On the phonemes in "phonemic awareness"

This post is a response to a recent post on the SPELLtalk discussion from Dr. Jeannine Herron, a reading specialist active in reading education. The original post is in the SPELLtalk archive here, for people who have access. For those who don't, I've reproduced it here. 

Dear Jeannine,

I read your post about phonemic awareness with great interest. You expressed concern about the fact that there is an ongoing debate about PA instruction. You articulated your understanding of why PA instruction is important and invited comments about whether people agreed with it or not. This was an admirable thing to do: thank you. You’re obviously committed to reading education and to achieving consensus about what works and why. Engaging in thoughtful, civil discussions about challenging issues is essential, of course, and I am responding with that in mind. Although I’m responding to your post, I’m addressing views that are shared by many people, which you’ve helpfully summarized. 

You’ve articulated a common rationale for PA instruction:

• • Spoken words are composed of phonemes, which are discrete segments of sound.
  • Phonemes are used in speaking and listening, without conscious awareness.
  • Graphemes (letters and letter combinations such as SH) represent these sounds.
  • Learning to read requires bringing phonemes to conscious awareness.
  • Which requires instruction.

A child who knows the phonemes of English (in this sense) will then be able to learn correspondences between units in the written code (graphemes) and units in speech (phonemes). This provides a way to link print and sound, a crucial step in learning to read.

Continuing with this reasoning, the logical sequence for teaching reading is then to start with the 44 or so phonemes. Learning the phonemes is like learning the alphabet, but for speech instead of print. Phonemes can be taught using oral language tasks such as deciding whether two spoken words start (or end) with the same phoneme, or hearing a word and then saying it with one of the sounds deleted. These tasks can also be used to assess children’s progress. A child who exhibits a requisite level of phonemic awareness as measured by such tasks is then ready to begin learning grapheme-phoneme correspondences (“phonics”). 

I will refer to this as the “Phonemes First” approach. I’ve tried to describe it as fairly and accurately as I can. Stated this way, it makes a lot of sense. But there’s a problem: the approach is based on mistaken beliefs about phonemes.

Here is some important information about phonemes:

• • Spoken words do not consist of spoken phonemes. (They are not, as you said, “individual sounds you make with your mouth when you pronounce a word”.)
  • Phoneme representations in the brain do not arise just from using spoken language. Using speech doesn’t rely on unconscious knowledge of phonemes (that needs to be brought to conscious awareness for reading).  
  • Rather than being a prerequisite for reading, phonemes are an abstraction–an idealization–that results from engaging in a variety of activities in which print and sound are paired. These activities importantly include ones involving explicit instruction. Rather than being the precursor to reading, “phonemic awareness” results from progress in learning to read alphabetic writing. In other words, the Phonemes First approach gets the causality backwards–phonemes are a result of sound-print pairings and learning to read, not a prerequisite for reading.

Speech does not consist of a sequence of discrete sounds like beads on a string. We think about it this way but that is an illusion–a very helpful one–that resulted from learning to read an alphabet. We learn to treat speech as if it consists of discrete sounds. That happens through learning about print, which shapes the way the sounds of words are represented in the brain, creating a phonemic level of representation (see illustration; click to embiggen).

The phonemic level develops for a computational reason: so the brain can efficiently connect patterns in print (one type of code) and sound (a different kind of code). The reader can then perform tasks such as reading words aloud or hearing a spoken word and spelling it. (This system is also tied to meaning, which I've left aside for brevity.)  I described the development of the phonemic abstraction in Seidenberg (2017): ‘What is it like to be the word “bat”?’ (pp. 26-29) and “Becoming a reader” (pp. 119-121). I’m attaching the excerpts here and here.

The Phonemes First approach would be entirely sensible if spoken words actually consisted of discrete phonemes. Phonemes could be taught (or “brought to conscious awareness”) and, importantly, there would be a good reason to do so: because they are used in pronouncing spoken words. Reading aloud–generating pronunciations of words from print–is an important task for beginning readers. The way to enable this would be to teach the components: first the phonemes, then the graphemes, and then the mappings from one to the other. Specifying the order in which to teach these components of reading is helpful for teachers. Following this logic, we might conclude that it’s important to teach all of the phonemes rather than only some, just as we teach all of the letters in the alphabet. 

The fact that speech doesn’t actually consist of discrete phonemes should raise questions for everyone about this approach. Teaching children to pronounce and use 44 phonemes in preparation for reading literally isn’t possible because speech doesn’t consist of discrete, pronounceable units to teach. The sounds that are being taught in phonemic awareness instruction aren’t phonemes and they aren’t the building blocks of spoken words. They are bits of sound that have been artificially isolated from speech in order to be pronounceable. We can ask whether these sounds need to be taught, but the rationale can’t be because readers need to know phonemes, the building blocks of spoken words.  

The illusion that spoken words consist of discrete sounds is very powerful. Most people don’t give it a moment’s thought: they assume that spoken words consist of the spoken equivalents of letters. It isn’t so.  

If you need further proof, do the experiment yourself using software that comes with your computer, such as Quicktime:

Take a word like BAT, record a sound for each letter. Use the “phonemes” (sounds) that are taught in PA activities if you know them. Then edit the three sounds together in the right order. The combination of the three sounds will not sound like “bat”. Such sounds are not the building blocks of spoken words. 

People are putting a lot of time, energy, and money into teaching the 44 phonemes, going so far as to police using the “correct” pronunciations (as in this example). Teaching children that the pronunciation of B is “buh” is an error (on this view) because that spoken syllable contains a vowel that isn’t part of the “b” phoneme itself. The teacher is encouraged to practice saying just the initial phoneme and to use that in class. 

These recommendations reflect the deep entrenchment of the phonemic illusion. The exercise treats “buh” as if it consists of two sounds, the “real” pronunciation of the “b” phoneme and then a vowel. A person might think there are two discrete sounds there because they can spell words like BIT and BUT.  In reality, perception of the initial sound as “b” depends on information in the “vowel” part. If you try to say “b” without the “vowel” part, you corrupt the pronunciation of the “b”. What’s left is an unnatural sound, not the sound at the beginning of “buh” or “bat”. 

Again, if you need proof, do the experiment yourself.  Record a spoken word (such as “bat”), and then try to divide it into three parts to make the “phonemes”. You won’t be able to do this. You can also do the experiment with a recording of “buh”. If you try to delete the “uh”, you take out part of the “b”.  There’s a recording of me trying to do this here.

To summarize, skilled reading depends on treating spoken words as if they consist of phonemes. This abstraction underlies our abilities to read and spell thousands of words fluently and accurately. Rather than being a precursor to reading, the abstraction emerges as children learn to read alphabetic writing. People who read non-alphabetic scripts (such as Chinese) and people who are illiterate do not develop this abstraction. Conversely, for people who learn to read an alphabet, performance on “phonemic awareness” tasks such as phoneme deletion and substitution relies on knowledge of spelling. 

Rather than treating speech as if it consists of discrete segments, Phonemes First instruction is based on the mistaken idea that speech actually consists of such segments. It doesn’t, and so the sounds being taught are artificial units rather than components of natural speech. Such instruction also reverses cause and effect, treating phonemic awareness as a precursor to reading rather than a consequence.  

Going back to your post, Jeannine, I’ve tried to correct a misunderstanding about phonemes and phonemic awareness that is very common and at the core of a popular but questionable instructional practice. I’ve probably gone into both too much detail and too little. I recorded several talks about these issues, which go into even more detail, in case you’re looking for it! The talks and slides are here. If you or anyone else has questions about this material, or think it is incorrect in some way, or want to bring up other considerations, I would be happy to engage in  discussion. This can be done on SPELLtalk, or here on this website, where comments are open. Comments that engage the issues rather than merely express approval or disapproval are welcome there.

This post raises additional questions that go well beyond the ones raised in your post. I hope there will  be an opportunity to have an exchange of ideas about them as well in a suitable forum. Here are a few:

First, what are the implications of these observations for instruction? I’ve observed that a practice in the science of reading is based on a misunderstanding of what phonemes are and where phonemic knowledge comes from. What are the implications for what to do and what not to do in the classroom? Are there any significant implications or is this just a quibble about the definition of phoneme? Perhaps the sounds that are being taught don’t fit the textbook definition of phoneme but they are based on the same idea, and isn’t that good enough for instructional purposes? In my view, the issues are not about a definition; they are about what to teach. Teaching children the letters of the alphabet, their names, and an associated sound for each (such as "buh" for B) is essential. Beyond that, is there any reason to teach any sounds of words in isolation or independent of spelling? In my view, no. What should be taught instead? In my view, the emphasis should be on teaching the correspondences between spellings and the natural, authentic pronunciations of words chosen so as to lead the child into treating words as if they consist of segments. That is a topic for another time. 

A second question is how the idea that phonemes could and should be taught became a tenet of the science of reading. The fact that speech does not consist of spoken phonemes is not controversial in linguistics (see the attached excerpt from a scholarly encyclopedia of linguistics, which uses the term “segments” instead of “phonemes”).*  But don’t people also say “the phoneme is the minimal unit of spoken language,” as exemplified by the contrast between words such as “bat” and “pat”?  No wonder people get confused. I think that it is important to understand what happened not to assign blame but to improve how research gets translated into educational practice. That too would be a good topic for a future discussion.

Thanks again for your post and for kicking off this discussion.

Mark S. 

* The text refers to segments rather than phonemes because elsewhere in the same resource it says, "Since about 1950, it has been known that there are various problems with the concept of the phoneme.... [Such] arguments against the phoneme were widely accepted, leading to the near-universal rejection of the phoneme as an element in linguistic theory.  Still, the word "phoneme" continues to be used as a convenient way of talking about speech sounds....".