To　explore　the　impacts　of　variables　on　receptor　model　results　in　source　apportionment　for　soil　pollutants,　the　sampling　data　set　of　soil　heavy　metals　in　the　middle　and　upper　reaches　of　Le＇an　River　was　used　as　the　input　data　set　for　the　typical　receptor　model　(PMF　model).　After　obtaining　the　results　of　basic　scenarios　by　PMF　model,　local　sensitivity　analysis　method　was　introduced　to　study　the　sensitivity　of　variables　on　PMF　diagnosis　and　source　identification.　The　six-factor　scenario　was　the　best　result　for　the　simulation　of　PMF　base　model,　Cu,　Mo,　Na2O,　As,　Mn　and　Cd　in　the　soil　were　the　sensitive　variables　and　also　the　main　loading　elements　in　each　factor　profile　(i.e.　the　typical　pollutants　of　each　source).　There　was　a　significant　difference　on　the　sensitivity　for　these　variables:　the　total　sensitivity　of　Cu　and　Mo　are　much　higher　than　that　of　the　other　variables,　reaching　12.1　and　8.2　respectively.　Therefore,　it　revealed　that　the　sensitive　variables　should　be　the　specific　pollutants　when　applying　the　receptor　model　for　source　apportionment,　and　the　data　quality　was　an　important　factors　affecting　of　the　reliability　of　source　apportionment.　©　2019,　Editorial　Board　of　China　Environmental　Science.　All　right　reserved.