Contribution
The thesis argues that translation needs no general theory of its own. Relevance theory already explains it: most translation is interpretive use of language across a language boundary, and the principle of relevance supplies a notion of faithfulness that is general and yet fitted to each text and audience (pp. 3–4, 190). It distinguishes indirect translation, which may resemble the original loosely, from direct translation, which aims at complete resemblance, and shows the second to be the limiting case of the first (pp. 254–255). It is the doctoral version of the account published in 1991 as Translation and Relevance: Cognition and Context (LIT-784).
Key insight
A translation is not judged by how much of the original’s wording or propositional content it keeps. It is judged by how far the assumptions it communicates to its audience overlap with those the original communicated, and in which respects that overlap matters to the audience. Wording matters only as a clue to interpretation: a translation that changes the words and keeps the interpretation keeps what counts.
Assumptions
- Relevance theory, as in Sperber and Wilson (1986). Communication is ostensive-inferential. Every act of ostensive communication carries a presumption of its own optimal relevance: adequate contextual effects for no unjustifiable processing effort (pp. 50–52).
- Thoughts and utterances have propositional forms, and a representation can be used descriptively, as true of a state of affairs, or interpretively, in virtue of its resemblance to another representation (pp. 56, 60–61).
- Concepts have logical, encyclopaedic and lexical entries (pp. 209–211). Connotation, register, the force of greetings and proverbs are placed in the encyclopaedic entries of the words themselves (pp. 229–240).
- The translator has no access to the audience’s cognitive environment, only beliefs about it, and the account predicts failure where those beliefs are wrong (p. 180).
Key results
- Interpretive resemblance between propositional forms (p. 57). After Wilson and Sperber (1988), two propositional forms “interpretively resemble one another in a context C to the extent that they share their analytic and contextual implications in the context C.” Resemblance comes in degrees, and two thoughts with no analytic implication in common can resemble each other through a shared contextual implication (pp. 59–60).
- Resemblance between utterances is a matter of what is communicated (pp. 70, 74–75). Between utterances, resemblance “is generally a matter of what overlap there is in the sets of assumptions that are communicated by each utterance rather than merely a matter of sharing analytic and contextual implications” (p. 70). “Since the linguistic form of an utterance can influence its contextual effects, identity in propositional form between two utterances is not a sufficient condition for identity in their interpretations” (p. 72): “So Charles has lost his car keys” and “After all, Charles has lost his car keys” share a propositional form and differ in interpretation. Gutt amends the definition so that each utterance is interpreted in its own context: “two utterances interpretively resemble one another to the degree that the sets of assumptions which each conveys in its own cognitive environment overlap, that is, interpretive resemblance between utterances depends on the degree to which they share their explicatures and/or implicatures” (pp. 74–75). The worked case is a report of “The back door is open”. In a new context the report “We should close the back door” resembles the original more closely than a word-for-word repetition does (pp. 73–74).
- Faithfulness as optimal resemblance (pp. 162–164). “In interpretive use the principle of relevance comes across as a presumption of optimal resemblance”: what the reporter conveys is presumed to resemble the original, and to do so with adequate contextual effects and no gratuitous processing effort. The translator’s goal is “to produce a receptor language text such that its interpretation resembles the interpretation of the original in those assumptions that make it adequately relevant to the receptor language audience”, expressed so that “these assumptions are recoverable by the receptors without unnecessary processing effort” (p. 164).
- Closer is better, other things equal (pp. 178–180). Of Knight’s rhyming translation of Morgenstern against four alternatives, “the more closely it resembles the original, the better” when processing cost is equal (p. 179). Where more resemblance costs relevance, the less resembling rendering “will usually be the one required for successful communication”. Scott-Moncrieff’s “thou” for Stendhal’s “tu” kept the pronoun’s number and lost the intimacy it implied. To English readers it conveyed archaic or church language, at extra processing cost (pp. 165–170, 179–180).
- Translation principles are applications of relevance to audiences (pp. 181–190). Beekman and Callow’s guidelines, Newmark’s rules for text types, Namy’s for interpreters and Savory’s twelve contradictory principles are each the principle of relevance applied to a class of audience. They are resolved by qualifying each “when required for consistency with the principle of relevance” (p. 190). The account is “context-specific in that the search for consistency with the principle of relevance always brings in the particular cognitive environment of the audience addressed. All this is achieved without recourse to typologies of texts, communication acts and the like” (p. 190).
- Direct translation by analogy with direct quotation (pp. 197–202). Indirect quotation depends on resemblance in cognitive effects, direct quotation on resemblance in linguistic properties (p. 198). Linguistic properties cannot be kept across languages, but “the point of preserving stylistic properties lies not in their intrinsic value, but rather in the fact that they provide clues that guide the audience to the interpretation intended by the communicator” (p. 199). Silt’i clefting can do what English focal stress does (pp. 200–201). Direct translation “calls for the preservation of all communicative clues” and lets the receptors reach the intended interpretation “provided they used the contextual assumptions envisaged by the original author” (p. 201).
- Communicative clues, by source (chapter 6). Clues come from the semantic representation (pp. 202–213), from syntax (Dickens’s juxtaposed clauses carry an echoic, ironic reading that conjoined clauses lose, pp. 215–217), from proper names (pp. 222–224), from pragmatic connectives (an English translation of Hebel turns an anaphoric German so into a consequential English so, pp. 224–229), from formulaic expressions and proverbs, from onomatopoeia, from register, and from rhyme and rhythm (pp. 229–246). A formulaic expression is translated by one that carries the same encyclopaedic information: the Amharic greeting, literally “May he give (you) health on my behalf”, by “hello”, and German Frisch gestrichen by “Wet paint” (pp. 231–235). Explicating weak implicatures is not direct translation, because those implicatures depend on context (p. 246).
- Direct translation defined (p. 254). “A receptor language utterance is a direct translation of a source language utterance if and only if it purports to interpretively resemble the original completely in the cognitive environment envisaged for the original.” Whether two utterances share their clues is checked “by checking whether they give rise to the same interpretation when processed in the same cognitive environment” (p. 253). So direct translation depends on interpretive use, and “communicative clue” has no theoretical standing of its own (p. 255). Direct and indirect translation are one continuum of interpretive resemblance, of which complete resemblance is the limit (pp. 254–255).
- Against equivalence (pp. 22–27). “Function” is now the most widely accepted frame for translation equivalence (Levý, House, de Waard and Nida) (pp. 22–23). On House’s model it supports comparison but not evaluation: “the notion of ‘equivalence’ itself may not be truly evaluative in nature but merely comparative” (p. 27).
- Against dynamic equivalence (pp. 122–129). Nida and Taber aim at the original’s “message”, but their constraint to “linguistic translation” forbids supplying the contextual assumptions or implications that receptors lack (pp. 123–124). Achieving “a high degree of equivalence of response” may be possible in near-primary communication situations “but it seems unrealistic for secondary communication situations with significant differences in cognitive environment” (pp. 126–127).
Claims
Not filled: the note is Skimmed.
Concepts
- Interpretive use: entertaining or uttering a representation in virtue of its resemblance to another representation, not as true of a state of affairs (p. 60).
- Interpretive resemblance: between propositional forms, shared analytic and contextual implications in a context (p. 57). Between utterances, overlap of the explicatures and implicatures each conveys in its own cognitive environment (pp. 74–75).
- Optimal resemblance: the form the presumption of relevance takes in interpretive use; Gutt’s notion of faithfulness (p. 162).
- Primary and secondary communication situations: whether the audience uses the contextual assumptions the communicator envisaged (pp. 118–120).
- Communicative clue: a property of the original stimulus that guides the audience to the intended interpretation (p. 199). It is defined by its effect, and derived from interpretive use (p. 255).
- Direct and indirect translation: complete interpretive resemblance in the original’s envisaged cognitive environment, against resemblance in relevant respects for the receptors’ (pp. 202, 254).
Connections
It applies Sperber and Wilson (1986) and Wilson and Sperber (1988), under Wilson’s supervision. Its foils are the equivalence theories: House’s functional equivalence and covert translation, Levý’s functional hierarchies, Nida and Taber’s dynamic equivalence and Beekman and Callow’s idiomatic translation. Of these the record holds Nida unread (LIT-791) and now Nida’s “Principles of Correspondence” as reprinted (LIT-1864, read alongside on 2026-10-30). Nord’s functionalism (LIT-829, unread) is not discussed in the pages read. The 1991 book (LIT-784) grew from this thesis.
Bearing on the record
- CLAIM-501: met in substance, by the thesis. Its defeat condition is
“a reading of Gutt on which interpretive resemblance is judged by
closeness of wording or of propositional content”. The thesis says the
opposite in so many words. Resemblance between utterances is overlap in
the explicatures and implicatures each communicates in its own context
(pp. 74–75), and “identity in propositional form … is not a sufficient
condition for identity in their interpretations” (p. 72). It counts
irony (pp. 215–217), intimacy (pp. 165–167), register and formality
(pp. 238–240) and the force of a greeting (pp. 229–232) as parts of the
interpretation a translation is to keep, and these are of the kind the
line calls stance and footing. Even direct translation, the strictest
notion, fixes its clues by their effect on interpretation, not their
form (pp. 253–255). So on this reading the account predicts that judged
pragmatic observables beat text similarity, and the claim’s
extrapolation holds. Three cautions for the status review:
- the thesis makes no prediction about human fidelity ratings, so the prediction is still the record’s inference from the account;
- fidelity is relative to a cognitive environment, the receptors’ for indirect translation and the original’s for direct, so raters’ context enters what the account predicts;
- this is the 1989 thesis, and the book (LIT-784) stays unread.
- CLAIM-205: partly, for its relevance-theoretic half. The claim says a relevance-theoretic account “need not be committed to only one context or a static stance profile”. For indirect translation, which Gutt calls the general case, the thesis agrees outright. Faithfulness is “determined for each occasion by the principle of relevance and the cognitive environment mutually shared” (p. 63). The account is “context-specific … without recourse to typologies” (p. 190), and the same original calls for different translations for different audiences (pp. 161–164, 181–190). Direct translation does fix one context, the one envisaged for the original (p. 254). The thesis proposes no operationalization, so it does not bear on the “rich contextual baseline” CLAIM-205 proposes.
- Not for the anthology. The thesis carries no instruction for machine-learning practice. Its footnote on simulating translation in artificial intelligence (p. 256, n. 2) is a remark about what such work would need.
Limitations
- Skimmed, not read. Chapter 3 (covert translation), most of chapter 4 and the second half of chapter 7 (pp. 258–298, on using the original cognitive environment, partial resemblance and the risks of failure) were not read. The account of when a translation succeeds is in those pages.
- The thesis, not the book. The 1991 book and its 2000 second edition revise this text. Nothing here says how.
- OCR. The text layer has errors (for example “Lev?” and “LevI” for Levý). Quotations were checked against the sense, but not against the page images.
- No empirical test. The account explains translators’ and critics’ judgements after the fact. The thesis tests none of its predictions on readers.
Open questions
- Does the 1991 book keep the definition of resemblance between utterances as overlap of communicated assumptions, each in its own context (pp. 74–75), and the definition of direct translation (p. 254)?
- How would a rater’s fidelity judgement be indexed to a cognitive environment, as the account requires, in a study of the kind the pragmatic-transport line proposes?