Exploring Fitts’ Law: An Evaluation of Pointing Devices Chen-Che (Eric) Ma University of Washington Information School Seattle, WA [email protected]

Jimmy Bao Nguyen University of Washington Information School Seattle, WA [email protected]

ABSTRACT

This case study evaluates three pointing devices using Fitts’ Law in an attempt to determine which device is superior in terms of throughput. Throughput is measured in bits/second, and is calculated by using either the average Index of Difficulty (ID) divided by the average Movement Time (MT), or 1 divided the constant ​b​. The three devices that were used in this study were a Logitech MX Master Mouse, an HTC Vive headset and controller, and Xbox 360 controller. Using the FittsStudy software, six participants completed eighteen point-and-click task trials on each of the three devices. Following the tests, we found that the mouse performed the best, while the Xbox 360 controller performed the worst. Author Keywords

Fitts’ Law; HTC Vive; Input Devices; Mouse; Xbox ACM Classification Keywords

Pointing Devices, Pointing, Virtual Reality INTRODUCTION

Developed in 1954 by Paul Fitts, Fitts’ Law models throughput in aimed movements [1]. It is very powerful due to its ability to provide simple comparisons. The equation itself reads as the following: MT = a + b log2(A / W + 1), where MT is Movement Time, ​a and ​b are constants, ​A ​is the distance from the target, and ​W is the width of the target. In general, Fitts’ Law shows that Movement Time can be decreased by decreasing the distance from the target, or increasing the width of the target. The purpose of this study was to evaluate three different pointing devices in the context of Fitts’ Law in order to determine which device is superior. The devices we used were a Logitech MX Master Mouse, an HTC Vive headset and an Xbox 360 controller. The Logitech MX Master is a Bluetooth mouse that operates similarly to the standard mouse. The HTC Vive uses a headset to produce a virtual reality environment in which the computer screen can be mirrored into. It utilizes a motion controller that uses ray casting in order to point, and a trigger button to click. The Xbox 360 controller, while primarily used for playing video games, can also control the

Kendall Reonal University of Washington Information School Seattle, WA [email protected]

Rebecca Ta University of Washington Information School Seattle, WA [email protected]

cursor on a computer screen with a third-party program. METHOD Participants

Six people participated in the study. The average age of the participants is 22.667 years (s=1.03). Four of the participants were male, while the other two were female. All of the participants use a mouse on a daily basis, but their level of experience with the other devices varied. Apparatus

To run the tests, we used the FittsStudy software. FittsStudy was developed by Dr. Jacob Wobbrock for conducting pointing tests. The devices we tested were a Logitech MX Mouse, an HTC Vive headset and controller, and an Xbox 360 controller. We used SteamVR to control the desktop cursor with the HTC Vive, and the Keysticks software to control the desktop cursor with the Xbox 360 controller. All of the tests were conducted in the University of Washington Information School VR Lab, using a desktop computer. The specifications of the computer are the following: Windows 10 Enterprise, Intel i5-6600 3.30 GHz, 16gb RAM, Nvidia GeForce GTX 1070, and a Dell 27’’ 2560x1440 IPS Monitor. Procedure

One participant was tested at a time. Before each test, participants were briefed on the purpose of the study. They were told that the pointing devices were being tested, and not them. They were then introduced to the FittsStudy software, and we demonstrated a few trials as examples in to help the participants become more familiar with the program and their tasks. The FittsStudy software uses two ribbon targets for point-and-click tasks. The target that needs to be clicked is highlighted in blue, and alternates between the two targets twenty-three times per trial. Each trial produces targets with a different width and amplitude. After 414 trials per participant, the test ends. For the Xbox 360 controller and the HTC Vive, participants were instructed beforehand on how to use the devices to point and click. The Xbox 360 controller used the right joystick to move the cursor, and made a clicking selection with the left trigger. The HTC Vive, on the other hand, used the headset to view the computer desktop in a virtual

environment, and a single controller to control pointing and clicking. Before beginning the tests, participants were also given the opportunity to practice interacting with the software using the devices. This was important to make sure that each participant had fair results since several of the participants had not used an Xbox 360 controller or HTC Vive before. Afterwards, the tests began. The order of the device was chosen at random with a die. Rolling a one or two would start the test with the mouse, three or four would start the test with the Xbox 360 controller, and five or six would start the test with the HTC Vive. After the first device was tested, the last two devices are assigned even or odd numbers. Since Fitts’ Law assumes a ~4% error rate, participants were told to aim for cumulative error rate around 4%. After each trial, FittsStudy displays the cumulative error rate in red. If participants had an error rate below 4%, they were asked to speed up. If they had an error rate higher than 4%, they were asked to slow down. Participants completed 414 trials per device. After completing 414 trials for a single device, the participants would move on to the next device. Since there were three devices in our study, each user completed 1242 trials. With six participants, we collected 7452 total trials. Of the twenty-three clicks per trial, the first three clicks were marked as practice clicks, so a total of 972 clicks did not count in our analysis. That resulted in a grand total of 360 counted trials per device for each participant, and 1080 counted clicks per participant. 6480 total trials were used in our analysis.

Figure 1. Average movement time with standard deviations between the Logitech MX Master, HTC Vive, and Xbox 360 controller Error Rate

The mouse had an average error rate of 5.0926% with ​±1 standard deviation of 7.0063%. The HTC Vive controller had an average error rate of 5.4167% with ​±1 standard deviation of 6.5910%. Lastly, the Xbox 360 controller had an average ​error rate ​of 5.1528% ​with ​±1 ​standard deviation of 6.5914%.

After completing the full test, the participants were thanked and released. RESULTS Movement Time

The mouse had an average movement time of 710.23ms with ​±1 ​standard deviation of 295.8233ms. The HTC Vive controller had an average movement time of 942.668ms with ​±1 standard deviation of 594.6712. Lastly, the Xbox 360 controller had an average movement time of 1669.0074ms with ±1 ​standard deviation of 1510.3457ms.

Figure 2. Average error with standard deviations between the Logitech MX Master, HTC Vive, and Xbox 360 controller Throughput

Tables 1-3 contains the data for calculating throughput in two different ways: ID ₑ / M T [2], and ​1/b ​[3]. Under both of these throughput calculations, the mouse had the highest throughput at 4.8881 bits/s with ​±1 ​standard deviation of 0.2750 bits/s with the first equation, and at 5.4027 bits/s with ​±1 ​standard deviation of 0.7326 bits/s with the second equation. The HTC Vive controller had the second highest throughput at 4.4888 bits/s with ​±1 ​standard deviation of 0.5022 bits/s, and at 2.3697 bits/s with ​±1 ​standard deviation of 0.5582 bits/s. The Xbox 360 controller had the lowest throughput at

2.2605 bits/s with ​±1 ​standard deviation of 0.5255 bits/s, and at 1.6823 bits/s with ​±1 ​standard deviation of also 0.5582 bits/s.

5

-553.76

367.0653

0.7999

5.3146

2.7243

6

-771.0322

482.3333

0.9219

3.984

2.0733

Mean

-696.0465

439.9381

0.8542

4.4888

2.3697

StdDev

267.1413

92.5868

0.0457

0.5022

0.5582

Table 2. Data from the Fitts’ Study software for the HTC Vive

Figure 3. Average throughput with standard deviations between the Logitech MX Master, HTC Vive, and Xbox 360 controller Participan t

a

1

92.0595

2

b

Pearson

TP

TP

r

IDe/MT

1/b

188.544 6

0.9326

4.6497

5.3038

48.7084

179.777 4

0.9718

5.1547

5.5624

3

91.7149

168.296 4

0.9246

5.0884

5.9419

4

-28.8184

228.820 2

0.9784

4.5633

4.3702

118.746 4

152.667 6

0.9579

5.2326

6.5502

6

15.7834

213.329 9

0.9074

4.6400

4.6876

Mean

56.3657

188.572 7

0.9455

4.8881

5.4027

StdDev

50.6242

25.8536

0.0258

0.2750

0.7326

Table 1. Data from the Fitts’ Study software for the mouse a

1

-618.2248

2

b

a

1

-1186.326 3

2

b

Pearson

TP

TP

r

IDe/MT

1/b

1078.012 4

0.8542

1.568

0.927 6

-477.966

493.901

0.9035

3.0479

2.024 7

3

-410.0996

592.8512

0.9381

2.2723

1.686 8

4

-68.209

403.2957

0.8962

2.6786

2.479 6

5

-206.9798

506.9116

0.9447

2.3475

1.972 7

6

-1053.581 2

997.8305

0.8614

1.6484

1.002 2

Mean

-567.1937

678.8004

0.8997

2.2605

1.682 3

StdDev

414.5837

260.8237

0.0343

0.5255

0.558 2

Table 3. Data from Fitts’ Study for the Xbox 360 controller

5

Participan t

Participan t

Pearson

TP

TP

r

IDe/MT

1/b

424.5737

0.8747

4.1866

2.3553

-1017.2317

528.4638

0.8239

4.7055

1.8923

3

-286.8931

302.9028

0.8243

4.6523

3.3014

4

-929.1374

534.2898

0.8802

4.0898

1.8716

DISCUSSION

Of the three devices we tested, the Logitech MX Master Mouse was the best across each category. It had the highest throughput with a low standard deviation, which suggests that it is the most efficient at selection tasks. Not surprisingly, the Xbox 360 controller had the worst performance with the highest error rates, longest average movement time, and lowest throughput. This was expected because out of the three devices, the Xbox 360 controller offers the least control over the cursor’s movements. Users often experienced difficulty controlling slight movements. All of the participants frequently overshot the ribbon and constantly had to adjust the cursor in order to make an accurate selection. However, it may be possible to achieve a better performance if we lower the sensitivity of the controller, which was set to the default sensitivity of Keysticks during testing. Unexpectedly, the HTC Vive performed almost as well as the mouse. It had similar movement times and throughput, but a slightly higher average error rate than both the mouse

and the Xbox 360 controller. The good performance could possibly be attributed to our procedure in testing the HTC Vive. Ray casting is not very stable. The HTC Vive controller moves frequently when being used, especially when pressing a button. It can be very difficult to select the ribbon when it has a small width because the ray cast’s contact point covers more than the whole ribbon on the virtual desktop screen. However, we had our participants rest the controller on their laps, using their hands to move the controller. If we forced our participants to hold the controller straight out, the results could have been worse due to fatigue. They were also told to stay seated in the same place—about 7 feet away from the virtual desktop screen—without moving forward or backwards, which helps ensure consistency between trials. Nonetheless, the challenges with ray casting may have contributed to the higher error rates observed for the HTC Vive—even with our testing technique. For further testing, it would be interesting to investigate how different screen sizes and distances of the virtual desktop screen affects pointing tasks. These results were highly favorable to the mouse, implying that the mouse is a superior pointing device. This may be due to a variety of reasons. All participants were regular mouse users; some had little or no experience with the other devices. This greatly decreases the level of difficulty in using the mouse. Compared to the two other devices, there is less mental preparation to transition from the intended motion of the cursor into the physical movement of the hands. In addition, the mouse is incredibly user-friendly for pointing tasks. Unlike the Xbox 360 controller, rapid movements of the cursor can be accomplished with precision. From our findings, the mouse was the most efficient device, and the Xbox 360 controller the worst performing. These results justify the mouse’s continued use today, and also suggest that the other controllers need further development to better support input. CONCLUSION

We have presented a study of Fitts’ Law through evaluating three different pointing devices. We have also presented a method to calculate the superiority of pointing devices using Fitts’ Law; it takes into consideration of different variables such as movement time, error rate, and throughput. From our study, we have shown pointing devices that use joysticks such as Xbox controllers are inferior to other pointing devices. At the same time, we also have shown that a mouse is the most superior pointing device between the tested devices because of its shorter movement time and higher throughput value. Lastly, we have shown that ray casting in a virtual environment may

be an acceptable method of controlling a desktop cursor under certain conditions. We hope that from the result of this case study, people will not take the mouse for granted and understand that in comparison to other devices, the mouse is still one of the most efficient pointing devices ever invented. ACKNOWLEDGEMENTS

We would like to give thanks to the volunteers of our study, to the University of Washington Information School for letting us borrow the virtual reality equipment, to our instructor Martez Mott for the lectures on Fitts’ Law, and to our teaching assistant Abdullah Ali for grading our in-class quizzes about Fitts’ Law. REFERENCES

1.

M. Paul Fitts. The information capacity of the human motor system in controlling the amplitude of movement, Journal of Experimental Psychology, v.47 n.6, p.381-397, June 1954 from http://dx.doi.org/10.1037/h0055392

2.

R. William Soukoreff, I. Scott MacKenzie, Towards a standard for pointing device evaluation, perspectives on 27 years of Fitts' law research in HCI, International Journal of Human-Computer Studies, v.61 n.6, p.751-789, December 2004 [doi>10.1016/j.ijhcs.2004.09.001]

3.

Shumin Zhai, Characterizing computer input with Fitts' law parameters: the information and non-information aspects of pointing, International Journal of Human-Computer Studies, v.61 n.6, p.791-809, December 2004 [doi>10.1016/j.ijhcs.2004.09.006]