A 2007 experiment became one of the most cited results in infant research. Babies watched a character struggle up a hill, then reached for the puppet that had helped rather than the one that had hindered. Replications since have come out mixed. So 37 labs on five continents ran one , standardized version, testing 1018 infants between five and ten months old. What did the big version show?
The preference was not there. In the , 49.34 percent of infants chose the Helper over the Hinderer, which is indistinguishable from chance. In the , with an inanimate object pushed up or down, 55.85 percent chose the pusher, and that differed neither from chance nor from the social figure. The did not produce the effect it is famous for.
The 2007 experiment is a landmark. Infants watched one character try and fail to climb a hill while a second helped and a third hindered, and the babies then reached for the Helper. It was taken as evidence that social evaluation starts inside the first year of life.
The trouble is what came next. Replication attempts gave mixed results, which is the situation a big coordinated study exists to resolve. Using the ManyBabies framework for team-based science, 37 labs across five continents ran one preregistered, standardized protocol. They tested 1018 infants aged five and a half to ten and a half months, of whom 567 met inclusion.
In the social condition, 49.34 percent of infants preferred the Helper. In the nonsocial condition, where an inanimate object got pushed up or down rather than a climber helped, 55.85 percent preferred the pusher. Neither proportion differed from chance, and neither differed from the other, so the social effect was absent and so was the nonsocial comparison it would have been judged against.
Read the conclusion precisely. This is evidence against prosocial preferences in the hill paradigm, and the authors put it as the effect being weaker, absent, or developing later than previously estimated. That is narrower than saying babies have no moral sense. The study is also the first of its kind to use an active behavioral measure, manual choice, at this scale. That makes it a proof of concept for how the field might retest its other landmarks.
wat. Across 37 labs and more than a thousand babies, the helpful puppet got chosen 49 percent of the time, which is a coin.
wat wat
A finding this widely taught deserves a test this size. The conclusion is not that infants lack social evaluation, but that this paradigm does not reveal it, and that the effect is weaker, absent, or arriving later than anyone estimated.
How do we know?
The ManyBabies consortium ran a preregistered, standardized replication across 37 labs on five continents, testing 1018 infants aged five and a half to ten and a half months and including 567 in the analysis.
What is still unknown
This retires one paradigm rather than the question. A manual-choice measure may also miss a preference an infant holds without reaching for anything.
Learn something?
Where to next
What happened when the Dunning-Kruger effect was run on random data?replication
What was left of the marshmallow test when they ran it on ten times as many children?replication
What does the familiar version of the Milgram experiment leave out?replication