Many standard HCI software usability tests adapt well to testing the layout of control within an input device (particularly when the physical input gestures are visually guided as they are with standard remote controls). Candidate layouts can even be software-prototyped and tested using a touch screen. However, these standard tests adapt less easily to testing other attributes of input devices. This article describes some useful techniques which specifically address the usability problems faced in input device design. It illustrates how usability testing can guide (rather than just validate and confirm) design decisions and become an integral part of the design process. This article suggests ways of designing your usability tests such that the information you obtain is most applicable to future design iterations.
Input device design is concerned with sensing and assigning meaning to physical movements. Yet device development takes place on four levels:
This article discusses usability tests which can help guide design decisions at each of these four levels.
This article begins by discussing some of the problems with usability testing of input devices, describes a case study of three input device usability tests undertaken at Compaq, discusses two discount usability techniques (the Wizard of Oz, and Checklists), discusses the value of performance measures, and the relevance of performance models to usability inspections.
Discount usability
|
Performance measures
|
Performance models (usability inspection)
|
This paper assumes familiarity with many usability testing techniques. Refer to these sites and books for clarification of methodologies, definitions of unknown terms, or step-by-step descriptions of each technique. Detailed descriptions of how and when to implement various usability tests are available from many sources. Two of the best books on the subject are:
Two excellent online sources are:
Though concealed from the user, the transducer, firmware, and device driver significantly affect the input device's usability.
| Components of Input Devices | |
| Transducer | Senses changes in the physical properties of the input devices (such as position) and converts these into electrical quantities (e.g., voltage) |
| Firmware (On board Microprocessor) | Often converts the signal from Analog to Digital and may use a transfer
function (a mathematical model) to derive the controlling action (usually cursor control)
from the input signals Generates interrupts signaling that new data is available from the device |
| Device Driver | Reads the data from the port on the computer and generates events which are handled by the operating system or the applications |
We can distinguish indirect (manipulative, or mediated) input devices (e.g., joystick, trackball, mouse, keyboard, tablet, etc.) from direct input techniques (i.e. touch screen, lightpen). Pressing buttons on a stylus is indirect, while using the pen on the screen is direct. With direct input techniques users needn't monitor the cursor's position, but can "touch" the object directly. With mediated input, instead of pointing to what we want, we manipulate a computer-monitored control system (such as a mouse) to position a cursor over the item we want.
Keywords: usability testing, input devices, human computer interaction, design, development, overview of techniques for assessing input devices, industrial design, manipulative gestures
| "... the design methods for pointing devices ... seem to lack principle and are not guided by ergonomics research. Developers go through cycles of building a device, informally observing how a few users perform (usually part of the development group or their friends), tweaking various parameters and then re-testing. The cycle is repeated until a vague level of satisfactory performance has been achieved." – (Douglas & Mithal 1997:6) |
The design and evaluation of novel physical input devices presents unique challenges within the field of HCI. The guiding mantra of HCI – "Test early, test often" – applies equally to input devices. Yet tests which provide meaningful usability information to device designers are difficult to achieve, particularly in the early stages of development. This is largely because:
Usability tests can easily reveal when an input device does not meet the users needs, the challenge is to find out why, and to use this information to guide future iterations of the design.
After measuring the participants' hands for breadth and length, Compaq had users test:
| 14 | Non-functioning models provided by the Industrial Design Group Various sizes and designs (from conventional to non-conventional) All were painted the same metallic Grey |
| 5 | Standard Writing Pens |
| 3 | Mechanical Pencils |
| 3 | Functioning competitive products |
Instructions: Participants were asked to...
Task: Participants were asked to...
The preferred shapes were then refined for the next iteration of testing.
Three candidate pen tip materials and three competitive pens were used on six different writing surfaces.
For each pen:
Results: Researchers were able to determine which combinations (of pen tip and writing surface) were perceived favorably and how the candidate tips compared to the competitive benchmarks. Researchers were also able to drop those writing surfaces with poor user acceptance from consideration.
In 1991 Compaq conducted usability tests on various key switches. In addition to carefully recording and classifying keying errors (i.e., thiis type of error may suggest too little key-resistance or key-travel, while ths type may suggest too much) and comparing typing speeds, user preference played a major role in their usability testing:
Post-commercialization performance tests on several recent "innovations" in the input device industry (the Microsoft Inteli-Point mouse, and the keyboard-embedded IBM Track-Point isometric pointing stick) show that these devices are actually less efficient than their predecessors (Zhai et al., 1997; Douglas et al., 1997). Despite both products' commercial success, this seems to suggest that all is not well with early-design-usability-testing of novel input devices.
The more you have invested in development, the greater the pressure to commercialize a product. It is expensive, painful, and sometimes impossible to start over once have gotten to the point of functional models.
| "... by the time we can user test the interaction with functional models, the form factor that supposedly should support the interaction is already frozen." – Gomoll & Wong (1994), explaining Apple's User-Aided Design strategies. |
Discount usability tests are quick, inexpensive, and capable of providing valuable information before committing to a design direction. Discount usability has made substantial inroads into the software industry and many techniques are thoroughly documented. Some of these techniques can be adapted to the input device design process with minor modifications (sometimes the only modifications are in the types of things you look for). An example of this is The Wizard of Oz technique.
The Wizard of Oz technique uses a human to simulate the response (or part of the response) of the system. Often users are unaware (until after the experiment) that the system was not real. This technique can allow you to user test device concepts and experiment with techniques which would take months, and sometimes years to fully implement.
Typically the "wizard" sits in a back room, observes the user's actions, and simulates the system's responses in real-time. For input device testing the "wizard," will typically watch live video feeds from cameras trained on the participant's hand(s), and simulate the effects of the observed manipulations.
This technique is typically more useful for gathering information and observations on which to base early design decisions, than it is for product refinement.
When the input sensing is simple (e.g., operates in a single dimension) or involves gross motor movements. The "wizard" has to be able to quickly and accurately discern the user's intent. Hence this technique would be ineffective if you were developing a 6-axis 3D input puck. The output must also be sufficiently simple that the "wizard" can simulate or create it in real time. Hence if the output is intricate tactile feedback, the Wizard of Oz technique may be impracticable.
The Wizard of Oz is a very flexible technique principally because of the flexibility of the human "wizard". Feedback generated by the "wizard" needn't even be computer mediated (e.g., force feedback could potentially simulated through a series of levers physically manipulated by the "wizard").
Depending on how far along you are in the design process the Wizard of Oz technique can allow you to
The Wizard of Oz technique should provide you with valuable information on which to base future designs.
This technique should be approached as an tool for information gathering. Decide what you want to find out, and design your tests and your prototypes around this.
| "Prototypes for usability tests should be designed as tools for exploring and learning (observing), rather than as potential final designs. First learn about the interaction, then derive the industrial design from this information" – Gomoll & Wong (1994) |
The cognitive effort needed for successful use varies between manipulative techniques. Strategically designed prototypes will vary the manipulative techniques in an attempt to determine which technique corresponds most closely to the task.
Three Approaches to Prototyping |
|
| Rapid Prototyping |
|
| Reusable Prototyping Evolutionary Prototyping |
|
| Modular Prototyping Incremental Prototyping |
|
Brief: We want to enable users to scroll through documents using physical manipulation and this new functionality should be integrated within a traditional mouse. We will use the Wizard of Oz technique to gather information on the usability of various control mechanisms and manipulative gestures.
We will mock up several prototypes by adding various non-functional control actuators to otherwise identical off-the-shelf mice. Possible actuators include:
Though ergonomic or cost analyses might enable you to rule out some of these concepts out up-front, if testing them might provide valuable information about the interaction they should still be included. Our objective is to determine the kinds of physical manipulations which best represent the scrolling task, and tests will be more informative if you have many styles represented. If time constraints and prototyping costs force you to limit your testing to a short list, select those devices which will tell you the most about the interaction.
Checklists help ensure that you consider usability principles in your design. For software, these checklists usually form the basis of a usability inspection. I have been unable to find similar checklists pertaining to Input Device Usability. The collection of questions bellow have been culled and rephrased from the literature, and call attention to some common usability shortcomings in input devices.
Begin by deciding upon the usability checklist(s) you'll use to judge the attributes and interaction the device. You may want to tailor the guidelines to suit the exact issues faced by the device's user.
Running through a list of pointed questions such as these can be useful at many points in the design process. Early on checklists may aid choosing between several directions under consideration, later on they may form part of a formal usability inspections or evaluation.
| Input Device Usability Checklist | |
| Layout: |
|
| Product Semiotics: |
|
| Movement Characterization: |
|
| Feedback: |
|
| RSI/Ergonomics: |
|
| Design of the controls: |
|
Performance models help usability inspectors understand principals of input movements and can contribute to input device usability in a number of ways:
Developed by Card et al. (1980; 1983) the Keystroke Level Model (KLM) is an engineering approximation of human performance. The KLM decomposes a sequence of interface actions into unit-tasks, each of which are given a standard time. The sum of these unit-task times yields the task time (assuming consistent error-free use of the device(s)). The KLM does not factor in extraneous times (e.g., unscheduled pauses), and assumes that the unit-tasks will be performed in a specific manner and order.
There are two groups of unit-tasks:
The KLM's principal use for input device developers is as a tool to factor device-switching times into performance models for various input device. For example, if you were interested in estimating what performance improvements you could expect if you eliminated device-switching time, the KLM would allow you to do this without engaging in a controlled experiment. Designers can use this model to predict practiced human performance with an 80% accuracy (Baber, 1997).
The Network Model extends the basic KLM to accommodate probabilities and variation in performance including considering error. The network model overcomes the sequential limitations of the KLM. Performance is analyzed as a network of possible routes between unit tasks rather than a rigid sequence. The number of possibilities varies at each juncture during the task. Though much more complicated than the KLM, the Network Model represents human behavior more accurately. The Network Model also allows researchers to predict and compare times for alternate routes through the network.
Fitts' law predicts overall movement time for pointing. Movement time is calculated as a log inverse relationship (known as the index of performance [IP]) between the target distance and the target width. Smaller targets and longer distances take longer to execute.
Fitts' law has been experimentally confirmed (and coefficients have been determined) for the following pointing devices: mouse, touch tablet, isotonic and isometric velocity-controlled joysticks (track-point), trackball, and head and eye trackers (Douglas et al., 1997). With the correct coefficient, pointing times can be accurately predicted given the target distance and width.
Though the IP can be useful in comparing pointing speeds between devices, cross-experiment comparison of IP's should be done with great care. Due to differences in experimental task, limbs used, features of the devices, and skills of participants, Fitts' Law experiments do not provide an 'absolute' measure of device performance. Having said that, however, a table summarizing data from Fitts' Law studies of various pointing devices can be found in Baber (1997:181), and may prove useful in gauging the results of your own Fitts' Law tests.
Fitts' law describes only the movement time, and ignores what happens during the movement. Hence Fitts' law can be used to compare and benchmark devices, tasks, or limbs (within experiments), but cannot explain any resulting performance differences.
For input device developers and designers, Fitts' Law provides an excellent means of gauging where your prototype or device stands in relation to a benchmark (e.g., the mouse, or a competitive product). Software is available (via anonymous ftp) to facilitate this testing (see below), so running a Fitts' Law test on pointing devices is simply a matter of connecting the device(s) to your PC, running several participants, and analyzing the data-files. Rough benchmark tests needn't involve numerous participants, but if reliability is important your testing procedures will have to be much more rigorous. Examples of Fitts' Law experimental methodologies can be found below.
Psychomotor models describe the manner in which people plan and execute movements and are principally developed by psychologists. Relevant work in this area dates back as early as 1899. For a summary of the Psychomotor models relevant to Input device Design, refer to Douglas & Mithal (1997:11-34). The psychomotor model most relevant to input device design is the Stochastic Optimized Submovement (SOS) which is discussed below.
Stochastic Optimized Submovement (SOS) is a microanalytic description of rapid aimed movement. It defines the pointing action as a sequence of submovements for which it predicts the relative size and accuracy.
| SOS Model of Human Motion
During Pointing (after Walker et al. 1993) |
|||
| Movement Phase | Description | Target Attribute(s) Influencing Performance | Timing |
| Initiation | visually select target; define path to the object (aim the cursor, spatial planning of movement) | None | 30ms (constant) |
| Execution | consists of a submovement, pause, submovement, pause | Width & Distance | 331ms to 628ms |
| Verification | measures the time between the end of the mouse movement, and the mouse click (selection) | Width | 157ms to 227ms or 20% to 36% of the total pointing time |
Understanding these models of human motion can furnish directions for improving pointing device performance (for example, Walker et al. (1993) observed that reducing verification time could significantly reduce total pointing time).
Psychomotor studies are perhaps most relevant when refining transfer functions.
The transfer function is the mathematical model which derives cursor control (other types of control include document scrolling, and playback speed controls for audio) from the input signals received from the transducer. For pointing devices, transfer functions are typically linear (the displacement of the device is directly related to the displacement of the pointer), or non-linear (the speed of the pointing gestures influence the distance of cursor travel). Non-linear transfer functions can reduce the mouse's footprint, or the number of "retakes" necessary with a trackball, yet they have been shown not to improve pointing speed with a mouse (Jellinek & Card, 1990).
Gain describes the ratio between the control movements and the cursor movement (for a mouse this would be the magnitude of displacement of the mouse and the displacement of the cursor, respectively). However when sensing properties other than displacement (e.g., isometric pointing devices -- which translate force into displacement) the effectiveness of the transfer functions has a much greater impact on the device's usability (Rutledge & Selker, 1990).
Transfer functions do not yet factor the cursor's position within the field of use (e.g., the screen), or its proximity to "clickable" targets into the mathematical model. One could imagine how transfer functions which incorporated this information might speed pointing tasks. For instance, researchers have proposed that a "gravitational" force which pulled the cursor towards "clickable" items might reduce pointing times.
Usability testing will be increasingly important to refining and fine-tuning the transfer functions of future input devices. Yet as the transfer functions increase in complexity, informative usability testing becomes more difficult.
An understanding of models of movement is useful in evaluating the performance of transfer functions. Usability inspectors should understand how the transfer function can affect and enhance performance. For example Rutledge and Selker (1990), experimentally determining the force-to-motion transfer function to optimise the pointing speed of their Pointing Stick found the following characteristics to be desirable:
They then used these characteristics to guide further refinement of the transfer function.
| Compatibility (after Baber, 1997:99-101) |
|
| Movement | Movement compatibility occurs when the manipulative action corresponds to
the user's intentions. For example, the movement in the cursor corresponds to the manipulative movement of the mouse. However the mouse is manipulated within the horizontal plane of the mousepad, while the cursor typically moves within the vertical plane of the screen. Movement compatibility describes the relationship between the manipulation and the movement. |
| Spatial | Spatial compatibility occurs when the location of the control establishes its relationship to the object it controls. (i.e. it could be next to the object, or mirror the spatial arrangement of the object) |
| Operational | Operational compatibility occurs when the object's design clearly communicates its relationship to the user (e.g., how it is held in the hand) and its affordances (e.g., how it is manipulated) |
| Conceptual | Conceptual compatibility occurs when the level and type of feedback is sufficient to resolve ambiguity yet not cause distraction |
| Modality | Modality compatibility occurs when the feedback and the controlling action operate within the same modality. Visual feedback to physical manipulation is an instance where the modalities are incompatible. |
| "With the best-designed products, we often do not realize we are even using a product, so smoothly does the performance of a task appear. As task performance "breaks down", so our attention focuses on the tool rather than the task." – (Baber, 1997:82) |
Characterization and classification can be useful in input device usability inspections. The technique essentially takes stock of the possibilities, carefully analyzing the characteristics of each. This helps ensure that better matches between task and technology are not get overlooked.
Five levels of analysis can prove helpful in determining the appropriate gesture & sensing mechanism for the task. These levels characterize candidate:
| Characterization of Input Types | ||
| Input Type | Examples | Characteristics |
| Continuous | Pointing Scrolling |
|
| Discrete | Keystrokes |
|
| Classification
of Gestures (after Koons et al. 1993) |
|
| Symbolic Gestures | convey linguistic meaning |
| Iconic Gestures | describe specific properties of an object |
| Deictic Gestures | point to an object within the environment |
| Pantomimic Gestures | act out or mime the use of a tool (e.g. miming the use of a chisel, or miming rotating selected object) |
| Note: Though, currently, gestures used to physically interaction with the computer hold minimal semantic content, this may change in the near future. | |
| Characterization
of Isometric & Isotonic Sensors after Douglas & Mithal (1997:40) |
|
| Isometric |
|
| Isotonic |
|
| Control
Resistance Type Control resistance is the principal determinant of the "feel" of the device |
|
| Resistance Type | Characteristics & Examples |
| Elastic (spring loading) | Spring loading is most often used insure the control mechanism returns to a central position on release and is most likely to be used in velocity control devices |
| Static and Sliding Friction | Though unavoidable in most controls, friction provides a valuable form of feedback for input devices |
| Viscous Damping | Movement is resisted by a viscous substance |
| Inertia | Weight and size |
| Analysis of Joystick Types | ||||
| Cursor Movement derived from... | Lever movement | Means of sensing | Type of kinesthetic feedback | |
| Switch-activated | Activation of switches (limited to 8 directions of movement) | Lever moves in the direction of the supplied force provided the force is sufficient to activate the switch | Switches | Possibly auditory (switch activation) and tactile displacement |
| Displacement | lever displacement | proportional to the applied force (remains in the end condition) | Displacement | |
| Spring-loaded | lever displacement | proportional to the applied force (returns to the center when released) | Resistance to applied force (from the springs) as well as displacement. | |
| Rate Controlled | As force increases the cursor appears to accelerate | Resistance to Pressure | ||
| Isometric | Pressure (force) applied to the lever | no appreciable movement | Strain gauges - Force Sensing Resistors | Pressure as kinesthetic feedback (isometric) |
Deciding what to measure is often the most important consideration when using performance tests to get usability information. Quantitative measures are useful as benchmarks (see Fitts' law, above) or as a way to gauge improvements between iterations. However performance measures can sometimes also provide valuable information on which to base design decisions.
For example, the following performance measures could be used to guide future iterations of a novel keyboard design or layout:
Other performance measures are listed below:
| Measures
for Research From (Baber, 1997:58) |
|
| Time-based |
|
| Task-based |
|
| Accuracy |
|
| Subjective |
|
| 1 A regressive
action is one which takes the user back to a previous system state. 2 A failed action is one which does not result in the fulfillment of the users goals (regardless of whether or not the goals could be deemed correct). 3 Users could either not understand system prompts or incorrectly interpret the feedback presented to them. |
|
| Measurable Human Performance Goals | |
| Learning |
|
| Performance |
|
| General |
|
| Comfort |
|
| Reliability |
|
| Performance
Measures (Douglas & Mithal, 1997) |
||
|
Pointing devices focus primarily on:
|
Less commonly examined are:
|
Analysis of errors can be an extremely informative. Compaq, for example, logged incidences of over 30 different types of typing errors when conducting their keyboard usability tests. Errors occur when human performance breaks down, detailed analysis of errors can help determine what role the device played (if any) in this failure. The SHERPA table below provides a useful template on which to base error classification for the input device being tested.
| SHERPA (Systematic Human Error Reduction and Prediction Approach) (after Embry, 1986) |
|
| Planning errors | |
| P1 P2 P3 P4 P5 |
|
| Action errors | |
| A1 A2 A3 A4 A5 A6 A7 A8 A9 |
|
| Checking errors | |
| C1 C2 C3 C4 C5 |
|
| Retrieval errors | |
| R1 R2 R3 |
|
| Information communication errors | |
| I1 I2 I3 |
|
| Selection errors | |
| S1 S2 |
|
There are basically two approaches to understanding usability through movement looking at the gestures, and looking at the output.
Gestures:
|
Output:
|
Micro-movement Analysis: Real-time performance is analyzed at the milli-second level with detailed time and displacement data. This kind of analysis can be useful in understanding why an input device performs poorly for a continuous tasks, such as pointing (Douglas & Mithal, 1997).
There is around 20 years of research in the ergonomics literature on issues related to computer input. This can be a valuable source of usability information to guide input device design. Below are summarized some of ergonomist's foremost usability concerns.
Repetitive Strain Injury (RSI) is a form of Cumulative Trauma Disorder (CTD) resulting from repeated exposure to doses of strain, which, if experienced in isolation, would be innocuous. The immune response (swelling) combined with the anatomy of the wrist and hand compounds the effects of this strain.
Wrist postures have an enormous impact on the strain caused by repetitive hand and finger motions. As the wrist moves into greater extension, the increase in carpal tunnel pressure is almost linear (Hedge et al., 1996). Repetitive ligament movement under such conditions injures the tissue. The body’s natural response to injury is swelling. This further aggravates the problem by creating still more pressure within the carpal tunnel.
Many of the factors contributing to RSI (e.g., body posture, frequency of breaks, stress levels, and variety in the job) are beyond the designers control. However designers can, choose product forms which deter wrist postures which further aggravate the strain of repetition.
The following wrist postures concern Ergonomists the most:
Forearm rotation is also a (lesser) cause of strain.
Muscle strain can be measured using a technique called electromyography. Accurate electromyographic readings, however, are hampered by interference from skin, problems with precise receptor placement, and variation in levels of habitual muscular tension between participants. Posture can be measured using various sensing technologies. For example Microsoft used VPL Fiber Optic Data Gloves in usability tests of the Microsoft Mouse 2.0. This allowed them to measure joint angles in user's hands while using the Microsoft mouse and competitive products and compare this data with neutral hand postures.
Contact points between the device and the hand, wrist, and fingers should also be carefully considered. Even light pressure on certain areas of the hand can restrict blood flow, or cause undue rubbing on tendons. Both of these increase susceptibility to RSI.
People contort their hands and bodies with varying degrees of discomfort to align themselves with their input devices. Wrist postures and input actions aren't all that contribute to input-device-induced repetitive strain. For example, the constrained posture and lack of mobility imposed by the keyboard and mouse contribute significantly to the risk of RSI. Areas pay attention to include:
All of these may lower the user’s concentration and productivity.
Muscles are designed for movement. Static posture is significantly more strenuous than movement and restricts blood flow to the strained area(s). Input devices whose use promotes static loading will increase users' susceptibility to RSI. Typing on conventional keyboards, for example, statically loads the knuckle-base, shoulders, wrists, and arms.
Fine motor activity tends to induce static muscle loading, particularly when:
Sustained static loading leads to muscle fatigue and reduces circulation to the affected areas thus increasing the hazard of RSI.
Usability tests are vital to practitioners in the field of input device design, yet the tests need to be designed so as to gather richer information from the participants. Usability tests needn't be costly, however, put a lot of thought into the planning to insure that the information you gather is pertinent to the design questions at hand.
HCI Bibliography Bibliography and review of HCI books and other reference
Baber, Christopher. Beyond the Desktop: Designing and Using Interaction Devices. San Diego: Academic Press, 1997.
Balakrishnan, R., & Patel, P. (1998). The PadMouse: Facilitating Selection and Spatial Positioning for the Non-Dominant Hand. Proceedings of ACM CHI'98 Conference, pp. 9-16.
Beringer, D.B. & Peterson, J.G. (1985). Underlying Behavioral Paramaters of the Operation of Touch-Input Devices: Biases, Models and Feedback. Human Factors, 27(4), pp. 445-458.
Buxton, W., & Myers, B.A. (1986). A Study in Two-Handed Input. in Human Factors in Computing Systems: CHI'86, pp. 321-326. New York: ACM.
Brunner, H., & Richardson, R. (1984). Effects of Keyboard design and typing skill on user keyboard preferences and throughout performance. In D. Tattooed and C. McCann, (eds.), Proceedings of the 1984 International Conference on Occupational Ergonomics. Rexdale, Ontario: Human Factors Association of Canada, pp. 267-271.
Card, S.K., English, W.K. & Burr, B.J. (1978). Evaluation of Mouse, Rate-Controlled Isometric Joystick, Step Keys, and Text Keys for Text Selection on a CRT. Ergonomics, 21(8), pp. 601-613.
Card, S.K., Moran, T., & Newell, A. (1980). The Keystroke Level Model for User Performance Time with Interactive Systems. Communication of the ACM, 23, pp. 396-410.
Card, S.K., Moran, T., & Newell, A. (1983). The Psychology of Human-Computer Interaction, Hillsdale, NJ: LEA.
Card, S.K., MacKinlay, J.D., & Roberson, G.G. (1990). The Design Space of Input Devices. in Human Factors in Computing Systems CHI'90, pp. 117-124. New York: ACM.
Card, S.K., MacKinlay, J.D., & Roberson, G.G. (1991). A morphological analysis of the Design space of Input Devices. ACM Transactions on Information Systems, 9, pp. 99-122.
Chen, C., Burrastero, S., Tittiranonda, P., Hollerbach, K., Shih, M., &Denhoy, R. (1994). "Quantitative Evaluation of 4 Computer Keyboards, Wrist Posture and Typing Performance". Proceedings of the Human Factors and Ergonomics Society 38th Annual Meeting 1994, Santa Monica CA. Human Factors and Ergonomics Society, pp. 1094-97.
Cohen, Leon (1970). "Interaction Between Limbs During Bimanual Voluntary Activity". Brain, 93, 259-272.
Conrad, R. (1967). Performance with push-button arrangements. Het PTT-Bedrijfdell, 15, pp. 110-113.
Conrad, R., & Longman, D.J.A. (1965). Standard Typewriter versus Chord Keyboard: An Experimental Comparison. Ergonomics, 8, pp. 77-88.
Cooper, William E., ed. (1982), Cognitive Aspects of Skilled Typewriting New York: Springer-Verlag.
Douglas, Sarah A. & Mithal, Anant Kartik. The Ergonomics of Computer Pointing Devices. New York: Springer, 1997. TK7887.5.D665 1997
Embrey, D.E. (1986). SHERPA: a systematic human error reduction and prediction approach. Proceedings of the International Topical Meeting on Advances in Human Factors in Nuclear Power Systems. Knoxville, Tennessee.
Fleischer, A.G., & Becker, G. (1986). Free hand-movements during the performance of a complex task. Ergonomics, 29, pp. 49-63.
Garrett, J.W. (1971). The human adult hand: some anthropometric and bio-mechanical considerations. Human Factors, 13, pp. 117-131.
Gardiner, Margaret M. & Christie, Bruce. Applying Cognitive Psychology to User-Interface Design. Toronto: John Wiley & Sons, 1987. TK7887.5.A66 1987
Gill, R., Gordon, S., Dean, S., & McGhee, D. (1993). Integrating cursor control into the computer keyboard. Proceedings of the 35th Annual Meeting of the Human Factors Society. Santa Monica: CA: Human Factors Society, pp. 256-260.
Glencross, D.J., & Barrett, N. (1989). Discrete movements. In D.H. Holding (ed.), Human Skills. Chichester: Wiley.
Haller, R., Mutschler, H., & Voss, M. (1984). Comparison of Input Devices for Correction of Typing Errors in Office Systems. in Proceedings of the Interact 84 Conference, First IFIPS Conference on Human-Computer Interaction, pp. 177-182.
Hargreaves, W., Rempel, D., Halpern, N., Markison, R., Kroemer, K., & Litekwa, J. (1992). Towards a more Humane Keyboard. CHI'92. Reading, MA: Addison-Wesley. pp. 365-368.
Harrison, B.L., Fishkin, K.P., Gujar, A., Mochon, Carlos, & Want, Roy (1998). Squeeze Me, Hold Me, Tilt Me! An Exploration of Manipulative User Interfaces. Proceedings of ACM CHI'98 Conference, pp. 17-24.
Hedge, A, McCrobie, D., Morimoto, S., Rodriguez, S., & Land, B. (Jan 1996). "Toward pain-free computing". Ergonomics in Design, Vol. 4, No. 1, pp. 4-10.
Jacob, R.J.K., Sibert, L.E., McFarlane, D.C., & Mullen, M.P. (1994). Integrality and Separability of Input Devices. ACM Transactions on Computer-Human Interaction, 1(1), pp. 3-26.
Kabbash, P., Buxton, W., & Sellen, A. (1994). Two-Handed Input in a Compound Task. in Human Factors in Computing Systems CHI'94, pp. 417-423. New York: ACM.
Kabbash, P., MacKenzie, I.S., & Buxton, W. (1993). Human Performance Using Computer Input Devices in the Preferred and Non-Preferred Hands. in Human Factors in Computing Systems CHI'93, pp. 474-481. New York: ACM.
Kawachiya, K., & Ishikawa, H. (1998). NaviPoint: An Input Device for Mobile Information Browsing. Proceedings of ACM CHI'98 Conference, pp. 1-8.
Koons, D.B., Sparrell, C.J., & Thorisson, K.R. (1993). Integrating simultaneous input from speech, gaze and hand gestures. In M.T. Maybury (ed.), Integrated Multimodal Interfaces. New York: Academic Press.
Lessenberry, D.D. (1928). Analysis of Errors. Syracuse, NY: L.C. Smith and Corona Typewriters.
Loeb, K. (1983). Membrane Keyboards and Human Performance. The Bell System Technical Journal, 6, pp. 1733-1749.
Loricchio, D.F. (1993). A comparison of keyboard-integrated pointing devices. In G. Salvendy and J.J. Smith (eds.), Human-Computer Interaction: Software and Hardware Interfaces. Amsterdam: Elsevier.
Mackinlay, J., Card, S.K., & Roberson, G.G. (1990). A Semantic Analysis of the Design Space of Input Devices. Human-Computer Interaction, 5(2-3), pp. 145-190.
Massaki, F., & Yasuhito, S. (1994). FingeRing: a full-time wearable interface. CHI'94. New York: ACM, pp. 81-82.
Motoyuki, A., & Soto, S. (1994). A Multi-Modal Mouse with Tactile and Force Feedback. International Journal of Human-Computer Studies, 40(3), pp. 443-453.
Norman, Donald A. (1993). Things that make us smart - Defending human attributes in the age of the machine. Reading, MA: Addison-Wesley.
Rabbitt, P.M.A. (1978). Detection of errors by skilled typists. Ergonomics, 21, pp. 945-958.
Ratz, H.C., & Ritchie, D.K. (1961). Operator Performance on a Chord Keyboard. Journal of Applied Psychology, 45, pp. 303-308.
Rutledge, J.D. & Selker, T. (1990). Force-to-Motion Functions for Pointing, in Proceedings of INTERACT '90: The IFIP Conference on Human-Computer Interaction, pp. 701-705, 1990.
Sellen, A.J., Kurtenbach, G.P, & Buxton, W.A.S. (1992). The Prevention of Mode Errors through Sensory Feeback. Human-Computer Interaction, 7(2), pp. 141-164.
Shaffer, L.H. (1975). Control processes in typing. Quarterly Journal of Experimental Psychology. 27, pp. 419-432.
Siebel, R. (1962). Performance on a five-finger chord keyboard. Journal of Applied Psychology, 46, pp. 165-169.
Verwey, W.B. (1993). Effects of extended practice in a one-finger keypressing task. Acta Psychologica, 84, pp. 179-197.
Zhai, S., Smith, B.A., and Selker, T. (1997). Improving Browsing Performance: A Study of Four Input Devices for Scrolling and Pointing Tasks, in Proceedings of INTERACT '97: The 6th IFIP Conference on Human-Computer Interaction, pp. 286-292.