Language learning and innateness: some implications of Compounds Research

Area of Science:

• Psycholinguistics
• Computational Linguistics
• Cognitive Science

Background:

• The formation of English noun compounds often disfavors regularly inflected plural modifiers (e.g., *rats-eater) compared to singular (mouse-eater) or irregular plural forms (mice-eater).
• This pattern has historically supported dual-mechanism theories of lexical representation, positing distinct innate mechanisms for regular and irregular word forms.

Purpose of the Study:

• To challenge the prevailing dual-mechanism theory of lexical representation in psycholinguistics.
• To propose and validate an alternative account for the acceptability of noun compound modifiers based on constraint satisfaction.

Main Methods:

• Analysis of large-scale English language corpora to identify patterns in noun compound formation.
• Behavioral experiments to assess human judgments of noun compound acceptability.
• Development and testing of computational models simulating linguistic acquisition and processing.

Main Results:

• Empirical and computational evidence indicates that the rule-versus-exceptions dichotomy makes inaccurate predictions regarding noun compound acceptability.
• The proposed constraint satisfaction model, integrating semantic, phonological, and other factors, provides a more accurate account of the observed data.
• Learned constraints, acquired through general-purpose learning algorithms from input data, effectively explain modifier acceptability.

Conclusions:

• The regular/irregular distinction in lexical representation is unnecessary and potentially misleading.
• A unified constraint satisfaction framework, sensitive to various linguistic factors and learned from input, better explains noun compound formation.
• This approach offers a more parsimonious and empirically supported explanation for psycholinguistic phenomena related to word formation.