diff --git a/docs/specs/voice-treatment-entry/spec.md b/docs/specs/voice-treatment-entry/spec.md
index bc3c176..f34c713 100644
--- a/docs/specs/voice-treatment-entry/spec.md
+++ b/docs/specs/voice-treatment-entry/spec.md
@@ -691,6 +691,18 @@ other type keeps a plain teeth row and has no prosthesis row at all. The merged
 The chart catches a misheard tooth number at a glance; the list confirms the material. Either
 one alone is the weaker check.
 
+> **Accepted consequence:** on a prosthesis dictation the teeth are reachable only through this
+> row. `voiceRowAvailability` sets `teeth: false` for a lab-dependent result, so no plain teeth
+> row exists and `selection.teeth` never becomes true. Untick the merged row — which releases the
+> decision-41 lock — then untick the type row, and the detail applies with no teeth at all.
+> `master` kept them, because the rows were separate.
+>
+> This is decision 39 doing what it was chosen for, and it follows from *the sheet is a contract*
+> below: the row that displayed those teeth was unticked, so applying them anyway would apply
+> something the clinician refused. The cost is real and accepted — a clinician who wants only the
+> teeth from a prosthesis dictation, intending to pick materials by hand, has to tap them in. If
+> that turns out to matter in use, the fix is a spec change, not a patch.
+
 ### The sheet previews the stack that will actually land
 
 The merged row builds its preview through `applyLeafToJobs` — the same function the manual