those of adults without autism?
Participants’ and controls’ scores on the continuous AAI scales were compared using
Mann-Whitney U tests, summarised in Table 2. Participants’ and controls’ scores differ
significantly (p < .05) on reflective function, coherence of transcript, coherence of mind
and unresolved loss.
Although the two-category chi-square test had shown no significant difference in
attachment security between participants and controls, the fact that some disorganised
controls had a secondary classification as secure had led to concern that this might falsely
inflate any difference between the groups. That is, those classified as secure would be
expected to have higher scores for coherence and reflective function than those classified
as insecure. Because of this, the analyses were repeated with only the 17 participants and
11 controls whose primary AAI classification was insecure or disorganised. In this
subsample, participants and controls still differ significantly on reflective function (p < .
05, effect size 0.75), coherence of transcript (p < .01, effect size 0.95) and coherence of
mind (p < .01, effect size 0.88). However, the difference on unresolved loss is no longer
significant. The larger effect sizes seem to be because the three secure autistic participants
scored high on these three scales relative to the rest of the group, so reduced the difference
from controls.
Although there is no space in the present paper for a discussion of the quality of
the participants’ responses to the AAI, brief excerpts are provided in Table 4 to give a
flavour of the transcripts.
-------------
Table 4 about here
18