The limitations and advantages of any given measure of performance used to classify students as good or poor readers have not been investigated thoroughly. A study was conducted to determine what happens on several dependent measures of comprehension when the same students are separated into good and poor readers based on different criteria. Thirty-eight sixth grade students, 18 of whom were designated by their teachers as poor readers, participated in the study. The students were administered the Gates-MacGinitie Reading Comprehension Test and tests of speed and accuracy in word recognition for high and low frequency words. The students read and listened to passages, two designed to be easy and two to be difficult, that had been equated for difficulty. Following the reading and listening exercises, the students were given free recall tests and multiple choice tests of literal comprehension. The results revealed that 40% of the students shifted categories from good to poor readers or vice versa, depending upon which criteria were used for classification. The findings suggest that educators and researchers should be wary of relying on any one classification system. (FL)