TY - JOUR
AU - Elgin, Peter, D
AB - Abstract This study examined users' schemata of hypermedia. It is frequently assumed that users' schemata contain spatial information about how the pages of a website are interconnected. However, it is not clear how these schemata could contain such information when none is presented to the user while he/she is exploring the website. Unfortunately, there has been little research addressing this assumption. Toward that end, the reported study examined the mental representations (i.e. schemata) acquired when using hypermedia by systematically varying the interconnections within a website while holding the information that the website contained constant. Analyses of 40 participants' drawings of the website's organization indicate that drawings largely reflected conceptual (i.e. semantic) relationships, and not the true nature of the website's interconnections. In light of this research, it is suggested that we reevaluate the conjecture that hypermedia is mentally represented in ways similar to the physical world. 1 The pervasiveness of hypermedia Hypermedia can consist of interconnected web pages, which may contain text, images, audio, and/or video. These web pages are linked to one another, so that after a user selects a link to another web page, then the information from the selected web page can be viewed. In this way, one can move non-linearly through a body of information to find the specific information that one desires in an efficient manner. Supported by the growth of the Internet, hypermedia has become increasingly important in our day-to-day lives. For example, military and civilian organizations are looking more and more to hypermedia-based solutions to information dissemination problems. In addition, colleges and universities are turning their attention to hypermedia-based courses, wherein students learn via course content presented on the Internet (Parlangeli et al., 1999). These examples demonstrate how hypermedia is becoming increasingly pervasive. Accordingly, the usability of hypermedia becomes a very important issue. One way to enhance the usability of hypermedia is by examining what information gets processed and subsequently retained when one explores a hypermedia-based system. 2 Exploration leads to schemata of hypermedia Users explore hypermedia in order to attain certain goals (e.g. find a particular piece of information). This exploration involves selecting links at various web pages that will ultimately lead the user to the goal-relevant resource. Based on the links chosen, a user will develop a schema, i.e. a mental representation, concerning that particular hypermedia network. Neisser's (1976) model of schema development provides a possible framework for this process. He describes a cycle that continuously modifies a schema of the environment. Specifically, he argued that one explores the environment in order to sample the information that is available about things that are in that environment. This information, in turn, modifies our schema of that environment. This newly modified schema then directs further exploration of the environment, thus completing the cycle. To understand how this might apply to a hypermedia-based system, consider a user who is trying to find product information on a corporate website. In this scenario, the user might explore the hypermedia system by selecting a particular link. For example, the user might select a link that is titled ‘Products’. Having done so, a new web page would be presented, which would contain new information, possibly presenting a list of the company's current products. This new information would, in turn, modify the user's schema related to the hypermedia system. For example, the user might see that the web page only contains links to information about currently available products. If the user is seeking information about an older product, she will now modify her schema to accommodate the fact that the desired information is not in that portion of the website. Finally, the user's newly modified schema will direct further exploration of the website. Specifically, she might select the back button to take her to the previous page so that she could choose a link to a different portion of the website. This general description captures the essence of the diverse set of user search and browsing strategies (Catledge and Pitkow, 1995). This example demonstrates the impact of a user's schema on the usability of that website. If the information presented leads the user to develop an inappropriate schema, then she will likely have difficulty finding the desired information. In addition, if this user has developed an inappropriate schema of the organization of the website, then even if she did happen upon that information, she might have a difficult time finding the information again at a later date. Thus, as certain authors have previously argued (e.g. Dillon et al., 1990), the schemata that users develop are critically important for hypermedia usability. 3 A spatial metaphor and its implications for schemata of hypermedia 3.1 Evidence for a spatial metaphor Hypermedia is generally considered to have spatial qualities, e.g. height and depth. This assumption permeates the writings of researchers who study the use of hypermedia. For example, Chignell and Waterworth (1997) have argued that “the overall metaphor of multimedia and hypertext is generally assumed to be spatial. In this spatially oriented view of multimedia, browsing is a process of navigating through some information structure” (p. 1817). Likewise, Dillon et al. (1990) state that although “theoretical work in navigation is primarily concerned with travels through physical space such as cities and buildings it does offer a perspective that might prove insightful to the design of hypertext systems, where navigation is conceptualized as occurring through an information space” (p. 588). In a similar vein, Stanton (1990) suggests “that the electronic environment supports the analogy of navigation based on the physical environment” (p. 170), which suggests that the spatial abilities we employ to navigate in the physical environment are useful for exploring a hypermedia-based system. Similarly suggestive comments can also be found in the work of other authors in this area (e.g. Nielsen, 2000; Mayhew, 1999). Thus, it appears that people working with hypermedia conceptualize it in spatial terms. 3.2 From metaphor to schemata Certain hypermedia researchers have taken the spatial metaphor one step further by proposing that users' schemata contain spatial information. Some authors have merely speculated that users' schemata may contain spatial information. For example, Dillon et al. (1993) provide a framework for how spatial schemata could develop through exploration of a website. These authors do, however, suggest that the users' schemata could contain non-spatial (e.g. conceptual) information as well. Other researchers, however, have been much more adamant in asserting that schemata contain spatial information. For example, Kim and Hirtle (1995) state that “research in human spatial processing and navigating in physical environments can be applied to the problem of disorientation in hypertext systems.” (p. 239). Likewise, Stanton and Baber (1992) argue that “it appears that individuals attempt to create representations in the form of survey-type maps for orienting and navigating around hypertext” (p.163). These sentiments are also echoed by other researchers in this area (e.g. Calvi, 1997; Edwards and Hardman, 1989). Thus, it appears that many researchers in this area have taken the spatial metaphor and assumed that users process information in ways that are consistent with that metaphor. 4 Our view We challenge the assumption that people process and retain spatial information when exploring hypermedia. Thus, we challenge the assumption that users' schemata of hypermedia systems contain spatial information. Specifically, we argue that hypermedia is inherently non-spatial. Physical environments have spatial qualities, such as depth and direction. These qualities allow an actor to physically move about within that environment. On the contrary, hypermedia does not have direction or depth. Some might conceptualize ‘moving deeper’ within a hierarchically organized website, however, such an abstraction is qualitatively different than depth in the physical environment. Probably the most vivid demonstration of the non-spatial nature of hypermedia pertains to its exploration. When one explores the physical environment, one physically moves from point A to point B. This movement provides actors with a continually changing point of view, which contains a great deal of spatial information (Gibson, 1966, 1979). On the contrary, one explores hypermedia by selecting a particular link, which results in a particular web page being brought forth (i.e. displayed). This occurs relatively instantaneously. Thus, there is no actual movement through any kind of space as a user explores a hypermedia network. Accordingly, this exploration cannot offer spatial information, only information that an instant change has occurred in response to the user's link selection. It is illogical to assume that users incorporate spatial information into their schemata of a hypermedia system, when no such information is present. To have such information be part of a schema, one would have to create the spatial information from an understanding of the spatial metaphor and then encode that artificially created spatial information as part of the schema. For example, a user would have to visualize the network of interconnections that exists within the hypermedia network and encode the metaphorical relative locations and depths of various pieces of information as part of his/her schema of that system. We feel that it is unlikely that most users would do this. However, as was noted previously, our view appears to be the minority, thus it behooves us to seek a test of our assertions. 5 Testing for spatial schemata Based on our contention, it is necessary to examine whether users do, in fact, form spatial schemata of hypermedia. To address this issue, one could ask whether the same mechanisms help people to develop spatial knowledge of the physical world and metaphorical spatial knowledge of hypermedia. That is, do manipulations that help users learn the spatial layout of a physical environment also help them learn the metaphorical spatial layout of a hypermedia network? Two manipulations that have been shown to improve spatial knowledge of physical environments are (1) providing individuals with maps of the physical environment and (2) allowing exploration of the environment (Thorndyke and Hayes-Roth, 1982; see also Schwartz and Kulhavy, 1988; Thorndyke and Goldin, 1983). 5.1 Map exposure Allowing someone to use a map of a physical layout helps to develop their spatial knowledge of that area (Schwartz and Kulhavy, 1988; Thorndyke and Hayes-Roth, 1982). This has lead some hypermedia researchers to suggest that designers should provide maps of their websites (e.g. Nielsen, 2000). However, the utility of such sitemaps for navigation is debatable (for evidence in favor, see Billingsley, 1982; McDonald and Stevenson, 1998; for evidence against, see Dias and Sousa, 1997; Stanton et al. (1992)). In addition, a few studies have looked at the effect of site maps on one's schema of a hypermedia system. These studies suggest that sitemaps do not necessarily improve users' mental representations. For example, Stanton et al. (1992) had participants draw their understanding of how a hypermedia network was interconnected and compared the drawings of participants that used a sitemap to those that did not use a sitemap. The drawings of participants who did not use a sitemap were more accurate than the drawings produced by participants who did use a sitemap. This suggests that using a sitemap might have actually impaired the development of a schema of that hypermedia network. However, a limitation of these studies is that participants have discretionary use of the sitemap. This would allow participants to use the sitemap as a cognitive prosthesis, which suggests that they might not have expended the effort to develop a schema. Thus, it is unclear whether poorly developed schemata are attributable to reliance on sitemaps, or because schemata of hypermedia are not fundamentally spatial and thus sitemaps are not helpful for their development. 5.2 Exploration In addition to map exposure, simply allowing someone to explore a physical layout helps to develop their spatial knowledge of that area (Thorndyke and Goldin, 1983; Thorndyke and Hayes-Roth, 1982). A few studies have examined the schemata acquired after exposure to a hypermedia network that did not include a sitemap (e.g. Gray, 1990; Shapiro, 1999). By not including a sitemap and allowing participants to freely explore a website, one can examine the schemata that are formed without concern about participants' reliance on a sitemap as a mental prosthesis. However, these experiments did not evaluate the accuracy of the participants' schemata. That is, were participants' schemata consistent with the actual structure of the interconnections of the hypermedia network? Thus, in order to assess whether exploration leads to the development of spatial schemata of hypermedia, more research is needed. 6 The reported experiment The purpose of the reported experiment was to determine whether users' schemata of a hypermedia network contain accurate information about the organization of the website's interconnections (hereafter referred to as its ‘connection-structure’). We examined this issue by allowing participants to freely explore a website that did not contain a sitemap. After this exploration period, we tested the participants' schemata for evidence of information about the connection-structure by means of a drawing task. The information on the website was held constant, but the number of levels within the connection-structure hierarchy varied between groups. This provided us with a means to determine whether participants were encoding information about the connection-structure into their schemata of the website. Schemata were then evaluated by asking the participants to draw the hypermedia system's connection-structure. We concur with Shapiro (1999) that using participants' drawings provides “only a snapshot of their conceptualizations at one moment in time” (p. 232). However, we believe that, for hypermedia, the drawing task is well suited for initially determining representational skills when compared to other task analysis techniques serving the same purpose (see Seamster et al., 1997, for an overview of such task analysis techniques). If the users base their schemata on the website's connection-structure, then the number of levels that they draw should increase with the number of structural levels that exist within the website. If they base their mental representations on non-spatial information, then there should be systematic errors in the number of levels that were drawn (i.e. errors of omission or commission based on the website's connection-structure). 7 Method 7.1 Participants Participants were 40 students (30 women and 10 men) at Kansas State University fulfilling credit for a general psychology course. They were all between 18 and 24 years of age (M=19.15) and had 1–8 years of experience using the World Wide Web (M=3.80). Of the total time they spent online, an average of 36% was devoted to using the World Wide Web. Although participants' experience using the World Wide Web was quite diverse, there were no systematic differences between treatment groups (see Section 8 for details). 7.2 Apparatus and stimuli A Pentium II PC, with a 19-in. monitor, displayed the stimuli. All web pages were saved and accessed on the system's hard drive in order to eliminate the possibility of inconsistent download times. Four test websites (1, 2, 3, and 4 level hierarchical structures) were developed. The test websites contained computer-generated images. These websites were modeled after existing websites that present such material (examples of these sites can be found by searching the World Wide Web using keywords such as ‘web design graphics’). Images were chosen as the sites content so that the participants would spend more time navigating the website and less time reading or studying text. For the test websites, the images were organized into five image type categories (i.e. bullets, backgrounds, lines, buttons, and clipart). Each of these categories contained 13 color categories (i.e. red, orange, blue, yellow, green, white, black, gray, brown, magenta, cyan, purple, and other), each of which contained 21 images for each image type/color combination. All experimental conditions featured the same images and were categorized in the same way. However, the way the content was divided between various web pages (i.e. the connection-structure) differed between conditions (see Fig. 1). The connection-structures of the websites for each condition and the web browser used are discussed separately in the following sections. Fig. 1 Open in new tabDownload slide Illustration of the procedures necessary to navigate the test websites with their overall connection-structure. Fig. 1 Open in new tabDownload slide Illustration of the procedures necessary to navigate the test websites with their overall connection-structure. 7.2.1 One-level hierarchy The website used for the one-level condition (1L) was a completely networked website with every page connected to every other. The content was organized by dividing each menu with nested tables into image types and colors (see Fig. 2); five tables for image types, each with 13 nested tables for colors, and 21 links (using the numbers 1–21 for the text) within each image type/color for the images. When participants selected an image number from the menu, a new page was presented with the selected image and an image label (e.g. ‘9008.tif’) at the top of the page. This structure had 1365 pages all featuring a different image. Fig. 2 Open in new tabDownload slide An example of the nested tables used in the test website. Fig. 2 Open in new tabDownload slide An example of the nested tables used in the test website. 7.2.2 Two-level hierarchy The two-level condition (2L) was the same as 1L except clicking a link changed the display to a page featuring only the selected image and a link labeled ‘Back’. Therefore, this was a 1×1365 page connection-structure (1 top-level node and 1365 bottom level nodes). 7.2.3 Three-level hierarchy The three-level condition (3L) had separate pages for each image type category, but colors were presented in nested tables after an image type category was selected. Therefore, participants selected an image type category from the first page, an image number from the next page, and then saw a page featuring the selected image and a ‘Back’ link. This yielded a 1×5×273 connection-structure (1 top level node, 5 category nodes, and 273 bottom level nodes). 7.2.4 Four-level hierarchy The four-level condition (4L) was the same as the 3L condition with the exception that it contained separate pages for each color category. Therefore, participants selected an image type category, then a color category, and then an image number. This series of actions resulted in the presentation of the page featuring the selected image with a ‘Back’ link. This yielded a 1×5×13×21 connection-structure (1 top level node, 5 image category nodes, 13 color nodes, and 21 bottom level nodes). 7.2.5 Web browser A simplified web browser (partially shown in Fig. 2) was created to eliminate (1) variation due to browser preference and (2) the opportunity to use the browser's address bar to infer the website's connection-structure. This web browser had no address bar or menu items. Participants could only interact with the browser through the use of the back, forward, stop, and reload buttons. In addition, this web browser made it impossible for participants to access websites other than the one they were instructed to navigate. 7.3 Procedure Participants were randomly assigned to one of four conditions so that all conditions contained 10 participants. They were asked to imagine that they were designing a website and that they were to find computer generated art to use on it. They were also informed that the computer display would be recorded while they explored the website. This recording was made to insure that all participants gained adequate experience with the website (i.e. they explored more than two image type categories) and to measure the number of end-nodes, or individual images, that they viewed during the session. Participants were informed that they would have 5 min to explore the website. This relatively short time interval was considered appropriate because this amount of time has been sufficient for the development of spatial schemata of physical spaces (e.g. McNamara et al., 1989; Thorndyke and Hayes-Roth, 1982), and therefore, should have been sufficient in the present experiment. In addition, because images were used instead of text, 5 min was considered ample time to explore much of the test website. When the directions were understood, the experimenter left the room and the participants began to freely explore the website. After the exploration period, the experimenter reentered the room and asked the participant to stop. The web browser was closed and the participant was given instructions for the drawing task (see Appendix A). To summarize these instructions, the participant was asked to draw the connection-structure of the website with shapes representing pages (i.e. network nodes) and arrows between the shapes representing the connections (i.e. links) between the pages. In addition, they were asked to label each shape with a descriptive word or phrase. It was emphasized that they were to draw what they visited and to make inferences about what they did not visit based on their experience with this website. They were explicitly told not to draw the screen layout, graphics, or other specific content, only to draw labeled nodes and links. They were also informed that they would be videotaped while drawing the map of the website. This videotape, however, was only used for exploratory purposes. After they acknowledged that they understood the directions of how to draw the connection-structure of the test website, they were tested to prove that they understood (see Appendix A for more details on this procedure). In the rare case they failed to prove their understanding, they reread the appropriate portion of the directions until they understood. After demonstrating that they understood the task, the experimenter gave them a pencil and a blank sheet and instructed them to begin. The experimenter then left the room and allowed them to spend as much time as they wanted to complete the task. 7.4 Data coding Two judges individually coded the drawings without prior knowledge of which condition a drawing was associated with. To code the number of levels represented, the judges counted the number of nodes along all possible paths from the top of the hierarchy (i.e. where participants indicated that they began) to the end-nodes (where each branch ended) until all possible paths were accounted for. Because the drawn connection-structures were asymmetrical for some participants, two measures came from this data coding: mean number of levels drawn and the maximum number of levels drawn ( and , respectively). In all cases, the two judges agreed upon the mean and maximum number of levels that were drawn (see Fig. 3 for an example of a drawing coded as four levels). Fig. 3 Open in new tabDownload slide A drawing of the test website's connection-structure made by a participant in the 1L condition. Fig. 3 Open in new tabDownload slide A drawing of the test website's connection-structure made by a participant in the 1L condition. Two error measures were also calculated for each participant. For the mean errors per participant (Merror), where Levelscondition is the number of levels in the website that the participant explored. Similarly, maximum errors per participant (Maxerror) was calculated as Both the Merror and Maxerror measures yield a positive value for errors of commission (i.e. the participant added a level to the structure) and negative values for errors of omission (the participant excluded a level from the structure). Both error measures and both level measures were included in the analysis to determine if averaging across paths for asymmetrical drawings would yield different results than using the longest path only. 8 Results and discussion This experiment examined whether participants recall the connection-structure of a website. If participants' schemata were consistent with the connection-structure of the test websites, then one would expect the number of levels drawn to reflect the number of levels in the test website. It was initially observed, however, that all of the drawings were notably similar regardless of the number of levels within the test website. Specifically, most of the drawings were hierarchies with either three or four levels (e.g. image category nodes, color nodes, image nodes, and sometimes a ‘home page’ node). This observation was corroborated with a multivariate analysis of covariance (MANCOVA), which determined the effects of exposure to the different test websites on participants' schemata, while controlling for individual differences in experience using the World Wide Web (hereafter called expertise) and number of end-nodes visited within the test website. The four dependent measures obtained from participants' drawings were , , Merror, and Maxerror. Because there was substantial variation within participants' expertise, number of years online was used as a covariate. Number of end-nodes visited within the test website during the experiment was also used as a covariate because as the website's structure includes more levels, the number of visits to end-nodes (i.e. where the pictures were within the test website) decreased. This is to be expected because websites with more levels require more steps to get to end-nodes than websites with less levels (Fig. 1), thus affording less time to view end-nodes1 1 The number of end-nodes visited did differ between groups, F(3,36)=24.30, p<.001, η2=0.67. The group means for 1L, 2L, 3L, and 4L were 49.20, 29.40, 31.40, and 22.50, respectively. . In addition, there was a great deal of variation in the number of end-nodes visited within groups. Unadjusted and adjusted means for all dependent variables in the following analysis are shown in Table 1. MANCOVA results revealed a significant effect for the four test website groups, Wilks' Λ=0.19, F(12,82.31)=6.11, p<0.001, multivariate η2=0.43. Neither of the covariates significantly influenced the combined dependent variable (expertise: Wilks' Λ=0.76, F(4,31)=2.45, p<0.05; number of end-nodes visited: Wilks' Λ=0.83, F(4,31)=1.54, p>0.05). Table 1 Unadjusted and adjusted means (adjusted means are in parentheses; means with matching superscripts are significantly different (p<.05)) Group . MLevels drawn . MaxLevels drawn . MError . MaxError . 1L 3.27(3.36) 3.50(3.58) 2.27(2.37)a,b 2.50(2.40)c,d 2L 3.16(3.14) 3.40(3.40) 1.16(1.14) 1.40(1.44)e 3L 3.53(3.50) 3.60(3.54) 0.73(0.70)a 0.80(0.75)c 4L 3.87(3.82) 4.00(3.98) 0.33(0.28)b 0.20(0.30)d,e Total 3.46(3.46) 3.63(3.63) 1.12(1.12) 1.23(1.23) Group . MLevels drawn . MaxLevels drawn . MError . MaxError . 1L 3.27(3.36) 3.50(3.58) 2.27(2.37)a,b 2.50(2.40)c,d 2L 3.16(3.14) 3.40(3.40) 1.16(1.14) 1.40(1.44)e 3L 3.53(3.50) 3.60(3.54) 0.73(0.70)a 0.80(0.75)c 4L 3.87(3.82) 4.00(3.98) 0.33(0.28)b 0.20(0.30)d,e Total 3.46(3.46) 3.63(3.63) 1.12(1.12) 1.23(1.23) Open in new tab Table 1 Unadjusted and adjusted means (adjusted means are in parentheses; means with matching superscripts are significantly different (p<.05)) Group . MLevels drawn . MaxLevels drawn . MError . MaxError . 1L 3.27(3.36) 3.50(3.58) 2.27(2.37)a,b 2.50(2.40)c,d 2L 3.16(3.14) 3.40(3.40) 1.16(1.14) 1.40(1.44)e 3L 3.53(3.50) 3.60(3.54) 0.73(0.70)a 0.80(0.75)c 4L 3.87(3.82) 4.00(3.98) 0.33(0.28)b 0.20(0.30)d,e Total 3.46(3.46) 3.63(3.63) 1.12(1.12) 1.23(1.23) Group . MLevels drawn . MaxLevels drawn . MError . MaxError . 1L 3.27(3.36) 3.50(3.58) 2.27(2.37)a,b 2.50(2.40)c,d 2L 3.16(3.14) 3.40(3.40) 1.16(1.14) 1.40(1.44)e 3L 3.53(3.50) 3.60(3.54) 0.73(0.70)a 0.80(0.75)c 4L 3.87(3.82) 4.00(3.98) 0.33(0.28)b 0.20(0.30)d,e Total 3.46(3.46) 3.63(3.63) 1.12(1.12) 1.23(1.23) Open in new tab An analysis of covariance (ANCOVA) was conducted on each dependent variable as a follow-up to the MANCOVA. These tests yielded no differences on the and measures, Fs(3,34)=1.13 and 0.78, ps>0.05, respectively. However, groups did differ on the Merror measure, F(3,34)=2.74, p<0.01 (partial η2=0.32), and were linear (p<0.001), suggesting that the drawings were based on something other than the connection-structure of the website. In addition, groups differed on the Maxerror measure, F(3,34)=3.31, p<0.01 (partial η2=0.33), and were linear (p<0.001). Pairwise comparisons of group means for each dependent variable, using a Bonferroni correction, are shown in Table 1. These analyses indicate that participants' drawings did not reflect the connection-structure of the test website. In fact, participants' drawings, particularly in the 1L and 2L conditions, had nodes (i.e. web pages) and links included that did not exist in the connection-structure of the system. Because participants had a general tendency to draw three or four levels, the results suggest that they based their representations on information other than the connection-structure of the site. Furthermore, based on inspection of the drawings, it appeared that they had encoded or recalled categorical information rather than the connection-structure of the test website. To investigate this possibility, a post-hoc analysis was conducted on the content of the nodes drawn. The hypermedia used in the current experiment presented participants with three categories of information: images, colors, and numbers (see Figs. 1 and 2). Although a rather crude measure, it was noted whether or not some sort of image, color, or number representation existed in the node labels of the drawings. This was done by noting if any node labels included terms that were descriptive of an image category, color, or number. It was possible for participants' drawings to have a single node containing all the aforementioned criteria if, for example, a node was labeled ‘Blue Background 2’. Likewise, a node could meet none of the criteria if it was arbitrarily labeled with something like ‘C’. From this coding, two participants' (5% of the sample) labels of nodes were undeterminable. Of the remaining 38 participants, 27 participants (67.5%) had no errors and 11 (27.5%) had one omission (e.g. omitting labels indicating color categories, labels for pages with the images, etc.). Generally, these errors were spread equivalently across groups, whereas the aforementioned MANCOVA results indicate that errors in the number of levels drawn differed across groups. As noted, this pattern suggests a tendency to mentally represent categorical information. In other words, this suggests that it is not pertinent for users to recall what steps are necessary to navigate the site (i.e. the connection-structure) only the decisions about what content or concept category to view next (i.e. how concepts are related). For example, to view a blue background, one does not need to know exactly what steps are necessary to get there, just what categories to choose when presented with various options. Thus, the number of levels drawn appears to have reflected these categories because most of the drawings had between three and four levels. This further suggests that the schema of hypermedia are based on content, not the connection-structure of the system. 9 Conclusion The results presented here support the notion that users do not form accurate schemata of the connection-structure of a website. Although participants were explicitly told to draw the connection-structure, they appeared to have drawn the conceptual relationships within the test website. This may explain some of the discrepancies in results about the utility of site maps. That is, if a site map does not reflect the user's (or domain's) conceptual relationships (i.e. a conceptual mental model), then the map's utility is lessened. This was noted by McDonald and Stevenson (1998), who reported that although maps enhanced performance, “…the [connection] structure reflected the conceptual structure” (p. 139). Therefore, the results they obtained may be a product of mental representation of conceptual relationships, and not a schema of the hypermedia's connection-structure. This is consistent with the task analysis literature (e.g. Seamster et al. 1997) and with Otter and Johnson (2000) in that the connection-structure of a website should be consistent with the users' conceptual mental models to enhance its usability. These conceptual models can be obtained with task analysis techniques (e.g. card sorting techniques) that probe the way users organize activities and information (Mayhew, 1999). In addition, Otter and Johnson (2000) proposed a mental model-based method to support hypermedia designers “in modeling the categorical structure that potential users of the system already have of the domain” (p. 3). One potential criticism of the current study is that participants were only allowed 5 min to explore the website. One could argue that using such a brief exploration period did not allow participants the requisite time to develop an accurate schema of the website. However, research has shown that cognitive maps (i.e. spatial schemata of physical environments) can be acquired within this amount of time (e.g. McNamara et al., 1989; Thorndyke and Hayes-Roth, 1982). Moreover, the use of images rather than text as the primary content of the test websites allowed the participants to explore a lot of the website (see Footnote 1). In addition, the participants' drawings did not seem incomplete, as would be expected if they simply did not have enough time to learn the overall connection-structure of the test website. Rather, the drawings depicted the qualitatively different conceptual relationships of the website's contents. For example, recall that in the one-level condition, the connection-structure consisted of a completely networked design, wherein no hierarchy exists and each web page is linked to every other web page. A participant in this condition (see Fig. 3) drew nodes representing categories of information on the website (i.e. backgrounds and bullets). In addition, this individual drew a hierarchical relation between these categories and the various sub-categories embedded within those categories (e.g. different colors, etc.). This is qualitatively different from what was presented to that participant. The other participants' drawings are replete with similar examples. Therefore, it appears that it was not the case that participants formed incomplete representations of the connection-structure. Rather, it appears that participants encoded the information in a qualitatively different way than anticipated based on the spatial metaphor. The current study demonstrated that participants did not form schemata of the connection-structure of the website that they explored. However, they did form some kind of mental representation and, although we did not manipulate the conceptual relationships of the test websites, the evidence suggests that participants used the conceptual relationships to form their mental representations (i.e. schemata). Therefore, we suggest reevaluation of the assumption that hypermedia is mentally represented as an environment having similar spatial properties to those found in the physical environment. Appendix A Instructions to participants (parenthetical remarks are instructions to the experimenter) For this task, you will be asked to draw the website you just explored. In order to draw the website you must draw nodes. Nodes are symbols—any shape you want to use—that represent separate web pages within the website. Therefore, for every time the screen changed, that is you went to a different web page, or you know that clicking something would have changed the screen's display, a different node must be drawn to represent that web page. Nodes are not drawings of what is on the web page, like text and graphics, but rather they are just representations that web pages exist. Although you did not actually visit all the web pages on the website, on a blank sheet you will draw ‘nodes’ representing each web page as you remember them or as you can assume them to be based on your experience with the website. For example, Fig. A1 shows a drawing of a website that has nine pages (point to each of them while counting them aloud). ‘Dogs Page’ and ‘Cats Page’ are linked together so that cats can be accessed from dogs and dogs can be accessed from cats. ‘Birds Page’, however, can only be accessed from the cat page. Note the use of arrows as links. Fig. A1 Open in new tabDownload slide Example connection-structure used to instruct participants. Fig. A1 Open in new tabDownload slide Example connection-structure used to instruct participants. Nodes should be titled with a descriptive word or phrase and links, which are represented as the arrows between the nodes, should have arrows indicating how the pages are linked. In Fig. A1, note that a line with one arrow end shows that the link to ‘Birds’ is on the cat page, but the bird page is does not have a link to the cat page, or any other page. You will not remember all of the pages, but try to complete this task as thoroughly as possible. If you cannot remember what was on a page or you did not visit a page but still know that it exists, just use your best guess based on your experience. If you run out of space, you can use the other sheets of paper beside you. There is no time limit, so feel free to use as much time as needed. I will be waiting in the hall outside the door. If any questions arise during the task, please feel free to ask. When you are done, go to the hall and get me. Do you have any questions? (If yes, answer questions by re-reading the appropriate part. If no, proceed). Before you start, I need to quiz you on your understanding of how these drawings work. For all questions, you will imagine that you have just arrived at a particular node within the website drawn in Fig. A1. I will ask you how you would get from one node to another only using the links that are on the particular pages represented by the node you are at. In other words, you cannot use the web browsers' back or forward buttons, only the links. (If any participants are incorrect on any of the following items, explain why and then try the next item. Repeat these questions until they respond correctly to all items. When they have completed all items correctly, they can begin drawing the test website). If you are viewing the dog page, how would you get to the bird page? (Should say: “Click on the cat link and then the bird link.”). If you are viewing one of the cat pictures, how would you get to a dog picture? (Should say: “Click on the cat link, then the dog link, then one of the dog picture links.”). If you are viewing one of the dog pictures, how would you get to a cat picture? (Should say: “Click on the dog link, then the cat link, then one of the cat picture links.”). If you are viewing the bird page, how would you get to the dog page? (Trick question because the bird page has no links on it. Should say: “It cannot be done.”). References Billingsley, 1982 Billingsley P.A , Navigation through hierarchical menu structures: does it help to have a map? , Proceedings of the Human Factors Society 26th Annual Meeting 1982 ) 103 107 OpenURL Placeholder Text WorldCat Calvi, 1997 Calvi L , Navigation and disorientation: a case study , Journal of Educational Multimedia and Hypermedia 6 ( 1997 ) 305 320 OpenURL Placeholder Text WorldCat Catledge and Pitkow, 1995 Catledge, L., Pitkow, J.E., 1995. Characterizing Browsing Strategies in the World-Wide Web. In: Berners-Lee, T., Kroemker, D. (Eds.), The Third International World-Wide Web Conference: Technology, Tools and Applications. Darmstadt, Germany, April 10–14. Chignell and Waterworth, 1997 Chignell M Waterworth J , Multimedia Salvendy G The Handbook of Human Factors and Ergonomics Second ed 1997 Wiley-Interscience , New York 1809 1861 OpenURL Placeholder Text WorldCat Dias and Sousa, 1997 Dias P Sousa P , Understanding navigation and disorientation in hypermedia learning environments , Journal of Educational Multimedia and Hypermedia 6 ( 1997 ) 173 185 OpenURL Placeholder Text WorldCat Dillon et al., 1990 Dillon A McKnight C Richardson J , Navigation in hypertext: a critical review of the concept , Human–Computer Interaction—INTERACT'90 1990 ) 587 592 OpenURL Placeholder Text WorldCat Dillon et al., 1993 Dillon A McKnight C Richardson J , Space—the final chapter or why physical representations are not semantic representations McKnight C Dillon A Richardson J Hypertext: A Psychological Perspective 1993 Ellis Horwood , New York 169 191 OpenURL Placeholder Text WorldCat Edwards and Hardman, 1989 Edwards D.M Hardman L , Lost in hyperspace: cognitive mapping and navigation in a hypertext environment McAleese R Hypertext: Theory into Practice 1989 Intellect , Oxford 105 125 OpenURL Placeholder Text WorldCat Gibson, 1966 Gibson J.J , The Senses Considered as Perceptual Systems 1966 Houghton Mifflin , Boston Gibson, 1979 Gibson J.J , The Ecological Approach to Visual Perception 1979 Houghton Mifflin , Boston Gray, 1990 Gray S.H , Using protocol analysis and drawings to study mental model construction during hypermedia navigation , International Journal of Human–Computer Interaction 2 ( 1990 ) 359 377 Google Scholar Crossref Search ADS WorldCat Kim and Hirtle, 1995 Kim H Hirtle S.C , Spatial metaphors and disorientation in hypertext browsing , Behaviour and Information Technology 14 ( 1995 ) 239 250 Google Scholar Crossref Search ADS WorldCat Mayhew, 1999 Mayhew D.J , The Usability Engineering Lifecycle 1999 Morgan Kaufmann , San Francisco McDonald and Stevenson, 1998 McDonald S Stevenson R.J , Navigation in hyperspace: an evaluation of the effects of navigational tools and subject matter expertise on browsing and information retrieval in hypertext , Interacting with Computers 10 ( 1998 ) 129 142 Google Scholar Crossref Search ADS WorldCat McNamara et al., 1989 McNamara T.P Hardy J.K Hirtle S.C , Subjective hierarchies in spatial memory , Journal of Experimental Psychology: Learning, Memory, and Cognition 15 ( 1989 ) 211 227 Google Scholar Crossref Search ADS PubMed WorldCat Neisser, 1976 Neisser U , Cognition and Reality: Principles and Implications of Cognitive Psychology 1976 Freeman , San Francisco Nielsen, 2000 Nielsen J , Designing Web Usability: The Practice of Simplicity 2000 New Riders , Indianapolis Otter and Johnson, 2000 Otter M Johnson H , Lost in hyperspace: metrics and mental models , Interacting with Computers 13 ( 2000 ) 1 41 Google Scholar Crossref Search ADS WorldCat Parlangeli et al., 1999 Parlangeli O Marchigiani E Bagnara S , Multimedia systems in distance education: effects of usability on learning , Interacting with Computers 12 ( 1999 ) 37 49 Google Scholar Crossref Search ADS WorldCat Schwartz and Kulhavy, 1988 Schwartz N.H Kulhavy R.W , Encoding tactics in the retention of maps , Contemporary Educational Psychology 13 ( 1988 ) 72 85 Google Scholar Crossref Search ADS WorldCat Seamster et al., 1997 Seamster T.L Redding R.E Kaemph G.L , Applied Cognitive Task Analysis in Aviation 1997 Avebury , Aldershot, England Shapiro, 1999 Shapiro A.M , The relevance of hierarchies to learning biology from hypertext , The Journal of Learning Sciences 8 ( 1999 ) 215 243 Google Scholar Crossref Search ADS WorldCat Stanton, 1990 Stanton N.A , Supporting learners' strategies in computer based training Winterburn E Realizing Human Potential 1990 Kogan Page , London 166 172 OpenURL Placeholder Text WorldCat Stanton and Baber, 1992 Stanton N.A Baber C , An investigation of styles and strategies in self directed learning , Journal of Educational Multimedia and Hypermedia 1 ( 2 ) 1992 ) 147 167 OpenURL Placeholder Text WorldCat Stanton et al., 1992 Stanton N.A Taylor R.G Tweedie L.A , Maps as navigational aids in hypertext environments: an empirical evaluation , Journal of Educational Multimedia and Hypermedia 1 ( 1992 ) 431 444 OpenURL Placeholder Text WorldCat Thorndyke and Goldin, 1983 Thorndyke P.W Goldin S.E , Spatial learning and reasoning skill Pick C.P Jr. Acredolo H.L Spatial Orientation 1983 Plenum Press , New York 195 217 OpenURL Placeholder Text WorldCat Thorndyke and Hayes-Roth, 1982 Thorndyke P.W Hayes-Roth B , Differences in spatial knowledge acquired from maps and navigation , Cognitive Psychology 14 ( 1982 ) 560 589 Google Scholar Crossref Search ADS PubMed WorldCat © 2002 Published by Elsevier Science B.V.
TI - Users' schemata of hypermedia: what is so ‘spatial’ about a website?
JF - Interacting with Computers
DO - 10.1016/S0953-5438(02)00011-5
DA - 2002-10-01
UR - https://www.deepdyve.com/lp/oxford-university-press/users-schemata-of-hypermedia-what-is-so-spatial-about-a-website-b1tpHo6z0K
SP - 487
EP - 502
VL - 14
IS - 5
DP - DeepDyve
ER -