What “the science of reading” actually means
You have probably seen the phrase on a school newsletter or an app store listing. Here is what it refers to, what it does not, and which parts of it are settled.
Start with what it is not
It is not a teaching scheme. There is no box you can buy with “the science of reading” printed on the side, and no one owns the phrase. Any company can put it in an advert.
What it refers to is the pile of research built up over about fifty years on how children actually learn to read: reading scientists, cognitive psychologists, speech and language researchers, and a lot of classroom trials. Some of that research is settled to the point of being boring. Some of it is still argued over. This page tries to keep the two apart.
So when a school or an app says it follows the science of reading, the useful question is not whether they say it. It is what they do with children.
The five parts of reading
These came out of the United States National Reading Panel and have shaped teaching in English-speaking countries ever since. They grow together rather than in single file.
Hearing the separate sounds inside a spoken word: that “sun” starts with /s/, that “cat” is made of three sounds pushed together. No letters involved yet. This is done with ears only.
Which letters stand for which sounds, taught in a deliberate order rather than as they happen to come up. This is the part with the strongest evidence behind it, and the part most often skipped.
Reading smoothly enough that the effort of working out the words stops eating the attention needed to follow the meaning.
How many words a child knows the meaning of. A child can decode “brittle” perfectly and still have no idea what has just been said.
Understanding, questioning and remembering what the text said. Everything above exists to make this possible.
The one idea worth taking away
In 1986, two researchers published a short paper proposing something they called the Simple View of Reading[1]. It says that understanding what you read is the product of two things multiplied together: decoding (working out what the words on the page say) and language comprehension (understanding language when you hear it spoken). A longer study followed that tested the model against real children[2], and the idea has been used in hundreds of studies since [3].
Multiplied, not added. That is the whole point, and it is the part most summaries lose. If either number is zero, the answer is zero. If either number is small, no amount of the other one rescues it.
Why this explains the thing that worries you
A child who sounds out every word on the page and then cannot tell you what happened in the story is not failing at reading in some mysterious way. One side of the multiplication is strong and the other is weak.
The reverse happens too, and looks completely different: a child with a rich spoken vocabulary who loves being read to, but who stalls on the page, has the language and not the decoding.
Both children need help. They need opposite help. That is why “my child is behind at reading” is not one problem with one answer.
One honest caveat, made by a reading researcher writing about the model thirty years on: because the two halves are usually drawn as equal boxes, people assume they are equally simple to teach [3]. They are not. Decoding is a bounded job. Language understanding is not bounded at all, and it has proved much harder to shift.
What the numbers mean
Research about teaching gets reported in effect sizes, and they are worth two minutes of your time, because they are how you tell a real finding from a press release.
An effect size puts different studies on one scale. It asks: how far ahead did the group that got the teaching end up, measured in the spread of the scores rather than in marks or reading ages? By rough convention, 0.2 is small, 0.5 is moderate, 0.8 is large. An effect size of 0.4 means the average child in the taught group finished ahead of about two-thirds of the children in the comparison group. Real, useful, not miraculous.
Now the actual figures. The peer-reviewed write-up of the National Reading Panel's phonics review pooled 66 treatment-and-comparison pairs from 38 experiments and found an overall effect of 0.41 for systematic phonics against unsystematic or no phonics [4].
Systematic phonics against unsystematic or no phonics, pooled across 38 experiments. A moderate effect.
When the teaching began in kindergarten or first grade. Kindergarten on its own was 0.56, first grade 0.54.
When it began after first grade. Still real, clearly smaller. Timing is the single biggest lever in the whole review.
The effects were biggest where you would hope: 0.67 for reading regularly spelled words and 0.60 for reading made-up words, the test of whether a child can tackle something never seen before[4]. That is what phonics is for.
And the review is honest about its limits, which is a good sign in a piece of research. Phonics did not help older low-achieving readers whose difficulties included cognitive limitations, where the effect was 0.15 and not distinguishable from nothing [4]. Phonics is a strong tool for one specific job. It is not a treatment for every reading problem a child can have.
Where this leaves reading apps
We sell a reading app, so treat the next paragraph as evidence against our own interests, and check the source.
A 2025 review pooled 119 studies from 105 papers on educational technology used for literacy with children in kindergarten through fifth grade, published between 2010 and 2023. The effects were positive: 0.33 for decoding, 0.30 for language comprehension, 0.23 for reading comprehension, 0.81 for writing[5].
Then the reviewers did the thing that separates a careful meta-analysis from a marketing chart. They pulled out only the studies using standardized tests, the ones not written by the people who built the software, and every figure dropped: 0.23 for decoding, 0.12 for language comprehension, 0.14 for reading comprehension, 0.34 for writing [5].
Put the 0.33 next to the 0.41 that systematic phonics scored in the phonics review, and the 0.23 next to it as well[4]. Software lands somewhere below teaching, and further below it on the strictest tests. Be careful with that comparison, though: these are two separate reviews of different studies, not a trial that put an app and a teacher side by side. We did not find such a trial. If you see anyone quote a clean ratio between the two, they have invented it.
What we take from that is the reason Plimmery is built to sit alongside a grown-up rather than instead of one. It is also why nothing on this site says a child will learn to read from a screen.
Questions parents ask
Is the science of reading a teaching scheme I can buy?
- No. It is the accumulated research on how children learn to read, from cognitive psychology, linguistics and classroom trials. Anyone can say their product is built on it, which is why it is worth asking what a product actually does.
How much does teaching phonics in a set order help?
- Across 66 treatment-and-comparison pairs drawn from 38 experiments, systematic phonics beat unsystematic or no phonics by an effect size of 0.41, which counts as moderate. Starting early mattered: 0.55 when it began in kindergarten or first grade, 0.27 when it began later. [4]
Why can my child sound out every word and still not follow the story?
- Because understanding what you read depends on two things multiplied together: working out what the words say, and understanding language when someone speaks it. Strong sounding-out with weak language understanding still leaves a child who has read the page and missed the point. [1][2][3]
Can an app teach my child to read on its own?
- The honest answer is no. A review of 119 studies of software used for literacy in kindergarten through fifth grade found an effect of 0.33 on decoding, and 0.23 once only standardized tests were counted. Phonics taught by a person scored 0.41 in a separate review. Software helps. It does not replace a teacher or a parent. [4][5]
What is an effect size?
- A way of putting different studies on one scale so they can be compared. As a rough guide, 0.2 is small, 0.5 is moderate and 0.8 is large. An effect size of 0.41 means the average child who got the teaching finished ahead of roughly two-thirds of the children who did not.
Keep reading
- The order children learn to read in, and which parts of that order are settled.
- Teaching letter sounds, including why some letters are much harder than others.
- How Plimmery works, if you want the product rather than the research.
Sources
Every source here was read before it was cited. Follow any number in the text to land on its entry.
- [1]Gough, P. B., & Tunmer, W. E. (1986). Decoding, reading, and reading disability. Remedial and Special Education, 7(1), 6-10. doi:10.1177/074193258600700104
- [2]Hoover, W. A., & Gough, P. B. (1990). The simple view of reading. Reading and Writing, 2(2), 127-160. doi:10.1007/BF00401799
- [3]Catts, H. W. (2018). The simple view of reading: advancements and false impressions. Remedial and Special Education, 39(5), 317-323. doi:10.1177/0741932518767563
- [4]Ehri, L. C., Nunes, S. R., Stahl, S. A., & Willows, D. M. (2001). Systematic phonics instruction helps students learn to read: evidence from the National Reading Panel's meta-analysis. Review of Educational Research, 71(3), 393-447. doi:10.3102/00346543071003393
- [5]Silverman, R. D., Keane, K., Darling-Hammond, E., & Khanna, S. (2025). The effects of educational technology interventions on literacy in elementary school: a meta-analysis. Review of Educational Research, 95(5), 972-1012. doi:10.3102/00346543241261073