Compensation plays an important role in advanced transaction models, cooperative work and workflow systems. A schema designer is typically required to supply for each transaction another transaction to semantically undo the effects of . Little attention has been paid to the verification of the desirable properties of such operations, however. This paper demonstrates the use of a higher-order logic theorem prover for verifying that compensating transactions return a database to its original state. It is shown how an OODB schema is translated to the language of the theorem prover so that proofs can be performed on the compensating transactions.