The earliest and least visual way of representing (and accessing) the location of files (=documents) in a computer was to show them at the ends of paths through a hierarchical tree branching from a root node (Unix) or particular partition letter (MS DOS). This tree with its files at the end of its paths was called a file directory, and a file's location was given as the path from the root through the various intervening named nodes (subdirectories) and terminating in the file's name. The sample screen at the left is of ZTree for Windows at work (an XTree clone); in XTree-type representations, the files were not represented as leaves but in the separate lower window. That is, these files are the contents of the C:\Program Files\Xing\AudioCatalyst\ subdirectory. Each node/subdirectory is said to contain the nodes and/or files that branch from it. XTree was breakthrough technology for the desktop computer in the later 1980s, since it made it possible to grasp the layout of an entire partition (or directory system), where before one relied on memory aided by much rehearsal viewing the contents of individual subdirectories.
The storage of paper documents in manila folders in drawers in multi-drawer cabinets was a natural metaphor for storing electronic documents and some version of this metaphor has been a part of most graphical user interfaces since the 1980s. The fit is less than perfect in one regard, however, which is that files several subdirectories down end up being contained in folders which are themselves contained in folders which are themselves contained .... This difference led one web design guru (Alan Cooper) to say, "Programmers are very comfortable with nested systems, where an instance of an object is stored in another instance of the same object. Humans, on the other hand, generally have a very difficult time with the idea. (...) In a file cabinet, you never see folders inside folders, Pendaflexes inside Pendaflexes or file drawers inside file drawers."
Figure 6.18 is a mapping of the same subdirectory ("collage")with a program called TreeMapper developed by Ben Schneiderman and his associates at the University of Maryland. TreeMapper represents files as areas and subdirectories as framed boxes packed with the files. Label, color, and size of the file-areas can display attributes such as name, type of file or date modified, and size of file. The map at the left uses color for type of file, since the label=name of the file does not include its extension, and size. Within each layer of structure, the components are packed for size (here). Further, more exact text and numerical information for a file is displayed in a pop-up window when the file is selected. This two-dimensional display enables us to use our pattern recognition capacities (after becoming accustomed to the display). Here for example it is very easy to see where images with different graphic extensions are located and the relative amounts of space they take up. At "mr99" we are looking at files four levels down in the subdirectory "collage". Even the tiniest file gives pop-up info when moused over, left-clicking on a subdirectory zooms into it and right-clicking zooms up a level. It is thus possible to represent the relative locations of a very large number of files on one screen, more than in a file-and-folder style table of contents, especially one like Windows Explorer that separates files and folders and only displays the files in one folder at a time. It should be noted, however, that Schneiderman apologizes for violating the best principles of design in using increase in the area of rectangles (two-dimensional figures) to signify greater file length (a one-dimensional scalar). Such a move represents one more de-spatializing of space, but one that is becoming standard. Martin Wattenberg's spectacularly successful Map of the Market (www.smartmoney.com/marketmap/), which is based on TreeMapper, uses area to signify size of companies (measured in terms of gross revenues or some such metric).
Why, oh why, however do they call Schneiderman's treemap, which so resembles an Iowa countryside seen from the air in the summer, a tree? It would appear for the same reason that graph theory uses the term tree for hierarchical structure, in which case we may say the term in this use has lost almost all of the imagery associated with the big plants that make up a woods. Let us call the node-and-edge abstraction a "geometrical tree." The term family tree is also used for diagrams that are very schematized, as for example in Dan Short's Indo-European Language Tree frequently cited on the Web (and there are numerous others). Occasionally one comes across something more analog and treelike which does have some attractions which the nodes and edges approach lacks.
Actually, this fine speciman which originally appeared in Thomas Gamkrelidze and V.V. Ivanov's lifted from the Armenia Highland site looks more like an anemone to me, or an alien life-form, than it does a tree. It is given standard tree colors, however, with brown trunk and yellow branches with green tips. Note that the sub-family names are not written at nodes but in the middle or alongside the branches, thus avoiding the imagery of nodes or boxes that languages just popped up with out warning, as it were. That is, the image of a language as a discrete terminus of an edge misses the sense of languages as constantly changing and of emergence of one language from another. Indeed this tree (or anemone) strongly conveys flow and gradual emergence and lifetimes of a language or language family.The tree image also supports the notion that the area of current growth is the ends of the branches and leaves and that the trunk is the area of earliest growth. It thus discreetly conveys chronological ordering (directedness) without scaling languages to a time line. Jacques Bertin suggests using edge length in genealogical trees to represent each individual's length of life, but languages, unlike individuals, do not have datable births and deaths (278). One rarely sees Bertin's suggestion implemented, however.
Before leaving the Indo-European language family, we can compare these maps to the starburst representation printed for many years on the back endpaper of the American Heritage Dictionary, an outline of which is reproduced at the left. This is basically a starburst diagram showing descent from a common center. It is also laid out west to east, so that descent is also diffusion over geographical space (it "loosely suggests geographical distribution," the legend says). Note too the varying font sizes and thickness of the edges with no legend to help us; thick + big signify "major" by general graphic rules, but it is not clear whether the same criterion (or criteria) for "majorness" applies in each case of big. (The close reader may have noticed that these diagrams differ in content as well as metaphor, and this would be even more evident if more of the first diagram were reproduced. The major difference concerns the first branching from the root: Short's tree uses the old bifurcation into two forks (centem and satem languages); Gamkredlidze and Ivanov argue for four main branches; American Heritage's map gives eight main branches and 5 minor ones.) Though uncertainties such as this remain, the American Heritage map is an amazing blending of three types of relations among languages: family and group relations, geographical distribution, and diffusion/differentiation over time from a central homeland.
File directory trees are always drawn with the root at the top of the page and is often referred to as the top of the directory. Clearly no chronological ordering is implied (top = earliest in time? chronological time generally runs top to bottom or left to right in European printing; geological time, however, almost always runs bottom to top.) The root is the top because it is the starting point for a search, and the vertical axis does not represent a quantitative or other continuum. Vertical order of files does signify if the directory has been sorted for name, date, type, or size, but even so, the order is only relative, not scalar. Even before the file-and-folder gui's, subdirectory names were written to the right and below the directory containing them (i.e., they step down to the right), so that the horizontal dimension does signify depth of embedding.
Some things lend themselves even less to tree diagramming than language families. Where there is merger and re-convergence, the defining formal criterion of a tree ("unique path between two points") collapses and we get occasional wonders of the topiary art like the one at the left. This somewhat humorous tree by Miguel Covarrubius ("The Tree of Modern Art -- Planted 60 years ago") appeared in Vanity Fair in 1933, a couple of years before Alfred Barr made his famous "Diagram of Stylistic Evolution from 1890 until 1935" a highly re-converging figure that uses the names of schools and movements of art rather than artists for its node labels. One might even question why Covarrubius chose a tree, especially a literal one, as the metaphor, but trees do express continuity and change, and gradual process, and indeed such an oak tree would take about 60 years to develop.
So powerful and compelling are tree images that they are often used to represent hypertext webs, formally quite different structures to which we shall shortly turn.
We have pushed the notion of "map" pretty far, but they are still at least flat, and the viewer's position toward them is stable and outside them. With 3D maps, we even give up those features. In recent years, many programs have been developed to represent file systems in three dimensions with the viewer doing fly-throughs of the virtual file worlds. Linux is particularly well supplied with these maps, as they are nice ways for a young programmer to learn graphics. File system space is generally black and without distinguishing marks. The names of files and directories appear in that space usually with some icon attached to the name. Some of these spaces have a floor or floor grid that provide an orienting surface and horizon (and support the illusion of 3D space), but others emulate the vast, featureless expanse of outer space. Often what is visible is only one level in the file hierarchy, that is, the files and subdirectories found there; in order to explore the contents of subdirectories, the viewer has to go to or pass through the subdirectory node, at which point the contents of the subdirectory become visible. The image at the left is a screen capture from a little prototype program called Portals: it uses the metaphor of a colonnade for the entry from a portal (the name of the subdirectory, represented as two column with a lintel overhead). As you emerge from the colonnade-passage, you see the various subdirectories of that subdirectory ranged in an arc before you, with the files that are in that subdirectory represented by their file names in red. Portals is being developed as part of a larger project "Threedsia" and at present doesn't do anything useful other than display one's file system as a colonnaded landscape to move through.
Yusuke Shinyama's program Xcruise adopts a deep space metaphor for a filing system. Subdirectories appear at first in the distance as labelled small ellipses ("galaxies") which grow larger if you approach them and envelop you if you enter them (i.e., enter the subdirectory). The files appear as spheres with diameters proportional to their file sizes; the color differences seem to be made for visibility and a splashy liveliness. While you are inside a galaxy/subdirectory, you can see nearby bodies in other galaxies: this adds an extra layer of complexity for the viewer, who may accidentally leave the galaxy she is in without realizing it, though her location is plotted in the upper left-hand corner of the screen. Also the lack of floor disorients and the compass bearings on vertical and horizontal axes do not help me very much. Xcruise also models one of the ways file directories are not like trees, namely their use of symbolic links to connect two files in any two arbitrary places in the structure. These links are represented as "worm holes"—long curving, sweeping green lines which bundle quite densely in a Unix file system around certain system and shared library directories.
Figure 6.25 of the subdirectory containing this chapter, which was produced by Leander Seige's tdfsb, takes us back to a grounded display, and indeed a sort of city layout with labelled file icons, proportional to size, displayed on ranks and files (the translucent yellow columns are HTML files). The display only gives one subdirectory at a time, with silver spheres at certain places to provide jumps ("warps") to directories above and below. It is thus more like a department store than a city (no spheres happen to appear in this snapshot). Directional lighting also gives orienting cues. Image files are displayed on slabs with some depth to them (unless they are transparent GIFs, in which case they float in position) and can be approached and viewed as if one were walking by them in a gallery (and they were all large canvases, or Jeff Wall's light boxes). In effect, then, a directory with images, especially one largely or wholly devoted to them, becomes a walk-through gallery which replicates the experience of seeing something from a distance (down a corridor, say) that looks interesting and moving and turning toward it to see it better, all with a few other pieces in view.
One of the commonly claimed benefits of displaying abstract structure is "insight." These immersive, 3D spaces challenge our models of insight, which is generally thought of as something more stable and enduring than may be apparent in the surface distraction and clutter of things, something that is more fully integrated into the larger scheme of things. This sort of black space that envelops the viewer provides a relatively new experience for viewers, especially the experience of maneuvering your view point around in it. Such maneuvering with mouse or arrow keys has become a specialty of the young, however, and those who have been raised on the Quake 3D graphic engines would get the movements fairly quickly, though the standard games always have key features in the various rooms of their mazes, not just black, or black with a floor grid. Those that have played with Flight Simulator or engaged in space dog fights should be better at maneuvering in featureless space, but for the rest of us, inhabiting these spaces is anything but automatic. In any case, 3D displays are unlikely to challenge standard file directory displays as ways of finding documents and launching programs.
Figure 6.26 is the view-screen of HotSauce, a browser plug-in developed by Apple and then abandoned in 1997 in the throes of financial crisis. It is displaying its approach to the web site Phonetic Resources. As it "enters" the cluster of links and pages, the first ones flow "past your head." HotSauce is actually a link mapper, and so here provides a transition to the next section. The colors represent the levels of linkage from the top page. It does allow you to open the pages the links point to, so it was intended as an alternative to the tables of contents often found in the left column of web pages. (Further snapshots of HotSauce in operation can be found in Dodge and Kitchin, pp. 134-5; Kahn and Kryzsztof, p. 114).
It is easy to imagine the corporate cost-cutter drawing a bead on HotSauce. What is its business concept? How much, for that matter, will people pay to be able to view the document names in their hard drives in 3D? Probably very little, which suggests that the making of devices to contemplate the orderly/disorderly worlds of hard drives is probably best left to Open Source programmers or to art, as in the next example.
Mary Flanagan's "[Phage]" is a Macromedia application that runs on Windows computers, reads their hard drives, and displays fragments of the texts, sounds and images that it finds there. These fragments appear a few at a time against a completely black field, spinning slowly and drawing closer to the viewer and then fading or spinning off, while at the same time, bits of sound files are played in the earphones (earphones recommended!) along with a regular ticking sound synchronized with the movements of the text and image swatches. The user has no control over the display once the program is launched. [Phage] is the anti-mapper to all mappers, since it presents the hard drive's contents as disconnected, unrelated fragments rather than parts of connected structures, their only location "the depths" out of which they rise and into which they return. Because of the fragmentation and rotation, one struggles to recognize the contents and recollect how, when, and why they came to be on one's hard drive.
[Phage] won a prize in the Digital 2000 ASCI competition; on www.maryflanagan.com it comes with some critical commentary (cited and linked) and an unusually detailed statement by the artist (who is also a Professor of Art at the University of Oregon) describing its meaning and purpose. Jon Ippolito, Jury member and now Media Curator of the Guggenheim, speaks of [Phage] as a psychoanalyst for the computer plumbing the depths of its unconscious, and Montfont (in Britannica.com) speaks of receiving a "rare glimpse into the turbulent unconscious of our computers—our prosthetic minds." In her own writing about [Phage], Flanagan offers a number of metaphors for a poetics of this experience, but she does not use the "unconscious mind of the machine" (which misidentifies the mind in question). Rather, she offers that of the "phage" (a virus that destroys and reconstitutes), "virtual palimpsest," and "feminist map of the machine." Here is one paragraph which combines some of these:
In this case, our environment contains our own artifacts, separated from our own ordering power. [phage] allows the user to experience his or her computer memory as a palimpsest of life experiences rather than experience the computer as simply a tool for daily use. By mapping a user's unique encounters—through images, downloads, web sites visited, emails—it creates spatial memory maps that not only reflect the user's interactions, but to a larger degree the user's redefinition of self within in technoculture.
"A map!" I can imagine someone saying, "How can anything so formless and constantly shifting be a map? This poetics of the machine stretches language to the limit!" And indeed it does, here to the point of equating mapping with "making a visual display of something." But we have been willing to tolerate that, as we have been talking about mapping abstract spaces. We have come close to this limit with Apartment, where a number of the mapping rules are somewhat occult (for example, those about adjoining rooms, and about selection of images in the 3D view) but we believe those rules are fixed, because the same words entered produce the same layout and 3D display each time. The extra challenge to our common-sense notion of mapping here is that the algorithm for finding, selecting, and displaying the fragments, as well as fragmenting them, seems to involve a considerable element of randomness: each time [Phage] is started on a machine, it produces different results (though with some common elements, of course, especially sound clips). Not only are our artifacts separated from our own ordering power, but we aren't sure what power is ordering them, or to what degree they are ordered.
Flanagan now offers [collection], an extension of [Phage] which links the data flowing from its scan of your hard drive to similar flows from other users located elsewhere. Thus you can see a composite flow which alternates the flows being merged from different computers. [collection] thus uses the internet "allegorically," she says, as a "collective memory space." So you can if you wish activate [collection] on your (Windows) machine and plug into a virtual anima mundi, or as one might say, Flanagan's Wake. In a recent twenty-minute session, I spliced data with MARY, amphale, megaserver, mimi, Christiane, brothercat, treestump, Keaton, fmiller4, and chris.
3D representations lead to immersion—moving of the view point into the file system (instead of opening more folders within folders as you "drill down"--presumably a data-mining metaphor). They do give a sense of experiencing the collection of files in a new way. Even with the radically altered orientation of the viewer, however, it is recognizably the same structure we usually experience via tree directories. There are landmarks, familiar subdirectories and file names, to the usually-somewhat-disorderly groupings we have made as we work on a project. With Flanagan's mapping we have a strong sense of experiencing objects on our own hard drive, but it is data stripped of its naming and ordering structures. It is a decomposed tree, or razed house, stirred by the winds and bursts of old, half-forgotten voices. If the darkness is taken to be some part of mind not illuminated by the light of full ordering consciousness, then the preconscious mind as Freud conceived it might be more apt. With [collection], on the other hand, one might think of the little fragments as IP-packets gone astray and the black screen as the external "cyberspace" through which they flow. The same visual effects are read quite differently, depending on which frame is invoked.