Editorial notes:
This is the first in a series of 30 Science News, documented mostly in Evernote between April 2013 and October 2014.
Concepts: Neural decoding of dreams. Also: hypnagogic hallucinations.
Source: A News and Analysis and an (original research) Report, both in Science (April 5, 2013).
- History: A 2005 study developed an algorithm that could read fMRI recordings to determine which one of 8 grid perspectives a subject was looking at.
- Experiment setup: Calibrated algorithm with report-coupled-fMRI recordings during stage 1 sleep. Subjects were woken up after sleep onset (stage 1, during which people experience hypnagogic hallucinations, cf. dreaming proper which occurs in REM) and asked to report “dreams”. Collected about 200 reports from each of three subjects, and grouped frequently repeated visual elements into 20 broad categories tailored to each subject (eg. ice-pick, key, and plunger, are all implements).
- Result: Algorithm could predict “dream” content with 60% accuracy.
- An implication: Dreams (or at least hypnagogic hallucinations) are not made up upon awakening.
