Cursor control by mouse and keyboard
Citation for published version (APA):Vereijken, B. (1987). Cursor control by mouse and keyboard: an experimental comparison. (IPO-Rapport; Vol. 615). Instituut voor Perceptie Onderzoek (IPO).
Document status and date: Published: 01/10/1987 Document Version:
Publisher’s PDF, also known as Version of Record (includes final page, issue and volume numbers) Please check the document version of this publication:
• A submitted manuscript is the version of the article upon submission and before peer-review. There can be important differences between the submitted version and the official published version of record. People interested in the research are advised to contact the author for the final version of the publication, or visit the DOI to the publisher's website.
• The final author version and the galley proof are versions of the publication after peer review.
• The final published version features the final layout of the paper including the volume, issue and page numbers.
Link to publication
General rights
Copyright and moral rights for the publications made accessible in the public portal are retained by the authors and/or other copyright owners and it is a condition of accessing publications that users recognise and abide by the legal requirements associated with these rights. • Users may download and print one copy of any publication from the public portal for the purpose of private study or research. • You may not further distribute the material or use it for any profit-making activity or commercial gain
• You may freely distribute the URL identifying the publication in the public portal.
If the publication is distributed under the terms of Article 25fa of the Dutch Copyright Act, indicated by the “Taverne” license above, please follow below link for the End User Agreement:
www.tue.nl/taverne
Take down policy
If you believe that this document breaches copyright please contact us at:
openaccess@tue.nl
providing details and we will investigate your claim.
Rapport na. 615
Cursor control by mouse and keyboard
Beatrix Vereijken
by
mouse and keyboard
An experimental comparison
xxxx
Beatrix Vareijken
October 1987
by
mouse and keyboard
An experimental comparison
scriptie Psychologische Funktieleer
begeleiding: Prof. dr. A.J.W.M. Thomassen
Katholieke Universiteit
Nijmegen
Dr.
O.G.
Bouwhuis
Drs. J.M.M. Hanssen
Instituut voor Perceptie Onderzoek
Eindhoven
Eindhoven, oktober 1987
Beatrix Vareijken
Abstract Preface
1. INTRODUCTION
1.1. Hlstorical overview
1.2. The development of cursor control devices 1.3. Critica! evaluation of earlier experiments 1.4. The present experiment
2. METHOD 2.1. Subjects 2.2. Task 2.3. Material 2.4. Procedure 3. QUALITATIVE RESULTS
3.1. Results of the questionnaire
3.2. Correlation between quantitative results and subjectlve preferences 4. QUANTITATIVE RESULTS
4.1. Impravement of performance wlth practice In posltloning time 4.2. Overall speed
4.3. Effect of the variables on access time 4.4. Effect of the varlables on posltionlng time 4.5. Effect of the variables on typlng time
4.6. Significant Interactlens between the varlables 4.7. Errors 1 2 3 3 3 5 8 9 9 9 9 12 14 14 14 15 15 16 17 18 21 21 26
5.1. Reaction time analysis following Donders 27
5.2. Positioning time analysls 29
5.3. Comparison of devices 31
5.4. Access time versus typlng time 33
5.5. External validity 33
5.6. lmprovements to the cursor control devices 34 5.7. Suggestions for further research 35
6. CONCLUSION 37
7.REFERENCES. 38
Appendix A: Instructien 40
Appendix B: Questionnaire plus results 42 Appendix C: Overview of the analysls of varianee 45 Appendix D: Significant Interactlens between the variables on the time components 48 Appendix E: Calculatlon of the requlred number of keystrokes In the case of the
Abstract
In a simulation of ed~ing tasks three cursor control devices have been compared, i.e., Sun mouse, Sun arrow keys, and Teletex arrow keys. The speed and accuracy with which subjects can select targets on a screen (varying in size and location on the screen) and reptace them by P has been investigated. Results show that positioning times are shortest overall w~h the mouse, and longest tor the Sun arrow keys. Furthermore, the mouse has the lowest error rates, the Teletex the highest Results of a questionnaire show that the rnouse is the device preterred most by the subjects. Positioning times tor the rnouse can be accounted for by Fms· Law. In the case of the arrow keys, pos~ioning times are shown to be proportional to the number of keys-trokes required to reach the target.
PREF ACE:
I wish to thank all the people at IPO and especially from the
Cogcom-group for the support they gave me. A special word
of thanks is given to my supervisors, and to Bart Kramer, An
Olijslagers, and Karl Kusters.
1. INTRODUCTION
1.1. Historical overview
During the past forty years, computers went through a rapid development. Around 1946, the first generation of computers was designed, consisting of many thousands of electron tubes. This made these first computers very large (about 150m2) and heavy (about 30 tons). They had only little storage capacity and were very expensive. These were serieus disadvantages, which did not make the first computers suitable for widespread use. In the mid fifties the second genera-tion computers evolved. lnstead of using electron tubes. these computers consisted of transis-tors and PCBs (Printed Circuit Boards). This ledtoa smaller volume and less energy consump-tion. and to a much higher processing speed. With the introduetion in the sixties of !Cs (lntegrated Circuits) and the possibility of teleprocessing, the third generation computers arose. These computers we re rnuch better and cheaper. Consequently. they we re no Jonger only attractive for the large industries, but for small companies and private users as wel!. During all this time, the ma in focus was on development and improvement of the computer itseH. Develop-ment of software, more carefully designed user interfaces, and input-output devices was some-what neglected. But for the last ten years, both computer manufacturers and researchers have paid more attention to these aspects. This resulted, amongst other things, in a fast development of cursor control devices. The cursor indicates the place on a terminal screen where the next character wiJl be entered. The cursor is movedover the screen by a cursor control device. We will turn to the development of cursor control devices in the next section.
1.2. The development of cursor control devlces
At first, the user had at his disposal only four arrow keys on the keyboard. With these keys, the cursor could be moved on the screen in horizontal and vertical directions. Some keyboards became extended with text keys. These keys can move the cursor toanother character, but also toanother word, line, or paragraph. later, more devices were developed. For example, the joys-tick, a small strain gauge with a rubber bar mounted on it. Applying force to the bar in a certain direction causes the cursor to move in the similar direction. Another device is the tracker ball. Rotatien of the ball by the palm of the hand causes the cursor to move on the screen in the same direction as the hand rotating the tracker ball. The knee control is a simHar device. With this device, the user has to push a lever with his knee. A sideward motion of the knee moves· the cursor horizontally, while vertical cursor movement is controlled by an up-and-down motion of the knee.
Besides cursor movement, some cursor control devices were extended with a specific function. Examples of these are the lightpen and the mouse. A lightpen is a hand-held pen with a light-sensitive element which detects the emission from the phosphor of a screen display. The mouse is a hand-held device separate from the keyboard, that can be moved across a surface or a tablet. Both devices can be used to position the cursor on the desired location by making a direct (lightpen) or indirect (mouse) movement towards that location, and thus select the charac-ter or tigure of incharac-terest. So the user does not need to code the commands in
a
menu any Jonger, or to make multiple key-pressing (like the "crtl + character") to give commands. He can simply select the desired cernmand by directly pointing at it.Computer manufacturers, who, through the years, placed more and more emphasis on a smooth user interface, have increasingly supplied their customers with the mouse as a standard means of c:ommunication with the computer. They gradually extended its capacities beyond pointing and selection to include graphical and editing purposes, without it being clear, however, whether the mouse was indeed the best device for all those applications.
The widespread use of the mouse on the one hand, and the obscurities about its characteristics on the other hand, has, in the course of recent years, led to many experiments and reports that compare the "advanced mouse" with the "old fashioned arrow keys". We shall now take a closer look at characteristics of arrow keys and mouse, and compare the findings in the litera-ture about these two devices, starting with the Jatter.
The mouse
Several reports find the mouse the best overall cursor control device (English, Engelbart, & Ser-man, 1967; Card, English, & Burr, 1978; Card, Moran, & Newell, 1983). They regard the mouse as fast, compatible, and least prone to errors. Furthermore, they agree that the mouse has the highest learning speed. Other advantages of the mouse are that it can work in small places, because it can be lifted and repositioned; the control-display gain can be modified, i.e. the amount of cursor movement per unit of mouse movement can be changed; and the operator can usually locate and move the mouse while stilllooking at the screen (Salvendy, 1987).
Other studies do not find the mouse to be the best device (Engel & Granda, in Embley & Nagy, 1981; Karat, McDonald, & Anderson, 1984; Ewing, Mehrabanzad, Sheck, Ostroff, & Shneider-man, 1986). They say that the mouse is not the fastest device overall, and that it is most difficutt to use for novice users.
The mouse also seems to have mechanica! problems, most of which are summarized by Thomassen, Teulings, & Schomaker (1987) and Salvendy (1987). First of all, the mouse is remote from the keyboard, occupying additional space on the desk next to the terminal and the keyboard. Furthermore, only movements made directly on the desk or tablet are registered. Jumps of the mouse above the desk are merely neglected. Also, the mouse is sensitive to orien-tation: when the mouse is positioned on the desk obliquely, movements with the mouse on the desk in a straight line willlead to diagonal movements on the screen. Furthermore, the mouse typically has lower resolution capabilities than for instanee digitizers, and is not well suited for single character data entry. Finally, the mouse is operated in a working plane that deviates in position, orientation, and dimension from the screen where the effect is intended and effectuated.
We think that the mouse has also problems related to motor aspects of its use. First of all, in· order to grasp the mouse, one hand needs to leave the keyboard. This c:ould lead to interter-enee with tasks like editing. Furthermore, the mouse does not occupy a stationary position, but lies 'somewhere' on the desk. When moving the mouse, it can run off the desk. When this is the case, the mouse needs to be Jifted and repositioned on the desk. While doing this, the connee-tion wire can be caught by a corner of the deskor by an object on the desk, or the wire could occupy the space where the mouse should be placed. This could also lead to interference. Arrow keys
Arrow keys have several advantages. First of all, they are positioned on the keyboard, which has several c:onsequences: they have a stationary position; they do not need additional space on the desk; they are not sensitive to orientation; and the hand can stay on the keyboard in
order to operate them. Furthermore, they require only simpte motor actions (Chambers & Stock-bridge, 1970). So novice users learn to use them quickly, and need only short training periods (Embley et al., 1981). Besides, performance is stabie over a wide range of conditions: the slope of the keyboard, the force necessary tor key depression, the key displacement, and the type of kinesthetic feedback trom key actuation may be varied within a wide range without affecting per-formance (Aiden, Daniels, & Kanarick, 1972; Conrad, 1966). Several studies regard arrow keys as quick (e.g. Karat et al., 1984). Chapanis and Kinkade (1972) found arrow keys to be quick and taking up little space.
Chambers et al. (1970) wrote in their review on push-buttons that there exists a tendency to operate them with a ballistic-Bke movement resulting in increased speed, but reduced accuracy. Ewing atso found that arrow keys lead to more errors and are more visually demanding than the mouse (Ewing et al., 1986). Accuracy also decreasas with the removal of visual feedback (Seibel, 1972). Furthermore, there are some studies that find arrow keys not to be so quick. Engel et al. (in Embley et al., 1981) tor example, found that arrow keys are slow, in particular when the cursor is rnoved over larger distances. Another important disadvantage with arrow keys is that there is no standardization regarding the spatial position and the lay-out of the arrow keys (Embley et al., 1981; Seibel, 1972; Moore, 1974). They can be positioned directly beside the alpha-numeric keys, or be embedded in a cluster of function keys. They can be arranged clockwise, or three keys in a row with the fourth key above the second key. There can be tour arrow keys, one tor each orthogonal direction, or just two keys, in which case the remaining direcUons require the use of the shift key. Furthermore, awkward reaches and chorded key strokes are more difficult than other single strokes, teading to more errors and reduced speed (Seibel, 1972). And finally, like the mouse, the arrow keys are operated in a working plane that deviates from the screen.
Concruslons
We have to conetude that the opinions about arrow keys and mouse are not unequivocal, and often even contradictory. Since the mouse is continuously getting more functions and applica-tions, and is regarded by more and more people to be the standard, and thus the best means of communication, it is important to compare the qualities of the mouse once more with the arrow keys, especially in the domain of editing, to see if and when the mouse is better. We suspect that the mouse need not be the best overall device in this kind of application.
From a critica! analysis of forrner experiments, we wiJl discuss point by point the aspects which are important for a carefully designed experiment, and indicate and justify the conditions we adopted in the experiment we performed.
1.3. Critical evaluation of earlier experiments
Spatlal arrangement of the arrow keys
In an experiment, Goedwin (1975) compared pointing times (in this case, the time needed to position a cursor at specified locations on the screen) tor three devices: lightpen, lightgun, and arrow keys. Her choice of arrow keys, the Sanders 720 arrow keys, is peculiar for two reasons. First, their spatial arrangement does not correspond to the direction of the cursor movement, which may cause confusion in the subjects. Second, in some cases the keys had to be used simultaneously with the shift key, which reduces speed (Seibel, 1972). Not surprisingly, she found the arrow keys to be much slower than the ether two devices. In the present study, only
keyboards will be used whose arrow keys have a compatible arrangement, which in this context means arrow keys whose spatial arrangement corresponds to the direction of cursor movement Edltlng task
In the Goodwin (1975) experiment, the taskof the subjects was to move the cursor to the target, and reptace the target with an "x". However, an "x" can be entered with one hand. So, the experiment did not really simulate the altemating pointing and bimanual text entry typical of edit-ing. In our experiment, subjects had to type a capita! "P", which requires the use of both hands (shift key and P key).
Independent varlables
Most reported experiments do not vary one or more of the variables that are, in our opinion, important. In several experiments where pointing time was studied, the size of the target was not varied (Goodwin, 1975; Karat et al., 1984; Ewing et al., 1986). We expect that a large target is easier to select than a smaller one. There might even be an optimum target size, such that performance does not improve any more with larger target sizes. Furthermore, the distance from the initia! cursor position to the target was often held constant (Goodwin, 1975; Karat et al., 1984; English et al., 1967). And finally, only Card et al. (1978) varied the movement angle between initia! cursor position and the target. In the present study, all three variables will be con-trolled simultaneously: target size, target distance, and movement angle.
Dependent varlables
In 1967, English et al. compared the pointing times of five different devices: lightpen, Grafaeon tablet, joystick, mouse, and knee control. Since they were specifically interestad in editing, they divided the total reaction time in two parts:
1) the access time of the devices, which is the time necessary for the hand to Ie ave the key-board and to take hold of the pointing device, up to the start of the cursor rnovement, and 2) the motion time, beginning with the first cursor movement and ending with the select
act ion.
In the present study, the total reaction time will be divided into access time, positioning time (equivalent to motion time) and a third part: typing time, which will be measured from the moment that the selection button has been pressed until the "P" has been typed.
In the next section, we will first describe Card's experiment (Card et al., 1978). Because their study was so carefully designed, and bears much similarity to ours, we will do this in more detail than the experiments reported thus far. Subsequently, we will turn to the design of our own experiment.
Card's experiment
Card et al. used an experimental design, equivalent to that of Englishet al. (1967), to compare mouse, joystick, arrow keys and text keys with respect to text selection on a screen. They not only varied target size, but also distance to target and the angle of hand movement They meas-ured the speed of Jearning, the homing times (equivalent to English et al.'s access time). the positionirYJ times (equivalent to English et al.'s motion time), and the number of errors. They carefully eliminaled learnirYJ effects by training the four subjects with each device until position-ing time was no longer decreasposition-ing with practice. They found that practice caused more
impravement in the mouse and text keys than on the other devices. The mouse was overall the fastest device, the arrow keys were the slowest The positioning time for the mouse increased approximately with the log of the distance. For the arrow keys, the time increased rapidly as the distance increased (see Figure 1).
5 ... mouse ~ stepkeys 0+---~~--.-~--r--r--r-~ 0 4 8 12 16 distance (an)
Figure 1.1. Positioning time as a tunetion of distance, trom the study of Card et al. (1978}.
The time for the rnouse and the arrow keys decreased with the log of the target size (see Figure
2). 5 4 :§: ~ 3 ~ ~
6
2 E.,
8.
0 0....
2 :& + mouse-o- step keys
4 6 8
target size (characters)
10 12
Figure 1.2. Posltioning time as a tunetion of target size, trom the study of Card et al. (1978}.
The angle of hand movement caused no significant effect in the case of the mouse. The effects of direction with the other devices were not great: for the arrow keys, direction amounted to only 9% of the mean positioning time. The rnouse had the lowest overall error rate, viz., 5%; the arrow keys had the highest, viz., 13%. Card et al. then demonstrated that the relation between positioning time, target size, and target distance obeyed a version of Fitts' Law (Welford, 1968), which relates positioning time (T) to distance to target (0), target size (S), and two constants (K0 • K), as fellows:
T
p=
K
0+
K .
2I
o
g
(DIS
+
0. 5)
s.
( 1 -1)
values of an optima! device with respect to the information processing capacities of the eye-hand guidance system. In the extreme case of one character targets 16 cm. distant, positioning times for such an optima! device, the mouse, and the step keys are about 1.32 s., 1.41 s., and 3.15 s. respectively.
1.4. The present experiment
The present study differs from the Card et al. experiment in several aspects. First, a different set of devices will be measured: the Sun mouse, the Sun keyboard, which has a traditional layout of the arrow keys, and the new Philips Teletex keyboard, which has larger arrow keys in a circular arrangement, and the possibility of diagonal cursor movement See Figures 2.3 and 2.4.
Second, a third type of time maasurement wiJl be made beside access and positioning time: viz., typing time. Thus, the total time needed to complete a trial will be divided into three separate measurements: access time, positioning time, and typing time. We included this third time com-ponent to see whether it varies in the same manner as access time when conditions are changed. lf this would turn out not to be the case, it would be an indication of extra interterenee with the data entry part of the task.
Furthermore, intheir experiment Card et al. used wordsas targets. But since we vary the target size in our experiment, we would need different words for different target sizes. To exclude any linguistic and visual effects, we decided to use sequences of xs as targets.
After the subjects have completed the experiment with all three devices, they wiJl be given a questionnaire probing the subjects' device preferences. Although several studies suggest that operator preferences do not correlate significantly with measured performances (Aiden et al., 1972; Ewing et al., 1986; Kinkead, 1975), we included this small additional effort in our experi-ment to investigate the relationship between subjective preferenee and measured performance. Finally, we will make an attempt to fit the data in a theoretica! model, which wiJl allow us to predict performance in general terms.
2. METHOC
2.1. Subjects
Twelve subjects performed the experiment, six male and six female. Typing experience was a requirement, because we wanted to simulate a real life editing task. Since we wanted to investi-gate learning effects, we selected subjects who were novices on all three devices. All subjects were righthanded.
2.2. Task
Before the beginning of each trial, the subjeet's left hand was above the shift key, and his right hand above the space bar. As soon as he detected the target, he had to select it, and replace it by a capita! p (P). He had to do this in the following way: first strike the space bar with his right hand, then reach for (access) the pointing device with the same hand, and position the cursor on the target. Thus having positioned the cursor, he had to select the target by pressing the left select key and replace it with a P by simultaneously pressing the shift key with the left hand, and the P key with the right hand. After one second, the next target appeared.
2.3. Material
The workstatlonIn our experiment, we used the Sun 3/50 workstation, which runs an enhanced version of the 4.2 BSD UNIX system (see Figure 2.1).
Figure 2.1. Sun workstation.
Sun workstations are especially oriented towards applications in engineering, CAD (Computer Aided Design), CAM (Computer Aided Manufacturing), graphics, and so on. This expressas itself in a large 36 x 28 cm., bit-mappad monochrome display having 1152 by 900 pixels. The display screen has its own memory - the frame buffer - to provide high-speed data transfer to the screen. In our experiment, the default value was black characters on a white background. Polntlng devlces
Three pointing devices were tested. One continuous device (the Sun r12ouse) and two key operated devices (the Sun keyboard and the Philips Teletex keyboard).
Mouse
In this experiment, we used the Sun mouse for practical reasons: we wanted to use the advanced Sun wol1<station, which had connected to it the optica! Sun mouse. The dimensions of the mouse are 6.6 centimeters (cm)(horizontal) by 9.7 cm. (vertical). lt has three buttons; the left-hand and the central buttons were in this experiment equally defined as selection buttons. The right-hand button kept its original tunetion (i.e., menu selection). When the experiment was running, it could be aborted by popping up a menu with the righthand button. The mouse lay on a tablet to the right of the keyboard, connected to it by a wire. The dimensions of the tablet are 27.8 cm. (horizontal) by 22.8 cm. (vertical).
Figure 2.2. Mouse.
When the mouse is moved across this tablet, two intra-red lights sense the (x,y) motion of the mouse. When the mouse moves, a cursor moves simultaneously on the screen, two units of screen movement tor one unit of mouse movement
Arrow keys on the Sun keyboard
The arrow keys are the four keys found on many CAT terminals. See Figure 2.3.
DO OOI
11
11
11
11
IDDD DDD
DODDDDDDDDDDDDDDDDDD
DO DDDDDDDDDOEJDDD D.DD
DO DDDDDDDDDDDDIReturn
I
D~D
DO
IShiftIDDDDDDDDDDDD
~(:]~
OO[spacebar
JDD
D~D
Figure 2.3. Sun keyboard.
They move the cursor in each of four directions, as indicated by the arrows. Pressing one of the keys with a horizontal arrow moves the cursor 0.25 cm. (i.e., one character) to the right or to the left. Pressing a key with a vertical arrow moves the cursor 0.5 cm. (i.e., one character) up or down. Holdi!"Q down one of the keys tor more than 0.061 seconds (s) causes it to enter a
repeating mode: 5.9 centimeters per second (cmls) vertical movement and 4.1 cmls horizontal move ment. The central key in between the arrow keys (labelecl R11) was defined as the selec-tion key for two reasons. Just like in the case of the mouse, it is close to the cursor control keys, and it must also be operated with the right hand.
Arrow keys on the Phlllps Teletex keyboard
The Teletex keyboard was for the purpose of the present experiment, connected to the Sun 3/50. The arrow keys on this keyboard work on the same principle as the Sun arrow keys. Pressing down one of the keys with a horizontal arrow moves the cursor 0.25 cm. to the right or to the left. Pressing a key with a vertical arrow moves the cursor 0.5 cm. up or down. But the Teletex arrow keys have two distinct features. First, they are much larger, have different shapes, and have a circular arrangement. Second, pressing down one key with a vertical and one key with a horizontal arrow simultaneously moves the cursor 0.6 cm. in a diagonal direction with an angle of 60 degrees. Holding down one or two of the keys for more than 0.25 s. causes it to enter a repeating mode: 7.1 cmls vertical movement and 3.8 cmls horizontal movement For pressing two adjacent keys simultaneously this speed is 8.0 cmls. The central key in between the arrow keys was, in analogy to the Sun keyboard, defined as the selection key. See Figure
2.4.
0 DDDDDDOOOODODD
D
DOOD
0 DDDDDDDDDD@JDDD
D
0 DDODDDDDDDDDDLJ 0 •
Á0
Dl
ID
0
Figure 2.4. Teletex keyboard.
Targets
The targets varied in three aspects, viz., in horizontal size, in distance to initia! cursor position, and in location on the screen. There were four target sizes: 1, 2, 4, and 10 characters, i.e., sequences of xs. The distance between initial cursor position and target was defined as the dis-tanee between the lower left corner of the cursor and the nearest point of the target. See Figure 2.5. There were five different target distances: 1, 2, 4, 8, and 12 cm. The location on the screen was defined by the angle between a horizontal line through the cursor in its initial posi-tion, and a straight line through the cursor (in it's original position) and the target. See Figure 2.5. There were eight different angles, viz., 0, 45, 90, 135, 180, 225, 270, and 315 degrees. In sum, this resulted in a total of 4 x 5 x 8
=
160 unique, ramc:lomly ordered stimuli./ / / ··~~···~ / / / / //distance / / / / / / /II.JCAa-.IIUUI..1L" / / / ) a n g l e
Y---Figure 2.5. Screen during a trial with definition of distance and angle.
2.4. Procedure
The subject was seated in front of a Sun display screen for output, a Sun or a Philips keyboard for input, and one of the three devices for pointing at targets on the screen. The distance between a subjeet's eyes and the screen was approximately 55 cm. The intensity of ambient lights was 1 OB Lux. Reflections on the screen from ambient lights we re avoided. On each trial, the cursor was displayed at the center of the screen. We used a black vertical rectangle as the cursor: black, because of the white background of the screen, and rectangular, because that is the shape of the characters (i.e., non-capita! x) we used for targets. Following the cursor's appearance, the target was displayed on the screen. lts position was random within the
con-straints mentioned above. ·
Each subject repeated the experiment with each device in a complete counterbalanced design. On each device, the subjects accomplished 40 warm-up trials plus 6 blocks of 160 test trials, each block lasting approximately 16 minutes. The order in which the stimuli appeared was ran-domized, and different for every block. Across conditions and subjects, the order of stimuli in the 6 blocks was identical. At the end of each blockof 160 trials, therewas a break of 25 seconds, in which the subject was given feedback on the average number of errors for that block. The three sessions were for each subject divided over three separate days. In each of the sessions, they worked with one of the devices. They were paid f7.50 per hour fortheir participation in the experiment.
The subject is given feedback in order to maintain his motivation (Card et al., 1978; Poulton, 1974). The emphasis in the instructien is on accuracy. Therefore, we only give the subject
feedback on his number of errors, not on his average positioning time. A trial is recorded as an error when the cursor is not accuratety positioned on the target (in the present experiment, the cursor must cover at least 1/4 of the target, i.e., an x, to be correct), and when the subject does not follow exactty the instructions, for exampte, hitting the space bar with his left instead of his right hand, pressing a key not firmty enough so that the computer does not register any input, etc. We coutd in principle atso give the subject feedback on every trial, for example by inverting the colours of target and background when the target is setected correctty. But in this way, sub-jects might be tempted to try to correct their errors, teading to additionat stages (beside access, positioning, and typing stages), which would interfere with the task. Therefore, we only gave the subjects feedback at the end of trial btocks. There are saverat raasons for breaking down the total numbers of trials into blocks of 160 trials. Since there are 160 unique stimuli, which all appear exactly once in a block, the blocks are comptetety comparabte. This in contrast with Card et al. (1978), who had 200 unique stimuli, but blocks of 20 trials. Furthermore, we did not want to interrupt the subjects too often, because that may interfere with their task.
Measurements
All errors were recorded, and classified as being pointing errors or instruction errors. In the case of an instruction error, the experimenter recorded the exact nature of the error.
Time measurements were made in three separate parts. Access time was recorded from the moment the space bar was struck until the first movement of the cursor. Positioning time was recorded from the first cursor movement until the select key was pressed. Typing time was recorded from the moment the select key was pressed until the p-key was pressed. See Figure
2.6 for a summary of these time components.
lXXX
I
Pressing of the spacebar hand movement te deviceI
L!
First cursor movement I lXX hand movement to target•·
Pressing of the select key movement to alphanum. keys·I·
Pressing of shift endP
~acce"
llme-+P••Itlonlng l l m • + typlngtime~
Figure 2.6. Summary of a subjects actions during a trial, and
3. QUALilAliVE RESUL TS
3.1. Results of the questionnaire
At the end of each session, subjects were given a questionnaire probing the subjects' opinion about the task and the instruction. After the third session, they were also asked to write down the advantages and disadvantages they experienced with each device, and to give their device preferences. The resutts of the questionnaire are summarized below. See Appendix A for the questionnaire and the subjects' responses.
In summary, most subjects find the instructions easy to follow. In genera!, they find the task easy, but boring and tiring. The mouse is regarded as the tastest device, but also as a device difficutt to use: it is difficult to position the cursor exactly on the target with the mouse, and sometimes it must be lifted and repositioned. Both keyboards are judged to be to slow in gen-era!, with the Sun keyboard being regarded as being too fast with smal! movement distances. Several subjects say that they can wor1< very accurately with the Sun arrow keys. But they also find it a disadvantage that they could not move the cursor in a diagonal direction. The possibility of diagonal cursor movement with the Teletex keyboard is seen as an advantage by most sub-jects. The mouse is by far the most preferred device, the Sun arrow keys the least preferred. atthough the subjective preferenee differences between the two keying devicesis only smal!.
3.2. Correlation between quantitative results and subjeelive preferences
Table 3.1 gives the combined figures of quantitative results and subjective preferences.Table 3.1. Subjective preferences and quantitative results.
~
Mouse Teletex SunMouse 10 1 1 Teletex 2
6
4 Sun 0 5 7 Total 12 12 12 Total 12 12 1236
When there is no relation at all with quantitative results, the value of 4 is expected in each cell. A goodness-of-fit test between observed and expected values is based on the quantity
x 2
=
1:
(OI
~i
Et)2
(3-1)where
X2
=
a value of the random variabie X2 whose sampling distribution is approximated very closely by the chi-square distribution.Ei .. the expected value for the ith cell.
o, ..
the observed value for the ith cell.Calculation of chi-square shows that there is a significant correlation between subjective prefer-ences and objective performances (
x
2 .. 22; df • 4; p < 0.001). This is in contrast with what is4. QUANTITATIVE RESULTS
In the present chapter, the results of the experiment will bedescribed in separate sections. Sec-tien 4.1 gives the impravement of performance with practice on the three devices. SecSec-tien 4.2 presents the overall speed of the three devices, and gives an overview of the access times,
positioni~ times. and typing times. In sections 4.3 to 4.6 the main effects of the three variables - distance to target, target size, and approach angte- on access time, positioning time, and typ-ing time are described. The significant interactions between the variables are discussed per dev-ice in sectien 4.7. Insection 4.8 the error scores are given. Unless mentioned otherwise, all indi-cated differences are statistically significant at the p < 0.01 level. See Appendix B for an over-view of the results of the analysis of variance.
4.1. Impravement of performance with practice in positioning time
We haved looked at the practice effect on positioning time in order to make a decision about the data to include in the finat analysis. As can be seen in Figure 4.1, performance improves most with practice in the case of the mouse. The mouse also has the largest overall improvement.
5 -o- aun ... teletex ~ 4
"'*"
mouse ~ 3~
A 11 11 ';= ~6
2~
i
..
..
..
..
0 0 2 3 4 5 6 7 block numberFigure 4.1. Leaming curves for the pointing devices.
Each point on the graph is the average of a block N of 160 contiguous trials from which errors have been excluded x 12 subjects. Since every block c:ontains the 160 unique stimuli, they are completely c:omparable. As was shown by De Jong (1957), the improvement rate of positioning time as a function of the amount of practice can be approximated by a power tunetion
where
Tn .. estimated positioning time on the Nth block of trials N .. trial block number
a ..
an empirically derived constantTaking the log of both sides produces a linear equation
log
Tn=
log
11-
a.log N
(4-2)The parameters T1 and a as determined by regression analysis are given in Table 4.1, along with the standard error (se) and the squared multiple correlation (R2).
Table 4.1. Leaming curve parameters.
Device T1 (S) Cl Equation S8(S)
R2
Mouse 2.12 0.18 TN • 2.12 N·0 ·18 0.014 0.95 Sun 3.44 0.08 TN • 3.44 N-o.os 0.012 0.82 Teletex 3.29 0.07 TN • 3.29 N-0 -07 0.003 0.98
The high percentage of varianee explained by equation (4-2) is partly due to the low number of data points, i.e. the number of bleeks.
Table 4.1 shows that the learning parameter a with the mouse is more than twice as high as with the keyboards. Furthermore, even on block number 1, when there has been no practice yet, the mouse is considerally faster than the keyboards. The Teletex keyboard is atways a little tas-ter than the Sun keyboard.
Subsequently, two-tailed t-tests were performed. For the mouse and the Sun arrow keys, bleeks 4, 5, and 6 are not significantly different amongst themselves (p < 0.05). For the Teletex arrow keys, bleeks 3, 4, 5, and 6 are not significantly different (p < 0.01 ). So, unless indicated other-wise, the remaining analysis wilt be based on the data of the last three bleeks, excluding those trials on which errors occurred.
4.2. Overall speed
Table 4.2 gives the access times, positioning times, typing times and totaltimes for the devices, averaged over all the distances, target sizes, and movement angles.
Table 4.2. Overall times.
Movement time for non-error trials.
Device Access time Positioning time Typing time Total time M SD M SD M SD M
Mouse 0.70 0.19 1.59 0.61 0.77 0.29 3.05 Sun 0.73 0.25 3.15 1.57 0.63 0.25 4.52 Teletex 0.75 0.24 3.00 2.06 0.67 0.28 4.42
All the ditterences within the three time components among all device pairs are significant (p <
0.01, using t-tests). The ditterences between the access times and between the typing times of the three devices are relatively small compared to the larger differences between the positioning times. The Teletex has the slowest access, the mouse the fastest, even though in the latter case, the distance to betraveled is largest.These findings are in contrast with these by Card et al., who found the mouse to have a relatively slow access. See Table 4.3. Furthermore.the values Card found for the access times are much shorter than the access times found in the present experiment. In total, the mouse is by far the fastest device, the Sun arrow keys the slowest, at only a small distance from the Teletex arrow keys. See also Figure 4.2 for a diagram
Table 4.3. Overall times, from the study of Card
et
al. (1978). DeviceMouse Step keys of the time c:omponents.
5 4 0 Access time
M
SD
0.36 0.21 mouse 0.13 0.30 sun device Positioning timeM
SO
1.29 2.31 0.42 1.52 Iele te x Total time M 1.66 2.51 • acx:ess time 11 positioning timeEl
typing timeFigure 4.2. Diagram for the overall times.
4.3. Effect of the variables on access time
We shall now look at the effect of the variables on access time, i.e., the time necessary for the hand to leave the keyboard and to take hold of the pointing device. In the experiment this is defined as the time between hitting the space-bar and the first movement of the cursor.
Effect of dlstance
As can be seen in Figure 4.3, access time initially decreasas with increasing distance, but when distance gets larger, access time starts to increase again. Analysis of varianee shows that the. effect of distance is significant within every device at the p < 0.01 level. lt appears that espe-cially at small and large distances subjects need more time before the first cursor movement is made. But the differences, though significant, are relatively small.
Effect of target slze
This effect is significant at the p < 0.01 level only for the Teletex arrow keys. Access time decreasas slightly with increasing target size, with access time being relatively long at a target size of 1 character. But again, the ditterences are relatively small. See Figure 4.4.
1,0 0,8
;
0,6 - teletex ~ sun 0,4 ,.. mouse 0,2 0,0 +-....-,.-..--,--.--.--...,_".-"-T"-...---r---, 0 2 4 6 8 10 12 14 distance (cm)Figure 4.3. Access time as a tunetion of distance.
1.0 0.8
--.
•
IC ~ 0.6I
.,
.,
...
lala te x 0.4§
al 0.2 0.0 0 2 4 6 8 10 12 size (charaders)Figure 4.4. Access time as a tunetion of target size. Effect of approach angle
The effect of approach angle on access time is given in Figure 4.5. At all diagonal angles, access is tastest tor the mouse, and slowest for the Teletex. Within both keyboards, access is
tastest at horizontal angles - 0 and 180 degrees. This is in contrast with the effects of approach angle on positioning time, where times are shorter with targets at verticat angles. See section. 4.4.
4.4. Effect of the variables on positioning time
We shall now look at the effect of the variables on positioning time, i.e., the time between the first cursor movement and the selection of the target.
Effect of dlstance
Figure 4.6 shows the effect of target distance on positioning time. At all distances, positioning time is shortest with the mouse, and its advantage increases with increasing distance. The positioning time for the mouse increases approximately with the log of the distance. The time for the arrow keys increases rapidly, approximately linearly, as the distance increases. At a distance
1,0 0,8 0,0 • mouse 0 90 180 270 45 135 225 315 angle (degrees) • -.letex 0 90 180 270 45 135 225 315 angle (degrees)
EJ
sun o eo 180270 •5 135225315 angle (degrees)Figure 4.5. Access time
as
a tunetion of approach angle.5
..
~ ~ sun~
3....
teletex ~ ~...
mouse ~ 2...
~ &8.
0 0 2..
6 8 10 12,..
distanoe (an)Figure
4.6.
Positioning time as a tunetion of distance.of 1 cm. the Sun arrow keys are significantly slower than the Teletex arrow keys. This is prob-ably due to the speed with which the Sun arrow keys enter the repeating mode, viz., 0.061 s. The resulting overshoot makes the cursor much harder to position over smal! distances with the Sun arrow keys than with the Teletex arrow keys. At distances larger than 4 cm. the Teletex is significantly faster than the Sun arrow keys, probably owing to the possibility of diagonal cursor. movement This advantage increases over larger distances.
Effect of target slze
The effect of target size on positioning time is given in Figure 4.7. From size 1 to 4, positioning time rapidly decreasas for all three devices. This decrease is largest for the Sun arrow keys, again owing to the relative difficulty of moving the cursor over very smal! distances. For targets bigger than 4 characters, positioning time hardly decreasas for all three devices. Apparently, it is only important whether the target is 1, 2, or more characters long. Exactly how much more seems to be irrelevant. At all target sizes, the mouse is significantly faster than the keyboards. These ditterences are all significant at level p < 0.01 using t-tests.
-o- sun
"*
teletex ... mouse li~----,·~---..
0+-~-.--~---,--~T-~~~~ 0 2 4 6 8 10 12target slze (c:haractars)
Figure 4.7. Positioning time as a tunetion of target size.
Effect of approach angle
Figure 4.8 shows the effect of approach angle on positioning time. At every angle, the devices ditter significantly (p < 0.01 using t-tests) from one another.
5
• mouse • teletex
El
sunJ
3 Dl ~6
2 ~&.
1 0 0 90 180 270 45 135 225 315 angle (degrees) 0 80180270 45135225315 angle (degrees) 0 90180270 45135225315 angle ( degrees)Figure 4.8. Positioning time as a tunetion of approach angle.
Also, analysis of varianee shows the angle makes a significant difference (p < 0.01) within every device. This is in contrast with Card et al. who found no significant effect for the mouse. The. differences for the mouse found in the present study, however, are small. At angles of 0, 45, and 90 degrees, it takes slightly shorter to position the cursor. According to Salvendy (1987) and McCormick (1970) controlled arm movements that are primarily a pivoting of the elbow, with fairly nomina! upper arm and shoulder action, tend to take less time than those with a larger degree of upper arm and shoulder action. Minima in positioning time would thus lie between 0 and 110 degrees, and between 180 and 290 degrees, and maxima at 145 and 325 degrees. This is confirmed for the angles of 0, 45, and 90 degrees, where positioning times are shortest But at 225 degrees the longest positioning time is found. Wrth the two keyboards, there is a significant difference between positioning times at orthogonal angles - 0, 90, 180, and 270 degrees - and at diagonal angles - 45, 135, 225, and 315 degrees -, the latter having much longer positioning times. Furthermore, the positioning times at vertical angles - 90 and 270 degrees - are shorter than at horizontal angles - 0 and 180 degrees. In this experiment the
cursor moved faster in vertical than in horizontal direction. The speed in vertical direction with the Sun arrow keys is 5.9 crnls in a repeating mode, and in horizontal direction 4.1 crn!s in a repeating mode. The speed with the Teletex arrow keys in vertical direction is 7.1 cm/s in a repeating mode, versus 3.8 crnls in a repeating mode in horizontal direction. Furthermore, the speed of cursor movement for the Sun in horizontal direction in a repeating mode is faster -viz~.
4.1 crn!s - than for the Teletex - viz., 3.8 crnls. lndeed, the horizontal angles - 0 and 180 degrees- are the only angles in this experiment where the Sun is taster than the Teletex.
4.5. Effect of the variables on typlng time
The most marked feature is that the mouse has longer typing time than the keyboards. See Table 4.2 and Figure 4.2. For none of the devices are the effects of target distance and target size significant at p < 0.01. Figure 4.9 shows the effect of approach angle on typing time. This effect makes a significant ditterenee only for the keyboards (p < 0.01 ).
1.0
•
Ielelexm
sun 0,8 §: 0,6 ~ ';:: ~ 0,4l
0,2 0,0 0 80180270 451352253150 0 80180270 45135225315angle (degrees) angle (degrees)
Figure
4.9.
Typing time as a tunetion of approach angle.With the Sun arrow keys, typing time at angles of 90, 270, and 315 degrees is a little shorter. With the Teletex arrow keys typing time is longest at angles of 180 and 225 degrees. But the differences induced by direction are relativety smalt.
In sum, the ditterences induced by the variables on typing time are smalt, and only the effect of approach angle is significant at p < 0.01. There is however, a marked difference with access time. The mouse always has the tastest access, the Teletex always the slowest This in contrast with typing time, which is always longest with the rnouse, and shortest with the Sun.
In the next section, significant interactions on the time cornponents between the variables wiJl be discussed per device.
4.6. Significant interactions between the variables
Mouse: significant lnteracttons between dlstance to target x approach angle
A significant interaction on positioning time between these variables was found with the mouse (F= 2.82; df= 28, 308; p < 0.01). At-test shows that at the distances 1 and 4 cm., there are no significant differences. At these distances,the positioning tirnes for the eight angles converge. The reach of wrist movements is approximately 4 cm. So, for bigger rnovements the elbow and shoulder will have to move as well. This could be the reason for the shorter positioning times at
distances up to 4 cm. At a distance of 12 cm. the positioning time is significantly shorter at an angle of 0 degrees (p < 0.01) than at all other angles. See Appendix 0, Table 0.1.
The interaction between distance to target and approach angle on access time is also significant (F= 2.27; df= 28, 308; p < 0.01). Using at-test the most striking feature is the rapid increase in access time at distances greater than 8 cm. at angles of 90 and 180 degrees, and a long access time at an angle of 270 degrees at distances larger than 4 cm. Apparently, subjects need most time before the first cursor movement with targets at orthogonal angles (except for 0 degrees) and larger distances. See Appendix 0, Table 0.2.
Mouse: significant lnteractlons between dlstance to target x target slze
There is a significant interaction between distance to target and target size on access time (F=
3.77; df= 12, 132; p < 0.01). An inspeetion of the data revealed no large ditterences (see Appen-dix 0, Table 0.3), except fora rapid increase in access time for 1 character targets at distances larger than 8 cm.
Mouse: significant Interactlens between approach angte x target size
This interaction on access time is also significant (F= 2.88; df= 21, 231; p < 0.01). In general, access time decreases with increasing target size, but there are a few exceptions. The access times with angles of 0 and 135 degrees are Jonger with target size 2 than with a 1 character tar-get. The times with angles of 45, 135, 180, and 270 degrees increase with increasing target size. See Appendix 0, Table 0.4.
Sun: significant Interactlens between dlstance to target x approach angle
There is a significant interaction between these variables on positioning time (F= 12.07; df = 28, 308; p < 0.01). After performing t-tests, it appears that the eight approach angles fall apart in three separate groups: the diagonal angles - 45, 135, 225, and 315 degrees -, the horizontal angles- 0 and 180 degrees- and the vertical angles-90 and 270 degrees. Figure 4.10 shows positioning time with the three groups as a tunetion of distance to target.
6 -+ diagonal -o- horizontal
"*
vertical o+-~~--~~~~~~~-r~~ 0 2 4 6 8 10 12 14 distanoe (cm)Figure
4.1
0. Positioning time as a tunetion of distance and approach angle {Sun).At distances larger than 2 cm., the times with diagonal angles gradually increase. But at a dis-tanee of 1 cm., positioning time is relatively longer. This probably reflects the rapid overshoot of the Sun arrow keys. A problem the subjects face twice with targets at a diagonal angle. This
makes those targets extra hard to select at small distances. At distances of 1 and 2 cm. the positioning times with the horizontal angles are shorter than with the vertical angles, but at larger distances this difference inverts. Apparently, cursor positioning over small di stances is short er in horizontal than in vertical directions. At larger distances, especially at 12 cm., the difference between the cursor movement speed in horizontal and vertical direction becomes more impor-tant. Not only does a single keystroke move the cursor twice as far vertically than horizontally, but also in a repeating mode the movement speed in vertical direction is almost 44% faster than in horizontal direction. So, times with vertical angles are shorter than with horizontal angles with increasing distance. There is one significant difference between the angles of 90 and 270
degrees, viz., at a distance of 12 cm (t= 4.35; df= 286; p < 0.01 ), where positioning time at an angle of 270 degrees is shorter than at an angle of 90 degrees.
There is also a significant interaction between target distance and approach angle on access time (F= 2.08; df= 28, 308; p < 0.01). Generally, access time decreasas with increasing dis-tance. See Appendix 0, Table 0.5. At distances larger than 8 cm, the access time for the diago-nat angles gradually increases. For the orthegonaf angles, this increase is rapidly. See also Fig-ure 4.3 and text above.
Sun: slgnHicant lnteractlons between target dlstance x target slze
This interaction on positioning time is significant as well (F= 2.62; df= 12, 132; p < 0.01). Using t-tests, there appears to be hardly any ditterenee between target sizes of 4 and 10 characters. So, in Figure 4.11, positioning timesforthese two target sizes have been averaged.
6 5 -o- size 1
"*
size2 ... size4+10 0+-~~--r-~~-,--~~-r--~ 0 2 4 6 8 10 12 14 distanc:e (an)Figure 4.11. Positioning time as a function of distance and target size (Sun).
In general, the times with target sizes of 1 and 2 characters are significantly Jonger than with larger target sizes. A target of 2 characters seems to be relatively harder to select at a distance of 1 cm. than a target of 1 character. lf this is related to the rapid overshoot of the Sun arrow keys, it is unclear why rapid overshoot has not the same effect on a 1 character target.
Sun: significant lnteractlons between approach angle x target slze
This interaction on positioning time is significant at the p < 0.01 level (F= 12.14; df= 21, 231 ).
After performing t-tests, the eight angles cluster into the same three groups as above: horizon-tal, vertical, and diagonal angles. See Figure 4.12.
1t
appears that target size has no significant effect when the target is approached vertically. This is to be expected, since with a target at a8
...
diagonal 5...
vertical ~ -o- horizontal ~..
~.r
3~
6
tl!
i
2 0+----r·~-T--~--~~~----~ 0 2 4 6 8 10 12 target si ze ( c:haracters)Figure 4.12. Positioning time as a tunetion of approach angle and target size (Sun).
vertical angle the cursor always crosses the target in the height. The height of the target is always 1 character, no matter how many characters the target length is. This contrasts with hor-izontal and diagonal angles, which do profit from larger target sizes. For in those cases, the cur-sor crosses the target in it's fulllength, which makes it easier to select the target correctly. This can beseen in Figure 4.13.
x x)
•
I
I
I
---~~
I
I
I
I
I
xxxx~---1
---l
:select areaFigure 4.13. The advantage of larger target sizes tor targets at horizontal and vertical angles.
The times for both horizontal and diagonal angles decrease rapidly with increasing target size. Only at small target sizes are the times for vertical angles shorter than for horizontal angles, probably due to the higher speed of vertical cursor movement
There is also a significant interaction between approach angle and target size on access time (F= 2.31; df= 21, 231; p < 0.01). The times for horizontal angles are shortest The times with angles of 0 and 225 degrees, which have a relativa maximum at target size 4, decrease with target size 10. The times for the other angles, except for 90 degrees, increase when target size gets larger than 4 characters. See Appendix D, Table 0.6.
Teletex: significant lnteractlons between target dlstance x approach angle
This interaction on positioning time is significant at the p < 0.01 level (F= 12.85; df= 28, 308).
Using t-tests, the angles fall apart in three groups here as wen. See Figure 4.14. The times for the horizontal angles increase the most rapidly with increasing distance, for the vertical angles the slowest
e
5 -+ diagonal -o- horizontal ... vertical 0+-~~--~~~~~~~----~ 0 2 4 6 8 10 12 14 distanoe (an)Figure 4.14. Positioning time as a tunetion of distance and approach angle (Teletex}.
Here again, the vertical angles have an advantage over the horizontal angles, since the cursor speed is higher in vertical direction - viz. 7.1 cmls - than in horizontal direction - viz. 3.8 cm/s. The interaction between these variables is also significant on access time (F= 2.97; df= 28, 308;
p < 0.01). At distances larger than 8 cm. the access times for the angles of 0, 90, and 180
degrees increase rapidly. As was the case with the Sun, both very small and very large move-mentsneed the most time befere the start of the cursor movement See Appendix 0, Table 0.7. There is also a significant interaction between distance to target and approach angle on typing time (F= 1.87; df= 28, 308; p < 0.01). The typing times for the angles of 180 and 225 degrees rapidly increase when the distance gets larger, for the angles of 0, 45, and 90 degrees when distance gets larger than 8 cm. But these differences are, though significant, relatively smalt. See Appendix 0, Table 0.8.
Teletex: significant lnteractlons between approach angle x target slze
The effect of target size on positioning time for the three groups of angles is given in Figure
4.15. This interaction is significant at p < 0.01 (F= 6.83; df= 21, 231 ). The most marked feature here is the small increase in positioning time with increasing target size for the vertical angles, while for the ether angles positioning time decreasas with increasing target size. As was the case with this interaction with the Sun, the horizontal and diagonal angles profit trom larger tar-get sizes, but the vertical angles do not (see Figure 4.13).
There is also a significant interaction between approach angle and target size on access time ((F= 2.58; df= 21, 231; p < 0.01 ). At target sizes larger than 4 characters the horizontal and diagonal angles slightly increase, while the vertical angles slightly decrease. But these differences are, though significant, relatively small. See Appendix 0, Table 0.9.
4.7. Errors
e
...
diagonal 5 oO horizontal ~...
Y811icalI
4...,.
..
..
~ 3 G...o...6
~ D•
..
..
"
i
2 0 0 2 4 6 8 10 12target size (charactars)
Figure
4.15.
Positioning time as a tunetion of approach angle and target size (Teletex).Across all 6 trialblocks, the mouse has the lowest overall error rate, 1.65%. The overall error rate of the Sun arrow keys is 2.06%, and of the Teletex arrow keys 2.46%. See Table 4.4.
Table
4.4.
Errors.Device Overall error Pointing lnstruction ra te errors errors Mouse 1.65% 0.84% 0.81% Sun 2.06% 1.13% 0.93% Teletex 2.45% 1.04% 1.41%
These rates are much Jower than the rates that appear in the literature. Card et al. for example, found overall error rates of 5% and 13% for the mouse and step keys respectively. The low rates found in this experiment are probably due to the strong emphasis in the instructien to the subjects on accurate rather than tast cursor positioning. This explanation is in agreement with Daniels and Graf (1970, in Stewart,1973) who found instructions to be important. They claim that the specific wording of the instructions influence the results of the experiment. The error scores can be divided in two groups: pointing errors - these occur when the cursor is not accu-rately positioned on the target - and instructien errors - these occur when the subjects do not. follow exactly the instructions. With the Teletex, most of the errors are instructien errors. A closer look at the data reveals that of the 1.41% instructien errors of the Teletex keyboard, 45% is caused by not pressing the p-key firmly enough. The percentages of this kind of error for the mouse and the Sun keyboard are 37% and 25% respectively. Since the error rates and the ditterences betweenthem are so small, no further error analyses have been performed.
5. DISCUSSION
The first two sections of this chapter deal with two different kinds of analyses that have been performed on the positioning time data. In sectien 5.1 an attempt will be made to apply Donders' subtraction method to the Sun data, in particular to the positioning times found with targets at diagenat angles. Sectien 5.2 will describe positioning time analyses for the three devices in termsof Fitts' Law (in the case of the mouse) and in termsof the required number of keystrokes (in the case of the keyboards). In sectien 5.3 a comparison between the devices will be given. Sectien 5.4 wilt contrast access time with typing time. Extemal validity will be discussed in sec-tion 5.5. And finally, improvements to the cursor control devices and suggessec-tions for further research wilt be proposed in sections 5.6 and 5.7 respectively.
5.1. Reaction time analysls following Donders
Since the Sun arrow keys can only move the cursor in horizontal or vertical directions, targets in diagonal directions must be reached with a combination of these two orthogenat directions. The question arose, whether it is possible to add two positioning times for targets at orthogonal angles to form the positioning time for a target at a diagonal angle. This line of reasoning resem-bles the early work of Donders (1869, in W.G. Koster, 1969), based on the idea that the time between stimulus and response is occupied by a series of successive processes, or stages, with each component process beginning only when the preceding one has ended. He assumed that the durations of these successive stages are additive components of the reaction time. He then developed the subtraction rnethod: mean reaction times from two different tasks are compared, where one task is thought to require all the stages of the other task, plus one additienat stage. The difference between mean reaction times is taken to be an estirnate of the mean duration of the interpolated stage. Although Donders used the subtraction method to identify and study mental activities in the stimulus-response sequence, it might be possible to apply this method to the positioning times of targets at diagonal angles. In these cases, the subjects response con-sists of two sequences of keying movements, one horizontal and one vertical. For every angle in the experiment, the equations for positioning times are calculated according to
Tp
=K
0+ K.(Dx /0.25 + Dy/0.5) s.
(5-1)
using regression analysis (see sectien 5.2 for an explanation of this formula). Table 5.1 gives. the eight equations along with the standard error, squared multiple correlation, and the position-ing time at the average target distance (D .. 5.4 cm), tp . When adding the mean positioning times of two targets at diagonal angles, one adds twice the constant Ko including components as the time needed for the hand initially to reach the arrow keys and to make the selection. These components of course occur only once. So, the equations for positioning time with diago-nat angles should be
