[Paper Review] Natural selection. VI. Partitioning the information in fitness and characters by path analysis

Frank, Steven A.|arXiv (Cornell University)|Jan 22, 2013

Philosophy and History of Science21 citations

TL;DR

This paper introduces a path analysis framework that partitions phenotypic and fitness components into causal predictors—such as genes, maternal effects, symbionts, and social interactions—using information theory to interpret selection equations. It redefines heritability and Fisher’s fundamental theorem, offering a unified, causally interpretable foundation for kin selection and multivariate evolution.

ABSTRACT

Three steps aid in the analysis of selection. First, describe phenotypes by their component causes. Components include genes, maternal effects, symbionts, and any other predictors of phenotype that are of interest. Second, describe fitness by its component causes, such as an individual's phenotype, its neighbors' phenotypes, resource availability, and so on. Third, put the predictors of phenotype and fitness into an exact equation for evolutionary change, providing a complete expression of selection and other evolutionary processes. The complete expression separates the distinct causal roles of the various hypothesized components of phenotypes and fitness. Traditionally, those components are given by the covariance, variance, and regression terms of evolutionary models. I show how to interpret those statistical expressions with respect to information theory. The resulting interpretation allows one to read the fundamental equations of selection and evolution as sentences that express how various causes lead to the accumulation of information by selection and the decay of information by other evolutionary processes. The interpretation in terms of information leads to a deeper understanding of selection and heritability, and a clearer sense of how to formulate causal hypotheses about evolutionary process. Kin selection appears as a particular type of causal analysis that partitions social effects into meaningful components.

Motivation & Objective

• To provide a causal decomposition of phenotypes and fitness into component causes such as genes, symbionts, and social effects.
• To reinterpret traditional selection equations (covariance, regression, variance) through the lens of information theory for deeper mechanistic insight.
• To resolve conceptual confusion in kin selection by framing it as a causal modeling problem rather than a set of special rules.
• To replace outdated, fragmented models of social evolution with a general, unified framework based on explicit causal hypotheses.

Proposed method

• Applies path analysis to decompose phenotypic and fitness variation into distinct causal components.
• Uses the Price equation as a foundational framework to partition total evolutionary change into selection and non-selection components.
• Interprets covariance, regression, and variance terms in selection equations as information-theoretic measures of causal influence.
• Integrates information theory (e.g., mutual information, entropy changes) to interpret selection as information accumulation.
• Applies the framework to multiple characters and hierarchical systems, including social and kin interactions.
• Replaces heuristic models of kin selection with a formal causal structure that partitions social effects meaningfully.

Experimental results

Research questions

• RQ1 How can we formally partition the causes of phenotypic variation beyond standard genetic models?
• RQ2 How do selection, heritability, and fitness components relate when expressed in information-theoretic terms?
• RQ3 Can path analysis provide a causal interpretation of selection that clarifies kin selection and social evolution?
• RQ4 How does the decomposition of fitness into individual and social components improve the understanding of evolutionary processes?
• RQ5 What is the role of information theory in unifying the interpretation of fundamental evolutionary equations?

Key findings

• The paper shows that selection can be interpreted as the accumulation of information in populations, with the selection term in the Price equation corresponding exactly to information gain via fitness differentials.
• Phenotypic and fitness components are formally partitioned into causal predictors (e.g., genes, maternal effects, social interactions), enabling precise modeling of heritability beyond narrow genetic definitions.
• The framework reinterprets Fisher’s fundamental theorem as a statement about information transfer and decay in evolutionary systems.
• Kin selection emerges not as a special mechanism but as a specific application of causal path analysis, where social effects are partitioned into direct and indirect fitness components.
• The method allows for a general, information-theoretic interpretation of multivariate selection, replacing ad hoc models with a unified, causally coherent framework.
• By treating causal hypotheses as central, the approach reveals that all evolutionary models depend on prior assumptions about causality—making explicit modeling of causes essential for accurate inference.