A local woman with three days of training called once a month for nine months. Toddlers in rural Ethiopia ate more varied meals, but eggs and meat rose just as much without her

At around six months, breast milk stops being enough on its own, and stunting that sets in over the next year and a half is hard to undo. In rural West Gojjam, in Ethiopia's Amhara region, the paper cites stunting of 35.5% to 39.2% among young children. Advice already reaches mothers there at clinic visits. So a team working with Bahir Dar University tried something closer to home, and reported the results in PLOS One on 29 October: a monthly visit from a local woman, for nine months, on a weekend or a religious holiday, lasting 40 to 60 minutes.

The visitors were 30 team leaders from the Women's Development Army, the network of women who link households with Ethiopia's Health Extension Program and who normally work unpaid. Each had a three-day course and a one-day refresher four months later, and looked after 13 or 14 mothers. At each visit she asked what the child had been eating, picked two or three things to work on, and went through an illustrated guide the mother kept. The talk was built on the Health Belief Model, which holds that people change when they think a problem could reach them, think it matters, and believe they can do something about it. Thirty rural communities were randomised, 15 each way, covering 802 mothers of babies aged 6 to 8 months. The other 15 got routine advice during antenatal, postnatal and immunisation visits, with no guide and no follow-up.

The main measure was strict: solids started on time, five of eight food groups the day before, enough meals, and some egg or flesh food, all four together. It rose from 10.7% to 24.0% of mothers with the visits (94 of 391) and sat still in the comparison group, 8.5% to 8.6% (33 of 385). That 13.2-point gap is a shade above the 12 points assumed in the trial's sample-size sums. The adjusted odds ratio was 2.49, which the paper words as nearly 2.5 times more likely, though it is a comparison of changes in odds and not a simple ratio of chances. Dietary variety went from 14.0% to 41.7% against 13.5% to 26.2%, and a minimum acceptable diet from 13.5% to 37.9% against 13.2% to 20.3%.

The eggs and meat are where the story turns. The share of children eating them climbed from 27.7% to 49.9% with the visits and from 19.7% to 38.7% without, and the difference in those rises was nothing to rely on (odds ratio 1.10, 95% CI 0.72 to 1.70). Some of the climb on both sides may just be babies becoming toddlers who eat more, though that is our reading. The authors suggest counselling alone cannot fix what a household can afford or get hold of, and point to norms and household decisions as well. A similar split shows up in the mothers' answers: knowledge jumped (odds ratio 6.61), while the shift in attitudes fell just short of significance (1.45, 95% CI 0.99 to 2.12).

Hold the good news loosely on a few counts. Everything was recalled by mothers, over one day of eating, and those who had spent nine months being told what to feed may have reported accordingly; the authors raise this too. With only 30 clusters, the authors say they could not apply the usual small-sample correction, and the variety and meal results, whose intervals start at 1.04 and 1.01, were not adjusted for multiple comparisons, so treat them as supporting evidence. The groups also differed in ways the model did not adjust for: 55.4% of babies with the visits were boys against 43.9%, and 17.0% of those mothers were first-timers against 24.2%. And the question that matters most stays open. The trial registry lists children's length and weight as a secondary outcome, and this paper, which covers one objective of the trial, reports none. Even taken at face value, 76% of counselled children still missed the mark. What the trial fairly shows is that a modest, local, repeated conversation seems to shift what mothers say they feed, as far as a household's food supply allows.