Download Daniel C. Dennett - Consciousness Explained and more Study notes Communication in PDF only on Docsity!
HOW WORDS^ DO^ THINGS
WITH US
Language, like^ consciousness,^ only^ arises^ from^ the^ need,^ the necessity, of intercourse^ with^ others. KARL MARX,^1846 Consciousness generally has only^ been^ developed^ under^ the pressure of^ the^ necessity^ for^ communication. FRIEDRICH NIETZSCHE,^1882 8efore my teacher came^ to^ me,^ did^ not^ know^ that^ I^ am.^ I^ lived in a world that was a^ no-world.^ I^ cannot^ hope^ to^ describe adequately that unconscious,^ yet^ conscious^ time^ of nothingness.. Since^ I^ had^ no^ power^ of^ thought,^ I^ did^ not compare one mental^ state^ with^ another. HELEN KELLER,^1908
- REVIEW: E PLURIBUS^ UNUM? In chapter^5 we^ exposed^ the^ persistently^ seductive^ bad^ idea^ of^ the Cartesian Theater,^ where^ a^ sound-and-light^ show^ is^ presented^ to^ a^ sol- itary but powerful^ audience,^ the^ Ego^ or^ Central^ Executive.^ Even^ though we've seen for ourselves^ the^ incoherence^ of^ this^ idea,^ and^ identified an alternative, the^ Multiple^ Drafts^ model,^ the^ Cartesian^ Theater^ will continue to^ haunt^ us^ until^ we^ have^ anchored^ our^ alternative^ firmly^ to the bedrock of empirical^ science.^ That^ task^ was^ begun^ in^ chapter^ 6, and in chapter^7 we^ made^ further^ progress.^ We^ returned,^ literally,^ to
first principles:^ the^ principles^ of^ evolution^ that^ guided^ a^ speculative
narration of the gradual^ process^ of^ design^ development^ that^ has^ created our kind of consciousness.^ This^ let^ us^ glimpse^ the^ machinery^ of^ con- sciousness from^ inside^ the^ black^ box^ —^ from^ backstage,^ one^ might^ say, 227
228 AN EMPIRICAL (^) THEORY OF MIND in homage to the (^) tempting theatrical (^) image we are trying (^) to overthrow. In our brains (^) there is a cobbled-together (^) collection (^) of specialist brain circuits, which, (^) thanks to a faniily of (^) habits inculcated partly (^) by culture (^) and partly by individual (^) self-exploration, (^) conspire together to produce a (^) more or less orderly, more (^) r less effective, (^) more or less well- designed virtual (^) machine, the Joyce:in (^) machine. By yoking (^) these in- dependently evolved (^) specialist organs together (^) in common (^) cause, and thereby (^) giving their (^) union vastly erhanced (^) powers, this virtual (^) ma- chine, this (^) software of the brain, (^) performs a sort of (^) internal political miracle: It (^) creates a virtual captain (^) of the crew, without (^) elevating any one of them to (^) long-term dictatorial (^) power. Who's in charge? (^) First one coalition and then (^) another, shifting in ways (^) that are not chaotic (^) thanks to (^) good meta-habits (^) that tend to entrain (^) coherent, purposeful sequences rather (^) than an interminable (^) helter-skelter power (^) grab. The resulting (^) executive wisdoni (^) is just one of the (^) powers tradi- tionally assigned to the (^) Self, but it is an (^) important one. William (^) James paid (^) tribute to it when he (^) lampooned the idea (^) of the Pontifical (^) Neuron somewhere (^) in the brain. We (^) know that the job (^) description for (^) such a Boss (^) subsystem in the brain is (^) incoherent, but we (^) also know that (^) those control (^) responsibilities and (^) decisions have to be (^) parcelled out somehow in the brain. We (^) are not like drifting (^) 5hips with brawling (^) crews: we do quite well not just (^) staying clear of (^) shoals and other (^) dangers, but plan- ning (^) campaigns, correcting (^) tactical errors, (^) recognizing subtle (^) harbingers of opportunity, (^) and controlling (^) huge projects that (^) unfold over (^) months
or years. In the next few chapters we will look more closely at the
architecture of (^) this virtual machine, in (^) order to provide (^) some support — not proof^ — for (^) the hypothesis that ii (^) could indeed perform (^) these ex- ecutive (^) functions and (^) others. Before we do (^) that, however, we (^) must expose (^) and neutralize (^) another source of (^) mystification: the (^) illusion of the Central Meaner. One of (^) the chief tasks of the (^) imaginary Boss is (^) controlling com- munication with (^) the outside world. As we (^) saw in chapter (^) 4, ideal- ization (^) that makes (^) possible (^) assumes that (^) there is (^) someone (^) home doing the talking, (^) an Author of Record, (^) a Meaner of all the meanings. (^) When we go to interpret (^) a loquacious (^) vocal sounds, we (^) don't suppose (^) they are just random (^) yawps, or words (^) drawn out of a hat by (^) a gaggle of (^) behind-the-scenes partygoers, (^) but the acts (^) of a (^) single agent, the (^) (one and only) (^) person whose body is (^) making the sounds. (^) If we choose to (^) interpret at alt, we (^) have no choice (^) but to posit a person (^) whose communicative (^) acts we are (^) interpreting. (^) This is not quite (^) equivalent to positing an (^) inner system (^) that is the Boss of (^) the body,
230 AN EMPIRICAL THEORY (^) OF THE MIND on (^) the disk! Answer Y (^) or N. It would be a very (^) naïve user who thought the computer (^) actually meant to be (^) sc solicitous. Let me put (^) some words in the mouth (^) of a critic. Since (^) this par- ticular imaginary critic (^) will dog our discussions (^) and investigations in later (^) chapters, I will give him (^) a name. Otto speaks:
It was a cheap trick to call Shakey "he" rather than "it"; the trouble
with (^) Shakey is that it has no (^) real insides like ours; (^) there is nothing it is like to be (^) it. Even if the machinery (^) that took input (^) from its TV camera "eyeS' (^) and turned that input (^) into box-identification had been strongly (^) analogous to the machinery (^) in our visual sys- tems (and it wasn't), (^) and even if the machinery (^) that controlled its production of strings of English (^) words had been strongly (^) analo- gous (^) to the machinery in our (^) speech systems that (^) controls the
production of strings of English words (and it wasnt), there would
still have (^) been something missing: the (^) Middleman in each of (^) us whose (^) judgments get expressed when we (^) tell how it is with us. The problem with (^) Shakey is that its input and output (^) are attached to (^) each other in the wrong (^) way — a way that eliminates (^) the ob- server (^) (experiencer, enjoyer) that (^) has to lie somewhere between the visual input (^) and the verbal 3utput, (^) so that there is someone in there to mean (^) Shakey's worth; when they (^) are "spoken." When I speak, [Otto (^) goes on I mean what I say. (^) My conscious life is (^) private, but I can (^) choose to divulge certain aspects (^) of it to you. I (^) can decide to tell you (^) various things about my (^) current or past (^) experience. When I do this, I (^) formulate sentences (^) that I care- fully tailor (^) to the material I (^) wish to report on. I can go (^) back and forth between (^) the experience and the (^) candidate report, checking the words (^) against the experience to (^) make sure I have found (^) ies mots justes. (^) Does this wine have a (^) hint of grapefruit in its flavor, or does it seem to me (^) more reminiscent of berries? (^) Would it be more apt (^) to say the higher tone sounded (^) louder, or is it (^) really just that it seems clearer or (^) better focused? I (^) attend to my particular conscious (^) experience and (^) arrive at a judgment about which (^) words would do the (^) most justice to its (^) character. When I am (^) satisfied that I have framed an (^) accurate report. I express (^) it. From my in- trospective (^) report, you (^) can come to know about (^) some feature of my conscious (^) experience. As heterophenomenologists, (^) we need to divide (^) this text into two parts. We (^) put to one side the claims (^) about how the experience (^) of speak- ing seems to (^) Otto. These are inviolable; that (^) is how the experience
HOW WORDS DO THINGS^ WITH^ US^231 seems to^ Otto,^ and^ we^ must^ take^ that^ as^ a^ datum^ demanding^ an^ expla- nation. To^ the^ other^ side^ we^ put^ the^ theoretical^ claims^ (are^ they^ the conclusions of tacit arguments?) that^ Otto^ makes^ about^ what^ this^ shows about what is going on in^ him^ —^ and^ how^ it^ differs^ from^ what^ was going on in Shakey, for^ instance.^ These^ have^ no^ special^ standing,^ but we will treat them with^ the^ respect^ due^ all^ thoughtful^ claims. It is all very^ well^ for^ me^ to^ insist^ that^ the^ Middleman,^ the^ Internal Observer in^ the^ Cartesian^ Theater,^ must^ be^ eliminated,^ not^ found,^ but we can't^ just^ throw^ him^ away.^ If^ there^ isn't^ a^ Central^ Meaner,^ where does the^ meaning^ come^ from7^ We^ must^ replace^ him^ with^ a^ plausible account of how a meant^ utterance^ —^ a^ real^ report,^ without^ any^ scare- quotes — could get^ composed^ without^ needing^ the^ imprimatur^ of^ a solitary Central^ Meaner.^ That^ is^ the^ main^ task^ of^ this^ chapter.
- BUREAUCRACY^ VERSUS^ PANDEMONIUM One of the skeletons^ in^ the^ closet^ of^ contemporary^ linguistics^ is that it has lavished attention^ on^ hearing^ but^ largely^ ignored^ speaking, which one might^ say^ was^ roughly^ half^ of^ language,^ and^ the^ most^ im- portant half^ at that.^ Although^ there^ are^ many^ detailed^ theories^ and models of^ language^ perception,^ and^ of^ the^ comprehension^ of^ heard utterances (the paths^ from^ phonology,^ through^ syntax,^ to^ semantics^ and pragmatics), no^ one^ —^ not^ Noam^ Chomsky,^ and^ not^ any^ of^ his^ rivals or followers^ —^ has^ had^ anything^ very^ substantial^ (right^ or^ wrong)^ to say about systems^ of^ language^ production.^ It^ is^ as^ if^ all^ theories^ of^ art
were theories^ of^ art^ appreciation^ with^ never^ a^ word^ about^ the^ artists
who created it^ —^ as^ if^ all^ art^ consisted^ of^ objets^ trouvés^ appreciated^ by dealers and^ collectors. It is not hard to see why^ this^ is^ so.^ Utterances^ are^ readily^ found objects with which^ to^ begin^ a^ process.^ It^ is^ really^ quite^ clear^ what^ the raw material or^ input^ to^ the^ perception^ and^ comprehension^ systems^ is: wave forms^ of^ certain^ sorts^ in^ the^ air,^ or^ strings^ of^ marks^ on^ various plane surfaces. And although^ there^ is^ considerable^ fog^ obscuring^ the controversies about just^ what^ the^ end^ product^ of^ the^ comprehension process is,^ at^ least^ this^ deep^ disagreement^ comes^ at^ the^ end^ of^ the process being studied, not^ the^ beginning.^ A^ race^ with^ a^ clear^ starting line can at least be^ rationally^ begun,^ even^ if^ no^ one^ is^ quite^ sure^ where
it is going to^ end.^ Is^ the^ "output"^ or^ "product"^ of^ speech^ comprehension
a decoding^ or^ translation^ of^ the^ input^ into^ a^ new^ representation^ —^ a
sentence of Mentalese,^ perhaps,^ or^ a^ picture-in-the-head^ —^ or^ is^ it^ a^ set of deep structures,^ or^ some^ still^ unimagined^ entity7^ Linguists^ can^ de-
HOW WORDS DO THINGS WITH^ US^233 be involved in^ producing^ an^ error^ like^ "naming^ a^ wear^ tag'^ for^ "wearing a name tag.' Thanks to^ ingenious^ experiments^ that^ provoke^ such^ errors,^ and intricate analyses^ of^ what^ does^ and^ doesn't^ happen^ when^ people^ speak, progress is being^ made^ on^ models^ of^ the^ highly^ organized^ mechanisms that execute^ the^ ultimate^ articulation^ of^ a^ message^ once^ it^ has^ been decided that a particular message^ is^ to^ be^ released^ to^ the^ outside^ world. But who or what puts this^ machinery^ in^ motion?^ A^ speech^ error^ is^ an error in virtue of being^ other^ than^ what^ the^ speaker^ meant^ to^ say.^ What taskmaster sets the task relative to^ which^ errors^ such^ as^ the^ examples above are judged? What, if not the^ Central^ Meaner?^ Levelt^ provides^ us^ with^ a^ picture,
a "blueprint for^ the^ speaker":
In the upper left-hand corner^ a^ functionary^ who^ looks^ suspiciously^ like the Central Meaner^ makes^ his^ appearance^ in^ the^ guise^ of^ the^ Concep- tualizer, armed with lots of world^ knowledge,^ plans,^ and^ communi- cative intentions, and capable of^ "message^ generation."^ Levelt^ warns his readers that the^ Conceptualizer^ "is^ a^ reification^ in^ need^ of^ further
explanation" (p. 9), but he posits^ it^ anyway,^ since^ he^ really^ can't^ get
the process going, it seems,^ without^ some^ such^ unanalyzed^ Boss^ to^ give the marching orders^ to^ the^ rest^ of^ the^ team. Figure 8.
(^234) AN EMPIRICAL THEORY OF THE MIND How does it work? The underlying (^) problem will be clearer if we begin (^) with a caricature. The (^) Conceptualizer decides to perform a speech act, such (^) as insulting his interlocutor by (^) commenting adversely on the size of his (^) feet. So he sends a command to the (^) bureaucracy under his sway, the Public (^) Relations Department (Levelt's (^) Formulator): "Tell this
bozo his feet are too big!' The PR people take on the job. They find the
appropriate words: the second-person singular (^) possessive pronoun, your; a good (^) word for feet, such as feet; the right (^) plural form of the verb (^) to be, namely are; and the (^) appropriate adverb and adjective: (^) too big. (^) These they cunningly combine, with (^) the right insulting tone (^) of voice, (^) and execute: "Your feet are too big!" But wait a minute. Isn't that too easy? (^) When the Conceptualizer gave the (^) command (what Levelt calls the preverbai (^) message), if he gave it in English, as my caricature (^) just suggested, he's done all (^) the hard work, leaving little for the rest of (^) the team to do, except to pass it along with (^) trivial adjustments. Then is the preverbal message (^) in some other representational system or language? Whatever it is, it must be (^) capable
of providing the basic "specs" to the production team for the object
they are to compose and (^) release, and it must be couched in terms they
can "understand" — not English but some version of Brainish or Men-
talese. (^) It will have to be in a sort of language (^) of thought, Levelt argues, but (^) perhaps in a language of thought that is used (^) only to order speech acts, not (^) for all cognitive activities. The team (^) receives the preverbal message, a detailed (^) Mentalese order to make an English (^) utterance, and then it fills this order. (^) This gives the subordinates a little more to (^) do, but just obscures the looming regress. (^) How does the Conceptualizer figure out which words (^) of Mentalese to use to give the order? (^) There had better not be a smaller duplicate (^) of Levelt's whole blueprint (^) hidden in the Conceptualizer's message (^) generation box (and so on, ad infini- tum). (^) And certainly nobody told the (^) Conceptualizer what to say; he's the Central Meaner, (^) after all, where meaning originates. How then does the meaning (^) of an utterance develop? (^) Consider the following nesting of commands, (^) leading from grand overall strategy through (^) detailed tactics to basic operations: (1) (^) Go on the offensive! (2) Do something (^) nasty but not too dangerous (^) to him! (3) Insult him! (4) Cast aspersions on some aspect of his (^) body! (5) Tell him his feet are too big!
236 AN EMPIRICAL THEORY OF THE MIND operation composed of a series of primitive (^) instructions can (^) first trans- late the binary number into decimal notation (e.g., binary 00000110 = decimal 6) or into a letter of the alphabet via ASCII code (e.g., binary
01100001 = "a" and 01000001 = "A") and then output the result.
These subroutines are at (^) the heart of (^) the fancier output instructions found in higher-level programming languages, like Fortran or Pascal or Lisp. (^) These permit the programmer ':0 create further subroutines for building larger messages, fetching (^) long series of (^) numbers from memory and running them through the accumulator, translating them and writing the results on the screen or printer. For instance, a subroutine can make several trips to the accumulator for to plug into the blanks in You have overdrawn your account, Mr. by (^) $. Have a nice day, Mr.
— a "canned" sentence formula that itself is kept stored as a series of
binary numbers in the memory until subroutine determines that it is time to open the can. In this manner, a strict hierarchy of fixed routines can turn (^) sequences of (^) specific contents in (^) the accumulator into expressions that a human being can read on a screen or printer: "Do you want to save this document?" or "6 files copied" or "Hello, Billy, do you want to play tic-tac-toe?' There are two features of this process that are shared by Levelt's model: (1) the process takes an already determinate content as its input, and (2) the bureaucracy (^) — the "flow of control" in (^) computer-science jargon — has to have been carefully designed: all decision-making" flows hierarchically by a delegation of responsibility to subagents whose job descriptions dictate which bit of means/ends analysis they are authorized to perform. Interestingly enough, the first of these fea-
tures — the determinate content — seems to be endorsed by Ottos view
of his own processes: There is a determinate "thought" somewhere in
the Center, waiting to be "put into words." The second shared feature,
however, (^) seems alien: The hierarchy of routines that slavishly render that very thought in natural language have been predesigned by (^) some- one else — by the programmer, in the case of the von Neumann ma- chine, and (^) presumably by the combination of evolution and individual development in the case of the activities in Levelt's Formulator. The creative, (^) judgmental role that the thinker of the thought should play in getting the thought into words does not appear in the (^) model; it is either usurped by the Conceptualizer. who does all the creative work before sending an order to the Formulator, or it is implicit in the design of the Formulator, (^) a fait accompli of some earlier design process.
HOW WORDS DO THINGS WITH US 237 How else could ends and means be organized? Let's consider an opposing caricature: a pandemonium of word-demons. Here is how we talk; First we go into vocal noise-making mode — we turn on^ the horn: Beeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeep.... We do this for no good reason, but just because no good reason not to
do it occurs to us. The internal "noise" excites various demons in us
who begin trying to modulate the horn in all sorts of random ways^ by interfering with its stream. The result^ is^ gibberish,^ but^ at^ least^ it's^ En- glish gibberish (in English speakers): Yabba-dabba-doo-fiddledy-dee-tiddly-pom-fi-fi-fo-fum.... But before any of this embarrassing stuff actually hits the outside^ world, further demons, sensitive to patterns^ in^ the chaos,^ start^ shaping^ it^ up into words, phrases, clichés... And so, how about that?, baseball,^ don't^ you know,^ in^ point^ of fact, strawberries, happenstance,^ okay?^ That's^ the^ ticket.^ Well, then... which incites demons to make further serendipitous discoveries, aug-
mented by opportunistic shaping, yielding longer bits of more accept-
able verbiage, until finally a whole sentence emerges:
I'm going to knock your teeth^ down^ your^ throat! Fortunately, however, this gets set aside, unspoken, since at the same time (in parallel) other candidates have been brewing and are now in the offing, including a few obvious losers, such as You big meany! and Read any good books lately? and a winner^ by default,^ which^ gets^ spoken: Your feet are too big! The muse has failed our speaker on this occasion;^ no witty retort made it to the finals, but at least something halfway^ appropriate^ to^ the
speaker's current "mind-set" got blurted out. As the speaker walks away
after the encounter, he will probably resume the chaotic tournament, muttering and musing about what he should have said. The muse may
HOW WORDS DO THINGS WITH US 239 of (^) a questioner. (^) The residual problem was how to get rid of the clever questioner, a problem we^ postponed.^ Here^ we^ have^ the^ complementary problem: how to get answers to^ an^ eager^ flock of^ contestants^ asking
questions like "Why^ don't^ we^ say,^ 'Your^ mother^ wears^ army^ boots!'"
or (in another context) "Why don't we say, 'I seem to see a^ red spot moving and turning green as it moves'?"^ Two complementary^ prob-
lems — could they^ perhaps^ solve^ each other by^ being^ mated?^ What^ if
the word-demons are,^ in^ parallel,^ the^ questioners/contestants,^ and^ the content-demons are^ the^ answerers/judges?^ Fully^ fledged^ and^ executed communicative intentions — Meanings — could emerge from a quasi- evolutionary process of speech act design that involves the collabora- tion, partly serial, partly in^ parallel,^ of^ various subsystems^ none^ of which is capable on its own of^ performing^ —^ or^ ordering^ —^ a^ speech act. Is such a process really possible? There are a variety of models of
such "constraint satisfaction" processes, and they do indeed have strik-
ing powers. In addition to the various "connectionist" architectures of
neuroniike elements (see, e.g., McClelland and Rumelhart, 1986),^ there are other more abstract models. Douglas^ Hofstadter's^ (1983)^ Jumbo^ ar- chitecture, which hunts for^ solutions^ to^ Jumbles^ or^ anagrams,^ has^ the right sorts of features, and so do Marvin Minsky's (1985) ideas about the Agents making up the "society of mind" — which will be discussed further in chapter 9. But we must reserve judgment until models that are more detailed, explicit,^ and^ direcfly^ aimed^ at^ language^ production are created and^ put^ through^ their^ paces.^ There^ may^ be^ surprises^ and disappointments. We know, however, that somewhere in any successful model of language production we must avail ourselves of an evolutionary process of message generation, since otherwise we will be stuck with a^ miracle ("And then a miracle occurs")^ or^ an^ infinite^ regress^ of^ Meaners^ to^ set the task.1 We also know — from the research Levelt surveys — that
- Dan Sperber and Deirdre Wilson (1986) open up a new perspective on how we compose our communications by insisting on models of how things actually work, in the speaker and hearer, contrary to recent practice among philosophers and linguists. who have tended to wave their hands about the^ mechanisms while appealing to rational reconstructions of the supposed tasks and their (^) demands. This permits Sperber and Wilson to raise considerations of practicality and efficiency:^ least-effort principles, and concerns about timing and^ probability.^ They^ then^ show from^ this new^ perspective^ how certain traditional "problems" disappear — in particular, the problem of how the hearer finds the "right" interpretation of what the speaker intended. Although they do not pitch
240 AN EMPIRICAL THEORY OF THE MIND there are quite rigid and automatic processes that take over (^) eventually and determine the grammatical-to-phonological transformations that compose the (^) final muscular recipe for speech. The two caricatures de- fine (^) extremes along (^) a continuum, (^) from hyperbureaucratic to hyper- chaotic. Levelt's actual (^) model — in contrast to (^) the caricature I have used in order to make the contrast vivid — incorporates (^) (or can readily be made to incorporate) some of the nonbureaucratic features (^) of the opposing caricature: for example, there is nothing deep or structural preventing (^) Levelt's Formulator from engaging in more or less sponta- neous (unrequested, undirected) (^) language generation, and, given the monitoring loop through (^) the Speech-Comprehension System back to the Conceptualizer (see Figure 8.1), this spontaneous activity could play the sort of generating role envisaged for the multiple word-demons. Between the two caricatures there is an intervening spectrum of more realistic ways^ alternative models could be developed. The main ques- tion is how much (^) interaction is there between the specialists who de- termine (^) the content and style of what is to (^) be said and the specialists who "know the words and (^) the grammar"? At one extreme, the answer is: None. We could keep Levelt's (^) model intact, and simply supplement it with a pandemonium model of what happens (^) inside the Conceptualizer to fix the "preverbal message." In Levelt's model, there (^) is nearly complete separation between the pro- cesses of message generation (specs-setting) and linguistic (^) production (specs-meeting). When the first bit of preverbal message arrives at (^) the Formulator, it triggers the production of the beginning of an (^) utterance, and as the words get chosen by the Formulator, this constrains how the utterance can conUnue, but there is minimal collaboration on re- vision of (^) the specs. The (^) subordinate language-carpenters in the For- mulator are, in (^) Jerry Fodor's terms, "encapsulated"; in their automatic way, they do the best they can with the orders they receive, (^) with no ifs. ands, or buts. At the (^) other extreme are (^) the models in which words and phrases from the Lexicon, together with their sounds, meanings. (^) and associa- tions, jostle with grammatical constructions in a pandemonium, all "trying" (^) to be part of the message, and some of them thereby make a substantial contribution (^) to the (^) very communicative intentions that still fewer of them end up executing. At this extreme, the (^) communicative their model at the of evolutionary processes of the sorts we (^) have just been consid- ering, it certainly invites just such (^) an
242 AN EMPIRICAL THEORY OF THE MIND generation will turn out to involve Pandemonium — opportunistic, parallel, evolutionary processes — almost all the way down. The next section will review some of them briefly.
- WHEN WORDS WANT TO GET THEMSELVES SAID Whatever we may want to say, we probably wont say exactly that. MARVIN MINSKY (1985), p. 236 The Al researchers Lawrence Birnbaum and Gregg Collins (1984) have noted a peculiarity about Freudian slips. Freud famously (^) drew our attention to slips of the tongue that were not random (^) or meaningless, he insisted, but deeply meaningful: unconsciously intended insertions into (^) the fabric of discourse, insertions that indirectly or partially sat- isfied suppressed (^) communicative goals (^) of the (^) speaker. This standard Freudian claim has often been vehemently rejected by skeptics, (^) but there is something puzzling about its application to particular cases that (^) has nothing to do with one's opinion about darker themes of sexuality,^ the Oedipus complex, or^ death wishes. Freud discussed an example (^) in which a man (^) said Gentlemen, I call upon you to hiccup to the health of our Chief. (In German — the language actually spoken (^) in the example (^) — the word
for "hiccup," aufzustossen, was slipped in for the word for 'drink,"
anzustossen.) In his explanation, Freud argues that (^) this slip is a (^) manifestation of an unconscious goal on the part of the speaker (^) to ridicule (^) or insult (^) his superior, (^) suppressed by the social and political duty to do him honor. However,. one (^) cannot reasonably expect (^) that the speaker's intention to ridicule his superior gave rise originally to
a plan involving the use of the word "hiccup": A priori, there are
hundreds of words and phrases that can (^) more plausibly be (^) used to insult or ridicule someone... There is no way that a planner could have reasonably (^) anticipated that the goal of ridiculing or insulting its superior would be satisfied by uttering the (^) word "hic- cup," for exactly the same reason that it is implausible that the
HOW WORDS DO THINGS (^) WITH US 243 planner (^) would have chosen to use the word as an insult in the first place. The only process that could explain^ the^ frequency^ of^ serendipitous Freudian slips, they argue, is one of "opportunistic planning."
-. What examples like the above seem to indicate, therefore, is that the goals themselves are active cognitive agents, capable^ of commanding the cognitive resources^ needed^ to^ recognize^ oppor- tunities to satisfy themselves, and the behavioral resources needed to take advantage^ of^ the^ opportunities.^ [Birnbaum^ and^ Collins, 1984, (^) p. 1251 Freudian slips draw^ attention to^ themselves^ by^ seeming^ to^ be mistakes and not mistakes at the same^ time,^ but^ the^ fact^ (if^ it^ is^ one) that they satisfy unconscious goals does^ not^ make^ them^ any^ harder^ to explain than other word choices that fulfill several functions (or goals) at (^) once. It is about as hard to imagine how puns and other forms of intended verbal humor could be the result of nonopportunistic, encap- sulated planning and^ production.^ if^ anyone^ has^ a^ plan^ for^ designing witticisms — a detailed plan that actually works — there are more than a few comedians who would pay good money for if Birnbaum and Collins are^ right,^ creative^ language^ use^ can^ be accomplished only by^ a^ parallel^ process^ in^ which^ multiple^ goals^ are simultaneously on the alert for materials. But what if the materials themselves were at the same time on the alert for opportunities to get incorporated? We^ pick^ up^ our^ vocabulary^ from^ our^ culture;^ words^ and phrases are the most salient phenotypic features — the visible bodies — of the memes that invade us, and there could hardly be a more congenial medium in which memes might replicate than a language-production system in which the supervisory bureaucrats had partially^ abdicated, ceding a large measure of control to the^ words^ themselves,^ who^ in^ effect fight it out among themselves^ for^ a^ chance^ in^ the^ limelight^ of^ public expression.
- Levelt^ tells^ me^ that^ he^ himself^ is^ an^ inveterate^ pun-hunter^ (in^ his^ native^ Dutch), and he knows^ just^ how^ he^ does^ it:^ °By^ lifelong^ training^ I^ turn^ around^ just^ about^ every word (^) I hear. I then (quite consciously) check the result for^ its^ meaning.^ In^ 99.9^ percent of the cases there is nothing funny coming out. But one per thousand^ is^ fine,^ and^ those I express right away° (personal communication). This is a^ perfect^ example^ of^ von^ Neu- manesque problem-solving: serial, controlled^ —^ ond^ conscious?^ The^ question^ is^ whether there are other, more pandemonic, ways^ of^ generating^ wit^ unconsciously.
HOW WORDS DO THINGS^ WITH^ US^245 Habits of the Free Imagination, writes^ about her^ own^ process^ of^ com- posing short stories. Every story has a story. This^ secret^ story,^ which^ has^ little^ chance of getting told,^ is the history^ of^ its^ creation.^ Maybe^ the^ "story^ of the story" can never^ be^ told,^ for^ a^ finished^ work^ consumes^ its^ own history, renders it obsolete, a husk. [Hampl,^ 1989,^ p. 371 The finished work, she notes,^ is^ readily^ interpretable^ by^ critics^ as an artifact cunningly contrived^ to^ fulfill^ a^ host^ of^ sophisticated^ authorial intentions. But^ when^ she^ encounters^ these^ hypotheses^ about^ her^ own work, she is embarrassed: "Hampl" had precious few^ intentions,^ except,^ like^ the^ charlatan I suddenly felt^ myself^ to^ be,^ to^ filch^ whatever^ was^ loose^ on^ the table that suited my immediate purposes. Worse, the^ "purposes" were vague, inconsistent, reversible, under pressure. And^ who^ — or what — was applying the pressure? I couldn't say.^ [p. 371 How then does she do it?^ She suggests^ a^ maxim:^ "Just^ keep talking — mumbling is^ fine."^ Eventually,^ the^ mumbling^ takes^ on^ shapes that meet with^ the^ approval^ of^ the^ author.^ Could^ it^ be^ that^ the^ process Hampl detects on a grand scale in her creative writing^ is^ just^ an^ en- largement of the more submerged and swift process that^ produces^ the creative speaking of everyday life? The tempting similarity does not involve just a^ process^ but^ also a subsequent attitude or reaction. Hampl's confessional^ zeal^ contrasts with a more normal — and not really dishonest —^ reaction^ of^ authors to friendly interpretations by readers: these authors^ defer^ gracefully^ to the imputations of intent,^ and^ even^ willingly^ elaborate^ on^ them,^ in^ the spirit of "Hey, I guess that is what I was^ up to,^ all^ along!"^ And^ why not? Is there anything self-contradictory^ in^ the^ reflection^ that^ a^ certain move one has just made (in^ chess,^ in^ life,^ in^ writing)^ is^ actually^ cleverer than one at first realized? (For^ further^ reflections^ on^ this^ topic,^ see^ Eco, 1990.) As (^) E. M. Forster put it, "How do I know^ what^ I^ think^ until^ I^ see what I say?" We often do discover^ what^ we^ think^ (and^ hence^ what^ we mean) by reflecting on^ what^ we^ find^ ourselves^ saying^ —^ and^ not^ cor- recting. So we are, at least on those^ occasions,^ in^ the^ same^ boat^ as^ our external critics and interpreters,^ encountering^ a^ bit^ of^ text^ and^ putting the best reading^ on^ it^ that^ we^ can^ find.^ The^ fact^ that^ we^ said^ it^ gives^ it a certain personal persuasiveness or at^ least^ a^ presumption^ of^ authen-
246 AN EMPIRICAL THEORY OF THE MIND
ticity. Probably, if I said it (and I heard myself say it, and I didn't hear
myself rushing in with any amendments), I meant it, and it probably means what it seems to mean — to me. Bertrand Russell's life provides an example: It was late before the two guests left and Russell was alone with Lady Ottoline. They sat talking over the fire until four in the morning. Russell, recording the event a few days later, wrote, "I did not know I loved you till I heard myself telling you (^) so — for one instant I thought 'Good God, what have I said?' and then I
knew it was the truth." [Clark, 1975, p. 176J
What (^) about the other occasions, though, where we have no such sense (^) of a discovery of self-interpretation? We (^) might suppose that in these, (^) the normal, cases, we have (^) some intimate and privileged advance insight into (^) what we mean, (^) just because we ourselves are (^) the Meaners, the fons et origo of the meaning of the words we say, but such a sup- position requires a supporting argument, not just an appeal to tradition. For it could as well be the case that we^ have no sense of discovery in these cases just because it is so obvious to us what we mean. (^) It doesn't take "privileged access" to intuit that when I say, "Please pass the (^) salt" at the dinner table, I'm asking for the salt. I used to believe there was no alternative to a Central Meaner, but I thought I had found a safe haven for it. In Content and Consciousness I argued that there had to be a functionally salient line (which I called the awareness line) separating the preconscious fixation of communi- cative intentions from their subsequent execution. The location (^) of this line in the brain might be horrendously gerrymandered, anatomically, but it had to exist, logically, (^) as the watershed dividing malfunctions into two varieties. (^) Errors could occur anywhere in the whole system, but every error had to fall by geometric necessity — on one side (^) of the line or the other. If they fell on the execution side of the line, they were (correctable) errors of expression, such as slips of the tongue, malapropisms, mispronunciations. If they fell on (^) the inner or higher side of the line, they changed that which was to be expressed (the "preverbal message" in Levelt's model). Meaning was fixed at this wa- tershed; that's where meaning came from. There had (^) to be such (^) a place where meaning came from, I thought, since something has to set the standard (^) against which (^) "feedback" can register failure to execute. My mistake was falling for the very same scope (^) ambiguity that bedevils the interpretation of Abe Lincoln's dictum. There does indeed have to be something on each occasion that is, for the nonce, the stan-
