What it is
Two questions come before every other question about a fixed wording: is this stretch a unit at all, and how far does it run. Neither is answered by the string, because a sequence that recurs and one that happens twice look alike, and a unit carries no mark where it ends.
In practice both are settled by declaring a rule. The worked case is Pontrandolfo's judicial corpus, where a sequence is admitted as a phraseological unit only if it reaches five occurrences per two million words and appears in at least five different judgments, the second condition there to keep one judge's habits out of the inventory. What the same study does not do is fix a length. The shapes were set per category instead, from N+P and P+N+P prepositions to routine formulae of differing length. A declared rule can therefore settle unithood and leave extent open.
That gap is where this reference decides for itself. A combination that one term anchors belongs on the term record; a unit that no single term anchors belongs here, and extent separates them. No source consulted lays down a criterion dividing a polylexemic terminological unit from a specialized phraseological one, and the literature reports that the line cannot be drawn in principle, so the decision is attributed to nobody.
Where the fixity comes from
Nothing prescribes how far a unit runs. The warrant is attestation, and attestation is worth no more than the rule that admits a sequence in the first place.
- The corpus behind those thresholds holds 782 criminal judgments and 6,036,915 tokens in three subcorpora of roughly two million words each, in Spanish, Italian and English. Both conditions are stated rather than left to the reader, which is what makes the counts mean anything.
- What the thresholds establish is that a sequence recurs and is spread across texts. What they leave open is where it stops, and the study answers that once per category rather than once for all.
- An inventory is therefore an inventory of what a rule admitted. The doublet counts on the neighbouring record were produced under exactly these conditions, and a lower cut-off would return a different set.
- Granger and Paquot explain why that is not a defect: scope varies with the approach taken and the borders with neighbouring disciplines are fuzzy, so a declared rule beats an assumed boundary.
Consequences for translation
Deciding how far the unit runs is most of the work, and getting it wrong is invisible in the sentence where it happens.
- Cut the unit too short and the frame is recomposed at every occurrence, so one construction leaves the document in three shapes. Cut it too long and a free slot freezes with the fixed part, so a new article number produces a new translation.
- Extent is not preserved across a pair. One source word can face a three-word target unit and the reverse happens as often, so word counts and segment boundaries are evidence about the file and not about the unit.
- For a reviser the symptom is inconsistency across a document rather than a mistake in a sentence. Two renderings of one frame in one text is a finding about extent, worth recording even where each is defensible.
- A frequency figure travels no further than the rule that produced it, so quoting a count without its threshold turns a measurement into an assertion.
Examples
EN>NL
- Source: Notwithstanding paragraph 1
- Target: Niettegenstaande lid 1
- Comment: Rome I, Article 6(2). The unit stops before the slot: what is fixed is Niettegenstaande followed by whatever is being set aside. Take the whole opening as the unit and it has to be settled again at the next paragraph number.
EN>ES
- Source: Notwithstanding paragraph 1
- Target: No obstante lo dispuesto en el apartado 1
- Comment: Same instrument, same provision. The Spanish carries a nominalization, lo dispuesto en, with nothing corresponding to it in the English, so the target unit's extent cannot be recovered from the source. Composing from the English gives No obstante el apartado 1, which is Spanish and is not what the regulation says.
EN>DE
- Source: Notwithstanding paragraph 1
- Target: Ungeachtet des Absatzes 1
- Comment: The same provision again. The German connector is one word and governs the genitive, so the unit reaches past its own edge and fixes what fills the slot: des Absatzes 1 and not der Absatz 1. Extent is a span of words plus a demand on what follows.
References
- Granger, S. and Paquot, M. (2008). "Disentangling the phraseological web." In S. Granger and F. Meunier (eds), Phraseology: An interdisciplinary perspective, 27 to 49. Amsterdam: John Benjamins.
- Pontrandolfo, G. (2015). "Investigating judicial phraseology with COSPE: a contrastive corpus-based study." In C. Fantinuoli and F. Zanettin (eds), New directions in corpus-based translation studies, 137 to 160. Berlin: Language Science Press, at pp. 141, 144 and 146.
- Regulation (EC) No 593/2008 of the European Parliament and of the Council of 17 June 2008 on the law applicable to contractual obligations (Rome I), Article 6(2), English, Dutch, Spanish and German language versions.