Glossary
You are retro-stalemate (also retropat) when you cannot retract any (legal) move.
This is the simple retro-equivalent to being stalemate in forward play. The big difference is that being retro-stalemate is much more difficult to recognize and may even be the whole point of some retro-problems. As a consequence, being retro-stalemate has not such a very formal definition, and rather means something like "it is obvious you have no legal retraction available".
Such a loose and subjective definition is not a real difficulty because the terminology is mainly used in the solutions of problems, where e.g. a given retro-maneuver "fails because it takes too much time and the other side will be retro-stalemate".
Here is an example problem, where the threat of retro-stalemate is always pending:
T. Volet 1st Prize Die Schwalbe, 1980
8+16. On which squares did captures occur and in what order?
Basically, in a "Help-retro-stalemate in n", Black and White retract n moves each, cooperating and trying to reach a position where Bl. has no legal retraction (Bl. is retro-stalemate).
Here is an example:
Charles Kemp and Thomas Dawson Fairy Chess Review, 1943
8+12. Help-retro-stalemate in 3
Solution: retract -1. Kf7xNe7 Nc8-e7 -2. Kg8xRf7 Rc7-f7 -3. O-O Rd7-c7 and Bl. has no legal retraction (because his O-O must be legal).
However, there is a logical difficulty with this stipulation: if, in the end position, Black has no legal retraction, then clearly the last Wh. retraction was not "legal"... So this raises the question of "What are the allowed retractions?" Perhaps a clear-cut definition would be what Henrik Juel suggested, namely to read "help-retro-stalemate in n" as "Find a position where Bl. is retro-stalemate, such that the diagram position can be reached in n forward moves". (Of course, we assume that the position where Bl. is retro-stalemate is legal in other aspects, i.e. assuming Wh. played last.)