So over the last few weeks, I have been observing intense debates over the issue of “advanced phonemic awareness” and the efficacy of extended phonemic awareness. This debate appears to be in my opinion taking place on two levels. On the one hand, Dr. Brown, Dr. Shanahan, the esteemed Stephen Parker, and Dr. Kilpatrick, have had an academically rigorous, and polite discussion on the merits of the research for “Advanced phonemic awareness”. However, on the other hand, I have found this debate on social media to be less rigorous and nuanced. 

 

While I wanted to weigh in on this debate, I also thought this debate could highlight another issue I have noticed within the “Science of Reading” community. Within science, it is usually considered best practice to avoid speaking in absolutes and to instead speak in degrees of probability, based on the available evidence. And while, the prestigious researchers who have been engaging in this debate on an academic level, have been arguing with nuance, civility, and scientific rigor, I have noticed quite a few people on social media, making very confident, absolutist statements, on both sides, with little to no evidence. 

 

Over the last few years of researching and writing for PNG, I have begun to feel a part of the SOR community. However, I have noticed the trend of proclaimed members, sometimes viewing that community, as more of an identity, than a concept. If we want to call ourselves scientific, we need to be open to opposing views, adjust our opinions to the ever evolving evidence, and communicate with scientific nuance. 

 

That being said, I completely understand how people new to the science of reading can be swept up in their new found convictions. When I first started writing about the science of education, I was completely swept up in the Dunning Kruger effect, I knew too little, to know how little I knew and I wrote with an embarrassingly high level of confidence. I was completely swept with excitement and new found conviction. But truthfully, I shudder when I read some of the things I first wrote for my blog. Years now into this process, after reading hundreds of studies on the topic of education science, and writing hundreds of pages myself, I feel very humbled. I try my best to write with humility, to be open to criticism, and to speak in degrees of probability, as I know mistakes are inevitable. With this off-topic diatribe out of the way, I do want to say it is with extra humility that I write about this debate today, as I fully recognize that the actual scholars already having the debate are far more qualified, experienced, and knowledgeable than myself. 

 

That all being said, as I see it, there are four areas of debate within the community over phonemic awareness instruction:

 

1. Should we provide all students with “advanced phonemic awareness instruction”.

2. Should we teach sounds before we teach letters?

3. Should we provide older students with phonemic awareness?

4. Do dyslexic students have a greater need for extended phonemic awareness instruction?

 

1) Should we teach advanced phonemic awareness to all students?

 

Before attempting to answer this I want to point out that both Dr. Shanahan, and Dr. Kilpatrick currently reject the use of this term. Well Dr. Kilpatrick popularized the term, he has stated that he used the term interchangeably, with phonemic awareness proficiency and did not mean it as a type of instruction. However, other scholars have used this term to represent deletion and substitution instruction. The NPR meta-analysis, which is quite possibly the largest meta-analysis of the topic, looked at three different types of phonemic awareness instruction: segmenting, blending, and deletion. It found similarly high results for segmenting and deletion, with slightly lower results for blending. 

 

The NRP meta-analysis also gave us several other important clues for this debate. The NRP showed us that PA instruction yielded high results for Pre-Kindergarten, Kindergarten, and at-risk students. The meta-analysis showed moderate results for grade 1 and reading disabled students. The meta-analysis showed low  results for students above grade 1. Interestingly, it also showed results that diminished after 9 hours of instruction. (If you want to read more about the NRP meta-analysis of phonemic awareness Instruction, check out my more detailed summary of it here: https://www.pedagogynongrata.com/phonemic-awareness ). 

 

With the NRP results in mind, I think we can say with reasonable confidence that phonemic awareness instruction should not go past grade 2 for most students and that the scope of phonemic awareness should likely be limited. This is not to diminish the value of phonemic awareness instruction as indeed, the NRP meta-analysis actually showed higher impacts for it than phonics, for Kindergarten and Pre-Kindergarten students. However, it does appear that the value of this instruction quickly diminishes. I think this might be surprising to some members of the SOR community, but it appears that both Dr. Kilpatrick, and Dr. Shanahan actually agreed with this point. Indeed, Dr. Kilpatrick has made it clear in the recent “debate” that he does not believe extended PA instruction is necessary for any students except dyslexic students. 

 

2. Should we teach sounds before we teach letters?

 

I think this is the question I see most often discussed on social media. I often see people making strong claims for either sounds first or letters first. The NRP meta-analysis did examine PA with and without written instruction. Their results showed higher outcomes for instruction with letters, with a mean effect size of .88 for with letters and .55 for without.  I think realistically this is unsurprising though, considering that reading instruction with letters is more specific than reading instruction without letters. That being said, the studies included may not have made for a fair comparison. As ideally to study this we would want treatment groups that do PA instruction without letters, followed up by phonics instruction that included letters with a control group that used only phonics. To the best of my knowledge, this is not the comparison that was looked at in the NRP. Additionally, I want to make the point that the effect size of .55 does show a moderately high impact, indeed the mean effect size found for morphology in my secondary meta-analysis of the subject was .51. I often see people claiming students cannot learn to read properly, if they learn sounds first, or if they learn letters first. The NRP meta-analysis results do not support either argument, as students can clearly learn how to read either way. 

Dr. Kilpatrick also highlights that the studies in the NRP on PA instruction without letters, were specifically on whole class instruction, which he argues is different than what he is arguing for. Dr. Kilpatrick is arguing that students with Dyslexia, benefit from extended PA without letters, Indeed, Dr. Kilpatrick cites 6 papers that specifically look at extended PA instruction for dyslexic students. To attempt and assess the validity of Dr. Kilpatrick’s claim, I decided to do my own non-peer reviewed meta-analysis of the papers he listed. For studies with sample sizes below 50, I used a Hedge's g calculation. For studies with a sample size above 50, I used a Cohen's d calculation. 

 

Studies Included: 

Torgesen 1999: This paper was an RCT study of 180 kindergarten students, suspected of having Dyslexia. The treatment group was given mostly PA instruction, whereas the control group was given analytic phonics instruction. I found a mean effect size in favor of the control of .72, which is high.

​

Torgesen 2001: This paper was an RCT study of 60 students in grades 3-5, suspected of having dyslexia. The treatment group was given mostly PA instruction, whereas the control group was given analytic phonics instruction. I found a mean effect size of .23. which is barely significant.

​

Torgesen 2010: This paper was an RCT study of 108 grades 1-3 students that were suspected of having dyslexia. Both the control group and the treatment group included PA instruction and letters instruction. However, the treatment group received substantially more PA and less letters instruction. I found a mean effect size of .84, which is high. 

 

Wise 1999: This paper was an RCT study of 122 students in grades 2-5, suspected of having dyslexia. The control group received phonics instruction. I found a mean effect size of .75, which is high.

 

Torgesen 2003: Despite my best efforts, I could not locate this paper cited by Dr. Kilpatrick. I searched the ERIC database, Education Source, Sage Portal, and Google. 

​

Alexander 1991:  I excluded this paper from my analysis because it had no control group and sample size of 10.

Results: