An Agent-First Preference in a Patient-First Language During Sentence Comprehension

Area of Science:

• Psycholinguistics
• Neuroscience
• Computational Linguistics

Background:

• The incremental processing of language often favors an agent-first interpretation.
• This bias might stem from a fixed cognitive property or be learned from language usage frequency.
• Existing evidence is primarily from agent-initial languages, limiting understanding of cross-linguistic biases.

Purpose of the Study:

• To investigate the agent-first bias in language comprehension using electroencephalography (EEG).
• To examine how word order and referent type (human vs. nonhuman) influence interpretation in a patient-initial language.
• To differentiate between a fixed cognitive bias and a usage-frequency-based bias in sentence processing.

Main Methods:

• Utilized electroencephalography (EEG) to record brain activity during sentence processing.
• Investigated comprehension in Äiwoo, a language that defaults to a patient-initial word order.
• Presented sentence-initial nouns that were ambiguous between agent and patient roles, varying referent type (human/nonhuman).

Main Results:

• Nonhuman nouns were initially interpreted as patients, showing negativity when disambiguated to an agent role.
• Human nouns were interpreted as agents, even against frequency, eliciting an N400-like negativity when disambiguated to a patient role.
• The agent bias for human referents proved robust against usage frequency, unlike for nonhuman referents.

Conclusions:

• The agent bias appears to be a fixed cognitive property for human referents, resistant to frequency effects.
• For nonhuman referents, the agent bias is malleable and can be shaped by language usage frequency.
• Findings suggest a complex interplay between innate biases and learned linguistic patterns in sentence comprehension.