1Metaphors We Surf the Web By1Paul P. Maglio Teenie MatlockIBM Almaden Research Center Cognitive Psychology Program650 Harry Rd., NWED-B2 University of California, Santa CruzSan Jose, CA 95120 Santa Cruz, CA [email protected] [email protected] 1 With apologies to Lakoff and Johnson (1980).AbstractThe way people think about the World-Wide Web(WWW) has implications for the way that theynavigate it. In this paper, we discuss the nature ofpeople’s metaphorical conception of the WWW,as gathered from interviews with beginning andexperienced web users. Based on linguistic data,we argue that people naturally think of the web asa kind of physical space in which they move,although information on the web is not physical,and web users do not actually move.Nevertheless, such metaphorical thought ismotivated by the same basic image schemata thatpeople rely on to mentally structure everyday life.IntroductionHow do people naturally conceive of informationspaces, such as the World-Wide Web? It seemsobvious that the way people think about the webwould have implications for the way they use andnavigate the web, and thus for the design of toolsto facilitate information navigation. In this paper,we consider metaphorical conceptions of theWWW, as gathered primarily from language-usedata. Briefly, we argue that both experienced andinexperienced web users naturally talk about theweb in definite and consistent ways. For instance,people see themselves as moving towardinformation, rather than information as movingtoward them. Nevertheless, we found somedifferences between experienced andinexperienced web users in the consistency withwhich they talked about web activities. In theend, we argue that the particular language peopleuse is metaphorical and is motivated by basicimage schemata, which emerge from embodiedexperience (e.g., Johnson, 1987; Lakoff, 1987).Before presenting our data and argument indetail, we first discuss some prior researchconcerning the way people think about the web,along with some background on metaphor andthought.How People Remember the WebBased on data collected from people asked torecall specific WWW searches, Maglio andBarrett (1997a) argued that web navigation isconceived in terms of a cognitive map similar to acognitive map of physical space, that is, in termsof landmarks and routes (e.g., Anderson, 1980).In this study, experienced users searched the webfor answers to specific questions. To identify keycognitive aspects of their activities, users werefirst asked about their plans, and then theirbehavior was tracked while they searched. Then aday later they were asked to recall the steps theyhad taken in each of their searches the previousday, and finally to retrace their steps. Participantswere not warned on the first day that recall wouldbe required on the second day. This methodenabled Maglio and Barrett both to chart behaviorto uncover search tactics (using the behavioraltraces) and to extract some of the structure of theirinternal representations (using the recall data).The data showed that participants recalledonly a few of the sites they visited. Specifically,they remembered key nodes that led to the targetinformation. These nodes were called anchorpoints by analogy to the notion of anchor points inthe cognitive map literature (Couclelis, Golledge,Gale & Tobler, 1987). An anchor was defined asa node along a search path from which there is anunbroken sequence of links on successive pagesthat lead to the goal node (i.e., no URLs need tobe typed in or explicitly recalled). Oncetraversed, anchor points are recognized as lyingalong the path to the goal—even if the same pathis not followed to the goal in every case. For theparticipants in the study, searching on the second2day often meant finding anchors encountered onthe first day, rather than finding paths found onthe first day.A second observation that emerged from thebehavioral data is that individuals relied onpersonal routines when trying to find information.For instance, some participants routinely used aparticular search engine, such as AltaVista,whereas others routinely used a particularhierarchical catalog, such as Yahoo! It is notmerely that these searchers preferred to use oneapproach over another, but that theyconceptualized their search tasks in terms of theirfavorite routines. It often did not matter what wasactually done on the first day, the searchersremembered searching as if their personal routineshad been followed. On the analogy to cognitivemaps of physical space, personal routinescorrespond to the familiar routes that anindividual uses to get from one landmark (oranchor point) to another.If people mentally structure web use in thisway, tools for web navigation ought to present theweb in this way. Because individuals tend to usethe same search patterns over and over, andbecause they recall their searches in terms of theirstandard patterns—almost regardless of what theyactually did—Maglio and Barrett (1997b) built apersonal web agent to identify repeated searchpatterns and to suggest similar patterns for newsearches. Because people focus on key nodes oranchor points when recalling their searches, andbecause these structure memory for the searches,Maglio and Barrett (1997b) built a web agent toidentify the key nodes in finding a piece ofinformation, and to maintain personal trails interms of these.How People Talk About the WebThe key to designing information navigation toolslies in discovering how people naturally conceiveof information spaces. Technically, the WWW ispart of a network of geographically distributedmachines connected via wires. The informationaccessible by users of this physical network isorganized in a conceptual network of hyperlinksamong documents. Despite this actual structure,people’s conceptual structure of the WWW israther different.Matlock and Maglio (1996) found that webusers often refer to the WWW as amultidimensional (most commonly two-dimensional) landscape. Obtaining information inthis landscape is expressed as traversinginterconnected paths toward locations that containinformation objects, such as user homepages andcommercial catalog sites. Users say things suchas, “I went to his homepage,” and “I came back towhere I saw that picture.” Some of theseinformation objects are talked about as two-dimensional and others, as three-dimensional; forinstance, people say “in Yahoo!” which suggests athree-dimensional