided by storing keystrokes in a buffer until the system is ready to receive them. In a comparison of buffers storing 1, 2, 4, 6 and 7 characters, buffer length was found to interact with the speed of visual feedback in determining typing speed (Williges and Williges, 1981,1982). With a fast visual display, the buffer could be shorter without affecting typing speed. If the display was delayed by as much as 1.5 seconds, a larger buffer was required. Buffer length was found to be less important than speed of visual display: long delays in visual feedback resulted in slower performance regardless of the length of the buffer. Since computer response times often result in delayed visual feedback, a large buffer (of at least 7 characters) is recommended. Repeat features. Although there are no experimen­ tal data to support the need for key-repeat features, experience shows that "typamatic" keys are handy for many purposes, such as underlining and graphic sym­ bols used to make tables. On many computer key­ boards, all keys are implemented to be typamatic. The delay before creating repeated characters after the key has been depressed should obviously be neither too short nor too long, but the literature does not contain any specific recommendations for the optimal delay length. Color and Labeling Recommendations for the color and labeling of keys and keyboards are based on requirements for visibil­ ity and for coding purposes, rather than experimental investigations. Neither the keys themselves, nor the entire keyboard should be so shiny as to create a glare source. A "matte" to "silky matte" finish is usually

487 recommended. The names of keys should be legible and understandable to the user. A slightly rough finish on the keytops aids finger positioning, but should not reduce the legibility of the key label. Keys that are colored and grouped by function assist the search pro­ cess and make the keyboard aesthetically pleasing to the user. This is particularly helpful on computer key­ boards, because they are rarely used simply as touch typing devices for text entry (Cakir, Hart, and Stewart, 1980).

21.5 Innovations in Keyboard Design Over the last several years new developments in key­ board design have included physical reconstructions of the keyboard and modifications in the means of oper­ ating a keyboard. Some of these designs are presented in this section. Split and Tut Keyboards Studies of how people naturally hold their arms and hands compared to their postures at conventional type­ writers indicate that there may be components of fatigue inherent in the way conventional keyboards are designed (Kroemer, 1972; Nakaseko, Grandjean, Hunting, and Gierer, 1985; Zipp, Haider, Halpem, and Rohmert, 1983). The K-keyboard. Kroemer (1972) named his Kkeyboard for Klockenberg who published a critique of the standard typewriter in 1926. Based on experi­ ments and physiological analyses of typists, Klocken­ berg proposed splitting the keyboard into right and left halves and laterally angling them downward on either side to allow a more natural positioning of the hands (like a handshake as opposed to palm-down) (Kroe­ mer, 1972). The K-keyboard incorporates this fea­ ture (see Figure 2). The keys are arranged in straight columns, but the rows are curved to better fit the differ­ ent finger lengths. The space bars (one for each hand) are curved to fit the shape of the thumb. Kroemer has not yet determined the assignment of characters to keys. Experiments with the K-keyboard have shown that the degree of lateral tilt of the two halves does not affect key tapping frequency or errors. Of the an­ gles tested (zero, 30, 60 and 90 degrees from hori­ zontal), some typists preferred 60 degrees over no tilt (Kroemer, 1972). In another test, more errors were made on a standard QWERTY keyboard than on the K-keyboard with a 45 degree tilt (127 versus 77 er­ rors for every 1000 key taps). There were no differ-

CHAPTER 21. KEYS AND KEYBOARDS

488

Hinges

Figure 2: Kroemer's K-keyboard ences either in typing speed or in number of heart beats (a measure of "circulatory strain"). However, Kroemer also reports that participants terminated the task for different reasons. Users of the standard keyboard more often complained of "aches and pains," while Kkeyboard typists were more likely to report that they couldn't concentrate any longer (although some aches and pains were also reported by this group). Kroemer's research indicates that the K-keyboard may improve productivity by reducing errors, creating a more comfortable posture and reducing the amount of fatigue experienced. More work is needed, partic­ ularly in the area of assigning characters to the keys. It would also be interesting to observe the angle at which typists adjust the keyboard to suit their own preferences and whether this capability is important to typists. The STR keyboard. Nakeseko and his colleagues (1985) performed preliminary research on an exper­ imental split-half keyboard with a "hand-configured" design (see Figure 3). The keys are in curved rows and the columns are slightly offset. The upper corners meet to form a triangle, which results in a smaller dis­ tance between the two halves than on the K-keyboard. This keyboard uses the German QWERTZ assignment of letters to keys. Another feature is the use of a forearm-wrist support. Experiments have been done to test different open­ ing angles between halves, the angle of tilt, and the size of the forearm-wrist support surface (Nakaseko et al., 1985). Typists found the split keyboard acceptable and favored the design with lateral tilt of 10 degrees and an opening angle of 25 degrees. Although per­ formance data have not been reported, preferences and measures of arm and hand positions show a strong ef­ fect due to the use of a large forearm-wrist support. In fact, a traditional keyboard with a large support

was preferred by more typists than the experimental keyboard with a small support. The use of a large forearm-wrist support might also result in a less stress­ ful posture. Later work with this keyboard led to the development of a final product by Standard Telephon and Radio (STR) AG (Buesen, 1984). The STR key­ board won a design award at Ergodesign in 1984. A molded keyboard. Another invention is a keyboard that is molded to match the positions of hands and fingers, called the "PCD Maltron" keyboard (Malt and Litewka, 1983). It was developed by Lillian Malt in conjunction with Stephen Hobday of P.C.D. Limited. The company markets several versions for different purposes such as typewriter or computer compatibility, and one- handed or mouth-stick operation for handi­ capped users. The basic design is depicted in Figure 4. The two halves of the PCD Maltron keyboard are tilted laterally. Key heights are designed to match finger lengths. The layout of characters on the keys is based on principles similar to those espoused by Dvorak plus a few additional criteria. For instance, the vowels were found to be a major source of error on Dvorak's layout, so they are located as far apart as possible to avoid confusion (Hobday, 1981). For those who do not wish to change from QWERTY, an option is available for switching between Maltron and QWERTY with a "changeover key" (Hobday, 1982). The space between the two halves of the keyboard can be used for additional function keys. Controlled experiments with this keyboard have not been reported. However, initial tests (e.g., PCD Mal­ tron Ltd., 1982) with a small number of typists have led to claims that it: 1. can be learned in about a quarter of the time taken to learn QWERTY (5 hours to become competent as opposed to 15 to 20 hours for QWERTY); 2. allows a more comfortable typing position than standard keyboards; 3. results in a 20 to 40 percent increase in typing speed with an error rate of about .02 percent in a "production situation;" and 4. does not interfere with QWERTY typing skill when the typist switches back to a QWERTY ar­ rangement. Some versions of the design do have obvious advan­ tages, such as the possibility of one-handed operation. However, each Maltron keyboard is hand-assembled so that "the sculpted design is inherently more expen­ sive" (Lu, 1983).

21.5. INNOVATIONS IN KEYBOARD DESIGN

Figure 3: STR Keyboard of Nakaseko et al.

Figure 4: Basic Design of the PCD Maltron Keyboard

489

CHAPTER 21. KEYS AND KEYBOARDS

490 New Methods of Typing A one-handed chord keyboard. With traditional key­ boards, keys are pressed one at a time to create charac­ ters in sequence. Chord keyboards require the depres­ sion of several keys simultaneously to create a char­ acter as in striking a chord on a piano. Since each key can be used in various combinations to create differ­ ent letters, the number of keys on a chord keyboard can be greatly reduced. For instance, a five-key, chord keyboard can produce up to 31 characters. The main interest in chord keyboards has been for mail sorting tasks at the Post Office (Noyes, 1983a). The Microwriter is a single-handed, chord keyboard. It is marketed as the computer that fits in your pocket, because it "weighs only two pounds and is slightly larger than a paperback book" (C. Suchman, personal communication, January 5, 1983). It has five keys, one for each finger, with an additional command key operated by the thumb (see Figure 5). Letters of the alphabet, numbers and punctuation are created with combinations of the five basic keys. Word processing functions such as delete, insert and tab are performed with the aid of the command key. Other features of the Microwriter are its 12-character display and 8Kbyte memory. Since it is battery operated, it can be used anywhere.

Display Memory (internal)

Main Keyboard

Command Key

Figure 5: The Microwriter A small study was conducted with the Microwriter (Wheeler, 1980), but the advantages of the device were related more to its word processing functions than to the fact that it is a chord keyboard. It has been esti­

mated that the Microwriter allows one to type up to 150communication, January 5, 1983). The Microwriter may be useful for some special applications, particularly because of its capability for storage and interfacing with computers. However, it does have some basic ergonomie problems. For in­ stance, the size of the display is limiting and some characters are extremely difficult to key. A two-handed chord keyboard. The concepts in­ volved in the design of the Microwriter have been incorporated into a two- handed, chord keyboard for computer text entry (Gopher, Hilsernath, and Raij, 1985). The two halves of the keyboard are mirror images of each other and are connected in a lateral tilt arrangement. The chords for creating letters are exactly the same on each half, and correspond to the same fingers of each hand. Some tests have been conducted with people typing Hebrew text. The researchers report that people could memorize the codes for each character in about 30-45 minutes. After about 30 hours of training, typing rates were 38-42 words per minute. They estimate that with about 20 hours of practice, people can type 50% faster than their handwriting speed, which is about 20-22 words/min. There is also some indication that learning of the chord keyboard does not interfere with previous (e.g., QWERTY) typing skills. The two-handed keyboard may eventually lead to faster keying (with practice), although initial tests showed no differences in speed between one- and twohanded operation. However, the cognitive processes of two-handed chording appear to be quite different from those required for sequential typing. For exam­ ple, because each hand can make the same characters, it is possible to alternate hands for each pair of let­ ters. This can lead to some difficulties for words that the typists learned in a set sequence always beginning with the same hand first. Anecdotal evidence showed that at least one typist had problems with typing cer­ tain words. She put double spaces at the end of every word that had an uneven number of characters. The researchers have suggested that the two-handed, chord keyboard might even be used for a bilingual, data-entry system, with entry of one language on the right and the other on the left half. They have yet to demonstrate whether typists are capable of performing such a task. A wipe-activât ed keyboard. Another new method of keyboard operation is based on capacitance tech­ nology. Because no downward travel is involved, the "keyboard" can be designed as a flat tablet that the fingers glide across or wipe to create characters. An

21.7. ACKNOWLEDGEMENTS

491

0©©©ΘΘΘΘΘΘΘΟΘ©Θ θ©©Θ©©©©ΘΘ©©©θ ®©©©©©0Θ©©00<§> ©©©©©©©©©©©© SPACE

Figure 6: Montgomery's wipe-activated keyboard layout. alternative method of operation is to use a stylus to interact with the keyboard. Edward Montgomery (1982) of the University of Texas Health Science Center at Dallas has proposed such a keyboard. To take advantage of the wiping motion, Montgomery also developed a new character layout, which can be seen in Figure 6. With this ar­ rangement it is possible to create many words with a single wiping motion. For example, "TEE" is created with one motion from left to right. No tests (other than analyses comparing number of wipes to number of keystrokes required on traditional keyboards) have been reported to determine perfor­ mance or preference effects with this keyboard. Since it has no moving parts, this device can be produced in almost any size and can easily be used by handicapped people.

21.6 Summary Keyboard design has been influenced more by entiepreneurship and convention than by empirical test­ ing. Human performance data indicate that typing behavior is unaffected across a wide range of vari­ ables, so it is unlikely that further work in this area will uncover factors of significance to the physical as­ pects of typing. However, there is a need for study of the cognitive processes involved in the use of mod­ ern keyboards such as those found on word processors and computer terminals. For instance, little is known about the use of function keys for giving commands or formatting text. The extent of current work in this

area has been limited primarily to the arrangement of cursor keys (e.g., Emmons, 1984 showed that· a cross arrangement was better than a box layout especially for inexperienced users). With increases in the number of function keys and the advent of virtual keyboards (in which each key is reassigned functions corresponding to a variable display of the keyboard on the screen), the emphasis in keyboard research is shifting to the study of operator workload and visual search times.

21.7 Acknowledgements This chapter is a condensed and revised version of the Office Systems Ergonomics Report, Vol. 5, Num. 2, March/April 1986, published by The Koffler Group, Santa Monica, California. The author gratefully ac­ knowledges permission to use this material. The author also wishes to express her gratitude to Grayson Marshall of Los Angeles for drawing Figures 1, 2, 5, and 6.

21.8 References Abernethy, C.N. (1984). Behavioural data in the de­ sign of ergonomie computer terminals and worksta­ tions - A case study. Behaviour and Information Tech­ nology, 3(4), 399-403. Abernethy, C.N., and Akagi, K. (1984). Experimental results do not support some ergonomie standards for computer video terminal design. Computers & Stan­ dards, 3, 133-141.

492

CHAPTER 21. KEYS AND KEYBOARDS

Alden, O.G., Daniels, R.W., and Kanarick, A.F. (1972). Keyboard design and operation: A review of the major issues. Human Factors, 14(4), 275-293.

Davis, S. (1973). Keyswitch and keyboard selection for computer peripherals. Computer Design, 12(3), 6779.

Bayer, D.L. and Thompson, R.A. (1983). An experi­ mental teleterminal - The software strategy. Bell Sys­ tem Technical journal, January 1983, 121-144.

Deininger, R.L. (1960). Human factors engineering studies of the design and use of pushbutton telephone sets. Bell Systems Technical Journal, 39, 995-1012.

Brunner, H., and Richardson, R.M. (1984). Effects of keyboard design and typing skill on user keyboard preferences and throughput performance. In Proceed­ ings of the Human Factors Society 28th Annual Meet­ ing (pp. 267-271). Santa Monica, CA: Human Factors Society.

Dvorak, A. (1943). There is a better typewriter key­ board. National Business Education Quarterly, De­ cember 1943, XII-2, 51-58 and 66.

Buesen, J. (1984). Product development of an er­ gonomie keyboard. Behavior & Information Technol­ ogy, 3(4), 387-390. Burke, T.M., Muto, W.H., and Gutmann, J.C. (1984). Effects of keyboard height on typist performance and preference. In Proceedings of the Human Factors Soci­ ety 28th Annual Meeting (pp. 272-276). Santa Monica, CA: Human Factors Society. Butterbaugh, L. and Rockwell, T. (1982). Evaluation of alternative alphanumeric keying logics. Human Fac­ tors, 24(5), 521-533. Cakir, A., Hart, DJ. and Stewart, T.F.M. (1980). Vi­ sual Display Terminals. New York: John Wiley & Sons. Cassingham, R.C. (1986). The Dvorak Keyboard. Ar­ cata, CA: Freelance Communications. Clare, C.R. (1976). Human factors: A most important ingredient in keyboard designs. EDN Magazine (Elec­ trical Design News), 21(8), 99-102. Cohen, K.M. (1982). Membrane keyboards and human performance. In Proceedings of the Human Factors So­ ciety 26th Annual Meeting (p. 424). Santa Monica, CA: Human Factors Society. Conrad, R. and Hull, AJ. (1968). The preferred layout for numerical dataentry keysets. Ergonomics, 11, 165173.

Dvorak Developments. (Spring 1986). West Sacra­ mento, CA: Philip Davis. Emmons, W.H. (1984). A comparison of cursor-key arrangements (box versus cross) for VDUs. In E. Grandjean (Ed.), Ergonomics and Health in Modern Offices (pp. 214-219). London: Taylor & Francis. Emmons, W.H., and Hirsch, R.S. (1982). Thirty mil­ limeter keyboards: How good are they? In Proceed­ ings of the Human Factors Society 26th Annual Meet­ ing (pp. 425-429). Santa Monica, CA: Human Factors Society. Francas, M., Brown, S. and Goodman, D. (1983). Al­ phabetic entry procedure with small keypads: key lay­ out does matter. In Proceedings of the Human Fac­ tors Society 27th Annual Meeting (pp. 187-190). Santa Monica, CA: Human Factors Society. Galitz, W.O. (1966). CRT keyboard human factors evaluation: Study II. Univac, Systems Application En­ gineering, Roseville DOD, February 1966. Gentner, D.R. (1983). Keystroke timing in transcrip­ tion typing. In W.E. Cooper ÇEA.),Cognitive Aspects of Skilled Typewriting (pp. 95-120). New York: Springer Verlag. Goodman, D., Dickinson, J. and Francas, M. (1983). Human factors in keypad design. In Proceedings of the Human Factors Society 27th Annual Meeting (pp. 191-195). Santa Monica, CA: Human Factors Society.

Cooper, W.E. (Ed.). (1983). Cognitive Aspects of Skilled Typewriting. New York: Springer-Verlag.

Gopher, D., Hilsernath, H. and Raij, D. (1985). Steps in the development of a new data entry device based upon two hand chord keyboard. In Proceedings of the Human Factors Society 29th Annual Meeting (pp. 132136). Santa Monica, CA: Human Factors Society.

Davidson, L. (1966). A pushbutton telephone for al­ phanumeric input - two extra buttons. Datamation, April 1966, 27-30.

Hagelbarger, W. and Thompson, R.A. (1983). Experi­ ments in teleterminal design. IEEE Spectrum, 20(10), 40-45.

21.8. REFERENCES Hansen, M.D. (1983). Keyboard design variables eval­ uated during dual-task mode selection. In Proceedings of the Human Factors Society 27th Annual Meeting (pp. 924-927). Santa Monica, CA: Human Factors So­ ciety. Hiraga, Y., Ono, Y., and Yamada, H. (1980). An analy­ sis of the standard English keyboard (Technical Report 80-11). Department of Information Science, Faculty of Science, University of Tokyo, 7-3-1 Hongo, Bunkyoku Tokyo, 113 Japan. Hirsch, R.S. (1970). Effects of standard versus alpha­ betical keyboard formats on typing performance. Jour­ nal of Applied Psychology, 54, 484-490. Hobday, S. (1981). 'The quick brown fox../ or key­ board alternative. The Inventor, 21(1), 16-18 and 20. Hobday, S. (1982). The New Maltron Keyboard. Brochure produced by PCD Maltron Ltd., August 1982. Hufford, L.E. and Coburn, R. (1961). Operator performance on miniaturized decomalentry keysets. (NEL/Report 1083.) San Diego, CA: US Naval Elec­ tronics Laboratory, December 1961. Kinkead, R. (1975). Typing speed, keying rates, and optimal keyboard layouts. In Proceedings of the Hu­ man Factors Society 19th Annual Meeting (pp. 159161). Santa Monica, CA: Human Factors Society. Kinkead, R.D. and Gonzalez, B.K. (1969). Human fac­ tors design recommendations for touch-operated key­ boards - final report (Document 12091-FR). Min­ neapolis, MN: Honeywell, Inc., March 1969. Kroemer, K.H.E. (1972). Human engineering the key­ board. Human Factors, 14(1), 51-63. Long, J. (1976). Effects of delayed irregular feed­ back on unskilled and skilled keying performance. Er­ gonomics, 19(2), 183-202. Lu, C. (1983). Will there ever be an efficient keyboard? High Technology, January 1983, 22-23. Lutz, M.C. and Chapanis, A. (1955). Expected loca­ tions of digits and letters on ten-button keysets. Jour­ nal of Applied Psychology, 39(5), 314-317. Malt, L., and Litewka, J. (1983). An ergonomie ap­ proach to making correct posture possible for keyboard operators. Manuscript submitted for publication.

493 Michaels, S.E. (1971). Qwerty versus alphabetic key­ boards as a function of typing skill. Human Factors, 13(5), 419-426. Miller, I., and Suther, T.W. (1981). Preferred height and angle settings of CRT and keyboard for a display station input task. In Proceedings of the Human Fac­ tors Society 25th Annual Meeting (pp. 492-496). Santa Monica, CA: Human Factors" Society. Miller, I., and Suther, T.W. (1983). Display station anthropometries: preferred height and angle settings of CRT and keyboard. Human Factors, 25(4), 401408. Montgomery, E.B. (1982). Bringing manual input into the 20th Century: New keyboard concepts. Computer, March 1982, 11-18. Monty, R.W., Snyder, H.L. andBirdwell, G.G. (1983). Keyboard design: An investigation of user preference and performance. In Proceedings of the Human Fac­ tors Society 27th Annual Meeting (pp. 201-205). Santa Monica, CA: Human Factors Society. Nakaseko, M., Grandjean, E., Hunting, W., and Gierer, R. (1985). Studies on economically designed alphanumeric keyboards. Human Factors, 27(2), 175187. Nakatani, L.H., and O'Connor, K.D. (1980). Speech feedback for touch-keying. Ergonomics, 23(7), 643654. Norman, D.A., and Fisher, D. (1982). Why alpha­ betic keyboards are not easy to use: Keyboard layout doesn't much matter. Human Factors, 24(5), 509-519. Noyes, J. (1983a). Chord keyboards. Applied Er­ gonomics, 14(1), 55-59. Noyes, J. (1983b). The QWERTY keyboard: A re­ view. International Journal of Man-Machine Studies, 18, 265-281. Paci, A.M., and Gabbrielli, L. (1984). Some experi­ ences in the field of design of VDU work stations. In E. Grandjean (Ed.), Ergonomics and Health in Modern Offices (pp. 391-399). Philadelphia: Taylor & Francis. Paul, F., Sarlanis, K. and Buckley, E.P. (1985). A hu­ man factors comparison of two data entry keyboards. In Institute of Electrical and Electronics Engineers Symposium on Human Factors in Electronics.

494

CHAPTER 21. KEYS AND KEYBOARDS

PCD Maltron Ltd. (1982). PCD Maltron ergonomie keyboard. Brochure produced by PCD Maltron Ltd., October 1982.

Wheeler, P.S. (1980). The Microwriter as an educa­ tional aid. British Journal of Educational Technology, 11(3), 160-169.

PoUard, D., and Cooper, M.B. (1979). The effect of feedback on keying performance. Applied Ergonomics, 10(4), 194-200.

Williges, R.C., & Williges, B.H. (1981). Univariate and multivariate evaluation of computer-based data en­ try. In Proceedings of the Human Factors Society 25th Annual Meeting (pp. 741-745). Santa Monica, CA: Hu­ man Factors Society.

Roe, C.J., Muto, W.H., and Blake, T. (1984). Feed­ back and key discrimination on membrane keypads. In Proceedings of the Human Factors Society 28th Annual Meeting (pp. 277-281). Santa Monica, CA: Hu­ man Factors Society. Rosch, W.L. (1984). Keyboard ergonomics for IBMs. PC Magazine, 3(19), 110-122. Rosinski, R.R., Chiesi, H. and Debons, A. (1980). Effects of amount of visual feedback on typing per­ formance. In Proceedings of the Human Factors Society 24th Annual Meeting (pp. 195-199). Santa Monica, CA: Human Factors Society. Scales, E.M. and Chapanis, A. (1954). The effect on performance of tilting the toll-operator's keyset Journal of Applied Psychology, 38(6), 452-456. Strong, E.P. (1956). A comparative experiment in simplified keyboard retraining and standard keyboard supplementary training. Washington, DC: General Ser­ vices Administration. Suther, T.W., and McTyre, J.H. (1982). Effect on op­ erator performance at thin profile keyboard slopes of 5 °, 10 °, 15 ° and 25 °. In Proceeding of the Human Factors Society 26th Annual Meeting (pp. 430-434). Santa Monica, CA: Human Factors Society. Texas Instruments, Inc., (1983). VPI Keyboard Study, Design Development Center, June 1983. U.S. Navy Department (1944a). A practical experiment in Simplified Keyboard retraining: A report on the retraining offourteen Standard Keyboard typists on the Simplified Keyboard. Training Section, Departmental Services, Division of Shore Establishments and Civil­ ian Personnel, Navy Department, Washington, D.C., July 1944. U.S. Navy Department (1944b).A comparison of typist improvement from training on the Standard Keyboard and retraining on the Simplified Keyboard: A supplement to 'A practical experiment in Simplified Keyboard retraining'. Training Section, Departmental Services, Division of Shore Establishments and Civilian Person­ nel, Navy Department, Washington, D.C., 18 October 1944.

Wilüges, R.C., & Williges, B.H. (1982). Modeling the human operator in computer-based data entry. Human Factors, 24(3), 285-299. Yamada, H. (1980). A historical study of typewriters and typing methods: From the position of planning Japanese parallels (Technical Report 80-05, February 1980). Department of Information Science, Faculty of Science, University of Tokyo, 7-3-1 Hongo, Bunkyoku Tokyo, 113 Japan. Zipp, P., Haider, E., Halpem, N., and Rohmert, W. (1983). Keyboard design through physiological strain measurements. Applied Ergonomics, 14(2), 111-111.