Skip to content

Research Library · Study page

How do children learn to read? "Ending the Reading Wars" (2018), explained

Three reading researchers sum up how children learn to read. Kids must be taught how letters stand for sounds. But phonics is not all of reading.

Study Facts

Castles and others, 2018

Expert review

Question
How do children learn to read?

What it found

Most children must be taught how letters stand for sounds. But phonics is not all there is to learning to read.

EvidenceExpert view

Three researchers' summary of other studies. No pooled numbers of its own.

What it is
A summary of other research
Looked at
Decoding, fluent reading and understanding text
By
3 researchers, 3 universities
Paid by
Research councils in Australia and the UK, and the Leverhulme Trust
Conflicts
Not stated

Anne Castles, Kathleen Rastle and Kate Nation (2018). “Ending the Reading Wars: Reading Acquisition From Novice to Expert.” Psychological Science in the Public Interest, 19(1), 5–51. Read the free full text · Publisher's version
Free to read on the publisher's site. We read the authors' accepted manuscript, so a word here and there may differ from the printed version.

Last reviewed November 17, 2026 · Used in: Phonics or whole language: which works better?

Go deeper

The study in eight short parts. Tap a part to open it.

1

What is the study?

A long review by three reading researchers, from Australia and the UK. They call it a "tutorial review": a guided tour of the science of learning to read, from first steps to skilled reading. It is not a new study. It is not a search of every paper on the topic either.

Its title says what it hopes to do: end the "reading wars".

2

Why this matters to you

For a long time, a "pendulum" has swung between two ways of teaching, the authors write. In phonics, the sounds that letters make are taught clearly. In whole language, the child discovers meaning through experiences in a setting rich in reading and writing. The fight goes back over 200 years.

3

What was measured?

No children were tested. The authors sum up research on three steps:

  • Learning how letters stand for sounds, so a child can sound out words.
  • Reading words quickly and smoothly.
  • Understanding text.
4

Who was measured?

No one, directly. The paper measured no children of its own. It draws on many earlier studies.

It is about typical readers in ordinary classrooms. Children with learning difficulties are "beyond our scope", the authors write.

5

How was it measured?

  • A summary of other research. It gives no pooled numbers of its own.
  • Broad, not deep. "Given its breadth, our review is of course limited in detail", they write.
  • Boxes beside the main text take on common myths and teaching questions.
6

The results

Letter sounds must be taught. Learning how letters link to sounds is "necessary and non-negotiable" in an alphabet like English, they write. Phonics teaching does this best. Most children do not work it out on their own. Preschool children were taught to read fat and bat. Then they were asked if a new word said "fun" or "bun". Across more than 80 children, "virtually none" could tell.

Reading is not like talking. A child who hears rich speech will almost certainly learn to talk. "The same cannot be said for reading", the authors write.

Phonics opens many words. Learning one word can help with others. Knowing vet can help a child learn van. Learning words one by one, without the code, is "hard to imagine" scaling up to the tens of thousands of words adults know.

Three-cueing. In 1967, Ken Goodman called reading a "psycholinguistic guessing game". Later, a "three-cueing" approach spread. In it, beginners use meaning, grammar and letter clues together to guess a word. The authors see Goodman's mistake as teaching beginners what skilled adults do. They compare it to teaching piano by "putting a child in front of a Tchaikovsky score". The review describes three-cueing but reports no test of it.

Balanced literacy. The name is often used for programs with "a bit of everything", they write. Such programs typically have limited phonics with no set order.

Phonics is not all of reading. That reading is more than phonics is "a statement of the obvious to any reading scientist", they write.

  • Reading a lot matters. In one study of grade 5 children, those who read least met about 60,000 words a year. Those who read most met more than four million.
  • Spoken language matters. Words, grammar and story skills when a child starts school predict later understanding of text.

Findings that cut against a phonics-only view:

  • Teaching a small number of sight words "has its place" alongside phonics.
  • There is "a good argument" for decodable books at the very start. "Beyond the initial stages of reading, however, the case for decodable books weakens."
  • There is "not yet sufficient evidence" that one kind of phonics, called synthetic, should be preferred over another, called analytic.
  • Rewards for reading "may paradoxically have a negative impact" on a child's wish to read.
  • "Many outstanding questions remain" about how best to teach phonics in class.

England. State schools in England must teach phonics in a planned order when children start school. The share of children passing its phonics check rose from 58% in 2012 to 81% in 2016 and 2017. National tests at age 7 show small gains in understanding text. But "it is not possible to conclude" that phonics caused them, the authors write. One study found a longer-term gain at age 11 for children likely to start school as struggling readers, such as children new to English or from poorer homes.

7

Why this matters to you, again

  • Ask how letter sounds are taught. Most children "must be taught explicitly", the authors write.
  • Phonics first, not phonics only. They want "balanced instruction" to mean teaching time that shifts as the child grows.
  • Read a lot, and talk a lot. Seeing many words builds smooth reading. Spoken language builds understanding.

The paper's own limits:

  • Typical readers only. Children with learning difficulties are beyond its scope.
  • Broad, not deep. It is a tutorial review, not a search of every study.
  • England. It cannot say that phonics caused England's gains.
  • Three-cueing. It describes the approach but reports no test of it.
8

Who did the study, and who paid

Who did it. Three researchers. Anne Castles is at Macquarie University in Australia. Kathleen Rastle is at Royal Holloway, part of the University of London. Kate Nation is at the University of Oxford.

Who paid. Research councils in Australia and the UK paid for the work. So did the Leverhulme Trust.

Conflicts of interest. Not stated in the copy we read.