Dream reports collected after rapid eye movement (REM) sleep are typically longer, more vivid, and more story-like than those from non-REM sleep, but traditional measures may be confounded by report length. By analyzing 133 dream reports from 20 participants as non-semantic directed word graphs, researchers found that REM dream reports have greater connectedness—words recur with longer range—compared to N2 sleep reports. Graph measures predicted dream complexity, with higher connectedness and lower randomness linked to more complex reports. The largest connected component improved models using report length alone for predicting sleep stage and complexity. Graph analysis offers an automated method to complement traditional dream report analysis.
During non-rapid eye movement (NREM) sleep, the brain's large-scale functional networks show a surprising pattern: nearly all networks are most active during NREM stage 2, then abruptly lose activity in NREM stage 3. However, despite this high activity in stage 2, the functional connections and mutual dependencies between networks progressively break down as sleep deepens. This means that even though networks attempt to communicate during stage 2, the efficiency of information transfer is low. The findings advance neural models of sleep and consciousness by showing that network integrity, not just activity levels, is crucial for conscious awareness.