One of the most important findings to emerge from recent reading comprehension research is that there are large differences between tests in what they assess--specifically, the extent to which performance depends on word recognition versus listening comprehension skills. Because this research used ordinary least squares regression, it is not clear that the findings apply similarly to poor and good readers. The current study uses quantile regression to assess whether there might be differences within tests in the relative contributions of component skills as a function of performance level. There were 834 individuals (ages 8-18) who took 5 reading comprehension tests. Quantile regression showed that, for 3 of 5 tests, the contributions of word recognition and listening comprehension vary as a function of reading comprehension skill. These quantile differences hold across both younger and older readers. Our findings show that what skills a test assesses vary not only with the specific test used but also with how well the person performs.