A Puppet Show Designed to Find What an Autistic Child Sees Differently

A child sits at a small table while two researchers work a pair of dolls. Sally has a basket. Anne has a box. Sally puts a marble in her basket and leaves the room. Anne takes the marble out, hides it in her own box, and tucks the basket lid back down. Sally returns. "Where will Sally look for her marble?" the experimenter asks.

The correct answer is the basket. Sally believes — wrongly — that her marble is still where she left it. To answer correctly, the child has to hold two models of the world in mind at once: the actual location of the marble, and Sally's outdated belief about it. Choosing the basket means recognizing that other minds carry their own information, separate from yours. Choosing the box means treating Sally as if she had access to what you have access to.

The task descends from a 1983 paper by Heinz Wimmer and Josef Perner, who in turn drew on a thought experiment by Daniel Dennett. In 1985 Simon Baron-Cohen, Alan Leslie, and Uta Frith at University College London adapted it into the Sally-Anne version and ran it on 61 children: 20 with autism, 27 typically developing, and 14 with Down syndrome. Eighty-five percent of the typical children answered correctly. So did 86 percent of those with Down syndrome. Only four of the 20 autistic children — about 20 percent — passed.

The paper, titled "Does the autistic child have a 'theory of mind'?", reframed autism in cognitive terms: not a deficit of warmth, but a specific difficulty in modeling what is going on inside other people. Forty years on, more sophisticated tasks have qualified the picture, but the Sally-Anne remains the single best-known one-minute window into a piece of social cognition most people use without noticing.