Demolishing a Straw Man: Should We Teach Fluency?

What’s the relationship between oral reading fluency and reading comprehension? Does fluency instruction automatically lead to comprehension? Are reading comprehension and fluency independent processes? Reasonable questions like those abound about this aspect of learning to read. However, if you are seeking answers to such questions in the recent article on fluency and reading comprehension that appears in The Reading Teacher (RT, March 2009, vol. 62 (6), pp. 512-521), don’t bother.

  The problem is that the authors, M. D. Applegate, A. J. Applegate, and V. B. Modla, have made up a straw man (um, straw person) argument implying that someone out there supposes oral reading fluency alone is the necessary and sufficient condition for high-level reading comprehension. Their contention is that there is this (apparently stupid) group of teachers or scholars who believe that if kids can read text aloud quickly, accurately, and with proper expression, then they’ll comprehend anything no matter what their knowledge of vocabulary or the world or their other intellectual skills and predilections.

  Surprise, surprise, the authors manage to show that oral reading fluency alone is insufficient to guarantee reading comprehension. This isn’t especially difficult to show, but this study is poorly reported so the findings aren’t particularly persuasive or informative.

  One big problem with the data analysis: student selection is poorly executed, leading to some rather circular analysis (I got dizzy anyway). They only included students with a score of 16 or higher on their fluency rubric, but provided no information on the reliability or validity of this, and they didn’t indicate who administered the test and how inter-rater reliability was handled, nor did they provide any information about why 16 is a good level of fluency to select. Reliability and validity data for setting a specific cut score on a test are different than the reliability and validity of an overall test—this is particularly true in an instance like this where the authors tried to identify students with a narrow range of test performance. Why should we think that a 16 on this rubric is the level of fluency that students should attain?

  On other tests, fluency criteria are usually set against some outcome measures—typically reading comprehension. Here’s where it gets circular: a level of fluency on a test is usually set to make it possible to determine if, for most kids, fluency will be sufficient to allow reading comprehension success. That fluency cut is never set at a point where 100% of the kids at that level comprehend well (test data are too messy for that). So the authors have set a cut score that does not match with all kids being able to comprehend, and then they set out to prove that not all kids who reach this cut score can comprehend.

  The study divides students into three levels of comprehension but not in any clear way (and neither standard deviations nor standard errors of measurement for their tests are provided), so it is impossible to know how different these groups really are or whether other ways of dividing the data would help us to understand things any better. The study tries to analyze different question types, but without any discussion of the reliability or validity of these parts of the tests (you cannot simply divide up a test into parts and assume that the parts remain valid and reliable).

  Not surprisingly, they found that students who were fluent could for the most part answer questions accurately about text if the answers to the questions had been explicitly stated in text (for 143 of the 171 kids tested). They even found that more than two-thirds of these fluent kids could do high level comprehension tasks such as interpretation and critical response, but that the rest struggled with these aspects of higher-order thinking despite their fluency.

  In other words, for the most part, fluent readers were able to accomplish average levels of reading comprehension when the focus was on remembering or interpreting information in the text, but about 15% of these students had difficulty when asked to use the information in text to do other things. The accuracy of these figures is impossible to verify given the reporting flaws, but whether these numbers are accurate or not the authors are correct: fluency alone will not guarantee comprehension.

  The highest correlation I’ve seen between oral reading fluency and reading comprehension is about .85 (which is very high), but the correlations vary a bit and probably average a bit lower than that. But even a correlation that high only indicates that oral reading fluency predicts about 72% of the variance in reading comprehension scores. In the simplest theory possible, the one that claims that oral reading fluency simply and directly causes reading comprehension, that still leaves more than a quarter of the variation in reading comprehension performance to other variables. The imperfect correlation of fluency and comprehension alone tells me that some fluent readers will not comprehend very well—and that some disfluent ones will manage to understand text anyway. The high correlation on the other hand (especially from experimental studies in which children’s fluency is improved through some kind of intervention that leads to improved reading comprehension) indicates the importance of teaching fluency.

  The National Reading Panel reviewed the experimental studies on fluency instruction and found that, in 15 of 16 studies, fluency instruction led to improved reading comprehension. The one study that did not find comprehension gains from fluency instruction is the one that proves that fluency alone does not guarantee comprehension improvement. The other 15 studies are the ones that suggest it would be wise to make sure that kids are fluent.

  While teachers should know fluency instruction alone will not guarantee comprehension gains (nor will instruction in any other aspect of reading), these authors might want to spend some time in middle schools and high schools in cities like Cleveland and Chicago where large numbers and percentages of students lack the fluency levels commensurate with being able to get an author’s message or to do text level interpretation of text successfully. Teach fluency, but by all means, don’t just teach fluency.