Complexity in language acquisition.

Area of Science:

• Cognitive Science
• Linguistics
• Computational Learning Theory

Background:

• Traditional language acquisition research often focuses on information-theoretic problems, such as the lack of direct negative evidence.
• These approaches frequently neglect the probabilistic nature of human cognition and learning processes.
• Existing arguments conflate hypothesis classes with learnable concept classes, limiting conclusions about the learner.

Purpose of the Study:

• To re-evaluate the primary challenges in language acquisition by shifting focus from information theory to computational complexity.
• To propose that computational complexity offers a more direct and insightful approach to studying the language acquisition device.
• To explore the implications of complexity-based learnability for the nature of target grammars.

Main Methods:

• Analysis of existing arguments in language acquisition theory.
• Theoretical examination of the distinction between hypothesis classes and learnable concept classes.
• Investigation of computational complexity in constructing linguistic hypotheses from input data.

Main Results:

• Arguments based on information theory and negative evidence are flawed due to conflating hypothesis and concept classes.
• Computational complexity, not the absence of negative evidence, is identified as the central challenge in language learning.
• Learnability results informed by complexity studies suggest grammars must be objective.

Conclusions:

• Language acquisition is better understood by studying computational complexity rather than abstract learnable classes.
• The language acquisition device is better studied directly through complexity analysis.
• Objective grammars, based on language-internal properties, are favored by complexity-informed learnability, supporting data-driven learning.