How we grade evidence
The short version
Every claim on this site carries one of four grades. The grade tells you how much weight the claim can take.
| Grade | What it means | Example from this site |
|---|---|---|
| Strong | Several randomised trials or meta-analyses pointing the same way, or official guidance resting on them | Reading with your child improves language (a meta-analysis of 19 randomised trials) |
| Moderate | Consistent findings from studies that follow or observe families, or a small number of trials. Cause and effect not fully settled | Back-and-forth conversation predicts later language |
| Weak | Few studies, small samples, a single research group, or mostly expert opinion | Chores build self-control |
| Contested | Credible researchers disagree about the finding or what it means | Whether pretend play itself causes any gains |
The word carries the grade. The pencil ticks repeat it for print and greyscale: three for strong, two for moderate, one for weak, and a question mark for contested. Colour is never the only cue. Contested is not a rung below weak. It means good researchers disagree.
Three more labels you will see
- Reasoned, not tested. Something that follows sensibly from the evidence but that nobody has tested directly. Most of what I suggest for two brothers of different ages is in this group, because no research was found on it.
- Secondary. I confirmed the figure from an abstract, a press release or a summary, not from the paper itself.
- Not verified. I could not confirm it from the original source. It is either left out or carries this label.
In each article's source list, every source also shows what it was confirmed from: "read" (the source itself), "abstract" or "secondary".
How to read an effect size
Most findings in child development are small. A correlation of 0.1 to 0.2 is typical. That means a real pattern across thousands of children and very little power to predict what happens with any one child. Nothing on this site should be read as "do this and your child will do that".
The order I trust sources in
- Australian guidance: Raising Children Network, the Queensland child health record (the Red Book), the Australian 24-Hour Movement Guidelines, the eSafety Commissioner.
- The World Health Organization, the US Centers for Disease Control and Prevention, the American Academy of Pediatrics.
- Peer-reviewed reviews and meta-analyses.
- Single studies, with their size and limits stated.
Things that lower a grade
- The study only shows that two things go together. Families who read a lot differ from families who don't in many other ways.
- The people who tested a programme are the people who built it.
- The result was reported by the parents who took the course.
- The children studied were older or younger than the ones I'm writing about.
- The study is small, or has not been repeated.
What a grade does not tell you
It doesn't tell you whether something is worth doing. Wrestling with your kids has only moderate evidence behind it and is still a good idea. Pretend play has not been shown to build creativity and kids should still do it, because they like it. The grade is about the claim, not the activity.
When I change my mind
Each article says what would change the advice. When a better study turns up, the article is updated and the change is noted with a date.