The common patterns of nature

We typically observe large-scale outcomes that arise from the interactions of many hidden, small-scale processes. Examples include age of disease onset, rates of amino acid substitutions, and composition of ecological communities. The macroscopic patterns in each problem often vary around a characteristic shape that can be generated by neutral processes. A neutral generative model assumes that each microscopic process follows unbiased stochastic fluctuations: random connections of network nodes; amino acid substitutions with no effect on fitness; species that arise or disappear from communities randomly. These neutral generative models often match common patterns of nature. In this paper, I present the theoretical background by which we can understand why these neutral generative models are so successful. I show how the classic patterns such as Poisson and Gaussian arise. Each classic pattern was often discovered by a simple neutral generative model. The neutral patterns share a special characteristic: they describe the patterns of nature that follow from simple constraints on information. For example, any aggregation of processes that preserves information only about the mean and variance attracts to the Gaussian pattern; any aggregation that preserves information only about the mean attracts to the exponential pattern; any aggregation that preserves information only about the geometric mean attracts to the power law pattern. I present an informational framework of the common patterns of nature based on the method of maximum entropy. This framework shows that each neutral generative model is a special case that helps to discover a particular set of informational constraints; those informational constraints define a much wider domain of non-neutral generative processes that attract to the same neutral pattern.

💡 Research Summary

The paper tackles a long‑standing puzzle in many scientific domains: why do large‑scale phenomena—such as the age at disease onset, rates of amino‑acid substitution, and the composition of ecological communities—often conform to a handful of familiar probability distributions (Poisson, Gaussian, exponential, power‑law)? The author argues that the answer lies not in the specific mechanistic details of each system but in the informational constraints that govern the aggregation of countless microscopic processes.

A “neutral generative model” is defined as a stochastic system in which each underlying event follows an unbiased random rule (e.g., random network rewiring, fitness‑neutral amino‑acid changes, random species immigration/extinction). When many such events are combined, the resulting macroscopic pattern is called a “neutral pattern.” Classic neutral patterns correspond exactly to the four distributions mentioned above.

The theoretical backbone is the principle of maximum entropy (MaxEnt). Given a set of observable constraints—typically the mean, the variance, or the geometric mean—MaxEnt selects the probability distribution that maximizes uncertainty (entropy) while satisfying those constraints. The paper shows formally that: (1) fixing only the mean yields the exponential distribution; (2) fixing both mean and variance yields the Gaussian distribution; (3) fixing the mean and the logarithmic mean (geometric mean) yields a power‑law distribution; (4) fixing only the event rate (mean count) in a time‑or‑space window yields the Poisson distribution. In each case the neutral pattern emerges as the MaxEnt solution for a particular informational constraint.

Crucially, the author demonstrates that these informational constraints are far more general than the neutral mechanisms that originally inspired them. Any process—neutral or non‑neutral—that preserves the same limited statistics will be attracted to the same neutral pattern. Thus, selection, environmental heterogeneity, or other deterministic forces can be present, yet if the observable summary statistics remain unchanged, the system’s macroscopic distribution will still converge to the neutral form. This “attractor” property explains why neutral models often fit empirical data across disparate fields.

Empirical illustrations reinforce the theory. In epidemiology, the distribution of disease onset ages aligns with an exponential form when only the average onset age is known, and with a Gaussian form when both mean and variance are considered. In molecular evolution, neutral models of amino‑acid substitution predict exponential substitution times under a mean‑only constraint, while adding variance reproduces the observed Gaussian spread of substitution rates. In community ecology, preserving the mean species richness and the geometric mean of abundances leads to a power‑law species‑abundance distribution, matching field observations.

The paper’s central insight is that neutral generative models are special cases of a broader MaxEnt framework: they are tools for uncovering the underlying informational constraints of a system. By reframing model selection from “which microscopic mechanism is most realistic?” to “what summary statistics are reliably measured?”, researchers can construct models that are both parsimonious and widely applicable.

In the discussion, the author outlines future directions, including (i) systematic identification of additional informational constraints that give rise to other, less‑studied distributions; (ii) quantitative assessment of how non‑neutral forces modify the effective constraints; and (iii) integration of MaxEnt‑based priors into machine‑learning pipelines to improve generalization and avoid over‑fitting.

Overall, the paper provides a unifying, information‑theoretic perspective on why simple neutral models succeed so often, and it offers a powerful methodological blueprint for extending these ideas to more complex, realistic systems across biology, physics, and network science.