This study attempted to determine the personal and workplace variables which predict a special educator's decision to stay, transfer, or leave the classroom. A stratified random sample of 1,576 special education teachers in Florida was surveyed, resulting in 1,152 usable responses. Two models were estimated with forward stepwise logistic regression methods: the first to differentiate leavers (those no longer teaching in special education) from stayers and transfers, the second to differentiate stayers (teaching in the same school) from transfers (still teaching in special education but in a different school). For the first model, both current certification status and stress were significant predictors of likelihood of leaving special education. An inappropriately certified teacher with high stress had the highest probability of leaving (.43) while an appropriately certified teacher with little stress had the lowest probability of leaving (.11). For the second model, both climate and age were significant predictors of staying or transfer status. The probability of transferring increased with decreases in age or climate. Results raise questions concerning the practice of placing uncertified teachers in special education classrooms, particularly classroom situations where teachers are likely to experience high levels of stress. Attached tables list the independent variables used in the analyses, examples of questions and coding, and probabilities of leaving special education (by certification status and stress levels) or of transferring to another school (by age and climate). (DB)