Last year, I wrote a non peer-reviewed meta-analysis on the topic of decodable texts and admittedly, it seems this was my most controversial article. The results of this meta-analysis did not show a statistically significant benefit and most of my readers identify with the Science of Reading community (as do I), in which this pedagogical concept has very strong support. However, this meta-analysis was one of my first and admittedly, my skills in this regard have improved greatly over the last year. With this in mind, I redid the meta-analysis, to see if I would get any different results. 

 

Previously, I calculated the effect sizes with a Cohen’s average calculation. Whereas this time, I used Cohen’s d for the studies with samples above 50 and Hedge’s g for the studies with samples below 50. I searched the Education Source data-base, the Sage Pub database, and the first several pages of google. I also reviewed the studies I used last time. This resulted in me removing the 2001, J Hoffman study, as it used a correlation effect size and adding a quasi experimental study by Wiley Blevins, conducted in 2000.  

 

When I last looked at this issue, one criticism I received was that in order to properly examine the efficacy of decodable texts, we need studies that use phonics in both the treatment and control group, so as not to conflate the effects. I agree with this criticism; however, I did not explore the pedagogical methods used by authors in my previous analysis. This time, I controlled for this data, so we could see specifically what is the impact for adding decodables to a phonics program. 

 

In total I was able to find 7 relevant studies, 2 of which I was not able to find comparable effect sizes for. This makes this meta-analysis in general very weak, as not only is it not peer-reviewed, it is only drawing its effects from a handful of studies. That being said, this is likely why (to the best of my knowledge) there exists no peer-reviewed meta-analyses on the topic), as there is very little research in this regard. While this is a weakness for my analysis, it is also a weakness in the case for decodables as being part of the science of reading. I would argue that in order for us to call something science, it should have strong scientific evidence, IE multiple high quality studies showing the same effect. 

​
 

Studies Included: 

Juel 1990: This study was a quasi-experimental study with 93 grade 1 students. The teaching methodologies were not sufficiently detailed for me to know if both groups used phonics. This study did not have sufficient data for me to draw an effect size. However, the results appear to be negligible Indeed, the original authors wrote “In addition, although both groups appeared to finish first grade with equal amounts of letter-sound correspondence knowledge, children in the Economy series seemed to utilize it more in word identification [.....]The interpretation of the results of this study do not constitute advocacy of any one specific approach to beginning reading instruction”. 

​

Felton 1993: This study was an RCT study with 81 dyslexic kindergarten students. It was the only study I found with significantly positive results. However, this study also used a control group with Balanced Literacy instruction and an experiment group with phonics instruction. For this reason, I worry that this study is confounding the effects of phonics and decodables. As I would expect a phonics kindergarten group to outperform a Balanced Literacy kindergarten group, regardless of the type of readers used. This study compared Phonics and decodables to a business as usual (BAU) group, and to a Balanced Literacy Group. The effect size in comparison to the BAU group was .74. The effect size in comparison to the Balanced Literacy group was .32. If we exclude this study the mean result of my meta-analysis would be -.14, meaning the majority of my found mean overall positive effect comes from this one study alone. For these reasons, I think the study could be considered an outlier. 

 

Blevins 2000: This study was a quasi-experimental study, with 101 grade 1 students. This study was one of 2 studies that specifically compared a phonics group to a phonics group with decodable readers. For this reason, I would argue that this is one of the most important studies on the topic. However, it was not peer-reviewed and was not an RCT, making the results less scientifically valid. The study found a mean effect size of .16, which is a positive, but statistically insignificant result.  

​

Jenkins 2004: This study was a RCT study, with 95 grade 1 students. This study was the second study that specifically compared a phonics group to a phonics group with decodables. However, as this study was peer-reviewed and RCT, I would argue that it is the most valuable study on the topic. This study showed a mean result of .03, which suggests a very small but statistically insignificant impact. 

​

Mesmer 2005: This study was a quasi-experimental study, with 22 grade 1 students. The results of this study were -.04. However, the study did not specify the instructional methodology of the groups. For this reason, it is hard to make extrapolations based on the results. 

​

Price‑Mohr 2019: This study was a RCT study, with 36 kindergarten students. This study did not specify the teaching methodology of the groups. It found an extremely negative mean effect size of .73. It is very hard to make extrapolations about this study, as the teaching methodologies were not sufficiently specified. Moreover, this effect size was clearly an outlier.   

 

Results: