Abstract: Electroencephalography (EEG) has long been used to index brain states, from early studies describing activity during visual stimulation to modern work employing complex perceptual tasks. These studies shed light on brain-wide signals but lacked explanatory power at the single neuron level. Similarly, single neuron studies can suffer from inability to measure brain-wide signals. Here, we combined these techniques while monkeys performed a change detection task and discovered a link between EEG and a signal embedded in spiking responses. This ‘slow drift’ was associated with arousal: decreases in pre-stimulus α power/increases in P1 amplitude were accompanied by: 1) increases in pupil size, false alarm rate and saccade velocity; and 2) decreases in microsaccade rate and reaction time. These results show that brain-wide EEG signals can be used to index modes of activity acquired from direct neural recordings, that in turn reflect global changes in brain state that influence perception and behavior.
Introduction
Neuroscience often operates at two extremes. Electroencephalography (EEG) provides a "bird's-eye view" of brain-wide activity but lacks the resolution to see what individual neurons are doing. Conversely, single-neuron recordings provide exquisite detail but are invasive and can miss global, brain-wide shifts in state. A major challenge has been linking these two scales, particularly regarding "slow drift"—gradual, co-fluctuating changes in the firing rates of large populations of neurons. While slow drift is known to affect behavior, such as impulsivity or attention, it was previously unknown if this internal drift could be detected using non-invasive tools like EEG.
Methods
Researchers trained two rhesus macaque monkeys (Macaca mulatta) to perform an orientation-change detection task, identifying when a visual pattern tilted. To bridge the gap between scales, they employed simultaneous recording techniques:
- Invasive Recording: 100-channel "Utah" microelectrode arrays were implanted in the visual cortex (area V4) to monitor populations of neurons.
- Non-Invasive Recording: Simultaneously, EEG was recorded from 8 electrodes on the scalp.
To define "slow drift," the team applied Principal Component Analysis (PCA) to the residual spike counts of the V4 neurons, identifying the first principal component as the primary axis of drift. They focused on two specific EEG metrics: pre-stimulus alpha power (~8–12 Hz) and the P1 amplitude (the first major peak in the visually evoked potential). Analysis was conducted using a 30-minute sliding window to isolate slow, state-like changes over several hours.
Key Results and Insights
The study revealed a significant antagonistic relationship between alpha power and P1 amplitude. When alpha power is low, P1 amplitude tends to be high, suggesting that low alpha represents a release from functional inhibition, allowing the brain to respond more strongly to visual inputs.
Crucially, these EEG signals served as a proxy for the population drift in V4. Slow drift was negatively correlated with alpha power (median $r = -0.38$) and positively correlated with P1 amplitude (median $r = 0.31$). These fluctuations were tied to a global signature of arousal. When the EEG and neuronal drift indicated high arousal, the monkeys exhibited:
- Increased pupil size and faster saccade velocity.
- Faster reaction times.
- An increase in "false alarms," suggesting that arousal shifts the decision criterion (making the subject more "trigger-happy") rather than simply improving visual sensitivity.
These effects were specific to the alpha band; beta and gamma frequencies did not show the same consistent relationship with slow drift, ruling out broadband noise as a factor.
Discussion and Future Directions
While the link between EEG and internal neuronal drift is strong, the study is correlational. It remains unclear if the slow drift causes the EEG changes or if both are driven by subcortical arousal centers like the locus coeruleus. Future work could investigate if specific electrode locations provide better tuning to the slow drift of specific brain regions or if this link holds during more complex cognitive tasks.
Confirming this "global signature" in humans would be a major step for clinical applications, such as non-invasively monitoring fatigue, attention, or decision-making states in real-world settings.
Summary updated