translation-universals-and-the-third-code

name	translation-universals-and-the-third-code
kind	skill
status	ready
description	Review claims about translation universals (explicitation, simplification, normalisation) and the third code — checking they are treated as measurable hypotheses, not presumed features, and kept distinct from translationese used as a quality proxy. Use when a universals or third-code claim needs method and faithfulness review; corpus design for such a study goes to corpus-design-and-methodology.
provenance	{"principles":["P001","P079","P084","P114","P115","P121","P139","P143"],"claims":["C00490","C00491","C00492","C00494","C00495","C00496","C00497","C00498","C00499","C00501","C00502","C00503","C00504","C00509","C00515","C00543"],"evidence":[],"source_anchors":[],"authored_from_digest":"672fb77344abd286aeaa681de48f352180800807d13ff4fdac1706789707d936"}

Translation Universals And The Third Code

Purpose

This skill reviews claims about translation universals — explicitation, simplification, normalisation, and the distinctive distribution that forms the 'third code' — and checks that they are handled as hypotheses to be measured, not features presumed to hold. It requires the constructs to be operationalised distinctly, the third code to be kept apart from translationese, and universals to be scoped as tendencies that a language-pair or culture-specific explanation might beat.

When to use

• A study asserts or relies on a translation universal (explicitation, simplification, normalisation) and the claim needs testing against evidence rather than assumption.
• Explicitation, simplification, and normalisation are being treated as interchangeable labels instead of distinct, measurable constructs.
• A 'third code' finding is being conflated with translationese, or read as a quality defect rather than a describable systemic feature.
• A universal is being generalised beyond the corpus that produced it, or reported more strongly than its sample supports.

Procedure

1. Probe for the recurring candidate translation universals — a rise in explicitness (explicitation), disambiguation and simplification, a preference for conventional grammaticality, avoidance of source-text repetition, exaggeration of target-language features, and a distinctive distribution of common features — as hypotheses to be measured against evidence, without presuming any of them holds in the corpus under review (P001).
2. Operationalise explicitation, simplification, and normalisation as distinct measurable constructs instead of treating them as interchangeable labels (P079).
3. Keep universals and norms distinct: universal features arise from constraints inherent in the translation process and are taken to be culture-invariant, whereas norms occur consistently only within a particular socio-cultural and historical context and vary across languages and cultures — flag any move that generalises a culture-specific norm into a universal (P084).
4. Treat a distinctive translated-language profile non-evaluatively unless the evidence shows actual translator incompetence or error (P114).
5. Scope any proposed translation universal as a tendency, and check whether a language-pair- or culture-specific explanation fits the data better before accepting the universal reading (P115).
6. To assess whether a claimed universal is properly isolated, check that the corpus assembles texts translated into one target language from a variety of source languages, keeps only the patterns that recur across source languages and are absent or rarer in original target-language text, and cross-validates against corpora translated into other target languages before claiming universality (P121). This monolingual-comparable design isolates universals visible against original target text (e.g. simplification, normalisation, conventional grammaticality); but a candidate defined relative to the source — explicitation is a marked rise in explicitness against the source text (P001) — additionally needs source–target (parallel) comparison, so route explicitation-type claims to parallel-corpus evidence, not the comparable design alone.
7. Distinguish the third code from translationese: treat the third code as a systematic distribution of features arising from the confrontation of source and target codes that sets a translation apart from both its source and original target texts, and reserve the label translationese for cases where an unusual distribution is clearly the result of the translator's inexperience or lack of target-language competence — do not treat every distinctive distribution as translationese, and do not attribute a systemic pattern to individual incompetence or vice versa (P139). Name which sense of "translationese" is in play: an automatic "translationese" classifier or corpus "translationese score" measures a statistically distinct translated-vs-original profile — that is the third code (P139) or normal translated-language patterning (P114), not by itself the competence-caused deviation P139 reserves the label for, and not a quality verdict (P002).
8. Generalise a simplification hypothesis only after it has been tested across genres, language pairs, and interpreted texts, not from early or single-design evidence (P143).

Inputs

• The universals/third-code claim under review, the corpus it rests on, and how each construct was measured.
• The reasoning offered for the decision under review: the corpus, the orientation, the brief, and any quality claim made.

Output

Per finding: flaw, principle, correction, trade-off, next step, highest-impact first — this skill reviews, it does not produce the translation or decide publication.

Anti-patterns to flag

• Universal-vs-norm mixup: calling a culture- or period-bound preference a "universal" because it looks constraint-like. The separating test — does it hold outside this culture and period? — has not been run; untested, it is at most a candidate norm (P084).
• Confirmation without a rival test: treating supporting evidence for a universal as sufficient without checking whether a language-pair- or culture-specific account predicts the same data equally well. Undefeated-by-a-rival is the bar; merely-consistent-with-the-data is not (P001, P115).
• Single-corpus overreach: calling a pattern universal from one target language or one source-language pairing. The threshold is recurrence across several source languages into the same target language, cross-validated against corpora into other target languages — short of that, it is a target-language- or pair-specific result, not a universal (P121).
• Blurred constructs: using "explicitation," "simplification," and "normalisation" interchangeably for one observed shift. Explicitation is source-relative (needs parallel, source–target evidence); simplification and normalisation are target-relative (measured against original target-language text) — each needs its own operational test before the label is assigned (P079).
• Mislabelled distribution: reading an automatic translationese-classifier flag or corpus "translationese score" as diagnosed translator incompetence. The score measures a distinct translated-vs-original profile — third code, or normal patterning (P114) — not the competence-caused deviation the term translationese is reserved for, and never a quality verdict (P139).
• Two related overreaches: reading a distinctive profile as a quality defect without evidence of actual error (P114), and generalising a simplification finding beyond the genre, language pair, or mode (written vs interpreted) it was actually measured in (P143).

References

See ../../references/translation-quality-principles-index.md for the full principle catalogue grouped by skill, and ../../references/translation-quality-evidence-notes.md for how these principles are grounded and kept faithful to the sources.

Provenance

Derived from P001, P079, P084, P114, P115, P121, P139, P143, grounded in the distillation-only sources Kruger et al. (eds.), Corpus-Based Translation Studies, and Baker, Corpus Linguistics and Translation Studies. The frontmatter provenance block lists the exact principle and claim ids, which resolve into principles/principles.yaml and analysis/claims.jsonl.