The first instinct of a careful editor is to make the text better. The polisher’s instinct is to make only the changes that were asked for, and then stop. When a Ukrainian translation comes through for polish, we get three things: the original English article, the current translation, and the reviewer’s feedback. The temptation is to read all three and start improving. Almost everywhere we look, we can find a sentence that could be tightened, a word choice that could be sharper, a paragraph break that might land better one line earlier. We resist that. The discipline of polish is the discipline of restraint.
Why every edit is a risk
Every change to a translation is an opportunity for a new error. When we touch a sentence, we have to re-verify that the new version preserves meaning, register, theological precision, and Ukrainian fluency. The original sentence already passed at least three filters: the translator wrote it, presumably with attention; the reviewer read it and chose not to flag it; the source author wrote it knowing what they wanted to say. Our edit replaces all three filters with one: our own judgment in a single pass.
That math should make us cautious. It usually does, but not always. Editors are trained to improve text, and there is a real pull to demonstrate that we read carefully by leaving evidence everywhere. The polisher has to fight that instinct hard enough to leave large sections of text completely alone, including some that we might privately think could be a little better.
A reviewer’s silence is data. If they read the entire article and only flagged seven sentences, those seven are the contract. Everything else has been ratified, even if not enthusiastically. Treating ratification as approval is uncomfortable for a perfectionist, but it is what makes polish a distinct step from rewriting.
The cost of touching what wasn’t flagged
The most expensive edits we make are not the ones the reviewer requested. They are the ones we make on our own initiative, the “while I’m in here” changes. Three of these in a single pass and we have introduced material the reviewer never saw, never approved, and won’t see again unless the article goes back through review.
This matters for two reasons. The first is mechanical: changes outside the reviewed scope can break things. A sentence we reshape might rely on a connective phrase that holds the paragraph together. A word we swap might be a quiet repetition that the author used deliberately. A clause we relocate might fix one rhythm and ruin another. Even small structural moves can pull on something we did not realize was load-bearing.
The second reason is procedural. A polish that touches a hundred lines is not really a polish. It is a quiet second translation, smuggled in under a different name. The pipeline assumes the reviewer’s scoring still applies after polish. If we have rewritten material the reviewer already approved, that assumption breaks. The next person downstream, usually a publisher who runs final checks before posting, is looking at material that was never reviewed in its current form.
The honest answer, when we notice an unflagged sentence we want to improve, is to leave a note. We can flag it back for the next reviewer to consider, or accept that it is good enough, or escalate for a re-review if it is genuinely poor. What we should not do is silently fix it and ship.
When minimal isn’t enough
The opposite failure mode also exists. Sometimes a reviewer’s feedback is too narrow. They flag a symptom without seeing the underlying problem. A flagged sentence may need fixing, but the fix might propagate. If we change a term in one place, we have to harmonize it everywhere. If we restructure a paragraph to address a flow issue, the next paragraph may now read awkwardly.
The rule we use here is: edits propagate only when consistency forces them. If the reviewer asks us to change a theological term in paragraph three because the context demands it, but the same term appears in paragraphs one and seven, we change it in all three places. Not because the reviewer flagged paragraphs one and seven, but because translation consistency is a constraint that overrides the flag-based scope.
When in doubt, the question we ask is whether a different reviewer, reading the polished version, would agree that our additional changes were forced by the flagged ones. If yes, the propagation is justified. If we are reaching to defend an edit, we pull it.
The diff is the deliverable
We have started to think of the diff as the actual product of a polish pass, not the polished text itself. The polished text is what gets published, but the diff is what we can defend. A tight diff, small and targeted, with each change traceable to a piece of feedback, is the signature of a polish done well. A sprawling diff, even one that improves the text in many small ways, is a sign that the polisher overstepped.
This framing changes how we work. Instead of reading the translation and looking for things to fix, we read the reviewer’s feedback and look for the smallest set of changes that addresses it. The text is not a draft we are improving. It is a near-final document we are correcting at specific points.
Restraint is a strange thing to take pride in. But the more polishes we do, the more we trust that less is harder than more, and that the discipline of touching little is most of the craft.