The false belief task and its replication

Heinz Wimmer and Josef Perner proposed in 1983 a task that became standard: a character puts an object away, leaves, the object is moved in their absence, and the child is asked where the character will look for it.

Answering correctly requires distinguishing what one knows oneself from what the character believes. Three-year-olds fail overwhelmingly, five-year-olds succeed, and the switch falls around four in verbal versions.

A variant, known as the unexpected contents task, involves showing a box of sweets that actually contains pencils, then asking what another person will think is inside. It yields a comparable timetable of success, which strengthens the robustness of the result.

From 2005, several teams proposed versions based on gaze rather than verbal answer, measuring fixation time in front of a scene inconsistent with the character's belief.

The most cited study, published by Kristine Onishi and Renée Baillargeon in 2005, measured how long infants looked at a scene in which a character searched for an object in a place where they could not believe it to be.

Those protocols reported signs of attribution in infants as young as fifteen months, which gave rise to the hypothesis of two systems, one early and automatic, the other late and explicit.

Those early results are among the most disputed in developmental psychology. Multi-laboratory replication attempts published from 2018 did not find the effect, or found it far weaker than in the initial publications.

The classic verbal test, by contrast, replicates. That is a frequent configuration: the robust effect is the oldest and least spectacular, and the contested one is the finding that moved the timetable.

Succeeding requires understanding the question, holding the sequence in mind and inhibiting the obvious answer, which is to point to the object's actual location. Part of younger children's failure may stem from those demands rather than from the absence of the tested ability.

Simplified versions, reducing the language and memory load, obtain slightly earlier success, which supports that reading without establishing it. A meta-analysis by Henry Wellman, David Cross and Julanne Watson published in 2001 had already concluded that procedural variations shifted the age of success slightly without altering the general shape of development.

Cross-cultural comparisons find success on the task everywhere it has been administered, but the switching age varies by several months between populations.

Those gaps are associated, in several studies, with how often mental states are talked about in the child's surroundings. The association is measured; its direction is not established.