A training procedure was developed for teaching three year-old childre
n to pass a standard false belief task. A series of 4 video sequences
was developed, in each of which a child showed surprise following the
unexpected transfer of an abject. A group of 23 three year-olds who fa
iled a standard false belief task took part in 8 discussions of the 4
video clips, over a period of two weeks, and were compared with a cont
rol group of 23 task-failers who had 8 story reading sessions over the
same time period The training involved (a) interactive recall of the
initial events in each sequence, (b) factual questioning about the eve
nts and (c) an explanation of the thoughts and actions of the video pr
otagonists. The emphasis was on positive elaboration of children's ans
wers rather than on negative feedback as counter evidence. The trainin
g group performed significantly better at an immediate post-test, and
at a follow-up test two weeks later, on a similar as well as a general
isation task. Task success was unrelated to vocabulary (BPVS) score, b
ut was significantly related to success of responses during the traini
ng. The results indicate learning in the Training group of a discursiv
e format for talking about events useful for describing false beliefs.