RESEARCH
Correlated activity favors synergistic processing
in local cortical networks in vitro at synaptically
relevant timescales
Samantha P. Sherrill
1, Nicholas M. Timme2, John M. Beggs3, and Ehren L. Newman1
1Department of Psychological and Brain Sciences and Program in Neuroscience, Indiana University Bloomington,
Bloomington, IN, USA
2Department of Psychology, Indiana University-Purdue University Indianapolis, Indianapolis, IN, USA
3Department of Physics & Program in Neuroscience, Indiana University Bloomington, Bloomington, IN, USA
a n o p e n a c c e s s
j o u r n a l
Keywords: Neural information processing, Correlated activity, Computation, In vitro networks,
Effective connectivity, Cortex, Information theory, Synergy
ABSTRACT
Neural information processing is widely understood to depend on correlations in neuronal
activity. Tuttavia, whether correlation is favorable or not is contentious. Here, we sought to
determine how correlated activity and information processing are related in cortical circuits.
Using recordings of hundreds of spiking neurons in organotypic cultures of mouse neocortex,
we asked whether mutual information between neurons that feed into a common third
neuron increased synergistic information processing by the receiving neuron. We found that
mutual information and synergistic processing were positively related at synaptic timescales
(0.05–14 ms), where mutual information values were low. This effect was mediated by the
increase in information transmission—of which synergistic processing is a component—that
resulted as mutual information grew. Tuttavia, at extrasynaptic windows (up to 3,000 ms),
where mutual information values were high, the relationship between mutual information
and synergistic processing became negative. In this regime, greater mutual information
resulted in a disproportionate increase in redundancy relative to information transmission.
These results indicate that the emergence of synergistic processing from correlated activity
differs according to timescale and correlation regime. In a low-correlation regime, synergistic
processing increases with greater correlation, and in a high-correlation regime, synergistic
processing decreases with greater correlation.
Citation: Sherrill, S. P., Timme, N. M.,
Beggs, J. M., & Newman, E. l. (2020).
Correlated activity favors synergistic
processing in local cortical networks
in vitro at synaptically relevant
timescales. Network Neuroscience,
4(3) 678–697. https://doi.org/10.1162/
netn_a_00141
DOI:
https://doi.org/10.1162/netn_a_00141
Supporting Information:
https://doi.org/10.1162/netn_a_00141
Received: 23 Gennaio 2020
Accepted: 6 April 2020
Competing Interests: The authors have
declared that no competing interests
exist.
AUTHOR SUMMARY
Corresponding Author:
Samantha P. Sherrill
samfaber@indiana.edu
Handling Editor:
Jason MacLean
Copyright: © 2020
Istituto di Tecnologia del Massachussetts
Pubblicato sotto Creative Commons
Attribuzione 4.0 Internazionale
(CC BY 4.0) licenza
The MIT Press
In the present work, we address the question of whether correlated activity in functional
networks of cortical circuits supports neural computation. To do so, we combined network
analysis with information theoretic tools to analyze the spiking activity of hundreds of
neurons recorded from organotypic cultures of mouse somatosensory cortex. We found that,
at timescales most relevant to direct neuronal communication, neurons with more correlated
activity predicted greater computation, suggesting that correlated activity does support
computation in cortical circuits. Importantly, this result reversed at timescales less relevant to
direct neuronal communication, where even greater correlated activity predicted decreased
computation. Così, the relationship between correlated activity and computation depends
on the timescale and the degree of correlation in neuronal interactions.
l
D
o
w
N
o
UN
D
e
D
F
R
o
M
H
T
T
P
:
/
/
D
io
R
e
C
T
.
M
io
T
.
/
T
/
e
D
tu
N
e
N
UN
R
T
io
C
e
–
P
D
l
F
/
/
/
/
/
4
3
6
7
8
1
8
6
7
3
0
1
N
e
N
_
UN
_
0
0
1
4
1
P
D
.
T
F
B
sì
G
tu
e
S
T
T
o
N
0
7
S
e
P
e
M
B
e
R
2
0
2
3
Correlation favors computation in networks at synaptic timescales
Information processing:
The transfer, storage, E
computation (or modification)
of information.
Correlated activity:
The synchronous spiking of two
neurons.
Information transmission:
Here we use this terminology to
refer to the multivariate transfer
entropy obtained when two neurons
predict a third.
Redundancy:
A measure that quantifies the
overlapping information gained
about a third neuron by knowing
the spiking of two neurons.
INTRODUCTION
What role does the correlated activity among cortical neurons play in neural information pro-
cessazione? Correlated activity is ubiquitous throughout the brain, emerging from both external
stimuli and internal dynamics. Correlated activity is predictive of information processing (for
a review, see Salinas & Sejnowski, 2001). Tuttavia, the extent to which it is favorable for
information processing is not clear. What is needed to better understand the role of neural
correlations in information processing is a comparison of how the amount of correlated ac-
tivity between upstream neurons relates to the amount of resulting information processing in
cortical microcircuits.
The view that correlated neural activity is favorable for neural information processing is
widely held within the cognitive rhythms community and is based on the idea that correlation
facilitates both communication between circuits and the orchestration of processing within
circuits. Correlated neural activity, especially synchronous activity, is understood to generate
the coherent rhythms that are observed in local field potentials, electrocorticography, and elec-
troencephalography that are theorized to subserve specific computational or cognitive mech-
anismo (per esempio., Fries, 2015; Hasselmo et al., 2002; Hernandez et al., 2020; Honey et al., 2017;
Lisman & Jensen, 2013; Newman et al., 2014; Norman et al., 2006; Ward, 2003). Ample
empirical evidence derived from in vitro, in silico, and in vivo studies supports the importance
of synchrony for organizing information transmission in cortical circuits (Averbeck & Lee,
2004; Azouz & Gray, 2003; Fries, 2015; Poulet & Petersen, 2008; Salinas & Sejnowski, 2001;
Yu et al., 2008). The synchronization of neuronal spiking is indeed linked to higher order
cognitive and behavioral processes (Grammont & Riehle, 1999; Riehle et al., 1997; Vinck
et al., 2015). From this perspective, correlation is favorable for information processing. Yet,
even within the cognitive rhythms community there is recognition that excess correlation
can be unfavorable and that, in some circumstances, desynchronization supports informa-
tion processing better than synchronization (per esempio., Bastos et al., 2015; Hanslmayr et al., 2012;
van Winsun et al., 1984). The discrepancy between this and the standard view of the cognitive
rhythms community warrants reconciliation (Hanslmayr, Staresina, & Bowman, 2016).
The view that correlated activity is unfavorable for neural information processing is widely
held in the sensory processing and artificial neural network communities and is based on the
idea that correlation is synonymous with redundancy and thus reduces efficiency (Attneave,
1954; Barlow, 1961; Gutnisky & Dragoi, 2008; Schneidman, Bialek, & Berry, 2003; Shadlen
& Newsome, 1998). This bandwidth-limiting effect of correlation motivates the “redundancy-
reduction hypothesis” (Atick & Redlich, 1990; Attneave, 1954; Barlow, 1961; Shadlen &
Newsome, 1998). In this line of thinking, a normative goal of sensory information process-
ing is to reduce the redundancy of neuronal signals (Atick & Redlich, 1992; Barlow, 1961;
Field, 1987; Gutnisky & Dragoi, 2008; Laughlin, 1989; Rieke et al., 1995; van Hateren, 1992).
Many works, together, indicate that signal redundancy decreases from lower order sensory ar-
eas to higher order sensory areas (Berry et al., 1997; Chechik et al., 2006; Dan et al., 1998;
Doi et al., 2012; Montani et al., 2007; Nirenberg et al., 2001; Puchalla et al., 2005; Reich
et al., 2001; Reinagel & Reid, 2000; Zohary et al., 1994). From this perspective, correlation is
unfavorable for information processing. Yet, as in the cognitive rhythms community, there is an
indication that qualification is needed to this standard view, given recent empirical evidence
that correlated activity can also increase processing (Nigam, Pojoga, & Dragoi, 2019).
Network Neuroscience
679
l
D
o
w
N
o
UN
D
e
D
F
R
o
M
H
T
T
P
:
/
/
D
io
R
e
C
T
.
M
io
T
.
/
/
T
e
D
tu
N
e
N
UN
R
T
io
C
e
–
P
D
l
F
/
/
/
/
/
4
3
6
7
8
1
8
6
7
3
0
1
N
e
N
_
UN
_
0
0
1
4
1
P
D
.
T
F
B
sì
G
tu
e
S
T
T
o
N
0
7
S
e
P
e
M
B
e
R
2
0
2
3
Correlation favors computation in networks at synaptic timescales
Timescale:
The range of time in which delays
between spiking neurons were
considered.
Organotypic culture:
A cell culture derived from
tissue—that retains many of the
structural and functional properties
of the intact tissue.
Information:
The reduction in uncertainty,
typically measured in bits.
Synergy:
A measure that quantifies the
information gained about a third
neuron by knowing the spiking of
two neurons jointly.
Our aim with the work described here was to determine which of these two perspectives
better accounts for the relationship between correlation and synergistic processing—a com-
ponent of information processing—in local cortical microcircuits at synaptic timescales and
beyond. To accomplish this goal, we analyzed the spiking activity of hundreds of neurons
recorded simultaneously from each of 25 organotypic cultures of mouse somatosensory cortex
(Figura 1). In these recordings we identified hundreds of thousands of triads wherein individ-
ual neurons received significant functional input from two other neurons at synaptic timescales
(<14 ms). For each triad, we measured the total information about receiving neurons firing
that was carried in sending activity and then decomposed this into constituent
components: unique contributions of sender, redundancy between senders
(i.e., redundancy), synergy senders (i.e., synergistic processing). Across
triads, found that correlation low, but greater correlations
were predictive processing by neuron. When extended
the timescale analysis to consider triads at extrasynaptic windows (up 3,000 ms),
we observed a significant increase overall senders’ activity.
We also shift negative relationship sender synergistic
processing. Secondary analyses transmission indicated that
l
D
o
w
n
o
a
d
e
d
f
r
o
m
h
t
t
p
:
>
