A study tracking 296 Australian families from 6 months to age 4 reveals the causal relationship between parent-child conversation and language development.
Introduction: It Feels Like Common Sense, but It Was Never Proven
“Talk to your child as much as you can.” Every parenting book gives this advice. Intuitively, it makes perfect sense. But in science, “obvious” and “proven” are two very different things.
There have been plenty of studies showing that children whose parents talk to them more tend to develop language faster. The problem is that these findings are correlational. Parents who talk more also tend to be more educated, have more books at home, and be more financially stable. So it’s impossible to tell whether the child’s faster language development is truly because of conversation — or simply a byproduct of the broader environment.
Even prior intervention studies had gaps. Parent coaching programs successfully increased how much parents talked, but many never measured the child’s actual language outcomes. Getting parents to talk more and actually improving a child’s language are two separate claims.
To bridge this gap, an Australian research team conducted a large-scale, 4-year longitudinal study.
The Core Research Question
A team from the Telethon Kids Institute and several Australian universities asked:
“If parent-child conversation is consistently increased over time, does it actually improve the child’s language development?”
This is a question about causation, not correlation. And to answer it, the team deployed two key tools: a device that objectively measures conversation, and a statistical technique that estimates causation from observational data.
How They Did It
LENA: A Small Device That Records an Entire Day of Conversation
The study’s core tool was LENA (Language ENvironment Analysis), a small recording device placed in the child’s clothing pocket. It records up to 16 hours of audio, and proprietary software analyzes the recordings to calculate the number of conversational turns.
A conversational turn is a back-and-forth exchange — an adult speaks and the child responds, or the child vocalizes and the adult responds. If both parties produce sounds within 5 seconds of each other, it counts as one turn. The key here is that it measures interactive conversation, not one-sided speech directed at the child.
8 Measurements Over 4 Years, Every 6 Months
The team recruited 296 families from the LiLO (Language in Little Ones) birth cohort — full-term infants born in 2017 across South Australia, Western Australia, and Queensland.
- Home visits at 6, 12, 18, 24, 30, 36, 42, and 48 months — 8 time points in total
- At each visit: LENA recording + parent questionnaire (home activities, number of books, screen time, mental health, financial situation, etc.)
- At 36 and 48 months: children’s language was directly assessed using the CELF-P2 (Clinical Evaluation of Language Fundamentals Preschool-2), a standardized language evaluation tool
The CELF-P2 comprises three subtests — sentence structure comprehension, word structure, and expressive vocabulary — and provides scores standardized for the Australian and New Zealand population (mean = 100, SD = 15).
Estimating Causation from Observational Data
The team used a Marginal Structural Model, a statistical technique whose core idea works like this:
In standard regression, you “control for” confounders. But when time-varying confounders — say, home activity levels at 12 months — may themselves have been influenced by conversation levels at 6 months, simply controlling for them introduces bias.
Marginal Structural Models use Inverse Probability Weighting to remove this time-dependent confounding, creating a “pseudo-population” where — statistically speaking — everything except conversation levels is equalized. In essence, it constructs a hypothetical world in which the pure effect of conversation can be estimated.
Within this pseudo-population, five scenarios were simulated:
- What if all families conversed at the 5th percentile level?
- What if all families conversed at the 25th percentile level?
- What if all families conversed at the median (50th percentile) level?
- What if all families conversed at the 75th percentile level?
- What if all families conversed at the 95th percentile level?
Who Were the Participating Families?
- 53% of children were female; 50% were firstborns
- 56.1% of mothers had a university degree or higher
- Mean maternal age at birth: 31.1 years
- Mean language score at 36 months: 103.8 (SD = 12.5)
- Mean language score at 48 months: 103.5 (SD = 12.1)
- Only English-speaking households were included
Average daily conversational turns rose steadily from 331 at 6 months to 741 at 48 months.
Results: More Conversation, Higher Language Scores
Simulated Language Scores at 48 Months by Conversation Level
| Conversation Level | Mean Language Score | 95% CI |
|---|---|---|
| 5th percentile | 97.21 | 96.86 – 97.56 |
| 25th percentile | 99.97 | 99.62 – 100.32 |
| Median (50th) | 102.15 | 101.80 – 102.50 |
| 75th percentile | 105.52 | 105.17 – 105.87 |
| 95th percentile | 111.41 | 111.06 – 111.76 |
The gap between the 5th and 95th percentiles was 14.2 points — nearly a full standard deviation. The same pattern appeared at 36 months, but the gap was wider at 48 months. This means the cumulative effect of conversation grows over time.
Crucially, these figures were obtained after statistically removing the effects of parental education, financial situation, home activities, number of books, screen time, and parental mental health. They represent the pure effect of conversation itself.
The Same Pattern at 36 Months
| Conversation Level | Mean Language Score |
|---|---|
| 5th percentile | 100.42 |
| Median (50th) | 102.80 |
| 95th percentile | 108.51 |
The gap is smaller than at 48 months. The effect of conversation grows stronger the longer it accumulates.
Would 150 Extra Turns a Day Be Enough?
The team connected their findings to real-world programs. A library-based program in the United States (Beecher & Van Pay, 2020) used LENA feedback to successfully increase parents’ conversation, boosting daily turns by roughly 150 in just 9 weeks.
How much impact would that have? Mapped onto this study’s scenarios, it corresponds to moving from the 5th percentile level (about 138 turns/day at 6 months) to roughly the median (325 turns/day). If sustained long-term, this shift could yield a 0.32 SD improvement in language scores — approximately a 4.8-point increase.
In other words, a few weeks in a library program could produce a measurable difference. Not an expensive therapy program, but a small shift in everyday conversation habits.
Why “Conversational Turns” — Not a Barrage of Words
What this study measured was not the total volume of words a parent directed at the child. It was the number of back-and-forth conversational turns. This distinction matters.
LENA counts a conversational turn when an adult speaks and the child responds (including babbling) within 5 seconds, or the child vocalizes and the adult responds. Having the TV on in the background or talking on the phone next to the child doesn’t count.
The famous “30 Million Word Gap” study — claiming that children from low-income families hear 30 million fewer words by age 3 — received enormous attention, but recent research has increasingly pointed out that the quality of interactive conversation matters more than sheer word exposure. This study adds another piece of evidence in that direction.
Practical Implications
Start Conversing Before the Baby Can Talk
The study began measuring conversational turns at 6 months — an age when babies can’t speak yet. But parents responding to babbling, laughter, and sounds counted as conversational turns, and these predicted later language development. The research team emphasized that “conversation and interaction should be promoted before the child begins to talk.”
Everyday Conversation Habits Over Elaborate Programs
The effects shown in this study don’t come from “special education” — they come from everyday conversational back-and-forth. Talking to your child during meals, walks, or diaper changes, and responding when they react. Even adding just 150 turns a day — as in the library program — produced meaningful differences.
Universal, for All Families
The research team recommends universal interventions targeting all families, not just at-risk groups. The effect of conversation showed a dose-response relationship: no matter where you start, increasing conversation yields benefits.
Limitations
The research team openly acknowledged several constraints:
- Only quantity was measured, not quality. They counted conversational turns but did not assess how rich or complex the conversations were. The team acknowledged that “the quality of interaction — vocabulary complexity, grammatical diversity — may matter more than quantity.”
- Only English-speaking families were included. The results may not directly apply to bilingual households or other language environments.
- Bidirectional effects were not fully separated. As a child’s language develops, parents may naturally converse more — a reverse effect that wasn’t completely isolated.
- Effect sizes are small. While statistically significant, whether the differences are perceptible at the individual family level is debatable.
- The language assessment tool (CELF-P2) may lack sensitivity for subtle differences in a general population. It was originally designed for diagnosing language disorders.
Closing Thoughts
The message of this study is this:
Everyday conversation — talking to your child and responding to their reactions — drives language development. This is no longer “common sense.” It is now causal evidence.
The effect isn’t dramatic. But it is consistent and cumulative. And the most encouraging part is that producing this effect requires no special tools or expense. Talk to your child. When they respond — with anything — respond back. From 6 months on, every day, little by little. That’s what makes a measurable difference four years later.
Source: Brushe, M. E., Mittinty, M. N., Gregory, T., Haag, D., Lynch, J. W., Reilly, S., Melhuish, E., & Brinkman, S. A. (2025). The causal effect of parent–child interactions on child language development at 3 and 4 years. International Journal of Language & Communication Disorders, 60(3), e70045. https://doi.org/10.1111/1460-6984.70045