How Children Are Taught To Read

AFew questions in education have been argued over as long, or as bitterly, as how a young child should be taught to read. The dispute is usually reported as a straight contest between two methods. On one side stands systematic phonics, in which children are taught the relationships between letters and sounds in a planned sequence, and are given texts restricted to the correspondences they have already covered. On the other stands what is loosely called meaning-first instruction, in which children meet whole books early and are encouraged to use context, illustrations and sentence structure to work out words they do not recognise. The labels are convenient for journalists and for ministers, but they describe positions more than they describe classrooms. When 1,100 primary teachers in England were surveyed in 2016, most described their own daily practice as a mixture of the two, and only a small minority claimed to follow either programme without departing from it.

BThe case for phonics rests on a claim about what an alphabetic script actually is. Written English encodes speech sounds, imperfectly but systematically, and a child who has not grasped that principle is left treating every word as a picture to be memorised. Marta Devlin, whose trial in 2014 followed 4,200 five-year-olds across ninety schools, reported that pupils given a structured phonics programme were, after two years, reading unfamiliar words aloud more accurately than their peers; the gap she measured was equivalent to roughly four months of additional progress. Devlin is unusually careful about what that figure does and does not cover. Her outcome measure was word reading, not understanding, and she has said in print that the two should never be reported as though they were the same thing. The benefit, she adds, was largest among the children who arrived at school with the smallest spoken vocabularies.

CThe objection raised by her critics is not that phonics fails, but that it is routinely asked to do more than it can. Paulo Ferreira, who tracked a single cohort of 900 children from the age of six to the age of eleven, found that the early advantage in decoding had narrowed sharply by the fourth year of schooling, though it had not disappeared. His explanation was that comprehension rests on knowledge the reader brings to the page - of vocabulary, of the world, of the ways in which written sentences behave - and that such knowledge accumulates slowly, and very unequally between households. A child may pronounce every word of a passage about irrigation correctly and still take nothing from it. Ferreira has never argued that the early teaching was wasted; his position is that its benefits are conditional on the instruction that follows, and that where nothing follows, the advantage quietly drains away.

DRead carefully, the two positions are less opposed than the public quarrel suggests. Both researchers accept that letter-sound correspondences must be taught deliberately rather than absorbed; both accept that teaching them is not sufficient. What divides them is a matter of sequence and emphasis. Devlin holds that the decoding phase should be short, intensive and finished early, so that attention can move on to everything else; Ferreira maintains that the same hours yield more when knowledge-building and decoding run alongside each other from the first term. Neither has tested the other's preferred arrangement, and no trial has yet set the two sequences against each other within one group of schools. The disagreement is genuine, then, but considerably narrower than its reputation, and it concerns timing rather than substance.

EPolicy has moved faster than this caution would warrant. In 2012 England introduced a screening check taken by every six-year-old, in which a child reads forty items aloud, half of them invented words such as 'blad' or 'strope'. The invented items were the point: a real word can be recognised from memory, whereas a word that does not exist can only be sounded out, so the check was meant to show whether decoding itself had been mastered. Within four years the proportion of pupils reaching the expected standard rose from 58 per cent to 81 per cent. Whether the underlying skill had improved by anything like that margin is harder to establish, since schools knew the format, rehearsed it, and could lift scores by teaching towards it. Devlin herself has warned that such a check serves as evidence only for as long as it is not being prepared for.

FWhat most researchers now accept is a division of labour rather than a victory for either camp. Decoding is finite: there is a limited stock of correspondences to learn, and for most children they can be covered within about two years. Comprehension is not finite. It grows for as long as a person goes on reading, and it is bound up with everything else that person happens to know, which is why it resists tidy programmes. The practical consequence is that the two should not be competing for the same slot in a timetable. This is also where measurement is at its weakest. Decoding is captured well by standardised tests, because it can be scored objectively and quickly; the slow accumulation of knowledge that Ferreira describes has no equivalent instrument, and what cannot be measured conveniently tends to be neglected in the accounts that reach ministers.