Evaluation of the criticality of in vitro neuronal networks: Toward an assessment of computational capacity

Axel Sandvig; Gunnar Tufte; Hugo Lewi Hammer; Ioanna Sandvig; Kristine Heiney; Stefano Nichele; Vibeke Devold Valderhaug

arxiv: 1907.02351 · v1 · pith:YTTLNDWCnew · submitted 2019-07-04 · 🧬 q-bio.NC · cs.ET

Evaluation of the criticality of in vitro neuronal networks: Toward an assessment of computational capacity

Kristine Heiney , Vibeke Devold Valderhaug , Ioanna Sandvig , Axel Sandvig , Gunnar Tufte , Hugo Lewi Hammer , Stefano Nichele This is my paper

Pith reviewed 2026-05-25 08:35 UTC · model grok-4.3

classification 🧬 q-bio.NC cs.ET

keywords in vitro neuronal networkscriticalityavalanchescomputational capacityelectrophysiological recordingspower-law distributionsself-learning systems

0 comments

The pith

Avalanche size distributions in in vitro neuronal networks can classify them as critical or non-critical to indicate computational capacity.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper examines electrophysiological recordings from in vitro neuronal networks to determine whether they operate in a critical state. Criticality is assessed by checking if the size distribution of network-wide activity avalanches follows a power-law form. This classification matters because the work assumes critical networks are optimally positioned for computation. The analysis supports development of models that reproduce observed behaviors for eventual use in engineered substrates. It also distinguishes networks that appear healthy from those that appear perturbed.

Core claim

The paper reports preliminary results from analyzing avalanche size distributions in recordings of in vitro neuronal networks. Networks are classified as critical when the distribution matches expectations for a critical state, and as non-critical otherwise. This classification is presented as a way to identify networks expected to perform well on computational tasks, since criticality is taken as the regime best suited for computation.

What carries the argument

The size distribution of network-wide avalanches of activity; it serves as the indicator for whether a network occupies the critical state.

If this is right

Critical networks identified this way are expected to perform well on computational tasks.
The method enables selection of networks well-suited for use in computational models.
The same analysis can classify networks as perturbed versus healthy.
Results from this classification can guide reproduction of target behaviors in other substrates.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

The same avalanche-based test could be applied to compare multiple culture conditions to find which ones most reliably produce critical dynamics.
If critical networks prove computationally superior, the classification step could become a routine filter before using recordings to train models for nanomagnetic hardware.
Extending the analysis to track how avalanche statistics change over time in the same culture might reveal whether criticality is stable or transient.

Load-bearing premise

The size distribution of avalanches accurately identifies the critical state and that state reliably correlates with computational performance.

What would settle it

A direct test in which networks classified as critical by avalanche distributions show no measurable advantage over non-critical networks when both are evaluated on the same set of computational tasks.

Figures

Figures reproduced from arXiv: 1907.02351 by Axel Sandvig, Gunnar Tufte, Hugo Lewi Hammer, Ioanna Sandvig, Kristine Heiney, Stefano Nichele, Vibeke Devold Valderhaug.

**Figure 2.** Figure 2: Probability distribution functions for three representa [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗

read the original abstract

Novel computing hardwares are necessary to keep up with today's increasing demand for data storage and processing power. In this research project, we turn to the brain for inspiration to develop novel computing substrates that are self-learning, scalable, energy-efficient, and fault-tolerant. The overarching aim of this work is to develop computational models that are able to reproduce target behaviors observed in in vitro neuronal networks. These models will be ultimately be used to aid in the realization of these behaviors in a more engineerable substrate: an array of nanomagnets. The target behaviors will be identified by analyzing electrophysiological recordings of the neuronal networks. Preliminary analysis has been performed to identify when a network is in a critical state based on the size distribution of network-wide avalanches of activity, and the results of this analysis are reported here. This classification of critical versus non-critical networks is valuable in identifying networks that can be expected to perform well on computational tasks, as criticality is widely considered to be the state in which a system is best suited for computation. This type of analysis is expected to enable the identification of networks that are well-suited for computation and the classification of networks as perturbed or healthy.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Desk Editor's Note private letter to a colleague

The paper applies standard avalanche-size analysis to label their in vitro networks critical or not, but reports no computational-task results to support the claimed link to performance.

read the letter

The main thing here is a straightforward application of avalanche-size distribution to classify a set of in vitro neuronal recordings as critical or non-critical. They follow the usual power-law check on network-wide events and report which of their cultures fall where. That part is competent but not new; the method has been used on similar preparations for years. The paper does a clean job of describing the recordings and the classification step, and it is honest that this is preliminary work aimed at later modeling and nanomagnet substrates. Credit for that clarity. The soft spot is the leap from criticality label to expected computational capacity. The text states that critical networks should perform well on tasks because the literature says so, yet it contains no reservoir-computing runs, no task accuracy numbers, and no direct comparison between the two classes on any benchmark. That assumption is imported rather than tested, so the classification stands alone as a descriptive result. The work is therefore useful to labs already running similar cultures who want a quick criticality screen, but it does not yet give evidence that the label predicts better computation. A reader looking for new methods or new performance data will not find it. I would bring this to a reading group only if the group is specifically collecting in-vitro avalanche datasets. I would not cite it for any computational-capacity claim. It is too early for peer review; the authors should add at least one concrete task comparison before sending it out.

Referee Report

1 major / 1 minor

Summary. The manuscript reports preliminary analysis of electrophysiological recordings from in vitro neuronal networks to classify them as critical versus non-critical based on the size distribution of network-wide avalanches of activity. It positions this classification as valuable for identifying networks expected to perform well on computational tasks, on the grounds that criticality is widely considered optimal for computation. The work is framed as part of a larger project to develop models reproducing target behaviors for eventual implementation in nanomagnet arrays.

Significance. If the avalanche-based classification is reliable and the assumed correlation with computational performance holds, the approach could help select suitable biological or bio-inspired substrates for unconventional computing. The manuscript applies established avalanche-analysis methods without introducing free parameters or circular derivations. However, the primary claimed significance rests on an external assumption rather than demonstrated results.

major comments (1)

[Abstract] Abstract: the statement that the classification 'is valuable in identifying networks that can be expected to perform well on computational tasks' is not supported by any results in the manuscript. No reservoir-computing benchmarks, task-performance metrics, or comparisons between critical and non-critical networks on any computational measure are reported, so the link to computational capacity remains an external assumption rather than an internal finding.

minor comments (1)

[Abstract] Abstract: no details are provided on the specific methods, data analysis procedures, statistical validation, or results, which limits assessment of the reported analysis.

Simulated Author's Rebuttal

1 responses · 0 unresolved

We thank the referee for their review and constructive feedback. We address the single major comment below and propose a targeted revision to the abstract to better align the claims with the scope of the presented preliminary analysis.

read point-by-point responses

Referee: [Abstract] Abstract: the statement that the classification 'is valuable in identifying networks that can be expected to perform well on computational tasks' is not supported by any results in the manuscript. No reservoir-computing benchmarks, task-performance metrics, or comparisons between critical and non-critical networks on any computational measure are reported, so the link to computational capacity remains an external assumption rather than an internal finding.

Authors: We agree that the manuscript reports only the avalanche-based classification of network states and does not include any direct computational benchmarks, reservoir-computing tasks, or performance comparisons. The statement in the abstract is intended to reflect the established literature consensus that criticality is associated with optimal computational properties, which motivates the larger project. However, to prevent any misinterpretation that the present study empirically validates this link, we will revise the abstract to clarify that the classification is motivated by this literature rather than presented as a demonstrated result of the current work. The revised wording will emphasize the preliminary nature of the analysis and its role in identifying target behaviors for subsequent modeling. revision: yes

Circularity Check

0 steps flagged

No significant circularity; relies on external literature for key assumption

full rationale

The manuscript performs preliminary avalanche-size analysis on in vitro recordings using standard methods to label networks as critical or non-critical. The claim that this label identifies networks suited for computation is justified by reference to the external consensus that 'criticality is widely considered to be the state in which a system is best suited for computation,' not by any internal derivation, fit, or self-citation chain. No equations, parameter fitting, or load-bearing self-citations appear; the paper contains no reservoir-computing benchmarks or task-performance results. The derivation chain is therefore self-contained against external benchmarks and exhibits no reduction of outputs to inputs by construction.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

Insufficient information in the abstract to identify specific free parameters, axioms, or invented entities.

pith-pipeline@v0.9.0 · 5765 in / 902 out tokens · 19161 ms · 2026-05-25T08:35:10.351761+00:00 · methodology

discussion (0)

Reference graph

Works this paper leans on

24 extracted references · 24 canonical work pages

[1]

Computation at the edge of chaos: Phase transitions and emergent computation,

C. G. Langton, “Computation at the edge of chaos: Phase transitions and emergent computation,” Physica D, vol. 40, pp. 12–37, 1990

work page 1990
[2]

Stepney, S

S. Stepney, S. Rasmussen, and M. Amos, Computational Matter, 1st ed. Springer Publishing Company, Incorpo- rated, 2018

work page 2018
[3]

Computation in artiﬁcial spin ice,

J. H. Jensen, E. Folven, and G. Tufte, “Computation in artiﬁcial spin ice,” The 2018 Conference on Artiﬁcial Life: A Hybrid of the European Conference on Artiﬁcial Life (ECAL) and the International Conference on the Synthesis and Simulation of Living Systems (ALIFE) , no. 30, pp. 15–22, 2018

work page 2018
[4]

The science of self-organization and adap- tivity,

F. Heylighen, “The science of self-organization and adap- tivity,” in in: Knowledge Management, Organizational Intelligence and Learning, and Complexity, in: The En- cyclopedia of Life Support Systems, EOLSS . Publishers Co. Ltd, 1999, pp. 253–280

work page 1999
[5]

A review of mor- phogenetic engineering,

R. Doursat, H. Sayama, and O. Michel, “A review of mor- phogenetic engineering,” Natural Computing, vol. 12, pp. 517–535, 2013

work page 2013
[6]

An overview of reservoir computing: theory, applica- tions and implementations,

B. Schrauwen, D. Verstraeten, and J. Van Campenhout, “An overview of reservoir computing: theory, applica- tions and implementations,” in Proceedings of the 15th European Symposium on Artiﬁcial Neural Networks , 2007, pp. 471–482

work page 2007
[7]

Towards making a cyborg: A closed-loop reservoir-neuro system,

P. Aaser, M. Knudsen, O. Huse Ramstad, R. van de Wijdeven, S. Nichele, I. Sandvig, G. Tufte, U. S. Bauer, Ø. Halaas, S. Hendseth et al., “Towards making a cyborg: A closed-loop reservoir-neuro system,” in Proceedings of the European Conference on Artiﬁcial Life 2017 . MIT Press, 2017

work page 2017
[8]

[Online]

(2018) Socrates: Self-organizing computational sub- strates. [Online]. Available: https://www.ntnu.edu/ socrates

work page 2018
[9]

Self-organization and neu- ronal avalanches in networks of dissociated cortical neu- rons,

V . Pasquale, P. Massobrio, L. L. Bologna, M. Chiap- palone, and S. Martinoia, “Self-organization and neu- ronal avalanches in networks of dissociated cortical neu- rons,” Neuroscience, vol. 153, pp. 1354–1369, 2008

work page 2008
[10]

Self- organized criticality in cortical assemblies occurs in con- current scale-free and small-world networks,

P. Massobrio, V . Pasquale, and S. Martinoia, “Self- organized criticality in cortical assemblies occurs in con- current scale-free and small-world networks,” Scientiﬁc Reports, 2015

work page 2015
[11]

The functional beneﬁts of criticality in the cortex,

W. L. Shew and D. Plenz, “The functional beneﬁts of criticality in the cortex,” The Neuroscientist , vol. 19, no. 1, pp. 88–100, 2013

work page 2013
[12]

Self-organized criticality as a fundamental propertie of neural systems,

J. Hesse and T. Gross, “Self-organized criticality as a fundamental propertie of neural systems,” Frontiers in Systems Neuroscience, vol. 8, 2014

work page 2014
[13]

Neuronal avalanches in neo- cortical circuits,

J. M. Beggs and D. Plenz, “Neuronal avalanches in neo- cortical circuits,” The Journal of Neuroscience , vol. 23, no. 35, pp. 11 167–11 177, 2003

work page 2003
[14]

Self-organized criticality: An explanation of the 1/f noise,

P. Bak, C. Tang, and K. Wiesenfeld, “Self-organized criticality: An explanation of the 1/f noise,” Physical Review Letters, vol. 59, pp. 381–384, Jul 1987

work page 1987
[15]

Neuronal avalanches imply maximum dy- namic range in cortical networks at criticality,

W. L. Shew, H. Yang, T. Petermann, R. Roy, and D. Plenz, “Neuronal avalanches imply maximum dy- namic range in cortical networks at criticality,” Journal of Neuroscience, vol. 29, no. 49, pp. 15 595–15 600, 2009

work page 2009
[16]

Information capacity and transmission are maximized in balanced cortical networks with neuronal avalanches,

W. L. Shew, H. Yang, S. Yu, R. Roy, and D. Plenz, “Information capacity and transmission are maximized in balanced cortical networks with neuronal avalanches,” Journal of Neuroscience, vol. 31, no. 1, pp. 55–63, 2011

work page 2011
[17]

Self-organized criticality in developing neu- ronal networks,

C. Tetzlaff, S. Okujeni, U. Egert, W org otter, and M. Butz, “Self-organized criticality in developing neu- ronal networks,” PLoS Computational Biology , vol. 6, no. 12, 2010

work page 2010
[18]

Development of neu- ral population activity toward self-organized criticality,

Y . Yada, T. Mita, A. Sanada, R. Yano, D. J. Bakkum, A. Hierlemann, and H. Takahashi, “Development of neu- ral population activity toward self-organized criticality,” Neuroscience, pp. 55–65, 2017

work page 2017
[19]

Generating regionalized neuronal cells from pluripotency, a step-by- step protocol,

A. Kirkeby, J. Nelander, and M. Parmar, “Generating regionalized neuronal cells from pluripotency, a step-by- step protocol,” Frontiers in Cellular Neuroscience, vol. 6, p. 64, 2012

work page 2012
[20]

Predictive markers guide differentiation to improve graft outcome in clinical translation of hesc-based therapy for parkinson’s disease,

A. Kirkeby, S. Nolbrant, K. Tiklova, A. Heuer, N. Kee, T. Cardoso, D. R. Ottosson, M. J. Lelos, P. Rifes, 5 S. B. Dunnett, S. Grealish, T. Perlmann, and M. Parmar, “Predictive markers guide differentiation to improve graft outcome in clinical translation of hesc-based therapy for parkinson’s disease,” Cell Stem Cell , vol. 20, no. 1, pp. 135–148, 2017

work page 2017
[21]

Isolation of human in- duced pluripotent stem cell-derived dopaminergic pro- genitors by cell sorting for successful transplantation,

D. Doi, B. Samata, M. Katsukawa, T. Kikuchi, A. Morizane, Y . Ono, K. Sekiguchi, M. Nakagawa, M. Parmar, and J. Takahashi, “Isolation of human in- duced pluripotent stem cell-derived dopaminergic pro- genitors by cell sorting for successful transplantation,” Stem Cell Reports , vol. 2, no. 3, pp. 337–350, 2014

work page 2014
[22]

Power-law distributions in empirical data,

A. Clauset, C. R. Shalizi, and M. E. J. Newman, “Power-law distributions in empirical data,” SIAM Re- view, vol. 51, no. 4, pp. 661–703, 2009

work page 2009
[23]

Deep learning with cel- lular automaton-based reservoir computing,

S. Nichele and A. Molund, “Deep learning with cel- lular automaton-based reservoir computing,” Complex Systems, vol. 26, 2017

work page 2017
[24]

Reservoir computing using nonuniform binary cellular automata,

S. Nichele and M. S. Gundersen, “Reservoir computing using nonuniform binary cellular automata,” Complex Systems, vol. 26, 2017. 6

work page 2017

[1] [1]

Computation at the edge of chaos: Phase transitions and emergent computation,

C. G. Langton, “Computation at the edge of chaos: Phase transitions and emergent computation,” Physica D, vol. 40, pp. 12–37, 1990

work page 1990

[2] [2]

Stepney, S

S. Stepney, S. Rasmussen, and M. Amos, Computational Matter, 1st ed. Springer Publishing Company, Incorpo- rated, 2018

work page 2018

[3] [3]

Computation in artiﬁcial spin ice,

J. H. Jensen, E. Folven, and G. Tufte, “Computation in artiﬁcial spin ice,” The 2018 Conference on Artiﬁcial Life: A Hybrid of the European Conference on Artiﬁcial Life (ECAL) and the International Conference on the Synthesis and Simulation of Living Systems (ALIFE) , no. 30, pp. 15–22, 2018

work page 2018

[4] [4]

The science of self-organization and adap- tivity,

F. Heylighen, “The science of self-organization and adap- tivity,” in in: Knowledge Management, Organizational Intelligence and Learning, and Complexity, in: The En- cyclopedia of Life Support Systems, EOLSS . Publishers Co. Ltd, 1999, pp. 253–280

work page 1999

[5] [5]

A review of mor- phogenetic engineering,

R. Doursat, H. Sayama, and O. Michel, “A review of mor- phogenetic engineering,” Natural Computing, vol. 12, pp. 517–535, 2013

work page 2013

[6] [6]

An overview of reservoir computing: theory, applica- tions and implementations,

B. Schrauwen, D. Verstraeten, and J. Van Campenhout, “An overview of reservoir computing: theory, applica- tions and implementations,” in Proceedings of the 15th European Symposium on Artiﬁcial Neural Networks , 2007, pp. 471–482

work page 2007

[7] [7]

Towards making a cyborg: A closed-loop reservoir-neuro system,

P. Aaser, M. Knudsen, O. Huse Ramstad, R. van de Wijdeven, S. Nichele, I. Sandvig, G. Tufte, U. S. Bauer, Ø. Halaas, S. Hendseth et al., “Towards making a cyborg: A closed-loop reservoir-neuro system,” in Proceedings of the European Conference on Artiﬁcial Life 2017 . MIT Press, 2017

work page 2017

[8] [8]

[Online]

(2018) Socrates: Self-organizing computational sub- strates. [Online]. Available: https://www.ntnu.edu/ socrates

work page 2018

[9] [9]

Self-organization and neu- ronal avalanches in networks of dissociated cortical neu- rons,

V . Pasquale, P. Massobrio, L. L. Bologna, M. Chiap- palone, and S. Martinoia, “Self-organization and neu- ronal avalanches in networks of dissociated cortical neu- rons,” Neuroscience, vol. 153, pp. 1354–1369, 2008

work page 2008

[10] [10]

Self- organized criticality in cortical assemblies occurs in con- current scale-free and small-world networks,

P. Massobrio, V . Pasquale, and S. Martinoia, “Self- organized criticality in cortical assemblies occurs in con- current scale-free and small-world networks,” Scientiﬁc Reports, 2015

work page 2015

[11] [11]

The functional beneﬁts of criticality in the cortex,

W. L. Shew and D. Plenz, “The functional beneﬁts of criticality in the cortex,” The Neuroscientist , vol. 19, no. 1, pp. 88–100, 2013

work page 2013

[12] [12]

Self-organized criticality as a fundamental propertie of neural systems,

J. Hesse and T. Gross, “Self-organized criticality as a fundamental propertie of neural systems,” Frontiers in Systems Neuroscience, vol. 8, 2014

work page 2014

[13] [13]

Neuronal avalanches in neo- cortical circuits,

J. M. Beggs and D. Plenz, “Neuronal avalanches in neo- cortical circuits,” The Journal of Neuroscience , vol. 23, no. 35, pp. 11 167–11 177, 2003

work page 2003

[14] [14]

Self-organized criticality: An explanation of the 1/f noise,

P. Bak, C. Tang, and K. Wiesenfeld, “Self-organized criticality: An explanation of the 1/f noise,” Physical Review Letters, vol. 59, pp. 381–384, Jul 1987

work page 1987

[15] [15]

Neuronal avalanches imply maximum dy- namic range in cortical networks at criticality,

W. L. Shew, H. Yang, T. Petermann, R. Roy, and D. Plenz, “Neuronal avalanches imply maximum dy- namic range in cortical networks at criticality,” Journal of Neuroscience, vol. 29, no. 49, pp. 15 595–15 600, 2009

work page 2009

[16] [16]

Information capacity and transmission are maximized in balanced cortical networks with neuronal avalanches,

W. L. Shew, H. Yang, S. Yu, R. Roy, and D. Plenz, “Information capacity and transmission are maximized in balanced cortical networks with neuronal avalanches,” Journal of Neuroscience, vol. 31, no. 1, pp. 55–63, 2011

work page 2011

[17] [17]

Self-organized criticality in developing neu- ronal networks,

C. Tetzlaff, S. Okujeni, U. Egert, W org otter, and M. Butz, “Self-organized criticality in developing neu- ronal networks,” PLoS Computational Biology , vol. 6, no. 12, 2010

work page 2010

[18] [18]

Development of neu- ral population activity toward self-organized criticality,

Y . Yada, T. Mita, A. Sanada, R. Yano, D. J. Bakkum, A. Hierlemann, and H. Takahashi, “Development of neu- ral population activity toward self-organized criticality,” Neuroscience, pp. 55–65, 2017

work page 2017

[19] [19]

Generating regionalized neuronal cells from pluripotency, a step-by- step protocol,

A. Kirkeby, J. Nelander, and M. Parmar, “Generating regionalized neuronal cells from pluripotency, a step-by- step protocol,” Frontiers in Cellular Neuroscience, vol. 6, p. 64, 2012

work page 2012

[20] [20]

Predictive markers guide differentiation to improve graft outcome in clinical translation of hesc-based therapy for parkinson’s disease,

A. Kirkeby, S. Nolbrant, K. Tiklova, A. Heuer, N. Kee, T. Cardoso, D. R. Ottosson, M. J. Lelos, P. Rifes, 5 S. B. Dunnett, S. Grealish, T. Perlmann, and M. Parmar, “Predictive markers guide differentiation to improve graft outcome in clinical translation of hesc-based therapy for parkinson’s disease,” Cell Stem Cell , vol. 20, no. 1, pp. 135–148, 2017

work page 2017

[21] [21]

Isolation of human in- duced pluripotent stem cell-derived dopaminergic pro- genitors by cell sorting for successful transplantation,

D. Doi, B. Samata, M. Katsukawa, T. Kikuchi, A. Morizane, Y . Ono, K. Sekiguchi, M. Nakagawa, M. Parmar, and J. Takahashi, “Isolation of human in- duced pluripotent stem cell-derived dopaminergic pro- genitors by cell sorting for successful transplantation,” Stem Cell Reports , vol. 2, no. 3, pp. 337–350, 2014

work page 2014

[22] [22]

Power-law distributions in empirical data,

A. Clauset, C. R. Shalizi, and M. E. J. Newman, “Power-law distributions in empirical data,” SIAM Re- view, vol. 51, no. 4, pp. 661–703, 2009

work page 2009

[23] [23]

Deep learning with cel- lular automaton-based reservoir computing,

S. Nichele and A. Molund, “Deep learning with cel- lular automaton-based reservoir computing,” Complex Systems, vol. 26, 2017

work page 2017

[24] [24]

Reservoir computing using nonuniform binary cellular automata,

S. Nichele and M. S. Gundersen, “Reservoir computing using nonuniform binary cellular automata,” Complex Systems, vol. 26, 2017. 6

work page 2017