Properly addressing this issue requires a comprehensive
theoretical approach to the relationship between integrated
information, emergence, and memory (Balduzzi and
Tononi, unpubl.). The working hypothesis is as follows
(Tononi, 2004): In general, for any system, integrated in-
formation is generated at multiple spatiotemporal scales. In
particular, however, there will often be a privileged spatio-
temporal “grain size” at which a given system forms a
complex of highest �—the spatiotemporal scale at which it
“exists” the most in terms of integrated information, and
therefore of consciousness.
For example, while in the brain there are many more
atoms than neurons, it is likely that complexes at the spatial
scale of atoms are exceedingly small, or at any rate that they
cannot maintain both functional specialization and long-
range integration, thus yielding low values of �. At the
other extreme, the spatial scale of cortical areas is almost
certainly too coarse for yielding high values of �. Some-
where in between, most naturally at the grain size of neu-
rons or minicolumns, the neuroanatomical arrangement en-
sures an ideal mix of functional specialization and
integration, leading to the formation of a large complex of
high �.Similarly, with respect to time, neurons would yield zero
� at the scale of microseconds, since there is simply not
enough time for engaging their mechanisms. At long time
scales, say hours, � would also be low, as output states
would bear little relationship to input states. Somewhere in
between, at a time scale of tens to hundreds of milliseconds,
the firing pattern of a large complex of neurons should be
maximally predictive of its previous state, thus yielding
high �.