Riccardo FusaroliWe put the learned lessons into play in a new study on a longitudinal dataset with 66 infant-caregiver dyads (30 autistic), from 18-62 months, 6 visits per child; turn by turn data; clinical, linguistic and cognitive assessment (https://doi.org/10.1016/j.cognition.2018.10.022). Adequately modeling the data (ex-gaussian including overlapping, left): fast responses, speeding w age, surprisingly autistic children are faster! replicating methodological pitfalls (right): >1s, slower in autism.
Including socio-cognitive, linguistic and motor skills improves the model. The first reduces overlapping and long latencies, the latter two are related to faster responses. 13/
Including turn-by-turn dynamics further improve the model: children both auto-correlate (going slower at turn t is related to going slower at t+1) and follow the other. Auto correlation is stronger in children than adults, interpersonal adjustment is stronger in adults 14/
These findings mostly generalize to a second corpus (28 (12f) autistic children, 20 (12f) TD – 10yo on average; measured 7 times at a distance of 1 week). 15/
We are now working on adding measures of linguistic complexity and turn-by-turn contingency; and more excitingly on adopting and adapting formal mechanistic models from the non-human animal literature. That is a wholly fascinating field, both a source of inspiration for doing things better, but also a refreshing way to identify things that humans do that are not accounted in current non-human animal models. 16/
See the video (youtu.be/-X5Z6dQuiCk) for more (and keep an eye open for more preprints to come). Thanks to all my collaborators and the wonderful groups who invited me to present this stuff (and provided very useful discussion)
