Every topic on this site leans on studies. This page explains what kinds of studies exist, how much weight each kind deserves, and the rules we followed when turning research into advice. If you read one background page on this site, make it this one.
Each topic carries one overall rating. It reflects the weakest link behind the short answer — not the single best study we found.
| Rating | Meaning |
|---|---|
| A — Strong | Consistent findings across multiple high-quality randomized trials or reviews of trials |
| B — Moderate | Some randomized trials, or strong observational evidence with only minor inconsistencies |
| C — Weak | Observational evidence only, small samples, or studies that meaningfully disagree |
| D — Very weak | Expert opinion, tradition, or essentially no data — we say so plainly |
The words in each topic are matched to its rating: strong evidence gets "the evidence shows…", moderate gets "the evidence suggests…", weak gets "it's unclear…", and very weak gets "we simply don't know." We never hedge strong evidence and never oversell weak evidence.
Systematic reviews and meta-analyses. Researchers gather every study on a question, grade their quality, and combine the results. One good review is worth more than ten single studies, because single studies are often flukes.
Randomized controlled trials (RCTs). Participants are randomly assigned to a treatment or a control group. Randomization is the gold standard because it balances out everything we can't see — motivation, genetics, income — so any difference in outcome can plausibly be blamed on the treatment itself. But RCTs are expensive, often small, and can't run long enough to catch effects that take years to appear.
Cohort studies. Researchers follow a large group of people over time and compare those who happened to do X with those who didn't. Cohorts can be huge and run for years, which RCTs rarely manage. The catch: the groups weren't randomized, so differences in outcome might come from who chose X, not from X itself. Parents who breastfeed, for example, differ from parents who don't in education, income, and health habits — all of which affect child outcomes independently. This is called confounding, and it is the single biggest trap in parenting research.
Case–control studies. Researchers start with the outcome (say, a rare condition) and look backward at what the affected children were exposed to, compared with similar children who weren't. Useful for rare outcomes, but looking backward makes them vulnerable to faulty memory and biased records.
Cross-sectional studies. A snapshot at one moment in time: researchers measure X and Y simultaneously and report the correlation. They can't tell you which came first, so they can't tell you about cause at all — only that two things tend to travel together.
Case reports and case series. Detailed descriptions of one patient or a handful of patients. They can flag something worth studying, but they prove nothing: there is no comparison group, and unusual cases get written up precisely because they're unusual.
Expert opinion and tradition. What experienced clinicians believe, or what parents have "always done." Sometimes right, sometimes fossilized habit. We treat it as a starting suspicion, never as evidence.
Most of what parents want to know can't be tested in an RCT — you can't randomize babies to breastfeeding or formula, to daycare or home care, and randomize them ethically for long. So the evidence base for parenting is dominated by observational studies, with all their confounding. Add small samples, short follow-ups, publication bias (boring "no effect" results don't get published), and study populations that skew Western, educated, and affluent — and you get a literature full of contradictions. When studies on a topic disagree, we say so and explain why (different designs, populations, or eras) instead of quietly picking a winner.
Let's be explicit: AI was used extensively to build this site. That was a deliberate design choice, structured so that no single model's output goes unchallenged:
The aim is the best and latest evidence, always updated — with the working shown so anyone can check it.
No anecdote as evidence ("my friend's baby…"). No relative risk without a base rate. No "natural = safe" or "medical = risky" framing. No diagnosis or prescription — we inform; your clinician decides with you. And no moralizing about choices where the evidence is equivalent: feeding, delivery mode, childcare arrangements.
Every topic was researched with extensive AI assistance (see above) and reviewed and approved by the author — a parent who checked the papers, not a clinician. Nothing here is medical advice. The site is updated as new landmark evidence appears, and every substantive correction is logged in the topic itself, not silently edited away.