A neuro-computational approximation of the qualities of mental images

A neuro-computational method reveals mental images are better represented by blurred, low-contrast, and psychedelic-style features, supporting theories of reduced sensory quality.

• Why it matters: Understanding mental imagery is difficult due to its subjective nature, limiting traditional introspective approaches and hindering objective assessment of mental image qualities.
• What they did: Researchers collected extensive EEG data while participants imagined scenes, used AI to generate candidate images, simulated visual cortex responses, and measured their alignment with EEG signals to infer image qualities.
• The result: Findings demonstrate that mental images align more closely with blurred, low-contrast, and distorted images, enabling objective, hypothesis-driven insights into mental imagery without relying on subjective reports.