pictures of the eye area of human faces to emotion words. The “strange stories” task
(Наррё, 1994) was developed to be used with children with autism, and adapted for adults
(Jolliffe & Baron-Cohen, 1999). Participants are read a series of scenarios and asked a
single question at the end of each one to assess their comprehension. In the theory of mind
stories, success relies on understanding the intention behind a non-literal utterance (such
as a lie or a double-bluff). For this study, five theory of mind stories were selected, based
on which stories had most effectively discriminated people with autism from controls in
an earlier study (Brent, Rios, Шррё & Charman, 2004).
• Parallel interview
The parallel interview is included in the Appendix, and the rationale for creating it is
described in the Introduction. It was designed to follow the structure and demands of the
AAI as closely as possible, without activating attachment representations. It was
administered at the very end of the interview session, which meant it was at least two
hours after the AAI, to reduce the likelihood of participants noticing the similarities in the
interviews or referring back to their AAI responses. The parallel interview was scored on
the AAI state of mind scales; transcripts were not assigned to categories, because of the
lack of “experience” scales and the difficulty of scoring some state of mind scales based
on anything other than interpersonal experiences. Previous research in a non-clinical
sample (Crowell et al., 1996) demonstrated that overall classifications on a different
parallel interview did not overlap significantly with those on the AAI, and that coherence
scores from each interview were not correlated. Although classifications on Crowell et
al.’s parallel interview were based on state of mind scales, these scores were not reported
individually. However, for the parallel interview to give different classifications from the
AAI, it can be assumed that the state of mind scores were not substantially similar, and
this is what was tested in the present study. The parallel interview was coded by an AAI
10