In this paper, the detection of response patterns aberrant from the Rasch model is considered. For this purpose, a new person fit index, recently developed by I. W. Molenaar (1987) and an iterative estimation procedure are used in a simulation study of Rasch model data mixed with aberrant data. Three kinds of aberrant response behavior are considered: (1) guessing to complete the test; (2) guessing in accordance with the three-parameter logistic model; and (3) responding with different abilities on different subsets of items. The power in detecting such aberrants is evaluated in two cases: when item difficulties are known; and when item difficulties are estimated from the data, including aberrants. The results reveal that, in the latter case, the estimates of the model parameters are biased and that the power of the index, as a consequence, is reduced. It is shown that, by using an iterative procedure, the recovery of the power of the index to the level obtained by known item difficulties is achieved. Furthermore, depending on the type of aberrance, a considerable reduction of the bias in the model parameters is possible. Finally, it is confirmed that this new index allows detection of aberrant response patterns with better statistical properties than former person fit indices. Three data tables and eight graphs are included. (Author/TJH)