Sherrington's view of cognition resulting from computations performed by individual neurons passing signals over connections does not fully capture neural dynamics. The presentation discusses: 1) Evidence of population-level phenomena like cell assemblies that challenge the importance of neuron-to-neuron connections. 2) Sherrington's view represents cognition with dedicated circuits, but population activity can represent information in high-dimensional state spaces mapped to lower dimensional manifolds. 3) reconciling descriptions of neural activity as mass action in space but flows in time may lead to the emergence of spatiotemporal scales of organization across the brain.

1. (Rick Chapman)
Sherrington’s Legacy:
Past & Future
Emmanuelle Tognoli, Center for Complex Systems & Brain Sciences
Neuroscience + Philosophy Salon, May 21st, 2021, organized by Luiz Pessoa
• Redressed S’s view on
neural collective
behavior
• Manifold,
dimensionality
• Flows and mass action
in spacetime
• Organization across
scales.
#2viewsCog

2. Thus, in the Sherringtonian view, explanations of
cognition will always take the form of computations
performed by individual neurons and signals
passed over their connections.
On collective behavior of neurons
On cell assemblies
A host of population-level phenomena are often cited in
challenge to the importance of neuron to neuron
connections17. These include reverberating activity49, cell
assemblies50, central pattern generators51 and even
oscillatory communication between brain areas52.
Freeman, Basar,
Bressler, Tognoli and
many others.

3. [In] the Sherringtonian view […] cognition
results from dedicated circuits. This implies
that only the individual nodes can be
assigned semantic contents.
On representations
Fields and Space-Time
To perform computations, the
Sherringtonian view has only two
elements at the algorithmic level: nodes
and connections.

4. Sherrington Hopfield
Neurophysiology
Physics
Spikes
Waves
Single cells, types
Collective
behavior
High dimensional
state space
Lower
dimensional
manifold
Micro-
Macro-
Wittgenstein
slide

5. BIGGER STATE SPACE IS GOOD:
“The dimensionality of the population activity,
the number of independent ways that the
population activity can vary, was then
estimated. This dimensionality was estimated
to be higher when the mixed selectivity was
included. Importantly, the neural population
showed lower dimensionality on error trials
than on correct trials, a difference that
disappeared if non-linear mixed selectivity was
left out of the decoding.”
High Dimensional
state space
Lower
dimensional
manifold
Low dimension
Haken’s synergetics
BUT LOWER DIMENSIONAL
DYNAMICS IS ESSENTIAL FOR
EFFICACY:
“In addition, many novel models of
neural data involve uncovering low-
dimensional latent structure from
high-dimensional data53,54,55.”
SEE ALSO Scholz & Shoner, 1999; Babloyantz 1989;
Ege Atlan, Sara Solla et al. 2020

6. Mass Action, but no flow
Transfer Synchronization
space
Flow, but no mass action
TIME TIME
SPACE
Tognoli & Kelso, Frontiers in System
Neuroscience, 2014
Reconciling
flows and
mass action…
(Non-reconfigurable)
modularity in space (pontifical/grand-
mother cell, center) and
decomposability in time

7. To date, no one has yet described a way to
execute computations by fast reconfigurations
in the connections between nodes.
The brain may solve
challenges at the
ecological level by
putting together
computations from the
dynamics of subcellular,
neuronal, small circuit or
large population
components.
Tognoli & Kelso, Neuron, 2014
and Tognoli & Kelso, 2013
…lead to the emergence
of spatiotemporal scales
Kuramoto & Battogtokh’s Chimera

8. Sherrington’s Legacy:
Past & Future
Emmanuelle Tognoli, Center for Complex Systems & Brain Sciences
Brain Salon, May 21st, 2021, organized by Luiz Pessoa
• Redressed S’s view on
neural collective behavior
• Manifold, dimensionality
• Flows and mass action in
spacetime
• Organization across scales
#2viewsCog
The Sherringtonian view does not reflect
Sherrington
Manifolds are lower dimensions, non monotony
to heed
Relaxing from untenably strong modular views:
metastability is one way to execute motions
through a complex coordinative system
It leads to the emergence of spatiotemporal
scales, in which scale complexity is both spatial
AND temporal