Moran Process¶

Model a finite constant-size population as a discrete birth–death Markov chain in which one type is selected to reproduce and one individual to be replaced, yielding stochastic extinction or fixation under drift and selection.

Version: v1 · 2026-08-30 · History
Domain-specific #: 2309
Origin domain: biology
Subdomain: population genetics
Aliases: Moran model, Classical Moran process

Core Idea¶

The Moran process is a discrete-time stochastic model for a finite population of fixed size \(N\). In the classical two-type version, one individual is selected for reproduction and one for death or replacement at each step, so the count \(X_t\) of type \(A\) changes by at most one. Moran introduced the overlapping-generation random process as an alternative to models that replace an entire generation at once.^[1]

Under neutral sampling, when \(X_t=i\), the upward and downward transition probabilities are both \(i(N-i)/N^2\), with the remaining probability assigned to no change. Without mutation, states \(0\) and \(N\) are absorbing: type \(A\) ultimately becomes extinct or fixed. Its neutral fixation probability from \(i\) copies is \(i/N\).

With constant relative fitness \(r\) for type \(A\), birth chosen proportional to fitness, and uniform death, \(T_i^+=[ri/(ri+N-i)][(N-i)/N]\) and \(T_i^-=[(N-i)/(ri+N-i)][i/N]\). For \(r\ne1\), the fixation probability is \(\rho_i=(1-r^{-i})/(1-r^{-N})\); the neutral limit is \(i/N\). Frequency dependence, mutation, spatial structure, and alternative birth/death order define variants rather than consequences that may be imported silently.^[2]

Structural Signature¶

The finite population. Exactly \(N\) individual slots are represented throughout the classical process.
The heritable types. Alleles, genotypes, strategies, or abstract replicator types occupy the slots.
The count state. \(X_t\in\{0,1,\ldots,N\}\) records the abundance of a focal type in the well-mixed two-type model.
The reproduction draw. One individual is sampled, neutrally or with a stated fitness weighting, to contribute an offspring copy.
The death/replacement draw. One slot is sampled for removal or replacement so population size remains fixed.
The one-step locality. The focal count changes only by \(-1\), \(0\), or \(+1\) per update.
The Markov kernel. Transition probabilities depend on the current state under the declared fitness and update convention.
The boundary behavior. Without mutation, extinction and fixation are absorbing; mutation can restore movement from the boundaries.
The fixation quantities. Absorption probability and time summarize stochastic evolutionary outcomes.
The variant controls. Selection, mutation, frequency dependence, graph structure, and update order are stated explicitly.

What It Is Not¶

Not the Wright–Fisher model. Moran updates one birth–death event at a time rather than sampling a whole new generation synchronously.
Not a deterministic replicator equation. Finite-population sample paths fluctuate and can absorb despite identical expected drift.
Not genetic drift alone. Neutral drift is one regime; selection, mutation, and frequency dependence can be incorporated.
Not automatically a spatial graph process. The classical model is well mixed; evolutionary graph variants require location and replacement structure.
Not absorbing when recurrent mutation is present. Boundary states cease to be absorbing under two-way mutation.
Not a biological laboratory procedure. It is a descriptive mathematical model of hypothetical population-state transitions.

Scope of Application¶

The process is literal in theoretical population genetics and evolutionary dynamics when finite constant-size populations evolve through asynchronous individual replacement.

Neutral genetic drift. Calculating extinction, fixation, and absorption times in a finite population.
Selection. Comparing fixation probabilities under relative-fitness differences.
Evolutionary games. Allowing fitness to depend on current type frequencies.
Mutation-selection models. Adding transitions that prevent permanent absorption.
Evolutionary graph theory. Generalizing reproduction and replacement across a population structure.
Cancer and microbial evolution models. Describing abstract clonal competition under carefully declared population assumptions, without prescribing experiments.

Clarity¶

Specify population size, ploidy abstraction, types, state variable, time convention, reproduction weighting, death or replacement sampling, whether the same individual can be selected twice, mutation, update order, and spatial structure. Write the transition kernel before quoting fixation results. Distinguish unconditional absorption time from conditional fixation time and state whether formulas count events, generations, or rescaled time.

Manages Complexity¶

A population configuration collapses to a finite-state birth–death chain whose absorption probabilities and times can often be computed exactly. The model isolates drift and selection without tracking genealogies or all individual identities, and the tri-diagonal kernel turns many questions into recurrences rather than an enumeration of full population configurations. It also gives a controlled baseline against which spatial, frequency-dependent, or mutation-bearing variants can be compared one assumption at a time. That economy depends on fixed size, exchangeability, update convention, and type sufficiency; age structure, recombination, variable demography, environment, and spatial heterogeneity require explicit extensions. Even when two variants share the same fixation probability, they may have different absorption-time laws or sample-path behavior, so matching one summary is not model equivalence.

The process reduces a finite population configuration to a Markov state, often the count of one type under well mixing, and updates that state through one reproduction and one replacement event. This yields exact finite transition probabilities and makes extinction and fixation absorbing boundaries explicit. The compression is valid only when identities, locations, ages, or other hidden attributes do not affect future transitions beyond the recorded state. Selection changes reproduction weights, mutation can reopen boundaries, and graph structure can require a richer configuration state. Diagnostics therefore declare update order, fitness mapping, replacement sampling, time convention, and whether results are exact or derived from a diffusion limit. The model organizes stochastic evolutionary change without asserting that every real population is well mixed, constant in size, or governed by frequency-independent fitness.^[1]^[2]

Abstract Reasoning¶

Choose the finite population size and focal type state.
Declare the reproduction weighting and replacement rule.
Derive upward, downward, and self-loop probabilities from the current count.
Verify that each row of the transition kernel sums to one.
Identify absorbing or mutation-connected boundary states.
Solve the harmonic recurrence for fixation probability.
Compute or simulate absorption-time distributions under the same time convention.
Compare selection, mutation, structure, and update-order variants without conflating them.

Knowledge Transfer¶

The strict parent is Stochastic Process: \(X_t\) is an indexed family of random population states governed by one joint law, with a Markov dependence structure and sample-path/ensemble distinction. Evolutionary Algorithm and Genetic Drift are related applications or regimes, not the taxonomic identity.

Stochastic Process is the strict parent because the Moran trajectory is a time-indexed random state governed by a transition kernel. The transferable skeleton is finite state + random update law + repeated transitions -> distribution over paths and boundary outcomes. The population-genetic residue is constant size, reproduction coupled to replacement, heritable type, drift, selection, mutation, and fixation interpretation. A generic birth-death process can change total population size, while Wright-Fisher updating replaces a generation synchronously. Deterministic replicator dynamics can approximate frequency change but suppress finite-population path variability and absorption timing. Transfer is literal only when the declared random process preserves the Moran event structure or is explicitly identified as an approximation rather than the same model.

Examples¶

Canonical¶

For \(N=10\) and \(i=4\) under neutrality, \(T_4^+=T_4^-=4\cdot6/100=0.24\), and \(T_4^0=0.52\). The expected one-step count remains four, yet a realized path wanders until extinction or fixation. Its probability of eventual fixation is \(4/10\), demonstrating that zero expected drift does not remove stochastic absorption.^[2]

Mapped back: finite neutral population → birth and replacement draws → tri-diagonal Markov step → random walk in type count → extinction or fixation.

Applied / In Practice¶

A theoretical study compares a neutral type with one assigned relative fitness \(r\). It computes the one-mutant fixation probability from the exact transition kernel and then repeats the analysis after adding recurrent mutation. The mutation model is not reported as having absorbing fixation, because the boundary can be left; the two analyses answer different questions.

Mapped back: declared fitness and mutation regime → transition kernel → boundary classification → fixation or stationary analysis.

Structural Tensions¶

Constant size vs. demographic realism. Fixed \(N\) makes the chain tractable but suppresses population growth and bottlenecks. Diagnostic: Is replacement conservation defensible for the question?
Well mixing vs. population structure. Uniform draws erase location and network effects. Diagnostic: Does who can replace whom matter?
Exact finite dynamics vs. diffusion approximation. Continuous limits simplify analysis but alter time and boundary detail. Diagnostic: Are results exact for \(N\) or asymptotic?
Absorption vs. mutation-maintained variation. No-mutation chains fix, while recurrent mutation can yield an ergodic stationary distribution. Diagnostic: Can a lost type re-enter?
Autonomous process vs. generic stochastic process. Random indexed states travel; single birth–death replacement at fixed \(N\) defines Moran. Diagnostic: Does every update preserve population size and change the count by at most one?

Structural–Framed Character¶

The Moran process is structural-leaning. Given its update rule, state space and probability law are mathematical. Type choice, fitness mapping, update order, and whether fixed size is appropriate are modeler-framed. It is evaluatively neutral and descriptive. Its population-genetic birth–death semantics keep it domain-specific.

A model diagnostic reconstructs the one-step transition probabilities and verifies that they sum to one for every state, with correct behavior at extinction and fixation boundaries. Simulation summaries should agree with exact small-population calculations before asymptotic approximations are trusted. This validates the stochastic model while remaining entirely descriptive and nonprocedural about biological systems.

Structural Core vs. Domain Accent¶

The skeleton is finite state + random local update → Markov trajectory → boundary or stationary outcome. The accent is fixed population size, reproduction-death replacement, heritable types, selection, mutation, extinction, and fixation. Removing them yields a generic birth–death stochastic process.

Stochastic Process is the strict parent because the type-count trajectory is a time-indexed family of random variables under a single Markov law. The Moran residual is the constant-size individual birth–death update and its population-genetic boundary interpretation.

The prospective workspace queue contains one strict upward edge to prime:stochastic_process. No live DAG mutation is authorized.

Relationships to Other Abstractions¶

Current abstraction Moran Process Domain-specific

Parents (1) — more general patterns this builds on

Moran Process is a kind of Markov Process Prime

The accepted reference-grade review places Moran Process under Markov Process because the child instantiates or depends on the parent's broader structure while retaining its own constitutive identity.

Hierarchy paths (4) — routes to 4 parentless roots

Moran Process → Markov Process → Stochastic Process

Show alternative paths (3)

Neighborhood in Abstraction Space¶

Moran Process sits in a sparse region of the domain-specific corpus (87^th percentile for distinctiveness): few abstractions share its structure, so a faithful description tends to retrieve it precisely.

Family — Population Genetics & Selection (15 abstractions)

Nearest neighbors

Computed from structural-signature embeddings · 2026-09-08

Not to Be Confused With¶

Wright–Fisher model. Finite-population model with synchronous generation replacement.
Birth–death process. Broader class whose total size may change and whose rates need not encode reproduction plus replacement.
Genetic drift. Random allele-frequency change, one phenomenon represented by the neutral Moran process.
Replicator dynamics. Deterministic infinite-population frequency dynamics.
Evolutionary graph process. Spatial generalization with replacement constrained by graph structure.
Coalescent process. Backward-time genealogy model rather than forward count dynamics.

References¶

[1] P. A. P. Moran, ‘Random Processes in Genetics,’ Proceedings of the Cambridge Philosophical Society 54, no. 1 (1958): 60–71, https://doi.org/10.1017/S0305004100033193. registry ↩a ↩b

[2] Warren J. Ewens, Mathematical Population Genetics 1: Theoretical Introduction, 2^nd ed. (Springer, 2004), https://doi.org/10.1007/978-0-387-21822-9. registry ↩a ↩b ↩c