Abstract.A strategic internal auditing model is developed to analyze the use of discovery sampling to deter or detect abstraction of assets by an auditee. The analysis develops a game in which the auditee chooses the fraud level and an auditor chooses HB effort level (sample size). The auditee seeks to maximize expected successfui fraud net of a fixed sanction for detected fraud. The auditor seeks to minimize expected costs from sampling and fraud losses. Both simultaneous play aad commitment versions of the game are analyzed. For each version, pure strategies are optimal. In comparison with simultaneous play, the commitment version equilibriam results in greater audit effort and less fraud by the aaditee. Coraparative statics showing effects of sanction level and recovery rates are derived. For simultaneous play, optimal monitoring effort decreases with the sanctioji level and increases with the recovery rate; for the commitment version, these effects are reversed. The analysis demonstrates the sensitivity of results to the audit effort commuaicatioB arrangement and to the specific audit objective of fraud detectioB.