Evolution and evolvability  ·  Article

Evolvability

Evolvability concerns how living systems come to change by variation and selection, and why they do so with such success. The held texts approach it from several sides: the kinds of variation, accidental by-products, energetic niches and robust cell-level competencies. Fields and Levin (2022) argue that modular, self-modelling biological systems are highly evolvable.7

Earliest held
1859, Darwin
Most discussed in
Evolution of networks for body plan…, 2011
In the library
1,346 passages in 220 works
Rewritten
2026-11-03
01

Variation and by-products

James (1890) argued that some emotional reactions are accidental in origin. They are incidental to other reactions evolved for utility, and would never have been evolved independently for any utility they might possess.1 Bateson (1894) separated two classes of variation, meristic and substantive. In Bateson's account, either may occur independently of the other. Bateson called an appreciation of this distinction a first step towards understanding how the bodies of organisms are evolved.2 Lotka (1925) defined evolution as the history of a system in the course of irreversible transformation.3

02

Energy and variation

Sousa and colleagues (2013) argued that in a world of genes and proteins, evolution via variation and selection can readily supplant chemical determinism.4 Wallace (2013) described an energetic route for variation. Continuously arising mtDNA functional variation lets subpopulations of animal species occupy a range of energetic environments within the species' niche.5 Wallace added that subpopulations in peripheral energetic environments amass multiple mtDNA variants, and that the slower-evolving nuclear bioenergetic genes then follow.5

03

Robustness in network models

Biswas, Clawson and Levin (2022) compared memory in models of biological transcriptional networks with memory in random networks. The biological models were much more robust to edge perturbation.6 The authors suggested that evolutionary pressure to handle changing environments may explain this. They framed it as a possibility, not a result.6 They also stated that the role of trainability in evolution and development is as yet unknown.6

04

Competency and evolvability

Fields and Levin (2022) argued that systems whose parts can be changed in a modular way, by evolution and by engineers, help clarify why biological systems are so highly evolvable.7 Shreesha and Levin (2023) introduced cellular competency as another contributor to phenotypic fitness. They set it apart from organism-level learning, which needs mechanisms such as a nervous system that evolution must build. Cellular competencies come by default, because cells carry capabilities evolved when their ancestors lived independently.8 The authors present this as complementary to work on learning and on evolvability mechanisms.9

SourcesEach quotation was checked word for word against the passage it opens.
  1. many such reactions, incidental to others evolved for utility's sake, but which would never themselves have been evolved independentlyJames, W., 1890 · The Principles of Psychology, Vols. 1-2 · open at passage 2807
  2. An appreciation -of this distinction is a first step towards a comprehension of the processes by which the bodies of organisms are evolved.Bateson, W., 1894 · Materials for the Study of Variation Treated with Especial Regard to… · open at passage 92
  3. Evolution is the history of a system in the course of irreversible transformationLotka, A. J., 1925 · Elements of Physical Biology · open at passage 29
  4. In a world of gene and proteins, evolution via variation and selection can readily supplant chemical determinism.Sousa FL, Thiergart T, Landan G, Nelson-Sathi S, Pereira…, 2013 · Early bioenergetic evolution · open at passage 37
  5. The continuously arising mtDNA functional variation allows subpopulations of animal species to occupy a range of energetic environments within the species' niche.Wallace DC, 2013 · Bioenergetics in human evolution and disease: implications for the origins of… · open at passage 44
  6. The memory capability of biological models is also much more robust to edge perturbation than those of random networks, as indicated in Figure 7, perhaps due to evolutionary pressure to handle changing environments.Biswas S, Clawson W, Levin M, 2022 · Learning in Transcriptional Network Models: Computational Discovery of… · open at passage 42
  7. in which all of these aspects can be changed in a modular way: by evolution and by engineers. This helps clarify why biological systems are so highly evolvableFields C, Levin M, 2022 · Competency in Navigating Arbitrary Spaces as an Invariant for Analyzing… · open at passage 73
  8. In contrast, cellular competencies come “by default” because organisms consist of cellular components that already have many capabilities evolved during their ancestral lifetime as independent organisms.Shreesha L, Levin M, 2023 · Cellular Competency during Development Alters Evolutionary Dynamics in an… · open at passage 8
  9. Thus, it is important to begin to study how cellular competencies affect evolution, to complement approaches focused on learning, evolvability mechanismsShreesha L, Levin M, 2023 · Cellular Competency during Development Alters Evolutionary Dynamics in an… · open at passage 8
Linked ideas
Basal cognitionrelated to
Biswas et al. link pressure to adapt to variable environments to basal cognition, and robust memory in biological networks to that pressure.
Fields and Levin tie evolvability to agents modelling themselves without privileged access to self-world boundaries, which they say dissolves fixed boundaries.
Regenerationrelated to
Bongard and Levin expect work on overloaded functions in evolved and designed substrates to affect regenerative medicine.
Robustness to perturbation, as in Biswas et al.'s network memory, concerns keeping function under disturbance, which is the area homeostasis covers.
Xenobots reach a novel replication method with no evolutionary selection for being a Xenobot, which Levin uses to argue for plasticity of standard cells.
Wallace links mtDNA variants that changed membrane potential and Ca2+ metabolism to lineages that moved into new climates.
Passages link them only indirectly: Baluška & Levin say gene networks learn associatively, tying memory to evolvability; Levin 2023 cites Physarum learning as basal cognition for evolution of cognition. No direct link.
Loeb's account of heritable mutations versus non-heritable fluctuations bears on how variation can feed evolution.
Fields and Levin treat agents that build actionable self-models, and high-level reward spaces, as part of why biological systems evolve so readily.
Vitalism and mechanismin tension with
Sousa and colleagues call thermodynamics chemistry's deterministic version of natural selection, a mechanistic framing of how evolving systems arise.
Levin (2023) argues that voltage states give modularity, a known component of evolvability, between channel genes and anatomy.
Passages say neurons merely optimized ancient tricks (ion channels, gap junctions, neurotransmitters) for speed, and pre-neural signaling preceded nerves. That implies evolutionary co-option, but evolvability itself is never discussed alongside neurotransmitte
Levin 2023 argues cells' morphogenetic problem-solving competencies, inherited from unicellular ancestors, shape the evolutionary process itself, so collective cellular intelligence plausibly bears on evolvability. The passages stop short of a formal link.
Wallace treats mild mtDNA variants as functional variation that lets populations adapt to local energetic environments.
Random networks that often relax to fixed points bear on how readily evolved systems find workable solutions.
The evo-devo literature treats physical mesoscale forces as one source of developmental variation alongside gene regulatory networks.
Kouvaris and colleagues treat the capacity to vary usefully in new environments as a problem that selection alone seems unable to explain.
Friston 2013 suggests that persisting under the free energy principle implies adaptation and evolution; Levin 2019 and 2023 treat active inference as driving evolved homeostatic loops, and say its role in evolutionary advances needs study.
Feedback between ecological and evolutionary dynamics, as in Power and colleagues, bears on how populations shape their own capacity to change.
Levin's 2023 argument on why somatic experience cannot simply be written back into a genome bears on how lineages change.
Levin passages link the two: polyploid newt tubule formation shows downward causation toward one large-scale outcome, which 'evolution does' not hardwire; anatomical homeostasis coarse-grains robustness. Evolvability is implied, not directly analysed.
Kouvaris and colleagues link sparse GRN connectivity to enhanced evolvability as well as robustness.
Bose says nyctitropic movement evolved stepwise from light-induced differentiation, not 'for the purpose' of plant advantage, warning against circular final-advantage arguments; this contrasts purposive explanation with evolutionary-process explanation.
Levin's Darwin's agential materials asks how evolution searches morphology space so effectively; Kriegman and Lobo apply evolutionary algorithms in silico to design organisms and infer networks, a computational analogue of biological evolvability.
Only B's passage (Wolf 2008) mentions fitness-enhancing memory strategies as possibly evolved, but explicitly offers no evidence; no passage discusses variation, selection, or evolvability, so the link is weak.
Blackiston and colleagues say novel constructs help understand evolvability and plasticity beyond the Earthly data set.
Kashiwagi and colleagues suggest attractor selection could act as preadaptation for later signal-specific pathways, contributing to evolvability.
Kashiwagi and colleagues suggest attractor selection may be a preadaptation that contributes to evolvability.
Passages say synthetic biology and chimeric engineering blur evolved versus designed systems, and Davies lists evolvability and robustness as pattern-engineering goals; no passage directly links evolvability theory to tissue patterning.
Self-scaffolding is argued to raise evolvability by smoothing the search for fit bodies.
Davies & Glykofrydis list evolvability and robustness as goals for elaborating engineered patterns, while expanding the morphogenetic module toolbox. The passages never directly link evolvability to the cell death, fusion or adhesion modules.
Concerns the supply of variation, which Campbell links to the transmission term.
Hartl et al. use NCA models of morphogenesis with evolutionary algorithms to test how multi-scale cellular competency affects evolution itself, so NCAs serve as the tool for studying evolvability.
In silico models serve as testbeds for evolutionary dynamics: Hartl et al. run neuroevolution on simulated morphogenesis, and Budnikova and Lobo use evolutionary search scored against in vivo data. Neither passage links the concepts directly.
Kriegman and colleagues tie modularity to evolvability, since one subsystem can improve without disrupting the rest.
Evolution shapes morphostasis: Pio-Lopez says evolution prioritized development over long-term morphostasis, so aging follows without regenerative goals; Newman says canalizing evolution stabilizes phenotypes against developmental perturbation.
Modularity is argued to support evolvability, the property the network simulations examine.
Competency is proposed to change how readily evolutionary search finds good designs, which bears on how evolvability is measured.
That study also asks whether evolved modularity and robustness influence future evolvability.
Fast re-adaptation of an evolved NCA to higher noise bears on how readily solutions can evolve further.
Passages don't discuss evolvability directly; Etcheverry tests GRN robustness/versatility via push, wall, noise perturbations, while Budnikova uses evolutionary search and Padalko cites redundancy-driven robustness/versatility in evolved respiratory chains.
What evolves quickly in these simulations, such as simple networks or growing legs, bears on how readily a system can be improved.
Hartl et al. study how cellular competencies (multi-scale encoding) affect the evolutionary process itself; Pio-Lopez shows low noise yields direct encoding; Watson shows developmental gene networks generalise to novel phenotypes.
Kriegman and colleagues link modularity to evolvability, since one subsystem can improve without disrupting others.
Ten Tusscher & Hogeweg simulate evolving gene networks whose patterning uses a morphogen wavefront and segment-boundary genes, so evolvability is studied through wavefront segmentation. Other passages treat the segmentation clock only as robust or intelligent.
Redundant modules add evolvable parameters, so the topic bears on how readily cell-level policies can change over evolutionary time.
Modularity is argued to support both robustness and evolvability, by letting parts change without wrecking other functions.
Hartl et al. study how competencies of multi-scale agential components (B) affect the evolutionary process; Levin's Darwin's agential materials frames competent subunits as reshaping evolution, linking B to A.
Budnikova uses evolutionary search with graph-difference fitness for automated model discovery. Etcheverry maps GRN behavioral competencies with curiosity-driven exploration. Shreesha links cellular competency to evolutionary dynamics, so the two connect only
Ten Tusscher and Hogeweg reported that SF networks, with oscillating segment formation, were more evolvable than SS networks.
Variability of reduced parts bears on how much variation is available for change.
Both turn on 'versatility': Levin 2023 says the versatile bioelectric system was readily exapted for behavior, while Etcheverry defines versatility as reaching diverse goal states. No passage links them directly.
Where it is discussedPassages matching evolvability, evolvable, robustness, developmental plasticity
2011Evolution of networks for body plan patterning; interplay of modularity… · Ten Tusscher KH, Hogeweg P49
1930The History of Biological Theories · Rádl, E.45
2024Evolutionary Implications of Self-Assembling Cybernetic Materials with… · Hartl B, Risi S, Levin M44
1925Elements of Physical Biology · Lotka, A. J.36
2017How evolution learns to generalise: Using the principles of learning theory to… · Kouvaris K, Clune J, Kounios L, Brede…34
2016Evolutionary Connectionism: Algorithmic Principles Underlying the Evolution of… · Watson RA, Mills R, Buckley CL…32
2025AI-driven automated discovery tools reveal diverse behavioral competencies of… · Etcheverry M, Moulin-Frier C, Oudeyer…30
2012Evolution of associative learning in chemical networks · McGregor S, Vasas V, Husbands P…25
1890The Principles of Psychology, Vols. 1-2 · James, W.25
2025Aging as a Loss of Goal-Directedness: An Evolutionary Simulation and Analysis… · Pio-Lopez L, Hartl B, Levin M24