Gavagai, good God-agai, not again

This Anus is not rare in that it can speak. If given a chance, most Anuses can, and since virtually all are given the chance — nay, since all are required to — virtually all do. But sometimes one shouldn’t just listen to what an Anus has to say; one should reflect on the fact that it simply says anything. For it is not so much amazing that an Anus can speak well as it is amazing that an Anus can speak at all — and what’s more, that it can speak with so many different mutually exclusive tongues that are nevertheless largely inter-translatable. In this respect, a speaking Anus is a miracle (it is unique in nature) without really being a miracle in the strict sense of the term (every individual in the species can learn, and does learn, any one of about 6500 currently used languages) [1].

Such is the mystery facing any philosopher or scientist who reflects on language, reference, and translation. Or at least, it should be.

If one listens to Quine, the fact that any two people can speak about the same object is, conceptually speaking, nothing short of a miracle — literally, a miracle. Similarly, the fact that any two people can translate their languages into each other’s is equally miraculous, also in a literal sense. In fact, for Quine, at bottom these two problems are one of apiece: one manifestation of the problem of the indeterminacy of reference is the problem of radical translation, and vice versa, and for him, this dual-sided problem is insurmountable, and it has far-reaching implication. For in so far as any attempt to know expresses its truth(s) in language(s), the problem of reference in radical translation represents a permanent limitation on success, be it success in every-day common sense discourse, or success in the language of physics.

A lot, apparently, rides on this fact — so much so the Anus had to go another round.

Since it had already seen how Quine handled the problem of reference in The Roots of Reference, it wondered how he might have handled this related joining of “reference” and “radical translation” in Word and Object, fourteen years earlier. Perhaps, it thought, that formulation — one of the most famous in the anal-lytic philosophy of language — might be instructive…

The Anus was tight with anticipation…

As Quine sets it up, the problem of radical translation is relatively straightforward. Imagine a linguist travels to a strange land where the natives speak a new language, one never interpreted by anyone from his own language group. He wants to set up a translation manual. He meets the folk, a rabbit runs by, and one of the natives exclaims “Gavagai.” To begin translation, the linguist needs to know what “Gavagai” means; that is, he needs to know what “Gavagai” refers to — hence his problem of translatability is simultaneously a problem of reference, and it is “radical” in that his is the first encounter of anyone in his group with this particular language.

Quine then lays out how the linguist would proceed. First, he notes: “the recovery of the man’s current language from his currently observed responses is the task of the linguist who, unaided by an interpreter, is out to penetrate and translate a language hitherto unknown.” And per previous discussion, he also notes: all the linguist really has to go on are “the forces that he perceives impinging on the native’s surfaces,” and the native’s “observable behavior, verbal and otherwise.” Thus put, the problem of radical translation is simultaneously the problem of reference in an indeterminate sensory environment in which all that is known is known through “surface irritations” on the senses and the observable responses to them, again, verbal or otherwise. What’s more, these irritations represent the only stimuli, and the task of the translator is to determine the stimulus meaning that determines the association among them that is reference. Once reference is determined, translation can commence with all the due diligence required (which as it turns out, is extensive).

Good god, not again. Does the Anus even have to, because as posed the problem is so mis-specified no Anus could even begin to solve it. As a first step, then, the problem needs to be reformulated.

Put simply, in terms of straightforward observation, two men, both of whom already speak a language, want to understand one another. One sees what the other calls a “rabbit” run by and utters “Gavagai”; the other wants to know what this utterance at the time of this event means. Does it refer to the object, the rabbit as “a rabbit”? Does it refer to the rabbit as “dinner”? Does it refer to the rabbit as “rabbit as running”? Does it refer to the rabbit at all? And so forth. The linguist wants to establish this reference as a gateway to translation, i.e., as establishing a common ground between languages for transposing the words and rules from one language to the other. At the outset, though, the only means at his disposal are the language the native knows, the language he knows, both their observable behaviors, both their observable intentions, and the easily discernible objects in the common environment. Several runs of trial and error will be required to determine what “Gavagai” vis-à-vis ‘a rabbit running by’ means. Translation can then begin with all due diligence.

Framed in this way, the problem makes sense; it can be investigated scientifically, as it in fact has been — extensively. Framed in Quine’s way, well….

First, the Anus notes right of the bat that the problem of “radical translation,” as Quine formulates it, is neither all that “radical” nor really a problem at all. Or more precisely, any reader of Quine, any reader of this blog — in short, anyone who has acquired a language — has already solved it not just in being able to speak in accord with others about specific objects, but also by the simple fact of being able to speak in the first place. For in an abbreviated form, “solving” the “problem” is essentially what one did as an infant while leaning language and reference for the first time.

Second, even if it was a “problem” in the sense of asking how this learning comes about — and most emphatically it is an interesting scientific problem in that sense — Quine, once again, obliterates the issue. Like with the problem of reference in The Roots of Reference, he conceptualizes the problem of “radical translation” not only in the wrong way; his conceptualization renders inexplicable the observable accomplishments of language learning, including translation, and it renders unsolvable the very the formulation it sets for itself to solve. No wonder then, as he concludes, radical translation is insurmountable!

As a first step to seeing just how this obliteration occurs, it is illustrative to start by noting the similarity between his two formulations of the underlying problem of reference, in Word and Object here and The Roots of Reference fourteen years later.

In an important respect, the linguist trying to understand “Gavagai” from a native speaker is very much like the infant encountering “native” speakers in The Roots of Reference, in that the linguist knows little better than the infant just what “Gavagai” is. Is it “gav a gai”? Is it “gava gai”? Is it “gavagai”? And so forth. Phenomenally — what Quine means by “surface irritations (the Anus will come back to that in a moment) — “Gavagai” is a rhythmic stream of phonemes, not a sentence, or even a word, thus parsing that stream into words — Quine’s “observation sentence” — is the first task the linguist faces. In fact, this first task is paramount.

Quine sort of mentions this “phonetic” problem only to dismiss it as irrelevant, but of course that dismissal is an error because how the ‘observation sentence’ is phonetically structured contains important information for decoding what that observation sentence might mean. For instance, if it is one word, the interpretive bias may be that it only refers to the object as “Gavagai”, or “rabbit.” If it is several, maybe it refers to “rabbit running,” as with “gava gai”. Or maybe it refers to “dinner is escaping”, as in “gav a gai”…and so forth. Starting from “Gavagai” as just one “surface irritation” — as the basic unit of potential meaning — omits the observable fact that that surface irritations have a pattern and that that pattern, in important respects, can be dispositive for what that irritation ultimately means.

So, the linguist and the infant are faced with a similar problem right off the bat: parsing the phonetic stream of sound into a unit or units which in turn mean(s) something. Once this parsing is done, then — and only then — can the two engage in the problem of mutually understanding the reference of the observation sentence, i.e. only then does the problem of reference even arise [2].

But set this similarity aside. Set aside the fact that learning a radically new spoken language and learning a language as an infant are at first blush more alike than different. Grant that Quine can start where he does, with: 1) language ‘as such’ as already understood and 2) a known observation sentence with an unknown but determinable referent, with the unknown referent suggested at first only as a vocal utterance corresponding to a “surface irritation” on the part of the native (and presumably the utterance itself is only known via a “surface irritation” on the eyes and ears of the linguist). Scientifically there is nothing wrong with starting at a “middling point” in the conception of language and reference, and besides, Quine says elsewhere that he’s doing just that.

So, the Anus will assume, with Quine, that “Gavagai” is a one-word observation sentence that refers to an object or event, and that the linguist knows this. From there, how does Quine say the linguist must proceed?

Since all the linguist has to go on are the “the forces he perceives impinging on the native’s surfaces” and his “observable behavior, verbal or otherwise,” Quine says the linguist must first establish a code for affirmation and negation, then he must proceed by trial and error to determine whether or not the native associates “Gavagai” with a rabbit — in other words, when prompted, would he assent to “Gavagai” when the linguist would assent to “rabbit.”

At least that is what one would think, but Quine cautions otherwise. For he says: “it is important to think of what prompts the native’s assent to ‘Gavagai’ as stimulations and not rabbits” — a “stimulation” that “is best perhaps identified, for present purposes, with the pattern of chromatic irradiation of the eye.”

In other words, “translation at the present stage turns solely on the correlation [between] non-verbal stimuli” and the verbal utterance, and it is this correlation that defines the “stimulus meaning.”

Stimulus meaning in hand, then, the problem for the linguist is one of linking ‘the pattern of chromatic irradiation‘ of one’s own eye to the ‘verbal utterances’ of the speaker, such that the following law applies: “the native would assent to ‘Gavagai’ under just those stimulations under which we, if asked, would assent to ‘rabbit’ ” (emphasis added). With this granular approach in hand, one only needs the additional assumption that “microscopically the same irradiations would prompt him” to say the same thing, just in his language, even though “this conjecture rests wholly on samples where the irradiations concerned can at best be hazarded merely to be pretty much alike.”

Once all this is done, Quine says the linguist can proceed to determine experimentally whether or not “Gavagai” means “rabbit.”

(He goes on to add that these experiments would need to account for various confounds like the vividness and salience of stimuli, including central versus peripheral stimulation, since “ocular irradiation patterns are best conceived in their spatial entirety.” He even suggests that trial and error proceed by “evolving ocular irradiation patterns between properly timed blindfoldings,” and he warns that “we must allow for the speaker to change his ways,” i.e. for the stimulus meaning to change…the factors involved are quite extensive, and the tight [sic], logical [sic] reasoning [sic] required to eliminate them all takes up about fifty pages.

Although the details are far too numerous to recapitulate here, it turns out that learning a word in practical terms is an enormously elaborate experimental affair…)

Now…if what’s been indicated so far sounds like a bunch of half-baked, pseudo-rigorous, pseudo-scientific mumbo jumbo — and the Anus refers the reader to those un-recapitulated fifty pages — that’s because that’s precisely what it is. Not only does what is describe bear no resemblance whatsoever to the observable phenomena of the original situation — of reference and translation; it renders that problem unsolvable by stripping it of all the meaning that makes it the problem that it is.

In other words, instead of seeing the problem as two people trying to understand one another over a rabbit running across a scene, Quine turns the problem into one of isolating a physical stimulus from within a nexus of inert and living matter, subject to interactive forces acting in an arena of objective space and time, where miraculously somehow “meaning” emerges for the hearer because by some other miracle, it was also there for the speaker.

Why on earth would anyone want to do this?

This is not to say, of course, that matter in motion or irradiation is not happening in the original situation, but why turn a language problem into a physics problem — or in this case, a psychophysics problem? Formulating the problem in this way entirely misses the point; it mis-specifies it.

To see how — as though it’s not clear already — consider an analogy: determining the taste of a cake. Under Quine’s way of proceeding f, to know how a cake tastes, one would break down the ingredients into their chemical components, determine the transformation of these components under the controlled conditions of mixing and heat, then measure the reactions caused by these components on the taste buds of some in vitro tongue, then measure the reactions on those taste buds, then finally model how those nerve actions are organized in the brain. Once all this scientific and experimental work is done, only then can we say we know what the cake tastes like.

Of course, approaching the problem this way logically and literally makes no sense. The obvious alternative is just to bake the damn cake and taste it, then try to reconstruct what ingredients contributed to what flavors, and how. Quine simply obliterates this rather obvious methodological starting point for any scientific formulation of the problem of language learning with his behaviorist-cum-psychophysicist mumbo-jumbo. In it, he strips the situation of the meaning that defines the problem as the problem that it is, and this stripping pervades — even determines — his analysis in the rest of the book.

But what’s in name calling? Maybe he’s onto something. Maybe one should ignore the meaning inherent to the problematic situation as too “subjective” and formulate it instead in more objective, behavioral-psycho physical terms. Surely that’s Quine’s intent — to be objective. Maybe this more “objective approach” gets to the heart of the problem and offers the best truly scientific solution...

Except that it doesn’t, and it doesn’t because Quine’s error goes deeper than just a methodological mis-specification of the problem in behavioral-psycho physical terms; for those terms are not, in and of themselves, erroneous. Rather, Quine’s use of them begs the scientific question — or more precisely, his use presupposes exactly what it purports to explain, in that there is no way to go from “surface irritations” on the ear and “chromatic irradiations” on the eye to “sound of the word ‘Gavagai’” and “rabbit moving through a visual scene” without presupposing, in an ordinary sense, the sound and the spectacle that is to be explained by the behavioral-psycho physical terms.

To cast the point in Quine’s details…

…in order for the “chromatic irradiation” to be an object moving in the visual field, much less a rabbit moving across a scene, first one has to see a rabbit moving in the visual scene and then conceptually map — as it were — the pattern of “chromatic irradiation” onto that event, otherwise it couldn’t be an explanation of that event. In other words, starting with “surface irritations” and so forth in fact presupposes the ordinary meaning of the perceived terms in their explanation, so why not just begin by expositing that ordinary meaning by making it explicit? Why not first analyze and test its implications for the translation and reference problem, instead of turning to “explanations” of those terms as the first approach to that problem?

Logically, this is what one should do, first because the terms are presupposed anyway, and second because it goes nowhere to try to explain what you’ve tacitly presupposed in the first place. Without the common sense rabbit and word as referents for the theoretical account of a “stimulus meaning,” references to “chromatic irradiation” or “tympanic vibration” amount to explanations of, literally, nothing — or alternatively stated, it could be literally anything. The “surface irritations” would be just two events in space and time, events proximally indistinguishable from one another and from other nearby events, events which are unrelatable to the perceived rabbit and the spoken word except by some miraculous fiat, one that tacitly invokes what is driving the explanation anyway.

That’s not science. Science is not fiat and tacit invocation, and it simply will not do to generate explanations in this way. Just as opium doesn’t put people to sleep because of “dormitive power,” language isn’t learned via “surface irritations”). Quine, however, does philosophy-cum-science in this ways; he does nothing more than invent “stimulus meaning” by fiat when, after cautioning that it is important to start with “stimulations” and “surface irritations” instead of ‘rabbits’ and ‘words.’ In practice he is referring to rabbits and words all along, just through the back door, and the rest of his behavioral-psycho physics and pseudo-scientific mumbo-jumbo imports all its derived meaning through that unacknowledged back door [3].

Methodologically, then, Quine’s starting point fails. In fact, it must fail because it only moves in a tight little circle that presupposes in its explanation a solution to the very problem it purports to explain. Applied to this specific problem, the problem of translation, any attempt to determine the link between a word and object— and hence reference — by invoking the concept of “meaning” after “stimulus” in a combined concept stimulus meaning presupposes reference itself as the meaning of the stimulus. Thus, the “explanation” of reference through “stimulus meaning” is entirely circular; the explanans is tacitly contained in the explanandum.

Logically and conceptually speaking — despite all its apparent ‘circumspect scientific care’ and ‘rigorous analysis’ — Quine’s conceptual fiat and circular explanation amounts to trying to explain one side of a coin using the other without even looking at either side, all the while tacitly presupposing that one has done just that.

Explanations of that sort will always fail, and they fail because they neither explain (being mere ad-hoc constructions) nor formulate a real problem (the constructions conceal it). Scientific pretensions notwithstanding, they go no-where — or, as Dewey put it best, they “will consist simply in making explicit the assumptions which have tacitly been made in stating the problem — subject to the conditions involved in failure to recognize that they have been made.”

Sellars, as shown before, bases an Objective Realist epistemology on this kind of non sequitur. With the “problem” of radical translation and the roots of reference, Quine apparently does too.

None of this is to say that there aren’t interesting — even crucial — philosophical and scientific implications to the actual accomplishments of reference and translation, despite the inherent indeterminacy of both. There are, and with this in mind, the Anus can only speculate that the scientists who cite Quine’s “Gavagai” in their own work on this issue (there are a few) haven’t actually read how he poses the problem and purports to solve it. Instead, since he is a famous philosopher, they probably give him the benefit of the doubt of a rigorous philosophical formulation, one that analyzes the conceptual implications of two basic, widely noted observations: that reference, as conventional, is in an important respect indeterminate, and that translation, because of this (and for other reasons too), is never perfect or complete either [4].

This recognition, however, is wrong. In Word and Object Quine turns these rather mundane, everyday starting observations into an intractable philosophical puzzle by contorting them into mis-specified terms that both obliterate the underlying issue and render the puzzle insolvable. From this apparent unsolvability, he draws a whole nest of conclusions about translation, reference, truth and language.

Yet, as matter of fact, it is solved every day.

In this respect, then, Quine’s approach to translatabilty in Word and Object is much like his approach to reference in The Roots of Reference, in that it obliterates a problem. But in a deeper respect, it is even worse than being misguided in that way: here, it’s just plain stupid — “stupid” in that there is absolutely no evidentiary ground for reducing language comprehension to a physical and psycho physical process, one onto which meaning is grafted by some sort of miraculous fiat, a fiat that presupposes precisely what it sets out to explain — which means, scientifically speaking, the whole endeavor “explains” nothing. In fact, the move is so stupid it’s hard to believe anyone would even propose it, much less “one of the most influential philosophers of the twentieth century.”

Fortunately, occasional citations to him notwithstanding, the science addressing the issue underlying Word and Object has emerged entirely without Quine’s help, and what’s more, it never could have emerged had it relied on his anal-lysis to guide it.

One Anus to another, what’s the point of a philosophy when it’s as constipated as that? [5]

[1] The Anus omits, with sympathy, autistic and other disabled individuals from this list.

[2] The Anus is setting aside the problems with Quine’s idiosyncratic use of “sentence” for “word”, and vice versa, in his use of “observation sentence.” Like in The Roots of Reference, this conflation starts too high up the conceptual food chain, just in a different place. It is also ignoring the obvious possibility that the utterance isn’t an observation sentence at all but a different kind of utterance, maybe an invocation for a blessing from a dead ancestor, or what not. Since the oil has room for only so many fish to fry, the anus will stipulate, with Quine, that the basic, first unit of language is an “observation sentence,” be it one word or many, one that has, as it were, a ‘single’ referent.

It could also be noted that this native speaking linguist faces a special problem that pre-verbal infants don’t face, namely, discriminating non-native phonemes. For not long after learning their native phonemes — and certainly long before becoming competent language speakers, infants lose their ability to distinguish non-native phonemes in naturally spoken speech. On light of this, the native speakers are going to have to take specific measures to parse the phonetic stream into its proper inflections and words.

[3] Mutatis mutandis, in a perhaps interesting sense (though the anus doubts it), a psychophysiology of the translation situation could provide, if developed well enough, an isomorphic understanding of the situation as it actually occurs between the linguist and the native. But this isomorphic explanation can’t solve the problem of translation by itself, for the solution must be presupposed before the psycho physical explanation can make any sense as an account of it.

[4] To see that neither the formulation nor the implication is new to Quine’s “Gavagai,” the anus notes that Dewey long anticipated both the example and the significance of it for reference with his “table” example in Logic: The Theory of Inquiry (p. 59). There he points out the ‘indeterminacy of reference’ (and therefore translation) in a direct way that not only undercuts Quine’s methodological stupidity; it also pre-empts his ‘theory of meaning’ in with respect to ‘behavioral use’ by many years (the amplified discussion in Chapter 5 of Experience and Nature pre-empts it by even more). The anticipation and implications in both works are so obvious the anus refers the reader there instead of recapitulating the example and analysis here. For discussions of reference within the context of behavior and communicative intention, including references to competing theories and evidence, see Tomasello’s The Cultural Origins of Human Cognition and Constructing a Language. For the Anus, they represent as a good place to start as any in the massive literature on the topic, without representing a definitive endorsement of the theory.

[5] The Anus may not be entirely fair here. Perhaps with the love and patience of a wife one could look past the stupidity and find something enduring about language, truth, reference, and translatability in Word and Object. But since the Anus is already married, any such effort would amount to cheating. But maybe someone could love Quine and still get a lot out of the relationship…the Anus just doesn’t (want to?) see how.