This paper considers (1) the requirements imposed on data in order to conform to the Rasch model, (2) some common sources of departure from the model, and (3) a procedure for recognizing the occurrence of these disturbances. The specific disturbances discussed are guessing, practice, speededness, and bias. The observed characteristic curve for each situation is compared with the true logistic ogive and with the item characteristic curve that the Rasch estimation procedure would fit to such data. A convenient, interpertable form of residual between model and data is suggested. (Author)