The research says ask questions. Here is exactly which ones.

There is a version of reading aloud where the adult does all the work. You do the voices, you turn the pages, you get to the end, and the child lies there and receives it. It is lovely and it is also the least useful version.

The research on the alternative has been sitting in plain sight for nearly forty years, and the instruction at the heart of it is almost rude in its simplicity: read less, and ask more.

What the evidence actually says

The technique is called dialogic reading. The What Works Clearinghouse, the evidence review body run by the US Department of Education, describes it as a practice in which “the adult and the child switch roles so that the child learns to become the storyteller with the assistance of the adult who functions as an active listener and questioner”.

Its review found four studies that met its evidence standards and one that met them with reservations, covering over 300 preschool children, most of them from economically disadvantaged families. The verdict:

The improvement index on oral language averaged +19 percentile points, with a range from -6 to +48.

Two things to take from that pair of numbers. Nineteen percentile points for changing how you talk during a book you were going to read anyway is a very large return for no money and no extra time. And a range that includes -6 means it did not work everywhere, which is what an honest effect looks like.

The phonological processing result matters just as much. Dialogic reading does not teach a child the sounds in words. It is not a phonics programme and nobody should use it as one.

The vocabulary number

A 2008 meta-analysis by Mol, Bus, de Jong and Smeets in Early Education and Development pooled 16 studies comparing dialogic reading with ordinary shared reading. Looking specifically at expressive vocabulary, the words a child can actually produce, across nine studies and 322 children, it found an effect of d = 0.59, with a confidence interval from 0.44 to 0.75.

That is a moderate effect, and it is on the thing parents care most about: not whether the child recognises a word but whether they use it.

The two things to learn

The method is two acronyms. They look like jargon and they take about four minutes to learn.

CROWD: the five kinds of question

These are the Clearinghouse’s own descriptions.

• Completion. The child fills in the blank at the end of a sentence.
• Recall. You ask questions about a book the child has read.
• Open-ended. You encourage the child to tell you what is happening in a picture.
• Wh-. You ask who, what, where, when and why about the pictures.
• Distancing. You relate the pictures and words in the book to the child’s own experiences outside it.

Completion prompts are the easy entry point, because a child who knows a book will finish your sentence without being asked twice. Distancing prompts are the ones that do the most work, and they are the ones almost nobody uses: “that is what happened when Grandpa’s car broke down, isn’t it.”

PEER: what you do with the answer

A prompt on its own is a quiz. PEER is the loop that makes it a conversation.

• Prompt the child to say something about the book.
• Evaluate the response.
• Expand it, by rephrasing and adding to what they said.
• Repeat the prompt, so they can use the longer version.

Here it is on one page of one book.

You: What is the dog doing? (prompt) Child: Running. You: He is! (evaluate) He’s running away from the bees. (expand) Can you say that? He’s running away from the bees. (repeat)

That is four seconds and the child has gone from one word to seven, including a preposition and a reason. Do it two or three times a book, not on every line.

The Clearinghouse makes one more point about how this changes over time: as the child gets more familiar with a book, “the adult reads less, listens more, and gradually uses more higher level prompts”. You are supposed to be working yourself out of a job.

The part nobody puts on a poster

Two findings from the meta-analysis complicate the cheerful version of this, and they are worth knowing.

The effect was smaller for children aged four to five than for younger ones. The authors concluded that dialogic reading can change home literacy activities for families with two and three year olds. If you have been waiting until your child is old enough to discuss a book properly, the evidence says you have waited too long.

The effect was also smaller for children at risk for language and literacy impairments. The authors’ conclusion was blunt: it did not change home literacy activities for the families of children at greatest risk of school failure. So this is a technique that helps, and it helps the children who need it least by the most. It is not a fix for a gap, and anybody selling it as one is overselling it.

How to start tonight

Four rules, in order of how much they matter.

1. One prompt a page at most. Less, at first. The failure mode is interrogation.
2. Follow their attention, not the text. If they are staring at the cat in the corner, the cat is the prompt.
3. Expand rather than correct. “Running” is not wrong. “He’s running away from the bees” is more.
4. Stop when they want you to. Some nights a story is a story and the right thing to do is read it. We go through why the same book forty times is also doing real work in the same story every night, and how long any of this should take in how long should a bedtime story be.

If reading aloud has become a battle rather than a wind-down, prompts are not the first thing to fix. What to do when a child says reading is boring is the better starting point.

Why we care about this particular research

Taleium is built on one rule: the family decides what happens, and Taleium writes it down. A parent and a child say what the next chapter does, and the words come back in the language for the age they picked.

The reason dialogic reading is the research we keep coming back to is that it describes the same shape from the other direction. Both are the adult asking and the child answering. Both get worse the moment the grown-up takes over the story.

We are not going to claim that writing a chapter together has been through five controlled trials, because it has not. What is established is narrower and still useful: a child who is asked what is happening, and whose answer is expanded rather than corrected, ends up with more words they can use. That is true whether the picture is in a book somebody else wrote or in a story the two of you are making up.

If the number of words is what interests you, the million word gap has the arithmetic on how much a nightly book is worth before school. And if the question is whether any of this works at bedtime specifically, does a bedtime story help a child sleep looks at what the sleep evidence does and does not support.

The Clearinghouse’s review, including the CROWD and PEER descriptions quoted above, is published in full by the Institute of Education Sciences. The meta-analysis is Mol, Bus, de Jong and Smeets, “Added value of dialogic parent-child book readings: A meta-analysis”, Early Education and Development 19(1), pages 7 to 26, indexed on ERIC.