Saturday, December 19, 2015

The Global Landscape of Cognition - brilliant neuroscience paper from Hava Siegelmann and colleagues

Fig 6 of Taylor et al. ranking Paradigm Classes of
mental activity according to how connected they are
to primary sensory inputs
Hava Siegelmann sends me a link to the brilliant new paper "The global landscape of cognition: hierarchical aggregation as an organizational principle of human cortical networks and functions" she and her colleagues have produced, which is in Nature Scientific Reports.

Fig 7 of Taylor et al ranking paradigm classes from survey of
c. 500 participants. Comparisons with fMRI orders had +ve
linear relationships with low p values 
This provides a direct demonstration that human brain functions organize in global gradients of abstraction starting from sensory cortical inputs. The analysed fMRI databases (~17,000 experiments; ~1/4 of fMRI literature) and tested whether network depth predicted localization of abstract versus concrete behaviours over the whole set of studied brain regions. They developed a new cortical graph metric, termed network-depth, and objectively sorted stratified landscapes of cognition, starting from grouped sensory inputs in parallel, progressing deeper into cortex. Thus they "defined a hierarchically ordered connectome, revealing a related continuum of cognitive function".

They validated their results by surveying 500 people and asking them to rank these tasks on an abstract/concrete spectrum. This produced generally significant agreement with the computer generated levels. The p values were encouragingly low (4E-4 to 9E-9) though so alas were the r values (0.2 to 0.5). It would also have been interesting to compare these responses across different cultures. But this is the least important aspect of the study.

Certainly these results must be interpreted with due caution: the techniques they use are powerful and very interesting but (as is clear from Fig 6) the correlations on which they are based are not by any means perfect and as they note the linear regression they use for Fig 6 merely approximated the actual distributions - which are all shown as small subsets next to each word. But clearly this is a major step forward, and the kind of great, deep intellectual achievement that one would expect from Hava.

Very interesting that Naming (covert) has the highest level of abstraction. Not least because the one intellectual job God gives to Adam in Genesis is ... naming!


No comments: