This experiment has analyzed effects of three keyboard layouts on typing speed. Keyboards studied are:
The experiment consisted of four separate typing tests, three training periods, and three subjective satisfaction questionaire.
Four typing tests are identically administered. A subject is given a peice of paper containing typing data and they enter it into a testing program. An initial test is given on a standard keyboard (Keyboard A). This pre-test serves to familiarize subjects with test functionality. Assingment of keyboards for remaining three tests are based on a given User ID number.
Training periods allow subjects to become accustomed to a given keyboard. Subjects were encouraged to train for as long as they deemed necessary; but, most only trained for a few minutes. This amount of time is not long enough to become an expert on a given keyboard; but, it did serve to familiarize subjects with peculiarities of each keyboard.
A subjective satisfaction questionnaire was admistered during the course of test. The questionaire was presented after a subject had completed the typing test for a given keyboard. Each questionaire asks four questions concerning the subjects satisfactions with a given keyboard.
Experimental hypothesis were as follows:
For this experiment, each keyboard is considered a treatment (independent variable) and dependent variables are:
Typing speed is measured in words per minute and is determined from typing tests performance. User preference is determined from subjective satisfaction questionnaires given after keyboard usage. Error rate is a count of backspace key activations over test length. Error correction time is an average time for backspace key activation.
A pilot study served as a beta test for experiment testing software. Modifications made after pilot tests included refining error underlining and clearing of previous test text when returning from questionaire (instead of when you hit "Go" button. Consideration was given to shortening test input from it's current 155-197 words; but, we decided to leave at the current length because a good typist would be able get through a test in under three minutes.
A further analysis of pilot study results would have shown a costly error. Data generating functions appeared to be working properly; but, there was an error. This error surfaced after we had already run 6 subjects and those subjects had to be discarded.
Six subjects for this experiment were mostly University of Maryland students. Subjects, 3 males and 3 females, had varing exposure to a QWERTY keyboard. Since all subjects are performing all treatments, differing experience is not considered a threat to validity. A few subjects had experience with Keyboard C; but, none had any experience with Keyboard B. The subjects were assigned a unique ID numbers that determined keyboard presentation order.
| ID Number | Test 1 | Test 2 | Test 3 |
|---|---|---|---|
| 01 | A | B | C |
| 02 | A | C | B |
| 03 | B | A | C |
| 04 | B | C | A |
| 05 | C | B | A |
| 06 | C | A | B |
Experimental materials included three keyboards, testing program,
typing input data, and computer system. Keyboards are all commericially
availiable (click on the image to see a larger image).
This keyboard is a generic QWERTY keyboard. This keyboard, a Digital
model RT101, is similar most keyboards in use today. It served as
experimental control.
2.4.2 Keyboard B - "Split Space Bar"This keyboard is a generic QWERTY keyboard, space bar modification
withstanding. The space bar has been split in half with the left side
remapped to the backspace key. This keyboard is a commerically
available
from Compaq as model number 237835-301.
2.4.3 Keyboard C - "Angled Keys"This keyboard is a highly modified QWERTY keyboard. Modifications
take several forms. The most noticable modification is the split, sloped
key placement. There is also a built-in palm rest, enlarged spacebar and
slightly enlarged backspace key. Lock lights (Num, Caps, and Scroll)
have been moved to the center of the keyboard. This keyboard is commercially
availiable from Microsoft as model 59758 ("Natural Keyboard")
The testing program was developed in Visual Basic 4.0. The program was written specifically for this experiment and included four typing tests, three practice sessions. A novell game was also availiable to help subjects become accostumed to a unique backspace key positioning. The program also recorded data and wrote out a simple ASCII data file for analysis. The program also provided an on-line subjective satisfaction user questionaire.
Typing test data was provided to subjects on 8.5" x 11" paper. Text excerpts are from the Associated Press. These excerpts were selected to be as even as possible. The font used is Book Anitqua (14 Point) for greatest possible clearity.
| Test Number | Words | Character | Character/Word |
|---|---|---|---|
| Pre-Test | 197 | 932 | 4.73 |
| Test # 1 | 186 | 865 | 4.65 |
| Test # 2 | 155 | 714 | 4.60 |
| Test # 3 | 165 | 751 | 4.55 |
Computer systems used were Pentium-based PC's running Windows 95 with 17" monitors running in a resolution of 1024 x 768.
Experimental procedure was as follows:
Keyboard presentation order is pre-determined by a subject's ID number. It is designed to cover all permutations of keyboard ordering. A dialogue box prompts for switching of keyboards and tells which keyboard should be connected next. Six subjects allows for each permutation to be covered.
Training and testing period allowed a subject to take as much time as they need to become accustomed to a new keyboard. This time is subject determined. Training takes on two forms. A simple statement typing exercise allowed subject to become accustomed standard key placement. The statements were excerpted from "The Story of the Four Little Children Who Went Round the World", from Nonsense Songs, Stories, Botany and Alphabets by Edward Lear (1871) and were selected to because of their representational alphabet coverage. A space game was provided to encourage subjects to learn of space bar modifications
Two major problems were overcome during this experiment (keyboard and custom software).
The first problem had an easy solution. We spent a large amount of time trying to construct a keyboard with a split space bar. We found several commercially availiable systems; but, none mapped a space bar to anything else other than a space bar. After destroying 3 or 4 older keyboards, we discovered that Compaq had just come out with a keyboard that does precisely what we were trying to do.
The second major problem was not as easily solved. Creating the testing application took much longer than anticipated. The resulting application works well; but, required more resources than originally anticipated. Major obstacles included underlining of incorrect words and data generating functions. The first batch of data from subjects had to be discarded because of a software problem that failed to record backspace key timings.
Test length continued to be a bit of a problem. For poor typists, it seemed to take a very long time to complete (~45 minutes). Good typists could be done in 15 minutes. This is a problem we didn't resolve. One possible resolution would be to limit typing test to a specific time; but, that might not generate enough usage. Because testing time was so long and we had to discard subjects, we were only able to run six subjects after repairing software glitches.