Phonics – is there much evidence?

A small but exciting study was completed recently. Dr Marlynne Grant has been working with a cohort of students learning to read from Reception to Year 2 with her own systematic synthetic phonics programme called ‘Sound Discovery’. She followed a Reception class of 30 students in a catholic primary school designated for travellers of Irish origin. Here are a few of the findings:

  • By Year 2 the class were 28 months above their chronological age for reading and 21 months ahead for spelling. The overall achievement range was 7.07 years to 13.09 years for reading and 7.01 years to 14.09 years for spelling.
  • By Year 2 children eligible for Free School Meals were on average 24 months above their chronological age for reading and 20 months above for spelling.
  • By year 2 the boys in the class were on average 36 months above chronological age for reading and 27 months above for spelling.

So, given the number of times I have read the assertion that there is no evidence in favour of SSP (systematic synthetic phonics) to teach reading, surely SSP proponents might be expected to be crowing? Their ‘method’ finally has some proof.

There is a good reason why this study was greeted with great interest but not a fanfare. We’ve seen it all before. Dr. Grant’s study is actually small scale and not prepared for peer review. However, for those of us only too familiar with the success of SSP the results are unsurprising. They match the vast body of converging evidence that SSP is a better approach than the methods used in most British classrooms. Personally, I don’t leap to implement the latest piece of educational research because individual studies are narrow and classrooms and the children in them are complex. However,

“that direct instruction in alphabetic coding facilitates early reading acquisition, is one of the most well established conclusions in all of behavioural science” (Stanovich, Progress in Understanding Reading p415.)

In science it is rare that anything is proved in one defining study. Data is evaluated from dozens of experiments, each containing some flaws but providing part of the answer”

award winning cognitive psychologist Keith Stanovich explains in his fascinating article Using Research and Reason in Education p15-19. He goes on:

“This is a gross misunderstanding of scientific process… one experiment rarely decides an issue…issues are most often decided when the community of scientists gradually begins to agree that the preponderance of evidence supports one alternative theory rather than another… when a conceptual hypothesis survives many potential falsifications based on a different sets of assumptions we have a robust effect…taken collectively, a series of partially diagnostic experiments can lead to a strong conclusion if the data converge…”

He explains that:

“The principle of converging evidence has been well illustrated in the controversies surrounding the teaching of reading…the results of [a variety of] studies have been synthesized in several important research syntheses:                                                                                                                                                               (Adams, 1990; Ehri et al., 2001; National Reading Panel, 2000, Pressley, 1998; Rayner et al., 2002; Reading Coherence Initiative, 1999; Share and Stanovich, 1995; Snow, Burns, & Griffin, 1998; Snowling 2000; Spear-Swerling & Sternberg, 2001;Stanovich, 2000).                                                                                                                                                                             These studies were used in a process of establishing converging evidence, a principle that governs the drawing of a conclusion that a particular educational practice is research-based.”

A research synthesis draws on many studies, for example the National Reading Panel 2000 synthesis draws on the results of 66 comparisons from 38 different studies to indicate:

“… solid support for the conclusion that systematic phonics instruction makes a bigger contribution to children’s growth in reading than alternative programs providing unsystematic or no phonics instruction.”

So where does Dr Marlynne Grant’s study fit into this body of evidence?

To appreciate its place you need to be aware of some of the main areas where there has been converging evidence in favour of SSP (systematic synthetic phonics). I will only explain some in more detail but do follow the links.

  1. Evidence to disprove the theories behind the method used in the vast majority of British classrooms (see here and here ).

This approach is confusingly called ‘mixed methods’. It was theorised in the 1960s and 70s that fluent readers used prediction. Therefore in approximately 94% of British classrooms (see p61) although children are taught to recognise how letters correspond to sounds, when actually reading books children are encouraged to guess what words might be. They are taught to look for ‘cues’ such as knowing the initial letter in the word, overall shape of word, context and pictures. In fact the theories were wrong. Fluent readers don’t guess.

The term ‘mixed methods’ is very misleading. It is a method which encourages children to attend to word shape and context, at the expense of attention to all the letters in the word and their order. Saying you support ‘mixed methods’ is rather like telling people they can drive their car on the left and also drive on the right simultaneously. When reading you can teach children so: ‘eyes are darting about for clues’ (mixed methods) or ‘eyes go  left to right to decode’ (SSP). A child can’t do both simultaneously…

However, mostly the reading books used in schools aren’t designed to help children practise to read words left to right through decoding with the letter sound correspondences they are familiar with. The books are designed so guessing works. Many children initially learn this easy shortcut in preference to ‘sounding out’. It is those that gradually lose the guessing habit that will become fluent readers. Too many don’t – as the booming dyslexia industry demonstrates.

To be clear an approach to reading based on an entirely debunked theory is still the basis for reading instruction in possibly 94% of British schools. Ignoring all the evidence Reading Recovery, the National Literacy Trust, International Literacy Centre and United Kingdom Literacy Association still advocate ‘mixed methods’. It is this bizarre state of affairs that provoked Stanovich’s whole article above and Stanovich is a good starting point if you are interested in the research in this particular area.

2. Evidence that fluent readers need to recognise that words are composed of sounds and match letters, or groups of letters, to sounds. See here

3. Evidence that “the vast majority of school-age struggling readers experience word-level difficulties” (Fletcher et al., 2002; Torgesen, 2002).

This “bottleneck” at the word level is thought to be particularly disruptive because it not only impacts word identification but also other aspects of reading, including fluency and comprehension (LaBerge & Samuels, 1974). According to Torgesen (2002), one of the most important discoveries about reading difficulties over the past 20 years is the relationship found between phonological processing and word-level reading. Most students with reading problems, both those who are diagnosed with dyslexia and those who are characterized as “garden variety” poor readers, have phonological processing difficulties that underlie their word reading problems (Stanovich, 1988)” (p.179). [Links and more detail here]

4. Some evidence that reading is accomplished with letter-by-letter processing of the word. See here p26.

‘Fluent readers do perceive each and every letter of print. Thus, we can distinguish casual from causal, grill from girl, and primeval from prime evil.’

5. Evidence that learning to recognise whole words is ineffective. See here 

“The evidence against such an account of reading behaviour is by now incontrovertible. Accurate and fluent word decoding is a pre-requisite for efficient reading for interest and information”  ‘Dyslexia, literacy and psychological assessment‘, the British Psychological Society 2005 p26

6. Evidence showing that countries with ‘shallow orthographies’ (words are easy to decode) do not have a problem with ‘dyslexia’ as we know it. See here

7. Evidence that decoding should to be taught separately from comprehension. See here and here.

Those opposing SSP criticise it for focusing on children’s ability to decode the text rather than comprehend the text’s meaning. These criticisms again ignore the overwhelming evidence regarding decoding and comprehension.

a. It seems perverse to criticise a method that nearly eliminates dyslexia type problems (see links in part 7) because it doesn’t also solve children’s comprehension problems.

b. There is lots of research showing children will comprehend writing if it is within their spoken vocabulary and understanding. This is known as ‘the Simple View of Reading’. There is little difference in comprehension whether a passage is read to you, or you read it yourself but you can only lift the meaning from the words if you are able to decode them. Therefore to improve comprehension it is essential that children have broad and rich exposure to written literature, much of it at the higher level of children’s oral comprehension, to enrich their vocabulary and knowledge of the world.

Critics of SSP suggest it leads to children decoding without attending to the meaning of the words because they are not using ‘prediction’ to read. However, fluent readers don’t guess and it is simply incorrect to suggest SSP leads to poorer comprehension than other methods. In fact comprehension is generally better than average with SSP (see studies in next section) because children read more books so learn more about the world. All children should have exposure to plenty of good literature to improve comprehension then the choice is between a method leading to most children decoding effectively or one where too many struggle to decode and thus comprehension is a real challenge.

 So the research on phonics is by no means minimal…

However, all this does not prove that specific SSP approaches do actually work better. So finally we have:

8. Evidence that in the classroom SSP leads to results that are startlingly better than is generally expected.

Marilyn Jager Adams’s writes in her 1990 book synthesizing research to that date, Beginning to Read: Thinking and Learning about print:

‘Perhaps the most influential arguments for teaching phonics are based on studies comparing the relative effectiveness of different approaches to teaching beginning reading. Collectively these studies suggest, with impressive consistency, that programs including systematic instruction in letter-to -sound correspondences lead to higher achievement in both word recognition and spelling, at least in the early grades and especially for slower or economically students.’ (p. 31)  

Studies that leap to mind are the enormous ‘Project Follow Through’ and the Clackmannanshire longitudinal study but there are many others, see here and here.  Dr Grant’s study is interesting because it breaks down the performance of different at-risk groups. Her findings suggest a full study would be fruitful. What does she say led to those results? Simply good SSP from the beginning of Reception (waiting until the phonics screening check in year 1 leaves many behind). Children that struggled were not given ‘other methods to use’ (because they don’t work) they had extra time in catch up groups with the same programme to consolidate their learning.

When SSP approaches are taught faithfully, systematically, and without guessing from contextual cues or an emphasis on learning some words by sight, the results can be astounding. Surely things have to change? Can those blinded by ideological bias, misinformation and habit really continue to ignore the weight of evidence and stop others from finding out? Can children be deprived of that most vital skill and precious gift by those unwilling to even look into more successful methods?


Thanks to Anna Worthington and others at the Reading Reform Foundation for links and helpful comments.


32 thoughts on “Phonics – is there much evidence?

  1. Really important posting – thank you so much.

    Please check out your very last link which is not currently active and looks like it may well be Dr Grant’s paper!

    Will refer to this posting widely and add it to my Phonics International message forum where I have linked to other excellent postings that you have written.

    I shall give this particular posting it’s own thread because it already links to excellent evidence and information.

    Thank you once again.


  2. Despite all the evidence, how much do teachers in England know and understand about the research on reading, and how much do they understand and adhere to the systematic synthetic phonics teaching principles?

    Readers of your blog posting might also be interested on the questions which I raise in response to three important, but different, reports published in May 2014 regarding the state of play in England:

    “Debbie Hepplewhite looks at the progress, practice and problems of synthetic phonics teaching in schools

    Three different, but inter-related, reports on synthetic phonics were published in May 2014. All three reports are interesting and informative but, in some ways, they leave us with more questions than answers. They certainly raise serious questions regarding early literacy provision for children generally and for widely recognised vulnerable groups:

    *do teachers embrace in full the systematic synthetic phonics teaching principles described in government guidance and in the core phonics programmes that they purport to follow?

    *what does the widespread objection to the 40-word Year 1 phonics screening check actually reflect?

    *what approach and programmes really serve children best, particularly those who are slower to learn or with special needs?”

  3. I agree that the research supports SSP, but how do we teach words that cannot be decoded phonically without mixing methods and teaching them by sight?

    1. Of course our writing system is sound to symbol which means all words can be decoded. However, I presume you are asking about words that have a complex or less common code? Have a look at some of the most successful SSP appproaches that are linked to on this blog and what they do. Generally they use simple code when first learning and gradually teach the complexities once simple code is grasped. If there are words with more complex code that children come across generally they should decode what they can and the teacher will explain the complex aspect of the word.
      This blog gives a really useful explanation:

  4. Hate to say this, because I’m convinced by the SP evidence, but a study that begins “The aim of these studies is to demonstrate…” and doesn’t appear to have a control, isn’t going to carry much weight with the sceptics.

    1. Of course alone this study, or any other, can’t be used to prove anything much. That is exactly why I explained the idea of ‘converging evidence’ and how Stanovich explains that such information, despite drawbacks, can be one (very small) part of a big picture.

      1. I don’t think this study should be discounted, but I get the impression that Stanovich is thinking of well-designed experiments or trials forming the converging evidence, not people’s personal observations.

  5. In the absence of the research community and universities taking an interest in objective and comparison trials of teaching programmes based on existing evidence to date, then what remains is a responsible approach to look at practices/programmes relative to ongoing results across all types of identified groups in schools to contribute to the overall picture.

  6. You will also find that leading systematic phonics programmes’ authors are supportive of one another, and confident about the effectiveness of all the programmes, as we are so aware that it is the systematic code-based approach with cumulative word/sentence/text level resources plus decodable reading books – and no multi-cueing guessing strategies – that are at the heart of all the success across all the various groups of children. Thus, there is a set of teaching principles that we identify as being key to thorough, non-damaging teaching and that we all share in common.

  7. There has to be a distinction made between the recognition that synthetic phonics is a good tool for decoding regular words and the RRF corollary that it should be taught first, fast and exclusively in reading instruction, as advised in England currently. Research on the value of systematic synthetic phonics needs to acknowledge that in England it is being enforced in a specific form through funding and testing. Does the research you have quoted investigate this specific form, or is it more concerned with process than method? For instance, Stanovich shows that experienced readers do not use context, but does not say that, therefore, beginner readers should not use context. Adult readers do not generally use context because they do not need to, as they know words when they look at them. Publishers of decodable reading schemes naively try to replicate this situation (no context needed) for beginners without acknowledging either that some words are not decodable through phonics or that the aim of reading instruction is automatic recognition of words, not phonic decoding.

    By the way the Clackmannanshire study is not without its faults:

    1. First I do wonder why you are not really keen to find out how Grant (and others) were able to get such stunning results? Whether or not you suspect SSP was responsible for those results they cry out to be investigated further if you care about childrens’ literacy.
      There are a lot of different points you raise and without writing many more blogs I could not address them all coherently. However, I think my blog does already address some of the issues you raise.
      The whole argument about ‘converging evidence’ is that evidence from a range of areas moves us towards a particular judgement even when research in one narrow area is not enough. This is the case with your suggestion that early readers ‘need’ prediction. I’d point you towards the Stanovich links as a starting point but don’t have time currently to compile a similar blog on this one issue. This piece of research
      is in precisely the right area, demonstrating that early readers apply the correspondences they encounter in new contexts. There is actually lots of relevant research but I am afraid it is a while since I’ve read it all and I haven’t time currently to dig into it again.
      Just a few quick points I can make:
      Phonics is not a tool for ‘regular words’ because ALL words are ‘regular’ in the sense that our writing system is a sound to symbol one. All words are made up of single letters or groups of letters that represent the sounds of speech. It is just that some code is more complex than other code. That is why children start with simple code amd gradually learn more complex code when learning with SSP. It also explains why other methods such as prediction are distraction because they divert attention from what readers need to appreciate to become proficient – that very relationship between sounds and symbols. Those helping struggling older readers using SSP are well aware of the problems caused by encouraging unhelpful guessing (eyes darting around the page rather than left to right decoding.)
      SSP is not being enforced in a specific form, there are an enormous range of SSP programmes that all meet government guidelines. However, ‘mixed methods’ are not advised even though (as I say in the blog) possibly 94% of schools use this approach instead of SSP.
      ‘Adult readers do not use context because they know what the word looks like’. Here you make a presumption dealt with in the blog. Adult readers use phonic knowledge to read and there are links to research on this in my blog and just to clarify if necessary, they also don’t use shape of words to read (there is plenty of research on this).
      Regards the Clacks study the major ‘fault’ found with the Clacks study was that comprehension did not improve as fast as decoding skill. However, that is because any method by which reading is taught never could have been a really significant factor in improved comprehension as I explain in my blog. As I say in my blog, to dismiss an approach when it has such success eliminating dyslexia type problems, when other methods can’t get better reaults for comprehension anyway, seems perverse. Here is a response to the articles criticising the clacks study:

      Click to access Torgerson%20article.pdf

      The only other thing to explain is that, as Stanovich says, no study is without faults which is why scientists use converging evidence.

  8. As I made no reference to my keenness, or otherwise, about finding out how Dr. Grant’s results were obtained, your off the wall implication that I do not care about children’s literacy has no grounding. It is simply irrelevant to any argument arising from my comment, so let’s put that behind us first. Neither can I find anything in my comment asserting that early readers ‘need’ prediction. So that’s another matter we can drop by the wayside.

    The link that you supplied does not have any bearing on the notion of converging evidence, but what I would say on that score is that converging evidence is useful, so long as the individual pieces of evidence used are sound. Where a study is not sound it cannot contribute to a valid body of converging evidence (multiplying by 0 can only ever give 0). The Clackmannanshire longitudinal study, for instance, was not reviewed in the Torgerson review of the literature (2006) on synthetic phonics because it did not meet the research standards they were applying.

    The link you gave took me to a paper familiar to me. I agree that it is an interesting piece of work. I would ask you to look at that paper again and think about why it includes the following:

    “The development of orthographic knowledge appears to be not entirely parasitic on decoding ability” p186

    “the third analysis revealed that orthographic knowledge predicted a significant amount of variance in orthographic learning, over and above the contribution of target decoding accuracy” p196

    I think you are getting confused about orthographic learning and decoding. The self-teaching hypothesis depends on the reader decoding novel words correctly and then remembering their exact spellings, even when there are alternative spellings possible within the ‘alphabetic code’. This is orthographic knowledge and it is word-specific. While not wanting to understate the importance of phonic decoding let’s be careful with the detail.

    1. My comment on attitudes to Dr Grant’s was indeed irrelevant to the discussion but interesting to me, for the reasons I outlined.
      I ‘m sorry if I presumed you were arguing for the prediction of mixed methods because you are critical of SSP, it is not an unreasonable assumption but I stand corrected.
      Stanovich goes to some trouble to explain that all individual pieces of evidence are inevitably flawed so I suppose it depends what you mean by ‘sound’. I suspect you are setting the bar unrealistically high. I think you are quite keen to discredit the Clacks study with an implication of unreliability. Once again I’d point out that fascinating lack of interest in HOW such brilliant results were achieved for so many children that were in schools that had previously been so much less successful. The results must have happened in some way. If we set your own interests and motivations aside, the anti SSP literature is very keen to find ways to discredit the Clacks study but I never read about any curiosity to find out (if not SSP) what did go so well for all those children.
      Finally, although a detailed discussion of the paper I linked to is interesting it was not the point of my link. My whole blog was to explain that it is totally ridiculous to look at what can summised from one paper alone. Each adds one extra and sometimes very small piece to the jigsaw. The link was to give an example of the sort of jigsaw pieces there are relevant to the issue you raised, not to ‘prove’ anything. As I said, to do justice to the wideranging and converging research on this area would take a vast amount of time I do not have.

  9. You seem determined to regard scepticism and a questioning attitude as an obstacle to making progress in teaching methods, as if being eager, positive and upbeat makes one right. This isn’t true. It is essential that we interrogate the ‘evidence’ and use research standards. There is a responsibility to point out flaws which make conclusions insecure. Where there are impressive improvements in children’s progress it is essential that the research is able to accurately and dispassionately account for it without making unwarranted assumptions.

    It is actually a responsibility to be critical of SSP when it is being pushed so relentlessly by government, programme writers, publishers and the rest.

    Part of the underlying problem is the false dichotomy set up between SSP and ‘mixed methods’. It appears that the SSP side takes an entrenched for fear of ‘mixed methods’, and therefore they fear finding out what SSP cannot do. Finding out what SSP cannot do is essential to accurate knowledge of what reading really is. The Stanovich paper is useful to that task, but only if you are prepared to look at it straight.

  10. It was having a questioning attitude that explains why I didn’t just accept the orthodoxy in the vast bulk of schools and looked at the evidence for myself.
    In fairness I’m sure we both have a very questioning attitude 🙂

  11. Why no mention of the NFER latest report into the phonics screening test? It found 60% of teachers of reading in England claimed to use synthetic phonics first and fast. However, their responses to questions about the use of other methods contradicted this. NFER concluded the majority of teachers recognised the importance of phonics instruction but combined it with other methods.

    Click to access Evaluation_of_the_phonics_screening_check_second_interim_report_FINAL.pdf

    Dr Grant’s study, although widely publicised, was too small to have any significance. This tweet from Ben Goldacre (“Bad Science”) sums it up:

    @bengoldacre “Teachers. If a non-randomised study of 30 kids in one class counts as significant evidence, your sector is broken.”

    1. Thanks for commenting. I very much do mention that NFER study in the blog, saying precisely your point, that most teachers do not teach SSP, but mixed methods. I them go on to explain why this is a problem.
      I also think you have misunderstood the whole point of my post in your second comment. My whole post is to explain how one small study could never be significant evidence. I know that Dr Grant is keen that, given the success she had, she can get funding for the sort of study Ben Goldacre would like to see more of.
      However, we do not await this possible study to be able to decide on the efficacy of SSP or to appreciate just how contradictroy to converging evidence current use of mixed methods is – as my blog outlines.

  12. “So, given the number of times I have read the assertion that there is no evidence in favour of SSP (systematic synthetic phonics) to teach reading”

    I don’t recall reading this assertion and I have recently done a fair bit of reading around the subject.

    Is it possible that this is actually is a myth.

    I know also that Ben is a genius and a guru and I am not, but I have a bit of a problem with this statement….

    “If a non-randomised study of 30 kids in one class counts as significant evidence, your sector is broken”

    What a silly thing to say. I believe a non randomised study of 30 kids in one class provides very significant evidence of what is happening in your class of 30 kids. Indeed it is my class of 30 kids that I am teaching and I need to understand.

    Clearly a non randomised study of 30 kids in one class would not give evidence that would be generalizable across the whole country in all contexts etc but for me this is not an issue.

    For me, taking the results of a large scale randomised trial and finding out about the sample in general and then applying this to Sydney who sits in my class but isn’t represented by your sample is an issue.

    If my son or daughter has a splinter, I look at their hand and identify it thus. I then use tweezers to remove it. I don’t go looking for large scale randomised trails about the efficacy or otherwise of different approaches to removing splinters from the hands of children. I just do it. As it is with much of teaching. I look at the kid, decide what needs to be done and do it. It’s not rocket science.

    I explain/demonstrate. I then ask the kid to explain/demonstrate. If they can we move on, if they can’t we repeat the process with some changes of approach possibly. And we go on until they can explain/demonstrate. People do tend to know what works, because it works.

  13. Reblogged this on and commented:
    For those who have questions about the evidence that phonics works.

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s