Contribution
A short essay that gives translation theory its most cited taxonomy and puts translation at the centre of linguistics. It defines meaning semiotically, as translatability into further signs; it distinguishes three kinds of translation; and it relocates the untranslatability debate from what languages can express to what their grammars force speakers to express. That last move is the essay’s substantive thesis.
Key insight
Languages do not differ in what can be said in them but in what must be said. A grammar poses yes-or-no questions (completed or not? male or female? one, two or more?) that every sentence has to answer. Translation between languages whose grammars pose different questions is therefore not loss alone: it also forces the translator to supply answers the source did not give, and to leave unanswered questions the source did answer.
Assumptions
- A semiotic theory of meaning. Meaning is not the referent but the sign’s translatability into other signs (Peirce’s interpretant): “no signatum without signum”. The essay argues this against Russell with ordinary-language examples, not by a theory of reference.
- A cognitive/poetic split. In its cognitive function language is “minimally dependent on the grammatical pattern”; in jest, dreams, magic, everyday verbal mythology and poetry, grammar carries “high semantic import”. The essay’s conclusions about translatability differ on the two sides of this line, and where the line falls is not specified.
- Metalanguage. Speakers can talk about their language, so vocabulary can be revised and redefined; this is what makes deficiencies repairable.
Key results
The essay argues; it does not prove or test. Its theses:
- T1 (three kinds). Intralingual translation (rewording), interlingual translation (translation proper), intersemiotic translation (transmutation): interpretation of verbal signs by signs of the same language, another language, or a nonverbal sign system.
- T2 (no full equivalence of code-units). Within a language synonymy is not complete equivalence (“every celibate is a bachelor, but not every bachelor is a celibate”); a code-unit is fully interpreted only by a message about it. Between languages there is “ordinarily no full equivalence between code-units”, but “messages may serve as adequate interpretations of alien code-units or messages”. Translation “substitutes messages in one language not for separate code-units but for entire messages”; it is reported speech, “two equivalent messages in two different codes”.
- T3 (expressibility). “All cognitive experience and its classification is conveyable in any existing language”, by loanwords, loan-translations, neologisms, semantic shifts and circumlocutions (Chukchee “rotating nail” for screw, “writing soap” for chalk); “any assumption of ineffable or untranslatable cognitive data would be a contradiction in terms”. A missing grammatical category is translated lexically (Old Russian dual: “two brothers”); “and/or” can be rendered in languages that lack it.
- T4 (obligatory categories). After Boas, the grammatical pattern “determines those aspects of each experience that must be expressed”. “Languages differ essentially in what they must convey and not in what they may convey.” Translating into a language with a category the source lacks forces a choice (dual/plural for “She has brothers”) or an explicit disjunction (“either two or more than two”). For “I hired a worker” a Russian must choose perfective or imperfective (nanjal / nanimal) and a masculine or feminine noun (rabotnika / rabotnicu), and the Russian still does not say whether I “hired” or “have hired”, or “a” or “the” worker.
- T5 (chains). Because the two grammars pose “quite different sets of two-choice situations”, “a chain of translations of one and the same isolated sentence from English into Russian and vice versa could entirely deprive such a message of its initial content” (Karcevski’s image: a circular series of unfavourable currency transactions). “But evidently the richer the context of a message, the smaller the loss of information.”
- T6 (grammar in verbal mythology). Grammatical gender shapes personification: a 1915 Moscow Psychological Institute test found Russians personify Monday, Tuesday and Thursday (masculine nouns) as male and Wednesday, Friday and Saturday (feminine) as female; the fallen knife and fork superstition; Repin and German die Sünde; Pasternak’s My Sister Life defeating a Czech translator, since Czech život is masculine. Constantine the Philosopher’s preface to the Old Church Slavonic Gospels already names the problem and resolves it in favour of “cognitive values” over words.
- T7 (poetry). In poetry “verbal equations become a constructive principle of the text”; every constituent of the code carries “autonomous signification” and paronomasia reigns. “Poetry by definition is untranslatable. Only creative transposition is possible”: intralingual, interlingual or intersemiotic (into music, dance, cinema, painting). Traduttore, traditore rendered as “the translator is a betrayer” loses its paronomasia, and a cognitive attitude would ask: “translator of what messages? betrayer of what values?”
Claims
| id | claim | strength | support |
|---|---|---|---|
| C1 | Interlingual equivalence holds, when it holds, between messages, not between code-units | moderate | argument with lexical examples (cheese/syr, bachelor/celibate) |
| C2 | Any cognitive content is expressible in any language, with lexical means filling grammatical gaps | weak | asserted; examples of neologism and circumlocution; no counterexample considered |
| C3 | Languages differ in what their grammar obliges a speaker to convey, so translation forces added choices and leaves some source information unexpressed | moderate | worked examples (dual/plural, aspect, gender, tense, articles), after Boas |
| C4 | Repeated translation of an isolated sentence between two languages can lose its initial content, and richer context reduces the loss | weak | stated by analogy (Karcevski), with no demonstration |
| C5 | Grammatical gender shapes how speakers personify and mythologize | weak | one 1915 test cited without data, and anecdotes |
| C6 | Poetry is untranslatable; only creative transposition is possible | moderate (definitional) | follows from the stated account of poetic function |
Concepts
- intralingual translation / rewording: interpretation of verbal signs by other signs of the same language.
- interlingual translation / translation proper: interpretation of verbal signs by some other language.
- intersemiotic translation / transmutation: interpretation of verbal signs by signs of nonverbal sign systems.
- code-unit: a word or idiomatic phrase-word, “a code-unit of the highest level”; contrasted with a message.
- equivalence in difference: the relation between two messages in two codes that translation aims at; called “the cardinal problem of language”.
- creative transposition: what remains possible where translation is not: recasting a poem into another poetic shape, language or medium.
Connections
The essay sits between Peirce’s semiotics (meaning as interpretant, via Dewey’s account) and Boas’s view of grammatical categories as obligatory, and it opposes Whorf’s “Mr. Everyman” inference from unlike grammars to unlike facts. It anticipates Jakobson’s own “Linguistics and Poetics” (1960) on the poetic function, here stated briefly. In the record, Nida’s (LIT-791) formal and dynamic equivalence and House’s (LIT-789) overt and covert translation are later answers to the question this essay calls “equivalence in difference”; neither was read.
Bearing on the record
- Produces THEORY-170, the obligatory-categories thesis (T4 with T5), stated with its evidence and what it does not establish. The record held no claim of this kind.
- The manuscript’s argument (
what-survives-translation). Its claim codes are left to the coordinator; the bearings, stated here so a reader can check them:- On the claim that a translation can broaden readings along one dimension while narrowing them along another: T4 is an early, concrete statement of it at the level of grammar. The Russian rendering of “I hired a worker” fixes aspect and sex the English left open, and leaves open tense and definiteness the English fixed.
- On the central thesis (“not the wording or the proposition”): C1 is a classical ancestor of “not the wording”, but Jakobson’s unit is the message’s cognitive content, not its pragmatic force. He does not address footing or illocution, so the essay gives nothing to the claim that propositional sameness is insufficient.
- On reconstruction chains: T5 is a 1959 statement that back-and-forth translation can lose content, with context as the damping factor. It is an assertion with an analogy, not evidence.
- On cross-modal renderings: T1 names intersemiotic translation as one of three kinds of translation, which is prior art for treating cross-modal rendering under one account. The essay says nothing about what such renderings preserve.
- On fidelity as one goal among adaptation, critique and parody: T7 is a narrow precedent. For poetry Jakobson replaces fidelity with “creative transposition”, and his closing question (“betrayer of what values?”) makes fidelity relative to a chosen value. He does not discuss adaptation, critique or parody.
- On significance fixed by contrasts within a system: the essay’s semiotics is Peircean (meaning as translation into further signs), not Saussurean differential value. It is compatible with that claim but does not support it.
- No instruction for machine-learning practice; nothing for the anthology.
Limitations
- An essay of eight pages argued from examples; nothing is tested.
- T3’s universal expressibility is asserted, and the essay’s own exceptions (poetry, jest, magic, verbal mythology) are large and not bounded.
- The chain claim (T5) is a “could”, with no example traced through several steps.
- The 1915 personification test is cited without data or a source.
- It treats cognitive content as the thing to be preserved and does not consider pragmatic force, register or the social act of the utterance.
Open questions
- Whether round-trip chains between languages with mismatched obligatory categories do lose the source’s distinctions at the rate T5 implies, and whether added context damps the loss: a controlled study of translation chains would close it.
- Where the cognitive/poetic line falls, and whether ordinary conversational language (stance, politeness) sits on the cognitive side or with “verbal mythology”.