The normal effects of aging include some decline in cognitive, perceptual, and motor abilities. This can have a negative effect on the performance of a number of tasks, including basic pointing and selection tasks common to today's graphical user interfaces. This paper describes a study of the effectiveness of two interaction techniques: area cursors and sticky icons, in improving the performance of older adults in basic selection tasks. The study described here indicates that when combined, these techniques can decrease target selection times for older adults by as much as 50% when applied to the most difficult cases (smallest selection targets). At the same time these techniques are shown not to impede performance in cases known to be problematical for related techniques (e.g., differentiation between closely spaced targets) and to provide similar but smaller benefits for younger users.
Graphical interfaces, user input, interaction techniques, pointing and selection tasks, Fitts' law, aging.
Older adults are the fastest growing segment of the population in the United States and Europe. By the year 2000, the World Bank estimates that 20% of the population in countries with market economies will be over 60 years of age. It is known that as people age, their cognitive, perceptual, and motor abilities decline, with negative effects on their ability to perform many tasks [5]. Computers can play an increasingly important role in helping older adults function well in society. Despite this, little research has focused on computer use of older adults.
Graphical interfaces contribute to the ease of use of computers. Interfaces based on windows, icons, menus, and pointing (so called WIMP interfaces) allow fairly non-trivial operations to be performed with a few mouse clicks. However, for older adults there is evidence that even simple mouse clicks can be quite challenging [8,9].
The trade-off between the accessibility of targets and the amount of information presented is a fundamental issue in human-computer interface design. People have particular difficulty working with small icons often utilized to fit a lot of information into one display. Two examples of small targets in common use today are toolbar icons and border manipulation handles.
The ease (speed and accuracy) with which a user can select an icon depends on the size of the icon and the distance the cursor must be moved. The relationship between the difficulty of hitting a target and target size is quantified by Fitts' law [1]. The variable part of Fitts' equation is called the index of difficulty. The index of difficulty is log2(2D/W), where D is distance to the target and W is width of the target. The index of difficulty increases as the width decreases. Increased difficulty causes decreased performance, as measured by speed and accuracy.
There is an extensive literature demonstrating that the ability to make small movements decreases with age [10]. This decreased ability can have a major effect on the ability of older adults to use a pointing device on a computer. Some of our previous work [8,9] has shown that even experienced older computer users move a cursor much more slowly and less accurately than their younger counterparts. In addition, older adults seem to have increased difficulty (as compared to younger users) when the targets become smaller. In our most recent work, we found that older computer users took over twice as long and made more than five times as many cursor positioning errors as younger users when moving to small icons [8]. Yet the icons used in that study were larger than some of those found on current interfaces. For older computer users, difficulty in positioning a cursor can be a severe limitation.
There are several possible solutions to the problem of decreased ability to position the cursor with a mouse. One obvious solution is to increase the size of targets on computer displays (e.g., Macintosh folders may be viewed using large or small icons). Greater size decreases Fitts' index of difficulty, resulting in greater speed and accuracy. However, this can often be counterproductive since less information is being displayed, requiring more navigation. Another option is to arrange the icons against solid borders that do not allow cursors to overshoot the target. This design increases speed and accuracy by making the depth of the target appear infinite along some trajectories [6]. Of course, not every icon can be next to a solid border.
Both of these approaches are possible but not very practical. They each require substantial change to user interface layout and/or interaction techniques for older users. Further these adaptations may have less desirable effects for the majority user population (because, for example, less information can be immediately accessible). A more feasible approach would be to provide new interaction techniques which operate within existing layouts, with only small changes to isolated portions of the software, and which have at least neutral effects (if not providing slight improvements) for younger adults. This paper considers two such techniques: area cursors and sticky icons.
An area cursor is a cursor that has a larger than normal activation area. Most cursors have a single point or "hot spot" which serves as the point of activation for the cursor. Area cursors simply have larger hot spots. There is evidence that performance with area cursors is better than performance with regular cursors for some target acquisition tasks [2,11]. Improved performance follows the principles of Fitts' law, where W represents the width of the cursor if the cursor is broader than the target. For example, if one is trying to select the edge of a box, the edge may be only one or two pixels wide. With a one-pixel hot spot on the cursor, the user must make a very finely controlled movement with the mouse because the effective target width is one pixel. This movement is very difficult for older adults [8]. With an area cursor that has a larger hot spot (say 12 by 12 pixels), the movement is much easier because the effective width of the target is now 12 instead of one pixel. Area cursors would clearly aid older computer users.
There are several potential problems associated with implementation of area cursors. First, large opaque cursors may block the area of the display behind them. This can obscure screen information relevant to the task. Second, area cursors are not useful for tasks that require discrimination between two icons or buttons that are located close to one another. This situation causes a problem because most current area cursors react unpredictably if more than one icon is within the activation hot spot.
Researchers have studied a translucent cursor design to remedy the problem of large cursors that obscure potentially important screen information. The results of their experiments suggest that viewing information behind translucent cursors does not hinder performance [11]. We utilized this translucent design in our study and investigated a new feature that should improve performance for discrimination tasks with area cursors.
The area cursor in our study was designed to have two hot spots &endash; a single point hot spot, as in a conventional cursor, and a larger hot spot, of the same size as the area cursor. When there is a single target within its boundary the cursor behaves like an area cursor. The point hot spot takes effect when more than one target falls within the cursor's bounds. In this mode the user is able to manipulate small clustered icons and text with the area cursor as if it were a regular cursor. We predicted that this design would increase performance for hitting isolated small targets without hindering performance on tasks that require discrimination between two icons placed closely together.
Another design feature that could aid performance is the use of sticky icons. Sticky icons are designed to have an automatic reduction of the cursor's gain ratio when the cursor is on the target icon. The gain, measured in units of mickeys, determines the number of pixels moved on the screen in response to a single increment of movement by the physical device. By decreasing the gain locally, one can make the effective size of the target larger, making it easier to stop the cursor on the icon. This makes the icon "sticky."
A study by Keyson investigated the effects of decreased gain toward and into targets and increased gain out of and away from targets. This gain structure requires users to increase movement of the mouse (or track-ball in Keyson's example) to approach targets and to decrease movement of the mouse to leave targets. This creates targets that are functionally "sticky." Keyson found that this type of dynamic gain structure significantly improved performance by decreasing cursor positioning time [3].
A problem with the use of sticky icons as described is that there is often interference due to intervening icons. Unless the trajectory is selected very carefully we will have cases where the cursor moves over icons that are not the target of the movement. If all icons are sticky, cursor movement will be retarded by every icon along the way, leading to slower cursor positioning in general.
Ideally we would like only the target icon to be sticky. While the goal of a cursor movement cannot always be predicted, it can be estimated with some accuracy. Our approach to this problem was to make the stickiness of the icon dependent on changes in cursor velocity. Previous work has shown that cursor movements are made up of an initial high velocity pulse, followed by slower corrective sub-movements to reach the target icon [4,7]. Using this information we designed a system in which the stickiness of an icon was dependent on a velocity profile of the cursor movement. In our design, gain was constant until the velocity of the cursor dropped to 30% of its peak velocity during the movement. Once this occurred, and the cursor was over an icon, the gain was adjusted downward to make the icon sticky.
To test our predictions, we compared six types of cursors:
1. Pointer cursors without sticky icons and
without adaptive-gain,
2. Pointer cursors with sticky icons and without
adaptive-gain,
3. Pointer cursors without sticky icons and with
adaptive gain,
4. Area cursors without sticky icons and without
adaptive-gain,
5. Area cursors with sticky icons and without
adaptive gain, and
6. Area cursors without sticky icons and with
adaptive-gain.
The area cursors were translucent and equipped with a dynamic hot spot. All cursor types were tested with all combinations of (a) three distances from the target, and (b) four widths of targets. Each distance/width combination was tested in a condition with (i) no intervening icons, (ii) with intervening icons placed close, or (iii) relatively far from the target icon.
The participants were 16 younger and 16 older adults. The younger adults (mean age = 23.4 years) were recruited undergraduate and graduate students from the Georgia Institute of Technology. All students were regular users of computers with a mouse as the pointing device. The older adults (mean age = 70.1 years) were recruited from the older adults subject pool of the psychology department at the university. The older adults had all participated previously in a study that provided three hours of training using a computer mouse followed by a three-hour experiment that required them to use the mouse as a pointing device. The older adults were paid US $50 for their participation in the experiment. Younger subjects were either volunteers or received course credit for their participation. An incentive of $25 per 8 participants was awarded based on overall performance in the experimental task.
The experiment used a seven-factor, mixed factorial design. The within-subjects variables that constituted task type were movement distance (D=100, 200, 400 pixels), target width (W= 4, 8, 16, and 32 pixels), and screen environment (none, near, or far intervening icon). Type of cursor (area and pointer) and type of icon (normal and sticky) were also within-subject variables. The between subjects variables were age and dynamic-adaptive gain. Half of the subjects had dynamic variable gain based on the speed of the cursor toward the icons, half did not.
The experiment was run on 386 based IBM PC clones with 14-inch SVGA color monitors, in VGA mode with 640 X 480 pixels. Participants used a Microsoft two-button bus mouse. The mouse was placed at the subject's discretion on the table. The left mouse button was used for the experiment. Normal mouse gain was set at a constant 1 mickey to 3 pixels ratio for all conditions. The bus mouse recorded position changes at 5 ms intervals.
Figure 1: Two cursor types and activation pattern of area cursor with adaptive hot-spot when located over two icons. The selected icon is shown in reverse video.
The area cursor was a 12 X 12 pixel box with a simple border containing a cross-hair. When the cursor was in area cursor mode, any part of the 12 X 12 cursor could activate an icon. When the area cursor spanned two icons, only the icon under the center cross-hair was activated. The pointer cursor appeared similar to the area cursor but the cross-hair did not have a border and the activation area was always at the center of the cross-hair (see Figure 1).
Sticky icons were created by adjusting the mouse-to-cursor gain function. The normal gain was set at 3 to 1. When the activation area of the cursor moved over a sticky icon, the gain ratio was lowered to 1 to 1. This dynamic gain function applied as long as the cursor hot spot was over the icon. Once the hot spot moved off the icon, the gain ratio returned to the standard 3 to 1 ratio.
The dynamic gain for sticky icons was adjusted based on the relative change in the velocity of the cursor as explained below. The gain was lowered to 1 to 1 when two conditions had been met:
1. The cursor was located on an icon; and
2. The velocity of the cursor was less than 30% of the peak velocity
attained during the movement.
The screen environment was varied by the presence or absence of an intervening icon between the starting location of the cursor movement and the target icon. This was done by displaying a tower of icons (equal in width to the target icon) that went from the top to the bottom of the screen. This ensured that the cursor would be moved across the intervening icon. The intervening icons varied in distance from the target icon on the dimension of the primary direction of movement. The intervening icon tower was located either next to the target icon with no space separating it from the target icon, or 50 pixels from the target icon. The intervening icons also became sticky based on the dynamic-gain function described above.
Subjects participated in four, 30 - 45 minute sessions. Each session consisted of 30 blocks of trials. There were 10 blocks for each of the three levels of intervening icons. The levels of intervention were: (i) No intervening, (ii) Intervening icons located 50 pixels away from the target, (iii) Intervening icons located immediately before the target. Each block consisted of 12 trials, one for each movement distance by target width combination. The order of blocks and trials was randomly chosen.
Two types of cursors used: Area cursors and Pointer cursors, as discussed. Half of the subjects used the area cursor for the first session and half of the subjects used the pointer cursor for the first session. All subjects used alternating cursor types on subsequent days. The order of usage was counterbalanced according to a Latin squares design.
Immediately before the first session, each subject was given written instructions regarding the use of the apparatus and the type of cursor assigned to them that day. For subsequent sessions, the participants read a shorter version of the instructions that explained the type of cursor assigned to them. The first session began with two blocks of practice trials (72 trials) to familiarize them with the particular cursor-type and icon-type configuration assigned to them that day. Other sessions began with 36 practice trials. Performance during these trials was measured and used as feedback for the participants.
At the beginning of each block, a text box appeared that read "click here to begin next block". The block of trials began when the participant pressed the mouse button with the cursor inside this text box. Each trial began with a home box on the left edge of the screen and a target box on the right side of the screen. The square target appeared with its center horizontally displaced D pixels away from the center of the home box. The target was a red outlined box. The intervening icons, when present, were located between the home box and the target icon .
The initiation of each trial began when the participant positioned the cursor in the home box and pressed the mouse button. The trial ended when the participant released the mouse button inside the target icon for a "hit" or outside of the target box for a "miss". Missed trials were randomly replaced in the remaining trials to be performed for that block.
The computer displayed points earned for each trial. The point system was designed to reward subjects equally for speed and accuracy. Missed targets resulted in an automatic decrement of 40 points. Points were earned for hit targets that were reached in a timely manner. Participants received 1 point for each 100 milliseconds that the total positioning time for that trial was below a preset value. This value was specific to each D/W combination and differed for the two age groups. Subjects lost 1 point for each 100 milliseconds that the total positioning time was above that preset value. The number of points that could be earned per trial was roughly equivalent for all levels movement difficulty. The participant's incentive to earn points was a $25 reward for the participant with the most earned points at the end of the experiment. This reward was given to one of every eight participants.
The computer began recording mouse movements after the cursor was positioned inside the home box and the left mouse button was pressed. The computer recorded mouse positions for the horizontal and vertical planes at 5 ms intervals until the mouse button was released.
The data were analyzed using two seven-factor mixed analyses of variance. The dependent variables were mean movement time for correct trials and percent correct trials. These analyses revealed over thirty significant effects (main effects and interactions). Most of the significant effects were for time, as accuracy varied little.
For both age groups, accuracy was above 96 percent and there was no evidence of a speed-accuracy trade off. We report only tests of specific effects that are directly related to our research questions and that were significant at the p < 0.05 level. The complete analyses of variance tables can be obtained from the authors.
Question 1: Did the use of an area cursor improve performance for older computer users?
The analyses of movement time clearly showed the beneficial effects of using an area cursor versus a pointer cursor. There was a significant main effect of cursor type (F (1,28) = 59.71, p < .001). There were also two significant , two-way interactions: cursor type and age ( F (1,28) = 6.14, p < .05); and cursor type and target width (F (3,84) = 38.74, p < .001). In addition, there was a significant three-way interaction of cursor type, target width and age (F (3,84) = 4.20, p < .01). As can be seen in Figure 2, younger adults made faster movements than did older adults, and movements were faster with an area cursor than with a pointer cursor for both age groups. However, the older adults benefited more from the area cursor when moving to small targets than did the younger adults.
Figure 2. Mean positioning time for the two area groups for pointer and area cursor by width of target icon.
Question 2: Could the area cursor be used to discriminate between adjacent icons?
To address this issue we looked at the interaction between intervening icon and cursor type. This interaction was significant (F (2,56) = 35.51, p < .001). Follow-up tests revealed that the area cursor led to faster movements when there was no intervening icon or when the intervening icon was far from the target icon. More importantly, there was no difference between area cursors and pointer cursors when the intervening icon was next to the target. This shows that both younger and older adults were able to use the area cursor with the adaptive function to select small targets in a crowded environment.
Question 3: Did the use of sticky icons aid performance?
The analyses of icon type and adaptive gain function both yielded effects that approached significance (p < .10). There was however, a significant interaction of icon type and target width (F (3,84) = 19.24, p < .001). Follow-up tests revealed that there was a beneficial effect of sticky icons when target width was at its smallest as compared to the largest target size. As can be seen in Table 1, older adults seem to benefit more from sticky icons and the adaptive gain function than younger adults. However, the beneficial effect of the adaptive gain function should be interpreted with caution as this was a between subjects variable and this group was faster in all conditions.
Table 1. Means for positioning time and hit rate for the two age groups for the six cursor types.
|
|
Younger |
Older | ||
|
|
Time (ms) |
Hit Rate |
Time (ms) |
Hit Rate |
|
Pointer Only |
759 |
95.0 |
1893 |
95.0 |
|
with sticky |
712 |
97.4 |
1869 |
96.3 |
|
and adaptive |
743 |
97.7 |
1485 |
95.5 |
|
Area Only |
639 |
96.9 |
1658 |
96.3 |
|
with sticky |
596 |
97.9 |
1596 |
97.7 |
|
and adaptive |
591 |
99.0 |
1203 |
97.9 |
Question 4: How much can cursor positioning performance be improved for older adults?
In order to determine the amount of improvement in positioning that can be gained from the various design interventions, we compared mean positioning time for older and younger adults on the two extreme types of cursor systems: "normal" cursor, where there is a pointer cursor, non-sticky icons and no dynamic gain function; and "fully augmented" cursor, where there is an area cursor, sticky icons and the dynamic gain function. Comparing movement performance for these two extremes we see a clear advantage for the fully augmented cursor (means = 591 for younger, 1204 for older) compared to the normal cursor (means = 759 for younger, 1894 for older). The advantage for using the fully augmented cursor is even greater if we compare the means for the condition with no intervening icons and the targets were smallest. For older adults movement time was 2421 msec for the normal cursor and only 1204 msec for the fully augmented cursor, a drop in movement time of more than 50%. The same pattern of results was found for younger adults (mean for normal = 941 msec, 600 for fully augmented) although the effects were not as dramatic.
This study clearly demonstrates that cursor positioning among older computer users can be improved. As we reported previously [8], older computer users position the cursor much more slowly than younger computer users and have great difficulty making correct movements to small targets. However, the time required to move the cursor to a small target was cut by over 50% when using a fully augmented pointing system as compared to the system used on most computer systems.
The study also shows that both older and younger computer users can use the area cursor effectively. When an icon was not in close proximity to another icon, cursor positioning time was much faster with the area cursor. When there were two icons next to one another, both older and younger computer users could select a single icon as quickly and easily with the area cursor as the pointer cursor. The results show that use of an adaptive hot-spot with the area cursor seems to have eliminated one drawback to it use. The results of the study also show that sticky icons can be beneficial, especially for older adults. The effects are not as dramatic as those for the area cursor, but there was a clear benefit in positioning time with smaller targets. As this is a particular cursor positioning problem with older adults, the use of sticky icons could help alleviate this problem.
In addition to the obvious benefits for positioning time that result from this system, this approach to solving age-related problems in computer usage can be easily implemented on current interfaces. Rather than making changes to all software, these options can be made part of mouse driver software and will then apply to all programs used.
This work was supported in part by a grant from National Institutes of Health (National Institute on Aging) grant no. P50 AG11715 under the auspices of the Southeastern Center for Applied Cognitive Aging Research (one of the Edward R. Roybal Centers on Applied Gerontology Research), by the National Science Foundation, under grants IRI-9500942 and CDA-9501637, and by the Intel Corporation.