Typing in tandem: Language planning in multisentence text production is fundamentally parallel

Area of Science:

• Cognitive Psychology
• Computational Linguistics
• Psycholinguistics

Background:

• Classical serial models propose text production involves discrete planning pauses followed by output bursts.
• Parallel models suggest planning can occur concurrently with the execution of previous text segments.

Purpose of the Study:

• To computationally model and compare serial versus parallel processing accounts of multisentence text composition.
• To analyze keystroke data to determine which model better explains writing behavior in children and adults.

Main Methods:

• Bayesian mixed-effects models were applied to six datasets of child and adult keystroke data.
• Interkey intervals were modeled using single (serial) and mixture (parallel) distributions.
• Analysis focused on intervals before sentences, before words, and within words.

Main Results:

• Model comparisons provided strong evidence favoring the parallel processing view across all datasets.
• While sentence-initial pauses were longer, many interkey intervals were too short for extensive planning.
• Findings contradict serial processing explanations and support parallel text composition.

Conclusions:

• Multisentence text composition is better explained by parallel processing models.
• Writing involves concurrent planning and execution, challenging the notion of distinct, serial planning phases.