Preference and Performance:
An empirical experiment that investigates a possible connection between
preference for certain spatial-temporal visualizations and performance of tasks
using these visualizations
Master Thesis
Computer-Mediated Communication Tessa Huisman
S2729458
Supervisor: L.M. Bosveld-de Smet Second reader: G.J. Mills
1
Preface
Voor u ligt de scriptie: Preference and Performance: An empirical experiment that investigates if there is a connection between preference for certain spatial-temporal visualizations and performance of tasks using these visualizations. Deze scriptie is geschreven in het kader van mijn afstuderen aan de masteropleiding Computercommunicatie aan de Rijksuniversiteit Groningen.
Het afgelopen jaar ben ik bezig geweest met het verzamelen en analyseren van data en het schrijven van de scriptie. De focus van mijn scriptie ligt op het onderzoeken van een connectie tussen de prestatie van gebruikers op het uitvoeren van taken met behulp van visualisaties en de voorkeur van gebruikers voor visualisaties. Het onderzoek dat ik heb uitgevoerd was complex, met name door de verschillende type data die gebruikt zijn. Samen met Daniël Houben en onder begeleiding van mijn scriptiebegeleider Leonie Bosveld-de Smet heb ik het experiment opgezet en is het gelukt om een uitgebreid en betrouwbaar experiment uit te voeren. Ik hoop met dit onderzoek bij te dragen aan bestaande literatuur over de evaluatie van spatiaal-temporale visualisaties.
Graag wil ik Leonie Bosveld–de Smet bedanken voor haar fijne begeleiding tijdens het hele scriptietraject. De begeleiding en feedback hebben bijgedragen aan het resultaat van deze scriptie. Ook wil ik graag mijn studiegenoten en goede vrienden Daniël, Dewi en Lisa bedanken voor hun steun, ideeën en gezelligheid. Niet alleen tijdens het schrijven van de scriptie maar ook tijdens het gehele mastertraject heb ik veel aan hen gehad. Als laatste wil ik mijn ouders, broertje en vrienden bedanken voor hun liefde en support. Dankzij hun motiverende woorden onvoorwaardelijke steun heb ik deze scriptie en mijn opleiding tot een goed einde weten te brengen.
Ik wens u veel leesplezier. Tessa Huisman
2
Abstract
This study aims to find out if there is a connection between the preference users have for certain temporal visualizations and their performance of tasks using these visualizations. Two spatial-temporal visualizations are evaluated based on three aspects of usability: effectiveness, efficiency and satisfaction. Two representations of spatial-temporal data are presented: a map representation, characterized by natural mapping of a location, and a Gantt chart, characterized by natural mapping of time. An empirical experiment was set up to evaluate the usability of both visualization types. Forty people participated in this study in a within-subject design. Participants had different genders, ages and educational backgrounds. All participants were rather novice users than expert users of the visualizations presented to them. Participants are exposed to two different visual representations and were requested to falsify or verify a variety of query statements. The query statements required cognitive operations and varied with respect to focus and reading level. In this study, effectiveness is measured as response accuracy and efficiency as response time. Satisfaction is measured by asking the participants to choose the visual representation of their preference. In a face-to-face semi-structured interview, participants had the opportunity to explain their preference in more detail. The objective result data show that both the Gantt chart and the map representation support users to verify or falsify query statements equally well. Users do answer the query statements faster when presented with the map representation compared to their response time when presented with the Gantt chart. Although it can be observed that users tend to prefer the map, no significant difference between the visualizations can be found. When it comes to preference, two dimensions are studied, namely insightfulness and attractiveness. Based on a qualitative analysis of verbal recordings of participants, significantly more participants consider the map more attractive than the Gantt chart. As for the insightful support participants get when answering the queries, no clear difference in preference has been observed. The Gantt is appreciated by users for helping answer queries with time in focus. The map on the other hand is more appreciated by users for answering queries with a location in focus. No possible connection between performance and preference is found in the analysis of the query statements. Two perspectives regarding this connection were tested; it is tested whether preference can be seen as a function of performance and whether performance can be seen as a function of preference. It can be observed that participants with a high response accuracy and a fast response time prefer the map, but no significant difference has been found. Remarkably, participants with a preference for the Gantt chart did not perform better on the Gantt chart. Instead, a negative association has been observed with participants who prefer the Gantt chart and their response times: they were significantly slower when answering Gantt queries than the participants who preferred the map.
3
Index
Preface
1
Abstract
2
List of Figures and Tables
6
Figures 6
Tables 6
1.
Introduction
8
1.1 Importance of usability studies of visual representations 8
1.2 Problem statement 8
1.3 Research Questions 9
1.4 Reading Guide 9
2. Theory
11
2.1 Visualizing spatial-temporal data 11
2.1.1 Mapping of data for visual representations 11
2.1.2 Visualizing objects, time and space 12
2.2 Map and Gantt chart 13
2.2.1 Map representation 13
2.2.2 Gantt chart 14
2.3 Querying spatial-temporal data 15
2.4 Usability of visualizations 16
2.4.1 Usability Engineering 16
2.4.2 Measuring usability of visual representations 17
2.5 Relation between subjective user satisfaction and objective task performance 17
2.5.1 Connection of subjective user preference and objective task performance in HCI studies 17 2.5.2 Connection of subjective user preference and objective task performance in visualization studies 18
2.5.3 Types of users 19
2.6 Expectations based on literature 20
3. Methodology
21
3.1 Design 21
3.1.1 Operationalization task performance 21
3.1.2 Operationalization preference 21
3.2 Participants 21
3.3 Materials 22
3.3.1 Mapping of data in visualizations 23
3.3.2 Visualizations 25
3.3.3 Complexity of visualizations 27
3.3.4 Data exploration tasks 27
3.3.5 Survey design 30
4
3.4 Procedure 31
3.4.1 Controlled setting 31
3.4.2 Paper and Pencil protocol 32
3.4.3 Procedure 32
3.5 Data Analysis 33
3.5.1 Task performance analysis 33
3.5.2 Preference data 33
3.5.3 Interaction of objective measures and subjective measures 34
4. Results
36
4.1 Overall performance 37
4.1. 1 Response accuracy 37
4.1.2 Response Time 38
4.2 Overall preference 39
4.2.1 Preference in relation to moderating variables 40
4.3 Connections between performance and preference 41
4.3.1 Preference as function of performance 41
4.3.2 Performance as function of preference 43
4.4. Results based on interview 46
4.4.1 Global results 46
4.4.2 Numeric results from the interview 47
4.4.3 Analysis of answers 49
5. Conclusion and discussion
52
5.1 Global Results 52
5.1.1 Performance 52
5.1.2 Preference 52
5.2 Connections between performance and preference 52
5.2.1 Response accuracy and preference 52
5.2.2 Response time and preference 53
5.2.3 Qualitative outcomes 53
5.3 Main outcomes of this study 53
5.3.1 Task performance 53
5.3.2 Connection between task performance and preference 54
5.4 Limitations and further research 54
5.4.1 Connections between performance and preference 54
5.4.2 Query statements and preference 55
5.4.3 Post task interview 55
5.4.4 Strategies 55
6.
References
56
Appendices
58
Appendix A | Finalized visualizations variants 58
A.1: Map 58
A.2: Gantt chart 60
Appendix B | Statements per statement type 62
5
B.2: Map 63
Appendix C| Informed consent 65
Appendix D | Codebook 66
Appendix E | Objective result data 68
E.1 Overall performance and preference per participant 68
E.2 Response accuracy of participants per visualization type 69
E.3 Response time of participants per visualization type 70
Appendix F | Outliers in objective result data 71
F.1 Response Accuracy 71
F.2 Response time 71
Appendix G | normal distributions 73
G.1 Overall response accuracy grouped by preference 73
G.2 Overall response Time grouped by preference 74
Appendix H: Transcribed intervies 75
Appendix I: Answers from interview 88
Appendix J: Analysis of interviews 90
6
List of Figures and Tables
Figures
Figure 1: Three parts of visualization according to Wang (1995) 11
Figure 2: An example of a Map representation 13
Figure 3: A simple map visualization designed by Kriglstein et al. (2016) 13
Figure 4: An example of a Gantt chart 14
Figure 5: A simple Gantt chart designed by Kriglstein et al. (2016) 14
Figure 6: Mapping of persons in both visualizations 23
Figure 7: Mapping of time in the Gantt chart 24
Figure 8: Mapping of time in the map representation 24
Figure 9: Mapping of location in the Gantt chart 24
Figure 10: Mapping of locations in the map representation 25
Figure 11: Simple example of the Gantt chart used 26
Figure 12: Simple example of the map representation used 26
Figure 13: Data exploration task distribution (Houben, 2020) 27
Figure 14: Example of layout of experiment 30
Figure 15: Testing environment during experiment 32
Figure 16: Distribution of response accuracy per visualization type 37
Figure 17: Distribution of response time per visualization type 38
Figure 18: Distribution of preference among participants 39
Figure 19: Distribution of response time and response accuracy of participants (N=38) color-coded by
preference 41
Figure 20: Wordcloud of interview 46
Figure 21: overview of names used for the visualizations by participants in a word cloud 46
Tables
Table 1: Question-types, their reading levels, the cognitive operations and their focus (Houben, 2020) 15
Table 2: Characteristics of all participants 22
Table 3: Variables and their corresponding values 25
Table 4: Differences in low and high complexity in data 27
Table 5: Query-statement formulation per type: the number of statements in category, rules, sentence
order and examples (Houben, 2020) 28
Table 6: Example of coding of the interview 34
Table 7: Response accuracy per visualization type 37
Table 8: Measures of response time per visualization type 38
Table 9: preference of participants (N=39) grouped by their gender 40
Table 10: Preference of participants (N=39) in relation to age 40
Table 11: Division of preference among education level 40
Table 12: Frequency table of the response accuracy of participants and their preference 42 Table 13: Frequency table of the response time of participants and their preference 42 Table 14: Response accuracy measures for overall response accuracy and per visualization type
grouped by preference 43
Table 15: Response accuracy measures per preference group, differentiated by visualization type 44 Table 16: Response time measures for overall response time and per visualization type grouped by
7 Table 17: Response time measures per preference group, differentiated by visualization type 45
Table 18: frequency of insightfulness per visualization 47
Table 19: Measures of most attractive visualization 47
Table 20: Example of analysis of remarks made by participants during the interview 48 Table 21: Frequency of positive and negative remarks made by participants in the interview 49
8
1. Introduction
1.1
Importance of usability studies of visual representations
Throughout the history of communication, visual representations have been used to communicate information or to tell stories. Visualizations are representations of information that exploit spatial lay-out in a meaningful way (Bosveld-de Smet, 2005). Gurr (1999) defines a visual representation as a collection of objects and some relations between these objects. The type of representation is determined by the characteristics of the symbols used to express these objects and relations. Hegarty (2004) points out that visual representations may be external or internal. The visual representations that are relevant for the purpose of this studyare external, two-dimensional, and static. These visual representations include realistic pictures, photographs, icons, maps, charts, graphs, diagrams etc. (Bosveld-de Smet, 2005). Different types of data can be communicated through visualizations; visualizations have the ability to show us relationships between objects without using intermediary abstract syntax that has to capture these relationships (Gurr, 1999). Visualizations can help people with solving tasks, for example a visualization of a map can help people to plan a route or to answer questions about locations on the map. Different data types require different forms of visual representations. When data are visualized the wrong way, the visualization will communicate the wrong information. Data should be visualized in such a way that it fits its communicative purpose. In some cases, the mapping of data is challenging due to the combination of data. Spatial-temporal data, which relate to space and time, are challenging data types. An example of spatial-temporal data is a person who changes locations over time. Visualizing spatial-temporal data can be done in multiple ways. Two possible information visualizations that support spatial-temporal data exploration are map-based and timeline-map-based representations. Both of these visualizations show the data in a different way; in the map-based representation, locations are arranged according to their positions in the real world. In the timeline-based representation, time is represented as a linearly ordered set of time points or time intervals. Section 2.2 will explain more about these visualizations.
It is important to examine whether data are visualized in a way that helps viewers to make sense of the data visualized. A way to gain insight into the usefulness of visualizations in data exploration tasks is to investigate the usability of visualizations. Usability is a notion that is generally associated with user interfaces (UI). Many human-computer interaction studies measure the usability of UIs. Usability is often defined as a construct with multiple dimensions. Effectiveness, efficiency and satisfaction are the dimensions used in the definition of usability of the ISO (International Organization for Standardization). Effectiveness and efficiency can be measured in an objective way. In this thesis, effectiveness will be operationalized as response accuracy and efficiency as response time. Satisfaction, on the other hand, is a subjective notion and can be measured with the aid of questionnaires about the user experience with UIs, or by sorting several UIs according to user satisfaction. This thesis examines user satisfaction of visualizations of spatial-temporal data, in relation to their effectiveness and efficiency. In visualization studies, satisfaction with visualizations is often referred to as preference. Two main dimensions of preference are measured: insightfulness and aesthetics. Section 2.4 will explain in more detail how the usability of visualizations can be measured.
1.2
Problem statement
There has been a line of research in UI studies as well as in visualization studies that have attempted to find a connection between satisfaction or preference on the one hand and efficiency and effectiveness on the other. The literature regarding UIs suggests that a connection can be found between preference
9 and task performance. Nielsen and Levy (1994) even state that performance can be seen as a function of preference, which means that based on the preference of users, the performance can be predicted. Some visualization studies, however, show that there is no connection between subjective user preference and objective task performance; even though others show that a connection does exist. This may not come as a surprise, as the use of visualizations differs fundamentally from the use of UIs. Humans interact with UIs to perform a computer task, such as calculating the average of a set of numeric data, substituting a certain word by another one, or storing a document in a specific folder. Usable UIs give feedback about whether the task has been performed successfully, whereas many visualizations are static and do not allow any interaction. Tasks performed with visualizations are different from the ones performed in UIs. And, as pointed out by Roberts, Gray and Lesnik (2013) there is no indication for users of visualizations when they perform their tasks successfully. So, on the one hand some visualization researchers find no connection between subjective preference and objective task performance (Roberts, Gray and Lesnik, 2013), while others do find a connection (Kessell and Tversky, 2010; Kriglstein, Haider, Wallner and Pohl, 2016). The reason for this disagreement could be explained by the types of visualizations that are studied. Roberts, Gray and Lesnik (2013) have done experimental studies with metro maps, while Kessell and Tversky (2010) and Kriglstein, Haider, Wallner and Pohl (2016) have compared various visualizations of spatial-temporal data. This study intends to contribute to existing literature about the usability of visualizations of spatial-temporal data. It reports on an empirical experiment measuring subjective preference and objective task performance (task effectiveness and task efficiency) with two visualizations of spatial-temporal data; a map-based and a Gantt-based one. As a follow-up, it investigates whether there is any connection between objective task performance and subjective preference, and to what extent.
1.3
Research Questions
The main question of this thesis is: Is there a connection between task performance and user preference for two different visualizations of spatial-temporal data, a map and a Gantt chart; and to what extent? To find an answer to this question, the following subquestions are formulated:
1. Which one of two visualizations of spatial-temporal data, map or Gantt chart, supports users best in performing data exploration tasks, with respect to response accuracy and response time?
2. Which visualization is most preferred?
3. Is there a connection between overall preference and task performance?
4. Do users’ explanations of their preference account for the outcome of question 3?
1.4
Reading Guide
Chapter 2 will provide a theoretical background for this study. It will show the components involved in visualizations in general and how data in the application domain are mapped to objects in the graphical domain. Chapter 2 will discuss spatial-temporal data and their possible visualizations in more detail, among which the map and the Gantt chart. Chapter 2 will also introduce the concept of usability in the context of UIs and visualizations; it will focus on objective task performance and subjective user preference. This chapter will also discuss related studies that have and studies that have not found a connection between preference and performance. Chapter 3 will discuss the methodological approach chosen to answer the research questions. An empirical evaluation experiment has been set up and conducted to obtain both objective and subjective measures. Chapter 3 will also discuss the experimental setup, the selection of participants, the materials designed for the experiment, the procedure, and how all the obtained result data are analyzed. Chapter 4 will describe the results of the experiment; the focus of this chapter will be on the comparison of objective task
10 performance and subjective user preference. In chapter 5, the results will be discussed and the research questions will be answered, based on the results. Further research will be suggested, and recommendations for the design of visualizations of spatial-temporal data will be indicated.
11
2. Theory
This thesis attempts to discover whether there exists a relation between user preference and task performance for two visualizations of spatial-temporal data. Two such visualizations are a map-based diagram and a Gantt chart. This chapter will start with a global overview of how visual representations in general are created (Section 2.1). Data in some application domain are mapped to objects in a graphical domain. The link between these two domains allows users to interpret the visualization. Spatial-temporal data are data that relate to objects, space and time. Two-dimensional static visualizations of these data are composite, as they have to combine mappings of objects, time, and space to spatial graphical objects. A map and a Gantt chart will be introduced as visualization options in section 2.2. The map represents space in an intuitive way, the Gantt chart represents time in a natural way. Both visualizations have their own strengths and weaknesses, in that they provide more or less support to users during spatial-temporal data exploration tasks. Different exploration tasks are introduced for spatial-temporal data, which can be questioned with respect to what, where and when. Data exploration tasks are formulated as statement queries with different foci. The queries relate to different search levels and cognitive operations. The details of the queries will be explained in section 2.3. These queries are part of the empirical usability evaluation reported on in this thesis. The concept of usability will be explained and the connection between objective measures (response accuracy and response time) and subjective measures (preference, aesthetics and insightfulness) will be discussed in the context of evaluation studies of user interfaces and of visualizations (Sections 2.4 and 2.5).
2.1 Visualizing spatial-temporal data
2.1.1 Mapping of data for visual representations
Wang (1995) shows that visual communication consists of three parts (1) the graphical domain (the picture); (2) the application domain (the problem and (3) the link (semantics of the picture).
FIGURE 1: THREE PARTS OF VISUALIZATION ACCORDING TO WANG (1995)
Figure 1 visualizes these three parts. The graphical domain contains all the graphical objects used in the visualization of the problem, which belongs to the application domain. The link that connects these two domains plays an important role, because it maps spatial information of the graphical domain to conceptual information in the application domain. According to Wang (1995), the link has to satisfy certain conditions to be helpful for the viewer to interpret the picture correctly. First, the link should associate the picture (graphical domain) with the subject matter in the application domain in a way which people view as natural. Second, the link should associate pictures with the subject matter in the application domain in a way which is not misleading. The link is misleading when the graphical objects in the graphical domain allow people to make inferences that do not correspond to the data in the application domain.
12
2.1.2 Visualizing objects, time and space
Spatial-temporal data are data involving space, time and objects. These three data types can be mapped to various different visual representations. In this thesis, we restrict these visual representations to two-dimensional static ones, in which all data get a spatial mapping, since all concepts and relations in the application domain are mapped onto a two-dimensional space, which can be a paper, a black or white board, or a computer screen.
Visualizing objects
Objects form one of the components of spatial-temporal data. In Kessell and Tversky’s (2010) study, objects are people moving through space over time. These objects are called agents. Several types of objects exist. Objects can be defined as physical objects, people, abstract concepts, tasks etc. (Kessell & Tversky, 2010, Kriglstein et al., 2016, Kriglstein et al., 2013). To refer to an object in visualization, different graphical objects, also called marks in the literature of information visualization, can be used. When there is an iconic relation, the mark in the visualization has a resemblance to the object that it refers to. Pictograms are an example of this kind of mark (van de Broek, 2010). More abstract marks can be used as well (Bertin, 1983). Such marks have a specific shape, which can be a circle or a rectangle. Marks can also get a color. Kessell and Tversky’s (2010) study used dots with different colors to mark different agents in one of the visualizations of spatial-temporal data that was part of their experiment.
Visualizing time
Time is a complex and a highly abstract concept (Kriglstein, Pohl & Smuc, 2013). Time can be visualized in different ways, as time, leading to animated, dynamic pictures, or as space, leading to static pictures. In this study, time gets a static spatial representation. Space is used as a metaphor of time. Time is naturally visualized as a line or as a cycle. Timelines are a powerful metaphor for visualizing events in a chronological order (Kriglstein, Phol and Smuc, 2013). Another natural way to map time is by using a cycle. A clock is an example of such a visualization. Similarly, the seasons are repeated in the same order, year after year, and are thus naturally represented as a cycle. When visualizing time as a timeline there are two options: time can be visualized as a time point on the timeline, or time can be visualized as an interval between two time points on the timeline (Kriglstein, 2013). The usage of time intervals is helpful in identifying the duration of events (Kriglstein et al. 2013). Therefore the usage of time intervals has an advantage over single points in time. This study will map time to a timeline consisting of chronologically ordered time intervals. A linear view is perceived as natural by most people, as it corresponds to a writing direction, which is from left to right for Europeans, but from right to left for people of the Arabic world (Boroditsky, 2001; Clark, 1973; Kessell & Tversky, 2010; Tversky et al. 1991).
Visualizing spatial data
To map space to a graphical object, it is important to see how spatial data are linked to the real world. Tversky (1981) points out that space can be mapped to objects that have both a horizontal and vertical orientation. The horizontal or east-west axis corresponds to the horizon and the vertical or north-south axis to gravity. As can be seen in Figure 2, the latitude is mapped vertically and the longitude is mapped horizontally. The latitude and the longitude are both used to localize a place. The most natural visualization for spatial data is a map representation. In maps, spatial information can be shown intuitively (Kriglstein, Haider, Wallner and Pohl, 2016).
13 FIGURE 2: AN EXAMPLE OF A MAP REPRESENTATION
2.2 Map and Gantt chart
Lohse et al. (1994) propose a classification of visualizations based on exploratory research of a large set of schematic visual representations. Among the main categories identified are graphs, tables, maps, diagrams, networks, icons, and time charts. Time charts display temporal data, they differ from tables in their emphasis on time. An example of a time chart is a Gantt chart, which is one of the visualizations focused on in this study. The other visualization focused on in this study is the map representation. Maps are symbolic representations of physical geography.
2.2.1 Map representation
A map representation is the predominant visualization to represent spatial data. Maps make it easy for people to identify the location of an object. Maps have a metaphor with the actual geographical position (Kriglstein et al., 2016), which makes them an intuitive representation of space. However, map visualizations also have a weakness. Maps are not optimized for the representation of time. A map easily gets cluttered, when several points in time are represented both statically and spatially (Kriglstein et al., 2016, Tversky, 2010). Figure 3 provides an example of the map representation used in Kriglstein et al.’s (2016) study.
FIGURE 3: A SIMPLE MAP VISUALIZATION DESIGNED BY KRIGLSTEIN ET AL. (2016)
Kriglstein et al. (2016) study made use of four different visualizations of spatial-temporal data, a map-based visualization, a Gantt chart, and two matrix-map-based visualizations. The map-map-based representation encoded (1) the location of a person, and (2) the time interval(s) associated with that person at that location, as shown in Figure 3. The location maps to a circle at the geographic x- and y-coordinates associated with that location. Each person is represented with a unique color in the circle of the location visited. If locations are visited by different persons, the circle is split into evenly sized sectors, in which each sector represents a visit of a person (Kriglstein et al., 2016, p. 237). Times of visit are represented as numeric annotations next to the location circles.
14 Kriglstein et al. (2016) designed an experiment in which participants were asked to perform tasks about group meetings with all four visualizations. One of the outcomes of their experiment is that participants perform worst with the map. This result was expected by the researchers since the map visualization involves the least appropriate mapping of time (Kriglstein et al. 2016, p. 247).
2.2.2 Gantt chart
A Gantt chart provides an accurate way to represent temporal data, as it involves timelines (Kriglstein et al., 2016). Due to the popularity of Gantt diagrams, the learning effort of this type of visualization is relatively low (Kriglstein et al. 2013). The Gantt chart presents temporal data from left to right. The Gantt chart is originally designed for visualizing the temporal scheduling of project tasks and subtasks. Figure 4 shows an example of a simple Gantt chart, in which the orange bars visualize the tasks that have to be performed within the time intervals indicated horizontally.
FIGURE 4: AN EXAMPLE OF A GANTT CHART
According to Kessell and Tversky (2010), line visualizations such as Gantt charts can be used very well for displaying a combination of time, space and agents. Kriglstein et al. (2016) also used a Gantt chart in their study. They did an experiment in which participants had to identify the moving of groups through time and space. An example of the Gantt chart as used by Kriglstein et al. (2016) can be seen in Figure 5. Notice that the Gantt visualization is rather abstract, and not optimized for the representation of space.
FIGURE 5: A SIMPLE GANTT CHART DESIGNED BY KRIGLSTEIN ET AL. (2016)
The x-axis in the Gantt chart represents one-hour intervals and the y-axis represents the locations, listed alphabetically in this specific design. Kriglstein et al. (2016) concluded from their research that the Gantt chart performed better than the map visualization, but worse than the other two matrix-based visualizations.
15
2.3 Querying spatial-temporal data
Spatial-temporal data can get different visual representations. A map-based representation and a Gantt-based one are two options among various other possible representations. It is not clear which visual representations are the most suitable to use to make sense of spatial-temporal data. It is important to understand what kind of information retrieval tasks are supported by which visualizations of spatial-temporal data. Adrienko et al. (2003) give a clear overview of exploration tasks that are relevant for spatial-temporal data. They introduce the question types and reading levels, which they prefer to call search levels, proposed by Bertin (1983) for visualizations of arbitrary data. Question types refer to components in data. According to Adrienko et al. (2003), the question types can be divided based on two main cognitive operations; identification and comparison. Furthermore, Bertin (1983) defines three reading levels, namely the elementary, intermediate and overall reading level. The reading level indicates if a question refers to a single data element (elementary level), a group of elements (intermediate level) or all elements (overall or superior level) involved in the type(s) of data visualized.
Peuquet (1994, discussed by Adrienko et al., 2003) restricts Bertin’s propositions to spatial-temporal data which involve three components: space (where), time (when) and objects (what). She distinguishes three basic categories of questions that may be asked about these components (Adrienko et al., 2003, p. 508):
When + where what: Describe the objects or set of objects that are present at a given location or set of locations at a given time of a set of times;
When + what where: Describe the location or set of locations occupied by a given object or set of objects at a given time or set of times;
Where + what when: Describe the time or set of times that a given object or set of objects occupied a given location or set of locations.
Table 1 shows the complete collection of question types, their reading levels and cognitive operations associated with them. The last column refers to the data type in focus, which can be either a location, a time or an object. A more extensive explanation of the querying of spatial-temporal data can be found in the master thesis of Houben (2020).
TABLE 1: QUESTION-TYPES, THEIR READING LEVELS, THE COGNITIVE OPERATIONS AND THEIR FOCUS (HOUBEN, 2020) Question type Reading level Cognitive operation Focus when + what where
Elementary Identify Location
where + what when
Elementary Identify Time
when + where what
Elementary Identify Object
when + what where
Intermediate Identify More than one location
where + what when
Intermediate Identify More than one time
when + where what
Intermediate Identify More than one object
when + what where
Overall Compare All locations
16 when
when + where what
Overall Compare All objects
2.4 Usability of visualizations
2.4.1 Usability Engineering
Usability of user interfaces of computer systems can be defined as the effectiveness, efficiency and satisfaction with which specified users can achieve specified goals in particular environments (Faulkner, 2000, referring to the ISO). Faulkner (2000) discusses how effectiveness, efficiency and satisfaction can be measured in usability studies. Effectiveness can be defined as the accuracy and completeness with which users achieve specific goals with the system. There are different ways to measure effectiveness of computer systems:
- The success and failure ratio in completing selected tasks
- The frequency of use of various commands or of particular language features/functions - The measurements of user problems
- The quality of the output
Efficiency is described as the accuracy and completeness of goals in relation to resources provided by the system and the environment. The following techniques are possible to measure efficiency:
- The time required to perform selected tasks
- The number of actions required in order to perform a task - The time spent looking for information in documentation - The time spent using online help
- The time spent dealing with error
The measures obtained for effectiveness and efficiency are objective rather than subjective. Measuring user satisfaction is a challenging problem however. Satisfaction is defined as the comfort and acceptability of the system (Faulkner, 2000, p. 115). It is nearly impossible to objectively measure satisfaction. Rating is an often used measurement, however, this is difficult since human response is very varied. When rating an interface on a ten points scale, one person’s 10 could likely be another person’s 7. However, a useful measurement of user satisfaction can be made if the measurement is based on observations of user attitudes towards the system (Faulkner, 2000). A user’s attitude can be measured with a questionnaire. In such questionnaires, the user can rate his or her attitude towards a system on a Likert scale (for instance on a five points Likert scale where 1 stands for a very low score and 5 for a very high score). For user interface studies a special questionnaire has been developed, the Questionnaire for User Interaction Satisfaction (QUIS).
Measuring the usability of a user interface is different from measuring the usability of visualizations. The most important difference is that visualizations are often static, unlike most user interfaces. User interfaces, if well designed, give immediate informative feedback on a user’s action. This feedback can help a user to perform his or her task successfully. When performing a task with a static visualization, a user does not know if the task has been completed successfully. Measuring the performance of visualizations is therefore also different. In visualization studies, satisfaction with visualizations is often referred to as preference for one visualization, since most studies evaluate two or more visualizations. The next section will explain how task performance and user preference is measured in many visualization studies.
17
2.4.2 Measuring usability of visual representations Measuring task performance
Usability of visualizations is often measured by looking at the task performance of users with these visualizations. The user can perform multiple tasks. In the context of usability of visualizations, it is common to use information retrieval or data exploration tasks. These tasks are given to users to execute with specific visualizations. Task response accuracy and task response time are often used to measure task performance. Kessell and Tversky (2010) measure the usability of visualizations with the aid of a questionnaire containing query statements. In this study subjects have to decide whether a statement is correct or incorrect, while using specific visualization variants. At the same time it was recorded how long it took subjects to verify or falsify these statements. Kriglstein et al. (2016) asked subjects to perform three tasks to identify groups in visualizations. The tasks and visualizations were distributed online. Among other things, response accuracy was measured. Response time was not measured. Roberts et al. (2013) measured the usability of metro maps in three different experiments. In experiment 1 they gave subjects 2 metro maps and let them plan 8 journeys for each map. Their focus was mainly on response time, it was measured how fast subjects could plan a successful journey with different designs of metro maps. In experiment 3 Roberts et al. (2013) gave all subjects three designs of metro maps of Berlin. The subjects had to plan journeys with all three maps. In this experiment, Roberts et al. (2013) measured the response accuracy and response time. The researchers recorded how many mistakes were made in journey planning and how fast subjects could plan a successful journey.
Evaluating subjective preference
To measure the satisfaction with a visualization, subjects are often asked to rank visualizations or to fill out a questionnaire. Kessell and Tversky (2010) asked for 11 statements taken from the task, which visualization was preferred. Subjects were forced to choose one of two visualizations. This is a method often used in visualization studies. Researchers test one or more visualizations and force subjects to make one single choice. Roberts et al. (2013) also used this method in their first experiment. In experiment 2 and 3 Roberts et al. (2013) developed an extensive questionnaire, in which they asked the subjects to rate the visualizations based on perceived usability and attractiveness. An example of this questionnaire can be found in Appendix H. Kriglstein et al. (2016) asked subjects to rank the visualizations from least preferred to most preferred.
2.5 Relation between subjective user satisfaction and objective task performance
How user satisfaction or user preference is related to task performance is a recurring question addressed in both user interface and visualization evaluation studies. The question can be formulated in a two-fold way: does satisfaction / preference affect task performance, and how, or the other way around. This section will discuss user interface and visualization evaluation studies that did research on the relationship between objective task performance and subjective user satisfaction.2.5.1 Connection of subjective user preference and objective task performance in HCI studies
Based on a meta-analysis of usability studies of human computer interfaces from 1982 to 1991, Nielsen and Levy (1994) conclude that there is a connection between performance and preference. They included studies in which performance was measured by the time taken by users to complete tasks, and preference was measured by asking users’ opinion after performing the tasks. Furthermore, Nielsen and Levy (1994) state that preference can be seen as a function of performance. This statement is based on the assumption that causality exists from experienced performance to expressed preference. One can also consider performance as a function of preference. This idea is based on the assumption that, if people like a system, they perform tasks better with it (Nielsen & Levy, 1194).
18 In visualization studies there is no agreement about the relation between subjective user preference and objective task performance. In visualization studies it has often been shown that it is very difficult for users to see if their performance with the visualization is successful, since most visualizations do not interact with users. This is also supported by the study of Morse et al. (1998) who did research on information retrieval interfaces, such as text lists, an icon list, a table, a graph and a ‘spring’ display. They did not find a connection between task performance and user preference. Most participants preferred visual representations over textual representations, even though participants performed better on the textual representation.
2.5.2 Connection of subjective user preference and objective task performance in visualization studies
Dissociation between user preference and task performance
In visualization studies the results regarding the relation between user preference and task performance are divergent. A large group of researchers are unable to find any relation. In three experiments, Roberts et al. (2013) show the absence of any association between user preference and task performance. In the first of three experiments, Roberts et al. (2013) examined the usability of an octolinear and a curvilinear design of a metro map. They come to the conclusion that even though participants perform significantly better on a curvilinear metro map, their preference is divided among the different designs. In the second experiment Roberts et al. (2013) wanted to know the reason of this preference. It was expected that the design that was experienced as most attractive was also seen as the most usable one. They asked participants to answer questions in which the perceived usability and attractiveness of the design of metro maps was measured. The octolinear design received both the highest perceived usability and the highest attractiveness score. However, a relation between the perceived usability and attractiveness was not found. While the curvilinear design had a high score on attractiveness, the design had a low score on perceived usability. And while the multilinear design had a high perceived usability, the design had a low score on attractiveness. In their third experiment Roberts et al. (2013) looked at the relation between usability ratings and objective task performance measures. Participants were asked to perform tasks with the help of different map designs. Based on this experiment, the researchers conclude that there is no difference in the objective measures, but that the usability ratings are different for the designs. Based on the results of the three experiments executed, Roberts et al. (2013) conclude that, before performing tasks, people already have a preference for a design and that task performance does not affect this preference. According to Roberts et al. (2013), preference is based on a combination of (1) expectations with regard to design rules, and (2) ideas about attractiveness of maps. Furthermore, the researchers discuss in their article that there is dissociation between objective measures and subjective measures, since there are limited cues for whether a performance is successful or not. Therefore it is inevitable that visualizations will be evaluated according to their superficial surface properties.
Hegarty, Smallman, Stull and Canham (2009) show similar results. They, too, did not find any connection between user preference and objective task performance measures. Their study reveals that users prefer extensive map representations, in which realism is included. Adding realism to weather maps has a disadvantageous effect on the task performance of naïve users. Adding realism increases the completion time and decreases the answer correctness.
19 Association between user preference and task performance
Even though the previous examples indicate that a connection between preference and performance does not exist, some researchers have found a match between participants’ preferences and their task performances. Kessell and Tversky (2010) performed an experiment in which they used two different forms of information visualizations, one with lines and one with dots, both displaying spatial-temporal data. Although the results of this study indicate that a connection between preference and performance exists, it is not clarified why participants prefer a certain visualization over another. Kriglstein, Haider, Wallner and Pohl (2016) also researched, among other things, the performance and preference based on a set of different information visualizations of spatial-temporal data, among which the Gantt chart and a map representation. They also observe a match between preference and performance. Their research shows that three of the visualizations used in their experiment (Augmented Matrix, Matrix
and Gantt) performed better than the fourth visualization, a map representation. These findings are in
line with the preference of participants. The map was preferred the least. According to Kriglstein et al. (2016), participants found it difficult to perform the tasks, involving group meeting detection, with the map. The Gantt chart, on the other hand, performed better than the map, because participants liked the readability of the Gant chart. At the same time, participants indicated that they had a hard time discerning locations in this visualization type.
Britton et al. (2002) investigated the relationship between user preference for sequence or collaboration diagrams in UML and accuracy of understanding information contained in these diagrams. Results of this study show that user preference for one of the two diagrams types, before carrying out the task, was not reflected in an improved performance with that type of diagram. However, after carrying out the task, user statements about which type of diagram they preferred to work with, were matched by improved performance with that type of diagram.
2.5.3 Types of users
Toker, Conati, Carenini and Haraty (2011) indicate that the literature on objective and subjective evaluations of visualizations can be conflicting and inconclusive. Toker et al. (2011) believe that this may be attributed to the fact that visualizations are very often designed without consideration of different user characteristics. Different user characteristics might explain why some studies do find a connection between subjective user preference and objective task performance and others do not. Toker et al. (2011) focus on four user characteristics: perceptual speed, verbal working memory, visual working memory and user expertise. In their study, it is emphasized that certain user characteristics do have a significant effect on task efficiency, user preference and ease of use. They conclude that perceptual speed influence the actual performance, whereas visual working memory and verbal working memory influence the subjective preference and ease-of-use. Hegarty et al. (2009) also indicate that user characteristics should be taken into account in evaluations of visualizations. Many evaluations use expert and not naïve (novice) users. They point out that, even though experts prefer simple systems, naïve users prefer displays that simulate the real world. Novice users differ from expert users in the way they perform and evaluate visualizations. According to Roberts et al. (2013) novice users will evaluate a visualization based on its superficial surface properties rather than on more subtle aspects of its design contributing to its usability. Novice users have a bias towards the most attractive visualization, considered from an aesthetic point of view, or the visualization they are most familiar with. In line with the findings of Roberts et al. (2013), experts seem to be better in predicting their perceived performance, since they pay attention to subtle aspects in the visualization designs. Kriglstein et al. (2016) have also found a connection between task performance and preference. In their study their group of participants consists of computer science students; this group
20 is probably more familiar with different designs of visualizations, and is better in predicting their performance with them.
2.6 General concluding remarks
On one hand, researchers report a dissociation between preference and task performance (Roberts et al., 2013; Hegarty et al., 2009). On the other hand, there are also researchers who report a connection between preference and task performance (Kessell & Tversky, 2010; Kriglstein et al., 2013; Britton et al., 2002). Kriglstein et al. (2013) and Kessell and Tversky (2010) find an alignment between performance and preference with respect to visualizations of spatial-temporal data.
Since there is no agreement about the connection between performance and preference within the visualization literature, this research attempts to contribute to already existing literature (i) about the usability of visual representations of spatial-temporal data, by investigating the usability of two popular spatial-temporal visualizations: the Gantt chart and a map representation, and (ii) about connections between subjective preference and objective task performance in general, by exploring whether any match can be found in the specific cases of a Gantt chart and a map, representing spatial-temporal data. Nielsen (1994) specifies that there are two ways in which the connection between performance and preference can be viewed. Performance can be seen as a function of preference and preference can be seen as a function of performance. Both perspectives will be analyzed in this study.
21
3. Methodology
3.1 Design
In order to answer the research question and the subquestions (see section 1), an empirical experiment was set up. To minimize random noise there was chosen for a within- subject design. All participants were exposed to two different visualizations of spatial-temporal data. The independent variables in this study are the visualization type (Map/ Gantt chart), the complexity of the visualizations (simple/complex) and different types of data exploration tasks. A questionnaire was created in Qualtrics, in order to collect data for quantitative data analysis. The questionnaire included thirty-two query statements; the statements focused on either time, object or location (see section 3.3.4). The participants were asked to verify the query statements with the aid of a scenario depicted in the map representation or Gantt chart. For each query, response time and response accuracy were recorded. After the digital questionnaire, a face-to-face posttask interview with the participants took place. The interview focused on the subjective experiences of participants, while performing taks with the two visualization types. Participants were asked to indicate which visualization they preferred and to substantiate their preference. The experiment took place on a one-to-one basis in a controlled setting.
3.1.1 Operationalization of task performance
Task performance was measured on the basis of effectiveness and efficiency. The number of correct answers measures the effectiveness. The more correct answers a visualization get the more effective the visualization is. Efficiency is measured by the time it takes to complete a task; the faster a task is completed the more efficient the visualization is. To measure the time needed to complete a task, a page-timer was used (subsection 3.3.5). Effectivity will be referred to as response accuracy and efficiency will be referred to as response time.
3.1.2 Operationalization of preference
To measure the satisfaction with the visualizations a semi-structured interview was set up. Right after the participant completed the digital questionnaire, the participants were asked to state which visualization had their preference. They were first asked to make one single choice. After they stated their overall preference, the interview took place. The participants were asked to choose which one of the visualizations they thought was the most insightful and which one the most attractive. Again, the participants were requested to make a single choice. For all questions the participants were urged to motivate their choice (see subsection 3.3.6).
3.2 Participants
Participants were personally asked by the researchers. They were sampled by convenience. Most participants who agreed to participate in the experiment were familiar with the city of Groningen. All are native Dutch speakers, with a normal reading proficiency. This is important for the basic understanding of the tasks they had to perform in the experiment. There were no other requirements for participating in this experiment. In total forty (n=40) participants were enrolled in this experiment. Table 2 gives an overview of the characteristics of the participants participating.
22 TABLE 2: CHARACTERISTICS OF ALL PARTICIPANTS
N (%) Gender Male Female 18(45) 22(55) Age <25 26-35 36-45 46-55 55-65 65+ 19(47.5) 4(10) 3(7.5) 5(12.5) 7(17.5) 2(5) Level of Education None High School MBO HBO bachelor WO bachelor Master PhD 0(-) 2(5) 9(22.5) 17(42.5) 2(5) 10(25) 0(-) Current Situation Student Full-time job Part-time job Entrepreneur Looking for a job Housewife/man 16(40) 8(20) 12(30) 1(2.5) 0(-) 3(7.5)
Familiarity with city of Groningen Very unfamiliar Unfamiliar Neutral Familiar Very Familiar 1(2.5) 2(5) 6(15) 16(40) 15(37.5) Color blindness Yes No 0 (-) 40(100)
3.3 Materials
For this research, different types of materials were created. Two different types of visualizations were designed; a Gantt chart (see section 3.3.2.1) and a map representation (see section 3.3.2.2). Both visualizations focused on time, location and object, but the data were mapped differently in both visualizations. The design of the visualization types are based on the research of Kriglstein et al. (2016). In this section, it will be explained how the variables stated above were displayed in the
23 visualizations, how the visualizations have been created and the variations that are made of the standard visualizations. Furthermore, this section will explain the typology of the query statements and how these statements were formulated (section 3.3.4). After answering the query statements, participants were asked to answer demographic questions (section 3.3.5) and after that they were asked to participate in an interview (section 3.3.6). Section 3.3.7 will explain which tools were used to carry out the experiment.
3.3.1 Mapping of data in visualizations Data in the application domain
In the application domain of the visualizations, three types of data were central: Time, Location and Objects (or Agents). The variables are categorical (people and places) and interval (time). The data were created based on logical scenarios. The data consists of people (objects) changing locations over time. The data were entered randomly in Gantt charts. With these data, it has been taken into account that people cannot be at two locations at once. A person can be absent from the data at a random point in time. This section will explain more about the variables used in the visualizations and how they were represented in the visualizations.
Objects
This study makes use of persons as objects. Persons were chosen because of the recognizability with location changing. The designed visualizations represent twelve different persons. The names chosen for the persons are short and distinguishable from each other, to eliminate the chance of confusion with persons as a confounding variable. On both of the visualizations the names of the persons are shown in a legend on the right top corner, each name has a corresponding color (Figure 11). At some points in time persons can be absent from the visualization, this means that they are not present at either one of the locations.
24 Time
A time span of twelve hours has been selected, from 9 AM till 9 PM. This particular time span is chosen to correspond to a realistic day planning of a person. In the Gantt, time is represented as a linear interval on the x-axis, as can be seen in Figure 7.
FIGURE 7: MAPPING OF TIME IN THE GANTT CHART
In the map, time is represented in a pie chart. The size of the segments shows proportionally the duration of the time a person is on that particular location (Figure 8).
FIGURE 8: MAPPING OF TIME IN THE MAP REPRESENTATION
Location
Twelve locations were selected to be represented in the visualizations. The following criteria were used to select these locations: (1) All locations have to be in the city center of Groningen; (2) All locations have to be recognizable for the participants, therefore commonly known locations were selected; (3) All locations should be distinct from each other. In the Gantt chart, locations are listed alphabetically on the Y-axis, as shown in Figure 9, which represents only part of the Gantt chart created.
25 In the map-based representation, the locations are shown as circles on a geographic map of the center of Groningen. The name of the location is below the circle (see Figure 10).
FIGURE 10: MAPPING OF LOCATIONS IN THE MAP REPRESENTATION
Table 3 shows all the variables and their corresponding values presented in the visualizations. TABLE 3: VARIABLES AND THEIR CORRESPONDING VALUES
Variables Values
Locations (name of location) Academiegebouw, A-Kerk, Concerthuis, Grote Markt, Harmoniegebouw, HEMA, Martinitoren, Museum, Pathé, Politiebureau, Station, Vismarkt Time (timespan of one hour) 9-10, 10-11, 11-12, 12-13, 13-14, 14-15, 15-16,
16-17, 17-18, 18-19, 19-20, 20-21
Agents (name of person) Arie, Bert, Dewi, Jim, Julia, Kim, Lieke, Lisa, Ron, Roos, Sam, Stefan
3.3.2 Visualizations
The visualizations used for this experiment were a Gantt chart and a Map, both visualizations display spatial-temporal information, that represent persons’ daily location changes. The spatial-temporal data represented in the visualizations were created based upon the guidelines stated in section 3.3.1, the actual data that represents on which time agents are on a specific location is randomly represented in the visualizations. It has been taken into account that an agent can only be on one location at a time. It is possible that an agent is at a certain time not present at either one of the locations. This has been done so the Low Complex (simple) visualizations would not have been too easy to interpret and so the High Complex visualizations would not have been too complex to interpret. Appendix A shows all the visualizations used during the experiment.
Gantt chart
A Gantt chart is often used to illustrate a project schedule or timeline. For this research, the Gantt chart has been transformed to fit the spatial-temporal data. To design the chart Microsoft Excel was used. As illustrated above, the x-axis represents a linear timeline (9:00 to 21:00) with intervals of one hour and the y-axis represents the locations in alphabetical order. To represent a person at a certain location from a certain timespan, colored bars are used. Figure 12 shows an example of the Gantt chart used.
26 FIGURE 11: EXAMPLE OF THE GANTT CHART USED
Map representation
The map that has been designed for this research, as shown in Figure 12, is a realistic map of the city center of Groningen, obtained by Google Maps. The map has been created with the use of Adobe Illustrator CC. To create the map with the spatial-temporal data all labels and other distracting elements have been removed from the map. Next, the spatial-temporal data have been put in the map, the locations on the map have been marked with their name and each location has a chart. The pie-chart shows the people present at the location, by showing their color in the pie-chart. The timespan has been written next to the segment representing a person and the size of the segment gives an indication of how much time a person is on a specific location. If a person visits a location multiple time, multiple time spans have been given next to the segment and the segments shows a collection of the total time spent.
27
3.3.3 Complexity of visualizations
To test if there is a difference between the Gantt chart and the Map when the complexity of the visualization is higher, we decided to create a variation in the data. A simplified version and a complex version were used. This variance in complexity is on the level of the application domain, which means that the maximum amount a person changes location differs among the complexity variances. In the simple version a person changes location maximum two times, while with the complex version a person can change a maximum of six times. Table 4 shows an overview of both complexities.
TABLE 4: DIFFERENCES IN LOW AND HIGH COMPLEXITY IN DATA
The maximum amount a person can change location
Simple 2
Complex 6
For the Gantt visualization this means that the visualization itself (Graphical domain) is being altered to fit the amount a person can change location, for the simple version 20 rows and for the complex version 40 rows. The map does not add extra elements to fit to the complexity, however the simple version only allows for a maximum of two persons at the same location at any given time being represented in the map.
3.3.4 Data exploration tasks
To perform the experiment 32 data exploration tasks were formulated as query statements. All statements were based on three elements, time, location and agents. The answer on the statements could be either juist (right) or onjuist (wrong), however to reduce the correction for guessing the option geen idee (no idea) was also given. For each complexity, five different types of data exploration tasks were created. Figure 14 gives an overview of the distribution of the data exploration tasks.
FIGURE 13: DATA EXPLORATION TASK DISTRIBUTION (HOUBEN, 2020)
All of the statements were in Dutch, since this is the native language of all participants. All statements were based upon research of Adrienko et al. (2003), more information about this approach can be found in section 2.2.3. Table 5 shows the categories created for the statement queries and shows the rules and order of the queries. In Appendix B, all statement queries used for this research can be found. More details about the statement queries can be found in the research of Houben (2020).
28 TABLE 5: QUERY-STATEMENT FORMULATION PER TYPE: THE NUMBER OF STATEMENTS IN CATEGORY, RULES, SENTENCE ORDER AND EXAMPLES (HOUBEN, 2020)
Query Statement category
N
Object in focus 8 Rule
Order Example
Always one object, one location and a timeframe*
object – timeframe - location
Julia is vandaag** in ieder geval van 11 tot 12 uur bij de Pathé
(Julia is today at least from 11 a.m. to 12 a.m. at the Pathé) Time in focus (multiple objects) 4 Rule Order Example
Always a timeframe, one location and multiple objects
timeframe - objects - location
Van 12 tot 13 uur zijn Lisa en Julia op de Grote Markt.
(From 12 to 1 pm Lisa and Julia are at the Grote Markt) Time in focus (cardinality) 4 Rule Order Example
Always a timeframe, one location and cardinality of objects
timeframe - objects - location
Van 15 tot 16 uur zijn er precies 3 personen op de Vismarkt
(From 3 to 4 pm there are exactly 3 people at the Vismarkt) Synchronization on locations 8 Rule Order Example
Always a location and cardinality of objects (no timeframe is given)
location - object (N)
Bij de A-kerk komen vandaag precies 5 personen
(Exactly 5 people visit the A-Kerk today) Sequence of
objects
8 Rule Always one object and two locations. Locations are ordered linear in time
29 Order
Example
object - locations
Bert komt vandaag in ieder geval in de HEMA en op het station
(Today, Bert will at least visit the HEMA and the station)
* Timeframe is one hour between 9AM and 9PM, i.e. from 10 till 11AM. ** Today is used to create less ambiguity about the timeframe
