Investigators of individual differences in reading acquisition sometimes compare groups of different ages matched on some aspect of reading performance level. These reading-level-match (RL-match) designs can provide a convenient, powerful, and appropriate way to study atypical performance. However, their usefulness has been diminished by researchers' acceptance of myths about the designs' characteristics. RL-match designs have been used as if they (a) were more similar to true experiments than other correlational designs, (b) required only informal sampling procedures, (c) were improved by case-by-case matching, (d) made unequal groups equal, (e) served unique theoretical purposes, (f) created equality between groups in underlying processes whenever a performance-level match existed, (g) yielded some patterns of results that always were interpretable and others that never were interpretable, and (h) were equally valid across all possible matching criteria. Examples from the reading literature are given to illustrate the hazards of accepting these myths and to suggest more realistic alternatives. Relevant statistical and design principles are summarized.