Effects of Scrolling and Display Size on Working Memory Performance

(1)

Effects of Scrolling and Display Size on

Working Memory Performance

by Ruben ten Hove University of Groningen Faculty of Economics and Business MSc Technology and Operations Management

July 2013

Lijnbaanstraat 22d 9711 RV Groningen r.h.ten.hove@student.rug.nl

(2)

Abstract

Increasing data in- and outflow in organisations, due to uncertainty, seems inevitable. This increase in information processing requires higher efficiency in desktop workplaces. Working memory performance was compared between

a large display and a small display which required scrolling. A quasi-experiment was conducted on 34 students who played an adaptation of the game Memory twice. A NASA Task Load Index based questionnaire, adapted

as a Raw TLX, asked participants after each game about their task load experience. Results showed that the large display showed an 11% lower

average time between clicks. Furthermore, the questionnaire showed a reduction in task load when using a large display. Other measured variables are found insignificant and are discussed. Managerial implications are that the

use of large displays can be recommended for their positive effects. Further research opportunities are using touch interaction or a wide range of display

(3)

1. Introduction

Increasing data in- and outflow in organisations seems inevitable, which directly affects its employees. Global supply chain integration, competition playing field expansion, intercultural relationships, etc. increase uncertainty (Matanda & Freeman, 2009). As

uncertainty increases, more information requires processing, which demands that to achieve the same level of performance with the same amount of labour, the used tools need to become more efficient (Galbraith, 1974). This broad and pervasive progression is prevalent in modern businesses, demanding changes in the internal systems of organisations (Daft & Lengel, 1986). Research into specific aspects such as workspace design supports such changes as these may provide the required improvements sought for. This study focuses on workspaces using a display for information retrieval.

Displays are key in human-computer interaction as without, information from and about a system would be indirect (for example, unexplained noise in a machine without an error message displaying the source) and would require unnecessary interpretation. Their main function is thus providing information to the user of the system. Next to its

(6)

Many interface designs are still limited to these innovations, which are part of the “desktop” paradigm, which in turn is based on the WIMP (Windows, Icons, Menus and Pointing devices) paradigm (Cantoni, Cellario, & Porta, 2004). The WIMP paradigm requires interfaces to consist of windows as information

containers, icons as interaction elements, menus for structure and a pointer to complement the keyboard. This paradigm provides a stable face to computing, but cannot scale to the myriad of computer applications and form factors in the future (Turk & Robertson, 2000).

As technology progresses, the size constraint becomes less of a problem. Displays become cheaper, larger and allow higher pixel density. Increasing the size of a display, and thus effectively allowing more information to be displayed with the same density, has implications for the WIMP paradigm. Appending display size increase to the WIMP paradigm allows to mitigate one of the key constraints in human-computer interfaces. This constraint is that many applications, especially when required simultaneously, do not fit on a single display (Henderson & Card, 1986). This addition allows to a certain extent, making the requirement of paging, using windows or using scrolling obsolete. Having all information directly visible greatly reduces the requirement of strategically and efficiently placing documents and structuring the accessibility of applications. This reduction is one step towards the natural interaction paradigm, where instead of devices such as keyboards and mice, human cognition is the tool for human-computer interaction (Dam, 2000) (Cantoni et al., 2004). Natural interaction is further explained in paragraph 2.1.

There are two ways to increase the amount of information space directly available on a display. One is to increase pixel density, the other is to increase the physical size of the

(7)

display, effectively making it larger. Increasing pixel density allows more data and fine details to be visible at once. Less virtual navigation (e.g. scrolling) is required between information. Downsides to this are that more physical navigation may be required (e.g. moving the mouse) because displays reach beyond human visual acuity and users need to pan and zoom in to be able to distinguish all information. This may result in frustration and disorientation making it an ineffective change (Ball & North, 2005). An advantage of increasing display size is that it allows to retain pixel density, while still allowing more information to be displayed at once. Prior research has demonstrated that larger displays improve 3D spatial navigation (Czerwinski, Tan, & Robertson, 2002) (Tan, Gergle, Scupelli, & Pausch, 2004) (Tan, Gergle, Scupelli, & Pausch, 2003) and 2D data navigation (Ball & North, 2005). These studies increased display size and resolution, without explicitly specifying if pixel density remained the same or played a role in the study. Furthermore, these studies did not determine what role display size change had in the found performance difference. To discover the specific role of display size in performance, this study focuses on increasing display size, while keeping the pixel density and information density the same. A fundamental research question is derived from this theoretical gap:

How does the physical change of display size affect performance?

(8)

completed without the involvement of working memory, making it a critical component of cognition (Jonides et al., 2008). Its importance is shown by this role, but this importance is hardly found in current literature where the role of working memory in display size is tested. This study focuses directly on fundamental this human cognition. Working memory

encapsulates short-term memory storage as well as processing of information for complex cognitive tasks (Baddeley, 1992). This is further explained in paragraph 2.2. This study addresses the lack of focus on working memory performance, expanding the previous research question:

How does the physical change of display size affect working memory performance?

This research question will be fully operationalized in the hypothesis, which is the fundament of the quasi-experiment conducted to answer the question. The hypothesis is construed at the end of Chapter 2.

Research into the fundamentals of display size dates back to the 1960s. Studies were conducted using tachistoscopes. These attempted to discover the effect on working memory of display size by varying the amount of shown elements and having participants memorize a short flash of information (Estes & Wessel, 1966) (Estes & Taylor, 1966). These studies are notable for their simplicity in testing fundamental human cognitions. In later decades, display size research focused on the effect on other specific human cognitions, by varying character density and measuring legibility (Duchnicky & Kolers, 1983).

Contemporary technology consists of displays in sizes ranging from a 2 centimetre versatile wristwatch to beamers having a tolerable quality projection size of 20 meters diagonally1. This suggests that many people now know of, and part of those people are used to various display sizes. Observing this in conjunction with the 1980s revolution of the

1_{Based on current consumer product availability}

(9)

“desktop” paradigm, implies that the target audience of computing has broadened over the past decades. This paradigm made the computer accessible to non-experts and usable for many kinds of purposes (Cantoni et al., 2004). Furthermore, this audience has undergone many changes in interaction and information processing in computing. As the experience with and purpose and technology of displays changes over decades, research into the effects of display size changes requires a revision.

This study combines a more extensive version of the previously discussed

(10)

2. Literature Review

This section provides the theoretical foundation of this study. It reviews current literature on relevant subjects and clarifies the gap found in literature this study aims to fill. First, existing literature on natural interaction is reviewed. Second, working memory is discussed. Third, literature on physical display size differences is examined. Fourth, the currently researched influence and effects of information movement in the form of scrolling is contested. Fifth, user task load experience is examined. Sixth, predicted effects are

explained. Seventh, and last, the hypotheses is construed, creating a fundament for the experiment.

2.1. Natural Interaction

Talking, using gestures, shifting their gaze and listening (simultaneously) is what humans consider as natural interaction (Dam, 2000). Natural interaction in computing is an approach to interaction that has often been argued as beneficial compared to indirect

interaction such as using a mouse for pointing and a keyboard for textual input (Forlines, Wigdor, Shen, & Balakrishnan, 2007). This becomes more apparent as natural interaction in computing is becoming more commonplace. A

sub-trend of this is the increase in display size (Dam, 2000), meaning that for interaction to become ‘natural’, display of information needs to become larger. ‘Large(r)’ is not defined with specific dimensions because the relative difference is studied. Optimal display size for natural

interaction is dependent on the environment, and is outside the scope of this study.

The familiar desktop computer interface, based on the WIMP paradigm, is ill-suited for large display sizes. Such interfaces contain interface elements requiring too much spatial

Figure 2. Large display in a

(11)

movement (using a pointer) to interact and rely on the keyboard for rapid interaction. Even though Graphical User Interfaces (GUIs) are very powerful due to their complexity and possibilities, the interaction style contrasts with the natural/casual style of interaction on larger displays (Guimbretière, Stone, & Winograd, 2001).

One of the main characteristics of an effective natural interface is reduction of cognitive load. This is possible by reducing the amount and complexity of interface and interaction elements, bringing human-computer interaction models closer to the user. Such direct manipulation paradigms have become common, and are experienced as natural, in the form of iconic interfaces (Cantoni et al., 2004). This allows for a straightforward

interpretation of the perceived interface and helps people to rely on interaction modalities they are used to, such as touch, speech, gestures, etc. (Valli, 2008). Furthermore, a transfer from indirect to direct (natural) interaction requires little translation in current interfaces since interactions are essentially the same (there was a click/touch at a certain location) (Valli, 2008).

The change from indirect to natural interaction seems the way to go when observing these advantages, but evidence exists that task input performance is worse when using direct (natural) input compared to using indirect input, when just a single point of interaction is required (Forlines et al., 2007). This means that when making successive interactions (not simultaneous), a mouse has shown to perform better. Missing in the study by Forlines et al. (2007) is that they do not take into account that this may not be the case when using a larger display. As a step towards determining the role of larger displays within natural interaction, the performance of an important human cognition is studied, while requiring a single point of indirect input. This human cognition is studied for its substantial role in information

(12)

2.2. Working Memory

Many studies have attempted to discover the roles, functions, limitations and possibilities of short-term and long-term memory. An important paradigm-shift occurred when the dichotomy of short-term and long-term memory was negated, indicating a continuum in memory storage and processing (Melton, 1963). This view on memory

functionality was adopted with the term working memory, which encapsulates a brain system providing temporary memory storage as well as processing of information for complex cognitive tasks (Baddeley, 1992).

Working memory has limitations, which has consequences for how these should be accounted for. When tasked with storing a single unit or chunk of information in working memory, to be recalled within a short period of time, similar information input (interference) causes forgetting of the initially stored information (Atkinson & Shiffrin, 1971), especially when the cognitive constructs overlap (Borst, Taatgen, & van Rijn, 2010). This is seen as one of two main reasons for forgetting. The second is decay, where as time passes, information in memory erodes and becomes less available for later retrieval. Decay is found to be a

controversial explanation for forgetting. Most studies on this subject fail to show how decay occurs or explain decay with observations resembling interference. Jonides et al. (2008) argue that random interference can be interpreted as decay. Due to the smooth addition of stochastic noise into the information to be remembered, this information loses its integrity and will become harder to remember. This interference due to noise, which degrades memory, is often (falsely) interpreted as decay.

(13)

As working memory is a limitation that cannot be improved at the source, we can only attempt to make working memory lean. Lean gives a clear directive as it involves removing waste. An example of removing waste is reducing the interference from paging or moving windows, which inserts noise and causes decay (wasted time, multiple overlapping windows, requiring scrolling, searching for the right button, etc.) which in turn causes the user to forget (part of) the previously visible information. This study focuses on paging and compares it to scrolling within a page.

In business contexts much larger heaps of information require memorisation and much more interference occurs. A more complex and common example is writing a

comprehensible, clear, concise and substantiated literature review (for example on the subject of working memory, such as this paragraph). This task comprises of reading several texts while memorizing and synergizing their relevance. Interference occurs due to using the same faculty2 for different information processing tasks. Part of the interference depends on the strategy of synergizing source documents. Another source of interference lies in having to switch between different documents and the synergized output.

Display size is expected to have a role in decreasing the amount of waste due to interference caused by constraints in the physical size of a display. This study attempts to discover the interference caused by a small and a large display in working memory when attempting to solve a simple but memory intensive task.

2.3. Display Size

Different display features influence the experience of the user, such as size,

information density, colour reproduction, pixel pitch, aspect ratio, etc. Size and information density are the most important for the transfer of information. Colour reproduction, meaning

2_{An inherent capability of the mind, such as speech, reasoning, memorising, etc. (Dictionary.com, 2013)}

(14)

how well a display performs at showing the intended colour, is important in fields such as image and video editing. But these fields are not of interest as this study focuses on the transaction of information in general, such as text, shapes and charts.

Studies on the effects of display size have been done on general task performance (Ni et al., 2006), spatial orientation (Tan et al., 2004) and some attempt to discover more

fundamental tasks, such as navigation (Ball & North, 2005) or text comprehension (Dillon, Richardson, & McKnight, 1990) (Bridgeman, Lennon, & Jackenthal, 2001).

These studies that found improvements when increasing display size also increased the resolution of the display, both as one independent variable. Thus they did not focus on

changing display size, without changing detail of or adding detail to information. If the amount of information stays the same but the resolution is increased, the higher detail of what is rendered has a significant effect on the comprehension of the shown information (Sheedy, Subbaram, Zimmerman, & Hayes, 2005). The figure on

the right provides an example of comprehensibility between different display resolutions, while keeping the

information density the same. Such detail allows for higher legibility and thus readability, increasing speed and accuracy (Ziefle, 1998).

Another method found in many studies ignoring the effects of resolution is changing the resolution to simulate a change in size. This is found to have a detriment effect on legibility (Sheedy et al., 2005) which negatively influences the task on the smaller display, increasing the risk of a type I error.

This study focuses on if increasing information availability affects working memory performance. The requirement of having to move information to be able to see all the

required information is removed on the larger display, without changing any characteristics of

Figure 3. The same two letters in

(15)

the information itself. Moving information on the small display is possible using scrolling, which is further elaborated in the next paragraph.

2.4. Scrolling

As was stated in the introduction, scrolling is one of several methods to overcome the problem of low space availability for information display. Scrolling has several disadvantages compared to paging.

Scrolling is found to interfere with task

execution, as well as being detriment to successful task completion (Bridgeman et al., 2001). Furthermore, time used on scrolling is found to be around 13% of the total time performing a task (Byrne, John, Wehrle, & Crow, 1999). A prominent disadvantage of scrolling is loss of context in working memory, which results in confusion and frustration (Ball & North, 2005). This may explain the performance decrease found in other studies, where users needed to regain context and memorise the content again after scrolling. Similar effects were found, where scrolling reduced the ability to understand complex texts compared to paging, especially when working memory capacity was lower (Sanchez & Wiley, 2009). When comparing the interference of paging and of scrolling, we see that scrolling causes more interference in working memory. Context is lost which causes loss of memorisation, comprehensibility is reduced, and a lot of time is used on waiting, which gives time for random interference (noise). This strongly relates scrolling to working memory and provides a direction of how to test the influence of scrolling on working memory performance.

Even though many disadvantages exist, scrolling is still used in many applications and has seen little reduction. Such reductions seem to be limited to users who heavily rely on

(16)

interface simplification, for example older users (Hawthorn, 2003). What was found to be a neutral effect of scrolling is that users aren’t reluctant to it (Byrne et al., 1999). Furthermore, when using scrolling, users have more control over the shown information, compared to using paging. Paging only allows full replacement of information, while scrolling can be limited to just one line. While this study will not focus on multiple pages, this effect was found to contradict the idea of full replacement having a higher negative impact on performance. A full new page shows more new information, expecting it to causing more interference than just the controlled addition of several new lines of text. In contrast, users were found to prefer paging and have also shown higher performance when using paging (Schwarz, Beldie, & Pastoor, 1983).

In short, scrolling is found to be inferior in performance, but have no direct significant negative effect on user task load experience. This study researches the role of scrolling to determine what influence it has on working memory performance and on user task load experience. How to determine user task load experience is explained in the following paragraph.

2.5. User task load experience

(17)

increasing detail of the same information has a positive effect on task load (Sheedy et al., 2005), but adding more information is expected to have a negative effect since interference in working memory may be much larger due to a larger amount of input.

Measuring task load will be done using a questionnaire based on an adaptation of research and development on the NASA Task Load Index (NASA-TLX) by Hart and

Staveland (1988). The focus of this questionnaire on the personal experience of a participant on the load of performing a task, and its proven reliability shown by its use in many well-executed and well-cited experiments make it a good choice for using it for this purpose.

2.6. Predicted Effects

As scrolling was found to influence working memory performance, scrolling is used to influence the task and enhance the effect on working memory performance. The following quantitative effects are predicted to show in the experimental results, based on the previous literature review. These are not directly related to working memory, as it is not possible to actually measure brain activity during this experiment. As the task is performed in working memory, the following measures will show the influence on working memory.

1) The average time between actions is expected to be shorter as besides moving the mouse, scrolling is not required to choose another tile.

2) As a summation of the previous, having to move the viewport using scrolling requires more time (Sukan, Feiner, Tversky, & Energin, 2012). It is expected that participants will have a lower total completion time using the larger display.

(18)

interference is expected to have a general detriment effect on performance, and quantitatively show more errors in the given task.

Therefore it is predicted that a larger display will show fewer mistakes in the

experiment, since participants no longer need to search in a movable viewport. Participants don’t have to memorise the relative location of a tile that is hidden but has been flipped before, only its absolute location.

2.7. Hypothesis

Based on the literature review and these predicted effects, a hypothesis is formulated on which the experimental design is based. It includes the performance when doing a task requiring storage and processing in working memory, and compares these results between a setting of using a small desktop size display which requires scrolling to perform the task and a large display, displaying all information at once.

Working memory intensive tasks performed using a large display, compared to using a small display:

• require fewer operations; • take less time;

(19)

3. Method

This chapter elaborates on the method of collecting data for the construed hypothesis. A quasi-experiment was designed which loads a participants working memory. First, the sample of the population is described. Second, the experimental design is made clear. Third, the dependent variables are given. Fourth, tools for analysis and reporting are described.

3.1. Sample

34 students of the University of Groningen participated in the experiment for which they received a gift in the form of a consumable as a reward. The age ranged between 20 and 29 with one outlier of 36 (mean = 24). None of the participants had any visual impairment which made the game visuals illegible. All participants had experience using a computer for various tasks, such as text processing, communication and playing games.

3.2. Design

During the quasi-experiment, participants had to play an adaptation of the well-known game of memory 3. This game was chosen for its simplicity, which seemed key in previous research on display size, in order to keep focus on what the experiment is about. For

example, a fundamental and well-cited study by (Estes & Taylor, 1966) consisted of using a tachistoscope to show participants visual elements for a brief moment, to test how many were correctly observed.

3_{The game can be accessed on www.rhtenhove.nl/expthesis2013/exp1/index.php. Changing exp1 to exp2 shows}

the small display version.

(20)

In the small display situation, participants played the game on a 17.3 inch widescreen display using a non-native 4:3 resolution of 1024 times 768, creating a pillarboxed (vertical black bars) image. In the large display situation, participants played the game on a 70 inch projection of the game using a native resolution of 1024 times 768.

On the distance between a user and the display are no clear and definite guidelines available. Distance between the user and the small display was 70 centimetres. For the projection the distance was 3 metres. Only a mouse was used to interact with the game.

Game. As a pre-experimental test, three people played the large-display version of

the game on a display using similar situational characteristics as all other participants were in, using random card layouts and sizes of 20, 24, 28 and 32 cards. Using 24 cards was found to be optimal ensuring participants would not get bored, perform less over time or lose focus.

Participants having affinity with certain graphics, are strongly aided in remembering specifics about the graphics (Chase & Simon, 1973) (Baddeley, 2010). In order to minimize the influence of having a relation with any graphics shown, playing cards were used as the graphic on the tiles. Such playing cards only contain simple shapes and characters, still allowing a wide range of different graphics and minimising the probability of influence of affinity. These cards do contain a fair amount of detail, which may cause interference, but this is the same for each participants and is expected to only enhance the strain on working memory to perform the task.

Figure 5. Colour image of a pillarboxed

(21)

Other differences compared to the original Memory game are:

1) All participants played the game in the same layout. This reduced the element of luck where participants could have a ‘lucky deck’. Only between the different display sizes was a different layout.

2) In the game on the small display, users had to play within a viewport showing only 50% of the playing field. They used vertical scrolling to move the viewport. This was added to simulate a shortage of visual display surface space. Halving the visible part of the game requires the participant, in order to view the entire game, to move required information completely out of view. This increases the pressure on memorising

relative locations, as context and absolute locations may be lost. Making the visible area even smaller would further increase this reliance, but also increases the risk of shifting the game from trying to remember tile locations to trial and error matching, as the game becomes too hard.

3) Users were presented with all tiles for five seconds. This was implemented to further reduce the element of luck in the game. The game on the small display first showed 100% of the game with all tiles and was reduced to 50% after the five seconds passed.

The game was located on an internet server, but when the actual game was played, the entire game was loaded into the browser. This ensured no delays or problems for the

(22)

Questionnaire. After each game, a questionnaire (in Dutch) is filled in by the

participant. This questionnaire can be found in Appendix A. The questionnaire is based on research and development of the Task Load Index by NASA (NASA-TLX) by Hart and Staveland (1988). The questionnaire examines task load experience when using the two different display sizes. Several changes were made to accommodate for the experiment. The NASA-TLX is divided into six categories, to be rated on a scale by a participant, which are explained in thorough detail. From these categories and explanations, relevant questions were extracted and formulated to fit the experiment (for example, emphasis was given on mental demand, while physical demand was reduced to a scale on “looking”). In this same process, five cognitive constructs were extracted from Hart and Staveland's (1988) their research on the NASA-TLX which were found to be relevant to the experiment.

The results of this questionnaire allow to infer a difference in task load experience between the two display sizes, while the quantitative results of playing the game provide performance results. When comparing performance and task load experience results, it can be inferred if display size change provides a positive and strong enough performance difference to justify a difference in task load experience. These results can be enhanced using pairwise comparisons by adding scales to each questions in which a participant can choose the importance (weight) of a question (S. Hart & Staveland, 1988). Yet this is used very infrequently for their added complexity to the questionnaire, and this weighting process is commonly eliminated. The questionnaire is as such referred to as Raw TLX (RTLX) (S. G. Hart, 2006). For this reason, the weighting process will be eliminated in this questionnaire.

Final question. During the experiment, a qualitative question was added to further

(23)

experience after the experiment. The question asked was: “How and why does the experience

differ between sessions?” This question is a within subjects design as its results are used to

determine the experience difference when a participant changes between the two different display sizes. Based on previously discussed literature, it can be expected that the smaller screen provides a better experience. These are more common and the storage limit of our working memory does not directly allow storage of all tiles. These would require a more efficient storage method, such as chunking. As the results were already available at the time of writing, this is further elaborated in the discussion section.

3.3. Dependent Variables

The dependent variables are divided between performance and task load experience. The first three variables are measured during the game to quantify the performance of participants. The fourth variable, task load, is measured using the questionnaire.

Total time. The time required to finish the game is measured from the first card flip

until the game is finished. This is measured as it is expectated that participants will require less time on the larger display since they do not need to use scrolling to access all tiles.

Click frequency. Click frequency allows inferences on strategy (clicking until it

works or trying hard to remember tile locations) and error frequency as only 24 clicks are required to finish the game, which consists of 24 tiles.

Average time between clicks. This variable can be calculated using the measures

(24)

Task load. The questionnaire was measured for its internal consistency using

Cronbach’s Alpha. If this is sufficient (α >= .7), it allows inferences on task load using all questions (Kline, 2000).

3.4. Analysis

(25)

4. Results

This section summarizes the main results of the experiment. First, the possible existence of a learning effect is determined in both the performance and the questionnaire results. Second, the results on performance are summarised. Third, the task load results of participants using the two different devices are shown. The following table defines the different groups of participants. The diagram on the right shows the order in which participants played the game in two situations.

Table 1

Legend of data groups

Group Name Description

S1 Small first Small display data of participants playing their first game.

L1 Large first Large display data of participants playing their first game.

S2 Small second Small display data of participants playing their second game. L2 Large second Large display data of participants

playing their second game.

4.1. Learning Effect Analysis

All participants played a session of two games, which may cause a learning effect. They may become less anxious about the experiment and/or develop a strategy to play the game. The influence of this effect could cause Type I and Type II errors. The learning effect is determined using an independent samples t-test between subjects.

Table 2 shows the data and results. Only the variable ‘click frequency’ reveals a learning effect, t(26.59) = 2.61, p = .015. Testing for a learning effect when all results are combined shows no significance. The significance of the learning effect in the variable ‘click frequency’ requires it to be analysed between subjects as the risk of a Type I or Type II error

Figure 6. The order in

(26)

is too large. All other variables are analysed within subjects (small compared to large display), since their learning effects are insignificant.

Table 2

Data and results of learning effect analysis

Dependent variable Mean Mean difference

S1 S2 L1 L2 Small Large

Click frequency 57.88 72.71 68.24 69.30 14.82* 1.06

Total timea 108.15 116.87 108.74 104.50 8.72 4.24

Average time between clicksa 1.87 1.65 1.61 1.50 0.23 0.11

Task loadb 2.97 2.98 2.55 2.62 0.01 0.07

Notes: *p = .015. aValues are in seconds. bThe means are the averages of the 10 task load relevant questions, further explained in paragraph 4.3.

4.2. Performance

Performance during the game is divided in a) the amount of times clicked on a tile, b) the time required to solve the game, and these together as c) the average time between clicks. The tables shows the mean, standard deviation and confidence interval. The means and confidence intervals are visualised in graphs. All tests are two-tailed since a decrease in performance may occur and is of equal importance as an improvement.

Click frequency. This variable is

analysed between subjects to prevent a Type II error, because a learning effect was found. The chart on the right visualises the

difference between S1 and S2 and how this data could cause erroneous results if a within subjects analysis was performed. Within subjects analysis combining performance

40 50 60 70 80 90 S1 L1 S2 L2 Cl ick s Game

Click frequency

(27)

results of S1 & S2 as well as L1 & L2 showed no significant results. A between subjects analysis showed a difference between S1 and L1 of 10 (15%), t(32) = 2.06, p = .048. Between S2 and L2, no significant difference was found. These results show that a within subjects analysis would have caused a Type II error. Table 3 shows the data and results of the between subjects analysis. These were analysed using an independent samples t-test.

Table 3

Data and results of click frequency analysis

Game Mean Std. dev. Conf. int. Mean difference

Small Large Small Large Small Large

First 58 68 12.28 12.08 5.84 5.74 10*

Second 73 69 19.96 19.30 9.49 9.17 4

Note: *p = .048.

Since no learning effect was found for the three dependent variables on the following pages, the performance results of each display size is combined for a within subjects analysis, to provide a larger dataset for analysis.

Total time (in seconds). Table 4 shows the

data and results. A paired samples t-test showed no significant difference between the two display sizes,

t(33) = -0.95, p = .349.

Table 4

Data and results of total time analysis

Mean St. dev. Conf. int. Mean difference

112.51 106.62 52.04 42.05 17.75 14.35 5.89 90 100 110 120 130 Small Large Se co nds Game

Total time (in seconds)

(28)

Average time between clicks (in seconds).

Table 5 shows the data and results. A paired samples t-test shows the average time is 0.2 seconds (11%) lower on the large display, t(33) = -2.31, p = .027.

Table 5

Data and results of average time between clicks analysis

Mean St. dev. Conf. int. Mean difference

1.76 1.56 0.70 0.49 0.24 0.17 0.2*

Note: *p = .027

4.3. Questionnaire

Table 6 on the following page shows the data and results. Between S1 and L1, the cognitive constructs mental workload and looking were significantly lower in group L1. Between S2 and L2, time pressure, uncertainty and decision making showed a significant higher result in group L2. All other questions between S2 and L2 also produced higher, but insignificant results. These were analysed using an independent samples t-test.

1,3 1,5 1,7 1,9 2,1 Small Large Se co nds Game

Average time between

clicks (in seconds)

Figure 9. Average time between clicks

(29)

Table 6

Data and results of the questionnaire

Question Meana _{Mean difference}a

S1 L1 Mental workload (Q4) 3.41 2.76 0.647* Looking (Q11) 3.24 2.06 1.176** S2 L2 Time pressure (Q2) 2.24 3.00 0.765*** Uncertainty (Q5) 2.12 2.82 0.706**** Decision making (Q8) 2.00 2.65 0.647***** Notes: *p = .041. **p = .008. ***p = .041. ****p = .049. *****p = .044.

a_{Values are mean scores on a 5-point scale (1 = low, 5 = high).}

The internal consistency of the questionnaire was analysed using Cronbach’s Alpha. The results of question 3 have been reversed (1 becomes 5, 2 becomes 4, etc.) to account for its reversed scale. The questionnaire is found to be internally consistent in every group, showing the highest α value when question 3 is deleted. This is expected since the other questions focus on task load during the game, while question 3 focuses on the experience

after the game, which may have made it less consistent with the other questions. Table 7 on

the following page shows the internal consistency analysis results in which question 3 has been deleted.

Table 7

Internal consistency analysis using Cronbach’s Alpha

S1 S2 L1 L2

α = .854 α = .851 α = .844 α = .818

α = .864a α = .834a

(30)

Since the internal consistency shows a α values in the range 0.8 ≤ α < 0.9, it is classified as good, allowing the questions to be combined and used as a measure of task load (Kline, 2000). A paired samples t-test between S1 & S2 and L1 & L2 showed a significantly lower task load when using a large display, t(33) = -2.259, p = .031. Data and results are depicted in table 8.

Table 8

Data and results of questionnaire analysis

Meana St. dev.a Mean

differencea Small Large Small Large

2.97 2.58 0.69 0.64 0.39*

Notes: *p = .031. aValues are mean scores on a 5-point scale (1 = low, 5 = high).

4.4. Final Question

(31)

5. Discussion

Analysing the results of this experiment shows several interesting insights into

behaviour and experience of the participants between using a large and a small display. First, performance and questionnaire results will be discussed. Second, managerial and theoretical implications are discussed to determine how managers can use these results and how literature can build upon these results with several examples of future research opportunities. Third, limitations of this study are described. Fourth and last, concluding remarks are given.

5.1. Results

The results answered the following research question:

How does the physical change of display size affect working memory performance?

which was the basis of the hypothesis:

Working memory intensive tasks performed using a large display, compared to using a small display:

• require fewer operations; • take less time;

• induce a lower task load experience.

The experiment has shown that the performance increased on one out of the three performance based variables, and the overall task load decreased when using a larger display. All other variables had an insignificant difference and none of the variables showed a

decrease in performance or an increase in task load. The results provide several insights which require further discussion.

(32)

significantly lower (better). They indicated they also had significantly less difficulty with looking (better). This may be explained with that having a better overview lightens mental workload and makes participants more active, using less strategy and more trial and error.

Another related contradiction is that when participants played the game for the second time, the click frequency was lower (better) when using a large display, yet several task load questions scored significantly higher (worse). This means that on the larger display,

participants performed better on the game, yet on the questionnaire they indicated that the large display caused a higher load on decision making and a stronger feeling of time pressure and uncertainty. These effects may have a reciprocal nature making the source of a change harder to determine, but certain effects can be explained using literature. The increase in difficulty with decision making can be explained by the significant increase of feeling time pressure (Ordonez & Benson, 1997) as well as the significant increase of feeling uncertainty. Decisions may be based on decisions in the first game or on unknown probability

distributions, for example luck, which are a source of uncertainty (Busemeyer, 1985). The increase in uncertainty on a larger display contradicts findings by Ball and North (2005), who concluded that a larger display is less stressful and results in a stronger feeling of certainty.

All other results have been analysed using a within subjects analysis because there was no learning effect found in these variables. The results of the average time between clicks showed a significant decrease when using a larger display, which is in favour of the

(33)

Summary. One dependent variable (click frequency) supports the hypothesis when

playing the game for the first time. Two dependent variables (average time between clicks and task load) support the hypothesis comparing the small and large display within subjects. One variable (time required) showed no significant effect. These results allow to infer that using a larger display requires less operations and users perceive a lower task load during working memory intensive tasks, accepting the hypothesis.

The results of the final question, which was asked to the last 15 participants, explain the small differences found in performance and task load between the two display sizes. These insights are a basis for further research as these cannot be further studied within the finished experiment. Although the sample size is relatively small (15), their opinion shows that the insignificant performance difference in total time may be explained by that the larger display does in fact decrease total time required, but the amount of information to be

(34)

5.2. Implications

The intention of this study, explained in the introduction, is to determine if working memory performs different between small displays using scrolling or large displays. In Chapter 2 current literature on relevant subjects was reviewed, which presented a gap in literature where there was no clear definition of the role of display size in general

performance or how it can add to the natural interaction paradigm. Furthermore, most studies did not focus on fundamental human cognitions but used broad tasks without clear cognitive applications. This study focused on the fundamental human cognition working memory because of this lack in literature and its importance in nearly every task.

Managerial Implications. Most workplaces have a small display to communicate

information to their user. What this study shows is that using a large display decreases task load and increases performance when doing working memory intensive tasks. These

improvements may not be significant enough for a single user, but when multiple employees are working on multiple tasks and multiple applications, its significance shows that a larger display benefits the organisation. The main addition is that a larger display will reduce the amount of interference in working memory, reducing the frequency of forgetting which in turn reduces the amount of errors made, and the time required to reacquire information. Another main addition is the reduction in task load, reducing the amount of mental strain that needs to be delivered, without reducing the task output.

Theoretical Implications and Future Research Opportunities. It was shown that a

larger display size, which removes the requirement of scrolling to view all the required

(35)

simple and obvious, such as increasing the sample size or adding dependent variables. Others may be more interesting and are depicted below.

• Use touch interaction instead of a mouse. Comparing results with this study yields insight into the effects on working memory of display size change in multiple interaction scenarios.

• Further study the balance of effects from increasing display size and removing the requirement of scrolling.

• Systematically increase or decrease interference on working memory, potentially amplifying the found effects. Since there are many sources of interference, an effect of interference can relatively easy be found (for example by adding audible noise). • Determine the interference effects of the independent variables in this experiment. For

example, scrolling may cause interference due to added interaction elements and unexpected visual changes due to movement or technological limitations.

• Study the difference between paging and scrolling using modern interfaces and visual aids. Animations were found to have an impact on scrolling (Klein & Bederson, 2005). This might show how scrolling can be made less detriment on performance.

5.3. Limitations

(36)

case. As such, a recommendation is to repeat the experiment with subjects that would benefit of the as of yet insignificant improvements found.

It was a limitation as well as a strength that participants did not first do a preparation session. The limitation is the influence of a learning effect. This may have provided more significant results as a lower variance can be expected. A strength in this is that it also shows how larger displays perform when people haven’t performed a task on it before. This allows inferences for situations where a large display may be used just once by the same person.

Not using a display’s native resolution can have a negative effect on the legibility of information (Sheedy et al., 2005). A display having the same native resolution as the projector could not be found. As a result, the small display showed an image in the correct aspect ratio, with the same detail, but these had slightly blurred pixel edges due to

enlargement of the image.

Playing with 24 cards was found to be the best choice in a short pre-experimental test with three people. Even though this highly exceeds the amount of items most people can store in their working memory, a higher amount of cards may show a stronger effect.

Question three from the questionnaire was omitted because its results were insignificant, did not add to or explain any effects and has shown to decrease the internal consistency of the questionnaire. This is further explained in paragraph 4.3.

5.4. Conclusions

(37)

possibilities of displays and implications found in the fundamental human cognition working memory.

The experiment conducted for this purpose has shown that a large display makes higher working memory performance possible. The average time between clicks was reduced by 11% when using a large display. Click frequency was 15% lower on the large display compared between subjects doing the game for the first time during their session.

Furthermore, it was found that when using the large display, the task load experience by participants was higher on some elements, but found an overall significant decrease. This study accepts the hypothesis construed in the introduction and allows to conclude that working memory intensive tasks performed using a large display do indeed require fewer operations and induce a lower task load experience, but do not show a significant decrease in the total time required.

(38)

6. References

Atkinson, R., & Shiffrin, R. (1971). The control processes of short-term memory. Psychology

Series.

Baddeley, A. (1992). Working memory. Science, 255(5044). Baddeley, A. (2010). Working memory. Current biology, 20(4).

Ball, R., & North, C. (2005). Effects of tiled high-resolution display on basic visualization and navigation tasks. In CHI ’05 extended abstracts on Human factors in computing

systems - CHI '05 (p. 1196). New York, New York, USA: ACM Press.

Borst, J. P., Taatgen, N. a, & van Rijn, H. (2010). The problem state: a cognitive bottleneck in multitasking. Journal of experimental psychology. Learning, memory, and cognition,

36(2).

Bridgeman, B., Lennon, M. L., & Jackenthal, A. (2001). Effects of screen size, screen

resolution, and display rate on computer-based test performance. Applied Measurement in … (pp. 1–23).

Busemeyer, J. R. (1985). Decision Making Under Uncertainty : A Comparison of Simple.

Journal of Experimental Psychology: Learning, Memory and Cognition, 11(3).

Byrne, M., John, B., Wehrle, N., & Crow, D. (1999). The tangled Web we wove: a taskonomy of WWW use. In Human Factors in Computing Systems: Proceedings of CHI 99.

Cantoni, V., Cellario, M., & Porta, M. (2004). Perspectives and challenges in e-learning: towards natural interaction paradigms. Journal of Visual Languages & Computing,

15(5).

Chase, W., & Simon, H. (1973). Perception in chess. Cognitive psychology, 61.

Cowan, N. (2001). The magical number 4 in short-term memory: a reconsideration of mental storage capacity. The Behavioral and brain sciences, 24(1).

Czerwinski, M., Tan, D., & Robertson, G. (2002). Women take a wider view. In Proceedings

of the SIGCHI … (p. 195). New York, New York, USA: ACM Press.

Daft, R., & Lengel, R. (1986). Organizational information requirements, media richness and structural design. Management science, 32(5).

Dam, A. Van. (2000). Beyond wimp. Computer Graphics and Applications, IEEE, (February).

(39)

Dillon, A., Richardson, J., & McKnight, C. (1990). The effects of display size and text splitting on reading lengthy text from screen. Behaviour & Information …, 9(3).

Duchnicky, R., & Kolers, P. (1983). Readability of text scrolled on visual display terminals as a function of window size. Human Factors: The Journal of the Human Factors and

Ergonomics, 25(6).

Estes, W. K., & Taylor, H. A. (1966). Visual detection in relation to display size and redundancy of critical elements. Perception & Psychophysics, (68).

Estes, W. K., & Wessel, D. L. (1966). Reaction time in relation to display size and correctness of response in forced-choice visual signal detection. Perception & Psychophysics, 1(5). Forlines, C., Wigdor, D., Shen, C., & Balakrishnan, R. (2007). Direct-touch vs. mouse input

for tabletop displays. Proceedings of the SIGCHI conference on Human factors in

computing systems - CHI ’07.

Galbraith, J. (1974). Organization design: an information processing view. Interfaces, 4(5). Guimbretière, F., Stone, M., & Winograd, T. (2001). Fluid interaction with high-resolution

wall-size displays. In Proceedings of the 14th annual … (pp. 1–140).

Hart, S. G. (2006). Nasa-Task Load Index (NASA-TLX); 20 Years Later. Proceedings of the

Human Factors and Ergonomics Society Annual Meeting, 50(9).

Hart, S., & Staveland, L. (1988). Development of NASA-TLX (Task Load Index): Results of empirical and theoretical research. Human mental workload.

Hawthorn, D. (2003). How universal is good design for older users? Proceedings of the 2003

conference on Universal usability - CUU ’03.

Henderson, D. A., & Card, S. (1986). Rooms: the use of multiple virtual workspaces to reduce space contention in a window-based graphical user interface. ACM Transactions on

Graphics (TOG), 5(3).

Jonides, J., Lewis, R. L., Nee, D. E., Lustig, C. a, Berman, M. G., & Moore, K. S. (2008). The mind and brain of short-term memory. Annual review of psychology, 59.

Klein, C., & Bederson, B. B. (2005). Benefits of animated scrolling. In CHI ’05 extended

abstracts on Human factors in computing systems - CHI '05 (p. 1965). New York, New

York, USA: ACM Press.

Kline, P. (2000). The Handbook of Psychological Testing (2nd ed., p. 744). London: Taylor & Francis.

Matanda, M. J., & Freeman, S. (2009). Effect of perceived environmental uncertainty on exporter–importer inter-organisational relationships and export performance

improvement. International Business Review, 18(1).

(40)

Ni, T., Bowman, D., & Chen, J. (2006). Increased display size and resolution improve task performance in Information-Rich Virtual Environments. In Proceedings of Graphics

Interface 2006 (pp. 139–146).

Ordonez, L., & Benson, L. (1997). Decisions under time pressure: How time constraint affects risky decision making. Organizational Behavior and Human Decision …, 71(2).

Sanchez, C. a., & Wiley, J. (2009). To Scroll or Not to Scroll: Scrolling, Working Memory Capacity, and Comprehending Complex Texts. Human Factors: The Journal of the

Human Factors and Ergonomics Society, 51(5).

Schwarz, E., Beldie, I., & Pastoor, S. (1983). RESEARCH NOTE A Comparison of Paging and Scrolling for Changing Screen Contents by Inexperienced Users. Human Factors:

The Journal of ….

Sheedy, J. E., Subbaram, M. V., Zimmerman, A. B., & Hayes, J. R. (2005). Text Legibility and the Letter Superiority Effect. Human Factors: The Journal of the Human Factors

and Ergonomics Society, 47(4).

Sukan, M., Feiner, S., Tversky, B., & Energin, S. (2012). Quick viewpoint switching for manipulating virtual objects in hand-held augmented reality using stored snapshots. In

2012 IEEE International Symposium on Mixed and Augmented Reality (ISMAR) (pp.

217–226). Ieee.

Tan, D. S., Gergle, D., Scupelli, P. G., & Pausch, R. (2004). Physically large displays improve path integration in 3D virtual navigation tasks. In Proceedings of the 2004

conference on Human factors in computing systems - CHI ’04 (pp. 439–446). New

York, New York, USA: ACM Press.

Tan, D. S., Gergle, D., Scupelli, P., & Pausch, R. (2003). With similar visual angles, larger displays improve spatial performance. In Proceedings of the conference on Human

factors in computing systems - CHI ’03 (p. 217). New York, New York, USA: ACM

Press.

Turk, M., & Robertson, G. (2000). Perceptual user interfaces. Communications of the ACM,

43(3).

Valli, A. (2008). The design of natural interaction. Multimedia Tools and Applications, 38(3). Ziefle, M. (1998). Effects of Display Resolution on Visual Performance. Human Factors: The

(41)

7. Appendix A. – Questionnaire

VRAGENLIJST

Deelnemer #: Leeftijd:

Omcirkel bij elke vraag het nummer wat het beste bij je oordeel past. Gebruik de schaal rechts voor je oordeel.

Een hoger getal betekent een sterker oordeel. Bijv.: 5 betekent een hoge belasting en 1 vrijwel geen belasting.

#1:

Vraag Schaal

In hoeverre vond je de taak moeilijk/complex? 1 2 3 4 5

Hoeveel tijdsdruk ervoer je? 1 2 3 4 5

Hoe goed denk je dat je in het volbrengen van de taak was? (1=zeer slecht, 5=heel goed) 1 2 3 4 5 Hoe hard moest je werken (mentaal) om je gewenste niveau van prestatie te bereiken? 1 2 3 4 5

Hoe onzeker voelde je je tijdens de taak? 1 2 3 4 5

Hoe gespannen voelde je je tijdens de taak? 1 2 3 4 5

Hoeveel moeite had je bij de volgende punten? -

Effects of Scrolling and Display Size on Working Memory Performance