Your company produces phone/fax machines, all which use a hardware phone pad. People can use the phone pad to enter alphabetic characters, which is used to create an on-line directory containing people's names, phone numbers, and notes on them. Users can then scroll through the directory and direct dial a person. Through caller id, the details of the caller are automatically displayed whenever the phone rings. Customers usually rave about these features.
You have been placed in charge of producing a new version of your company's phone/fax machine.This version will have a small screen, a mouse, and no hardware keyboard and/or keypad. Typing of numbers and text will be done by displaying a simulated keyboard on the screen, and by selecting keys with the mouse (called mouse-typing).
The problem is deciding what keyboard design to use. On one hand, you can simulate the conventional phone pad already used on your fax machines and all telephones (below left). People type letters by multiple clicks of keys, and by using the '>' to move ahead a letter. For example, typing 'hi' involves pressing '4' twice (for h), the '>' (for moving forward a character), and pressing '4' three times (for i). Alternatively, you can simulate an alphabetic keyboard (below right). However, your marketing department has convinced your boss that standard qwerty keyboards should not be used, as they believe many users of this product are not experienced typists and that all people will be fairly fast on one of the other layouts.
The choice of alphabetic or phone keyboard is not clear, for their are several tradeoffs.
You have decide to run an experiment comparing typing speed and error rate on these keyboards, and will use the results to make recommendations about which layout should be used in the product. To set the scene, some background to existing keyboard layouts and computing without keyboards is described below.
Early typewriter mechanisms were quite crude. Typebars (the levers that actually struck the paper) tended to jam, particularly when two nearby keys were hit in rapid succession. The first typewriters had keyboards that were laid out alphabetically, and fast typists would routinely run into jamming problems.
The Qwerty layout (below middle-left), named for the first six keys in the top-left row, was developed in the late 1870's to minimize the jamming problem. It increased the spacing between common pairs of letters so that sequentially-struck keys would not jam. It was also developed for two-finger typing; touch typing was unknown until the early 1900s. The Qwerty layout is found on virtually all of today's computers.
Today, typewriter mechanics and jamming are no longer a concern, and several alternate layouts to Qwerty exist. Alphabetic keyboards (above right) that lay out the keys in alphabetic order have recently appeared on several devices, such as pocket computers, calculators and children's toys. The idea is that non-typists can find keys more quickly because they know the alphabet. Surprisingly, most studies have found that typing speeds on alphabetic keyboards are lower than that of the Qwerty, even for non-typists.
The Dvorak keyboard (below, bottom-left), developed in the 1930s, was based upon time and motion studies of touch-typing of English text. Its design simplifies a typist's finger motion, increases the amount of time spend on the home (middle) row, and balances the amount of work done by all fingers. Frequently used letters are placed on opposite halves of the keyboard to maximize two-handed typing sequences (which is faster than one-handed typing of characters in sequence). Studies contrasting Dvorak and Qwerty typing confirm that Dvorak is indeed faster after training; however, there is some disagreement on the amount of improvement.
We are beginning to see more and more computers without keyboards. Examples are:
In spite of the prevalence of graphical interfaces, many situations still require text input (as in the fax machine product above). The most promising way to do this without a keyboard is by character recognition, where you use an electronic pen to print letters. The computer either recognizes the letter shapes or the order of strokes that make up the letter, and replaces it with a typeset version of its guess. The catch is that computers using character recognizers are still fairly rare, are notoriously inaccurate (especially with messy writers), and require users to train themselves or the character recognizers. Also, handwriting tends to be slower than typing, especially for lengthy passages.
Another option to text input is mouse-typing, where a person does the equivalent of one-finger typing of keys on a screen instead of touch-typing a physical keyboard. Because mouse-typing uses only one hand, we can consider radically different keyboards. In an earlier 481 class, for example, a circular mouseboard (above, middle-right)was tested that places letters in a circle, with most frequent letters having the larger keys and being closer to the center (the idea was that it should be easier to select frequent keys). Another class tested an alphabetic strip keyboard (above, top) that presented the keys in alphabetic order across a single row.
It isn't immediately obvious which of the qwerty, dvorak, alphabetic, circular, phone, or strip layouts work best for mouse-typing. Consider the following points:
Previous 481 studies have shown that people consistantly mouse-type faster on the Qwerty than on the other keyboards (in spite of what marketting says!). However, we still don't know how well people perform on the phone pad, and the alphabetic keyboard. Which do you think is best?
2.1 Hypothesis. You will be collecting data and comparing through statistics the mouse-typing speeds on the alphabetic and phone layouts. Your professor will do a more complex analysis that takes into account a subject's typing proficiency.
Null Hypothesis (for you): There is no difference between people's mouse-typing speed when using an alphabetic and a phone pad layout on a simulated keyboard. People are expected to have several years of experience typing with standard Qwerty keyboards and with a mouse.
Null Hypotheses (for your professor): People's skill-level in traditional typing has no effect on their proficiency with mouse-typing with either of the layouts.
2.2 Subjects. Subjects are members of your group, all University of Calgary CPSC 481 students. You may use other subjects as well, but make sure they have equivalent computing and typing experience. Subjects are expected to have several years of typing experience, although a broad range of typing speeds are expected.
2.3 Materials. Custom software was developed that presents a soft keyboard showing only the letter keys and a space bar on a screen. Particular keyboard layouts can be selected through a pull-down menu. The keyboard is unresponsive until a button labeled "Press here to begin" is selected. The keyboard remains active until the button, now labeled as "Press here when done" is selected again. Data is automatically collected that indicates the start and end time of each typing trial, the type of keyboard, and the text actually entered.
The computer used is a Sparc II color machine. You install and execute software as follows
Appendices to this report provide instructions to subjects, text for copy typing, consent forms and questionairres, and individual work sheets. The TAs also have a folder of readings about typing speeds and experimental design that you may wish to use for background reading.
2.4 Method. Each subject goes through the following steps.
a) Prolog:
b) Trial 1-2: Typing speed with physical keyboard
c) Trials 3-8: Typing speed with the two keyboards
d) Post-session
Each subject will be identified by a unique number. For example, if your group number is "B02-03", the third subject you run will be Subject B02-03-3.
You will be required to submit your numeric and qualitative results to the TA electronically (Appendix 4). Your numeric data should be recorded in EXACTLY the format described on the worksheet. Your TA may ask for additional details.
Your TA will collect the results of all the groups, and will make them available to you electronically. The TA will show you how to calculate a t-test contrasting the typing rates on the two layouts.
As well, you will plot a scattergram that contrasts subjects' typing speed on the conventional typing test (the X-axis) with their typing speed on each of the other "soft" keyboards (the Y-axis). For each scattergram, draw a straight line that best "fits" the result. (I will be doing this through statistics as well, through something known as a regression analysis.)
Your write up must include a table of summary statistics and descriptions of the t-test result. Your TA will show you what is needed.
You will write a paper following the style specified in the accompanying document "How to structure reports on experiments in human-computer interaction". This assignment specification gives you a bit of a head start for the introduction, description of the experiment, and expectations in the results section. You can use some of this material in your report (but in your own words!). Do not copy out the method or exact instructions; you can refer to this page instead. Your TA will give you additional information, and you are expected to flesh out your report where necessary.
It is up to you to interpret what the results mean. You should discuss this with your group, and think about it. I am particularly interested in the impact your findings have to the application constraints described in this scenario, the general use of soft keyboards, the implications to other practitioners, your critical reflections, and a research agenda of future study and issues. Use both the qualitative and quantitative findings of your study, as well as the extra readings that are available through the TA.
I will also give you my results of your experiment contrasting people's normal typing speed to their mouse-typing speed. Use this in your discussions section, particularly in the "relation to other work" sub-section. Your TA will also tell you the results of similar experiments. If your results differ from the published ones, you should discuss what you think the causes are.
Include your rough data and worksheets in an appendix. This should include the work sheets where you recorded pre- and post-experiment answers to questions, subject's comments, etc.
Your conclusion should make a recommendation and should justify your results.