Target Size Study for One-Handed Thumb Use on Small Touchscreen Devices Anna Sergeeva University Bremen Address sergeeva055@gmail.com Optional phone number ABSTRACT
In our experiment we replicated a study by Pekka Parhi, Amy K. Karlson, Benjamin B. Bederson from 2006 [1] to see if its results are still valid today after mobile interfaces evolved over eight years. We found that nowadays users are more experienced with touchscreen devices and use them intuitively. The results defined new minimum size of 6.8 mm for screen with size of 58.6 mm by 115.2 mm, that is useful for most common gestures with a one-hand use of device. In comparison, it was 9.2 mm for discrete tasks and targets of 9.6 mm for serial tasks in 2006. Keywords
one-handed thumb use; touchscreen devices; button size; mobile; target size; keyboard; layout; HCI. INTRODUCTION
Nowadays about 1 700 million people use smartphones. In 2014 usage of mobile phones exceeded PC internet usage for the first time. [6]
Denitsa Todorova University of the Arts Bremen Address deny.todorova@gmail.com Optional phone number It changes the way people interact with devices: they often do it staying in public transport, walking, driving, holding other objects in hands, etc. In this situation most of the time users have only one hand free to operate mobile devices. [3] In the previous study from by Pekka Parhi, Amy K. Karlson, Benjamin B. Bederson [1] a different element size for mobile touch screens was used. The aim of the test was to find the optimal sizes of buttons when people use a touch screen mobile device using one hand. Participants received two tasks. The first one was a discrete target task - it explored single button tasks such as activating buttons, radio buttons or checkboxes. The second one was a serial target task - exploring sequences of taps used, for example, in text entry. Since holding a device in one hand constrains thumb movement, the target positions were varied and user performance depended on screen location was also tested. The results showed that speed improved as targets grew, but there was no significant difference in error rate between target sizes from 9.6 mm in discrete tasks and targets from 7.7 mm in serial tasks. As a conclusion based on the subjective ratings and the findings on hit response variability the authors recommend target size of 9.2 mm for discrete tasks and of 9.6 mm for serial tasks. OUR HYPOTHESIS
The raise of touch-screen devices’ performance, and their wide spread among contemporary population has changed the feelings of users about comfortable element sizes since 2006 to 2014. METHOD Participants
Figure 1. Mobile Exceeds PC Internet Usage for First Time in History in the beginning of 2014 [6]
Twenty participants (9 male, 11 female) were recruited personally and via internet. The announced restriction was that people have to be right-handed users. The age of the participants varied between 22 and 29 years, with a mean of 25.6 years. Differences with previous research: all the participants are experienced with touchscreen devices. As in the previous test, participants were asked to rate how often they interact with touchscreen devices (from “never”
to “every hour”). 15 people use smartphone every hour and more often, 4 participants use smartphone several times per day and 1 person uses smartphone once a month (in comparison to 2006, with 18 participants experienced only with keypad-based handhelds, and only 5 users with experience with touchscreen-based handhelds). 12 participants favor one-handed use of the device, 5 users prefer both handed use and 2 people prefer both way of use. Two-handed use is practiced mostly for using web-browser and entering text. Participants received small gifts for their time. For each participant we measured hand width and thumb length. Thumb length varied between 90 and 135 mm (μ=108.35 mm, σ = 9.38), and hand width varied between 80 and 100 mm (μ=86.4 mm, σ = 0.027).
Apparatus
In order to test buttons on a contemporary device the experiments were performed on one of the most popular smartphones in 2014, Apple iPhone 4S. [5] It has screen resolution of 640×960 and dimensions of 115.2 x 58.6 x 9.3 mm.
session time, including instruction, data collection and questionnaire was approximately 10-15 minutes. Discrete targets task
The main objective of this phase (phase 1) was to determine the most comfortable target size for contemporary mobile devices. That helps to define recommendations for situations in which single-target interactions appear, such as single buttons, checkboxes, radio buttons. We tested 5 button sizes: 3.8 mm, 5.8 mm, 6,8 mm, 7.7 mm, 9.6 mm. To test the offset with which right-handed people tap and to find in which corner it is easier to tap the screen was divided into 3x3 sections which makes 9 places in the center of which we positioned our targets. This means that 5 x 9 different screens were tested. To test how precisely the participants tap on the right target (the one marked with “x”) the target was surrounded by other buttons (marked with letters from ”a” to “j”). After tapping on the right place the button provided visual feedback (changed its color) and the screen was changed to the next one.
The study interface and control interface was developed using Solidify prototyping tool [2].
Procedure
For the experience the same testing method was used as in the experiment from 2006 [1]. The study was equipped with a mobile application prototype imitating different size keyboards located at the different parts of the screen. Keyboards represented 4 original button sizes from previous research [1] 3.8 mm, 5.8 mm, 7.7 mm. 9.6 mm, and in this experiment we added one new size - the recommended size for Apple-devices [4] – 6.8 mm. The study was divided into two parts. In the first part users were asked to do the discrete target test and in the second part - the serial target test. During both phases participants were asked to follow condition to use phone with their right hand and the thumbfinger only. After completing the test participants were asked to fill in a questionnaire of demographic data and device use, as well as their subjective rating of the interaction experience. Performance was assessed by statistical data collected by the application (time, spent on each page, numbers of clicks per page, accuracy of clicks and heat maps). The total
Figure 2. Two screens from the discrete target task (with grid shown on top). For each of the button sizes the screen was divided into 3x3 areas and we tested every button size in the middle of each of those 9 rectangles.
Serial targets task
The aim of this phase (phase 2) was to determine the most comfortable target size for serial target tasks. Such situations could be typing on a keyboard or entering a sequence of numbers. The same 5 button sizes were tested: 3.8 mm, 5.8 mm, 6.8 mm, 7.7 mm, 9.6 mm. The screen was divided into 2x2 sections in which the keypad was placed. Very close to it was put the sequence of four numbers which had to be typed and the feedback of the already chosen buttons. Also “End� button was placed close to the numbers, it had to be pressed at the end of the sequence which moved the participant to the next button size and keyboard position. Again, like in the previous task, every button gave visual feedback when it was successfully tapped. The total amount of screens was 4 x 5 x 4 different screens for this task.
Figure 4. Mean total time between the release of one button tapping the following one for each target size in the discrete target phase. Number of errors for each target size in the discrete target phase. b.
Figure 3. Two screens from the serial target task (with grid shown on top). For each of the button sizes the screen was divided into 2x2 areas in which the keyboards were placed. Very close to every keyboard was the sequence of the 4 numbers that had to be tapped.
Discrete Task Percent Error
After the test we also observed where exactly errors happened. We studied if changing the button size affects the error rate but with the discrete target task the results showed that all mistakes happened when the layout was changed, not when size was changed.
RESULTS a.
Discrete target results
Task time was analyzed with one way analysis of variance (ANOVA) with factors of target size 3.8, 5.8, 6.8, 7.7 and 9.6 mm. The results showed 1% level of statistical significance. A surprising effect of target size, (F(4,40) = 3.49, p = 0.0154) was observed. Unlike in the previous research the time for reaction with the increasing button size does not get lower. The important effect is that error rate gets significantly lower with the bigger sized buttons. With buttons sizes above 7.7 mm the error rate almost stays the same, on average below 3.33%. We relate this result to the fact that people need to adjust themselves to the new keyboard layout. Another noticeable fact is that people make mistakes with getting a new layout. That means people have particular expectations related to a previous position of buttons when the screen is changed.
Figure 5. No significant offset was observed to be caused by the right-handed use of the device.
No significant offset was observed to be caused by the right-handed use of the device. This leads to the conclusion that the users felt equally comfortable touching any of the screen areas.
d.
Serial Task Percent Error
After the test participants were asked which was the smallest of the 5 keypad sizes they felt most comfortable. For the discrete target task 10 participants voted for the size of 5.8 mm, 5 preferred size of 6.8 mm, 3 voted for 3.8 mm size, and sizes of 7.7 and 9.6 mm got one vote each. The highest amount of mistakes was observed with the two smallest sizes (3.8 mm and 5.8 mm). Despite most of participants voted for the size of 5.8 mm, this size still has a significant rate of mistakes and we consider it as not completely comfortable. c.
Figure 7. Number of errors for each target size in the serial target phase.
Serial target results
The results from the serial task showed higher significance level than in the discrete target tasks, (F(4,95) = 6.37, p = 0,0001). Generally, errors declined as key size increased. The keypad with the smallest key sizes (3.8 mm) caused significantly more errors to be made than those with key sizes ≼ 5.8 mm. No differences between error rates for the other key sizes were significant. Very similar were the results about mean total time, where the most significant difference was the highest time for the smallest target size (3.8 mm) and no significant difference for the buttons above 5.8 mm was observed.
Figure 8. Difference between error rates in the smallest and biggest button size. As with single targets, tapping on serial targets shows no significant offset due to the right-handed usage of the device.
Figure 6. Mean total time between the release of one button tapping the following one for each target size in the serial target phase.
As we can see in Figure 6, the time needed for tapping sequence of buttons decreased with increased button sizes. As well, the amount of rate decreased as the key sizes, and thus total keypad size, grew. As in the previous part, change of a layout in a sequence-target task creates more mistakes.
Participants were asked which was the smallest of the 5 keypad sizes they felt most comfortable. Most of the testes voted for the size of 6.8 mm (11 people), the second comfortable size was 5.8 mm (4 people), 2 people voted for the size of 7.7 mm, 2 people chose size of 3.8 mm, and only 1 person prefers the biggest size of 9.6 mm. Even though the error rates and mean time per screen stop decreasing significantly above 5.8 mm, actually most of the participants felt comfortable with at least 6.8 mm targets. We interpret here that using a sequence-tap task people need bigger button sizes due to their position close to each other. Also, the biggest amount of mistakes was done with the 2 smallest sizes. DISCUSSION
The results of our research shows that user preferences related to element size have changed since 2006. In 2006, when the smallest recommended button size was 9.2 mm for single-target tasks and 9.6 mm for serial-target tasks,
now people feel and perform well with buttons with 5.8 and 6.8 mm sizes. We could speculate what would be the reasons for the participants to work easier with smaller buttons in 2015. We suggest two reasons: on one side, users nowadays are more experienced than before. All the participants of our experiment own smartphones with touchscreens and use them at least several times per day. The second reason could be the increased screen density in the screens nowadays. The iPhone 4S which we used has a 326 PPI (128 PPCM) resolution and the HP iPAQ h4155 has a screen with 114.2 PPI only. We also relate our findings to the fact that users are more precise when they have only one target for tapping. In a sequence task people need bigger buttons due to the position of elements near each other. That is why here participants voted for the size of 6.8 mm. Our recommendations might be different for touchscreen devices with screen size different from 115.2 mm by 58.6 mm, and screen resolution different from 960 by 640 pixel. CONCLUSION
In our aim to check if user preferences changed since 2006 we replicated a two-part study on interaction between target size and task performance in single- and multi-target tasks. We tried to determine optimal target sizes for one-handed use of touchscreen devices and wanted to compare the results with a previous study conducted in 2006. Our study was divided into two parts - we studied in detail the relation between target size and task performance in single and multi-target tasks. Based on the results of this experiment we would recommend 6.8 mm minimum button size both for discrete tasks and serial tasks as the smallest targets with which the users can feel comfortable and perform well.
ACKNOWLEDGEMENTS
We thank all the participants from University of the Arts Bremen and University Bremen that took part in our experiment and our teachers, Frederik Pollman and Rainer Malaka from University Bremen, who provided helpful comments on conducting this experiment and a place for testing. REFERENCES
1. Pekka Parhi, Amy K. Karlson, Benjamin B. Bederson Target Size Study for One-Handed Thumb Use on Small Touchscreen Devices. MobileHCI'06, September 12–15, 2006, Helsinki, Finland. 2. Solidify is a tool for creating clickable prototypes, validating user flows on any device by performing user tests in person, remotely or on testers own time to get the needed feedback. http://solidifyapp.com/ 3. Antti Oulasvirta, Sakari Tamminen, Virpi Roto, and Jaana Kuorelahti Interaction in 4-Second Bursts: The Fragmented Nature of Attentional Resources in Mobile HCI. CHI 2005, April 2–7, 2005, Portland, Oregon, USA. 4. iOS Human Interface Guidelines https://developer.apple.com/. 5. Web Trends Report Q2 2014, http://www.netbiscuits.com/. 6.comScore Media Metrix Multi-Platform & Mobile Metrix, U.S., March 2013 June 2014. http://www.comscore.com/Products/AudienceAnalytics/Media-Metrix-Multi-Platform