children's navigation of hyperspace – are spatial skills ... - CiteSeerX

CHILDREN’S NAVIGATION OF HYPERSPACE – ARE SPATIAL SKILLS IMPORTANT? S. Jones and G. E. Burnett School of Computer Science and Information Technology The University of Nottingham, UK {sjj, geb}@cs.nott.ac.uk

ABSTRACT Hypertext is becoming increasingly popular as a platform for educational material, allowing the user autonomy and flexibility in choosing a route through the presented information. However, the required decision-making process places extra cognitive demands on the user, and this may result in disorientation and the phenomenon known as “lost in hyperspace”. Individuals with high spatial ability appear to demonstrate superior navigational skills within hypertext, completing tasks more quickly and with fewer errors then those with low spatial ability. They tend to form more accurate internal representations, or cognitive maps, of hypertext systems that correspond better to the underlying physical structure. Little research has been carried out with children to assess their formation of cognitive maps of hyperspace. In this study, 32 children aged 10-11 years from a primary school in the UK were given search tasks to complete on an environmental Web site. Various measures were made of their navigational efficiency, their degree of lostness, and their ability to complete a map of the routes they had traversed. Those with high spatial ability completed the tasks in shorter time, became lost less frequently, and completed the maps more accurately. This paper discusses the implications of these results to the success of hypertext learning environments for learners with low spatial ability. KEY WORDS Hypertext, spatial skills, cognitive maps, web-based education, children

1. Introduction Educational material is increasingly being presented in the form of hypertext for students to locate information and problem solve. Some of the benefits of hypertext are that readers have several options and alternative routes through the information to choose from, and the autonomy to make the choice rather than being forcibly directed along one path as in a linear book-type model [1]. Hypertext may be envisaged as a platform for hosting multiple texts of the same topic and the resultant

interactivity has been shown to have a positive effect on learning outcomes [2]. However, these benefits come at a price. Similar to navigation in the real world, users may be considered to be navigating through “hyperspace”, the computer environment of hypertext documents [3]. With the availability of a choice of routes through hypertext comes extra cognitive demands, such as planning and decisionmaking, that are not normally required in a linear singleroute format. These demands require extra cognitive resources that may overwhelm the user and result in a sense of disorientation, a phenomenon known as “lost in hyperspace” [3,4]. A badly designed hypertext document will result in a higher number of users becoming disorientated. However, all things being equal, there are various measured individual differences between users that determine how successful they are at navigating around a piece of hypertext. Nielsen [5] analysed 92 existing hypertext studies and found individual differences to be one of the most important issues in hypertext usability. Age, gender, cognitive style, spatial ability and previous search experience have been shown to impact on success in hypertext navigation [1]. Spatial ability is a cognitive characteristic that gives a measure of the ability to conceptualise the spatial relationships between objects [6]. Individuals with high spatial abilities tend to excel in real-world navigational tasks, and there is an increasing collection of research looking at the impact of spatial skills on navigating in virtual environments, including hypertext. Westerman and Cribbin [7] had participants search for objects in a virtual representation of a database. They found that the low spatial ability group performed less well, taking longer and becoming lost more often. Keeble and Macredie [8] state that to browse hypertext effectively, users need good spatial ability so they do not become “lost in hyperspace”. Chen & Rada [9] carried out a meta-analysis of 23 experimental studies on interacting with hypertext. Users with high spatial ability completed tasks more quickly than users with lower spatial ability, and committed fewer errors on first attempts. Hook et al [10] looked at various cognitive abilities, and found that

spatial ability was the only one that impacted on task completion time in a hypermedia environment, those with high spatial skills finishing in a shorter time. It has been proposed that people take note of features in a real or virtual environment that then allows them to develop a “cognitive map” of that environment, and this facilitates negotiating a path to the destination. Cognitive maps have been defined as mental models of spatial information [11]. Edwards and Hardman [12] gathered data that suggest readers form cognitive maps of hypertext systems. Spatial ability is an important consideration, with those scoring high on spatial ability tests developing a more complete cognitive map of an environment, thus facilitating improved navigation and negating the need to access the content table as frequently [13]. Disorientation will result from the reader’s cognitive map of the hyperspace not corresponding sufficiently with the actual physical representation [7,14].

Little research has been carried out on how children develop a cognitive map of a hypertext system. Various studies have looked at how children search for information on the Internet [21], with some demonstrating gender differences in navigation patterns [22], but these studies concentrated on the use of search engines rather than the navigation of a single Web site. It has also been shown that girls demonstrate a higher propensity than boys for becoming lost and not knowing what to do next [23]. The current study carried out a hypertext navigation exercise with children in the UK. The aim was to determine if there were any relations between their spatial ability and various measures of navigational efficiency and effectiveness, including a measure of lostness, and scores achieved in construction of sketch maps as a marker of cognitive map formation.

Researchers have derived some idea of disorientation, or lostness, based on subjective feelings gleaned from questionnaires [12,15,16]. However, Smith [17] argues subjective responses are inadequate to gain a true measure of a reader’s lostness, and suggests a quantifiable measure of decline in performance is more relevant. The accepted measures of efficiency and effectiveness, such as task completion times and counting the number of errors, are inadequate for establishing the lostness of a user. More quantifiable lostness metrics have been developed based on parameters such as the number of nodes visited relative to the optimal path, repeat visits to nodes and the number of incorrect nodes visited [7,17,18].

2. Method

Various attempts have been made to understand a user’s cognitive map of an electronic environment. The difficulty is in translating an internal map into an external representation that is difficult to articulate. Some success in determining mental models of a hypertext system has been gained from card sorting exercises, with the nodes represented by cards with page titles, or screen dumps [18]. The cards are sorted to parallel the structure of the hypertext. Other groups have utilised sketch mapping, with lines joining boxes to represent the structure of the whole or part of a hypertext system [14,15,19]. The problem with this tool is that, without the use of a template, users produce very different sketches, due either to a range in drawing ability or varied understanding of how to represent the information. This makes quantifiable comparisons problematic. Sheard and Ceddia [14] classified the maps drawn by their participants into four groups (for example: single screen, series of screens) and could not give a weighting to the accuracy of the drawings within each group. Another issue is that, without cue cards, memory may be the variable that is being measured [19], and participants may be relying on semantic relatedness to compile the maps rather than truly understanding the networked nature of the information space [20].

Individual differences in spatial skills were measured using the Spatial reasoning tests devised by nferNelson [24]. The tests assess the child’s ability to carry out a range of cognitive tasks concerning shape, space and mental rotation, and provide a nationally standardised score with an average of 100.

2.1 Participants Participants consisted of 32 pupils from a primary school in Nottingham, UK. The pupils, comprising 16 girls and 16 boys between the ages of 10-11 years (mean age = 10.83 years, SD= 0.48), had similar reading abilities as measured by a standardised test, and comparable computer experience as measured by a questionnaire. 2.2 Materials

The children were given tasks requiring them to navigate round an environmental Web site called EEK! (“Environmental Education for Kids”, available at http://dnr.wi.gov/eek/). This site was chosen because it is designed for children of the age group to be studied, and consists of a large, highly networked structure with many cross-links between Web pages. Although the site was colour-coded for the different topics, there was no site map to aid navigation. The hypertext environment was configured so visited links were not indicated. This was so the children could not follow the path of a previous attempt or previous participant by picking the used links based on colour. The site had been downloaded onto a CD, so there were no issues with alterations to the site or with variations in download times. None of the children had visited the Web site before.

2.3 Procedure The spatial ability test was administered to the participants a week before the study commenced. The study was carried out with individual children in a separate room with no disturbances. They were given a maximum of 45 minutes to complete the exercise, enough time to finish all the tasks set, but were informed there would be a prize for the fastest time to completion. They were required to locate the answers to two familiarisation questions and four test questions without using the Web site search facility. All answers were at the top of the relevant Web pages and once they had located the answer, they had to inform the experimenter verbally. This was so writing speed was not introduced as an extraneous variable. The four test questions had more than one possible route through the site to the same answer page. After they had located one route, they were informed of how many routes there were and instructed to find the others. Once completed, they had to locate all routes to the four questions again as fast as possible. The whole exercise was screen recorded using the commercially available Camtasia software. Once the exercise was finished, the participants were asked to compile a map of the routes they had followed. The procedure was demonstrated for the first two test questions, and then the children completed the map for the remaining two questions without recourse to the Web site. The nodes were represented by small post-it notes with the title of the respective Web page printed on them. The post-its could be easily transferred around the chart, and were initially arranged in a random order (the same order for each participant), with some nodes being included that were not required for the map. After placing the nodes, the children represented the links with arrowed lines drawn with erasable ink.

locating the answers to the questions (L1) and the repeat exercise (L2). 2.4.2 Time taken and navigation speed The time taken to complete the four questions was measured for both the first pass through the site (T1), and the repeat visit (T2). Similarly, the average time per click was recorded (click 1 and click 2). 2.4.3 Layer traversal and backlooping The frequency of visits to the different depths in the site was recorded for the initial exercise. For the purpose of analysis, these results were divided into two layers: shallow (home, levels 1 and 2) and deep (levels 3,4 and 5). The frequency of use of the back button (back) was also scored.

3. Results 3.1 Map Scores and lostness The relationship between map scores and spatial scores was investigated. There was a strong positive correlation between the two variables (r=0.75, n=32, p