The aim of this study was to analyze the cursor trajectories of adolescents with cerebral palsy (CP) when using a mouse for point-and-click computer tasks. By identifying some of the factors limiting cursor movement and gaining a better understanding of human movement, it will be possible to design more accessible computer interfaces.
This study evaluated cursor trajectories of 29 individuals with bilateral CP who had different levels of upper limb function as measured by the Manual Ability Classification System, and compared the results with those of 12 adolescents with typical development.
Among adolescents with typical development, movement time increases linearly as the index of difficulty increases (Fitts' law); however, this linearity was not apparent in adolescents with bilateral CP.
Interfaces for members of the population are designed around Fitts' law, with low precision requirements at indices of difficulty lower than 4. We found that interactive displays for adolescents with CP should be limited to an index of difficulty of 2.
Cerebral palsy (CP) is a common cause of motor dysfunction affecting children and adults and is an umbrella term for a group of disorders of movement and/or posture which includes spasticity, dyskinesia, and ataxia. Until recently, it was believed that the prevalence rate of CP had remained steady at 2 to 2.5 per 1000 live births, but a 2002 report suggests that the prevalence is higher, nearer 3.1 per 1000 births.
The majority of children with CP have some degree of upper limb involvement, which can be classified using the Manual Ability Classification System (MACS). This is a self-report scale and is wide-ranging, from MACS levels I to V. Table 1 provides descriptions of the functional levels.
Table 1. Descriptions of the five Manual Ability Classification System (MACS) levels
Handles most objects but with somewhat reduced quality and/or speed of achievement
Handles objects with difficulty; needs help to prepare and/or modify activities
Handles a limited selection of easily managed objects in adapted situations
Does not handle objects and has severely limited ability to perform even simple actions
Computers are often used to increase interaction in education, work, and entertainment, but many adolescents with CP find interacting with computers difficult because of accessibility issues. We conducted a systematic review to identify the technologies currently available or in prototype form. In the course of the review it became apparent that many devices are only at the prototype stage and have been designed without engaging users and therapists in their development, thus reducing acceptability. A survey of computer users in various MACS levels showed that it is likely that adolescents functioning at MACS levels I, II, and III will access a computer by using a mouse, while those who function at MACS levels IV and V use alternative access devices, such as joysticks, or track balls, etc.
The influence of differences in upper limb function on mouse movement efficiency is not well known, though there is evidence that movement time is much slower than in those with typical development.[6-9] Approximately 65% of children with CP are classified as MACS levels I to III. This suggests that the design of interfaces to enable effective mouse control and cursor movement is important, possibly more so than the development of new devices.
The International Standards Organization has developed a method to evaluate input devices for computer use, including measurements of efficiency, comfort and effort, using tasks based on Fitts' law. Fitts' law is often used to model the act of pointing during human–computer interaction, in this case by use of a pointing device (e.g. mouse or joystick) to manoeuvre a virtual cursor to an object on the screen.[11, 12] The law describes a linear relationship between the time taken to perform the movement (MT) and the size and the distance, that is the ratio between the centers, of the targets. It can be represented as:
where a is the intercept and b is the slope of the regression line, with ID being the theoretical index of difficulty for a pointing task. The ID is calculated by:
where A is the centre-to-centre distance between targets and W is the width of the targets. Fitts' law predicts a trade-off between the speed and accuracy of visually guided movements, with the time taken to move to a target being a function of both the size and distance of the target. Generally, the larger the target and the smaller the spacing between targets, the easier it is to select.
Fitts' law has been extensively used in the field of human–computer interaction to study aiming tasks and has been shown to apply across many different user groups (e.g. the elderly, individuals with spastic hemiplegia), a variety of input devices (e.g. CrossScanner [R. J. Cooper & Associates, Lake Forest, CA, USA], ASL Mouse [Adaptive Switch Labs, Inc., Spicewood, TX, USA], joystick) and different parts of the human body (e.g. brain, hand, eyes). Of interest, the early work by Card et al. using Fitts' law to assess the ease of use of different input devices is credited with influencing the choice of the mouse as the preferred input device by Xerox and then Apple. Since then, many studies of conventional mouse and touch pads using Fitts' law have been conducted.[23, 24]
As far as the authors are aware, no study to date has examined cursor control in individuals with bilateral CP of different MACS levels in order to better understand how to improve computer interfaces to meet the abilities of users with motion impairment. Although various algorithms have been developed,[25, 26] the design of these has been based on users' perceived needs from an engineering design perspective, rather than a user-centred approach based on observed requirements. Information that provides estimates of speed when undertaking point-and-click tasks across impairment levels will enable development about general recommendations of interface design. This paper seeks to understand how individuals with CP of different MACS levels approach computer point-and-click tasks.
While Fitts' law is not the only model of human movement to evaluate effective computer access, the International Standard, based on evidence-based practice of interface design, was deemed to be the best representation of the industry standard.
Based on Fitts' law and the increased level of impairment across the MACS levels, our hypotheses were that (1) within each impairment level, the response time and movement time would increase linearly as the index of difficulty increased and the average speed would decrease and (2) across MACS levels, response time and movement time would increase while average speed would decrease as impairment increased.
This study was approved by both the Northern X Regional Ethics Committee in Auckland, New Zealand and the Bloorview Research Ethics Board in Toronto, Canada. The inclusion criteria for the study were that participants had to have a diagnosis of CP, be aged between 13 and 25 years, and be comfortable accessing computers with the one-handed use of a computer mouse.
Twenty-nine participants with bilateral CP were recruited from Rehabilitation Centres in Auckland, New Zealand or Toronto, Canada, from January to December 2009. Each participant with CP self-reported his or her MACS level. The distribution of MACS levels was as follows: adolescents with typical development, n=12; MACS level I, n=3; level II, n=14; level III, n=9; and level IV, n=3. No participants were MACS level V as they tend not to use a mouse for access (unless using a foot). Participants of MACS level I more often have hemiplegic CP than other forms of CP. As we examined only individuals with diplegic CP, fewer participants were available that fitted within the MACS level I category. Twelve adolescents with typical development were recruited as comparison participants.
Computer task test
Using the Movement Time Evaluator software, participants conducted a computer target selection task with indices of difficulty of 1.6, 1.9, 2.2, and 2.5. These indices of difficulty were chosen to be within an attainable range for adolescents with CP, but also to provide sufficient data to develop a Fitts' law regression model between index of difficulty and movement time (linear fit to find slope, b). Movement time (the time interval from a mouse click on the home block to mouse click inside the target block) and distance trajectories were collected. From these data, additional speed measures[9, 28-31] were calculated, including response time (the time difference between the first mouse click and the onset of mouse cursor movement) and the average speed (the average value of distance over time).
The participant used a computer provided by the researcher running on Windows XP home edition OS software with mouse operation and a standard mouse (Lenovo M/N: LXH-MOAFUO USB [product number 25007694]). For each trial, the participant positioned the cursor on a ‘home’ target located in the centre of the screen. Upon clicking the mouse button, the ‘home’ target disappeared and another ‘final’ target appeared. The participant moved the cursor to within the ‘final’ target and clicked the mouse button. This target disappeared and the ‘home’ target reappeared.
A bidirectional discrete tapping task indicated by the International Standards Organization Guidelines to evaluate computer task performance was undertaken to evaluate cursor movement between two targets of the same size. The location of the final target was randomized such that it appeared in each direction, 0°, 90°, 180°, and 270°, from the ‘home’ target a total of five times for each block of trials for a total of 20 trials. All trajectory movement was recorded at a frequency of 100Hz (10ms).
Each participant completed four blocks of 20 trials (A–B–A–B) with indices of difficulty of 1.6 (80 × 80 pixels) and 2.2 (30 × 30 pixels), which were counterbalanced among participants – this gave a total of 10 trials per participant per direction for each size of target. This was followed by an additional four blocks (C–D–C–D) with indices of difficulty of 1.9 (46 × 46 pixels) and 2.5 (22 × 22 pixels), which were also counterbalanced. It was anticipated that some participants may not be able to reach the ‘final’ target at a high index of difficulty (2.5). The experiment was designed such that if the accuracy was poor in the ‘B’ trials, the participant would not attempt the C–D–C–D block as the ‘D’ trials would require too much control (making the participant frustrated and disillusioned). A Visual Studio C# code was developed to analyze the trajectories based on the path evaluation measures.
For each trial, the response time, movement time, and total distance were recorded. From the distance and movement time, the average speed was calculated. Average speed gave an indication as to whether quick, inaccurate movements or slow, controlled movements had been used. For example, although the same movement time for the same size and distance of target might be the same in two individuals, one might move the cursor a longer distance in reaching the final destination (for example, overshoot and have to correct). In addition, the number of clicks outside the boundaries of the target in a given trial before a successful click was recorded as the number of error clicks for each participant. A successful click within the target was required before the next trial could begin.
Statistical analysis included repeated measures three-factor analysis of variance modelled in sas software (SAS Institute Inc., Cary, NC, USA) using the mixed model with independent variables angle, MACS level groups (with adolescents with typical development as a group) and the index of difficulty. This model also accounted for unequal variances among groups. To avoid the contribution of the ‘learning effect’, only the second block at each index of difficulty was used in the statistical analysis. Each of the dependent variables of response time, movement time, and average speed were modelled separately. Movement time and average speed were both log transformed to ensure normal distribution. Outliers were identified, using residual analysis and Cook's D, and removed. Scheffé's post hoc analysis teased out interaction effect differences identified in the model (p<0.05).
Error clicks were recorded for all participants such that the following error rates were observed (% error): adolescents with typical development (11.7%) and adolescents with MACS level I (15.3%), level II (22.3%), level III (25.6%) and level IV (41.3%).
Table 2 shows the significant contributions of angle, MACS level, and index of difficulty for response time, movement time, and average speed.
Table 2. Statistical results for response time, movement time, and average speed
Increasing index of difficulty within each MACS level group
The results of the analysis of response time, movement time and average speed with increasing index of difficulty at each MACS level are discussed in detail.
Within each MACS level group, there was no change in response time as the index of difficulty increased (p=0.08).
Adolescents with typical development displayed an increase in movement time as the index of difficulty increased (see Fig. 1, where different letters represent significant differences). Among adolescents with CP, the movement time increased at indices greater than 2 in the MACS levels II and III groups, but there were no significant differences in movement time observed across the indices of difficulty in the MACS levels I and IV groups.
Within the groups of adolescents with typical development and of adolescents in MACS levels II and III, the average speed decreased significantly from indices of difficulty 1.6 to 1.9. In all groups, there was no significant difference between indices of difficulty 2.2 and 2.5 (see Fig. 2).
Increasing index of difficulty across impairment levels
A main effect of participants' MACS level was observed (F4,5.65=7.06, p=0.02), with a significant increase in response time as the MACS level increased from I to IV. Response times were shorter (512ms) in comparison participants than in participants in MACS level II (1106ms, p=0.04), and those in MACS level III trending to the same (1596ms, p=0.07). The main effects of the index of difficulty and the angle of direction were non-significant (Table 2).
A main effect was observed for the index of difficulty of the task (F3,3138=75.93, p<0.001), the participant's MACS level (F4,4.75=18.58, p=0.004) and the angle of direction (F3,3136=8.54, p<0.001).
The interaction effect of index of difficulty and MACS level showed that at each index of difficulty there was a significant difference between the groups of adolescents with typical development and those of MACS levels II and III, but no difference between adolescents with typical development and adolescents in MACS level I, or between participants of MACS level I and those of MACS levels II, III, or IV (Fig. 3).
The interaction of MACS level and angle showed that, for all directions, movement time was longer for adolescents in MACS level III than for adolescents with typical development (p<0.01 at 0° and 90°, and p<0.04 at 270° and 180°).
The interaction of index of difficulty and MACS level for average speed showed that, at an index of difficulty of 1.6, average speed was higher for adolescents with typical development than for adolescents in MACS levels II or III (p<0.01 and p=0.04 respectively), and the same was observed at index of difficulty of 2.2 when comparing adolescents of MACS level II and adolescents with typical development (p=0.03).
The interaction of MACS level and angle showed that the average speed of participants in MACS level II was lower than that of typical adolescents at directions of 0° and 90° (p=0.03) while participants in MACS level III were slower at 0° (p=0.04).
It was hypothesized that impairment levels have a strong influence on evaluation measures. For example, a higher impairment level is associated with greater movement difficulty; thus, movement time and response time are expected to be longer. Although movement time was expected to be greater for adolescents with CP, it was also expected that Fitts' law would still apply.
Within each MACS level group
The first hypothesis suggested that, within each MACS level group, as the index of difficulty increased, both the response time and movement time would also increase and the average speed would decrease. This defines the Fitts' law model and has been shown to be true in participants with typical development when undertaking point-and-click tasks.[23, 32] Smits-Engelsman et al. found that children with congenital spastic hemiplegia adhere to Fitts' law in a visually guided tapping task in which the user targets an object with the hand. However, a computer task requires a spatial mapping between the movement of the hand and the movement on the screen that is more difficult than a visually guided tapping task. Before mouse control was commonplace, Bravo et al. used a touch board in conjunction with a computer to play a game requiring participants to target screen icons. They found that individuals with spastic quadriplegia did not obey Fitts' law.
The present study was based on 29 participants with bilateral CP performing computer point-and-click tasks. We found that Fitts' law does not adequately describe the speed–accuracy trade-offs when targeting icons on a screen for participants with CP at any of the MACS levels in this study (Fig. 1). The response time appears to be unaffected by the increase in the index of difficulty, and the movement time and average speed do not appear to increase linearly as indicated by Fitts' law. Below an index of difficulty of 2, the tasks are equally difficult, and the same is observed above an index of difficulty of 2. This observation is consistent with the results observed by Gump et al. and Bravo et al. and suggests that when conducting computer tasks, adolescents with bilateral CP do not appear to adhere to Fitts' law.
Across MACS level groups
As hypothesized, among children with decreased upper limb ability (higher MACS levels), there was tendency for response time and movement time to increase while speed decreased. The data also became more variable as impairment increased. While the data for adolescents in MACS level I were not significantly different from those obtained in adolescents with typical development, they were also not different from the data for participants of other MACS levels. In our case, the edge-to-edge distance between targets was similar across the indices of difficulty and all participants benefited from larger icons rather than smaller ones.
There is a lack of evidence to support the use of Fitts' law to design a human–computer interface for individuals with CP. Interface designers choose an index of difficulty trade-off based on ‘screen real estate’ and usability. The International Standards indicate that task precision (the measure of accuracy required for a point-and-click task) with an index of difficulty of less than 4 is ‘low’, while any task precision greater than 6 is ‘high’.
In the case of individuals with bilateral CP, it appears that the most universal interface design would be one that limited the index of difficulty to a maximum of 2. A jump in difficulty occurs above index difficulty 2 for participants of both MACS levels II and III. This value is substantially lower than the ‘low’ precision requirements identified in the International Standards.
According to Soukoreff and MacKenzie, Fitts' law can be applied to movement in which mid-trial pauses do not exist (these pauses violate the requirement for the law that movements be rapid) and the error rate is lower than 4%. In our case, the error rate was much higher than 4%. In addition, that users with motion impairments experience instability of movement and pauses is to be expected. This suggests that our participants' movements do not meet the initial criteria for Fitts' law and that, perhaps, another model of movement is required.
Examining cursor control at four different indices of difficulty, we found that response time and movement time were longer and average speed of cursor movement was slower in individuals with bilateral CP (especially for those in MACS levels II and III) than in adolescents with typical development. The model of movement did not appear to follow Fitts' law at any of the MACS levels.
The results of this study suggest that, while interfaces designed to meet the needs of the general population are based on Fitts' law to evaluate trade-offs between speed and accuracy, interfaces for adolescents with CP should be designed with an index of difficulty of no greater than 2.