that the AAI still manages to tap into a specific state of mind with respect to attachment.
This is an indicator of the discriminant validity of the AAI.
Importantly, there was no overlap between the scales that correlated across the
different interviews and those that differentiated autistic participants from non-autistic
controls. As predicted, there were clear differences between participants and controls in
the two coherence scores and reflective function, while most other scales did not even
approach significance, particularly when secure participants were excluded. Reflective
function is a measure of mentalising, a central difficulty for people with autism, so it was
expected that the participants would score lower than the controls on this measure
(although, surprisingly, reflective function scores did not correlate significantly with the
theory of mind measures). The coherence scores are based on Grice’s (1975)
conversational principles of quality, quantity, relation and manner; many participants
appeared to struggle particularly with quantity (responding either at great length or
extremely tersely), and relation (moving repeatedly onto irrelevant topics).
Coherence and reflective function scores were used as an index of attachment
security. Coherence of transcript was correlated with ADOS communication score, which
is understandable as both are measures of appropriate reciprocal conversation, although
they focus on slightly different aspects (the ADOS score includes nonverbal
communication). It is interesting that coherence of transcript also shows some suggestion
of a correlation with full-scale IQ, rather than verbal IQ; the interpretation of this is not
clear, and as it did not reach significance it needs to be replicated. Both of these findings
are consistent with the hypothesis that AAI security would be related to autistic
symptomatology and IQ, as is the case with autistic children’s security in the Strange
Situation. There is little evidence here, however, that theory of mind measures are related
to attachment measures, even to reflective functioning which is explicitly a measure of
21