Skip to main content

Not long ago, I dared question the evidence supporting the teaching of “advanced phonemic awareness.” That elicited negative reactions from some educators who found my assertions threatening. Their notion was that if I raised doubts about this, then balanced literacy, whole language, cueing systems, and the fall of the Republic would be manifest upon the land.

I’m not saying that there aren’t balanced literacy fanatics who cheer when I tell the truth about something like that (I wish they were as happy when I explain why it’s a bad idea to teach kids to guess words from context in place of reading).

Okay. I say there is no evidence for teaching advanced phonemic awareness and some readers are ticked off and dismissive and others are chanting, “Hooray for our side.”

But this hypothesis — that we may need to teach advanced PA — is too important to leave in this fog of virtue signaling and side choosing. 

This idea of advanced phonemic awareness was popularized by David A. Kilpatrick, a psychology professor at the State University of New York at Cortland. David has had a long career in school psychology and academia and may be the best-read scholar when it comes to things alphabetic.

How better to evaluate this idea than to turn to the source himself? He, more than anyone, knows from whence it came and the evidence, if any, that supports it. We recently spent a couple hours exploring these issues, and I reread the relevant portions of his landmark book [from 2015], Essential of Assessing, Preventing, and Overcoming Reading Difficulties (opens in a new window).

David thought a major clarification was needed. In his 2015 book, he used the terms “phonemic proficiency” and “advanced phonemic awareness” interchangeably to refer to a cognitive or linguistic skill. He never used these terms to refer to any specific instructional activities. Nevertheless, many educators have conflated advanced phonemic awareness with the notion that certain teaching activities, namely phoneme deletion and substitution, were required. That notion came from some readers of his book, not from David. He never claimed that all kids needed to be engaged in deletion or substitution tasks.

This misunderstanding has led David to abandon the term “advanced phonemic awareness” altogether. He still refers to the underlying ability that all students must develop as phonemic proficiency. When suggesting relevant instructional activities, he describes them specifically (e.g., blending, segmentation, deletion, substitution). The point is to distinguish the ability that students must learn from the instructional activities we use to promote that ability.

More interesting are his insights regarding phonemic proficiency.

He begins with Linnea Ehri’s theory of orthographic mapping. That theory “explains how children learn to read words by sight, to spell words from memory, and to acquire vocabulary words from print” (Ehri, 2014, p. 5). Basically, young readers need to develop sufficient knowledge of the letters and phonemes so that printed words and spelling patterns can be connected with phonological representations in the mind. It is these phonemic representations that are the anchors that secure that information in memory.

Readers only briefly and occasionally “sound out” words when they read. That would be too slow and laborious (a fact that troubles balanced literacy proponents). Orthographic mapping, like any kind of fast mapping (the concept was drawn from the language learning literature) is about getting information into memory quickly and effortlessly.

Not surprisingly, Dr. Ehri champions teaching letter names, spelling, and phonemic awareness so that students can fully segment (analysis) and blend (synthesis) the sounds. This means they could separate words into their phonemic parts proficiently and reassemble them. Those abilities play an obvious functional role in orthographic mapping. As the National Reading Panel reported, teaching PA improves reading achievement (NICHD, 2000).

That’s where most of us leave it and why I dared question claims that students may require more PA than that. And, that’s where it gets interesting …

You see, David isn’t like most of us. He didn’t see that research as complete. Oh, he bought into the idea that kids needed explicit instruction in PA, he just wondered if that was sufficient for everyone.

No idle wondering was this.

It led him back through a plethora of studies to see what was going on with kids who didn’t succeed and why some word-reading interventions ended up with better results than others.

He concluded that some kids with core phonological processing deficiencies needed more help developing these phonemic anchors. Meeting typical early learning criteria (e.g., knowing letter names and sounds, segmenting words phonemically) simply were not getting these children far enough.

Studies have reported that older students demonstrate much greater phonemic proficiency than younger students. That means the ability continues to grow beyond the period of explicit teaching that is now often provided in grades K–1. We have long known that PA and reading are reciprocal — PA development improves early reading and reading practice extends PA.

Kilpatrick’s hypothesis is that for most kids, developing PA to the point where they can fully segment words is all that is needed to get things rolling — the PA automaticity that supports orthographic mapping naturally develops from there. He concludes that “business as usual” reading and spelling instruction and reading and writing practice are all that are needed to keep PA proficiency developing. Except…

Except for those kids with core phonological deficits … the ones who simply don’t get enough PA support from usual reading experience. They would, he hypothesizes, benefit from from more extended PA instruction to promote the phonemic proficiency displayed by typical readers. The point of this isn’t to engage kids in particular kinds of practice (e.g., deleting phonemes, adding phonemes, reversing phonemes), though engaging in some of them may be part of such practice — David thinks that could be beneficial. No, the purpose is to enable orthographic mapping.

Is this a goofy theory? I didn’t think so, but who am I? I, again, went to the source. Who would have better insight into the value of this hypothesis than Linnea Ehri?

She has a somewhat different conception of the form that this advanced development may take but she agreed with many (most) of his predictions of what would work and with whom and why it would matter. So, not a crazy idea at all. (Maybe those people who claimed me to be an idiot might be onto something.)

Remember, however, my complaint about “advanced PA” wasn’t about its potential accuracy or its value as a hypothesis that could help advance our understanding. No, my problem was with folks who were extending PA instruction in the classroom for everyone, states adopting more extended or rigorous PA standards for all, and the like. Kilpatrick agreed with this concern. He said most students should not require this degree of PA training. He wasn’t abandoning the idea that students needed phonemic proficiency that exceeded the abilities to segment and blend, just acknowledging that those extended levels of proficiency develop naturally in most students. (Remember, his 2015 book was focused on struggling readers.) 

What of all of that?

David did look at a lot of studies and his ideas did emerge from the patterns he noticed in the word-reading intervention literature. To many that would seem to make his claims research-based and scientific. But a direct study that compared the efficacy of different kinds of PA instruction would be needed to convince the scientific community. David believes that the patterns he identified in existing studies provide valuable clues about best practices, not this is a “fully established scientific finding.”

For instance, Richard Feynman, the Nobel prize winning physicist wrote: “When you have put a lot of ideas together to make an elaborate theory, you want to make sure, when explaining what it fits, that those things it fits are not just the things that gave you the idea for the theory; but that the finished theory makes something else come out right, in addition.”

That just means that you need to test those patterns that you think you saw in the original data. In this case, if you taught PA to the point that some kids could do those more complex phonemic manipulations while staying to segmentation and blending with the others, you’d find out whether that kind of teaching was advantageous. It might be, but until such studies are done, we won’t know.

I wondered how Linnea Ehri felt about that. One could understand she might choose to embrace David’s hypothesis as a way of advancing her own theory.

Her response?

“David’s hypotheses need to be tested directly with controlled studies that examine whether training improves phonemic proficiency, the involvement of grapho-phonemic proficiency, and whether this facilitates sight word learning in typically developing readers and struggling readers.”

Nothing self-serving there. Just a scientist frankly evaluating the science.

Oh, and, perhaps more on point, is David Kilpatrick’s own take on the matter:

“We completely agree that a study is needed in which the only difference between groups is the nature and extent to which phonemic skills are developed. Indeed, that was supposed to happen last year with 200 severely reading disabled students. But due to COVID-19, that got scrapped.”

Maybe, I’m not the idiot or whole language devotee some of you presumed.

Where does that leave us?

1. Kilpatrick, Ehri, and Shanahan agree that there is substantial evidence showing a clear benefit from explicitly teaching young readers to perceive the sounds (phonemes) within spoken words and to link these sounds with letters and spelling patterns. Getting kids to the point of full segmentation in PA is a reasonable goal for regular classroom instruction.

2. We agree that the engine of learning words is not rote memorization but a deep knowledge and proficiency of phonemes and letters in words.

3. We also agree that there are some students who manage to accomplish those levels or degrees of PA efficiency but for some reason still fail to accomplish orthographic mapping.

4. We all agree that more needs to be done for those kids, but here there are some variations in our thoughts about what is needed.  

5. Kilpatrick has a cool theory that suggests that instruction could profitably automatize phonemic analysis skills — enabling orthographic mapping. That instruction would not be for all students but only for those who don’t get to that point of “sight word” reading. We all agree that his idea may be correct.

6. We also all agree that this idea is worth testing and that until it is tested, it has the status of a sensible data-based hypothesis, but that it is not a proven method for improving reading achievement. Until there is such evidence, “advanced PA mandates” and the like is the work of politicians and salesmen rather than scientists and educators.

Oh, and given the unfortunate misuse of the term “advanced phonemic awareness,” let’s join David and abandon its use. Rest in peace, advanced phonemic awareness.

Thanks, David and Linnea. Your generosity in helping me with this blog entry is deeply appreciated. Your integrity and deep devotion to a true science of reading is laudable. Of course, if I’ve gotten anything wrong here, that’s on me — not you.

Selected comments

Comment from Jessie

Thank you for this thorough explanation of the research that led to Kilpatrick’s hypothesis. As a follow up question, would there be any benefit of doing addition, deletion, or substitution tasks to directly support phonics instruction? (So not just skill drill on these practices for the sake of the practice itself, but rather as an entry point for print learning.) For example, would it help students to do a few warm up addition tasks orally first in preparation for a lesson on blends (so they hear that pattern of moving from a CVC word to a CCVC or CVCC word)? Similarly, would it be helpful to warm up with a few medial substitution tasks as students are first learning the VCe pattern (so they can hear the distinction between the closed syllable short vowel sound and the long vowel sound in the VCe pattern)? Thanks for sharing your expertise! 

Reply from Tim Shanahan

Jessie —

Nobody really knows. It is clear that analysis (segmenting) and synthesis (blending) have functional roles to play in reading and spelling. The idea that engaging in skills like deletion and manipulation is meant to help those kids who have phonological deficits to develop them – in other words, practicing with something harder that might stimulate the kinds of skills the other kids are picking up. I guess that means such additional practice could be helpful to a small percentage of your students – and then, again, since it hasn’t been tested, it could just be a waste of time (and a waste that would reduce the amount of decoding instruction you could be providing). If I were you, I would wait until the idea has been evaluated OR test it out in your own class to find out if you can see any difference in student progress (in other words, if you try it, try it with a grain of salt).


Comment from David A. Kilpatrick

Thanks Tim for your blog. I think it went a long way in terms of clarifying this issue. I really appreciate that.

I would, however, like to address something. You mention in one of your responses the idea about putting “belief above data.” To be clear, that’s not what is going on here. Quite the contrary, actually. My view on this issue emerged by complete surprise. I wasn’t looking for it but noticed it when sifting through the data in the intervention research literature (i.e., struggling readers) and considering standard score gains on nationally normed word identification tests reported in those studies. The data from the intervention studies induced the “belief.” Is it a hard scientific finding? No. Is it a demonstrably clear pattern in the scientific literature? Yes. It’s more than just a nice idea.

You and I both agree that a direct study of this phenomenon is needed, in which two groups of struggling readers get identical phonics instruction but the only difference between groups is the nature of the PA instruction. Where we seem to disagree is in what manner the existing intervention research should guide our efforts with struggling readers.

Studies that use phonics with no additional oral phonemic awareness typically show a benefit compared to a control group (usually of a whole language intervention). However, improvements on nationally normed tests of word identification in those studies are consistently modest, usually 3 or 4 standard score point gains. Even though national norms are far from a perfect point of reference, they let us know if a student is actually “gaining” on his or her peers.

Then, when we look at phonics intervention studies that add oral PA (or PA with non-lettered tokens) but only train to the point where students can segment and blend, the results range between 6 and 9 standard score points on normed word identification tests. But in studies that used phonics instruction and more rigorous phonemic awareness intervention (orally and with non-lettered tokens), usually involving deleting and substituting phonemes, standard score gains hover around 14-17 points on normed word identification tests. Some of these studies had control groups, some did not. But control group or not, we have to ask why one type of instruction consistently yielded substantially higher standard scores on the same type of word identification measure. We simply don’t see such strong gains using the less rigorous PA interventions.

These findings cut ACROSS studies, rather than represent the results from WITHIN a given study. So where we differ has to do with whether we put any stock in the fairly consistent pattern of standard score point outcomes on word identification tests.

I’m not suggesting that this pattern in the intervention research is a hard finding like the results of a random-control-trial (RCT) experiment. But given that teachers are looking for best practice, I’m suggesting it might be a good idea to emulate the studies with the strongest word identification outcomes rather than emulate the studies that consistently display lesser results. That seems reasonable to do while we await further research that may settle the issue.

I’m talking about struggling readers, not Tier 1 instruction.

Anyway, I wanted to be clear that we are not talking about beliefs or hunches or ideas untethered to data, but rather something that emerges from the data. We differ only on where to set the “best practice” bar. True, there is no RCT, but there is a clear pattern of empirical findings. I think emulating the studies with the strongest outcomes until further notice is a reasonable way to conceptualize best practice.

Again, I appreciate you clarifying the relationship between phonemic proficiency and orthographic mapping and explaining to folks that the term “advanced phonemic awareness” is a confusing phrase in need retirement.

Reply from Tim Shanahan

David —

I’m concerned about the issue that Nathan raises below. He isn’t saying that you went out looking for this pattern, but he is saying that he is skeptical about whether that pattern exists at all. Tiffany raises a real similar question. My original concern is that the idea of “advanced PA” and the associated activities that were presumed to teach it hadn’t yet been evaluated in classrooms and that I didn’t think that was good for kids. You appeared to share some of those concerns – believing people had conflated instructional activities with the cognitive changes you were trying to elicit and by generalizing something intended only for struggling readers to the whole population.

But these folks are questioning the pattern itself it seems to me. I very much like the suggestion that your next step should be to try to nail down the construct by conducting an appropriate meta-analysis to see if there is such a pattern, how reliable it is, how strong it is, etc. The nice thing about that suggestion is that it wouldn’t require a long wait for schools to be willing to be studied again due to COVID. Those data exist, you’e reviewed those thoroughly to your satisfaction and now you could put the idea up to scientific scrutiny at this point. I didn’t think of that myself but given how egregious some of the misinterpretations have been so far, I think it would either reassure some of those who have bought into those ideas (thinking them to have been nailed down) or it might do away with the issue altogether if Nate is correct about the pattern itself. Cool and helpful idea, I think.


Comment from Nathan Clemens

Thanks Tim, I agree with a number of points here particularly about the lack of evidence. A significant problem that must be acknowledged is that David has repeatedly framed his recommendations as being evidence-based. For example, from Kilpatrick & O’Brien (2019): “Yet when students receive more challenging phonemic awareness training, particularly using phoneme manipulation activities (phoneme deletion and substitution of phonemes within various positions within words), a greater degree of phonemic proficiency develops. This presumably allows students to more easily remember the words they read, resulting in the largest standard score point gains found in the intervention literature (e.g., Alexander et al., 1991; McGuinness et al., 1996; Simos et al., 2002; Torgesen et al., 1999, 2001, 2010; Truch, 1994)” (p. 203).

Just a cursory look at these citations reveals that none offer a basis for causal inference, nor were the studies even designed to test the hypothesis in the first place. Misrepresentations of the evidence base like this have been relayed to thousands of educators.

Comment from Tiffany K Peltier

Dave and Tim, thanks for your responses.

Dave, I agree that a systematic review of the literature would be a valuable thing to do before we can say this pattern you are proposing is actually a part of the science of reading.

I’ve seen states mandate some of these hypotheses you have made about phonemic proficiency, “advanced PA,” oral PA instruction, the PAST assessment etc., as part of *law* under the term “Science of Reading” (such as Oklahoma and Arkansas) which seems very problematic. I hope states read this and begin to shift their practices back to what we know from peer-reviewed studies.

I agree with Dr. Brady that some of the recommendations to teachers over the past couple of years have been problematic too, not only because a systematic review hasn’t been done, but also because the recommendations from the patterns you posit are contradictory to the research we have directly comparing instruction and intervention for students in PreK, K, or at-risk for reading difficulties. It would be great to be able to see all of the studies together to be analyzed in a systematic way.

Great conversation. Thanks for the blog!

Reply from David A. Kilpatrick

Tiffany and Tim,

Yes, of course a meta-analysis would be most useful here, which is why I’ve been working on one.

Back in 2013, I started doing a review of intervention studies for students with word-level reading difficulties. I was following the lead of Joseph Torgesen whose 2005 research synthesis that appeared as a chapter in Hulme & Snowling’s “Science of Reading” volume was, I believe, the only synthesis/review of word-reading intervention based upon standard score gains rather than effect sizes. It was in the course of working on that paper that I noticed the pattern I described above. I have assumed the reason this pattern had not previously been noticed was because all other reviews used effect sizes. Effect sizes do not represent a stable point of reference across studies and stronger or weaker performances by experimental groups can be minimized or exaggerated depending on the performance of the control group. Although the idea of using standard scores as the outcome variable has its own issues, it does not share this major drawback displayed by effect sizes. Given that no one looked at the research from this vantage point before (other than Torgesen), it should be at least somewhat unsurprising that something was noticed that previously had not been noticed. Torgesen did not notice this pattern because he did consider differences in the nature of the interventions but rather focused on testing out the metric of number of standard score points per hour of intervention across studies. While I was working on this project, I was invited to write a book that took up the better part of two years. The review did not get completed, but I summarized what I had found in that 2015 book. Earlier this year I resumed working on it. Given concerns voiced here, I’m happy to bump up that project on my priority list.

Comment from Lauren Thompson

Thank you for the clarifying post. I’m so glad that you got Dr. Ehri and Dr. Kilpatrick directly into the conversation. I have accepted that Kilpatrick’s theory is still a theory, but an interesting and productive theory. It’s got me going back to the NRP report and trying to parse exactly what the conclusions were about phonological and phonemic awareness instruction. Perhaps you/they can help.

1. The NRP report states in a number of places that studies showed that phonemic awareness activities that involved letters at some point, in some way, led to faster acquisition of reading skills. Most of the studies involved young children, who were perhaps still learning letter-sound correspondences. Did using letters with PA activities improve students’ sound-symbol knowledge, and thus improve there reading development, or did the letters clarify what phonemes are and help students to grasp the concept of phonemes? Both? Perhaps students with good sound-symbol knowledge don’t need to use letters to become better at phonemic awareness. Many of Kilpatrick’s One Minute Activities, which my students enjoy, are very difficult to do if letters are involved, because of the different spellings of rimes. Does the NRP report advise against doing phonemic awareness activities with sounds only with students who already know their ABCs?

2. The NRP report also repeats that doing a few phonemic awareness activities was better than doing many. What’s unclear is whether this means working on only a few types of activities in each session, or teaching only a few activities through the whole program. Kilpatrick’s program calls for teachers to focus on only two levels of activities in one session, one at the near-proficient level and another at a near-accurate level. Another popular phonological awareness program calls for teachers to present up to 10 different activities in one session (without letters, by the way). Do the findings of the NRP report support either approach? Or advise against them? The NRP mentions two activities in particular, segmenting and blending. Are these the only activities that are recommended? Or is this recommendation only for beginning students in K and 1?

Perhaps these could be two ideas for future conversations.


Reply from Tim Shanahan

Lauren —

The National Reading Panel found that phonemic awareness instruction in Grades K-1 and with older remedial readers with deficits in phonemic awareness benefited from explicit PA instruction. The instructional benefits were biggest for the normal range of K-1 students, they were somewhat smaller for at risk K-1 students, and with the least payoff for those older kids.The studies provided between 1-99 hours of PA instruction in these studies, with the best learning response from those who received between 14-18 hours of it. That would be comparable to 15 mins per day for a semester. The panel did not conclude that to be the right amount as it was recognized that some kids would not reach those blending and segmenting benchmarks and should receive more (and for kids who could do those things proficiently, the time could be less). The studies did not favor any specific lessons or lessons focused on specific PA skills — though getting kids to the level at which they could blend and segment was the target (perhaps some of those more complicated activities could stimulate kids to segment better, etc.). However, it was found that the instruction should stay simple (focusing on only 1-2 skills at a time to avoid confusion). PA worked best when it was combined with letter instruction, so no, it would make no sense to either wait for ABC knowledge to be there or to hold off on ABCs. Many people think of PA as being a kind of readiness step for explicit decoding, but that moving back and forth between PA and decoding appears to be better (I think the same can be said about the research conducted with preschoolers and kindergartners reviewed by NELP). I know the idea of including letters or in combining PA and phonics in those ways upsets some folks, but in the 51 studies that we reviewed that was best for children’s learning.

Research since NRP and NELP hasn’t really challenged those findings and until it does, I think those features of PA instruction are the best bet. David’s theories certainly suggest some other possibilities and when those are evaluated, it is possible that we may need to adjust our thinking.



See all comments and replies here › (opens in a new window)

About the Author

Literacy expert Timothy Shanahan shares best practices for teaching reading and writing. Dr. Shanahan is an internationally recognized professor of urban education and reading researcher who has extensive experience with children in inner-city schools and children with special needs. All posts are reprinted with permission from Shanahan on Literacy (opens in a new window).

Publication Date
November 15, 2021