I was recently asked about the difference of efficacy between Repeated Reading (RR) and the Iowa Reading Research's Varied Practice Reading sets. RR is a fluency exercise, in which teachers ask students to read the same text repeatedly. Students are often timed and encouraged to read as quickly as possible. RR is essentially a rote memorization teaching strategy for fluency. In many ways, RR could be called the current gold standard of fluency interventions. According to John Hattie, there are 159 studies to date on the topic, with an average effect size of .53, which is a moderate effect size. However, some meta-analyses have placed it at a much higher effect size. For example, the National Reading Panel meta-analysis found RR to have an effect size of .86. In my own research, I have generally speaking found RR studies to always have the highest effect sizes, when it comes to fluency interventions. I believe it is fair to say that RR typically is found to have a moderate to high effect size. 

 

That being said, the Iowa Reading Research Center (for who I have a lot of respect) has recently criticized the popular fluency intervention and stated “some forms of implementing this strategy can condition students to read in ways that are not authentic to how skilled readers approach texts.” as well as “Not only does the application to book reading lack an evidentiary basis, but there have been concerns raised about whether the gains students make in RR generalize or transfer to new passages” (IRRC, 2018). With this criticism in mind, they have developed their own fluency intervention, which they call Varied Practice Reading (VPR). Within a VPR protocol, students read three distinctively different texts that share 85% of the same words. Their hope is that this intervention builds fluency in a more authentic and adaptable way.

On the one hand, VPR does certainly sound a lot less boring for students, I do have some initial pauses for concerns. My biggest concern is the fact that the literature seems to show that the more repetitive a fluency intervention is, the more effective it is and VPR by definition is 25% less repetitive than RR. While RR meta-analysis has shown RR to have an effect size as low as .53, many studies of RR actually only have the students read the text two times. A meta-analysis of the topic by Therrion et al, in 2004, showed that having students with reading difficulties read the same text three times had an effect size of .95. This a large effect size and suggests that having students read the same text three times is a high yield strategy for fluency outcomes. Comparatively, Zimmermann et al, did a meta-analysis of non-repetitive reading fluency interventions in 2019, for students with reading difficulties, and their study found a cohen’s d effect size of .105. The Repeated Reading effect size in the 2004 Therrion et al, meta-analysis is more than 9 times higher.

The Iowa Reading Research Center recently conducted a study on the topic, in 2018, which I must admit was very well done. Their study had a sample of 827 4th grade students (which in my opinion is the ideal time to be conducting fluency interventions)  and directly compared repeated reading with varied reading. Each group had 30 sessions. At the end of the study, they found an effect size of .05 for VPR compared to RR and “concluded the Varied Practice group significantly outperformed the Repeated Reading group at posttest” (IRRC, 2018). 

 

However, I disagree with their conclusion. An effect size of .05 is very small. For comparison, if we look at John Hattie’s 2017 meta-analysis, he looks at 252 teaching influences and only 10 of those influences have effect sizes that small. If we look at the Cohens D guide to interpreting effect sizes (please see below), we can see that an effect size of .05 is classified as “weak”. I would have liked to have looked at their raw data, for both interventions; however, they only include their raw data for VPR and not for RR. That being said, if we calculate the effect size for VPR just based on the pre-test/post-test difference, rather than comparing it to RR, we get a super effect size of 2.01. Because the comparison effect size was so low, we have to conclude that both RR and VPR had extremely high effect sizes, if you only looked at their pre-test/post data. Ergo, I think it is likely logical to conclude from this study’s data, that both interventions are high yield strategies.