Blog/17 September 2026
Mind your ⟨p⟩s and ⟨q⟩s
“A place for everything, and everything in its place.”Proverb
Ask someone to lay the table for Sunday lunch and a surprising amount goes right without a word of instruction. The forks go on the left. The knives go on the right. Nothing ends up balanced on the gravy boat. (We all use a gravy boat, right?) Nobody has handed them a laminated diagram or signed them up for a thirty-minute Zoom lecture on table etiquette. They have simply eaten enough dinners to know where cutlery lives.
Now add soup to start and crumble to finish. The results become more inventive. The soup spoon is tucked in snugly beside the plate, the dessert spoon keeps it company, and the small fork and the large fork appear in whatever order they came out of the drawer. Strictly speaking, nothing is on the wrong side. Every piece is where cutlery is allowed to be. What has gone astray is which piece sits next to which, and that depends on what is being eaten and when. Working from the outside in is not something most people pick up by watching. Somebody usually has to explain it, often at a wedding, and often halfway through the starter.
Cutlery, in other words, runs on two kinds of rule. One says where a thing may go. The other says what it may go next to. The first is learned by looking. The second takes rather longer.
English spelling is laid out in the same way. Letters have their side of the plate: ⟨ck⟩ never starts a word, and ⟨j⟩ never ends one. They also have their neighbours: ⟨ck⟩ follows a single vowel letter, which is why back has it and beak does not. The research suggests children set out the first kind of rule largely for themselves, and are still shuffling the second for years.
The spelling rules themselves are familiar, and most teachers already teach them. What is less often noticed is that each one is doing both jobs at once.
Where letters stand
Four familiar rules
- ⟨ck⟩ never begins a word. There is no ckat. No ordinary English word breaks this rule.
- ⟨j⟩ never ends a word. That is why badge and bridge exist: when /dʒ/ comes at the end, English needs another spelling for it. The only exceptions are borrowings such as raj and hajj.
- ⟨y⟩ does ⟨i⟩’s job at the end of a word. English has cry, my and happy, and ⟨i⟩ appears at the end mainly in borrowed and shortened words: taxi, ski, mini, spaghetti, chilli. ⟨v⟩ behaves the same way, which is why have and give carry an ⟨e⟩. The exceptions are shortenings such as rev.
- ⟨ff⟩, ⟨ll⟩, ⟨ss⟩ and ⟨zz⟩ double at the end of a short word, straight after a single vowel letter: cliff, bell, miss, buzz. The exceptions are real but few: if, of, pal, yes, bus, gas, plus, this, quiz.
Two jobs at once
Two kinds of rule inside one
Linguists call these rules graphotactics: the conventions governing which letters, and which sequences of letters, a writing system allows. They are the written counterpart of phonotactics, the rules on how sounds combine in speech, and the two do not always agree. English speech happily ends a word with /v/ or /dʒ/; English spelling will not end one with ⟨v⟩ or ⟨j⟩.
Graphotactic rules come in two kinds. A positional rule says where a letter may sit. ⟨ck⟩ may not come first. ⟨j⟩ may not come last. A child only needs to know the letter and its place in the word.
A contextual rule says which letters may sit next to which. ⟨ck⟩ is used straight after a single vowel letter (back, neck), but not after a vowel digraph or a consonant (beak, bank). ⟨dge⟩ follows a single vowel letter (badge), and ⟨ge⟩ follows anything else (page, bulge). The doubled ⟨ll⟩ of bell depends on the single ⟨e⟩ before it, which is why feel has one ⟨l⟩. To apply a contextual rule, a child has to look at the neighbouring letters, not just at the position.
Fig. 1One spelling, two rules
Position
Where in the word may it sit?
ckat→catbackbaj→badge
Context
What does it sit next to?
back·beak·bankbadge·page·bulgebell·feel
In page the ⟨e⟩ after the consonant also marks the vowel, so the choice of ⟨ge⟩ depends on more than the letter before it.
Most of the rules above combine both. Where letters are allowed to stand is only half of what children need; what they are allowed to stand next to is the other half.
The evidence
What children learn without being told
Samara and Caravolas (2014) tested this directly. They invented new constraints, so no one could already know them, and built strings of letters that followed those constraints. The participants were 137 English-speaking children with an average age of seven and a half, and 113 adults. They were shown the strings while doing an unrelated task (spotting a red letter), and were never told there was a pattern. Afterwards they judged whether new strings looked as if they belonged.
Both groups learned both kinds of constraint above chance. The difference in how well they learned them was large. Accuracy was measured with d′, where zero means guessing.
Fig. 2How well each kind of rule was learned (d′)
Across everyone, positional rules scored 0.97 and contextual rules 0.19. Shorter exposure was 144 trials; longer was 288.
SourceSamara and Caravolas (2014), table 1. Values as published; bars drawn to a scale of 0 to 1.25.
The children’s rows matter most. Doubling the exposure improved their grasp of position. It made no difference to their grasp of context.
Real words
The same effect with real words
Artificial letter strings are one thing. There is also evidence that children apply positional knowledge to real spelling.
Children judged baff more word-like than bbaf, although nobody had taught them that English does not double letters at the start of a word.
Children were taught to read made-up words with doubled letters. Words with an illegal double at the start, such as RRUG, took longer to learn and were remembered worse. The more advanced children often wrote them with the double moved to the end, where English allows it.
Children read new words in stories. When a double had been placed where French does not allow it, straight after a consonant, children spelled it in a legal position a quarter of the time (guprrane became gupprane). A double at the very start of a word was not shifted, and was in fact remembered well, so the effect is real but not universal.
In each study, knowledge about letter positions changed what the children stored, without anyone telling them to change it.
The slower half
Context is learned too, but slowly
Children do learn contextual patterns in the end.
- Hayes, Treiman and Kessler (2006): seven-year-olds already used the preceding vowel to choose between spellings such as ⟨ch⟩ and ⟨tch⟩ at the end of made-up words. That sensitivity kept growing through the school years and into adulthood. The vowel’s letters mattered more than its sound.
- Singh, Wonnacott and Samara (2021): six- and seven-year-olds picked up a rule in which ⟨s⟩ or ⟨ss⟩ depended on the vowel before it, simply by being exposed to it. Children who were also told the rule outright did better still.
Letters that change their neighbours
The same principle, read the other way
Contextual rules do not only decide which letter to write. Many decide how a letter is read, and the spelling then bends to protect that reading. Three patterns show this especially clearly.
Soft ⟨c⟩ and ⟨g⟩
⟨c⟩ and ⟨g⟩ soften before ⟨e⟩, ⟨i⟩ and ⟨y⟩: cent, city, cycle; gem, giant, gym. Elsewhere they are hard: cat, cot, cut; gap, got, gum.
The spelling keeps faith with this in three ways:
- /k/ before those vowels is written ⟨k⟩: keep, kit, sky.
- A ⟨u⟩ is added to keep a consonant hard: guess, guide, biscuit.
- An ⟨e⟩ is kept to keep one soft: noticeable, courageous.
⟨c⟩ is very reliable; ⟨g⟩ much less so, with get, give, girl and gift all hard. The reason is history. Old English used ⟨c⟩ for both /k/ and /tʃ/, and never for /s/. The soft values came from Latin through French, and after 1066 French-influenced scribes brought the convention with them. Native words were respelled to fit: /k/ before ⟨e⟩ and ⟨i⟩ moved to ⟨k⟩ (cat but kitten, cow but kine), and /tʃ/ moved to ⟨ch⟩ (Old English cinn, now chin). That left soft ⟨c⟩ almost entirely to words from Latin and French, which is why it is so dependable. ⟨g⟩ had no second letter to move to, so many words of Germanic origin kept a hard ⟨g⟩ before ⟨e⟩ and ⟨i⟩: begin, get, give. A few took the French device of an added ⟨u⟩ instead: guess, guest, guilt.
Fig. 3What the next letter does to ⟨c⟩ and ⟨g⟩
Select a pair to see its words.
How it was countedAcross 30,945 common English words, with proper nouns removed, whose letters had been matched to their sounds. Each cell counts the words in which c or g, standing alone as a spelling, is followed by that letter. Words where the letter is silent or part of a longer spelling, as in exceed, are left out, so a few words fall outside both counts.
What /w/ does to the vowel after it
/w/ is made with rounded lips and the back of the tongue raised. During the 17th century that rounding spread to the vowel that followed it (Schlüter, 2012). The spelling was not changed to match, so three spellings are now read differently after /w/ from anywhere else. It is the sound that matters, not the letter: ⟨wh⟩ spells /w/ and ⟨qu⟩ spells /kw/, and both have the same effect.
- ⟨a⟩ says /o/, not /a/: was, want, swan, watch, what, squash, quality; compare cat and pant. The rounding did not happen before /k/ or /g/, and that is where nearly all the exceptions are: whack, wag, wax, quack, swagger. Of the 33 exceptions counted, 26 have /k/ or /g/ straight after the vowel; the rest include swam and wham.
- ⟨ar⟩ says /or/, not /ar/: war, warm, ward, reward, wharf, quarter; compare car, farm, hard. warm is one of the words Schlüter gives for the 17th-century change. Stressed exceptions are rare (quark). In the unstressed ending ⟨ward⟩ the vowel shrinks to a schwa instead: forward, backwards, awkward.
- ⟨or⟩ says /ur/, not /or/: word, work, world, worm, worth, worse; compare fork, storm, north. This one has a different history, ending in the vowel of bird and hurt, but for a reader the effect is the same. The exceptions are a handful: worn, sworn, quorum, whorl.
The pattern is dependable for reading, less so for spelling. /wur/ is usually written ⟨wor⟩, but not always: were, swerve, whirl, whirr, swirl, twirl, squirm, quirk.
Fig. 4After /w/: the pattern and its exceptions
How it was countedWords where the spelling follows /w/, whether written w, wh or qu, across the same 30,945 words. Examples are the commonest words in each group.
⟨q⟩ and its ⟨u⟩
⟨q⟩ is followed by ⟨u⟩: queen, quick, equal. This is the most reliable contextual rule in English. Of the 410 words in the same list that contain ⟨q⟩, all but one follow it with ⟨u⟩; the exception is burqa. Names such as Iraq and Qatar break the rule too.
The ⟨u⟩ is not always sounded, though. In later borrowings, mostly from French, ⟨qu⟩ and ⟨que⟩ spell plain /k/: conquer, mosquito, unique, cheque. The ⟨u⟩ stays even where it says nothing.
This too is a scribal habit. Latin wrote /kw/ as ⟨qu⟩, and French inherited it. Old English had little use for ⟨q⟩ and wrote ⟨cw⟩ instead: cwen for queen, cwic for quick. After the Norman Conquest, French-influenced scribes wrote ⟨qu⟩ more and more, and by about 1300 ⟨cw⟩ had gone.
Fig. 5What ⟨q⟩ and its ⟨u⟩ say
How it was countedWords containing q in the same list, sorted by what the spelling that begins with q says. Examples are the commonest words in each group.
Learned slowly, here too
Readers learn these patterns the same way spellers do: slowly. Treiman and Kessler (2019) found that children’s use of the letter after ⟨c⟩ and ⟨g⟩ grew gradually from age six to adulthood, and that even adults relied on it less than the vocabulary would justify. The authors suggest this is a place where explicit instruction may be needed. An earlier study tested the /w/ pattern itself, with made-up words such as squant and quarm: six-year-olds were already influenced by it to some extent, and the influence kept strengthening until about the age of ten or eleven (Treiman and colleagues, 2006).
And finally
Where teaching time goes
Taken together, the studies suggest this: reading appears to give children positional patterns largely for free. They supply contextual patterns much more slowly, and explicit teaching adds to what children pick up alone.
That has a practical consequence for how the familiar rules are taught. A child who writes ckat is rare. A child who writes bak, baje or fel is common, and each of those errors is a failure of context, not position. So when teaching ⟨ck⟩, ⟨dge⟩, ⟨tch⟩ and doubled final consonants, the teaching time is best spent on what comes before the spelling: one vowel letter or two, a vowel or a consonant. The position half can mostly be pointed out and left.
Some limits should be stated. Most of these studies use invented words and short exposure, and none of them is a classroom trial of one teaching approach against another. They show where learning comes easily and where it does not. They do not show which way of teaching the harder half works best.
I have not seen a teacher-facing article make this argument. The research to support it has been available for more than a decade.
Sources
References
- Cassar, M. and Treiman, R. (1997) The beginnings of orthographic knowledge: children’s knowledge of double letters in words. Journal of Educational Psychology, 89(4), 631–644.
- Hayes, H., Treiman, R. and Kessler, B. (2006) Children use vowels to help them spell consonants. Journal of Experimental Child Psychology, 94(1), 27–42.
- McArthur, T. and McArthur, R. (eds) (2005) Concise Oxford Companion to the English Language. Oxford: Oxford University Press. Entries for C, G and Q.
- Pacton, S., Sobaco, A., Fayol, M. and Treiman, R. (2013) How does graphotactic knowledge influence children’s learning of new spellings? Frontiers in Psychology, 4, 701.
- Samara, A. and Caravolas, M. (2014) Statistical learning of novel graphotactic constraints in children and adults. Journal of Experimental Child Psychology, 121, 137–155.
- Schlüter, J. (2012) Early Modern English: phonology. In A. Bergs and L. J. Brinton (eds) English Historical Linguistics: An International Handbook, volume 1, chapter 37. Berlin: De Gruyter Mouton.
- Singh, D., Wonnacott, E. and Samara, A. (2021) Statistical and explicit learning of graphotactic patterns with no phonological counterpart: evidence from an artificial lexicon study with 6–7-year-olds and adults. Journal of Memory and Language, 121, 104265.
- Treiman, R. and Kessler, B. (2019) Development of context-sensitive pronunciation in reading: the case of ⟨c⟩ and ⟨g⟩. Journal of Experimental Child Psychology, 182, 114–125.
- Treiman, R., Kessler, B., Zevin, J. D., Bick, S. and Davis, M. (2006) Influence of consonantal context on the reading of vowels: evidence from children. Journal of Experimental Child Psychology, 93, 1–24.
- Wright, D.-M. and Ehri, L. C. (2007) Beginners remember orthography when they learn to read words: the case of doubled letters. Applied Psycholinguistics, 28(1), 115–133.
Exceptions listed are the common ones, not a complete set, and the figures are as reported in the published papers. To see every spelling English allows for a sound, try the spelling chart; the word list builder finds words for any sound or spelling.
