Presently available test statistics for the Rasch model are shown to be insensitive to violations of the assumption of test unidimensionality. Two new statistics are presented. One is similar to available statistics, but with some improvements; the other addresses the problem of insensitivity to unidimensionality. (Author/JKS)