Asymmetric Constructability and the Whorfian Challenge to Translation — Epoche C2
Every finite clause of Tariana, an Arawak language of the Vaupés region in north-west Amazonia, must carry one of five suffixes recording how the speaker knows what the clause reports: by seeing it, by hearing or otherwise sensing it without seeing, by inference from visible evidence, by assumption from general knowledge, or by report from someone else. An English sentence handed to a translator almost never settles which of the five applies. The translator must nevertheless choose one, and in choosing supplies the Tariana text with information the English text did not contain. Aikhenvald, who worked with Tariana speakers over many years, reports that a wrongly chosen evidential is not heard as a stylistic lapse but as a lie. Asymmetries of that shape are the subject of this essay. The name proposed for them here is asymmetric constructability: the claim that where one language encodes a distinction in its grammar or basic lexicon and another does not, the second can usually construct what the first encodes, but the cost of construction in one direction bears no fixed relation to the cost in the other, and in some cases one direction is not a matter of cost at all but of the target grammar refusing to leave the matter open. The claim is offered as a reading of what is genuinely difficult in Whorf's challenge to translation, in place of the reading that the difficulty lies in divergent habitual conceptions of reality. Three things have to be established for it: what the universalist position it opposes actually asserts; what Whorf actually asserted, since neither of his two most-quoted examples now stands as he left it; and what quantity the asymmetry is an asymmetry in. The position under attack: Katz's principle of effability A thesis about the limits of translation needs an opponent with a statable commitment, and there is one. Jerrold Katz's principle of effability holds that each proposition is expressible by some sentence in any natural language. Katz advanced it in his 1978 paper on effability and translation as the ground of the possibility of translation: if every proposition has an expression in every language, then for any source sentence there exists a target sentence with the same propositional content, and translation is guaranteed a target to aim at. What the principle says and what it is often taken to say come apart at exactly the point this essay occupies. It asserts the existence of an expression. It says nothing about that expression's length, nothing about whether the language would ordinarily produce it, and nothing about whether the target sentence, in expressing the source proposition, is obliged to express others as well. A defender of effability can concede every empirical claim made below without retracting a word, which is why the essay's target is not the principle but the inference customarily drawn from it — that a guaranteed target makes the asymmetries a matter of stylistic inconvenience. What Whorf claimed, and what became of his two stock examples Before the alternative reading can be motivated, the received one has to be stated in Whorf's own terms rather than in the paraphrase this essay previously used. That paraphrase — that distinct linguistic systems entrench categorisations of reality so deeply that direct conceptual mapping becomes problematic — describes a doctrine about entrenched concepts. Whorf's mechanism in the 1939 paper on habitual thought and behaviour is narrower and more testable than that: grammatical patterning supplies analogies, and speakers extend those analogies to non-linguistic conduct. His evidence was drawn from his employment as a fire-prevention inspector: workers who smoked near gasoline drums labelled "empty", where the vapour left in a drum is more dangerous than liquid petrol, because "empty" carries an analogy of null, inert, nothing-there. The claim being made is about behaviour, and the behaviour was observed. That is a stronger form of the thesis than "habitual thought", and it is the form the received reading should be credited with. His two celebrated illustrations, however, have not survived, and the essay as it stood leaned on both. The first is Hopi. Whorf's claim, in the manuscript on the Hopi model of the universe, was that Hopi contains no words, grammatical forms, constructions or expressions referring directly to what English speakers call time, or to past, present and future. Malotki's Hopi Time of 1983 is a book-length demonstration that it does: temporal adverbials, units of time, tense and aspect morphology, and metaphors of temporal extent, documented from recorded speech and elicitation. The strong claim is false, and the contrast between a "temporal" and a "timeless" language cannot be used as a premise. What survives is the weaker and uncontroversial observation that Hopi and English grammaticalise different portions of the temporal field, which is a fact about obligatoriness rather than about the presence or absence of a concept — and obligatoriness is where this essay is heading anyway. The second is snow. Boas's introduction to the Handbook of American Indian Languages of 1911 gave four Eskimo stems: aput , snow on the ground; qana , falling snow; piqsirpoq , drifting snow; qimuqsuq , a snowdrift. Martin's 1986 article traced what happened to that figure of four as it passed through Whorf's 1940 essay into textbooks and then into journalism, arriving at counts in the dozens and hundreds with no new fieldwork behind any of them. Two consequences follow, and they pull in opposite directions. The forms this essay used, aput and qana , are Boas's own and are not in dispute, so the illustration does not have to be abandoned. But the framing does. Eskimo–Aleut languages form words by productive affixation, so "how many words for snow" has no determinate answer: the question presupposes an inventory of the kind these languages do not have, and a stem plus its derivations can be counted as one item or a