“One possible objection to SWI is that young children in Grade 1 who are just starting to read do not have the cognitive skills necessary to explore the interrelations between orthography, phonology, morphology, and etymology. Indeed, it is often claimed that morphological instruction should occur only after phonics… However, these claims and findings provide no evidence against the claim that SWI should be adopted from the very beginning.” Beyond Phonics: The Case for Teaching Children the Logic of the English Spelling System. Jeffrey S. Bowers and Peter N. Bowers
I am in favour of teaching children about morphology and etymology as part of a broad reading program. The program adopted by my own school does just this. However, I am doubtful about the idea that these concepts should be introduced at the very beginning of reading instruction.
I have already addressed the point that SWI is not the only way to teach morphology and etymology and the possibility that other approaches may be better, so I won’t rehearse that argument here. The Bowers brothers tend to cite a 2013 study by Devonshire, Morris and Fluck to support the case for morphological instruction from the very beginning of reading instruction. They point to the age of the children involved – between five and seven – and that the children apparently benefited from morphological instruction. However, age hardly matters. What matters is relative level of expertise which, in models such as cognitive load theory, effectively correspond with the presence of relevant schemas in long-term memory. These could be present in a four-year-old and absent in an 18-year-old. Critically, the children in the Devonshire et al. study had been given at least one, and up to two, years of a form of phonics instruction prior to receiving morphological instruction. So in the key sense that it has been cited i.e. as evidence supporting morphological instruction from the very beginning of reading instruction, it is entirely irrelevant.
Should a study be designed that addressed this key issue head on, I am doubtful it would find in favour of the SWI approach due to findings from basic cognitive science. However, before I examine why, it is worth pointing out that children who have not yet had a reading lesson are likely to be similar to children who have had lessons but failed to make progress in the sense that in both cases, their schemas for reading will be absent or only partially formed. We might therefore be able to infer something about initial reading instruction from studies that compare reading interventions for older struggling readers. I am aware of no study comparing SWI and systematic phonics as interventions for struggling readers, although one study involving Jeffrey Bowers did compare SWI with something called ‘motivated reading’ and found that the effects of both approaches were about the same. So that’s hardly overwhelming evidence for SWI.
One solid finding from basic cognitive science is that our working memory is extremely limited. The concept of working memory was developed by Baddeley and colleagues out of the earlier concept of short-term memory. The key difference is that working memory involves processing information in some way rather than just retaining it. Although we cannot grab a sharpie, draw a circle on a brain scan and say, ‘working memory is there,’ it is a robust construct that predicts a wide range of experimental outcomes. Currently, we believe that working memory can only process about four items at any one time.
Although fundamental to cognitive load theory, this finding sits outside the theory and is widely accepted. The role of cognitive load theory is to flesh out the implications for learning. Specifically, cognitive load theory suggests that when learning something we have not evolved to learn, such as learning to read, we must first process all of the items in working memory. However, as we learn, we build schemas in long-term memory that can be drawn upon to circumvent the limits of working memory. Oddly, it appears that we can fill-up our working memory to such a point that we can perform a task but not learn from it because there is no working memory available to help with schema building.
Cognitive load theory therefore predicts that it is optimal for learners to be asked to process only a few items at a time (no more than four) and that some or all of these must be relevant to schema building. Precisely how schemas are built and the role of working memory in building them is an area of some debate but regardless of the outcome of this debate, the limits of working memory still hold.
Now, picture a child on day zero of reading instruction. She knows none of the relationships between letters and sounds. If we are lucky, she may perhaps have received some phonemic awareness training in kindergarten (the ability to identify and manipulate individual sounds in spoken words) and she may have tried tracing some letters. But she cannot read at all. That is my definition of someone who is about to start initial reading instruction. That’s the very beginning.
I am not an expert in phonics and so those of you more expert than me will be able to suggest better approaches, but lets say I start teaching some simple CVC (consonant-vowel-consonant) words such as ‘cat’. There are three grapheme-phoneme correspondences (GPCs) in ‘cat’ – these are the correspondences between distinct sounds (phonemes) and the letter or letters that represent them in the word (graphemes). In ‘cat’ these are ‘c’, ‘a’ and ‘t’. Each of these is an item to process because our learner has no schemas to draw upon. Then there is a fourth item – blending them together. We have already reached our capacity limit and this probably explains why starting with CVC words is common in systematic phonics programmes.
How exactly could I add morphology and etymology to this? I could discuss ‘cats’ but I’ve then added one, and possibly two, more items to process. I have added the ‘s’ GPC and the concept of plurality – this may be something the child implicitly knows but mapping it to a particular letter will be a new item to process. And we have got nowhere near what I understand to be the core SWI practice of using word matrices.
With some effort, I can imagine an early approach to reading instruction that references morphology without overloading working memory, but it would be a weird kind of approach where, say, children would learn a series of sight-words and sight-morphemes. You may start, for example, by getting them to sight memorise ‘sleep’ and ‘s’, ‘ing’ and ‘er’ (although ‘ed’ wouldn’t help you much in this case). Such whole-word approaches have been largely discredited and I am not sure anyone advocates for them now. Even the dwindling band of proponents of whole language acknowledge the need for understanding GPCs, they just assume that children will pick-up this knowledge in context or implicitly. And proponents of SWI repeatedly insist that it teaches GPCs, claiming it is a misconception to suggest otherwise. In fact, the claim seems to be that SWI teaches GPCs, morphology, etymology and anything else you might imagine, right from the very beginning. It is difficult to picture how this is done without violating the four items limit of working memory.
In a lengthy response to one of my previous blog posts, Pete Bowers argues that SWI is consistent with cognitive load theory. He picks up on the idea of germane cognitive load – i.e. working memory resources that are used for schema building – and seems to suggest that SWI aligns with cognitive load theory because it generates this germane load. That may be true, but too much germane load will still violate the four items limit. This seems to be a slightly more sophisticated version of the ‘begging the question‘ syllogism that Jeffrey Bowers has deployed a number of times i.e.
- Learning is best when information is presented in an organised and meaningful way
- SWI presents information in an organised and meaningful way
- Therefore, SWI is best for learning
There are three problems with this argument. Firstly, learning is not best when too much information is presented at once, be it organised and meaningful or otherwise. Secondly, why should we assume that SWI is particularly organised and meaningful? Thirdly, why should we assume that alternatives to SWI such as systematic phonics (the clue is in the name) are not particularly organised and/or meaningful? After all, it’s pretty meaningful to be able to read stuff.
Pete Bowers goes on to give an example to support his contention that SWI is consistent with cognitive load theory. Bearing in mind my argument that cognitive load theory implies that teaching morphology from the very beginning of reading instruction would overload children, he chooses an odd example:
“Consider Cognitive Load Theory in the context of trying to teach a child who misspells the word “action” as *”acshun”. This mistake reveals accurate learning of possible ‘letter-sound correspondences’ (grapheme-phoneme correspondences). But if s/he has the phrase “sh sound” and “t sound” automated, we can see why drawing on learning from instruction would not provide guidance to get to the attested spelling. In this case the phrases “sh sound” and “t sound” represent an extraneous load to understand the spelling of this word. These oft repeated instructional phrases build well-integrated mental representations that then need to be hindered in order to get to the attested spelling of “action”.”
A child who can attempt a spelling of ‘action’ is someone who is already a pretty good reader. We are not talking about initial reading instruction here. In this case, I agree that some morphological instruction would help this child’s spelling ability and I would suggest that this may be a good stage to introduce it. It would certainly help with the interference a child would experience by having multiple ways of representing the same sound – although phonics does not teach that there is just one way, as Pete Bowers implies. I think that the critical message we are meant to take away from this example is that phonics instruction leads to this interference problem but SWI would avoid it. If true, we need to see exactly how. We need to see what SWI instruction looks like from the very beginning.
SWI sceptics such as me – or ‘phombies’ as some SWI advocates like to call us – have repeatedly asked for evidence on how SWI is used to teach initial reading. I have already examined two supposed examples that I found deeply unconvincing (here and here). Responding to my request, Jeffrey Bowers pointed me to some sources:
I have had a look at the sources and found a couple of videos, both from a place called The Nueva School. Nueva is a Californian school for gifted and high-ability learners. In order to be admitted, students need to pass an IQ test. This is interesting, because IQ (thought to measure something known as ‘general intelligence’ or ‘g’ in the literature) is highly related to working memory capacity. Perhaps working with such children has skewed the perspective of SWI advocates.
When you look at the videos, the children involved can evidently already read. For instance, consider the video below investigating the word ‘rain’. There is no instruction in how to read ‘rain’. The teacher simply asks one of the students to do so. And then off we go into morphology etc.
Yes, these are often very young children. But, again, that hardly matters. They also happen to be exceptionally gifted children. Moreover, if they can already read ‘rain’, perhaps because their parents have taught them to read at home, then this is not initial reading instruction. Age does not matter. What matters are the schemas in long-term memory.
Perhaps there is a form of SWI initial reading instruction that does align with cognitive science and that does not involve teaching children who can already read. However, until we see what it looks like, through an accurate description, lesson plan or video, I have no reason to suppose that any such form of SWI exists and teachers have no good reason to abandon systematic phonics in pursuit of a mere possibility.