How Culture and Biology Interact to Shape Language and the Language Faculty

Kenny Smith

TL;DR

This paper explores how human language, a uniquely complex communication system, evolved through a dynamic interplay between biological evolution and cultural transmission. It argues that cultural evolution of language can, in turn, shape the biological evolution of our cognitive capacities for language, creating a feedback loop that has shaped what language is and how we learn it.

What Problem Does This Paper Address?

The central question this paper tackles is how human language, with its open-ended expressivity and combinatorial structure, came to be. While we know language is transmitted culturally through learning and use, and that humans possess biological predispositions for language, the precise mechanisms by which these two forces interact to shape both language itself and the cognitive abilities that underpin it remain a significant challenge. The paper aims to synthesize recent research that models this gene-culture co-evolutionary process, moving beyond simply stating that both biology and culture play a role to exploring how they influence each other over evolutionary timescales.

Methods

The paper reviews a body of research that primarily employs computational modeling and experimental "iterated learning" paradigms.

  • Computational Models: These models simulate populations of learners and language users over many generations. They explore how linguistic structures emerge and stabilize based on factors like learning biases, communicative pressures, and genetic inheritance. Examples include models of vowel system evolution and the "moving target" model of language change.
  • Experimental Iterated Learning: This involves human participants learning and then transmitting a novel, simplified language to others, creating chains of transmission. Researchers then analyze how the language changes across generations, identifying emergent structural properties and the influence of learning biases. This method is used to study the emergence of compositionality and other linguistic features.

Datasets

Not discussed in the paper. The research relies on simulated data generated by computational models and experimental data collected from human participants in iterated learning tasks.

Key Insights and Ideas

  • Language as a Product of Cultural Evolution: Linguistic structures aren't static; they evolve through repeated cycles of learning and use, favoring variants that are easier to learn, produce, or are more useful for communication.
  • Gene-Culture Co-evolution: The evolution of language isn't just biology shaping culture, or vice-versa. It's a feedback loop: cultural evolution of language creates new selection pressures that can drive the biological evolution of language-related cognitive capacities.
  • The Role of Biases: Innate cognitive biases (e.g., a preference for simplicity) influence how languages are learned and transmitted, and these biases can be amplified or masked by cultural transmission.
  • Unmasking and Masking: Cultural transmission can "unmask" weak genetic biases, leading to rapid cultural convergence on a particular language type. However, as the language becomes universal, these biases can become "masked" from selection, as individuals can learn the language regardless of their specific genetic predisposition.
  • Compositionality as a Solution: The paper suggests that compositionality in language is a key innovation that balances competing pressures for simplicity in learning and expressivity in communication, providing the most efficient system.
  • Learning Shapes Learning: Unlike simpler models, the paper highlights that biases themselves can be shaped by linguistic experience, suggesting a more complex, hierarchical learning process.

Key Results

The paper synthesizes findings from various studies, rather than presenting a single set of quantitative results. Key qualitative findings include:

  • Emergence of Structure: Iterated learning experiments demonstrate that simple, arbitrary signals can evolve into structured, compositional systems over generations, even without explicit instruction.
  • Simplification and Degeneration: In the absence of communicative pressure, iterated learning can lead to the degeneration of linguistic systems, losing distinctions.
  • Compositionality Emerges with Communication: When communicative pressures (e.g., filtering ambiguity) are introduced, compositional structures tend to emerge, balancing learnability and expressivity.
  • Gene-Culture Feedback: Models show that cultural evolution can drive the selection of genes that support language learning and use, and vice-versa. For example, vowel systems evolving culturally can create selection pressures for biological adaptations in vocal tract and perceptual abilities.
  • Weak Biases Evolve Rapidly but Don't Become Strong Constraints: In certain scenarios (unmasking/masking), weak genetic biases can quickly lead to cultural universals, but these biases themselves don't necessarily become strong, fixed constraints due to cultural transmission.

Limitations and Future Work

  • Fixed Genetic Biases in Models: Many co-evolutionary models assume that genetic biases are fixed and unmodified by learning experience. The paper points out that for language learning, this is likely an oversimplification, as learning itself can shape biases.
  • Hierarchical Learning: The paper suggests that extending co-evolutionary models to account for hierarchical learning (where learning itself is learned) is a crucial next step.
  • Comparative Studies: More comparative studies are needed to understand whether humans have an unusual propensity for extracting higher-order generalizations compared to other animals.
  • Purely Cultural Shaping: The paper raises the possibility that purely cultural processes might shape higher-order learning and the capacities underpinning cultural evolution, which requires further investigation.
  • Moving Target Problem: While some recent work offers a more optimistic view, the challenge of how genes can evolve to track a rapidly changing linguistic environment remains an area of active research.

Summary updated

Abstract

Smith gives an excellent overview on research in language evolution, in which he discusses several recent models of how linguistic systems and the cognitive capacities involved in language learning m...
BibTeX
@inproceedings{smith2020how,
  author = {Kenny Smith},
  title = {{How Culture and Biology Interact to Shape Language and the Language Faculty}},
  booktitle = {Top. Cogn. Sci},
  year = {2020},
  url = {https://openreview.net/forum?id=ubDRgopuM0},
}
Figure