This is the second of three posts examining the Simple View of Reading, one of the pillars of the “science of reading” (SoR) approach to reading instruction. In the previous post I noted that the SVR makes an important point–that reading involves the child learning how print represents words in the spoken language they already know–and that people expect it to have further implications for instruction. Here I look at the contrast between what the authors of the SVR proposed and how it is being used in the SoR. In the third post I look at what later studies of the SVR showed, and consider what is gained by linking the SVR to other pillars of the SoR such as the Reading Rope (Scarborough) and the report of the National Reading Panel. I’ll end with some broader observations about the use and misuse of research in deciding educational issues.
What Is and Isn’t in the SVR
Here is how the SVR is typically presented for educators. This illustration is taken from an explainer from The Reading League, an NGO whose mission is furthering the SoR. (Click on figures to embiggen.)
Think of this picture as a snapshot of conditions that exist as children begin learning to read. They understand spoken language at a 5-6 year old level. They will be able to comprehend texts (at that level) if they learn how the written code represents words they know from talking and listening. The multiplication sign indicates that both components are necessary: reading comprehension is zero if either of the components is zero.
What are the instructional implications of this statement, beyond the important general point about the need for print instruction? To go any further we’ll need to consider a few more things: what, when, and how.
What means specifying what’s in the circles. “Word recognition” is a convenient bit of jargon but we need to say more about what it involves in order to decide what to teach and how to assess progress. The same holds for language and reading comprehension, although at the outset the main concern is reading words.
When means considering when specific types of knowledge and skills are acquired over time, a crucial issue for instruction. How much the child knows, what they need to learn, and what they are able to learn change over time. The SVR focuses on one developmental fact, that spoken language (ability to speak as well as comprehend) develops ahead of learning about print; therefore learning to read turns on gaining knowledge of print and its connection to spoken language. It doesn’t speak to how reading develops over the next several years (or language, either).
How refers to learning: how different skills and types of knowledge are learned. Learning is conspicuously missing from the “science of reading”–remember, the picture of SVR in Figure 1 is just a snapshot of the state of the child’s knowledge. It’s not a description of how they got there or how they progress. Humans have two main ways to learn: via explicit instruction and via implicit learning. Explicit instruction you know. Implicit learning is the way brains pick up on patterns in the environment, without conscious attention or awareness. We update this knowledge all the time, as we do what people do: read and talk, write a paper, drive a car, make dinner. Achieving the right balance between these two types of learning, and adjusting it as students progress, is crucial to helping them learn efficiently and effectively.
The SVR doesn’t address these what, when, and how questions. That is why it is a huge leap to go from the SVR to conclusions about instruction.
We could stop this discussion of the SVR here. Yes, it makes an important basic point that changes how people think about learning to read once they hear about it. No, it didn’t address other issues that are closer to the classroom. Fortunately, there are other places to look for that.
In my view, stopping here is the appropriate step, but I have to address some lingering concerns:
• Many people are trying to use the SVR as a framework, filling in some of the missing information. How is that going?
• What about the 150 studies supporting the SVR? Do they address the what, when, and how questions?
• How much is gained by coupling the SVR with other studies in the canon, such as the Reading Rope or the 5 Pillars of Reading from the NRP report?
Word Recognition and Decoding
Many people have recognized the limits of the SVR and tried to extend it in various ways. Doing this makes the limitations of the SVR as a framework for instruction more apparent. Dividing things up into print knowledge, language comprehension, and reading comprehension runs into problems–even though the basic point about the need for instruction about print is valid.
Let me illustrate this by asking: “word recognition”–what is that? Yes, I know: it’s recognizing words. But to take this to the classroom we need to get more specific. It’s not a trick question, but as you’ll see, it isn’t actually that easy to answer.
The print-related component is called “word recognition” but that isn’t right because beginning readers already know how to recognize words, from spoken language (researchers call this “auditory word recognition”). Reading builds on this existing skill. No problem, we change “word recognition” to visual word recognition (VWR). The language comprehension part on the right in Figure 1 is really “spoken language comprehension”, which includes auditory word recognition. (It also includes language production, just as the print part also includes writing, but these parts weren’t included in the simple view.)
With these clarifications we have Figure 2 (I’ve omitted the = reading comprehension part to save space):
Are we good? Well, you can’t teach “visual word recognition.” That is a label for what we want learners to be able to do. Teaching has to focus on the kinds of knowledge that are involved, mainly words and their properties, which can be taught.
Trying to do this within this version of the SVR creates additional instructional puzzles. Visual and auditory word recognition share a lot of parts–for example, meaning. We don’t represent the meaning of a word once for reading and another time for speech. Word meanings develop from using spoken and written language, and from experiencing the world (perceiving and acting). This common store of information is accessed in different ways in performing different tasks (reading, spelling, hearing, talking, others). Shouldn’t science-based instruction be consistent with these basic facts?
The same issue arises for phonology: it’s used in both reading and comprehending speech (also speaking). We therefore need phonology in both circles, but people don’t have one phonological code that they use in reading and a different one for listening and talking. Same for vocabulary and morphology. Same for essentially everything, even print.
Again: Trying to make sense of the SVR at the level of detail that’s required for instruction creates paradoxes and contradictions.
Bear with me just a little longer. I know this is a slog.
Decoding “Decoding”
What if we put “word recognition” entirely in the language comprehension part, and limit the print component to “decoding,” meaning using knowledge of the mappings between orthography and phonology (“phonics”) to pronounce letter strings (words like must and novel/nonce words like nust)? The purpose of phonics instruction is to enable this process. Decoding starts with overt pronunciation (“sounding out”), but readers soon begin to do this without speaking, using phonology, a mental code that is based on the pronunciations and sounds of words. I’ll call this “orth→phon” for short.
This is the other main interpretation of the SVR, as seen in illustrations like this one (from here), but also found elsewhere:
Call this SVR.D. This graphic does not say the same thing as Figure 1 because “decoding” as described here is not the same as “visual word recognition.” The latter refers to any process by which people recognize written words. “Decoding” refers here to something more specific, generating phonological codes from print, which only works for some words in English. Such codes can be used to pronounce letter strings aloud. They can also be used to recognize words if the phonological code matches one for a word that is known from spoken language.
“Decoding” in this sense and “word recognition” work with different kinds of spelling patterns. Here’s a summary:
Which version of the SVR are you using?
If you’re feeling confused about the term “decoding,” you’re not alone. I just described the traditional concept of “decoding”, which predates the SVR and has been the basis for phonics instruction in the US since the 19th century. In English, only words that obey phonics rules (sometimes called grapheme-phoneme correspondences) can be decoded. “Decodable texts” are mainly composed of words that use grapheme-phoneme correspondences/phonics rules that the reader is assumed to have been taught.
Decoding in this sense does not work for words whose pronunciations violate the rules. These were traditionally termed “sight words,” but they are also called “heart”, “trick”, or “snap” words, with various proposals about which words should be included, how they are read, and how they should be taught. Those details don’t matter here. Under all of these proposals there are exceptions that have to be learned and recognized some other way. If there is a rule for pronouncing but, cut, and nut, it will not work for put.
Gough and his colleagues used the term “decoding” in a different way: as a synonym for “word recognition.”
For the simple view, skilled decoding is simply efficient word recognition: the ability to rapidly derive a representation from printed input that allows access to the appropriate entry in the mental lexicon, and thus, the retrieval of semantic information at the word level. Hoover and Gough (1990), p.130.
This is how jargon gets in the way of understanding. Gough and colleagues were referring to the processes by which words–all kinds of words–are recognized. They were agnostic about the details. Unfortunately, they chose to label this component “decoding,” which already had a different, narrower meaning. This created massive confusion that has now spread to the “science of reading.”
For example: If you look closely, Figure 4 is attributed to Gough and Tunmer (1986) and Hoover and Gough (1990). The print component is labeled “decoding”, as in these articles. However, Gough and colleagues did not use “decoding” to mean the “ability to use sound-symbol relationships to read [some] words“. In fact, they pointedly questioned the relevance of this kind of decoding to beginning or skilled reading. Thus the illustration does not accurately represent Gough et al.’s own proposal.
Does this matter? Well, yes. The SVR is one of the main research studies on which the science of reading is based. These observations suggest that people do not agree on what it means. In Figure 4 it is being used to justify teaching children to decode in the traditional sense, which Gough and colleagues did not propose or endorse. That assumption is coming from somewhere else and requires its own justification. In other contexts, the component is taken as “word recognition” and used to sanction other methods.
Well, so what? Isn’t Figure 4 a reasonable place to start? Readers need to learn to decode (original sense) or something like it. We’ll also need to add procedures for dealing with the exception words. This can go in the print component of the SVR, sure. But where’s the payoff? Despite all this effort to amend the SVR, we are back at Square One: how do we teach children to read words quickly and accurately given the properties of written English? So far we’ve only managed to describe the problem, not how to solve it.
Getting a little tired of this? Me too. And yet the same issues arise if we examine other targets of instruction such as vocabulary and morphology. The SVR again does not directly address these topics and so carries zero direct implications for instruction. If we were to expand the SVR to incorporate these types of knowledge, where would they go? Like meaning, they seem to belong to both components. Does that mean we teach each of them twice? Note that morphology presents the same challenges as phonics: “rule-governed” patterns with lots of exceptions (see, for example, the past tenses of verbs). Syllables, too.
These are the puzzles that arose in 1980s research on learning to read. They are arising again in approaches to instruction based on this work–for example, debates about which words to treat as “heart” words or “sight words” and in how to teach phonics. The articles show Gough and colleagues struggling to describe “word recognition” and “decoding” using the concepts that were available at the time. They knew that readers needed to learn how spelling represents phonology and that at least some instruction was required in order to break the code. But instruction about what? They were skeptical of phonics rules, which seem like clumsy approximations of the complex system readers learn. Moreover, fluent reading seems very unlike applying rules to a series of letters. Having to memorize exceptions to the rules such as said and done creates an onerous burden for the learner and misses the fact that they overlap with many “rule-governed” words. Gough and colleagues’ use of “decoding” was confusing but perhaps understandable. They were using it to refer to word recognition procedures that they correctly intuited had to exist but couldn’t quite figure out.
A solution came a few years later when McClelland and I applied new ideas about how brains learn to reading. In this account, people learn a network of mappings between spelling, sound, and meaning. These mappings reflect patterns that exist across words. Unlike rules that apply every time a pattern is encountered, the mappings vary in frequency and consistency. This knowledge can be closely approximated by simple neural networks that learn based on similarity: learning about one pattern facilitates learning other, overlapping patterns. Such networks can easily learn patterns that are hard to describe in words, for example that –ook is pronounced one way in book, cook, and look but differently in spook, which overlaps with spoon and spool, which pattern with boon and cool. The same procedures are used in reading all words; there isn’t a separate one for words that violate the rules. The network uses what it has learned from words to process letter strings it has never seen before.
Phonics rules, on this view, are a convenient fiction. They put parts of this complex web of knowledge into words. They are not accurate descriptions of what is learned, but they are useful because they make explicit instruction about the system possible. Explicit instruction has two purposes. First, it draws attention to what the child can learn: orthographic patterns that map, with varying frequency and consistency, to phonological patterns (and thence to meaning, if the pattern corresponds to a word known from speech). Learning the rules themselves is less important than getting clued into the fact that such patterns exist. The second function of explicit instruction is to get the orth-phon system off the ground quickly. With some patterns established, learning based on the similarity across patterns can proceed, with less and less reliance on a teacher for feedback as knowledge accumulates.
This framework, which has many other bits, took a long time to fully develop. It runs counter to very strong intuitions about reading, and so took a while to gain broad acceptance among researchers. In my view, the “science of reading” is recapitulating this history. At the moment, it is focused on the Gough-era concepts and running up against the same puzzles. The good news is that there is a lot of very useful stuff to come.
Interim Summary
The point of this excursion was to examine what happens if one takes the SVR seriously as a framework for early reading instruction. Attempts to be more specific about the print component run into theoretical and practical complications. We know that the reader needs to learn to read and understand words quickly and accurately; that this involves linking written and spoken language; and that the properties of written English make this challenging. After stepping through possible extensions of the SVR, we arrive back at Square One: how to teach children a system with productive patterns but many words that violate the main patterns in varying degrees, including most of the most common ones. (I’ve also noted that there is a possible solution just over the horizon but for now we are focused on the SVR and other classic SoR studies.)
There are still a few loose ends. First, it’s said that 150 studies support the SVR. Do these studies address the what, when, and how issues? Second, do other studies that are pillars of the SoR, such as the Reading Rope, provide details that are missing in the SVR? How far can we get with what we’ve got? We need to look.