The effects of four training procedures on the match-to-sample performance of severely and profoundly mentally impaired students