Metaphors We Compute By

John M. Lawler

A Lecture delivered to staff
of the Informational Technology Division
of the University of Michigan

(this is the original version of the essay appearing in
Figures of Thought: For College Writers
by Dona J. Hickey; Mayfield Publishing, 1999)

Introduction
  • What this lecture is about

        I'm going to talk today about metaphors. Everybody here has heard and used the word metaphor plenty in the course of your educational experience (and the amount of educational experience in this room is pretty staggering, so that makes lots of uses of the word). To quote a famous sage, "It's a common word, something you use everyday." It generally gets stored in memory with all the other stuff you learn in literature classes, like simile, plot, characters, rhyme, meter, and so on. And then it gets forgotten, or at least not looked at often, until and unless you do something literary. I'm here today to suggest that in fact there is a human phenomenon (which I will call metaphor, though what name you give it doesn't really matter much) that is much more important to everybody than all this would imply.

        In particular, it's a very important thing for anyone professionally associated with computing to bear in mind. That makes this a sort of practical lecture. I think it's also an interesting topic in its own right, so I hope it will also be an intellectually stimulating lecture as well. We'll see, I suppose.

  • Why I'm giving this lecture

        I'm a linguist, which is to say I study language -- not so much to learn to speak lots of languages, but rather to try to figure out just how language works in the mysterious process of human communication. One of my special areas of linguistic research is Semantics, the study of meaning, and within that area I tend to specialize in Metaphor. That's not really saying much, because what I've discovered (and I'm far from the first to discover it) is that metaphor is such a pervasive phenomenon that specializing in it is not a very restrictive act. Metaphor is involved in practically everything.

        I'm also a computer fan. I started computing 25 years ago, when I was in graduate school, and I've been thinking about computers for most of the intervening years, even when I didn't have one of my own. Nowadays, when even someone on a professor's salary can afford to have a computer that's bigger and better than the monster that took up the basement of the Engineering School building at the University of Washington in 1967, I do something more than think about computing. But I haven't stopped thinking.

        When I got my first computer, I plunged in and learned as much as I could about it. I found that some of the things I'd figured out myself in the decade or so I was away from the computer world were right -- and that quite a few of those things were ridiculously wrong. I also noted that many of the things I was learning seemed familiar somehow, even though I had really never heard of them before. This puzzled me; enough to make me pay attention to how I was learning about computing. What I found was that I was using my linguistic knowledge to structure and understand (and store and access) the computing information I was learning. I was treating this as just another language to learn. That made it easy, since I already knew how to do that. In fact, computing was orders of magnitude easier to learn and understand than a natural language. This in turn made me wonder how I was doing this -- I mean, you're not really supposed to be able to pretend black is white and get away with it, let alone make it pay.

        What I was doing was using a metaphor. It turns out that that particular metaphor (COMPUTING IS LANGUAGE) is by no means the only useful one around, though it is a bit specialized in this culture, since American society is rather naïve about language. Since then, I've found lots more, and I've also found that people can often understand things better if they use a metaphor and know that they're using it. In fact, the more metaphors you use, the better. I'm giving this lecture, therefore, to encourage you all to:

    1. think about the metaphors you use in talking about computers, and to
    2. consider broadening your repertoire.

  • What I'm assuming about my audience

        I'm assuming that you are English speakers (native English speakers for the most part) and that you're interested in computers (this is not a very risky assumption). I'm assuming that you're professionally concerned with the problems of communicating about computers -- you're involved, after all, in educational activities that center around computing. I'm assuming you've all experienced (probably first in yourself and then in others) the particular species of confusion, irritation, and despair that we often call "computer anxiety". And I'm assuming that you've also experienced the difficulties that attend attempts to educate people about computing.

        I mention these assumptions because I'm going to cheat a bit in this talk. I'm not going to cite ergonomic or psychological studies to prove my contentions about how people think about computing. Instead, I'm going to appeal to your experience (or what I fondly believe is your experience); both your own personal history of computing and your experience in educating others in computing. This, you see, is how metaphors work -- we all believe, despite overwhelming evidence to the contrary, that we share some ideas and experiences, and it is by means of reference to those concepts that we manage to communicate. So I'm going to both talk about metaphors and use them today. I have no choice in the matter. Nobody does.


    Some Questions About Metaphors

    1. What's A Metaphor?

          You have to distinguish (or at least I find that I have to) between several different meanings of the word metaphor. To begin with, there is the human cognitive capacity for metaphor itself, which is what the lecture is all about. This phenomenon is real, but it's hard to get at -- very abstract -- and can't really be investigated by itself. To find out how it works, you have to look at more concrete phenomena. The most concrete is another use of the word metaphor, in which a word (say, spend) which is defined with respect to one kind of thing (in this case, money) is used in context with a completely different kind of thing (say, time, as in I spent two hours on that report.) I call this an instantiation of a metaphor. Note that it's linguistic in nature, and has to do with what words we use.

          The abstract phenomenon of metaphor and the instantiation of a metaphor are the abstract and concrete poles of the uses of the word metaphor. In between them is one other sense of the word, one that's very important because it represents the cognitive mapping that we use when we use metaphors, and because it controls or licenses the actual instantiations. I call this inbetween level the Metaphor Theme, and it is this that is frequently meant when people talk about metaphor. In the case of the instantiation above, the metaphor theme is TIME IS MONEY. This just means that the cognitive/semantic area (frame) of TIME is treated semantically as if it were the same as (there are actually differences, but never mind that now) the area or frame of MONEY. This is like a semantic equation (and there are differences there, too).

          Using a metaphor theme means that we can use words that are defined with respect to one frame in talking about concepts and words defined with respect to another. In effect, we enlarge the target frame. Thus (to continue with the example), we can talk about saving, losing, budgeting, wasting, and just plain having time, even though time is not really (literally) the kind of thing for which these predicates are defined.

      To summarize, the three levels are:
      1. Metaphor as a human cognitive phenomenon.
      2. Individual Metaphor Themes (e.g, TIME IS MONEY).
      3. Instantiations of metaphor themes (e.g, He spent an hour on that)

          What we are interested in here, then, is what metaphor themes exist that start off COMPUTERS ARE ... (or COMPUTING IS, etc.). We need to solve, in other words, a metaphoric equation; and, at that, a Diophantine one -- one that has a number of possible solutions. As is usually the case with such equations, not all of the solutions are equally useful.


    2. Why Are Metaphors So Important?

          It's not difficult to see that much of what we say about almost any topic is in fact metaphoric. That is, once we get attuned to metaphors -- get used to noticing when the language being used in some discourse is not literally defined with respect to the topic under discussion -- we begin to find them everywhere. This is one of those cognitive discoveries that can burst on someone as a satori-like experience, and in fact most experiential philosophies like Buddhism or Taoism have some fairly pungent things to say about metaphor, and its peculiar problems. I'll content myself here with repeating something from Chuang Tzu ...

      "Snares are used to catch game; once you have
      the rabbit, you can dispense with the snare.

      "Fish-traps are used to catch fish; once you have
      the fish, you don't need the fish-trap anymore.

      "Words are used to catch ideas; once you have
      the idea, you can throw the word away.

      "Oh, how I wish I knew someone who had thrown all
      his words away, so I could talk to him about ideas!"

      ... which sums up very neatly the problems of dealing with metaphor while using language. Words and phrases have to be defined in terms of some semantic frame, and the structure of that frame pretty much delimits what we can say and how we can view any topic. That makes it very, very difficult to talk about metaphor, and equally important to do so. In particular, choosing the right metaphor can make a terrific difference in how (or even whether) our attempts at communication are perceived.


    3. How Do You Find Out How They Work?

          There are many ways, but the one I use is linguistic in nature, since I'm a linguist. As they say in Linguistics, pick a language at random, say, English. And pick a topic, say, computing. Collect samples of discourse in that language about that topic. Then look at the words and phrases used, and seek out their literal senses (i.e, discover the frames where they are defined). Then posit metaphor themes that license such uses of the words and phrases. Then test these themes to see if you're right. We'll be seeing a number of themes today. I'm not going to try to categorize them; inventing categories for metaphors is a fun game, but ultimately frustrating, since they're always just more metaphors. We'll just stroll about and see what we find.


    Metaphors About Computers

  • Novelty and the role of metaphor

        It's a truism that new things are hard to talk about -- our experience moves much faster than our language does -- and few things are newer than computers. Just 60 years ago, there were NO computers in the world. Anywhere. There was talk about them in recondite academic technical circles, but the number of people who had even heard such talk (let alone understood it) was extremely small. I don't have to tell you about the results of the ensuing half-century; they're all around us. Naturally, we have had to cope, and the major way we have done so is by using metaphors.

        These metaphors have typically been invented spontaneously by people who understood some aspect of computing in order to communicate with others, to be able to tell them about something they didn't already understand. This might be a case of one hacker telling another one about a neat new algorithm, or it might be a case of somebody writing a users' manual. You can think of lots of other cases. The difference between the two cases I mention here is in the presuppositions that are assumed by the speaker (or writer) to be part of what the listener (or reader) believes. I say both speaker and writer because metaphors are by no means a matter exclusively of the written language; however, since writings are more permanent than speakings and therefore tend to dominate the evidence, we sometimes lose sight of the relative proportions of language use -- millions of words spoken for each one written -- and I want to emphasize here how important it is for you to understand that I'm talking about all language use here.

        Technical language is full of metaphors, but they are pretty impenetrable to outsiders. Think about what's meant by (say) signing on to a computer system. In real life, one signs on to a ship's company, or to a project, or to some other group; the metaphor is one of joining an enterprise and identifying with it by pledging with your signature, but if you weren't familiar with the details, you wouldn't have a prayer of understanding what's meant. Metaphors in this case are a part of the cultural context and serve as much to mark in-group status as to serve a more idealistic communicational function. And as with all such group-marking phenomena, there are dialects: in some computing environments you don't sign on, you log on. Or occasionally in. But that's sociolinguistics rather than semantics.

        My concern here, however, is not so much in these fascinating twists and turns of jargon as in the metaphors used with intent to communicate to people who can't be expected to know the details. Yet. This is a really serious problem, since it's not clear just what they can be expected to know, and any writing that is done without having in mind a clearly defined audience with clearly defined background knowledge is practically impossible to bring off successfully. As you all know.


  • Old myths and new

        One thing we can take for granted about our audience when we write or teach is that they are members of our culture. This isn't, of course, true any more about audiences in some media, such as the Net, but these are rapidly evolving their own cultures so they can play the same games. As such, they are parties to a number of communal jokes we play on one another for various purposes. One word for such jokes is Myth. A myth is a species of metaphor that is:

    1. Widely, even universally, known and used in a culture or subculture;
    2. Largely unconscious in nature [possibly because of (1)];
    3. Literally false, or even ludicrous, when spelled out.

    Myths can concern anything at all, and I don't really want to go into the subject too deeply here; as a colleague of mine once suggested, the grammar of mythology is a bloody business. Let's get back to computers.

        Computers are the subject of plenty of myths. They are new and therefore scary. Scary things need explanations; when we have an explanation, a label, we can put the scariness into a box and feel in control of it. This is a silly way to behave, of course, but it's pretty human. If computers hadn't been so damned useful, this wouldn't be a problem; on the other hand, we wouldn't be here discussing it, either. Since they have had such an effect on everyone's life, we need to take a look at the metaphors about them that have taken on the status of myth and to see what effects they've had.

    These myths fall into a number of categories:

    Deus Ex Machina

        The basic idea here is that computers, being powerful, mysterious, and omnipresent (and therefore very threatening), take on some or even all of the classical aspects of gods or demons. It's sort of a contrary of the well-known explanation for why dinosaurs are so popular -- they're said to be "Big, Dangerous, and Extinct", and thus a safe subject to fantasize about. Computers are big and dangerous, but very far from extinct, so many people feel threatened by them.

        This one had a lot going for it during the first part of the last half-century, since early computers were very large and remote, understood by only a few, controllable only by secret rituals, and ministered to by specially-trained (and -gowned) people who were already admitted to the mysteries, and governed the admission of others.

        For those in on the secrets, of course, this is a rather convenient view to encourage in others, and one often encounters it in large Data Processing enterprises. Likewise, for those on the outside, this myth makes computing look like a very dehumanizing activity, and their resistance to computing can take on religious overtones. Needless to say, all this makes life much more difficult for those whose job it is to de-mystify computers, since they can wind up coming across either like heretics or like soulless minions of the Devil. In short, this myth is not one we should encourage, not that I think anyone does.


    Mathematical Machines

        A related issue is the fact that American culture has never been fond of mathematics; 'Rithmetic is the last and least of the 3 R's, and the overwhelming majority of our compatriots have neither interest in nor understanding of anything that even looks mathematical. It is therefore a very simple matter to predict what the social response will be to any innovation that is billed as being primarily mathematical in nature.

        To add to this, there is a fairly common distrust of The Machine in our culture. Some machines we have to accept, just to get along -- the automobile, the telephone; some, like TV, seem to be insidiously dangerous; some, like nuclear weapons, are dangerous in ways that are far worse than insidious. Despite (or perhaps because of) our dependence on machines, American culture is resentful of them and often views them as dehumanizing.

        Put these two mythic viewpoints together, stir in some of the feelings of Deism mentioned above, and you get a real sense of what computer anxiety is all about. Since computers really are machines, after all, we don't have much choice about this. Of course, there are machines and there are machines.


    The Pathetic Fallacy

        Many of you will be familiar with this phrase as the classical name for what we now call anthropomorphic language, that is, the attribution of human qualities to non-human things. All languages have a vast repertoire of terms that refer only to humans and their activities, traits, feelings, appearance, intentions, etc. As a species, we are very narcissistic; human terms probably constitute at least half of the vocabulary of every language and are by any standards our favorite topics for discussion and writing. It's not surprising that we see them everywhere, or that we attribute them to something like a computer.

        It's the social role of a human that the computer (more correctly, the software on the computer) is expected to take on. There are lots of varieties of this myth because there are so many roles for humans: servants, confidantes, secretaries, bosses, friends, enemies, therapists, etc.

        Nevertheless, it's pretty obvious that expecting a computer to act or react like a human would is asking for trouble. Not that Eliza or Bob ever shied away from trouble.


  • Some examples

        In this section I want to give you a short tour of some of the uses of metaphor in computing. Each of the following is one example (where there are probably hundreds) of views of computing that are current in American culture. Each of the views (i.e, metaphor themes) licenses particular kinds of language to talk about computing and has particular consequences in the culture. Each has its problems, each its opportunities.

    The Servant Problem
    [THE COMPUTER IS A SERVANT]

        Few Americans have ever had servants; otherwise, this would be a very commonly-remarked phenomenon. As it happens, having servants is a mixed blessing. To begin with, servants belong to a different social class from their employers, and in most areas of the world, that means they speak a different language, or at least a dialect that can verge on mutual unintelligibility. Nor are they usually well-educated, nor do they always subscribe to the same cultural goals and standards as their employers. As a result of all this, one must often spend at least as much time and effort supervising a servant doing a task as one would spend doing it oneself. Many times it's much simpler to do it yourself.

        Now put this into a computer metaphor. Everybody would love a program that was a good servant; but like even the best of servants, such a program must be instructed on what to do and how to do it. And if you're still not satisfied with how the program or the servant does it, too bad. There is only so much any servant (or any program) can be expected to know or to learn. And now we come to the kicker -- computer programs are much more limited than humans, and can typically only do one kind of thing. To do another, you need a different program (or specialized servant), and that one can't speak the same language as your other one(s), and can't learn anything they already know. Therefore you must painstakingly learn yet another language, and yet another set of personal (programmatic) idiosyncracies in order to make it work for you. I won't even mention the effects produced by the American phobia for languages.

        This variety of anthropomorphic metaphor theme (THE COMPUTER IS A SERVANT) is reinforced by (among other things) software that uses first person pronouns, by interfaces that make the user learn a recondite and unchangeable set of terms for what the software can do, and by overly cute documentation that personalizes the name of the program.


    Running Fast
    [THE COMPUTER IS A RACE]

        Computers appear animated; that is, things seem to move about and responses to user stimuli can be noticed. Of the classic criteria of animateness (growth, ingestion, excretion, and irritability), good analogs exist for computing. In particular, the speed of computer response is an especially gratifying and very important phenomenon. I think it is no mistake that we use the transitive verb run to refer to the execution of computer programs.

        I imagine many of you here are familiar with the original title of the magazine Dr. Dobbs' Journal; it used to be "Dr. Dobbs' Journal of Computer Calisthenics and Orthodontia", with the epigraph "Running Light Without Overbyte" for those who weren't in on the joke. The English verb run is intransitive; that is, it doesn't use a direct object. It can be made into a transitive verb, though, and then it is causative, i.e, it means cause to run. One of its most common uses (celebrated in Dr. Dobbs' athletic reference) is in referring to racing animals, or in idioms like run me ragged, where what is connoted is the causing of very rapid performance.

        In this case, it is control of the (rapid, animated) behavior of the computer that we're talking about. To use run is to be a speed freak like those who constantly try to make everything go faster. Take a moment to reflect on how short a second is in normal human activities, but how l o n g it is when you're waiting for a computer response. As computer users, we've become addicted to instantaneity.

        But, to quote one of the maxims from Kernighan and Plauger's The Elements of Programming Style, it's important to "Make it right before you make it faster".


    Software Tools
    [THE COMPUTER IS A TOOL]

        Another Kernighan and Plauger book, Software Tools, is said to have started a revolution in software design. Whether true or not, it certainly was a clear case of a metaphor being used consciously. A MACHINE is a form of TOOL, and that's an extension of our manipulative ability -- functionally, something you use in your hand. Tools extend our ability to apply energy. There are, as we've discovered in the last couple of millenia, two kinds of energy involved in tools.

        One is the obvious physical sense of power, exemplified in the harder blows of a hammer as compared with a fist; the other is more subtle, and really refers to information. With a handsaw, for instance, a user supplies both types of energy simultaneously; with a power saw, on the other hand, the user supplies only the controlling energy, the information, while the motor provides the power; and with an automated drilling machine, both types are separately powered.

        You can probably see where I'm going from here. It's only a small step from a machine where one kind of energy is powered as information to one where the metaphor is turned back on itself and two kinds of information are powered separately. This leads directly to the distinction between algorithms and data, the abstraction of which is the core of computer science.

        For a technician, this is probably the most useful and productive metaphor available for computing. For anyone else, however, there are problems. Awareness of technical details is a plus for a technician; it's pretty often a hindrance to those interested only in using the machine for their own purposes. Something else has to be provided, yet, paradoxically, the technician is almost certainly the wrong person to decide what it should be.


    Car and Driver
    [THE COMPUTER IS A MACHINE]

        As I mentioned, allied to the TOOL metaphor is the MACHINE metaphor. I've mentioned the latent Luddite tendencies of many of our fellows; it's also true, though, that quite a few of us are very fond of machines. The best (and most locally relevant) example of this is, of course, the automobile. Our culture has not just accepted automobiles -- we've embraced them wholeheartedly. So it's tempting to use the metaphor of driving a car to refer to using a computer. There are some benefits to this; however, there are even more problems.

        The automobile is a machine that you don't need to understand in order to use. While there are plenty of people who enjoy tinkering with cars, many more want nothing to do with such activity. They want to use the car. This they can do, because functionally a car is simply an extension of a natural activity (movement) whose operation can be transferred from other learned activities. In short, you learn a high-level skill associated with the hardware -- you learn to drive a car. What you do with it is then open to your own intentions.

        Computers don't really work like this. Learning to run a computer doesn't help you at all in using it; in fact, it's not clear just what learning to run a computer might mean. Computers, unlike cars, have software, and the software is what you wind up using. The appropriate analogy to having a car would be if you had a programmable car. You would have a program that took you to the grocery store, another that took you to work, a third one -- that you customized yourself -- that took you to Grandmother's house, and so on. To go anywhere else (or to take a different route) you'd have to get a different program or change some installation constant. And it goes without saying that each program would behave differently, use different controls, display different messages, and so on. If cars really were the same kind of machines as computers, we'd never use them.


    The Desktop
    [THE COMPUTER IS A WORKPLACE]

        We're all familiar with the Macintosh Desktop and its origins in the Xerox Star and its copies in Windows, etc. And we've all had lots of discussions about how great an advance it is in user interface design (whether we believe that or not, there are enough folks who do to involve us in such discussions almost endlessly). You have to admire that kind of enthusiasm, and the products that evoke it. Nevertheless, we've not yet arrived at the perfect user interface.

        I'm not going to get involved in a debate on this here; I just want to point out a few things that I think aren't quite perfectly realized yet and see what you think. To begin with, the designers of the Mac were quite correct that many people seem to think visually and would welcome a visually-oriented interface. If the graphics quality isn't quite high enough to distinguish some of those icons, well, it'll get there, right? But there are others who don't think quite so visually, and who find the specificity of icons somewhat confining. The single fact that is true of all computer users (as of all human beings) is that they're all different.

        Even (perhaps especially) when dealing with people who are highly visually-oriented, there is the problem of information accessibility. I, for instance, happen to be a person who likes to have information resources visible; so I clutter my desk with things I refer to often, leave books I'm reading in places where I'll see them, post notices to myself, and so on. On any visual interface, you don't really get to see the information; you get to see the labels, and you must remember what's what. My Mac or Windows desktop doesn't really look like my real desktop, and it isn't nearly as useful. I doubt I'm the only person for whom this is true.

        What's really at issue here is what I call density. Not the physical concept of the same name, but a metaphorical one (what else?) dealing with access to information. I happen to like information to feel dense, like there's a lot of stuff in there (wherever there may be in the metaphor, not to mention stuff), like I can just reach in and grab it; which is why I like to see a lot of things at the same time, and have them interconnected if at all possible, whether or not they're logically or conventionally linked together. To a certain extent, this reflects how I think my mind works. Others prefer their information less dense, with fewer high-level nodes and less clutter overall, which probably reflects how they think their minds work. And for still others, this doesn't seem to be an issue at all. But there certainly is an enormous variability in how people deal with information density and access, and with how they externalize this in their interactions with computers.

        I don't really see a great deal that can be done about it, in fact, beyond making user interfaces as customizable and flexible as possible, and using a lot of synonyms when designing them. The point I want to make here is that diversity in personal styles of information management is not yet a well-known or -handled part of user interface design. There's always a big problem with adaptation; either you have to adapt yourself to the design of the computer (and you may not be able to do so usefully), or you have to adapt the computer to your own strategies (and this is a very difficult task at best). Mostly we try to do both, with quite variable degrees of success.


    File Systems
    [THE COMPUTER IS A FILING CABINET]

        The name file that is used for the most commonly used artifact of software is another thing that people have to get used to. The metaphor here is that of a business office, with a filing cabinet full of folders, each containing some kind of information, each with some kind of label.

        This is very misleading, though it's too late to do much about it, since the term is too firmly entrenched in technical jargon. The problem is that real files all hold the same kind of thing (legible papers), while computer files can hold anything at all, much of which isn't legible at all, at least by humans. The idea of putting executable code (for example) into a file is obvious enough, once you know something about computing; however, it's anything but obvious at first.

        Even after you get over this hassle, though, you have to learn (usually the hard way) about file formats, and about the hassles of trying to get information from one kind to another. The filing cabinet metaphor gets stretched too thin to be of use here; in fact, when I started looking for examples, I came on this one with a shock of recognition -- it's been ages since I thought about computer files as having anything to do with filing cabinets.


    Fun and Games
    [THE COMPUTER IS A TOY]

        The phenomenon of computer games is an unlooked-for one, and in my opinion one of the most sanguine examples of the serendipity we all expect to find in the information revolution. The first really successful personal computer company was Atari; of course, the Pong boxes weren't as neat as the Apple ][, but they paved the way for it, and significantly influenced its design. Not to mention the design of its successors. What would the Mac be like (would there even be a Mac, or an Apple?) if there had been no computer games? Looked at in this light, you can see the development of SpaceWar (the first visual computer game) at the MIT AI Lab as, in fact, a monumental advance in computer science.

        The reason I'm so optimistic about games and their ilk is that (as we all know) games are fun. Now we also know that computers are fun, but this is for some people a difficult proposition to swallow; even these folks, however, know that games are different. By definition, they're fun. And fun is precisely what we all need; by which subversive remark I mean that the protean promise of computing will never be kept unless the kind of enthusiasm and creativity we're willing to put into fun activities like games is routinely harnessed.

        We have a sort of problem here. The market forces driving the development of hardware and software are oriented to the world-view of business and conservative institutions, where Things Must Be Taken Seriously. This leads to speed, which we all welcome, of course, and sometimes efficiency, which has its place, but only rarely to fun, and therefore only rarely to real creativity.

        For instance, few of us are interested in yet another wordprocessor. On the other hand, I've often wondered what it would be like to do writing -- or, more interestingly, to learn writing -- on a word processor that had sound effects -- real bells and whistles. Pop! when you delete a word, Zzip! when you delete a line, and so on -- I leave the remainder of the design as an exercise for your imagination. The point is not that it might be more efficient, but that it might be more fun. And it might be a good idea to encourage fun in computing, just as it is to encourage it in education. Serious Business is good business, all right, but for repeat customers, fun sells better.

        One of the most hopeful signs I've seen in the computer world is the sense of humor that's evident everywhere. April 1st is the most important holiday on the calendar of the computer culture; I hope we keep it that way for a long time.


    Conclusion

        I've mentioned a number of metaphor themes that we use to approach computers as things and computing as activity. Some of these have Mythic status -- that is, they developed on their own, in the "cultural unconscious", and we have to deal with them, willy-nilly. Others are more or less conscious choices, made for particular reasons in particular contexts. There are still others that we haven't mentioned.

        I want to switch here from talking so much about the language used about computing to return to my own favorite metaphor theme, which could be roughly stated as COMPUTING IS A LINGUISTIC ACTIVITY. This is the other side of the coin, so to speak. There are plenty of things about computers for which at least some of the metaphor themes I've mentioned are not only appropriate, but productive; however, for one of the most important, I have some hopes for this one.

        The area of computing that (I think) everyone agrees needs the most work is user interface design. Progress in this area has led to such advances as the Mac interface; but there's more to the story. What it's all about is communication, and, while communication is not simply a matter of language, nevertheless, human language is the principal phenomenon of human communication. There are things that are known about it. These have been used in such areas of research as Natural Language Processing and Artificial Intelligence, but I want to turn it around and see what insights can be gained from looking at human interaction with computers as a linguistic process.

        To begin with, it is natural enough to view it this way, since keyboards are derived from typewriters, and those are used to produce written language. A naïve computer user will automatically use prior experience in typing and attempt to apply it to the task at hand. Early user interfaces that were oriented to a command line, in fact, explicitly attempted to use this, by making the command itself into a structure of Imperative Verb plus Direct Object, something derived directly from English grammar. This was an instantiation of the SERVANT metaphor theme, and the imperative is the form used to give orders.

        I could go into a lot of linguistic detail about the grammar of such commands; other forms that can appear are directly analogous to cases -- for instance, the DOS/Unix redirection arrows function precisely like the ablative and dative cases in Indo-European languages. More important, however, is the fact that, unlike most servants, most computer programs don't have much of any facility for what we call repair procedures -- i.e, what happens when we misspeak or see that we are misunderstood. It's here that the most important limitation of the principal myth about language in our culture comes out.

        This myth is called the CONDUIT METAPHOR, and it is particularly easy to see in distinguishing spoken from written language. In spoken language, we appear to be understanding a person through what they say; in written language, on the other hand, we appear to be dealing with the words themselves, and the literal meaning (the word literal itself simply means 'written') becomes a matter of very great importance. If we make a mistake in conversation, we can back up, restate, ask questions, pause, look dumb, or behave in a lot of different ways that can lead to clarification, rather like an elaborate error-trapping routine.

        In written language, however, we have much less to go on, and have consequently developed conventions for interpreting the writer's intentions. The Conduit Metaphor, which is a myth that is used to explain how we can communicate, even though we're not telepathic, supports the views:
  • that words and meaning are physical objects of the same type
  • that (literal) meanings are attached to words
    (the metaphoric attachment is that the meaning is inside the words -- note the use of such phrases as in a few words, empty words, full of meaning, and the like)
  • that communication is a matter of shipping the word strings over to the listener and having them unpack them.
  •     This is a pretty silly theory, particularly when applied to natural spoken language, but it has a certain utility in its application to writing. Written words are physical, after all, and written communication is a matter of exchanging strings of words. Since two of the three parts of the Conduit Metaphor seem to work well, it's a simple matter to assume the third -- that meaning is attached tightly to words (in fact, inside them), and that therefore any difficulties in understanding are due to the writer's improper use of words.

        Returning to computers, we see that the natural equation of computer interaction with written communication, coupled with the equally natural acceptance of the Conduit Metaphor to explain the functioning of written language, has led quite naturally to a situation in which computer software is doomed to behavioral flaws in its interaction with humans, since it is operating on the flawed assumption that meaning (which would really be better termed intention) is a literal matter. This isn't something that we can do a great deal about, of course, given the difficulties inherent in determining intentions and of making workable programs in the first place.

        Some solutions may be in the offing, and thinking about them is interesting. For instance, while speech recognition is very difficult and isn't going to become widespread soon, it offers some possibilities for getting away from the written language metaphor. It is startling to think that within a few decades, typing may be as rare a skill as knapping a flint arrowhead, driving a coach-and-four, shooting a flintlock musket, or solving an equation on a slide rule. We may be living in what might come to be called the Written Input Era; like the Vacuum Tube Era, one of the fascinating sidelights of technological history. Before we all learned how language really works.

        Of course, speech recognition doesn't provide the whole story; there is plenty of hard work to do in determining just exactly what intentions people can have in using computers. In our contact with new users, we've all come upon some strange ones. What I want to suggest today is that these conceptions aren't really as strange as we might consider them. They're normal ideas, based on normal expectations, which happen not to be met by currently normal hardware and software. We have little choice right now except to try to adapt the users to the machine; but the future holds some hope of being able to adapt the machines to the users.

        Provided, that is, we don't become wedded to our myths and embedded in our current metaphors. Some people (me among them) are often accused of Mixing Metapors. This is supposed to be a bad thing. I'll admit it can be a bit confusing, but I really think it's our only hope. The more different views you have of something -- and the more different the views are -- the more hope you have of understanding what the thing is really like. Of perceiving some aspect of its reality that isn't apparent in any of the individual views.

        The best metaphor I know of to explain this is the phenomenon of binocular vision, or stereo sound. We have two eyes and two ears, even though each one of them works fine alone. The other one isn't just a spare, though, because using them in parallel provides information about what is being perceived that isn't carried in either of the separate images. We perceive depth in visual or aural signals precisely to the degree we use separate, different signals and succeed in integrating them into a single percept. Nobody really understands what computers are, let alone what they can be; my suggestion is that we leave it that way, and continue coping with imperfect metaphors. The more the better. And of course, continue to have fun.


    Original lecture delivered 1987.     HTML version 1/27/99 John Lawler
    Copyright © 1999 The Eclectic Company     More Metaphor Links     The Chomskybot
    Links to classes that use this text at:   George Mason University     The Assignment
                                        Richmond University
                                        The University of Michigan