📄 evolutionary biology

Sexual Selection as a Mechanism of Evolutionary Information Preservation

This paper proposes and validates through agent-based simulation that sexual selection functions as a mechanism for preserving evolutionary information by sustaining historically adaptive traits via Fisherian co-evolution, thereby enabling lineages to respond more efficiently to fluctuating environments compared to random mating.

Original authors: Erden, Z. D.

Published 2026-04-17

📖 6 min read🧠 Deep dive

CC BY 4.0

Original authors: Erden, Z. D.

Original paper licensed under CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). ⚕️ This is an AI-generated explanation of a preprint that has not been peer-reviewed. It is not medical advice. Do not make health decisions based on this content. Read full disclaimer

The Big Idea: Evolution's "Time Capsule"

Imagine evolution as a massive, chaotic construction site. Usually, when the weather changes (the environment shifts), the workers (organisms) tear down old buildings and build new ones to fit the new climate. If the old weather comes back, they have to start from scratch, which takes a long time.

This paper proposes a fascinating new theory: Sexual selection acts like a "Time Capsule" or a "Library" for evolution.

The author, Zeki Doruk Erden, suggests that when females (or the choosing sex) are picky about who they mate with, they aren't just looking for the "best" partner for today. They are accidentally building a system that preserves blueprints for traits that were useful in the past, even if those traits aren't needed right now. This means that if the old environment returns, the population can "replay" those old blueprints instantly, rather than having to invent them all over again.

The Problem: Why Do We Have "Useless" Costly Traits?

Think of a peacock's tail. It's huge, heavy, and makes the bird an easy target for predators. It's a terrible survival tool. Yet, peacocks have them.

Standard biology says: "Females like the tail, so males with big tails get to mate, and the tail gets passed on." This is called Sexual Selection.

But this paper asks a deeper question: Why does this happen so intensely in the most complex animals? The author argues that the cost of these traits (the heavy tails, the loud songs) isn't just a side effect. It's a feature. It's the price of admission for a system that remembers the past.

The Mechanism: The "Lock and Key" of Memory

Imagine a lock and a key.

The Key (The Trait): A male has a specific feature (like a bright color) because it helped him survive a specific danger (like a certain type of predator).
The Lock (The Preference): Females evolve a "taste" for that bright color.

Once the females start preferring that color, something magical happens. Even if the predator disappears and the bright color is no longer needed for survival, the females still want it. The males keep growing the color because the females keep choosing it.

The Analogy:
Imagine a town where everyone used to wear heavy coats because it was freezing. Then, the weather gets warm.

Without Sexual Selection: Everyone stops wearing coats immediately because they are hot and uncomfortable. If winter comes back, everyone is freezing and has to buy new coats from scratch.
With Sexual Selection: The town has a "Fashion Council" (the females) that decides, "We love coats! They look great!" Even though it's warm, the town keeps wearing coats because the Council demands it.
The Result: When winter suddenly returns, the town is already dressed for it. They don't have to panic and invent coats; they just keep wearing what they already have.

The Experiment: A Digital Simulation

Since we can't wait thousands of years to see if this happens in real animals, the author built a video game simulation (an agent-based model).

The Setup: He created a population of digital creatures.
Phase 1 (The Storm): He made the environment harsh. Only creatures with a specific "superpower" (a specific trait) survived.
Phase 2 (The Calm): He turned off the storm. The superpower was no longer needed. In fact, it was slightly annoying (costly).
The Test: He compared two groups:
1. Random Mating: Creatures just picked partners randomly.
2. Picky Mating: Females were choosy and only mated with males who had the superpower.

The Results:

Random Group: As soon as the storm stopped, the superpower disappeared. The population "forgot" how to do it.
Picky Group: Even though the superpower wasn't needed, the females kept choosing males with it. The superpower stayed alive. In fact, in some cases, it got stronger because the females loved it so much.

The Twist: The "Library" is Stronger than the "Book"

The most surprising finding was about where the memory is stored.

The Trait (The Book): The actual physical feature (the tail, the color). This is fragile. If the environment gets too harsh, the trait might die out.
The Preference (The Library Card): The female's desire for the trait. This is incredibly durable.

The simulation showed that the females' preferences acted as a stronger, more durable archive than the physical traits themselves. Even if the physical trait faded a little, the "desire" remained. So, if the environment changed back, the females would immediately start pushing the males to bring the trait back.

It's like having a recipe (the preference) vs. the cake (the trait). If you lose the cake, you can bake a new one instantly if you still have the recipe. If you lose the recipe, you have to guess how to bake it again. Sexual selection keeps the recipe safe.

Why Does This Matter?

1. For Biology:
It explains why complex animals (like birds and mammals) are so good at surviving in changing worlds. They aren't just reacting to today's problems; they are carrying a "backup drive" of solutions from the past. This makes them more flexible and resilient.

2. For Artificial Intelligence (AI):
This is a huge deal for computer scientists. AI systems often suffer from "Catastrophic Forgetting." If you teach an AI to play Chess, and then teach it to play Go, it often forgets how to play Chess.
This paper suggests that if we build AI systems with a "sexual selection" mechanism (where the AI has "preferences" for certain solutions), it might be able to remember old skills while learning new ones. It could help robots or software adapt to changing worlds without losing their past knowledge.

The Bottom Line

Sexual selection isn't just about finding a pretty mate. It's an evolutionary backup system. By being picky, females create a self-sustaining loop that keeps useful traits alive in the population's "memory," ready to be used again if the world changes back to how it used to be. It turns evolution from a game of "start over" into a game of "remember and adapt."

1. Problem Statement

The paper addresses a fundamental gap in evolutionary theory regarding the persistence of complex traits. While established theories (Fisher-Muller, Red Queen) explain the benefits of sexual reproduction for recombination speed and parasite resistance, they do not fully account for why sexual selection (specifically asymmetric mate choice) is most intense in lineages with the highest morphological complexity and capability.

The author posits that existing explanations focus on immediate adaptive benefits but fail to explain the long-term structural function of sexual selection. The core problem is: How do sexually selecting lineages retain historically adaptive traits when environmental pressures fluctuate or disappear, preventing the loss of evolutionary information (a phenomenon analogous to "catastrophic forgetting" in machine learning)?

2. Methodology

The author employs an agent-based simulation to test the hypothesis that sexual selection acts as a mechanism for preserving evolutionary information.

Model Architecture:
- Population: $N$ males and $N$ females.
- Genetic Representation: Each individual carries two $n$ $n$ -dimensional real-valued vectors:
  - Trait Vector ( $T$ ): Expressed in males (phenotype), latent in females.
  - Preference Vector ( $P$ ): Expressed in females (mate choice criteria), latent in males.
- Inheritance: Mendelian per-component inheritance with additive Gaussian mutation ( $\sigma_T$ for traits, $\sigma_P$ for preferences).
Selection Mechanisms:
- Natural Selection: Males survive based on the alignment of their trait vector $T$ with an environmental vector $e$ . Fitness is maximized when $T$ aligns with $e$ , penalized by a cost coefficient $c$ for large trait magnitudes.
- Sexual Selection: Females select mates using a Boltzmann (softmax) distribution based on the dot product of their preference $P$ $P$ and the male's trait $T$ $T$ .
  - Choosiness Parameter ( $\tau$ ): Controls the strength of discrimination. $\tau=0$ implies random mating; $\tau>0$ implies preference for aligned traits.
Experimental Protocol:
- Two-Phase Environment:
  1. Selection Phase ( $d_s$ generations): Directional natural selection acts on a specific trait dimension $k$ .
  2. Neutral Phase ( $d_n$ generations): Natural selection is removed ( $e=0$ ), but trait costs ( $c$ ) remain. The goal is to measure how well the trait persists without external pressure.
- Metrics:
  - RAUC (Retention Area Under Curve): Average preservation of the trait over the neutral phase.
  - REP (End-Point Retention): The trait value at the end of the neutral phase relative to the peak at the end of the selection phase.
  - Preference Retention: Measured similarly for the female preference vector $P$ .
Parameters: The study sweeps across population sizes ( $N$ ), regime durations, selection strengths, trait dimensionality ( $n$ ), mutation rates, and choosiness levels ( $\tau$ ).

3. Key Contributions

Theoretical Framework: Proposes that sexual selection functions as an evolutionary memory system. Through Fisherian runaway co-evolution, the correlation between female preferences and male traits creates an internalized selection pressure that sustains traits even after the original environmental driver is removed.
Computational Evidence: Provides the first computational demonstration that sexual selection systematically preserves historically adaptive information across changing environmental regimes, outperforming random mating.
Dual-Layer Storage Mechanism: Identifies that adaptive information is stored redundantly in two layers:
- The expressed phenotype (Trait $T$ ), which is subject to natural selection costs.
- The latent preference (Preference $P$ ), which is not subject to direct survival costs and thus acts as a more durable "archive."
Cross-Domain Analogy: Draws a parallel between biological sexual selection and the problem of catastrophic forgetting in machine learning, suggesting that internal selective structures can prevent the erasure of past adaptations.

4. Results

Trait Preservation:
- In the absence of sexual selection ( $\tau=0$ ), traits decay monotonically during the neutral phase due to genetic drift and trait costs.
- With any degree of female choosiness ( $\tau > 0$ ), traits are significantly preserved. In many conditions, traits not only persist but are amplified (Retention > 1.0) due to ongoing Fisherian feedback loops.
- The advantage is robust across population sizes, selection strengths, and mutation rates.
Preference Durability:
- Female preferences ( $P$ ) are retained more strongly than the traits ( $T$ ) themselves.
- Because preferences do not incur the natural selection cost $c$ , they continue to grow or remain stable during the neutral phase, forming a robust "backup" for the trait.
Parameter Sensitivity:
- High Dimensionality ( $n \ge 8$ ): The preservation effect degrades as dimensionality increases due to "preference dilution" (noise in non-selected dimensions) and spurious runaway dynamics.
- Small Populations ( $N=50$ ): Genetic drift dominates, making the preservation effect detectable but unreliable.
- High Trait Costs: Strong natural selection against the trait during the neutral phase narrows the window for sexual selection to maintain the trait, but a preservation advantage remains.

5. Significance

Evolutionary Theory: The paper reframes sexual selection not just as a mechanism for immediate mate quality assessment, but as a structural mechanism for long-term evolvability. It explains why complex lineages (which exhibit intense sexual selection) are better equipped to respond to fluctuating environments: they retain a "record" of past adaptations that can be rapidly reactivated.
Artificial Life & AI: The findings offer a blueprint for designing artificial evolutionary systems and machine learning algorithms that resist catastrophic forgetting. By implementing internal "preference" structures that co-evolve with "traits," designers can create systems that retain historical capabilities while adapting to new objectives.
Future Directions: The paper suggests that empirical biologists should look for evidence of "environmental retention" in sexually selected traits that cannot be explained by current natural selection. It also calls for the integration of similar retention mechanisms into open-ended evolutionary algorithms to enhance long-term complexity accumulation.

In summary, Erden demonstrates that sexual selection creates a path-dependent evolutionary trajectory, where the co-evolution of traits and preferences acts as a buffer against environmental volatility, ensuring that valuable evolutionary information is not lost but preserved for future reactivation.