TY - JOUR
AU - Danielson, David, R
AB - Abstract Knowledge regarding how Web information-seekers behave with respect to the structures and cues they are provided with may shed light on general principles of navigation in electronic spaces, and assist designers in making more informed structural decisions. This study examines user movement through hierarchically structured Web sites and the behavioral effects of a constantly visible, textual contents list for relatively small sites or more extensive local views than are generally used on the Web today. The site overview resulted in users abandoning fewer information-seeking tasks. Users with such context dig deeper into the site structure, make less use of the browser's Back button, and frequently make navigational movements of great hierarchical distances. Navigational correlates of success and reported confidence for users with the overview differ from those without such context. Both with and without a constant overview, the relationship between the source and destination pages may help predict the amount of time spent at the destination. Experimental reports are preceded by a review of click-stream navigation behavior research. 1 Introduction With respect to Web navigation, there are few subjective experiences as satisfying as developing a thorough command and understanding of a space and moving freely and quickly through it with ease, while there are few subjective experiences as humbling as feeling utterly intimidated, disoriented, and simply lost. Designers of electronic information spaces have mighty big jobs indeed: to create an experience for navigators conducive to quickly feeling a command of the space, or to in essence feel as though they have that particular world in the palm of their hand. A common combatant to the problem of disorientation is the presentation of a site map, or an overview of the site structure. Such maps are standard ingredients in the Web development process (Lin et al., 2000; Newman and Landay, 2000), but their realization on the Web is far from pervasive. Instead, navigators are commonly left with some subset of the standard navigation aids: pervasive top-level links such as the site's home page, local views such as the current node's siblings or children, and breadcrumb lists. Moreover, the top-level links and local views are generally limited to around seven, often based upon the magic number proposed by Miller (1956). Navigators are unlikely to make use of browser-supported shortcuts (Catledge and Pitkow, 1995), so the available options lie in the provided links and the Back button. For this reason, as a navigator more thoroughly examines a small site or subspace of a larger one, the available movements in that space become more and more dominated by previously visited territory. The striking fact of hyperspace navigation is that where you haven't been will in general only affect you insofar as your current view of the world allows; if this view is primarily local with little global context, and you already have a thorough command of the local space, uncharted territory will look miniscule, when in fact there's a whole world out there. It is under these circumstances especially that we ought to envy the birds or at least enlighten ourselves with a map. This study is concerned with the pervasive presentation of site maps, addressing such questions as: What are the primary issues surrounding site maps, the hyperspace version of a bird's eye view of the world, on the Web today? How might a constantly visible site map affect the behavior of information-seekers with respect to the structure of a hyperspace? While the majority of studies in this research area focus on global, inter-site navigation to provide click-stream analyses of user movement on the Web, the current study focuses on single-session, within-site navigation. Global log file analyses provide extremely useful insights into fundamental issues of Web navigation, but site designers will benefit more substantially from within-site behavioral data, particularly when the types of user tasks are controlled. Moreover, as the field of information and navigation design on the Web matures, considerations of performance alone will not be enough, especially given the growing diversity of Web site domains and design schemes, and wide variety of task types. Instead, designers will benefit from navigational data upon which they may base design decisions in conjunction with the specific nature of their sites, users, and goals. Experimental research can aim to answer questions like ‘How will this particular design decision change the navigation behavior of site visitors?’ and allow designers to answer the question, ‘Are these particular changes what I really want?’ for themselves. 2 Navigation behavior Both Web research and investigations of closed hypertext systems provide useful insights into user navigation behavior. Differences between the Web and closed systems are most pronounced with respect to inter-site navigation, where the potential for variety in encountered hyperspace structures and designs is quite high. However, both inter- and intra-site Web navigation research may benefit from and extend closed hypertext research, insofar as both provide data regarding behavior in hyperspaces. Most navigation studies make a distinction between general purpose browsing, where the user consults sources with items of interest, and directed search, where the goal is known (Catledge and Pitkow, 1995). Navigation tasks begin with a decision regarding this split; the user decides whether she is looking for a specific item, a group of items, or general information about the contents of a space. Directed search may be further separated into subject-based exploration, in which the user attempts to gain a basic understanding of a specific subject area, and fact-finding missions, in which the user looks for a specific piece of information in the space (Newfield et al., 1998). Fact-finding missions have been classified by Choo et al. (2000) as formal search, signified by deliberate or planned efforts to find specific items. It is formal search with which this study is primarily concerned. Once a user decides on her navigation goal, she is likely to develop some sort of strategy for achieving it (Meyer et al., 1997). It is useful to first recognize that a single user's search strategy on the Web may constantly be in flux. Users often switch back and forth between an array of strategies in response to navigation aids and other stimuli provided by information designers. Perhaps the most widely accepted description of common Web navigation strategies is that of the hub and spoke methods proposed by Catledge and Pitkow (1995). Such strategies involve starting at a relative top-level node (or a hub), digging along some path from that hub (creating a spoke), and then returning to the hub to create another spoke. Hub and spoke traversal involves either backtracking or returning directly to a starting point. Such a return represents one example of an extremely common behavioral phenomenon on the Web: page revisitation. Tauscher and Greenberg (1997) reported that 58% of all page visitations are revisits, based on data collected over 6 weeks from 23 users of a commercial Web browser, and point out that Web browsing activity therefore qualifies as a recurrent system, or one in which ‘users predominantly repeat activities they had invoked before, while still selecting new actions from the many that are possible.’ Page revisitation was additionally examined in terms of frequency: 60% of pages were visited once, 19% were visited twice, 8% were visited three times, 4% were visited four times, and few were visited frequently. Tauscher and Greenberg further developed a recurrence distribution from the data, reporting a 39% chance that any given page would be a member of the set containing the previous six pages visited, showing that Web navigation involves a high degree of recency revisitation. Cockburn and McKenzie (2001) contend that navigators actually revisit pages at a much higher rate (81%) based on analyses of log files from longer time periods than those of the Tauscher and Greenberg study. The authors argue that the length of the navigation period impacts the revisitation rate (McKenzie and Cockburn, 2001). 2.1 The Back button The pervasiveness of Back button use has been well-documented, which should come as no surprise given the data regarding page revisitation; Back button clicks are necessarily revisits. Catledge and Pitkow (1995), in the first extensive study of Web browsing behavior, analyzed over 31,000 user navigational acts taken from 3 weeks worth of data (in mid 1994), and reported that Back button clicks comprised 41% of these acts, a staggering number. Subsequent research by Tauscher and Greenberg (1997) reported this number at 30%. The extensive use of the Back button occurs inspite of users' generally naı̈ve models of how it works (Greenberg and Cockburn, 1999). Navigators use the Back button to return to a landmark page (or hub, as previously discussed), often indicating that the user has either extracted all she wishes to extract from the current page, or that the page does not contain items of interest or the desired information. In directed search, a Back button click may represent a belief that the desired information was missed during extraction from a previously visited page. Commonly, navigators use the ‘click until the desired page is recognized’ strategy (Greenberg and Cockburn, 1999). Back navigation requires simple recognition of a visual stimulus (the page itself), rather than recognition of text (such as in using a browser's history list), or recall memory (such as in typing in a full URL). This may contribute to the strong user preference for the Back button over other page revisitation schemes. 3 Navigation design Design of navigation systems may be viewed as assisting users with their natural behaviors and movement strategies. That is, Web design in part involves anticipating movements based on behavioral data, and presenting navigation aids according to those anticipations. Navigation design answers three fundamental questions of navigation (Nielsen, 2000): Where am I? Where have I been? Where can I go? One of the first glaring facts of real-world Web sites is that many make use of hierarchies as the basis of their structure and organization, and thus benefit from our extensive exposure to hierarchical organization in everyday life. It is generally the job of information designers to build this structure, deciding on the number of top-level categories of the site, their subcategories, and so on. Web site structures, however, differ in important ways from the common hierarchies we encounter in the real-world, primarily due to the nature of hyperspace. Hyperspace nodes inter-connect in ways that often transcend hierarchical organization, lending themselves more obviously to a complex Web. However, the complexity of such inter-connections, even for small sites, makes visually presenting them to users impractical; their complexity may prevent a user from quickly discerning the general structure of the site, and thus what is presumably a primary purpose of a site overview is lost. 3.1 The site map Designers have a number of options for visually presenting the hierarchical organization of a site, and most fall within either (i) contents lists (such as an index or table of contents), and (ii) graphical representations. When one considers the common Web vocabulary, both fall under the term ‘site map.’ While this paper primarily concerns itself with the constant presentation of a contents list, a wide range of graphical site structure representations have been proposed, designed, and thoroughly reviewed. Indeed, graphical maps can and do make the limitations of hierarchical representation less obvious, as ‘an index may not highlight the links between nodes whereas a map would make these relationships explicit’ (McDonald and Stevenson, 1998). However, practical limitations such as page loading time often push graphic-heavy maps aside, and real-world sites appear to suggest information designers have sided with purely or predominantly textual presentations. Contents lists do not avoid these problems altogether, of course. Hyperspaces can be and often are much too large for a complete site map to be useful. Presenting more and more nodes in a site map is akin to hopping into a hot air balloon and ascending over the city for a bird's eye view. In physical spaces, “Increased altitude enables spatial relationships to be refined since more landmarks can be seen simultaneously, but increased altitude also decreases the strength of the stimuli” (Darken and Sibert, 1993). This description is no less accurate in hyperspace. Information designers must decide which ‘altitude’ is appropriate, weighing the tradeoff between providing a broad view showing a wide range of hierarchical relationships but running the risk of those relationships being lost in the shuffle, or providing a narrow view, ensuring that those relationships available will likely not be missed, but abandoning the opportunity to give a more extensive overview of the site structure. Information designers confront more fine-grained concerns regarding the display of textual site maps as well. First, designers make as many linguistic decisions as there are nodes in the hierarchy. With the bird's view of scanning a site overview, the few words used to describe each node can be crucial to the information-seeker's success (Norman, 1991). Secondly, information designers decide upon the general organization of the site—who is related to whom, who is linked to whom, and so forth. Decisions regarding this structure are often made earlier than they perhaps should be (Lin et al., 2000), and this can have serious implications on the perceptions of a navigator in the space. Suffice it to say that designers must take into account several factors when developing the link structure of a site, including content relationships amongst pages (Meyer et al., 1997), likely user interests given a particular location in the site (Nielsen, 2000), the balance of the tree, including how many nodes appear at each level, in each sibling group, and in what order (Norman, 1991), and the amount of information presented on each page and its viewing area, amongst others. To make an understatement, developing a well-designed site structure is not an easy task, or one that should be taken lightly. Thirdly, designers of textual site overviews must make information design decisions regarding how the separate levels of the hierarchy will be distinguished from one another, which visited and unvisited link colors, if any, they will use, and the fonts, sizes and other associated typography options. 4 Effects of textual Web context Providing contextual navigation aids on a Web site has a variety of effects on user behavior, and attacks the general problem of feeling lost in hyperspace. Most of the research concerning context effects focus on browsing more generally than information-seeking or formal search, but point the way towards research regarding more fine-grained analyses of formal search and fact-finding missions (Park and Kim, 2000). 4.1 Lost in hyperspace Disorientation is a common phenomenon on the Web. Disoriented Web navigators have trouble gaining an overview of the information structure before them (McDonald and Stevenson, 1998). Without proper contextual cues, users are often unable to determine where they are, find previously visited pages, or find desired information (Park and Kim, 2000). Because of the diversity and complexity of Web sites, and because a user's navigation history in relation to a site's structure quickly becomes difficult to retain in memory, context for the current location is often forgotten (Newfield et al., 1998). “Hypertext alone does not make it easy for the user to know what information is available or which parts of the text remain to be seen” (McDonald and Stevenson, 1998). Focus+Context approaches, such as Furnas' (1986) Fisheye views, for example, provide local navigation in addition to more global reorienting capabilities for the user, and have been shown to help reduce disorientation. Without such help from the information designer, users must simultaneously focus on remaining oriented in the information space and on achieving their goals. The additional effort and concentration necessary to maintain several tasks or trails at one time is known as cognitive overhead (Conklin, 1987). Structural contextual cues have been shown to reduce cognitive overhead and thus attack the problem of being ‘lost in hyperspace’ (Edwards and Hardman, 1989). 4.2 Performance A natural and rather practical effect of contextual navigation aids to look for are differences in user performance on either general knowledge of a navigated space or on specified tasks when such aids are present. WebTOC (Nation et al., 1997) creates a hierarchical listing of a site's contents, and has been found to be rather useful in informal settings. Improved performance with respect to navigation and memory of topics presented in an information space has also resulted from extensive textual context (Dee-Lucas and Larkin, 1995). Rada and Murphy (1992) demonstrated that overviews are particularly useful in supporting an understanding of the hyperspace's content domain. Moreover, McDonald and Stevenson (1998) found performance with contextual navigation aids to have particular types of interactions with user expertise for the material presented in the space. Non-knowledgeable users tend to rely more heavily on navigation aids than knowledgeable users, and the research suggests that maps are most useful when gaining familiarity with new material. Most interestingly, use of the contextual information appeared to help eliminate the gap between knowledgeable and non-knowledgeable browsers of the information space. McDonald and Stevenson suggested that the contextual information (textual context less so than an explicit map) gives the non-knowledgeable user a reflection of the conceptual structure of the material, which one might imagine is a primary difference between the knowledge of experienced and inexperienced people in a subject domain. While hypertext and Web research have shown that overview information can improve performance in a variety of ways, surprisingly little is known about the specific behavioral effects of pervasive overviews and site maps on the Web. Such information will assist information architects in making more informed design decisions, and allow them to better predict the effects of such decisions on user movement through their sites. 4.3 Usage In general, Web navigators can and do make use of a wide variety of navigation aids when provided. Experiments regarding structural contextual cues reported by Park and Kim (2000) found users to be both aware of contextual aids and willing to make use of them. Indeed, planned and exploratory searches within a Web space often begin with a search for a site overview (O'Day and Jeffries, 1993). With respect to the types of navigation, an index or table of contents is used more often during information search than during general purpose browsing (McDonald and Stevenson, 1998). While users make extensive use of overviews and other navigational support, recognize that the occlusion of such context can greatly reduce usage (Wright, 1991). This study provides experimental data on behavior when the contextual overview is constantly visible and available to the navigator. 4.4 Node visitation and revisitation A known effect of contextual information displays in browsing behavior, specifically in large information spaces, is that they cause navigators to visit fewer total nodes in the space (Park and Kim, 2000). These findings are intimately linked with the amount of backtracking one does in the absence of contextual information. The effect associated with fewer total node openings is fewer node revisits, commonly instantiated by the Back button. 5 Experiment This study concerns itself with the behavioral navigation effects of a constantly visible site map for relatively small sites and local views of large sites. The study generally examined user behavior in two conditions with respect to the structures provided to them by information designers in five real-world Web sites. 5.1 Participants Twenty randomly selected subjects, aged 18–40, participated in the experiment. (Analysis of age effects on browsing behavior may be found in Meyer et al. (1997), and suggest that these participants were likely more efficient browsers than a wider age sample would be.) The experiment used a between subject design. The participants were randomly divided into two groups: a control group and a constant site map group. Participants were also randomly given one of the five Web sites used to navigate, and tested individually. 5.2 Web sites used The five sites used in this study were selected to represent a range of topics, accessible to the general public. The sites concerned recreation, small business, law, leisure, and health, respectively, and ranged between 28 and 57 nodes in their hierarchical structures at the time of this study. The sites contained either two or three hierarchical levels. The specific sites used were: TopGolf (http://www.topgolf.co.uk/) Grassroots Enterprise (http://www.grassroots.com/) CA Small Claims Court (http://www.courtinfo.ca.gov/courts/trial/smallclaims/) Santa Cruz Beach Boardwalk (http://www.beachboardwalk.com/) Wellmed (http://www.wellmed.com/). 5.3 Constant site map The constant site overview acted as a separate frame, but its links were refreshed each time the main frame was refreshed. The browser window filled the entire monitor display, and was presented on the screen with a 1024×768 pixel resolution. The site map spanned the same vertical space as the site itself, and occupied approximately 270 pixels of horizontal space, as the left-most column of the screen, appropriate for supporting visual search performance (van Schaik and Ling, 2001). The site map was similar to the one provided within the site itself; neither structural nor linguistic decisions were imposed on its design. 5.4 Procedure Experimental sessions were separated into three sections: (1) participant instructions, (2) a 15-min navigation period, and (3) completion of a post-browsing questionnaire. Experimental tasks were fact-finding missions, such as ‘About how long does it take for an individual person to play three games of TopGolf?’ While navigating, screen monitor input was recorded and converted to standard video output for later viewing and coding. This option was preferred, as it was assumed to be less invasive than many other recording options. 6 Results In comparison to control participants, constant site map participants: Abandoned fewer information-seeking tasks. Showed different navigational correlates of success and reported information-seeking confidence. Dug deeper into the site hierarchy and spent more time at lower levels. Made less use of the Back button. Made more movements outside the immediate hierarchical family of the current node. Better estimated the number of distinct pages they had visited while navigating. 6.1 Navigational click-stream analysis There are three primary ways to analyze click-stream data, such as ‘Back button use,’ in Web navigation: (1) mean frequency, (2) mean proportion of some total, such as a user's total number of navigational acts, and (3) frequency and proportion across all navigators. The last approach is more common to global navigation analyses, often used when individual remote site users cannot be distinguished from one another. Table 1 summarizes the main click-stream data collected, paying attention to approaches (1) and (2) above. The data concerns page revisitation, browser-supported navigational acts (i.e. ‘Back’), hierarchical relationship of the source and destination pages (i.e. ‘Sibling’), hierarchical direction of movement (i.e. ‘Up’), and hierarchical level (i.e. ‘Top-level’) of page visits and time. In all, 700 navigational actions were recorded (386 control condition, 314 site map condition). Table 1 Click-stream comparisons between control and site map condition participants. (t-test p-values included; time measurements are in seconds) . Control . Site map . p . Control . Site map . p . . Mean# . SD . Mean# . SD . . Mean% . SD . Mean% . SD . . Navigation acts 38.6 9.4 31.4 10.2 0.119 Unique pages 16.9 5.1 17.8 5.4 0.707 44.00 8.49 58.70 15.5 0.020* Revisits 21.7 5.9 13.6 8.2 0.022* 56.05 8.49 41.30 15.5 0.020* Back 5.7 3.0 1.3 2.8 0.004** 14.02 6.15 3.04 6.06 0.001** Forward 0.0 0.0 0.0 0.0 – 0.00 0.00 0.00 0.00 – Sibling 20.1 5.1 13.2 7.3 0.026* 53.30 12.40 39.80 17.40 0.064 Child 10.9 6.1 6.2 5.9 0.099 27.70 11.20 17.70 12.20 0.094 Parent 3.3 2.6 1.8 2.2 0.180 7.69 5.33 5.15 4.78 0.278 Grandchild 0.0 0.0 0.0 0.0 – 0.00 0.00 0.00 0.00 – Grandparent 0.1 0.3 0.2 0.4 0.557 0.23 0.74 0.48 1.02 0.543 Distal 4.2 2.4 10.0 4.2 0.002** 11.83 7.55 36.90 24.00 0.010** Up 5.6 3.3 3.6 2.5 0.147 13.93 5.29 11.23 7.15 0.351 Down 12.4 5.2 12.3 4.6 0.964 31.33 8.44 40.09 9.63 0.045* Lateral 20.6 4.9 15.5 7.2 0.082 54.70 12.20 48.70 13.30 0.302 Top-level visits 19.1 5.7 9.8 8.7 0.013* 51.10 15.30 27.20 19.40 0.007** Lower-level visits 19.5 10.0 21.6 5.9 0.571 48.90 15.30 72.80 19.40 0.007** Top-level time 536.0 176.0 287.0 115.0 0.002** 59.60 19.50 31.80 12.80 0.002** Lower-level time 364.0 176.0 614.0 115.0 0.002** 40.40 19.50 68.20 12.80 0.002** Abandoned tasks 2.0 1.4 0.7 0.9 0.029* . Control . Site map . p . Control . Site map . p . . Mean# . SD . Mean# . SD . . Mean% . SD . Mean% . SD . . Navigation acts 38.6 9.4 31.4 10.2 0.119 Unique pages 16.9 5.1 17.8 5.4 0.707 44.00 8.49 58.70 15.5 0.020* Revisits 21.7 5.9 13.6 8.2 0.022* 56.05 8.49 41.30 15.5 0.020* Back 5.7 3.0 1.3 2.8 0.004** 14.02 6.15 3.04 6.06 0.001** Forward 0.0 0.0 0.0 0.0 – 0.00 0.00 0.00 0.00 – Sibling 20.1 5.1 13.2 7.3 0.026* 53.30 12.40 39.80 17.40 0.064 Child 10.9 6.1 6.2 5.9 0.099 27.70 11.20 17.70 12.20 0.094 Parent 3.3 2.6 1.8 2.2 0.180 7.69 5.33 5.15 4.78 0.278 Grandchild 0.0 0.0 0.0 0.0 – 0.00 0.00 0.00 0.00 – Grandparent 0.1 0.3 0.2 0.4 0.557 0.23 0.74 0.48 1.02 0.543 Distal 4.2 2.4 10.0 4.2 0.002** 11.83 7.55 36.90 24.00 0.010** Up 5.6 3.3 3.6 2.5 0.147 13.93 5.29 11.23 7.15 0.351 Down 12.4 5.2 12.3 4.6 0.964 31.33 8.44 40.09 9.63 0.045* Lateral 20.6 4.9 15.5 7.2 0.082 54.70 12.20 48.70 13.30 0.302 Top-level visits 19.1 5.7 9.8 8.7 0.013* 51.10 15.30 27.20 19.40 0.007** Lower-level visits 19.5 10.0 21.6 5.9 0.571 48.90 15.30 72.80 19.40 0.007** Top-level time 536.0 176.0 287.0 115.0 0.002** 59.60 19.50 31.80 12.80 0.002** Lower-level time 364.0 176.0 614.0 115.0 0.002** 40.40 19.50 68.20 12.80 0.002** Abandoned tasks 2.0 1.4 0.7 0.9 0.029* *Significant at p<0.05. **Significant at p≤0.01. Open in new tab Table 1 Click-stream comparisons between control and site map condition participants. (t-test p-values included; time measurements are in seconds) . Control . Site map . p . Control . Site map . p . . Mean# . SD . Mean# . SD . . Mean% . SD . Mean% . SD . . Navigation acts 38.6 9.4 31.4 10.2 0.119 Unique pages 16.9 5.1 17.8 5.4 0.707 44.00 8.49 58.70 15.5 0.020* Revisits 21.7 5.9 13.6 8.2 0.022* 56.05 8.49 41.30 15.5 0.020* Back 5.7 3.0 1.3 2.8 0.004** 14.02 6.15 3.04 6.06 0.001** Forward 0.0 0.0 0.0 0.0 – 0.00 0.00 0.00 0.00 – Sibling 20.1 5.1 13.2 7.3 0.026* 53.30 12.40 39.80 17.40 0.064 Child 10.9 6.1 6.2 5.9 0.099 27.70 11.20 17.70 12.20 0.094 Parent 3.3 2.6 1.8 2.2 0.180 7.69 5.33 5.15 4.78 0.278 Grandchild 0.0 0.0 0.0 0.0 – 0.00 0.00 0.00 0.00 – Grandparent 0.1 0.3 0.2 0.4 0.557 0.23 0.74 0.48 1.02 0.543 Distal 4.2 2.4 10.0 4.2 0.002** 11.83 7.55 36.90 24.00 0.010** Up 5.6 3.3 3.6 2.5 0.147 13.93 5.29 11.23 7.15 0.351 Down 12.4 5.2 12.3 4.6 0.964 31.33 8.44 40.09 9.63 0.045* Lateral 20.6 4.9 15.5 7.2 0.082 54.70 12.20 48.70 13.30 0.302 Top-level visits 19.1 5.7 9.8 8.7 0.013* 51.10 15.30 27.20 19.40 0.007** Lower-level visits 19.5 10.0 21.6 5.9 0.571 48.90 15.30 72.80 19.40 0.007** Top-level time 536.0 176.0 287.0 115.0 0.002** 59.60 19.50 31.80 12.80 0.002** Lower-level time 364.0 176.0 614.0 115.0 0.002** 40.40 19.50 68.20 12.80 0.002** Abandoned tasks 2.0 1.4 0.7 0.9 0.029* . Control . Site map . p . Control . Site map . p . . Mean# . SD . Mean# . SD . . Mean% . SD . Mean% . SD . . Navigation acts 38.6 9.4 31.4 10.2 0.119 Unique pages 16.9 5.1 17.8 5.4 0.707 44.00 8.49 58.70 15.5 0.020* Revisits 21.7 5.9 13.6 8.2 0.022* 56.05 8.49 41.30 15.5 0.020* Back 5.7 3.0 1.3 2.8 0.004** 14.02 6.15 3.04 6.06 0.001** Forward 0.0 0.0 0.0 0.0 – 0.00 0.00 0.00 0.00 – Sibling 20.1 5.1 13.2 7.3 0.026* 53.30 12.40 39.80 17.40 0.064 Child 10.9 6.1 6.2 5.9 0.099 27.70 11.20 17.70 12.20 0.094 Parent 3.3 2.6 1.8 2.2 0.180 7.69 5.33 5.15 4.78 0.278 Grandchild 0.0 0.0 0.0 0.0 – 0.00 0.00 0.00 0.00 – Grandparent 0.1 0.3 0.2 0.4 0.557 0.23 0.74 0.48 1.02 0.543 Distal 4.2 2.4 10.0 4.2 0.002** 11.83 7.55 36.90 24.00 0.010** Up 5.6 3.3 3.6 2.5 0.147 13.93 5.29 11.23 7.15 0.351 Down 12.4 5.2 12.3 4.6 0.964 31.33 8.44 40.09 9.63 0.045* Lateral 20.6 4.9 15.5 7.2 0.082 54.70 12.20 48.70 13.30 0.302 Top-level visits 19.1 5.7 9.8 8.7 0.013* 51.10 15.30 27.20 19.40 0.007** Lower-level visits 19.5 10.0 21.6 5.9 0.571 48.90 15.30 72.80 19.40 0.007** Top-level time 536.0 176.0 287.0 115.0 0.002** 59.60 19.50 31.80 12.80 0.002** Lower-level time 364.0 176.0 614.0 115.0 0.002** 40.40 19.50 68.20 12.80 0.002** Abandoned tasks 2.0 1.4 0.7 0.9 0.029* *Significant at p<0.05. **Significant at p≤0.01. Open in new tab Control participants showed significantly greater frequency and proportion of page revisits, consistent with past work in this research area (Park and Kim, 2000). Site map participants made significantly less use of the Web browser's Back button. Consistent with Catledge and Pitkow (1995), participants across the two conditions did not rely on the browser's Forward button (none of the 20 participants ever clicked on it). Site map participants showed significantly greater frequency and proportion of navigational movements outside the immediate family of the source node (‘distal’ movements), commonly instantiated by the available overview. Nonetheless, path ‘hops,’ such as to a grandchild or grandparent node, were extremely rare in both conditions. Site map participants made fewer top-level page visits and spent more time at least one level deep in the site hierarchy. A paired samples t-test revealed that the mean proportion of time spent at top-level pages for control participants (59.60, SD=19.50) was significantly greater than mean proportion of top-level visits (51.10, SD=15.30) for these participants (t-test, p=0.029). The same analysis for site map condition participants did not show this result (t-test, p=0.215). Although performance differences were not the focus of this study, control condition participants averaged two abandoned tasks (SD=1.4); site map participants averaged 0.7 (SD=0.9). Because a user was allowed to skip tasks when she did not believe she would find an answer, completed tasks cannot be compared reliably between the conditions. In the post-browsing questionnaire, users were asked to estimate the number of distinct pages they had visited while navigating. Site map participants provided better estimates of this number; their estimates significantly correlated with the actual number of unique page visits (r=0.647, p=0.043), while control participants' estimates did not show such a correlation (r=0.501, p=0.140). 6.2 Site map clicks During the 15-min experimental period, site map participants averaged 15.0 site map clicks (SD=10.1; max=31; min=3). These clicks comprised a mean of 53.49% of a user's navigational acts (SD=34.81%; max=100%; min=6.52%). The wide range in the extent of overview usage is testament to the cognitive variability in preference for maps (Streeter et al., 1985). Site map clicks primarily took the navigator away from the immediate family of the source node (51.3%), and were most commonly used to dig deeper into the site structure (46.0% to move down in the site, 84.7% to visit a page at least one level deep in the structure). Table 2 summarizes the 150 observed site map clicks. Table 2 Breakdown of 150 site map clicks observed across 10 site map condition participants . Raw . % . . Raw . % . Sibling 54 36.0 Up 14 9.3 Child 12 8.0 Down 69 46.0 Parent 5 3.3 Lateral 67 44.7 Grandchild 0 0.0 Grandparent 2 1.3 Top-level visit 23 15.3 Distal 77 51.3 Lower-level visit 127 84.7 . Raw . % . . Raw . % . Sibling 54 36.0 Up 14 9.3 Child 12 8.0 Down 69 46.0 Parent 5 3.3 Lateral 67 44.7 Grandchild 0 0.0 Grandparent 2 1.3 Top-level visit 23 15.3 Distal 77 51.3 Lower-level visit 127 84.7 Open in new tab Table 2 Breakdown of 150 site map clicks observed across 10 site map condition participants . Raw . % . . Raw . % . Sibling 54 36.0 Up 14 9.3 Child 12 8.0 Down 69 46.0 Parent 5 3.3 Lateral 67 44.7 Grandchild 0 0.0 Grandparent 2 1.3 Top-level visit 23 15.3 Distal 77 51.3 Lower-level visit 127 84.7 . Raw . % . . Raw . % . Sibling 54 36.0 Up 14 9.3 Child 12 8.0 Down 69 46.0 Parent 5 3.3 Lateral 67 44.7 Grandchild 0 0.0 Grandparent 2 1.3 Top-level visit 23 15.3 Distal 77 51.3 Lower-level visit 127 84.7 Open in new tab Site map clicks were used more commonly after the user had spent more time at the Web site. In the site map condition, mean elapsed time for site map clicks (557 s, SD=223) was significantly greater (t-test, p=0.007) than mean elapsed time for non-site map clicks (486 s, SD=247). Table 3 shows the proportion of site map clicks to non-site map clicks, by amount of time spent at the Web site. A significant interaction effect (χ2=4.556, p=0.033) was found between site map and non-site map clicks for the first and second halves of the experiment, with site map use doubling in the second half of the 15-min navigation session, and non-site map clicks increasing modestly. Table 3 Proportion of site map clicks to non-site map clicks by amount of time the user has spent at Web site Period (min) . Total . Site map . Non-site map . Site map proportion . First 3 min 27 9 18 0.33 3–6 60 22 38 0.37 6–9 78 41 37 0.53 9–12 69 37 32 0.54 12–15 80 41 39 0.51 Totals (full 15 min) 314 150 164 0.48 First half of study 124 50 74 0.40 Second half of study 190 100 90 0.53 Totals (full 15 min) 314 150 164 0.48 Period (min) . Total . Site map . Non-site map . Site map proportion . First 3 min 27 9 18 0.33 3–6 60 22 38 0.37 6–9 78 41 37 0.53 9–12 69 37 32 0.54 12–15 80 41 39 0.51 Totals (full 15 min) 314 150 164 0.48 First half of study 124 50 74 0.40 Second half of study 190 100 90 0.53 Totals (full 15 min) 314 150 164 0.48 Open in new tab Table 3 Proportion of site map clicks to non-site map clicks by amount of time the user has spent at Web site Period (min) . Total . Site map . Non-site map . Site map proportion . First 3 min 27 9 18 0.33 3–6 60 22 38 0.37 6–9 78 41 37 0.53 9–12 69 37 32 0.54 12–15 80 41 39 0.51 Totals (full 15 min) 314 150 164 0.48 First half of study 124 50 74 0.40 Second half of study 190 100 90 0.53 Totals (full 15 min) 314 150 164 0.48 Period (min) . Total . Site map . Non-site map . Site map proportion . First 3 min 27 9 18 0.33 3–6 60 22 38 0.37 6–9 78 41 37 0.53 9–12 69 37 32 0.54 12–15 80 41 39 0.51 Totals (full 15 min) 314 150 164 0.48 First half of study 124 50 74 0.40 Second half of study 190 100 90 0.53 Totals (full 15 min) 314 150 164 0.48 Open in new tab 6.3 Time spent on a Web page Several interesting factors predicting the amount of time a user would spend on a Web page were found, across the two conditions (700 total navigational page accesses). In particular, the way in which a user accesses a page may help predict longevity of page stay. Navigators spent less time on a page if the page was accessed via the Back button (t-test, p=0.000). There was a significant effect of hierarchical source–destination relationship on amount of time spent at the destination (one-way Analysis of Variance (ANOVA), p=0.000). Outside of immediate family (‘distal’) source–destination relationships resulted in the most time spent at the destination page. A top-level switch was defined in this study as any navigational movement that took the user from one main subspace of the site (determined by the structure of the hierarchical site map) to another. Navigators spent more time on a page if the page was accessed via top-level switch (t-test, p=0.005). There was a significant effect of source–destination direction of movement on the amount of time spent at the destination (one-way ANOVA, p=0.000). Downward movements resulted in the most time spent at the destination page. Overall, pages accessed in the control condition (386 navigational accesses) were visited for a mean time of 18.7 s (SD=20.4), while pages accessed in the site map condition (314 navigational accesses) were visited for a mean time of 24.4 s (SD=24.5), significantly higher (t-test, p=0.001). Including the first entry into the home page and each site reentry after a task was completed or abandoned, the 20 users combined for a total of 880 page visits (700 via navigation, 20 initial site entries, 160 site reentries). Of these, mean time spent on a first-time visit to a page (25.7 s, SD=24.7) was significantly higher (t-test, p=0.000) than mean time spent at revisited pages (17.1 s, SD=17.0). Mean time spent at top-level pages in the site hierarchy (17.5 s, SD=17.4) was significantly lower (t-test, p=0.000) than mean time spent at lower-level pages (23.8 s, SD=23.8). 6.4 Navigational correlates of success Information designers and log file analysts have a natural interest in the types of movement patterns and user navigation overviews that predict success and general satisfaction with their Web sites. In the post-browsing questionnaire, users in both conditions were asked to rate (on a 1–5 Likert scale) the extent to which they had felt confident they could locate the desired information in the site, and to what extent they felt their search strategies were planned; the latter therefore represented perceived planning of search. Success measures in this study included number of completed information-seeking tasks and number of abandoned tasks. Correlation analyses, summarized in Table 4, are done within condition for comparison. Table 4 Navigational correlates of information-seeking success, confidence, and perceived planning of search (Pearson correlation coefficients and p-values included) Control condition . . . r . p . Completed tasks Top-level visits 0.721 0.019* Completed tasks Upward movement (%) −0.693 0.026* Abandoned tasks Downward movement (%) −0.753 0.012* Confidence Navigation acts 0.779 0.008** Confidence Average page time −0.842 0.002** Confidence Revisits 0.804 0.005** Confidence Downward movements 0.823 0.003** Site map condition Abandoned tasks Lower-level visits 0.736 0.015* Abandoned tasks Planned search −0.630 0.051 Planned search Site map clicks −0.704 0.023* Control condition . . . r . p . Completed tasks Top-level visits 0.721 0.019* Completed tasks Upward movement (%) −0.693 0.026* Abandoned tasks Downward movement (%) −0.753 0.012* Confidence Navigation acts 0.779 0.008** Confidence Average page time −0.842 0.002** Confidence Revisits 0.804 0.005** Confidence Downward movements 0.823 0.003** Site map condition Abandoned tasks Lower-level visits 0.736 0.015* Abandoned tasks Planned search −0.630 0.051 Planned search Site map clicks −0.704 0.023* *Significant at p<0.05. **Significant at p≤0.01. Open in new tab Table 4 Navigational correlates of information-seeking success, confidence, and perceived planning of search (Pearson correlation coefficients and p-values included) Control condition . . . r . p . Completed tasks Top-level visits 0.721 0.019* Completed tasks Upward movement (%) −0.693 0.026* Abandoned tasks Downward movement (%) −0.753 0.012* Confidence Navigation acts 0.779 0.008** Confidence Average page time −0.842 0.002** Confidence Revisits 0.804 0.005** Confidence Downward movements 0.823 0.003** Site map condition Abandoned tasks Lower-level visits 0.736 0.015* Abandoned tasks Planned search −0.630 0.051 Planned search Site map clicks −0.704 0.023* Control condition . . . r . p . Completed tasks Top-level visits 0.721 0.019* Completed tasks Upward movement (%) −0.693 0.026* Abandoned tasks Downward movement (%) −0.753 0.012* Confidence Navigation acts 0.779 0.008** Confidence Average page time −0.842 0.002** Confidence Revisits 0.804 0.005** Confidence Downward movements 0.823 0.003** Site map condition Abandoned tasks Lower-level visits 0.736 0.015* Abandoned tasks Planned search −0.630 0.051 Planned search Site map clicks −0.704 0.023* *Significant at p<0.05. **Significant at p≤0.01. Open in new tab The two conditions did not share any correlations, reemphasizing the vast differences between their behavioral effects. Number of completed tasks in the control condition correlated positively with frequency of top-level visits, but negatively with proportion of upward movements in the space. Similarly, abandoned tasks correlated negatively with proportion of downward movements. Reported confidence in the control condition generally correlated strongly with rapid movement in the space, shown by positive correlation with frequency of navigational acts and page revisits, and negative correlation with mean time per page. Proportion of downward movements correlated negatively with number of abandoned tasks, and frequency of downward movements correlated positively and strongly (r=0.823, p=0.003) with reported information-seeking confidence. Interestingly, reported confidence and actual completed tasks did not correlate in the control condition (r=0.411, p=0.238), although they did in the site map condition (r=0.677, p=0.032). Number of abandoned tasks did not correlate with reported confidence in either condition. Perceived organization of search strategy did not quite correlate significantly with number of abandoned tasks in the site map condition, but approached a p<0.05 level at 0.051. 7 Discussion 7.1 Click-stream analysis and site map clicks Consistent with previous data on context effects, the constantly visible site map resulted in fewer Back button clicks. This result, just as it has in previous work in the area, had clear implications on the number of total movements made in the two conditions. The mean proportion of Back button clicks in the control condition (14.02%) actually suggests the five chosen sites provided a decent amount of navigational aid to users, if one considers the reports of 41% in 1995 and 30% in 1997. The focus on relatively small Web sites likely contributed. Additionally, it may be that Back button use is becoming less prominent, but a more probable conclusion is that Web designers are simply becoming wiser as the Web itself becomes older, providing more useful navigational support, reducing the need for browser-supported backtracking. The similarities between the mean revisitation proportion in this study for the control condition (56.05%) and the Tauscher and Greenberg study (58%) are particularly interesting in light of the Back button statistics. Back button clicks are a major factor in page revisitation, and in this study's control condition, they accounted for about half the proportion of movements as they did in the Tauscher and Greenberg study. One would expect, then, the revisitation number to decrease. The likely reason it did not is, once again, the focus of this study on relatively small Web sites. Small navigation spaces have the feature of becoming recurrent systems more quickly than large ones, since the fraction of total nodes that have been already visited in the system can increase more rapidly. Decreased Back button clicks compared to previous work in the area decreased page revisitation, but the focus of this study on small sites compensated, suggesting the close replication of the 58% revisitation rate is likely somewhat coincidental. In general, space restrictiveness may indeed be a factor in page revisitation rates, but further work is needed to fully understand its impact. Table 1 shows the predominance of long-distance movements in the constant site map condition, expressed by ‘distal’ movements, or a mean proportion of 36.9% of navigational acts. The availability of such movement was clearly taken advantage of by site map participants. Long-distance movements in a hyperspace have a number of potential causes; with respect to fact-finding, we may split these causes into two main types: The navigator feels the current subspace is not relevant to the desired information. The navigator feels the current subspace is relevant to the desired information, and that the destination node is related to the current position (and so relevant to the desired information). Both of these cases represent an enhancement in the navigational capabilities of a Web site visitor who has a constantly visible site map. In the former, the user abandons a subspace for a more promising area of the site. In the latter, the user is able to formulate and use a relevance relationship between nodes in the space that a navigator without the overview would not be able to formulate. The constant overview provides another type of user flexibility, namely, the freedom to examine multiple levels of site detail without a navigational act. In the control condition, a disproportionate amount of time was taken from top-level page visits compared to lower-level visits, although this effect did not occur for site map participants. Not surprisingly, control participants needed the additional time in, and additional visits to, a site's top-level in order to gain overview information that was constantly available to site map participants. The fact that navigators with a constantly visible overview will dig deeper into the site structure should be particularly instructive from an industry point of view: users will tend to spend time at detailed pages, confronted with more content from a site on a continual basis, requiring more extraction tasks. All of this means more time looking at your information, and potentially more time on your site. Depending on the type of site and likely user goals, this may or may not be desirable. Global analyses may be integrated with intra-site experiments such as this one to determine how navigational support differences across a wide range of sites a user is confronted with affect her movement behavior, and whether such effects are desirable given particular goals on the part of the site designers. 7.2 Time spent on a Web page The results suggest a user is likely to spend more time on a destination page if: The page is not accessed via the browser's Back button. The user is visiting the destination page for the first time. The hierarchical direction of movement from source to destination is downward. The destination page is deep in the site hierarchy. The source and destination pages have distal familial relationships in the site hierarchy. The page is accessed via ‘top-level switch.’ Multiple Back button clicks in rapid succession, and the button's common use in hub and spoke traversal (Catledge and Pitkow, 1995), likely contributed to result (i). Page accesses via Back button clicks necessarily represent page revisits, which also helps predict decreased time spent on the destination page, per result (ii). Result (iii) is consistent with lower-level pages in the site hierarchy involving more time on average, per result (iv). Both may speak to lower-level pages in a Web site typically containing more content and specificity than top-level pages. However, page word count was not analyzed in this study, and thus these are only speculative potential factors. Results (v) and (vi) may provide more interesting insights into user orientation in hyperspace. Distal movements outside the immediate hierarchical family of the source page and movements to different subspaces of a site typically result in greater Web interface changes, especially when the site designer provides primarily local structural navigation aids, such as the siblings and children of the current page. In particular, pages within a single subspace often share a visual style, such as a particular icon or color coding associated with that section of the site; a movement to another subspace is accompanied by changes in these visual features. Moreover, distal movements typically involve more drastic changes in the structural navigation aids, since the source and destination are likely to share fewer structural links. If we assume these types of movements do not inherently (or consistently) cause the user to focus on the content of the destination page more than she normally would, the difference in time spent results from the more dramatic changes in navigation support that result from these types of movements; they are more navigationally volatile, and force the user to focus on the navigation aids more than she normally would. Additional research is needed in support of these explanations, and to fully understand the effects of navigational volatility and its relationship to user perceptions and performance on the Web. The author's current research focuses on these issues. 7.3 Navigational correlates of success Most notably, the indications of success from a navigator's click-stream behavior may be dependant upon the specifics of the navigation design that user was confronted with during the navigation session; control and site map participants shared no navigational correlates of success. These results suggest that predictions of information-seeking success based solely on user movement, without consideration or control for navigational aids, are at least questionable. The reported confidence of control condition participants correlated well with rapid movement in the space, and frequent page revisits. Successful control participants needed to both gain overview information (frequent top-level page visits) and dig into the site structure (proportion of downward movements), which requires frequent page transitions. 7.4 Conclusions Information designers will benefit from data regarding the effects of differing navigational design decisions. Data showing user performance to be superior in one context over another is certainly useful, but does not address the full spectrum of designer concerns. This study provides data for single-session, within-site information-seeking, exposing some of the vast behavioral effects of presenting a constantly visible site overview of a small site or subspace of a larger one. Applying the general approach and methodology used in this study to other types of tasks and behaviors, such as exploratory search and general site understanding tasks, will make for interesting future research. This experiment first applies a fine-grained click-stream methodology to exploring issues of formal search, of great concern to designers of information-centric Web sites. Used in conjunction with data on other task types, information designers may develop a fuller picture of how particular design decisions will affect the behavior of their users. Acknowledgements Many thanks to Terry Winograd, Clifford Nass, and Todd Davies for valuable feedback throughout the development of this work, and to the referee of a draft of this paper for very useful comments. References Catledge and Pitkow, 1995 Catledge L Pitkow J , Characterizing browsing strategies in the World Wide Web , Computer Networks and ISDN Systems 27 ( 1995 ) 1065 1073 Google Scholar Crossref Search ADS WorldCat Choo et al., 2000 Choo C.W Detlor B Turnbull D , Information seeking on the Web: an integrated model of browsing and searching , First Monday 5 ( 2 ) 2000 OpenURL Placeholder Text WorldCat Cockburn and McKenzie, 2001 Cockburn A McKenzie B , What do Web users do? An empirical analysis of Web use , International Journal of Human–Computer Studies 54 ( 6 ) 2001 ) 903 922 Google Scholar Crossref Search ADS WorldCat Conklin, 1987 Conklin J , Hypertext: an introduction and survey , IEEE Computer 20 ( 7 ) 1987 ) 17 41 Google Scholar Crossref Search ADS WorldCat Darken and Sibert, 1993 Darken R.P Sibert J.L , A toolset for navigation in virtual environments , ACM User Interface Software and Technology'93 1993 ) 157 165 OpenURL Placeholder Text WorldCat Dee-Lucas and Larkin, 1995 Dee-Lucas D Larkin J.H , Learning from electronic texts: effects of interactive overviews of information access , Cognition and Instruction 13 ( 1995 ) 431 468 Google Scholar Crossref Search ADS 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 and Practice 1989 Intellect Books , Oxford 105 125 OpenURL Placeholder Text WorldCat Furnas, 1986 Furnas G.W , Generalized fisheye views , CHI'86, Human Factors in Computing Systems 1986 ) 16 23 OpenURL Placeholder Text WorldCat Greenberg and Cockburn, 1999 Greenberg S Cockburn A , Getting back to back: alternative behaviors for a Web browser's back button , Proceedings of the Fifth Annual Human Factors and the Web Conference 1999 OpenURL Placeholder Text WorldCat Lin et al., 2000 Lin J Newman M.W Hong J.I Landay J.A , DENIM: finding a tighter fit between tools and practice for Web site design , CHI'00, Human Factors in Computing Systems 2000 ) 510 517 OpenURL Placeholder Text WorldCat 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 ( 2 ) 1998 ) 129 142 Google Scholar Crossref Search ADS WorldCat McKenzie and Cockburn, 2001 McKenzie B Cockburn A , An empirical analysis of Web page revisitation , IEEE Proceedings of the 34th Hawaiian International Conference on System Sciences 2001 OpenURL Placeholder Text WorldCat Meyer et al., 1997 Meyer B.A Sit R.A Spaulding V.A Mead S.A Walker N , Age group differences in World Wide Web navigation , CHI'97, Human Factors in Computing Systems 1997 ) 295 296 OpenURL Placeholder Text WorldCat Miller, 1956 Miller G.A , The magical number seven, plus or minus two: some limits on our capacity for processing information , The Psychological Review 63 ( 1956 ) 81 97 Google Scholar Crossref Search ADS PubMed WorldCat Nation et al., 1997 Nation D.A Plaisant C Marchionini G Komlodi A , Visualizing Websites using a hierarchical table of contents browser: WebTOC , Proceedings of the Third Annual Human Factors and the Web Conference 1997 OpenURL Placeholder Text WorldCat Newfield et al., 1998 Newfield D Sethi B.S Ryall K , Scratchpad: mechanisms for better navigation in directed Web searching , ACM User Interface Software and Technology'98 1998 ) 1 8 OpenURL Placeholder Text WorldCat Newman and Landay, 2000 Newman M.W Landay J.A , Sitemaps, storyboards, and specifications: a sketch of Web site design practice , ACM Designing Interactive Systems'00 2000 ) 263 274 OpenURL Placeholder Text WorldCat Nielsen, 2000 Nielsen J , Designing Web Usability 2000 New Riders , Indianapolis, IN Norman, 1991 Norman K.L , The Psychology of Menu Selection 1991 Ablex , Norwood, NJ O'Day and Jeffries, 1993 O'Day V.L Jeffries R , Orienteering in an information landscape: how information seekers get from here to there , CHI'93, Human Factors in Computing Systems 1993 ) 438 445 OpenURL Placeholder Text WorldCat Park and Kim, 2000 Park J Kim J , Contextual navigation aids for two World Wide Web systems , International Journal of Human–Computer Interaction 12 ( 2 ) 2000 ) 193 217 Google Scholar Crossref Search ADS WorldCat Rada and Murphy, 1992 Rada R Murphy C , Searching versus browsing in hypertext , Hypermedia 4 ( 1 ) 1992 ) 1 30 OpenURL Placeholder Text WorldCat van Schaik and Ling, 2001 van Schaik P Ling J , The effects of frame layout and differential background contrast on visual search performance in Web pages , Interacting with Computers 13 ( 5 ) 2001 ) 513 525 Google Scholar Crossref Search ADS WorldCat Streeter et al., 1985 Streeter L Vitello D Wonsiewicz S , How to tell people where to go: comparing navigational aids , International Journal of Man–Machine Studies 22 ( 1985 ) 549 562 Google Scholar Crossref Search ADS WorldCat Tauscher and Greenberg, 1997 Tauscher L.M Greenberg S , How people revisit Web pages: empirical findings and implications for the design of history mechanisms , International Journal of Human–Computer Studies 47 ( 1997 ) 94 137 Google Scholar Crossref Search ADS WorldCat Wright, 1991 Wright P , Cognitive overheads and prostheses: some issues in evaluating hypertexts , ACM Hypertext'91 1991 ) 1 12 OpenURL Placeholder Text WorldCat © 2002 Published by Elsevier Science B.V.
TI - Web navigation and the behavioral effects of constantly visible site maps
JF - Interacting with Computers
DO - 10.1016/S0953-5438(02)00024-3
DA - 2002-10-01
UR - https://www.deepdyve.com/lp/oxford-university-press/web-navigation-and-the-behavioral-effects-of-constantly-visible-site-hlMwOrDu61
SP - 601
EP - 618
VL - 14
IS - 5
DP - DeepDyve
ER -