Sarah Sweeney: Way-Finding on the Web: Urban Planning and the Virtual Interface

This is available in printed form as part of the Conference Proceedings Book at Lulu.

Streaming audio for this panel is available here

When I open a new browser window I experience a simultaneous surge of excitement and dread. In the blank screen and blinking cursor there is the possibility of accessing billions of sites, but there is also the frustration that I do not know how to find most of them. I began this project thinking that I was alone in this frustration since there are so many critical and popular works extolling the virtues of our new personalized, more searchable, ever-changing web. However, as I kept looking, I found statistics and anecdotal evidence that suggested that there are many of us, users who are overwhelmed by the options and discouraged by our current tools for navigating among them. In this presentation I will speak as a web user and as an artist, nothing more. Kevin Lynchs seminal urban planning book, The Image of the City, will provide the

framework for much of my argument. It also serves as a guide for my tone, an approach he describes as speculative and perhaps a little irresponsible: at once tentative and presumptuous.1

In this talk I will take interface in a broad and malleable sense, as standing for the different technological elements that provide a connection between users and the websites they visit. I do not want to focus on the navigational elements that we use once we have made the connection to a website since this area has already been extensively covered within studies of usability. Instead, I would like to focus on the familiar yet commonly overlooked interfaces that we use to navigate our way to and between sites. These interfaces are difficult to define and identify since they can take many forms, including lists of links and form elements. They also have different levels of immersion within the web environmentsome exist as independent entities like search engine widgets or the address bar while others are embedded within websites and even within web content. While these interfaces differ in location and form, their common goal is the element that unifies them as a groupthey all work to help a web user find his or her way to a web site.

It is this same goal of finding ones way that Kevin Lynch studies in his 1960 book, The Image of the City. In this text Lynch uses this concept of way-finding as a means of assessing different systems of urban planning. Lynch suggests that [i]n the process of way-finding, the strategic link is the environmental image, the generalized mental picture of the exterior physical world that is held by an individual (IC 4). This image must have several qualities to have what he considers value for orientation in the living space. It must be sufficient, clear and well integrated, safe with the possibility of alternative actions, communicable, and adaptable to change (9).

Thinking about the web through Lynchs criteria initially suggests that the interfaces we most commonly use would not provide us with a strong environmental image. I would suggest that Lynchs first quality of sufficiency, which he describes as allowing the individual to operate within his environment to the extent desired (9), is perhaps the most essential quality in an operable environmental image and also perhaps the case in which web interfaces fall the most short. In the most obvious and basic case of this operation, finding ones way to a destination, these interfaces generally accomplish the goal in question without relying upon an environmental image. For example, the address bar, search engine interface, or a user-determined portal such as Alexa or Digg will very often take us directly to a variety of different destinations. The problem arises when we would like to return to this destination after some time has passed. Unless we bookmarked the destination or remember its address verbatim, without an environmental image to orient ourselves we must retrace the path we took exactly. However, since the path generated through these types of interfaces is generated at a specific moment in time it changes over time to reflect the changing relations within data on the web. The rankings of search engine results can change from one day to the next and what appears on the first page of Digg or as a top site on Alexa can change even more rapidly, making it difficult to retrace our steps in many cases. While a stable environmental image is not completely necessary for every type of navigation I would suggest that without it we lose a crucial element of control over our environment in some situations.

Just as time can change the way a single interface represents the relations between information, each interface also represents those relations differently based on its own organizational methodology. A search engine interface might rank and organize sites based on a page-ranking algorithm while a social interface like Alexa might organize sites based on a different criterion such as traffic. Thus the same site could appear through each interface in very different contexts. Because of the different ways in which these various interfaces mediate the relations of the web, any given site exists within a multiplicity of contexts, which works against the creation and integration of a clear, operable environmental image.

Without this clarity and integration, the communicability that Lynch also considers a crucial element of a successful environmental image becomes difficult. While it is relatively easy to communicate the address or name of a particular site, it is far more difficult to communicate the path necessary to reach a site without this type of specific information. In the absence of a clear and readable environmental image we cannot refer to common experiential or contextual clues that could help us to direct others when we no longer remember a precise destination. There are specific interfaces designed to help you share addresses without relying upon your memoryfor instance a personal bookmark manager or social bookmark managers such as del.icio.usbut they solve this problem without helping to construct an actual environmental image. I would suggest that while aiding our navigation, these interfaces force users into a more narrow and precise form of communication than might be possible with an environmental image.

Another quality Lynch considers important is the safety that comes with having multiple ways to reach a destination. He writes that an environmental image should be safe, with a surplus of clues so that alternative actions are possible and the risk of failure is not too high (9). He warns that [i]f a blinking light is the only sign for a critical turn, a power failure may cause disaster (9). Given the text-based nature of web navigation, so can a misspelled URL. This text-based nature creates a need for precision of language when we enter a search term or a web address. While Google has built in a corrective mechanism that asks you if you might mean to be searching for one thing rather than another, this is only a single option that is controlled by the technology rather than the user. Perhaps it is largely because of these limited options that, as one study suggests, only one in ten professionals always finds what he or she is looking for on the first attempt, while almost 70% of the same group admits to end[ing] up on sites they didnt expect to visit and are not relevant to their work.2

The last quality that Lynch cites as crucial to a functional environmental image is adaptability to change. Given the shifting nature of the web through the differences in time and interface I have discussed above, it would seem that our current web interfaces easily meet this criterion. However, I would suggest that this is not the kind of change Lynch is thinking of. While our current interfaces do indeed respond to the changing shape of the web, the changing elements of the image are often more numerous than the stable elements. For example, a search engine rearranges the results of the search with each new request. This constant rearrangement of our environment does not create an open-ended environmental image that is adaptable to change. Instead it makes the development and retention of an environmental image difficult if not impossible.

From this application of Lynchs theories two important conclusions start to emerge. First it becomes clear that the web interfaces that we commonly use make it difficult for a web user to produce a functional environmental image. For Lynch this would mean that while it is still possible to navigate this type of space, without an adequate environmental image we can only do so with what he describes as the cost of some effort and uncertainty and in even more extreme terms as strain, anxiety and even the terror that comes with complete disorientation (IC 5, 4). However, the second conclusion we reach is that these same interfaces accomplish many of Lynchs goals for an environmental image without actually producing one or causing the strain or anxiety Lynch predicts. It would seem at first glance that these two conclusions suggest that Lynchs mid-century theories of navigation within a physical space are inapplicable to the virtual space of the contemporary web.

However, I would suggest that the problem is not that Lynchs theories are not compatible with our experience of way-finding on the web, but rather that his theories do not take into account the way in which different processes of way-finding are influenced by different variables. Crucial among these variables is the degree to which we can articulate our destination. In 1960 the only possible process for way-finding was to navigate through ones environment from origin to destination regardless of the certainty with which one could identify that destination. Lynchs work could not anticipate a process of way-finding that allows for a direct connection between origin and destination made possible by the collapse of space within the virtual environment of the web at the end of the century. Even the fiction of instant connection to a destination popularized by Star Trek in the form of teleportation was six years away.

Thus while our ability to articulate a destination was not a central issue for Lynch in 1960, it is central today, as it determines how we choose between the different processes of way-finding. The degree to which we can articulate our preferences and its role in our decision-making process is discussed by Alexander Chernev in a 2003 study of product assortment. In this study Chernev finds that individuals with a salient ideal point face the relatively simple task of searching for the alternative that best matches their already articulated attribute preferences.3 I would suggest that we find the same processes at play on the web. In a situation where we can articulate the ideal destination, getting there is simply a process of finding a match. The existing interfaces accomplish this goal easily without reliance upon an environmental map. Moreover, to use an environmental map in this type of situation would only complicate the process and make it less efficient.

However Chernev also finds that individuals without an articulated ideal point face the more complex task of evaluating the available alternatives while at the same time forming the very criteria to be used in the evaluation process (PA 159). Similarly, on the web the task of finding a destination that is not highly articulated is far more difficult than finding an articulated one. In such a situation both our alternatives and the criteria derived from them are generated entirely by our interfaces. In this scenario we lose the agency we had in the case of the matching scenario. Here our destination is predetermined by the alternatives suggested by our interface using its unique organizational methodology. I would suggest that in this situation where our destination is not highly articulated a navigational interface that builds a stronger environmental image might allow us to regain some of this agency by allowing us to define our own alternatives and thus the criteria by which we determine this destination.

In addition to providing criteria by which to judge the strength of the environmental image, Lynch also outlines certain elements that the organizer of an environment, such as an urban planner or interface designer, could emphasize to produce a stronger and more legible environmental image, such as paths, nodes, edges, districts, and landmarks. I would now like to look at some of the web interfaces, real, imagined, and proposed, that employ these elements.

One of these elements is the district, which Lynch describes as a space that is recognizable as having some common, identifying character (IC 47). While there is no exact online equivalent, a broad application of this concept could include interfaces such as blogs in which links are arranged around an identity, usually an identity constructed by the producer of the site. The lists of sites included within such an interface form a sort of virtual district made distinct and recognizable by the identity that brought them together. The legibility and accessibility of this identity makes our choice between districts more deliberate and informed, thus giving us greater control over our environment.

Another element that adds legibility is the landmark. Lynch notes that the use of the landmark involves the singling out of one element from a host of possibilities (48). Two interfaces that differentiate between sites in this manner are Alexa and Digg. Both of these interfaces use the web community to identify sites that take on a different scale in the landscape of the web. While the temporal nature of this landscape makes it difficult to use for orientation over a span of time, at any given moment they allow us to create a clear environmental image that is common and shared.

While both web districts and landmarks provide contextual information that makes it possible to see distinct sets of data or pieces of the environmental image, without a spatial interface it is difficult to integrate them and see the connections between them. A spatialized environment would provide an opportunity for the paths, edges and nodes that create the continuity between pieces, making them more than a series of isolated entities. There is currently no interface that completely fulfills this possibility. However there are several promising interfaces that begin to suggest the potential of such a perspective.

One such space is the fictional world of the Matrix described by William Gibson in Neuromancer. This world is spatially organized around a graphic representation of data abstracted from the banks of every computer in the human system. It appears as lines of light ranged in the nonspace of the mind, clusters and constellations of data.4 Another such space is Second Life, described on its homepage as a digital world imagined, created, and owned by its residents.5 This three-dimensionally rendered world is structured around a geographical space that has been purchased and developed by the residents. Google Earth is another example of a spatialized interface, although it is organized around a much more familiar map of the actual physical world.

To this group I would also add an experimental interface named ColorColony that I have been developing with my colleague Danielle Laplante. Although we began work on ColorColony before I had read Lynchs work and thus it is not a direct response to his theories, it is a response to the same problems and issues Lynch documents in the least imageable urban environment in his study, Jersey City. Navigating the web we have often felt like one of the Jersey City residents Lynch quotes: Its much the same all overits more or less just commonness to me. I mean, when I go up and down the streets, its more or less the same thingNewark Avenue, Jackson Avenue, Bergen Avenue. I mean sometimes you cant decide which avenue you want to go on, because theyre more or less just the same; theres nothing to differentiate them (IC 31). ColorColony is an attempt to address this disorganized sameness through collections of websites organized around different identities or characters represented by color. Each collection or colony is spatially rendered as a grid within which there are clusters of similar sites that collect around hubs.

While there are different organizing factors at play in each of these interfaces information networks, imagined real estate, physical geography, or abstract colorsthey all employ spatial elements that work to create a continuous and communicable common ground. It is this common ground that I see as the crucial potential introduced by interfaces that allow us to envision a strong environmental image. This common ground allows us to define our own alternatives when searching for a destination but also allows us the stability to imagine new and unfamiliar destinations.

In addition to the navigational agency it provides, this common ground has important cultural and historical implications as well. It allows us to form shared memories and histories attached to groupings and spaces that are larger than individual sites. It also has a political and social dimension. Without it, our virtual world is self-defined, producing isolated, narrow experiences in which we only see what we want to see. Experiences which are unpleasant or foreign are rendered invisible or absent by the very fact that we do not look for them.

To suggest that the environmental image should be the primary or only means for structuring an interface would fall prey to this same type of narrow vision. However, I think that if we continue to work in the direction that these interfaces suggest to produce interfaces that can supplement our existing ones, I can envision a resulting environment that has the same potential as Lynchs ideal city, a space that not only offers security but also heightens the potential depth and intensity of human experience (IC 5).


Kevin Lynch, The Image of the City (Cambridge: MIT Press, 1960), 3. Subsequently cited in the text as IC.

Convera News and Events: Press Releases: Consumer Search Engines Leave Professionals at a Loss, says Convera Survey. 2006. 8 April 2007 <>.

Alexander Chernev, Product Assortment and Individual Decision Processes, Journal of Personality and Social Psychology 85.1 (2003), 159. Subsequently cited in the text as PA.

William Gibson, Neuromancer (New York: Ace Books, 1984), 51.

Second Life: Your World. Your Imagination. 2007. 9 April 2007. <>.