Modified COSMIC 22
response, showed no such association with ADOS-G overall scores. Furthermore, at the specific
item level, there was no association between similar pairs of M-COSMIC and ADOS-G items
evaluating neither the initiation of joint attention nor response to joint attention.
An ‘imperfect’ pattern of associations between the M-COSMIC and standardised
diagnostic assessment could be due to at least two factors. Firstly, different definitions of joint
attention are employed by the two instruments, with the ADOS-G requiring child direction to
code initiating joint attention whereas the M-COSMIC also accepts sharing of joint attention
here. Secondly, it should be remembered that behaviours of interest to each instrument are
measured across different settings, involving different social partners and different levels of
scaffolding provided for the child. The lack of association between items assessing response to
joint attention may have been due to the provision of a hierarchical sequence of presses for this
behaviour in the ADOS-G compared to the observation of naturally occurring (i.e.,
unprompted/pressed) behaviours in the M-COSMIC. However, clear specificity was shown in
the complete lack of association of any broad M-COSMIC code with the ADOS-G Stereotyped
Behaviour and Restricted Interests algorithm score.
6.2 M-COSMIC codes and standardised language measures
While verbal forms of communication measured by the M-COSMIC were shown to be
highly associated with scores on most standardised language measures, rates of non-verbal
communication use showed little association with language ability. Among the M-COSMIC
communication function codes, an overall pattern emerged such that ‘higher level’ functions
appeared to be more strongly associated with the standardised language scores than was the case
for ‘lower level’ communication functions. Specifically, while no significant association was
found between language scores and M-COSMIC rates of communication for behavioural