A six-week randomized controlled trial of 150 children in the UK, and what it revealed about the order between technique and frequency

Introduction: The Age of “Read It Like This”

At some point, a parent reading to a child gets anxious. They thought you just read the book, but the methods described in parenting books and videos are far more elaborate.

Ask questions. Expand your child’s answers. Use open questions. Don’t ask yes/no questions. Wait until the child speaks. Point at the picture and ask what it is.

This advice is not groundless. The technique called dialogic reading has been researched for over thirty years, and studies really have accumulated showing benefits to vocabulary, literacy, and school readiness. Public parenting policy in many countries has adopted it as the standard for parent education.

But look closely at the evidence and there is a gap. In most of the studies, the comparison group was “parents who did nothing.” Comparing children of parents who learned the technique against children whose households carried on as usual — of course a difference appears. Whether that difference came from the technique or simply from reading more books is not separated.

A collaboration between the University of Liverpool, the University of Manchester, and the Max Planck Institute for Psycholinguistics aimed straight at that gap.

The Core Question

“If you teach parents interactive reading techniques, does a child’s language develop more than from simply being read to a lot?”

Two further questions were attached:

  1. Does the technique work equally in all families regardless of socioeconomic background?
  2. Does it affect not only vocabulary but grammar and comprehension?

The first mattered especially. Existing meta-analyses reported smaller effects for lower-income children. Mol and colleagues (2008) reported an effect size of 0.13 for children at risk in language and literacy against 0.53 for those not at risk; Manz and colleagues (2010) found a similar gap, 0.14 in low-income families against 0.39 in middle and upper. And the field had attributed the cause to “less-educated parents being unable to execute the technique well.”

How They Studied It: Three Groups, Same Books, Same Number of Sessions

The heart of this study is its design. The team randomly assigned 150 children to three groups.

Group 1. Dialogic Reading Training

The classic technique developed by Whitehurst and colleagues. It follows the PEER sequence.

  • Prompt — get the child to say something about the book
  • Evaluate — evaluate the child’s response
  • Expand — expand on the child’s response
  • Repeat — have the child repeat the expanded phrasing

To this are added five frames for generating prompts (CROWD): completion, recall, open-ended, wh- questions, and distancing prompts that connect to experience outside the book.

Group 2. Pause Reading Training

Another technique designed by the team, weighted toward the adult stopping and waiting first.

  • Pause — leave room for the child to speak first
  • Respond — respond to what the child says or points at
  • Ask open questions — questions that don’t end in yes/no
  • Expand — rephrase the child’s words or add information

Parents were given five sentence frames for generating questions: “Why do you think ~?”, “Can you tell me about this picture?”, “What do you think will happen?”, “How did you know?”, “I wonder what’s happening here.”

Group 3. Active Control — This Is the Study’s Real Design

Control parents learned nothing about reading technique. They received only general advice:

  • Choose a space without TV or toys
  • Make a comfortable, cozy spot
  • Sit close enough that the child can see your face and the book
  • Let the child hold and turn the book if they want
  • If they don’t want to sit still, let them sit beside you
  • Enjoy it

But crucially, the control group received exactly the same 20 books as the other two groups and read exactly the same way — five times a week, two books a day, for six weeks. Each book was scheduled to be reread three times across the six weeks, order and all.

So the control group in this study is not “children who weren’t read to” but “children who were read to diligently, without the technique.” That distinction defines the entire result.

Design and Participants

  • Children: 150, aged 30–36 months (mean 32 months), 45% girls
  • Primary caregivers: 137 mothers, 10 fathers, 2 grandmothers, 1 childminder
  • Site and period: Northwest England, March 2015 – March 2018
  • Assignment: a random sequence generated by an independent researcher, sealed in envelopes. Even the opening of the envelope was recorded on video
  • Blinding: parents did not know that three groups existed; the researchers running post-tests and the video coders did not know the assignments
  • Preregistration: clinical trial registration complete (NCT02625584)
  • Actual reading: mean 50.6 sessions out of a maximum 60 (SD 14.5). Most families followed through conscientiously
  • Socioeconomic status: by caregiver education. Degree or above coded high SES; two or fewer A-levels coded low SES (63% of the population by the 2011 UK census falls in the latter)

What Was Measured

  • Expressive language: PLS-5 UK Expressive Communication — what the child can say
  • Receptive language: PLS-5 UK Auditory Comprehension — what the child understands
  • Sentence structure comprehension: CELF Preschool-2 UK Sentence Structure subtest — hear a sentence, choose the matching picture
  • Mean length of utterance (MLU): real conversation recorded and transcribed to measure how long the child’s utterances were (47-child subsample)

Parent reading behavior was measured too. Pre- and post-intervention reading videos were cut into 30-second intervals and each interval coded for whether the target behavior appeared. Inter-coder agreement was very high: 0.85 for dialogic behaviors, 0.91 for pause-reading behaviors.

Result 1: The Parents Definitely Changed

The training worked. Strikingly so.

In the dialogic reading group, the proportion of intervals containing target behaviors rose from 0.41 to 0.71. Over the same period the control group went from 0.42 to 0.39 — slightly down.

The pause reading group was more dramatic. From 0.17 to 0.59. The control group sat still at 0.25 → 0.24.

Both were highly significant (p < .001), and the models explained about 59% of the change in parent behavior. Six weeks of video training and weekly contact really did change how parents read.

An Interesting Side Finding

Parents who had been interacting less changed more (β = −0.73, p < .001). Parents already doing it well had less room to rise, and the training left more with those who started lower.

Result 2: But the Children’s Language Did Not Change

Here the study diverges from expectation.

On all four language measures, there was no statistically significant difference between the trained groups’ children and the control children.

  • Expressive language: the model did not even fit the data (R² = .04, p = .898)
  • Receptive language: the model fit (R² = .14) but the only significant predictor was the pretest score
  • Sentence structure comprehension: likewise, only the pretest score was significant (R² = .16)
  • Mean length of utterance: the model did not fit (p = .388)

Children in all three groups scored higher after six weeks. But that is because children this age grow over six weeks, not because of group assignment. Neither dialogic reading nor pause reading produced a better outcome than just reading.

The only meaningful predictor for receptive language and sentence structure was the starting point. Children who scored lower at the start rose more (β = −0.20, p = .001 / β = −0.35, p < .001). There was a catching-up current, but it had nothing to do with which group a child was in.

Result 3: This Is Not “The Sample Was Too Small”

The first thing to suspect in a null study is sample size. One hundred fifty is not small for language development research, but it may be too small to detect a small effect.

The team tested that suspicion themselves, because of the statistical principle that absence of evidence is not evidence of absence. To say there is no effect, you have to confirm the absence separately.

Two additional analyses were run.

Post-hoc power simulation. They calculated how many participants would be needed to obtain a significant result at the observed effect sizes. For the three standardized language measures, the required numbers ranged from 942 to over 10,000. Needing a sample that size means that even if an effect exists, it is not of a practically meaningful magnitude.

Equivalence testing (TOST). A method for directly testing the claim “no effect of 0.5 or larger exists.” On all three measures — expressive (p = .009), receptive (p = .008), sentence structure (p = .003) — effects of medium size or larger were statistically rejected. Only MLU was left undecided, because that measure’s sample was 47.

So this is not “we failed to find an effect” so much as “there was no effect of meaningful size.”

Result 4: The Assumption About Low-Income Parents Was Also Wrong

Differences by socioeconomic status appeared in neither direction.

Low-SES parents learned and executed the techniques just as well as high-SES parents. The children’s language gains were equal across the two groups.

This directly contradicts the standard explanation for the earlier meta-analytic pattern — the assumption that less-educated parents struggle to implement interactive reading. It was not an execution problem. Parents of every background did what they were taught, and children of every background showed equally no added effect.

Why Was There No Effect: The Identity of the Control Group

The team offers five possibilities. The two most important interlock.

The Control Group Was Already Reading Hard

All three groups read 20 books repeatedly for six weeks. Over 50 sessions on average. For control parents too, this was not ordinary life. The control children were also receiving an intervention — minus the technique.

How large that design difference is shows in another meta-analysis by the same team (Noble et al., 2019). Splitting book-reading intervention studies by control type:

Studies comparing against a passive control that did nothing: effect size 0.231 (p < .001) Studies comparing against an active control that did something: effect size 0.038 (p = .584)

The effect nearly vanishes. Meaning a considerable part of the “dialogic reading effect” prior studies had been finding may have been not the effect of the technique but the effect of more book reading itself.

The Technique May Add Nothing on Top of Frequency

The team’s fifth explanation summarizes as: teaching parents to read interactively may be no better than simply asking them to read more.

The team immediately applies a brake here, though. Interactive reading does contain elements known to help language development — expansion, recasting, open questions, joint attention — and research finds that the talk exchanged over a book is more varied in syntax and vocabulary than the talk during play. So the authors state plainly:

It would be premature to discard interactive shared book reading techniques wholesale on the basis of a single study.

Practical Takeaways

How Often You Read Comes Before How You Read

At least in the 2–3 year range, over six weeks, on the measures this study used. If worrying about executing the technique correctly reduces reading time itself, the order has been inverted.

One honest caveat has to be added. This study did not prove that reading a lot grows language, because there was no comparison group that read nothing. What this study can say is exactly one thing: when reading volume is equal, no additional contribution from technique was confirmed.

The Anxiety That You’re “Not Reading Properly” Rests on Weaker Ground Than You’d Think

Many parents check themselves against a list — how many questions asked, whether the child’s words were expanded well. This study shows the ground for clinging to that checklist is not strong. The advice given to control parents was far humbler: sit comfortably, let the child turn the pages, enjoy it.

Training Leaves the Most With Those Who Weren’t Doing It

Looking only at parent behavior change, parents who had barely been interacting gained most from the training. It did not move the children’s language scores, but it is a clue about whom parent education programs should target.

Limitations

The team documented these in detail.

  • Only two techniques were tested. Other forms of interactive reading may work.
  • Six weeks may be short. The duration followed convention in the existing literature, but moving standardized test scores may require a longer, stronger intervention.
  • Actual home implementation was not directly verified. Post-intervention videos confirmed the techniques were used, but whether reading looked that way for all six weeks rests on indirect evidence.
  • The instruments may have been insensitive. Instead of standardized global language tests, a sensitive measure like asking directly about words that appeared in the books might have produced different results.
  • MLU was analyzed on only 47 children, because of transcription costs; on that measure no conclusion about effect or no effect can be drawn.
  • The active control may itself be an intervention. The possibility of a Hawthorne effect — behavior shifting merely from awareness of being observed — cannot be excluded.

Even with these limits, this study followed the most rigorous design available in the field: a preregistered, double-blind randomized controlled trial. And when the desired result did not appear, it reported that.

Closing Thoughts

The message this study leaves is neither negative nor positive.

In reading to a child, technique may not be as decisive as we have believed.

Even the team kept its conclusion carefully open. They did not declare technique useless; they wrote that longer interventions and finer measurement are needed. But at least within what has now been confirmed, the story that a child falls behind in language because a parent failed to master “the correct method” rests on weak ground.

Under the conditions this study created — 20 books, five times a week, six weeks — the children in every group grew. Within that, no additional contribution from technique was found.

Sitting them on your lap, letting them turn the page when they want to, reading with a smile. That was the entirety of the advice the control group received, and their children were no different.

Source: Noble, C., Cameron-Faulkner, T., Jessop, A., Coates, A., Sawyer, H., Taylor-Ims, R., & Rowland, C. F. (2020). The impact of interactive shared book reading on children’s language skills: A randomized controlled trial. Journal of Speech, Language, and Hearing Research, 63(6), 1878–1897. https://doi.org/10.1044/2020_JSLHR-19-00288

This paper is open access under a Creative Commons Attribution 4.0 International license. Trial preregistration: NCT02625584