[Paper Review] Co-evolution of Language and of the Language Acquisition Device

Ted Briscoe|ArXiv.org|May 1, 1997

Language and cultural evolution11 references4 citations

TL;DR

This paper proposes a computational model of language acquisition using Generalized Categorial Grammar within a default inheritance hierarchy to formalize parameter setting in language learning. It demonstrates through evolutionary simulations that memory-limited learners with default parameter settings can emerge when exposed to a suitable linguistic environment, supporting a co-evolutionary account of language and the language acquisition device.

ABSTRACT

A new account of parameter setting during grammatical acquisition is presented in terms of Generalized Categorial Grammar embedded in a default inheritance hierarchy, providing a natural partial ordering on the setting of parameters. Experiments show that several experimentally effective learners can be defined in this framework. Evolutionary simulations suggest that a learner with default initial settings for parameters will emerge, provided that learning is memory limited and the environment of linguistic adaptation contains an appropriate language.

Motivation & Objective

• To develop a formal framework for parameter setting in language acquisition that reflects cognitive constraints and developmental patterns.
• To model how the language acquisition device (LAD) could co-evolve with human language through computational and evolutionary mechanisms.
• To investigate whether learners with default parameter settings can emerge under realistic learning constraints and environmental conditions.
• To provide a natural partial ordering on parameter settings using inheritance hierarchies in a computational grammar framework.
• To validate the model through evolutionary simulations showing emergence of default-based learners in appropriate linguistic environments.

Proposed method

• The framework uses Generalized Categorial Grammar (GCG) to represent syntactic structures and parameterized grammars.
• A default inheritance hierarchy is embedded in GCG to define a partial ordering on parameter settings, favoring default values in absence of evidence.
• Parameter settings are updated based on input data, with defaults applied when no specific evidence is available.
• The model simulates language acquisition under memory constraints, limiting the learner’s capacity to store and process linguistic input.
• Evolutionary simulations test whether learners with default settings can emerge when exposed to a language environment that supports such learners.
• The system evaluates learning success through grammatical accuracy and convergence on target grammars under constrained conditions.

Experimental results

Research questions

• RQ1 Can a language acquisition device with default parameter settings emerge through evolutionary processes?
• RQ2 Under what conditions does a memory-limited learner with default settings successfully acquire a target grammar?
• RQ3 How does the default inheritance hierarchy in GCG enable a natural partial ordering of parameter settings?
• RQ4 What role does the linguistic environment play in shaping the evolution of the language acquisition device?
• RQ5 Can computational models of parameter setting in GCG replicate empirically observed patterns in child language acquisition?

Key findings

• Evolutionary simulations show that learners with default parameter settings can successfully emerge when learning is memory-limited and the linguistic environment is appropriate.
• The default inheritance hierarchy in the GCG framework provides a natural and computationally effective partial ordering on parameter settings.
• Several experimentally effective learning algorithms can be formally defined within the proposed framework, supporting its plausibility.
• The model demonstrates that default settings are not arbitrary but can be evolutionarily stable under cognitive and environmental constraints.
• The framework supports a co-evolutionary account where language and the language acquisition device evolve in tandem, with the LAD adapting to the structure of available languages.
• The results suggest that default settings in language acquisition may be an evolutionary adaptation to limited memory and variable input.