By Neil Almond

In my previous article on spelling, Let’s Toark Abought Spelling: Palt Tough, I offered readers some core conceptual knowledge when it comes to understanding some principles of the alphabetic writing system and how this relates to spelling. Those fundamentals are:

1) A sound can be represented by 1,2,3 or 4 letters.

2) One sound can be represented with different spellings.

3) The same spelling can represent different sounds (known as code overlap).

In this final article, I want to talk about how we can get students really thinking about spelling sounds that they know multiple representations of. Most common errors, particularly at the end of KS2, come from when students are required to choose the spellings they know for a particular sound and what can be done to potentially increase the likelihood of students becoming good spellers.

One theory of how researchers believe that we become good spellers is through a process known as statistical learning. Statistical learning is a big-data game. Like the social media and internet giants that analyse searches, likes, posts and browsing habits in the background (usually done on servers owned by the companies) to build up profiles of ourselves to serve us with highly specific targeted advertisements, our brain works in a similar fashion when it comes to reading, and therefore, by extension, when it comes to spelling.

Becoming a highly skilled reader is big-data for people. The data we need to do this is not what our friends and family post and like over the internet, but rather the successful decoding of text.

Every time a word is read correctly, in the background, much like the servers of tech companies, our brains are constantly and without conscious awareness, scanning for statistical properties. Which statistical properties of English orthography do we subconsciously process?

We learn about patterns within words, for example that <y> at the end of the word is likely to spell the /ee/ sound. Patterns that could occur e.g. ‘florp’ could be a genuine spelling of a word should the word exist. But we also, and equally important, learn which probabilities are highly unlikely or impossible to occur within English orthography. For example, a true word is unlikely to be spelt ‘xplx’.

Linked to this is the idea of orthographic redundancies. Players of the internet game ‘Wordle’ will have experienced this redundancy property first-hand. Assuming you had correctly identified 4 of the letters ‘WORR_’ the statistical properties of what could be and what could definitely not be the final letter come to light. While there are 26 letters in the alphabet, we do not have to go through each of these letters individually to find the correct word.

Instinctively, we know that ‘x’, z’ and ‘u’ are unlikely to be correct. A bit more thinking and we can ascertain that <rr> of the spelling /r/ further reduces the possibilities of what the final letter could be (y). It is important to note that none of this will have been explicitly taught. It is the big-data game of learning to spell, in action.

We can use the above to our advantage in the following ways in making sure students become good spellers:

1) Feed the data machines (our brains) with plenty of text to read accurately. Reading mileage is important, but unlikely to be enough.

2) Do not provide incorrect ‘data’ to our brains. In order to learn these orthographic patterns we need a lot of correct data. Asking students to select words from a list where some have been spelt incorrectly will feed the brain bad data and could possibly interfere with the statistical learning process.

3) Jump start the statistical learning process by teaching the likelihood of a certain sound-spelling correspondence being used.

As mentioned previously, I see spelling lessons not as time to teach students to spell certain words, though there will be some of this, but to lift the veil of complexity of the English language. This means teaching all the sound-spelling correspondences. These should be organised in the following way:

“The first category comprises the major spelling alternatives beyond the basic-code level. Spellings that did not qualify as major were classified in one of three ways: as a special group (few words, improbable spelling, high frequency in print),…or omitted (singular spelling, low frequency in print).” (McGuinness, 2004)

Categorising sound-spellings in this way is useful. It helps us to decide which word needs to be learnt, to spell like those in the category of the special group ( <ea> spelling of /ae/break, steak)  and conversely which sound-spellings can be avoided ( <pph> as /f/ in sapphire) due to their frequency in print. Break and steak are likely to have high frequency in print but I can only think of a handful of examples where <ea> spells /ae/.

Likewise, sapphire has a low frequency so teaching <pph> as /f/ has a high time-cost benefit. These categories also help provide a ‘ceiling’ for what to include in the major spelling alternatives. While <pph> is a spelling alternative of /f/, I would not teach it as such as it is only used in one word.

The major spelling alternatives, which make the up the largest group of spellings, are what I would want to teach in terms of spelling frequencies, and not as learning to spell individual words. For example, here are all the major spellings alternatives for /ae/.

ey a-e a ay eigh ei ai

Students need to know all of these so when they come to a word like ‘investigate’ they stand the chance of spelling the /ae/ sound accurately. Revisiting the end of the of the part 2 of this series, we want students to realise that while they are not sure on how to spell the word investigate, they can spell many sounds of the word.

Teachers need to reinforce this idea by asking students which sound they are not sure how to spell. Assuming that the student is struggling with the /ae/ sound, the next step is to ask students which spelling is most likely. To do this, it is not enough for students to know the sound-spelling correspondences of the sound, but to know which spellings are most likely to be the case.

a-e ai a ay ei eigh ey

Here is the same list but arranged in order of most likely to least likely. This may be something that readers with a high reading milage may implicitly pick up (though in my experience, something they find difficult to convey) but what I believe need to be taught explicitly.

Students who know all the sound-spelling correspondences for a sound and know the order from most likely to least likely can then make educated ‘guesses’ as to how to spell the sound they are unsure of in the word they are struggling with. Armed with this understanding, they could write down the four most probable spellings.





Statistical learning tells us that the third spelling, given the position of the /ae/ sound, is unlikely to be correct.





From here, students could use their visual memory to see which spelling ‘looks’ correct – another reason why exposing students to plenty of print, and reading being part of spelling instruction, is crucial for teaching students to spell.

This means that the vast majority of spelling instruction needs to move away from the spelling of individual words and instead move to teaching students the sound-spelling alternatives in order for most likely to least likely and ensuring that students know these.


N.B. A possible arrangement of these spelling alternatives by most to least likely can be found in Diane McGuinness’Why Our Children Can’t Read, And What We Can Do About It: A Scientific Revolution In Reading.


McGuinness, D., 1997. Why Our Children Can’t Read, and what We Can Do about it: A Scientific Revolution in Reading. Simon and Schuster.

McGuinness, D., 2006. Early reading instruction: What science really tells us about how to teach reading. MIT Press.


Neil was a classroom teacher for 5 years before leading Teach and Learning in a small academy trust. Now he is a Deputy Headteacher on Thornton Heath. He regularly blogs and speaks at educational events around the country.

Write A Comment