Context and the activation of predictive inferences

Area of Science:

• Cognitive Psychology
• Natural Language Processing
• Computational Linguistics

Background:

• Reader inference is crucial for text comprehension.
• Both immediate context and earlier text influence predictive inference activation.

Purpose of the Study:

• To investigate how earlier text information and immediate context interact to influence predictive inference activation.
• To determine if earlier character trait information can override inferences suggested by the immediate context.

Main Methods:

• Two experiments were conducted using narrative passages.
• Experiment 1 manipulated character trait descriptions presented earlier in the text.
• Experiment 2 examined inference activation when immediate context conflicted with earlier trait information.

Main Results:

• The same immediate context activated different inferences based on prior character trait information.
• Earlier character trait descriptions activated specific inferences, even overriding the immediate context's suggested inference.
• Inference activation is a result of combined influences from immediate context and earlier text.

Conclusions:

• Reader inference is a dynamic process influenced by the interplay of textual information across different parts of a passage.
• Character trait information presented early in a text significantly impacts the activation of predictive inferences.
• Understanding these combined influences is key to modeling text comprehension and predictive processing.