Effects of Angle of Approach on Cursor Movement with a Mouse: Consideration of Fitts' Law

Transcription

1 Pergamon Computers in Human Behavior, Vol. 12, No. 3, pp , 1996 Copyright 1996 Elsevier Science Ltd Printed in Great Britain. All rights reserved /96 $ S (96) Effects of Angle of Approach on Cursor Movement with a Mouse: Consideration of Fitts' Law Thomas G. Whisenand and Henry H. Emurian Department of Information Systems, University of Maryland Baltimore County Abstract -- Research related to various aspects of human-computer interactions has become increasingly important to the design and implementation of effective interface systems and input devices. As the man-machine interface has become more direct, the analysis and application of models to account for this interaction have received increasing attention by investigators. In that regard, the speed and accuracy of human motor movements associated with computer input devices have often been modeled by Fitts' law. Most such analyses, however, have not considered the angle of approach as a factor. The present study investigated the interactive effects on movement time of the angle of approach, the size of the target, and the distance to the target for the use of a mouse to select icon-like targets presented on a computer's display screen. Angle of approach, target size, and target distance significantly affected movement time. Although the global R 2 based on a Fitts' law linear model was low in comparison with other studies, an analysis of the residuals showed that the linear model's effectiveness in prediction interacted with the other variables under investigation. Copyright 1996 Elsevier Science Ltd Requests for reprints should be addressed to either author at Human-Computer Interactions Laboratory, Information Systems Department, University of Maryland Baltimore County, Baltimore, MD

2 482 Whisenand and Emurian A 'mouse' is a handy device that allows a computer user to control a cursor or pointer on the display and to 'click' positions that determine the flow of information. English, Engelbart, and Berman (1967) developed an input device at Stanford Research Institute, Menlo Park, CA, and they referred to the device as a 'mouse' and to the cursor as a 'bug.' This was, perhaps, the first mention in the literature of the term 'mouse' as an input device. The term 'bug' for the cursor apparently did not survive. An interesting historical perspective of the development of the mouse and user-friendly interfaces is presented by Perry and Voelcker (1989). Under OS/2, Windows, and other operating systems, positioning a cursor with a mouse to select an icon initiates application processing or opens a processing window. The target area of the default icons generated by these operating systems is generally square and approximately 1.5 em per side (0.3 cm per side for minimized icons). With the proliferation of such windowing-type, icon-based software systems, the mouse has become the most utilized direct manipulation interface device supporting humancomputer interactions in all manners of tasks and settings. It is not surprising, then, that the speed and accuracy of a user's movements to accomplish pointing and selection tasks have become the focus of intense evaluation and modeling. The literature contains reports that compare the effectiveness of the mouse with other input devices and that propose and evaluate models to account for a user's movement time with the mouse. Early studies compared the effectiveness of the mouse with other devices to input information into the computer by selecting items on the VDT screen. English et al. (1967) compared the mouse with a joystick, a 'grafacon' potentiometer, and a fight pen for selecting characters and words on the screen. No statistical analysis of the data was reported, but the figures showed that the mouse was associated with the fastest selection time for 7 of 9 comparisons involving experienced and inexperienced subjects. Other comparative analyses include comparisons between high precision touchscreens and a mouse (Sears & Shneiderman, 1991), among touch, panel, mouse, and keyboard input devices (Karat, McDonald, & Anderson, 1986), among mouse, graphic tablet, and trackball (MacKenzie, Sellen, & Buxton, 1991), and among mouse, absolute touchpad, relative touchpad, trackball, displacement joystick, and force joystick (Epps, Snyder, & Muto, 1986). Additionally, Greenstein and Arnaut (1988) reported a survey of the earlier literature comparing and evaluating various input devices. Finally, Card, Mackinlay, and Robertson (1991) discussed the mouse and other input devices from the perspective of a morphological analysis of the design space of the devices, which represents input devices as 'points' in a parametrically described space composed of potential movements of the device together with operators that govern combinations of movements.

3 Angle of approach and Fitts" law 483 Although providing important evaluative guidelines regarding the relative effectiveness of various input devices, many of these studies were motivated, for the most part, by practical rather than theoretical considerations, and some studies lacked a framework for the emergence of first principles of interface design. As the stream of research matured, however, the similarity of moving a pointer from one screen position to another, using the mouse, to the classic reciprocal 'tapping' task reported by Fitts (1954) was recognized. Evaluations of Fitts' law ~, when applied to movements of the mouse pointer from one screen location to another, began to appear in the literature. Although traditionally stated as a 'law', the statement that the calculated 'index of difficulty' is linearly related to 'movement time' is a hypothetical relationship. The importance of developing such a model, however, is to allow prediction of the effectiveness of an interface whose design components and configurations are based on a 'first principle'. In that regard, Card, English, and Burr (1978) compared pointing times to textual targets of several sizes and distances among the mouse, joystick, step keys, and text keys. The mouse was the fastest input device, and when movement time was regressed onto the index of difficulty, the model (i.e., linear equation) based on Fitts' Law accounted for 66% of the variance in movement time. Other studies within this stream of research include comparisons among six cursor control devices (Epps, 1986) and among trackball, joystick, and mouse (Jones, 1991), based on Fitts' law models. MacKenzie (1992) discussed issues regarding the application of Fitts' law in human-computer interaction research, which involved a consideration of several input devices including the mouse. As windowing and icon-based information systems became widely available, comparative analyses began to be supplanted by studies of the effectiveness of the mouse, as the only input device, under varying conditions of interface design and presentation of information on the screen. Green and Barnard (1990) reported a decrease in selection time for abstract and representational icons when all icons remained in a fixed position on the screen. MacKenzie and Buxton (1992) reported that a mouse movement to a target was longer when the angle of approach was 45, in comparison with 0 and 90. An adjusted Fitts' law, reflecting considerations on the index of difficulty of target width in a two-dimensional task, accounted for 95% of the variance in movement time. Boritz, Booth, and Cowan (1991) reported differences in movement time for 'pieces' of a pie-shaped menu between 0 and 270 ; Fitts' law accounted for 97% of the variance in movement time. The research literature, then, supports the value of Fitts' law (or hypothesis) when applied to movements of the mouse pointer to a target, and some studies suggest that the angle of approach is a factor affecting movement time.

4 484 Whisenand and Emurian Since the computer user is generally faced with icons that are located at all angles and distances from a start position, it is important to extend the range of approach angles to target icons that include horizontal, vertical, and diagonal movements of the mouse under more varying target sizes and distances than have been reported previously. This study, then, focuses on the relationship between movement time and angle of approach for a mouse as the direct manipulation device for the selection of icon-like targets positioned along the axes mentioned above. The classic Fitts' factors are examined within the context of eight angles of approach, four target sizes, and five target distances. Finally, the effectiveness of Fitts' law to account for movement time is investigated under the various experimental conditions by undertaking a residuals analysis to determine the combination of factors that produces the lowest level of residuals and the optimal user performance. Subjects METHOD Nine male (mean age = 25.9 years) and 13 female (mean age = 27.5 years) undergraduate students in Information Systems participated for course credit (n = 22). All subjects reported over 30 days' experience with a personal computer (PC) and a mouse input device. Of the subjects, 21 were righthanded, and 1 was left-handed. Apparatus The experiment used an Eclipse PC 486/33 with 4M of RAM and an Orchid 280 F graphics board with 1M RAM. Microsoft's Visual Basic for Windows, 2.0, controlled the presentation of events and the recording of movement time and accuracy. The screen on the CTX Proscan 17-inch monitor was 31 x 23.5 cm (800 x 600 pixels). The background color for target selection was black, the starting block was gray, and the targets were yellow. Cursor control and target selection were controlled by a Logitech two-button serial mouse. A 'response' on the mouse was a 'click' down and up on the left button for target presentation and down alone for target selection. The subject and the experimenter (Thomas Whisenand) were located in a quiet room behind a closed door, with very light hall traffic and no phone or other interruptions. The PC was positioned on a 24 x 35 inch (60.96 x cm) desk-type work surface at a height of 26 inch (66.04 crn). All trials occurred during daylight hours, with daylight supplemented with overhead florescent lighting. Eye to screen distances were determined by the subject's

5 Angle of approach and Fitts' law 485 preference and comfort, and they ranged from 15 to 36 inch (38.10 to crn), with a mean of 22 inch (55.88 era). Procedure Participants were tested individually. Each subject was interviewed to gather demographic and experiential data. The subject read a single page of written instructions. The experimenter demonstrated and supervised a sample trial block of 16 target selections to familiarize the subject with the task and environment. Subjects were instructed to move the cursor and capture the target as quickly and accurately as possible. Subjects used their preferred hand to operate the mouse. Figure 1 presents a representation of the eight angles, four target sizes, and five target distances that were investigated. The index of difficulty (ID) is presented adjacent to or within each target. The starting block was always positioned at the center of the screen. Amplitudes (i.e., distance to the target) and angles were measured from the center of the starting block to the center of the target. Each trial presented an icon-like square target (0.25, 0.5, 1.0, or 1.5 cm/side) at a distance of 1, 2, 4, 8, or 10 cm, and at an angle of 0, 45, 90, 135, 180, 225, 270 or 315 from the starting point. Targets were yellow on a black background. Color selections were based on previous findings (Luria, Neff, & Schlichting, 1989) that color made no significant difference in search time, but a yellow on black combination was preferred by the users. This color combination is also consistent with the target/ background colors used in several previous experiments (Ostroff & Shneiderman, 1988; Sears & Shneiderman, 1991). Each subject completed four blocks of 160 trials. A block consisted of 160 randomly presented 'successful' trials representing each angle of approach, target size, and target distance combination (8 4 x 5 within-subjects design). For an individual trial, the subject positioned the cursor arrow tip within the starting block (black square in center of Figure 1) and, when ready, depressed (i.e., 'clicked' down-and-up) the mouse button; a target then appeared at its programmed angle, size, and distance. The subject then moved the cursor to the target as rapidly as possible and 'clicked' the mouse button to complete the selection. If the tip of the arrow was located within the target when the selection occurred, the subject proceeded to the next trial. If the tip of the arrow was outside the target when the selection occurred, a beep sounded, an error was recorded, and that trial remained in the pool of trials to be presented within that block. After these events, the subject proceeded to the next trial. A block of trials was not completed until all targets (160) were successfully selected. At the end of each block, the subject's performance feedback was

6 486 Whisenand and Emurian 0.5 cm [] [] 5.32 [] 0.25cm 3"~ 1 ~ [] 1.0cm O 13 [3 -~ 1.5 cm Figure 1. A representation of the eight angles, four target sizes, and five target distances investlgatad. The Index of difficulty (ID) is presented adjacent to or within each target. displayed with the number of errors and the mean movement time for that block. This use of feedback between successive blocks of trials has been suggested in similar research to enhance the motivation of subjects to perform optimally (Card et al., 1978). Between blocks, subjects were given an opportunity to comment on their interaction and to flex or stretch their arm and hand. The mean participation time for all subjects was 45 min, with a 1-min interval between blocks. The software automatically recorded the subject's identification number, trial block, target angle, target size, target distance, and movement time for every successful target selection. Movement time was measured from the time the cursor moved from the starting block to the selection of the target. This measure is identical to the procedure reported by Card et al. (1978). An error

7 Angle of approach and Fitts" law 487 consisted of a mouse 'click' when the cursor was outside the target. The response time data was accurate to 0.01 s. RESULTS The data analysis is based upon the MANOVA procedure recommended for within-subjects analyses (Maxwell & Delaney, 1992). All significant contrasts reported are significant under Bonferroni critical value corrections (p <.05). For clarity of visual inspection and interpretation, the figures connect data points with lines, even though the categories are discrete. Figure 2 presents mean movement time across the eight angles for each of the four target sizes. The overall means and standard deviations for target angle and size are also presented. The data set (n = 14,080) was reduced for multivariate analysis by averaging movement times over the five distances. MANOVA showed significant effects on movement time of target angle, F(7, 15) = 5.13, p <.005, and target size, F(3, 19) = , p <.001. MOVEMENT TIME a Z r,.) 0.8 LLI TARGET SIZE (cm) s~,e DEGREES Mean o., I i Angle Mean Figure 2. Mean movement time across the eight angles for each of the four target sizes. Overall means and standard deviations for target angle and size are also presented; H-V = horizontal and vertical angles, D = diagonal angles.

8 488 Whisenand and Emurian Figure 2 shows that mean movement time decreased as the target size increased. Pairwise contrasts showed significant differences between all pairs of the four target sizes. For angles, Figure 2 shows that movement times were generally slower for movements along the four diagonals (45, 135, 225, and 315 ), in comparison with the two horizontal (0 and 180 ) and two vertical (90 and 270 ) directions, although this tendency is graphically more robust for the overall means, in comparison with the means within the four target sizes. Although a complex contrast comparing the mean movement time between the four diagonal angles and the four other angles was not significant, F(1, 21) = 2.66, p >.05, the following pairwise contrasts were significant: 0 and 45, 45 and 180, 45 and 270, and 135 and 180. Figure 3 presents mean movement time for each of the four target sizes across the five distances. Figure 3 graphically shows that, for all target sizes, movement time increased as distance increased. The data set was reduced for multivariate analysis by averaging the movement times over the eight angles. MANOVA showed a significant effect of distance, F(4, 18) = 97.68, p <.001, and a significant distance by size interaction, F(12, 10) = 5.83, p <.005. The interaction is graphically evident in the transition in movement times between target distances of 8 and 10 cm. The increment between 8- and 1.6 MOVEMENT TIME CO t'~ 1.0 Z 0 to o.8 LLI O ,HI" I = = I I DISTANCE (cm) I TARGET SIZE (cm) :, ---t ~- - Figure 3. Mean movement time for each of the four target sizes across the five distances.

9 Angle of approach and Fitts" law cm distances appears greater for target size 0.25 cm in comparison with the increments for the remaining target sizes at those distances. This observation is confirmed by significant interaction contrasts (p <.05) for target sizes 0.25 and 1.0 cm interacting on movement time at distance pairs 1 and l0 cm, 2 and l0 cm, and 4 and l0 cm. Figure 4 presents mean movement time for all four blocks of trials across the eight angles. MANOVA showed a significant effect of block, F(3, 19) = 4.68, p <.05, and a significant block by angle interaction, F(16, 6) = 3.99, p <.05. This figure shows that, with some notable exceptions, movement times were generally faster in the later blocks of trials, and the overall means show an orderly decrease in movement times over successive blocks. A test for linear trend over blocks was significant, F(1, 21) = 14.62, p <.005. The block by angle interaction is graphically evident in the behavior observed during Blocks 2 and 3. During Block 2, movement times were roughly equivalent for angles 180, 225, 270, and 315, and during Block 3, the movement time was faster for angle 315, in comparison with Block 4. Additionally, interaction contrasts were significant (p <.05) for Blocks 1 and 3 interacting with movement times for angles 0 and 225 and angles 225 and 270. Figure 4 also graphically portrays the effect of angle on movement time, since each movement time that precedes a diagonal on the figure is MOVEMENT TIME 0.98 ~ CO 0.92 E3 Z LU " ".... ~/-.....'. _... o. "/ "'" "" i F J, i ~ I i DEGREES 0.82 BLOCKS gl-....&... " Figure 4. Mean movement time for all four blocks of trials across the eight angles.

10 490 Whisenand and Emurian almost always faster than the succeeding movement time across all blocks, with the exceptions that account for the interaction. To determine the effectiveness of a linear model, based on Fitts' law, to account for movement time in relationship to the variables under consideration in the present study, a residuals analysis was undertaken. A residual is the absolute difference between a predicted movement time (PMT) and the observed movement time (MT). The PMT, based on Fitts' law, was calculated by the following equation: PMT = (0.17)(ID) The slope and intercept parameters in the equation were obtained from the regression of observed MTs onto calculated IDs, and the parameters were significant (p <.001). The regression of MTs onto IDs yielded an R 2 of.43 (p <.001). Figure 5 presents mean residuals across the five distances for all four target sizes. Because a univariate ANOVA revealed no effect of angle on residual, the data were averaged over angle to reduce the data set size for the multivariate analysis. MANOVA showed a significant effect on residuals of distance, F(4, 18) = 40.39, p <.002, and size, F(3, 19) = 35.91, p <.002. The distance by size interaction was significant, F(12, 10) = 3.87, p <.05. Figure 5 shows that the magnitude of the residuals was generally related to the size of the target, with the smallest target having the greatest residuals 0.40 RESIDUALS I..~--I 0.30 Z 0"25 ~ e3o.=,... -m....--'"...a 0.25 ~. 4 ~ e- ~ 0.20 ~"..-- "~ I I I I I DISTANCE (cm) I TARGET SIZE (cm) = ---~... J, Figure 5. Mean residuals across the five distances for all four target sizes.

11 Angle of approach and Fitts" law 491 and the largest target having the smallest residuals across all distances. The distance by size interaction is graphically apparent in the differences among the four sizes across the five distances. Only the smallest target (0.25 cm) showed robust increases in mean residuals across the five distances, whereas the other 3 sizes showed unorderly and/or small changes in the magnitudes of increases, if any, across the five distances. Among the several significant interaction contrasts were the following: the and 1.0-cm targets interacting at the 2- and 10-cm distances, 1- and 10-cm distances, and 2- and 10-cm distances; and the and 1.0-cm targets interacting at the 1- and 10-cm distances, and 2- and 10-cm distances. Figure 6 presents the frequency of errors by all subjects across the four successive trial blocks; the means are presented in the filled bars. Mean errors over the four trial blocks decreased over each of the first three blocks and rebounded slightly in the fourth block. MANOVA showed a significant effect of block on errors, F(3, 19) = 3.53, p <.04. Pairwise contrasts showed a significant difference between Blocks 2 and 3. Significant (p <.001) intersubject product-moment correlations in error frequency were observed between successive blocks: Blocks 1 and 2, r =.69; Blocks 2 and 3, r =.89; and Blocks 3 and 4, r = ERRORS > Z LU 0 LU 8 n" LL v ~ ~ w oo SUCCESSIVE BLOCKS Figure 6. Frequency of errors by all subjects across the successive trial blocks; the means are presented in the filled bars.

12 492 Whisenand and Emurian DISCUSSION The results of this experiment show that the angle of approach affects the time required to move a cursor with a mouse from one position on a VDT screen to select a target at another position. Movements were generally faster along the two vertical and two horizontal directions in comparison with movements along the four diagonal directions. These results confirm and extend the findings of Boritz et al. (1991) who reported more rapid mousedriven cursor movements at 0, in comparison with 270, for item selection in a 'pie-shaped menu' that consisted of icon-like square targets similar to those used in the present study. The results of Boritz et al., which showed a difference in movement time only between 0 and 270 for 'small' targets close to a start block, contrast with the present findings, which showed movement time sensitivity to several pairwise comparisons across several target sizes and distances. The present results are also consistent with MacKenzie and Buxton (1992) who reported longer movement time for targets at 45, in comparison with targets at 0 and 90. Although accounting for similarities and differences in outcome among the several studies presents interpretive challenges, since there were differences in the task and selection events, taken together the studies indicate that angle of approach is a reliable factor affecting movement time. Distance to a target is included in studies of Fitts' law applications to models of movement time with a mouse because distance (i.e., amplitude) is a component in the equation for the ID. Intuitively, targets at a 'far' distance from a target should take longer to select than targets at a 'near' distance. In the present study, movement time increased as target distance increased, and decreased as target size increased. These effects are similar to those reported by Card et al. (1978) and Gillan, Holden, Adam, Rudisill, and Magee (1992) for mouse pointing time. Many studies have reported this effect, and Gillan et al. reported orderly increases in mouse movement time to a text object as distance increased for eight different horizontal sizes of text objects. The similarity in process and outcome for the distance variable contributes to the reliability and generality of other similar results across the studies. It was the case in the present study, however, that the residuals analysis showed that Fitts' law was a better predictor of movement time at the shortest distances and largest target sizes. This effect might be attributable, in part, to the manner in which subjects moved the cursor in relationship to different distances and target sizes. It was observed that many subjects made an initial downward movement away from the starting position before circling around and moving toward the target itself. Other subjects were observed initially to move the cursor in the direction of the immediately preceding target before correcting the motion in the direction of the

13 Angle of approach and Fitts' law 493 displayed target. These motions obviously complicate the application of a Fitts' law model (Fitts, 1954), which is intended to account for motions in a line along a horizontal plane, although ballistic movements of a stylus also contributed to the accuracy of the original model formulation (Gan & Hoffman, 1988). Many investigators have proposed an analysis of the components of a movement to a target, such that anticipatory and other 'preparatory' movement sequences are not included in the model (Gillan et al., 1992). An 'optimized-submovement' model, proposed by Meyer (1988), may permit a more robust linear modeling of the user's behavior than revealed in the present analysis, and Baccino and Kennedy (1995) have published a method to determine the dispersion, direction, and distance of mouse movements that may be useful within such modeling. Although the 1.5-cm target showed the lowest residual magnitude across the five distances, in comparison with all other target sizes, the magnitude of R 2 (.43) was small in comparison with other investigations. When Gillan et al. (1992) reanalyzed their data with revised hypotheses of the 'effective distance' to a text object, R 2 was.95 for a point-click mouse movement sequence. Further, Card et al. (1978) reported an R 2 of.83 for a point-dick sequence, Epps (1986) reported an R 2 of.70, and MacKenzie and Buxton (1992) reported an R 2 of.95 for an 'adjusted' target width for a point-click sequence. With particular reference to angle of approach, more recent analyses, which used a stylus task, have also emphasized the contribution of both horizontal and vertical constraint dimensions in the determination of a maximum ID (Hoffmann & Sheikh, 1994) that might best predict discrete movement time. Also considered is an 'effective shape' such that a square presented on an angle may have greater width of effectiveness across its diagonal (Sheikh & Hoffmann, 1994). Although R 2 was comparatively lower in the present study, the residuals analysis, based on the Fitts' law linear model, showed increases in residuals for increases in distance and decreases in size, and the significant distance by size interaction suggests a more subtle effect than revealed by a global R 2. The broad range of angles, sizes, and distances employed in the present study may require more practice for a user to achieve a level of performance effectiveness that would yield an R 2 similar in magnitude to those values just cited. Finally, previous investigations displayed the target prior to a subject initiating its 'capture,' thereby eliminating the necessity for locating the target on the screen prior to moving the cursor. The maximum error rate in Fitts' original reciprocal tapping task (Fitts, 1954) was 4.08% for the 1 LB stylus. Several studies (Crossman, 1960; Welford, 1960) have suggested the use of a corrected estimate of target width, which has the effect of normalizing the error rate at 4%. More recently, reported error rates for the mouse in target acquisition type tasks have been

14

15 Angle of approach and Fitts" law 495 Epps, B. W. (1986). Comparison of six cursor control devices based on Fitts' law models. Proceedings of the 30th Annual Meeting of the Human Factors Society, Santa Monica, CA, Epps, B. W., Snyder, H. L., & Muto, W. H. (1986). Comparison of six cursor control devices on a target acquisition task. Proceedings of the Society for Information Display, Fitts, P. M. (1954). The information capacity of the human motor system in controlling the amplitude of movement. Journal of Experimental Psychology, 47, Gan, K., & Hoffman, E. R. (1988). Geometrical conditions for ballistic and visually controlled movements. Ergonomics, 31, Gillan, D. J., Holden, K., Adam, S., Rudisill, M., & Magee, L. (1992). How should Fitts' law be applied to human-computer interaction? Interacting with Computers, 4, Green, A. J. K., & Barnard, P. J. (1990). Iconic interfacing: The role of icon distinctiveness and fixed or variable screen locations. Human-Computer Interaction -- INTERACT "90, Greenstein, J. S., & Arnant, L. Y. (1988). Input devices. In M. Helander (Ed.), Hana~ook of human computer interaction (pp ). Amsterdam: Elsevier Science. Hoffmann, E. R., & Sheikh, I. H. (1994). Effect of varying target height in a Fitts' movement task. Ergonomics, 36, Jones, T. (1991). An empirical study of children's use of computer pointing devices. Journal of Educational Computing Research, 7, Karat, J., McDonald, J. E., & Anderson, M. (1986). A comparison of menu selection techniques: touch, panel, mouse, and keyboard. International Journal of Man-Machine Studies, 25, Luria, S. M., Neri, D. F., & Schlichting, S. (1989). Performance and preference with various VDT phosphors. Applied Ergonomics, 20, MacKenzie, I. S. (1992). Fitts' law as a research and design tool in human-computer interaction. Human- Computer Interaction, 7, MacKenzie, I. S., & Buxton, W. (1992). Extending Fitts' law to two-dimensional tasks. Proceedings of the CHI'92 Conference on Human Factors in Computing Systems, MacKenzie, I. S., Sellen, A., & Buxton, W. (1991). A comparison of input devices in elemental pointing and dragging tasks. Proceedings of the CHI'91 Conference on Human Factors in Computing Systems, Maxwell, H. D., & Delaney, H. D. (1992). Designing experiments and analyzing data: A model comparison perspective. Belmont, CA: Wadsworth. Meyer, D. E. (1988). Optimality in human motor performance: Ideal control of rapid aimed movements. Psychological Review, 95, Ostroff, D., & Shneiderman, B. (1988). Selection devices for users of an electronic encyclopedia: An empirical comparison of four possibilities. Information Processing & Management, 24, Perry, T.S., & Voelcker, J. (1989, September). Of mice and menus: Designing the user-friendly interface. IEEE Spectrum, pp Sears, A., & Shneiderman, B. (1991). High precision touchscreens: Design strategies and comparisons with a mouse. International Journal of Man-Machine Studies, 34, Sheikh, I. H., & Hoffmann, E. R. (1994). Effect of target shape on movement time in a Fitts task. Ergonomics, 37, Welford, A. T. (1960). The measurement of sensory-motor performance: Survey and reappraisal of twelve years' progress. Ergonomics, 3,

16 本文献由 学霸图书馆 - 文献云下载 收集自网络, 仅供学习交流使用 学霸图书馆 ( 是一个 整合众多图书馆数据库资源, 提供一站式文献检索和下载服务 的 24 小时在线不限 IP 图书馆 图书馆致力于便利 促进学习与科研, 提供最强文献下载服务 图书馆导航 : 图书馆首页文献云下载图书馆入口外文数据库大全疑难文献辅助工具