PIPS: A Parallel Planning Model of Sentence Production

Area of Science:

• Psycholinguistics
• Computational Linguistics
• Cognitive Science

Background:

• Subject-verb agreement errors are frequent in sentence production.
• Previous research often used preamble completion tasks to study verb errors.

Purpose of the Study:

• To re-examine data from preamble completion tasks to identify other error types.
• To investigate the mechanisms underlying both verb agreement and preamble repetition errors.
• To test the Parallelism in Producing Syntax (PIPS) model's ability to explain these errors.

Main Methods:

• Reanalysis of data from 50 experiments comprising 102,369 observations.
• Exploration of the Parallelism in Producing Syntax (PIPS) model, a continuous-state stochastic dynamical system.
• Simulations of the preamble completion task using the PIPS model.

Main Results:

• The preamble completion paradigm elicits significant errors in repeating the sentence preamble, specifically in local noun number.
• Errors arise from partially activated non-target syntactic parses due to memory constraints and competing grammatical information.
• The PIPS model successfully simulates these errors, accounting for notional agreement and structural number conflicts.

Conclusions:

• Sentence production is susceptible to errors beyond verb agreement, including errors in repeating sentence constituents.
• Finite memory and prior experience in language processing lead to interference from non-target production plans.
• The PIPS model provides a viable framework for understanding the cognitive mechanisms underlying sentence production errors.