Reading intent 400 milliseconds before it moves

Reading intent before it moves

The headline result this week comes from human motor cortex, decoded in real time. A preprint reports that intracortical sensorimotor activity reliably predicts movement onset 250 to 400 milliseconds before the action happens — the brain's commitment to move, visible before the move. The state marking that commitment is task-independent: the decoder does not need retraining across different tasks. And in preempted trials, the decoded commitment aligns with people's own subjective reports of conscious intent 88 percent of the time. The philosophical argument about free will can wait; the engineering fact is that intent is now a readable, real-time signal. m:…9670

Reading intent needs clean signal, and the plumbing is improving in parallel. rtSOUND, a preprint method for real-time EEG noise suppression, uses spatial filtering to clean the data as it streams — at 5 kilohertz — and improves signal-to-noise in both TMS-evoked potentials and resting-state EEG. Real-time brain interfaces live or die on latency and noise; a spatial filter that keeps up with the stream at 5 kHz is infrastructure, not ornament. Together the two papers describe the closed loop getting tighter: quieter sensors, faster reads, intent decoded before the hand moves. m:…9325

Nudging sleep and perception

While the dish work trains wake-like activity, another preprint takes aim at sleep. Targeted memory reactivation — playing sound cues tied to learning while people sleep — reduces spindle and slow-oscillation density compared with regular sleep, and attenuates spindle-related cortical connectivity dynamics. That sounds like a cost, until you pair it with the correlational finding in the same study: higher spindle density and more slow-oscillation-spindle co-occurrence go with greater overnight motor memory gains. The same oscillations being modulated are the ones linked to the benefit, which makes the tradeoff the story. m:…9210

The effects are not only electrical. Learning-related diffusion MRI changes in the striatum and thalamus track the sleep oscillatory measures, which suggests the night's electrical choreography leaves structural traces in the circuits that learned. If you want a biomarker for whether a sleep-based memory intervention actually landed, the thalamus and striatum may be where to look. m:…9210

Over in the pharmacology corner, a psilocybin study adds a different kind of contingency: the drug decreases alpha-band power and increases gamma power at rest, but put participants in a structured sensory environment and the alpha-band connectivity changes seen with eyes closed get attenuated or even reversed. The neural signature of the drug is not just a property of the molecule — it is a property of the molecule in a room. Set and setting, rendered in hertz. m:…8390

The yardsticks keep moving

All of this steering demands better instruments, and the instruments are arriving. The Chronnectogram-ICA framework promises to track continuously evolving, overlapping functional brain states and their higher-order couplings — dynamics the authors say are invisible to discrete-state models. Applied to schizophrenia data, it finds a redistribution of connectivity regimes with both hyper- and hypo-volatility, depending on the measure. The field is finally building tools that assume the brain never sits still. m:…1786

And yet, in the same batch of preprints, a critique of resting-state fMRI argues that much of what the field has been calling brain dynamics may be thinner than it looks: a hidden Markov model with state-varying mean activity and variance accounts for most of the apparent dynamics, static functional connectivity plus HMM mean features predict behavior nearly as well as dynamic ones, and sliding-window analyses show little temporal structure beyond a static baseline in Human Connectome Project data. So neuroscience is simultaneously learning to drive the brain and questioning what it has been measuring. That tension — better steering, humbler yardsticks — is the actual state of the field this week. m:…9826