Marion Walton, Vera Vukovic
When discussing how people get the "scent" of information on the Web, and thus how best to design for them, we seldom consider the role of literacy. Literacy, in its broadest sense, is not a neat parcel of skills easily packaged and given to someone. Anthropologists have found that literate human societies have evolved a wide range of literacies, each a complex set of practices and beliefs about communication and knowledge associated with particular educational, linguistic, and social contexts .
As human-computer interaction (HCI) professionals, we often generalize about the habits associated with reading from the Web, not realizing that the practices we most often describe are those of an elite group of knowledge workers (to which we also belong). Although such workers may have diverse national origins, they come from relatively similar educational backgrounds. Our work with South African students from disadvantaged educational backgrounds identifies some of the challenges in assisting them in joining the ranks of these knowledge workers. Cultural differences make it difficult for these students to make the transition to Web use, although not in the stereotyped notion of ethnic or national cultures. Rather, we need to acknowledge that Web use is situated within distinct cultures of reading and writing.
As South African Minister of Communications, Jay Naidoo imagined information networks as magic pipelines connecting African people to unlimited supplies of liberating information . In this view, literacy is neglected, as it is in many discussions of the educational uses of the Web in developing countries. The notion of the "information superhighway" uses a common but misleading metaphor of communication, which assumes that information is transported from source to receiver in much the same way as a truck transports goods along a highway. Information is not a parcel of knowledge that can simply be transferred from sender to receiver. Connectivity is not the only ingredient of a knowledge economy.
Instead, politicians should emphasize that the interpretation and use of information are highly contextualized activities and should be investing in education along with infrastructure. The knowledge work performed on the Web is always associated with a particular, highly specialized domain and requires specific forms of literacy or communicative practices. These practices can include knowledge of the domain and its discourse, academic conventions, written English, and Western visual and user-interface design.
Studies of Web-reading behavior are widely cited, but they do not question whether literacy and schooling may be significant variables influencing the reading patterns observed in usability studies [5, 10, 12, 14]. In developing countries, this information is crucial. We need to recognize that current usability research (and current design practices) may be based on observations of participants from relatively similar, privileged, Western backgrounds. Our goal in this study was to document the skills and practices associated with Web use for a group of South African novice users.
Skilled Web users seem to master a range of specialized reading skills for efficient reading from the Web [5, 10, 12, 14]. Such practices, and associated assumptions about reading and sources, have probably been built over a lifetime of reading and evaluating large quantities of both printed and online academic and professional material. Skimming and scanning, habits associated with reading from the Web, are strategies acquired by people who areto use a slightly simplistic catchphrase"information rich" or are habitually confronted with large amounts of written information to read in a short time. These complex skills have visual, verbal, linguistic, and cultural dimensions.
Similar blind spots can be identified in more theoretical studies. For example, the theory of "information scent" studies the minutiae of information-oriented Web use. Detailed information about individual users' activity on the Web is mined to understand how people adapt to the flow of information in their environment, in the active construction of knowledge .
This theory is useful, but it still assumes that information, and its scent, is objectively "out there." Giuseppe Mantovani  cautions us that information seekers need particular diagnostic and strategic resources and that community and culture play a role in the construction of information.
Our experiences with South African students have highlighted the sociocultural component of Web use. We have found evidence suggesting a cultural dimension to information scent and information-seeking literacy practices. In using information, students need to move from school-based cultures to academic cultures in order to achieve social mobility. In many South African schools, interactions are oral and in the vernacular; literacy activities are centered on a few scarce and highly valued English-language print resources. In contrast, in predominantly English language-based academic subcultures, a vast range of print and electronic resources must be accessed, skimmed, evaluated, and synthesized.
As students learn to use Web-based resources, they need to master a wide range of new visual conventions. The hierarchical tree from literate Western culture pervades both the interfaces of Web sites and, oftenthrough databases and hierarchical file structurestheir underlying organization. Our study revealed the difficulties experienced by certain novice Web users in interpreting tree diagrams and the Web's visual navigational conventions (such as fish-eye views, breadcrumbs, and other implicit trees) that rely on an understanding of such hierarchies . Fish-eye views are hierarchical navigational devices that provide an overview of the major categories of content on the site, giving extra detail about the specific category that the user is currently navigating. Breadcrumbs (named for Hansel and Gretel's trail of breadcrumbs) are used by sites with hierarchically classified directories to indicate the positioning and line of descent of a particular resource within the directory's implicit tree structure. Other implicit trees use various layout devices to encode an implicit hierarchy. The remainder of this paper outlines the study and some of our major findings.
The study provides two profiles of 20 novice Web users recruited from a class of students taking an introductory computer literacy course at the University of Cape Town. The first profile involved an initial round of individual interviews, Web searching observations, and visual literacy tests.
Each student completed a set of exercises on visual literacy to investigate their understanding of common interface and navigation conventions on the Web, such as tabs and hierarchical file directory structures. In addition, we investigated:
- Tree diagrams and linear structures
- Breadcrumbs (which trace a path through an implicit hierarchy)
- Implicit navigational hierarchies (hierarchy is implied by layout rather than an explicitly drawn tree)
As students learn to use Web-based resources, they need to master a wide range of new visual conventions.
Students' responses were analyzed both heuristically (by interpreting students' diagrams and comments) and quantitatively. In order to assess their skill at skimming and scanning print materials, students were given a recent newspaper and asked to read it as they normally would. After one minute the newspaper was taken from them, and they were asked to sketch what they remembered of the paper.
Finally, each student generated a list of topics, based on which we asked them to retrieve and interpret information from a Web search engine or directory. We observed students during their first experience of searching the Web, using a combined "think aloud" and tutoring protocol as they attempted to find answers to these self-generated research tasks.
A second profile involved another round of individual interviews and observations of a new set of self-generated Web research tasks. Observations were coded to record all student errors.
At the start of the year, students battled significantly with navigation and were hampered primarily by their limited knowledge of basic conventions of graphical user interfaces (GUI). This limitation formed the largest category of coded errors. For example, extreme novices would not click on hyperlinks, thinking that they were merely underlined for emphasis, and did not know that they should type words into search boxes. They did not realize that they could scroll to see the parts of a Web page that extended beyond the screen, and often did not understand that pages that seemed blank were simply downloading slowly.
At the end of the year, all students but one had made significant strides in their general competence, confidence, and skill in Web use. None of the characteristic extreme novice user interface errors was visible, and students were generally comfortable in the Web environment. As a whole, their social lives had moved at least partly online, and the majority reported daily use of e-mail, chat, and Web-to-cell phone text messaging (Short Message Service, or SMS). For all but two students, their activities focused on communication and browsing (using the categories defined by Sellen, Murphy, and Shaw ). Information gathering, the key activity of knowledge workers, was almost entirely absent. The four students who used Web-based research for assignments had used sources recommended by tutors, and one student who had independently found and used an academically inappropriate Web source was penalized.
Unfamiliar Discourses and Classifications
Studies of hypertext have shown that prior knowledge of a domain helps readers draw inferences and bridge gaps between noncoherent bodies of text [2, 6, 9]. In other words, users' prior knowledge of a domain helps them to pick up the information scent that is needed for them to make the link and to construct their own coherence. Consequently, unfamiliar, domain-specific discourses and classifications can present critical barriers when Web use is unmediated by supportive communities of practice.
The novice Web users of our study experienced serious and complex problems when using the Web for academic purposes but were more at their ease when using it for everyday reasons (when they were more familiar with the discourses and classificational systems). These problems go beyond the well-documented sense of disorientation experienced by many users when navigating a hypertext. The issues that need to be considered here include:
- Familiarity with Web-specific visual conventions
- Experience in speed-reading practices
- Understanding the concept, scope, and purpose of a Web search
- Ability to make sense of incoherent hypertextual artifacts such as search results
- Barriers presented by the English language and specialized discourses
- Alien classification schemas
- Limited knowledge of academic domains
Visual Information Hierarchies
Visual exercises suggested possible cultural dimensions of the interpretation of common visual navigational conventions on the Web. For the users we studied, the problem was not the relatively superficial one of not recognizing icons such as tabbed files, folders, mailboxes, or trashcans (as discussed by Marcus ). Students were able to use visual metaphors such as tabs, despite their having encountered them only in their computer literacy course. More difficult challenges involved interpreting the meaning of hierarchical information structures.
As suggested by Kress and van Leeuwen , hierarchical classificational tree diagrams (Figure 1) are a culture-specific visual form. People need to understand the associations and hierarchies that implicitly connect the terms in the structure. For example, the conventional Western family tree structures families according to generations, with a patrilineal line of descent. This tree (grandparents, parents, children) is a basic metaphor underlying many of the logical structures we use to structure information.
Several student diagrams suggested the cultural specificity of this form. For example, Figure 2 is an example of the unconventional structure that one of the students, Khanyisile (all names have been changed to protect the privacy of the participants), produced when asked to draw a family tree. Interpretation is difficult, as in this case, particularly when the diagrams diverge from the Western conventions. However, she seemed to visualize her family as two lineages, one maternal and one paternal. Other students' diagrams seem to be structured around lines of gender or familial power. This set of diagrams suggests the diverse ways in which human societies organize and model the world, and these are not always identical to conventional Western structures.
Screen shots of implicit tree structures from domains relatively familiar to students (such as sport and music) were chosen as examples to determine the students' understanding of the various structures. Linear conventions (such as hypertext "next") were understood by three-quarters of the group. In contrast, the visual conventions expressing tree structures implicitly through layout (such as fish-eye view navigational devices) were understood by only a third of the students, and only a tenth of the students were able to interpret breadcrumbs.
Print Reading Practices and Source Awareness
A quarter of the students were observed to read Web sites line by line, often moving the mouse pointer slowly and painstakingly along lines of text on the screen. Their sketches of the front page of a newspaper suggest that this might be their mode of reading print as well. For example, another student's sketch suggested a possible linear reading pattern, since it focused on the top "kickers" rather than main headlines or pictures.
The artificial nature of the exercise and the need to remember and draw the page mean that this information is by no means conclusive evidence. Nonetheless, in combination with our observations of "mouse reading," the information does suggest the possibility of reading patterns different from those of elite knowledge workers. More conclusive and detailed evidence for linear reading patterns, however, would require the use of eye-tracking equipment.
Interestingly, only four students were able to correctly recall the name of the newspaper. Several students who included a newspaper name included an incorrect name, indicating that they did not consider the name of the newspaper important enough to consciously register it. This contrasts strongly with the intense source-awareness that characterizes the knowledge workers discussed by Sellen, Murphy, and Shaw  and the participants in Nielsen's Web usability study . A similar lack of awareness of sources bedeviled most of the Web searches we observed. Inappropriate sources were selected, and many students used sources from the United States or Europe, thinking that they referred to South Africa. Source awareness is a key difference between academic and university literacy practices, and students making the transition to academic culture need to consider the source of information at the same time as its content and to be alert to the subtle signs by which we identify sources on the Web.
Scope and Concept of Web Searches
Understanding the scope and concept of Web searches is difficult for someone with no sense of a database and little sense of the vast number of resources indexed in search engines. Students came from environments where information is a scarce commodity, and consequently, the most common problem we documented was the use of inappropriately general querieseven after the formulation of more specific search questions. For example, a student formulated the need to search for "more about African jazz," but translated this into the query "music." When interpreting search results, students missed potentially relevant sources because their goal in the search was to find a brief "point-form" summary similar to that provided by teachers at school, rather than the academic goal of synthesizing information from a wide range of sources. As Yusuf Sayed  points out, many South African students struggle with research because they "do not realize that it is unlikely that there [is] a single source containing all the answers."
Additionally, students seemed to assume that the computer understood and remembered their intentions and previous actions. For example, using a South African Web directory called Max, a student named Tumi attempted to locate the South African constitution. She misinterpreted Max's interface, which presents its hierarchical collection of topics visually, as a partial tree diagram. In the Law section of the directory, she chose Cases, rather than Constitutional Law, explaining that she was looking for "case studies." When questioned, she explained her logic: She imagined that all subsections of the tree had been generated from her previous search on the death penalty.
Breadcrumbs and Alien Categorization Schemes
When anthropologists study other cultures, they pay careful attention to classification schemes. The students in our study were often overwhelmed by the classification schemes used on both South African and international sites, and categorization was the third most frequent cause of error for the students. Even general Web directories such as Yahoo! and Google inherit academic classificational schemes and mingle these with new categorizations structured around the habits, interests, and beliefs of moneyed, usually Western users. To complicate matters further, the implicit hierarchy in the omnipresent breadcrumbs structuring convention was not clear to any of the students.
Searching for current South African exchange rates in the Max directory, Musonda, a student, read through the categories carefully. Initially unsure whether to choose the category Business and Finance or Industry and Trade, he first selected Business and Finance and then, somewhat despairingly, resorted to the catchall subcategory "Other Finance." Abandoning the site hierarchy, he tried the search engine, which returned several matches, including Travel and Tourism > Currency.
He ignored this link and followed a red herring. Max assumes that a user would be interested in currency primarily because of international travel, an activity that does not register strongly on Musonda's cultural radar. When I pointed out the link to him, he thought a bit before making the connection: "When you are traveling, you worry about currency." Musonda's facility with English-Zulu translation here needed to incorporate a more difficult translationtranslating between his search goals and those of the Web's paradigmatic, wealthy, middle-class, Western user.
Some of the challenges we have identified can be addressed through training and through Web designs that build maximal coherence, through the use of microcontent and good metadata, and through building sites that do not depend exclusively on hierarchical, tree-based conventions. However, some of the more crucial challenges are not likely to be solved at the interface level. We may weave information scent carefully into our Web designs, but in the last analysis, the scent is in the "nose" of the user. In developing contexts, the user's goals and practices may be vastly different from our assumptions, and they may not be able to crack the many codes by which we have encoded the scent.
1. Card, S., Pirolli, P., Van Der Wege, M., Morrison, J.B., Reeder, R.B., Schraedley, P.K., and Boshart, J. "Information Scent as a Driver of Web Behavior Graphs: Results of a Protocol Analysis Method for Web Usability." In Proceedings of the Conference on Human Factors in Computing Systems (2002). pp. 498-505.
2. Foltz, P. Comprehension, coherence, and strategies in hypertext and linear text. In Rouet, J., Levonen, J., Dillon, A., and Spiro, R. (eds), Hypertext and Cognition, Lawrence Erlbaum, Mahwah, NJ, 1996.
9. McNamara, D., Kintsch, E., Songer, N., and Kintsch, W. Are good texts always better? Interactions of text coherence, background knowledge, and levels of understanding in learning from text. Cognition and Instruction, 14, 1 (1996), pp. 1-43.
15. Walton, M., Marsden, G., and Vukovic, V. "Visual literacy" as challenge to the internationalization of interfaces: A study of South African student Web users. CHI-SA 2001. Available at http://people.cs.uct.ac.za/~gaz/papers/marion.pdf
Marion Walton is a senior lecturer and Web developer in the Multimedia Education Group, University of Cape Town, where she develops educational software and teaches courses in writing for the Web and new media. Her research focuses on usability and literacy issues in relation to Web use in South Africa, and on how literary approaches can help in understanding and designing user experiences of computer games, virtual reality, and new media in general.
After receiving an Honors degree from UCT in English and Medieval Latin, Vera Vukovic worked for two years as a research and development assistant in the Multimedia Education Department at UCT. In 2003 she will be following a Masters program in New Media at the UCT Institute for Film and New Media, where she hopes to expand her research interests in the use of digital art and animation within educational technology.
The authors are highly indebted to Gary Marsden for his ongoing guidance and assistance in this project. The project was made possible by a grant from the South African National Research Foundation.
©2003 ACM 1072-5220/03/0300 $5.00
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, to republish, to post on servers or to redistribute to lists, requires prior specific permission and/or a fee.
The Digital Library is published by the Association for Computing Machinery. Copyright © 2003 ACM, Inc.