Absence of Construct Validity in Standard False Belief Tasks



False belief tasks are used to assess theory of mind in infants, young children, individuals with autism, and chimpanzees. However, the construct validity of the tasks has not been established because it is possible to pass the tasks without attributing a false belief to the agent. The standard tasks have two response options, one corresponding to the agent's false belief and the other corresponding to the current real state of affairs. Children could simply reason that the agent cannot see the current real state of affairs, thus the agent does not know, and thus the agent will "get it wrong." Because the only "wrong" option in the standard tasks is also the false belief option, children using this reasoning would choose the false belief option by default. Fabricius and Khalil (2003) labeled this "perceptual access reasoning" (PAR) and tested it using 3-option versions of the false belief tasks where there were two "wrong" options: the false belief option and an irrelevant option. PAR predicts that young children who pass the standard, 2-option false belief tasks would choose randomly between the false belief and irrelevant options in the 3-option tasks. Fabricius and Khalil (2003) found evidence in favor of PAR; however, Perner and Horn (2003), found evidence against it and in favor of the validity of the false belief tasks. We conducted seven replication attempts (N = 165) using Perner and Horn’s methods, and we employed the "continuously cumulating meta-analytic” (CCMA) approach to supply a test of replication with high power. We found strong evidence that Perner and Horn's original findings failed to replicate, but Fabricius and Khalil's original findings did replicate, casting serious doubt on the validity of the standard false belief tasks. The videos and the data from the 7th study are presented here. The procedures from the other six studies are presented in the materials section. The video below shows a 4-year-old passing the standard, 2-option Typical box ("Smarties") false belief task but, in accord with PAR, passing one of the three 3-option tasks (Neutral box) and choosing the irrelevant alternatives in the other two 3-option tasks (Location and Typical box).
Date made available2014
PublisherNew York University Library

Cite this