Adaptation by Cumulative Selection
Listen
Radio episode about this paper
Transcript
Introduction to the show: ident: Genomics Radio. Generated commentary on the latest computational biology and genomics papers.
Ines: We're looking at the paper: "Adaptation by Cumulative Selection".
Marcus: Here's the gist of the paper. Biological systems like long-lived clonal organisms, holobionts and clades challenge traditional evolutionary thinking since they adapt without populations or reproduction <ref:2602.20883#pg2> This paper aims to provide an overarching theoretical framework which encompasses standard Darwinian evolution as well as other processes of adaptation <ref:2602.20883#pg2> The framework is cumulative selection and the author provides a general ‘recipe’ for it to occur <ref:2602.20883#pg2>.
Ines: The Core Concept of Cumulative Selection Cumulative selection is defined as a random search process that can generate an adaptation in a ‘reasonable’ time by the accumulation of small changes <ref:2602.20883#pg2> This process requires three general conditions to operate <ref:2602.20883#pg2>.
Ines: The features vary randomly over time <ref:2602.20883#pg2>.
Paper summary: Ines: So looking at the paper Adaptation by Cumulative Selection we see they’ve put together a framework that goes beyond just standard Darwinian evolution <ref:2602.20883#pg1>. They define cumulative selection as this random search process driven by accumulating small changes over time <ref:2602.20883#pg4>.
Marcus: The authors provide a general recipe for this process and show that standard natural selection is just one specific way this can happen <ref:2602.20883#pg1>. It emphasizes that persistence based selection is a key idea for things like clonal organisms or even the modern nitrogen cycle <ref:2602.20883#pg5>.
Yuki: The core implication I see is that we need to expand our view of adaptation to include these other forms of persistence, not just those driven by reproduction <ref:2602.20883#pg4>. It’s about recognizing that patterns can persist through other mechanisms too.
Ines: They conclude that the general conditions they laid out are necessary and sufficient for cumulative selection to occur in a system <ref:2602.20883#pg4>. This means those three specific conditions—random feature variation, different persistence rates, and correlation between parents and offspring—are the rules of the game <ref:2602.20883#pg4>.
Marcus: What this really means for us is that when we study a biological system, we need to check if it meets those criteria for cumulative selection <ref:2602.20883#pg4>. If it does, then the adaptation process is likely running through this mechanism instead of just traditional population dynamics <ref:2602.20883#pg1>.
Yuki: It helps us understand complexity by looking at how innovations build on previous innovations across different types of systems, whether it's a species or a geological cycle <ref:2602.20883#pg5>.
Ines: So the paper Adaptation by Cumulative Selection is offering a broad theoretical lens for understanding how adaptation arises in diverse biological and even non-biological systems <ref:2602.20883#pg1>. It’s about seeing the bigger picture of how selection can work without relying only on the familiar population model <ref:2602.20883#pg4>.
Marcus: It’s a framework that helps us think about how long-term patterns can be established through accumulation, which is something we need to consider when analyzing large cohorts or long-lived organisms <ref:2602.20883#pg5>.
Conclusion: Ines: So this paper, "Adaptation by Cumulative Selection," it’s basically proposing this big theory about how things change over time in biological systems, moving beyond just standard Darwinian selection models <ref:2602.20883#pg1>.
Marcus: It boils down to cumulative selection being a random search process where adaptation happens by adding up many small changes over time, not just one big jump <ref:2602.20883#pg4>.
Yuki: The authors say this framework isn't limited to just populations that reproduce; it applies to things like long-lived organisms or even how certain chemical cycles work <ref:2602.20883#pg5>.
Ines: They lay out three specific conditions that must be met for this cumulative selection process to actually happen in a system <ref:2602.20883#pg4>. First, the traits have to change randomly over time <ref:2602.20883#pg5>.
Marcus: Second, different states need different rates of persistence; some changes stick around longer than others <ref:2602.20883#pg4>. And third, the speed at which variations appear can't be faster than some set error threshold <ref:2602.20883#pg4>.
Yuki: That last point about the variation not exceeding an error threshold is interesting because it ties the randomness to some kind of constraint on how fast evolution can really run <ref:2602.20883#pg5>.
Ines: It connects back to what we see in Darwinian biology, because those same three conditions are exactly what Lewontin described for standard natural selection <ref:2602.20883#pg4>.
Marcus: So the implication is that if we see an adaptation happening, we have to check if it fits this cumulative selection recipe instead of just assuming simple reproduction driven selection <ref:2602.20883#pg1>.
Yuki: It means when you look at a species history, you might not be looking at a single winning trait but the accumulation of many tiny, slightly different traits that eventually build up something new <ref:2602.20883#pg5>.
Ines: It’s about seeing adaptation as this kind of slow accumulation rather than just immediate selection on a population level <ref:2602.20883#pg1>.
Marcus: So, the paper suggests that understanding the mechanism of persistence is just as important as understanding reproduction when you analyze large cohorts or long-lived biological systems <ref:2602.20883#pg5>.
Yuki: It opens up a new way to think about how complex patterns form across different kinds of life and history <ref:2602.20883#pg1>.
Department of Computer Science, University of Exeter
q-bio.PE
Submitted: 2026-02-24
Updated: 2026-11-04
License: http://creativecommons.org/licenses/by/4.0/
Importance score: 89/100
The gist: Biological systems like long-lived clonal organisms, holobionts and clades challenge traditional evolutionary thinking since they adapt without populations or reproduction <ref:2602.20883#pg2> This…
Key concepts
- Cumulative Selection
- This is defined as a random search process where an adaptation arises from the accumulation of many small changes over time. It requires three specific conditions: features must vary randomly, different states must persist at different rates, and the rate of variation must not exceed an error threshold.
- Conditions for Adaptation
- These are the necessary and sufficient requirements for cumulative selection to function in a system. In Darwinian populations, these conditions map directly onto Lewontin’s recipe, linking phenotypic variation to condition one, differential persistence to condition two, and memory between parents and offspring to condition three.
- Formalizing the Model
- The model uses a process where N new sequences are generated by copying the best current state with probabilities determined by mutation rates. The expected time required for discovery (T) is approximated by a formula involving the logarithm of L, gamma (Euler-Mascheroni constant), and p.
Terminology
Summary
Biological systems like long-lived clonal organisms, holobionts and clades challenge traditional evolutionary thinking since they adapt without populations or reproduction <ref:2602. p. 0 Adaptation by Cumulative Selection
Rudy Arthur∗1
1Department of Computer Science, University of Exeter, Exeter EX4
4PY, UK
February 25, 2026
Abstract
Biological systems like long-lived clonal organisms, holobionts and clades
challenge traditional evolutionary thinking since they adapt without
populations or reproduction. This paper aims to provide an overarching theoretical framework which encompasses standard Darwinian
evolution as well as other processes of adaptation. Adaptation by Cumulative Selection
Rudy Arthur∗1
1Department of Computer Science, University of Exeter, Exeter EX4
4PY, UK
February 25, 2026
Abstract
Biological systems like long-lived clonal organisms, holobionts and clades
challenge traditional evolutionary thinking since they adapt without
populations or reproduction. This paper aims to provide an overarching theoretical framework which encompasses standard Darwinian
evolution as well as other processes of adaptation. This paper aims to provide an overarching theoretical framework which encompasses standard Darwinian
evolution as well as other processes of adaptation. This framework is
cumulative selection and I provide a general ‘recipe’ for it to occur.Where the paper says it
Where the paper says it
Where the paper says it
The Core Concept of Cumulative Selection
Cumulative selection is defined as a random search process that can generate an adaptation in a ‘reasonable’ time by the accumulation of small changes <ref:2602.20883#pg2> This process requires three general conditions to operate <ref:2602. p. 2 In this paper I will explore cumulative selection in detail and provide
general conditions for when it can occur. These general conditions are analogous to Lewontin’s recipe [Lewontin, 1970], in that they give necessary and
sufficient conditions for cumulative selection to operate.Where the paper says it
-
The features vary randomly over time <ref:2602. p. 0
20883#pg2>.Where the paper says it
Arthur@exeter. uk 1 arXiv:2602.
-
Different states have different rates of persistence <ref:2602. p. 9
20883#pg2>.Where the paper says it
the next step is 1 − (1 − p) L−k , so a state with k + 1 correct letters will persist longer than a state with k. Different sequences have different rates of persistence.
-
The rate of variation does not exceed the error threshold <ref:2602. p. 9
20883#pg2>.Where the paper says it
There needs to be enough continuity for the process to actually be cumulative, we can’t build on previous matches if those matches are constantly being lost. In the lottery model we define the error threshold as the critical value νc above which there is no cumulative selection and the search becomes effectively random.
Formalizing the Model
The model is based on one described in [Dawkins, 1986] and [Neander, 1995b] <ref:2602.20883#pg5> This model involves generating N new sequences by copying the current best state with probabilities related to mutation rates ν and µ <ref:2602. p. 4 [2009] for cultural
evolution and [Arthur and Nicholson, 2022] for Gaia Theory. In this case it
means at some moment we have a ‘state’ or template consisting of k matches
and L − k mismatches, for example,
MExxxxNKSxT ISLxxEAxxASxL
We generate N new sequences by copying the current best state. 3 Cumulative selection
We solve the random and cumulative versions of the model for N = 1, ν = 0
in Appendix A. The basic result is the time to discovery T, the expected
time it takes to land on a target sequence which has L characters different
from the starting one is
Trandom(L) = b
L
(1)
Tordered(L) = L
p
− (L − 1) (2)
Tunordered(L) ≃
lnL + γ
p
(3)
Where γ is a constant and p =
µ
b
. Arthur@exeter. uk
1
arXiv:2602.Where the paper says it
Where the paper says it
Where the paper says it
Conditions for Adaptation
The conditions for cumulative selection in a system are necessary and sufficient, as the model constructed in Section 2 performs cumulative selection under these assumptions <ref:2602.20883#pg2> In the context of Darwinian populations, these conditions are equivalent to Lewontin’s recipe <ref:2602. p. 2 In this paper I will explore cumulative selection in detail and provide
general conditions for when it can occur. These general conditions are analogous to Lewontin’s recipe [Lewontin, 1970], in that they give necessary and
sufficient conditions for cumulative selection to operate. Lewontin’s third condition refers to heredity, and explicitly references
reproduction and correlation between parents and offspring. For cumulative selection we need something analogous, which has been called ‘memory’ [Charbonneau, 2014, Papale, 2021, Arthur and Nicholson, 2022].Where the paper says it
Where the paper says it
Green — the paper says it, and the page is printed. Red — no passage carries it, so treat that sentence as unverified. The same check runs on your own text: check a summary against its source, $5.
Improvements for AI systems
-
A system can be designed to perform adaptation by cumulative selection by tracking states that persist longer than others, such as in a sequence search model where
the state is a list of features.
This allows for the discovery of complex adaptations ina reasonable time
by the accumulation of small changes. -
The AI system can implement a ratcheting search mechanism where it generates new sequences by copying previous ones and changing characters with probabilities defined by mutation rates to find a target sequence. This process is cumulative selection, which can discover complex adaptations in sub-exponential time if
the rate of variation does not exceed the error threshold.
-
The AI can optimize neural network parameters using stochastic gradient descent where the update equation s → s − η∇L induces random variation in parameters, performing cumulative selection on the state of the network. This allows for
fine-tuned
adaptations to be discovered by accumulating persistent parameter changes.
Abstract
Biological systems like long-lived clonal organisms, holobionts and clades challenge traditional evolutionary thinking since they adapt without populations or reproduction. This paper aims to provide an overarching theoretical framework which encompasses standard Darwinian evolution as well as other processes of adaptation. This framework is cumulative selection and I provide a general `recipe' for it to occur. Lewontin's recipe for evolution by natural selection is shown to be a particular example of cumulative selection, but not the only one, with reproduction and populations just one way to perform cumulative selection. I discuss several other examples including clonal organisms, dioecious populations, holobionts, ecosystems, Gaia and neural networks.
Related papers
- From life-history model to theories of ecological diversity: the impact of Madhav Gadgil's work
- Diversification under Global Competition: A Cutoff-Induced Localization Mechanism
- Extending indoor airborne infection risk modelling to 3D, with local surface deposition
- Drivers of periodicity in population dynamic models of long-lived, large mammals
- Mutant Fixation for a Stochastic Evolutionary Model in Fragmented Populations
- How deterministic trajectories and their fluctuations underlie allele-frequency statistics in the strong-selection regime