Website Context and History: An Overview and Design Proposal Luc Tiemessen luctiemessen {at} gmail {dot} com ABSTRACT In this paper, we describe an overview of the work that is being done for supporting users while browsing the World Wide Web (WWW). We describe the problems users face when visiting multiple pages in different website environments. We evaluate different approaches to solve the “Lost in Hyperspace” (Conklin, 1987) problem. While evaluating several articles, we distinguished two different approaches; the first is the evaluation of the “Back” button and its implementation, the second is the facilitation of (navigational) context in order to give users handles for orientation within the hyperspace. At first these two approaches seem different but we think the functionality of both approaches overlap each other and can potentially support each other in their core function. Finally, we will take a look at how aesthetics can or should play a role in the design of the additional (navigational) context and propose an implementation for the back button and present guidelines for page designers. Keywords User interfaces, Browsers, Web Navigation, History and revisitation, Navigational Maps. the “disorientation problem”. We will give an overview of the work done in solving this disorientation problem where additional features are added to browsers in order to prevent users from having this problem. In addition, we discuss the back button for revisiting pages users have already visited. The “hub-and-spoke” navigation (Catledge & Pitkow, 1995) is one type in which the back button is very useful but fails to remember all the pages users visit. This is caused by the functional model behind the back button. The model behind the back button is a stack (Greenberg, Ho, & Kaasten, 2000) which in some circumstances pops a visited page of the stack and replaces it with a newer page. This behavior is not always desired. Revisiting pages is 58% of all the page visits web users do when they browse the web (Tauscher & Greenberg, 1997). Other recent experiments point out that this percentage can be between 38% (Herder, 2006, p. 128), and 80% (Cockburn & McKenzie, 2001) so with the millions of people browsing even more webpages every second of the day, any minor improvements on this mechanism is advisable. 2. LITERATURE STUDY 1. INTRODUCTION The WWW has expanded exponentially in the past decades, with enormous increase of web pages, images, and complex content. The amount of people using the web (users) has also increased enormously. One of the things that has been left behind are the web browsers and their user interaction design (UID). Browsers are constantly updating their engines with new support for W3C standards (W3C, 2008) and improving their preformance but the graphical user interface (GUI) has not evolved significantly (apart from being slimmed down). The “Back” and “Forward” buttons already appeared in the first major browsers (NCSA, 2009) and are still present in modern browsers like Firefox (Mozilla, 2009) and Safari (Apple Inc., 2009). Not much browsing functionality has been added in the past few years to aid its users while navigating. The “Lost In Hyperspace” (LIH) problem is known for years (Conklin, 1987) and is a problem that is possibly to be solved by the browser developers. Conklin states that users can lose track of their virtual environment, not being aware of their location on the page and navigational possibilities. He calls this Permission to make digital or hard copies of all or part of this work for personal or classroom use is granted without fee provided that copies are not made or distributed for profit or commercial advantage and that copies bear this notice and the full citation on the first page. To copy otherwise, or republish, to post on servers or to redistribute to lists, requires prior specific permission. 1st Twente Caput & Referaat Conference, Enschede 14 June 2004 Copyright 2004, University of Twente, Faculty of Electrical Engineering, Mathematics and Computer Science Lost in hyperspace occurs when users are not able to pinpoint their location in a website. This problem can have many causes. The most important, is the lack of contextual information and structure in the website. When users experience cognitive overload or disorientation on a webpage (Park & Kim, 2000), this is often caused by the lack of spatial and temporal context. Spatial context meaning awareness of the location on the webpage through structure or visual aids like site maps, navigation menus and index pages. Temporal context meaning awareness of the way that led to this situation by showing back and forwards buttons, visual navigation histories and bookmarks (Herder, 2006, p. 133). Much research is done on extending browsers with the capability of showing more contextual information. An overview of different research on designing for WWW navigation is presented by (Cockburn & Jones, 1992). They identify five issues in the design and functionality of the visualization systems they reviewed: 1. 2. 3. 4. 5. Characteristics of the visual representation; Scope of the subspace representation; Mechanisms for generating the visualization; Degree of browser independence; Navigation support facilities. These five issues form a nice starting point for our search for improvements in the design of WWW navigation as a whole (browser, website, user and their interaction with each other). When searching for research on browser capabilities, backwards navigation, user support, contextual information, and solving the lost in hyperspace problem, it soon becomes clear that a lot of research has been done varying from extending the browser with custom programmed tools (Cockburn, Greenberg, McKenzie, Jasonsmith, & and Kaasten, 1999; Medynskiy, Dontcheva, & Dontcheva, 2009) to the usage of the back button (Greenberg, Ho, & Kaasten, 2000; Greenberg & Cockburn, 1999; Orner & MacKenzie, 2006). One common behavior of users is the hub-and-spoke browsing method where users return to an index page with several links to “child” pages in order to search for specific information or just to scan the content of the entire website (Catledge & Pitkow, 1995). Thus, a hub contains a collection of links leading to a child website (“spoke”). When users search for information from a hub and return to this hub when no satisfactory information was found on the spoke, users go back to the hub and try to find the desired information. This going back to the original information point is called backtracking (Bieber & Wan, 1994). The amount of backtracking activities by users depends greatly on their task at hand. When users are executing a goal directed task, (i.e. finding a specific piece of information) backtracking is more common than when users are freely exploring the web (i.e. browsing for fun or scanning for news on news websites) (Herder, 2006, p. 128). In the next sections we will distinguish the back button and the context features. With context features we mean the contextual information displayed on the webpage to prevent users from being lost in hyperspace. The back button will be analyzed separately on its current behavior and related research. In the results section we try to find where these functionalities can join forces. 2.1 Back button As described above, of all the visits of web pages, almost 60% is a page that is visited before by the same user. This user accesses these pages though normal browsing (typing the appropriate URL or navigating via another hyperlink), bookmarks (3%), URL’s accessed via the browser’s history feature (<1%), the forward button (1%) and most importantly, the back button (30%) (Tauscher & Greenberg, 1997; Kaasten & Greenberg, 2000). With this vast amount of usage, one should expect users to fully comprehend functionality or model behind the back button. This is however, very untrue (Cockburn & McKenzie, 2001; Greenberg, Ho, & Kaasten, 2000). They have indicated that users cannot predict the behavior of the back button correctly (even sometimes when the users were certain about their answer) and use the back button until they visually recognize the page they are looking for. This asks of a serious rethinking about the usefulness of the model behind the back button, even more since the current model has a big disadvantage over other possible models. At this moment of writing, most modern browsers have implemented the same model behind the back button. The back button is not just an indicator to a simple list; the most fitting name of the underlying model of the back button is a “stack”. 2.1.1 Stack back Most modern browsers have implemented the stack as a model behind the back button. In this section we describe the actual functionality of this model and what the advantages and drawbacks are. The stack is analyzed by (Greenberg, Ho, & Kaasten, 2000) and works as follows: Figure 1 Browsing path Figure 1 shows the click path of a user. The black (straight) arrows represent the clicking action on a hyperlink; the grey (curved) arrows represent the users’ action of clicking on the back button of the browser (thus, not via internal hyperlinks). The user navigates from node A to B, to C, to D. Now the stack contains the following elements in chronological order (most recent on top/end): [A,B,C,D]. Then, the user uses the back button to go back to page B, on this page B, a link to page E is clicked and with that, the nodes C and D are pruned from the stack. The stack now looks like this: [A,B,E] and users are unable to get back to node C and D with only the back and forward buttons. The stack has great support for hub and spoke browsing and backtracking but we think the drawback of the pruning of the nodes might be one that is solvable with the right model, implementation and facilities. The stack is not the only model that has been a subject of research over the years. Recency is a variant which does not have the drawback of pruning but throws up other problems. 2.1.2 Recency back Recency is a model presented by (Greenberg, Ho, & Kaasten, 2000) which differs from a stack in a way that the most recently visited page is put on the end of a list. No matter how many times, a user clicked on the back button, the latest visited site is always on the end of the list. The biggest advantage of this is that no pages are pruned. Other research has showed that the Lost in Hyperspace occurs fewer times when the recency back model is used (Orner & MacKenzie, 2006). This phenomena can be explained by the simple fact that users are not looking for pages they cannot find since they are always (even if it takes a few more clicks) available. Recency back cannot be just an ordinary list with the most recently visited node on top. When users follow the path of Figure 1, when these users reach node D, the list looks like this: [A,B,C,D] to go back to node b many people would probably suggest two clicks back. This is wrong. Strictly taken, when clicking the back button from node D, the newest node on top of the stack would become C, resulting in the list to become [A,B,C,D,C] so far so good but now we can see the problem when we hit the back button for the second time. Since C is the new latest item and the user came from D, now, D is the most recently visited node. This model would switch endlessly from C to D and back. So to make this model usable we have to make some adjustments and Greenberg came up with multiple implementations (Greenberg & Cockburn, 1999). Their starting point was to remove the duplicates from the list, pruning al duplicates by keeping only one single copy in its most recent position in order to keep the list of items as small as possible. However, this did not solve the cycling problem of the last two nodes. Therefore they came up with three different implementations. • • • Adding spokes only Hub and spoke enhancement Temporal ordering enhancement Adding spokes only adds all new pages to the end of the list, removing the older ones in the list. This however results in bad performance for hub and spoke navigation. When only spokes are added to the end of the list, while hitting the back button to get back to the hub, users have to hit the back button n times, where n is the amount of spokes accessed from the latest hub. Hub and spoke enhancement solves the previous problem by adding both previous and current node to the end of the list. The temporal ordering enhancement is an implementation where the exact order of the previous nodes is preserved. This can be achieved by keeping track of a second list of visited nodes when users hit the back button. This second list can be added at the end of the first list and before the new node when the user clicks a new link. as showing the title of the recently visited pages). In addition they indicated where hub pages in the history list were located so users are able to quickly access the important pages. In their research they concluded that all these alternatives can produce unpredictable results for users. Most of these researches showed improved results on backtracking support, reducing “lostness” and revisitation. 2.2 Context features With context features we try to discover features designed by page designers and browser developers to support the user in keeping track of where they are in hyperspace. In the information visualization research area there has been a lot of work done on usability guidelines by Nielsen (2002), Shneiderman (1998) and many, many more. These guidelines give page designers and browser developers handles to design and develop a good user experience on the web by pointing out what they should keep in mind while designing or developing websites or browsing applications. Many of these rules are applicable to our problem of providing context to users (i.e. location, click path, origin, shortcuts, possibilities etc.). We will give an overview of some of the research done on improving contextual information for users within hyperspace. One of the many browser features researchers are addressing is the visualization of the browsing history. One could see it as a list with icons, text, images that represent the stack model underneath the back button. This gives the user information on how they have reached the current location and gives them the possibility to visually recognize recently visited pages (Cockburn, Greenberg, McKenzie, Jasonsmith, & and Kaasten, 1999; Chakrabarti, Gibson, & Mccurley, 1999; Ayers & Stasko, 1995). In their experiments they all implemented a sidebar or separate screen in which all visited pages were displayed with title, thumbnail and hierarchical structure. This gives the user the ability to view the path they followed to get to the current location, accompanied with information about how the pages visually looked by including thumbnails of the visited web pages. Other research proposed improvements on the back button (Milic-Frayling, Jones, Rodden, Smyth, Blackwell, & Sommerer, 2004; Orner & MacKenzie, 2006). These experiments made it possible to see the navigational history in textual form. Orner’s research gives users insight in the model behind the back button. In their research they use the recency back model and make it accessible with only the back and forward buttons, therefore facilitating users to maintain all of their browsing history and making it accessible with common control features (back and forward buttons). Although we think these solutions are somewhat interruptive, they show interesting approaches to maintaining the history list. Milic-Frayling’s research improved the back button with adding groupings to the buttons’ dropdown functionality (where the user can hold the back button down, showing a list of visited pages, almost all modern browsers have this functionality but only implemented Other research implemented a history search system for back navigation. (Won, Jin, & Hong, 2009; Kaasten & Greenberg, 2000) with thumbnails. These researches experimented with the improvement of the history list. The history list is the list of visited items and in most cases in modern browsers this is another list of visited pages than the stack list behind the back button. The history list is the complete list of most recently visited pages, in contrast with the back button; no nodes are pruned regardless of the browsing path. Their research improved the history with a search function, thumbnails and more contextual information about the visited webpage. Park & Kim (2000) evaluated the actual availability of contextual information in general, evaluating two different types of websites measuring the browsing efficiency and improving subjective ease of navigation. The measurements showed significant improvements of the performance of browsing tasks, regardless of the type of website. They carried out their research with only page design enhancements. No additional browser dependent application was written. They carefully thought about the possible navigational actions of the users and adjusted the page design to enable easier hierarchical navigation with links to siblings, parents, grandparents and children. This adding of important links can improve users’ question “what can I do?” or “where can I go?” another important question remains unanswered: “how did I get here?”. By adding temporal content to the page, users are able to answer this question (Theng, Thimbleby, & Jones, 1996). We evaluated research concerning the back button and contextual information. When we look at the many researches done by others we can see that some of the researchers who proclaim to deliver more contextual information, are actually visualizing the previously visited web pages (Cockburn, Greenberg, McKenzie, Jasonsmith, & and Kaasten, 1999; Kaasten & Greenberg, 2000; Milic-Frayling, Jones, Rodden, Smyth, Blackwell, & Sommerer, 2004; Tabard, Mackay, Roussel, & Letondal, 2007) accessible through the back button. When delivering information about the history of the visited web pages, the users get more contextual information. The only problem is that this information is not instantly visible or accessible via the back button although as few as possible actions should be carried out to reach this information (Tauscher & Greenberg, 1997). 3. RESULTS As described earlier in this paper, users do not have a good mental model about the functionality of the back button and use the button to go back (as if it was a plain list) and visually recognize the page they are looking for. When this page does not appear, users can feel lost in hyperspace. We think that if users have a poor mental model of the back button we should consider removing the big disadvantage of this stack namely, the pruning of visited pages. We think that the advantages of recency – being no loss of visited pages – will be bigger than their downside – being possible unpredictable behavior. Since users tend to “click-till-recognize” (ironically, caused by their poor understanding of the stack back button) we think the most important thing to do is to store every visited page. Given the fact that users tend to “click-till-recognize” we think we can support the users in giving information about the previously visited pages before they reach the desired page. By simply visualizing these pages by icons that represent the page (Teevan, et al., 2009) or by displaying a thumbnail of the entire page when only hovering over the back button. The hover action can initiate a small non-interruptive screen which shows the user an overview of the most recently visited web pages. by other browsers – was considered while designing the alternative. This possible solution obviously implies browser vendors to implement it. The browser should keep track of a list of URL’s and make screen captures of these websites and link them to their URL. Page designers cannot do this with the current technology, although Firefox’ extension library makes it easier for third party developers to add functionality to the browser but this is highly browser dependent. The list of design guidelines we propose in order to give users more recognition, handles and feedback is described below. Research has already indicated that it will be useful when for every page can be determined if it is a hub (Milic-Frayling, Jones, Rodden, Smyth, Blackwell, & Sommerer, 2004). We think the World Wide Web Consortium (W3C, 2008) can play a crucial role in this. If they make it possible for page designers and developers to manually indicate hub pages, this could be used by browser vendors to support users with visualizations. While searching for information on solving the lost in hyperspace problem we took a step back and came to the following conclusion: No matter how ingenious the implementations of the researchers were, many researches seem to ignore the role of the design of the web page itself since they try to develop browser extensions or external tools to add more functionality on the browser by implementing additional windows or additional interactive elements within the page. They did not question the possibility of having current technologies solve these problems with a few exceptions (Danielson, 2002; Hanson, 2007). Some research was done based on technologies that should (due to accessibility drawbacks) no longer be used like html frames (Chakrabarti, Gibson, & Mccurley, 1999). We think that many improvements can be made nowadays with design guidelines instead of additional features to resolve the lost in hyperspace problem because the development of CSS, JavaScript, and XHTML and Server Side Scripts like PHP or Python can deliver more information than currently used (in general). Not only the website technology has evolved, also the hardware technology has, users nowadays have more CPU power, more screen real estate and more bandwidth than several years ago. Bigger resolutions can lead to more detailed information, more CPU power can lead to more complex interaction patterns, and more bandwidth can lead to more interaction within a page. In unison of what other researchers argued (Theng, Thimbleby, & Jones, 1996) we think many of the lost in hyperspace problems are caused by bad page design. This is why we decided to formulate some design guidelines that meet today’s possibilities and will be suitable for page designers to use in their daily work. 4. DESIGN CONSIDERATIONS This section describes the design considerations that can be made when implementing our suggestions. As stated earlier, for the contextual information enhancement we will set up some guidelines for page designers. These guidelines will partly overlap with the many earlier written guidelines but we focused on the support of context information to prevent lost users. The back button design considerations result in some proposals for the implementation of our improvements. In designing the improvements, browser independence was taken into account. This means that no specific browser functionality – not shared We split this chapter in two sections. The first handles the visual design guidelines for contextual information, in the second we describe our reflections on redesigning the behavior of the back button. 4.1 Visual design guidelines Expose page origin and destination If the site facilitates a strong hierarchical structure, enable users to see it as such. Remember to show users the way they took to get where they currently are and think along with the users about their possible destinations and highlight these (Conklin, 1987; Weinreich, Obendorf, & Lamersdorf, 2001). This facilitates the user with a logical “back” or “undo” action. Visualize the page hierarchy Hierarchy or sections can be visualized with color or visual indentation. Visual distinction helps users to indicate they are in another section of the site (Conklin, 1987). Use color with tact Design visited links with different colors so users are visually triggered (Weinreich, Obendorf, & Lamersdorf, 2001). Keep navigation nearby On every location of the page, the user must have some clue on how to get to a location where he can choose which way he wants to go. This does not necessarily mean that the navigation should be visible on every spot on the page, though users should be able to reach the navigation with ease. Visualize impact If links are taking the user to another domain or window, clarify this with an icon (Weinreich, Obendorf, & Lamersdorf, 2001). Design for content Rethink the design after content is added; consider the contextual feedback on pages with a great vast amount of text, especially in these pages, users need contextual information. This can occur when a text is consuming all screen real estate leaving no place for navigational aids. One could think to divide the text into (logical numbered) subsections so users can see their progress or using only partial screen real estate to indicate where the navigation is located which is easily accessible when desired. This problem originates also in the conventional print medium where people can get lost in linear text (Conklin, 1987, p. 38). Although this issue can be addressed with previously described guidelines (Expose the page structure and Keep navigation nearby) we explicitly would like to describe that information on a page can grow exponentially in a short amount of time (considering big content management systems that allow multiple editors to work on one specific item). Do not take the back button for granted Facilitate users on the page with handles for back navigation; users want to feel in control (Shneiderman, 1998). Use images for branding Logo’s and color are well recognized by users (Teevan, et al., 2009) and usable for revisitation. Beautiful design is good design… Good aesthetics can have positive effects on usability. Even though beauty might be subjective, it actually can have a function of improving usability (Parizotto-Ribeiro & Hammond, 2005; Tractinsky, Katz, & Ikar, 2000). Separate content from navigation As much as possible, find one place for the navigational facilities, therefore users can concentrate on the content and when needed quickly recognize the navigation on the page. Design by page The amount of time spent on the page is dependent of many variables, one of them the level in the hierarchy (Danielson, 2002). Anticipate on the time spent on the page with visual clues for extra navigational items or letting users dive in the content. Also, evaluate your design for every page individually as if it is the first page the user visits, even if this page is deeply nested in your site structure. When your site is optimized for search engines, the search engines can pinpoint to one single page of your entire selection. In that case, users also have to be able to comprehend the site structure and their navigational possibilities. Design for your user Carefully think about your target Group when designing the page. Infants might have a different way of understanding navigation than elderly. Design a favicon Although the size is small, in history lists it is possible to stand out with your favicon. AJAX with thought With AJAX it is possible to request pages on your own websites without reloading, it is possible to enhance your internal links with a thumbnail of the target page. On every link this might clutter the interface but sparingly (i.e. with visited links) this can improve the revisitation within your website (Weinreich, Obendorf, & Lamersdorf, 2001). We think this list with design guidelines will give page designers handles for proper page design which minimizes lost in hyperspace problems for their visitors. 4.2 Functional design In this section we describe our considerations for redesigning the behavior of the back button. While evaluating web browsers we found one interesting visual design feature implemented on only one browser. Firefox 3.x (Mozilla, 2009) incorporated a big back button. Enabling users to find and use this button with more ease (Fitts, 1954). Figure 2 Firefox' bigger back button This relatively simple addition to the browser makes back navigation easier to access. Though the change seems marginal, we think this can greatly improve the overall usage of the back button since we stated earlier that every minor improvement can implicate great benefits because of the many times this button is used. We tried to come up with a simple yet effective, non real estate consuming, non interruptive, improvement of the back button while at the same time, giving users contextual information. This is where we think the lost in hyperspace problem and the back button can come together. When we thoughtfully design a feature that gives users feedback about the visited pages in a non-interruptive way while backtracking, this can support users while backtracking and show them contextual information on how they got to the current web page. While searching for a possible implementation we considered many alternatives and found a distinction between 2d and 3d solutions. A closely related topic is the performance of finding information within big datasets, filtering results to quickly access the desired information. We did not focus our research on measuring effectiveness of these implementations but we found an interesting conclusion about the use of 3d information visualization. Within small datasets (smaller than 100 items) 3d visualizations hinder the effectiveness of information finding within the available data (Cockburn & McKenzie, 2002). Since our research has shown that the amount of pages used in the history list should be between 6 and 10, the decision was made to propose a 2d suggestion. We decided to design an addition to the back button that would support the users while backtracking. This backtracking would be facilitated by the recency back model with temporal ordering and uses images of recently visited web pages so users can recognize the desired page and select it. Since users seem to have a poor mental model of the model behind the back button, we think giving insight in this additional model might confuse people. We think the model behind the back button is of minor importance for users as long as they are able to recognize the desired page. Another suitable candidate for supporting the back button is the history list. This is currently being implemented in several modern browsers where users are able to see the history list of most recently visited items when they “click and hold” the back button. It shows a list with page titles of the most recently visited items. We think the information given to users can be less interruptive and richer. We think the “click and hold” action is time consuming and interruptive and the list of page titles are not a good source of page recognition (Weinreich, Obendorf, & Lamersdorf, 2001). Therefore, we propose an overview of page thumbnails that appear while hovering the back button. Cockburn & Greenberg (2000) showed that textual history representation is not sufficient to quickly identify pages and in order to improve page recognition time; tools like ours should display individual pages in a way that aids rapid identification. We do this by making thumbnails of the visited pages. For determining the amount of thumbnails that should appear in the window, we decided the number should be between 6 and 10 items (Tauscher & Greenberg, 1997). They state that there is a 42% chance that the next URL the user would like to visit has not been visited before and a 43% change that this has occurred within the last 10 visits. As discussed in the beginning of this paper, 58% of all pages are revisits, so visualizing 43% means that more than 74% of all possible revisits are displayed within the last 10 pages. According to other research, long history lists are not worth considering since the screen real estate is (sometimes) limited and the cognitive overload of long lists is quickly achieved (Greenberg, 1993). The last design consideration was described by Tauscher & Greenberg (1997) where they proclaim that “Accessing the history mechanism should be minimally disruptive to the user’s current task.” With this last design consideration in mind we tried to design a solution that uses no additional screen real estate next to the browser, would not be disruptive to the users’ current task (namely clicking the back button) and would take cognitive overload into account. The result of the design process is shown in Figure 3. Figure 4 Design proposal for back button with hub enhancement (no exact temporal ordering) Figure 3 Design proposal for back button enhancement in Safari web browser Figure 3 shows a context menu with 9 thumbnails on a raster of 3 by 3 with the most recent visited page on the top left and the least recently visited page on the bottom right where the second most recent visited page is on the right of the most recent one. The menu has a pointing arrow towards the back button implicating their relation. In this situation, the context menu reserves space close to the back button (the closest horizontal section) for displaying three hub pages. These pages are separated with a line from other non-hub or less recently visited hub pages which order is based on recency. This solution might be confusing for users since the last visited page is not always located closest to the back button. Therefore, we propose another alternative which maintained the ordering of the first design proposal but explicitly highlighted the hub pages. While designing the panel, we put the “Settings…” button on the screen to think about useful settings users could define since they want to feel in control of the application (Shneiderman, 1998). The settings button could be used for the user to determine the amount of thumbnails shown in the menu, the size of the panel, the reaction time of the panel, disable the panel, set the opacity etc. We did however not evaluate these settings. In the example above, no hub and spoke enhancement is provided. We think the hub and spoke navigation could become available for users via the context menu. If it is possible to dynamically determine the hub page (or W3C comes with a standard for the page designers to describe if the page is a hub, this will be discussed later) Figure 4 illustrates how this can be used for easier access to hubs. To clarify the situation, we defined the third, fifth and sixth page in Figure 3 as hub pages. These three pages should become easier accessible for users to enhance hub and spoke navigation. Figure 5 Design proposal with hub enhancement and true temporal odering With this context menu for the back button we think it is easier for users to revisit web pages stored in the history list. Needless to say is that the way the context menu appears is of great importance. The usage of the back button should by no means be obstructed. When clicking the back button and the page takes some time to load, it might be possible to highlight (while the page or the context menu is still visible) the page in the context menu so every time the user hits the back button it gets information of where in the history list the user is going. 5. FUTURE WORK 6. CONCLUSIONS At first has to be determined whether the contextual menu proposed in this paper is beneficial for users in revisiting web pages. We would like to see tests on evaluating the performance of the contextual menu we proposed with users. With this information can be evaluated if the proposed design will support users in their page revisitation. If these tests prove otherwise, we should reconsider our design. The design guidelines have to be evaluated in order to see if the user and the page designer benefit from these rules while designing and visiting websites. Due to the enormous evolution of the web the lost in hyperspace problem still occurs since the amount of information users have to digest has also grown exponentially. With this vast amount of information, new and additional context has to be provided in order to prevent users from being disoriented in their web page. In addition, some other issues can be addressed. The size of the thumbnails might be subject for optimization where the drawback of used screen real estate is minimized and the visibility (where users can identify the page enough to determine the desired page) is sufficient. The contextual menu we proposed can also be evaluated when it is filled with the normal history list, leaving the back button with the stack back functionality for good hub and spoke support. The context menu than can only show the latest visited windows, maybe with hub enhancement. We have to determine if this mismatch between model of the back button and items in the context menu will not confuse users. The W3C can help browser vendors and page designers with delivering the standard in which page designers can indicate whether a page is a hub or a spoke or a page of another kind. When this feature can be styled with CSS, page designers can design specific links for navigating towards hubs. Currently page designers are able to define the type of page or their description, this is however metadata which is mainly used for search engine optimization. Obviously, internet users have to get used to these features. The generation of thumbnails we proposed opens a complete new array of possibilities if browser vendor also reveal these to the page designer who in this case can do many things with these thumbnails. For example, if the browser captures every visited page as a thumbnail, and keeps a list of the history with the URL and the corresponding thumbnail, page designers are able to use these thumbnails for internal links. In the case of revisits it is currently possible to style these links with CSS. If the user hovers over a visited link (this link must be in the history list) the browser vendor or the page designer (depending on the implementation) can display the stored thumbnail of the page next to the link in order for the user to preview the site he is going to visit. This can be very useful in navigational sections like menus, sitemaps, and breadcrumbs. When these features are combined, they give the user structural information and more information about where they are going. Especially for backtracking purposes this can be very useful. In this research we overviewed research currently done on the subject of supporting users to prevent the “lost in hyperspace” problem and discussed the functionality of the back button. When backtracking information, users tend to “click-tillrecognize” and since the current back button makes it possible to prune information from the list of recently visited items, we argued that a reconsideration of the back button was needed. Our main argument against the use of a stack related back button was the removal of nodes. Other research gave insight in alternative models of the back button. We chose the recency back button with duplicates removed and temporal ordering. Even though this model has its disadvantages, according to us it is a better alternative to the stack back model since all pages stay stored in the list of recently visited items. We also proposed design considerations based on findings in different research where researchers are trying to solve the problems with external tools, many of the problems are solvable with proper page design and thinking about the user of the website. Finally, we proposed a browser design implementation that makes the back button more usable by providing insight in the most recently visited pages on a non disruptive way. REFERENCES Adaptive Path Inc. (2009). Adaptive Path. Retrieved May 23, 2009, from Ajax: A New Approach to Web Applications: http://www.adaptivepath.com/ideas/essays/archives/000385.php Apple Inc. (2009). Apple - Safari. Retrieved May 23, 2009, from http://www.apple.com/safari/ Ayers, E., & Stasko, J. (1995). Using Graphic History in Browsing the World Wide Web. Atlanta: Georgia Institute of Technology. Bieber, M., & Wan, J. (1994). Backtracking in a multiplewindow hypertext environment. ACM European conference on Hypermedia technology (pp. 158 - 166). Edinburgh, Scotland: ACM. Catledge, L., & Pitkow, J. (1995). Characterizing Browsing Strategies in the World-Wide Web. The Third International WWW Conference. 27, pp. 1065-1073. Elsevier Science B.V. As seen in other research, colors and logo’s are a good reference point for users when they revisit web pages (Teevan, et al., 2009). The favicon, now used in almost every browser, is too small to contain the information needed for users to recognize the page; therefore the favicon could be enlarged to meet the size where it can be more usable for selecting sites from a history list. Chakrabarti, S., Gibson, D., & Mccurley, K. (1999). Surfing the Web Backwards. WWW 8 Conference, (pp. 1679-1693). Browsers can improve the display of links to a webpage that is stored in the bookmarks of the browser. They can be visualized differently from the other links on the page. This gives users feedback that they are able to visit a page they have visited before (this page might not be in the history list when users empty their history list, the bookmark list remains unchanged). Cockburn, A., & Jones, S. (1992). Design issues for World Wide Web navigation visualisation tools. Computer Science Working Papers. Hamilton, New Zealand: University of Waikato, Department of Computer Science. Cockburn, A., & Greenberg, S. (2000). Issues of Page Representation and Organisation in Web Browser's Revisitation Tools. Australian Journal of Information Systems , 7 (2), 120127. Cockburn, A., & McKenzie, B. (2002). Evaluating the Effectiveness of Spatial Memory in 2D and 3D Physical and Virtual Environments. CHI 2002. Minneapolis, USA: ACM. Cockburn, A., & McKenzie, B. (2001). What Do Web Users Do? An Empirical Analysis of Web Use. International Journal of Human-Computer Studies , 903-922. Cockburn, A., Greenberg, S., McKenzie, B., Jasonsmith, M., & and Kaasten, S. (1999). WebView: A Graphical Aid for Revisiting Web Pages. Australian Conference on Human Computer Interaction. Wagga Wagga, Australia. Conklin, J. (1987). Hypertext: An Introduction and Survey. IEEE Computer , 17-41. Danielson, D. (2002). Web navigation and the behavioral effects of constantly visible site maps. Interacting With Computers (14), 601-618. Fitts, P. (1954). The information capacity of the human motor system in controlling the amplitude of movement. Journal of Experimental Psychology , 47 (6), 381-391. Greenberg, S. (1993). The computer user as toolsmith: the use, reuse, and organization of computer-based tools. Cambridge University Press. Greenberg, S., & Cockburn, A. (1999). Getting Back to Back: Alternate Behaviors for a Web Browsers' Back Button. Human Factors and the Web. Gaitersburg, USA. Greenberg, S., Ho, G., & Kaasten, S. (2000). Contrasting StackBased and Recency-Based Back Buttons on Web Browsers. Alberta, Canada: University of Calgary. Hanson, N. (2007). Wayfinding and Navigation: A Look at Browsing Via Aerial Map Style. Retrieved May 23, 2009, from Nils Hanson: http://nilshanson.com/research.html Herder, E. (2006). Forward, Back and Home Again - Analyzing User Behavior on the Web. PhD thesis, University of Twente, Enschede, The Netherlands. Kaasten, S., & Greenberg, S. (2000). Designing an Integrated Bookmark / History System for Web Browsing. Western Computer Graphics Symposium. Panorama Mountain Village, BC, Canada. Kaasten, S., & Greenberg, S. (2000). Integrating Back, History and Bookmarks in Web Browsers. Extended Abstracts of the ACM Conference of Human Factors in Computing Systems. Alberta, Canada: ACM Press. Medynskiy, Y., Dontcheva, M., & Dontcheva, S. (2009). Exploring Websites through Contextual Facets. CHI 2009 - Web Searching and Browsing (pp. 2013-2022). Boston, USA: ACM. Milic-Frayling, N., Jones, R., Rodden, K., Smyth, G., Blackwell, A., & Sommerer, R. (2004). Smartback: Supporting Users in Back Navigation. WWW 2004 (pp. 63-71). New York, USA: ACM. Mozilla. (2009). Firefox web browser. Retrieved May 23, 2009, from http://www.mozilla.com/en-US/firefox/firefox.html NCSA. (2009). About NCSA Mosaic. Retrieved May 23, 2009, from http://www.ncsa.uiuc.edu/Projects/mosaic.html Nielsen, J. (2002, May 12). Home Page Design Guidelines. Retrieved May 23, 2009, from Jacob Nielsen's Alertbox: http://www.useit.com/alertbox/20020512.html Orner, D., & MacKenzie, S. (2006). A Comparison of Hierarchical and Linear Browser Back Menus. IADIS International Conference WWW/Internet 2006, (pp. 119-126). Parizotto-Ribeiro, R., & Hammond, N. (2005). Does Asthetics Affect the Users' Perception of VLE's. Artificial Intelligence in Education (AIED), (pp. 25-31). Amsterdam. Park, J., & Kim, J. (2000). Effects of Contextual Navigation Aids on Browsing Diverse Web Systems. CHI '2000 (pp. 257264). The Hague, Amsterdam: ACM. Shneiderman, B. (1998). Designing the User Interface: Strategies for Effective Human-computer Interaction. AddisonWesley. Tabard, A., Mackay, W., Roussel, N., & Letondal, C. (2007). PageLinker: Integrating Contextual Bookmarks within a Browser. CHI 2007 (pp. 337-346). San Jose, USA: ACM. Tauscher, L., & Greenberg, S. (1997). How people revisit web pages: empirical findings and implications for the design of history systems. Int. J. HumanÐComputer Studies , 47, 97Ð137. Teevan, J., Cutrell, E., Fisher, D., Drucker, S., Ramos, G., André, P., et al. (2009). Visual Snippets: Summarizing Web Pages for Search and Revisitation. CHI 2009 (pp. 2023-2032). Boston, USA: ACM. Theng, Y., Thimbleby, H., & Jones, M. (1996). "Lost in hyperspace": Psychological problem or bad design. APCHI'96, (pp. 387-396). Singapore. Tractinsky, N., Katz, A., & Ikar, D. (2000). What is beautiful is usable. Interacting with Computers (13), 127-145. W3C. (2008). World Wide Web Consortium - Web Standards. Retrieved May 23, 2009, from http://www.w3.org/ Weinreich, H., Obendorf, H., & Lamersdorf, W. (2001). The Look of the Link - Concepts for the User Interface of Extended Hyperlinks. 12th ACM conference on Hypertext and Hypermedia (pp. 19-28). Århus, Denmark : ACM. Won, S., Jin, J., & Hong, J. (2009). Contextual Web History: Using Visual and Contextual Cues to Improve Web Browser History. CHI 2009 (pp. 1457-1466). Boston, USA: ACM.