How Arousal Fluctuations Confound Resting State Neuroimaging

44 concepts35 papers1 proposal

Imagine asking a participant to lie perfectly still inside a dark, loud, yet strangely hypnotic MRI scanner for ten minutes and "think of nothing." What actually happens? They do not maintain a pristine, baseline cognitive state. Instead, they get bored, their minds wander, their eyes close, and crucially, they begin to drift into drowsiness or light sleep.

For decades, resting-state functional magnetic resonance imaging (rs-fMRI) and electroencephalography (rest-EEG) have treated this "resting state" as a stable baseline to uncover the brain's intrinsic wiring or identify biomarkers for disease. However, a growing body of research reveals a profound methodological crisis: transient fluctuations in a participant's physiological state—such as micro-sleeps, drops in vigilance, and circadian rhythms—exert a massive, unmodeled influence on brain connectivity, frequently confounding or entirely mimicking the neural signatures of interest.

The Mechanism: How Drowsiness Mimics Network Connectivity

Why does a simple drop in alertness distort neuroimaging data so severely? The answer lies in how the brain's global signal behaves during transitions in alertness.

As Xiao Liu et al. demonstrated in Subcortical evidence for a contribution of arousal to fMRI studies of brain activity, momentary drops in arousal trigger massive, brain-wide cortical signal spikes that co-occur with deactivations in deep-brain arousal centers. If a researcher assumes the subject is fully awake, these widespread, synchronized fluctuations are easily misidentified as stable, highly coordinated functional networks.

This instability directly undermines the study of time-varying functional connectivity (TVFC), which tracks how brain networks shift gears over seconds or minutes. As reviewed by Daniel J. Lurie et al. in Questions and controversies in the study of time-varying functional connectivity in resting fMRI, fluctuations in drowsiness and arousal are primary drivers of observed connectivity changes over time. Rather than reflecting dynamic cognitive processing or shifting thoughts, these time-varying signals are highly sensitive to non-neural artifacts and vigilance drops, which can produce false positives by mimicking structured connectivity changes.

Similarly, Marina Krylova et al. showed in Evidence for modulation of EEG microstate sequence by vigilance level that vigilance levels significantly modulate the sequence and duration of EEG microstates. This means that the assumption of a stable physiological baseline is routinely violated; what appears to be a meaningful shift in functional brain states may simply be a participant naturally transitioning from wakefulness toward sleep. This is further supported by the authors of Alpha EEG Activity and Pupil Diameter Coupling during Inactive Wakefulness in Humans, who noted that standard resting-state protocols frequently suffer from drowsiness or sleep onset due to their duration, interfering with the measurement of pure inactive wakefulness.

[Participant in Scanner] ──> [Vigilance Drops / Drowsiness]
                                     │
                                     ▼
                       [Massive Global Signal Spikes]
                                     │
                                     ▼
                [Misinterpreted as Stable Functional Networks]

The Illusion of Clinical Biomarkers

The most damaging consequence of this physiological confounding is its ability to masquerade as pathological or cognitive traits. When analyzing functional networks in clinical populations, it becomes incredibly difficult to separate a disease-specific signature from simple sleepiness.

For instance, Hossein Dini et al. noted in Dynamic Functional Connectivity Predicts Treatment Response to Electroconvulsive Therapy in Major Depressive Disorder that the absence of strict wakefulness monitoring introduces substantial noise into dynamic functional network connectivity, as drowsiness alters these patterns in ways that can be confused with disease-related states. This sentiment is echoed by Yiheng Tu et al. in Distinct thalamocortical network dynamics are associated with the pathophysiology of chronic low back pain, where fluctuations in wakefulness produce connectivity patterns that mimic pathological states, complicating the isolation of genuine neural signatures.

In large-scale studies, this issue scales up dramatically. Analyzing a population of over 10,000 children, Zening Fu et al. found in Cognitive and psychiatric relevance of dynamic functional connectivity states in a large (N > 10,000) children population that drowsiness or micro-sleeps introduce confounding connectivity patterns that mimic or obscure genuine biomarkers for mental health and cognition. Even in aging research, protocols that involve closing the eyes risk letting older adults drift into light sleep, which Kushal Borkar et al. warn in Ayu-Characterization of healthy aging from neuroimaging data with deep learning and rsfMRI can confound connectivity patterns and distort biomarkers meant to track healthy aging.

Environmental and Temporal Confounders

Beyond the immediate drop in alertness, a participant's broader biological rhythms act as hidden variables. The time of day a scan occurs can completely overshadow the stable behavioral traits a study aims to measure. Csaba Orbán et al. discovered in Correction: Time of day is associated with paradoxical reductions in global signal fluctuation and functional connectivity that scan timing exerts a significantly larger influence on resting-state functional connectivity than traits like fluid intelligence.

This temporal variance is tied directly to individual biology:

The Developmental Dilemma

The confounding effect of physiological state reaches a crisis point in developmental neurobiology. Infants cannot be instructed to lie still while awake, so they are almost exclusively scanned while asleep. Older children, however, are scanned while awake.

As Weiyan Yin et al. point out in Chart brain functional development from birth to 6 years of age, because functional connectivity is highly sensitive to the state of consciousness (awake vs. asleep), this methodological split makes longitudinal datasets across early childhood fundamentally incomparable.

Compounding this issue, infant scans themselves rarely monitor sleep stages. Wei Gao et al. in Temporal and Spatial Evolution of Brain Network Topology during the First Two Years of Life and Masoud Seraji et al. in Spatial development of brain networks during the first six postnatal months both emphasize that because brain connectivity varies drastically between active (REM) and quiet (non-REM) sleep, failing to record concurrent EEG makes it impossible to separate true developmental maturation from the transient effects of different sleep stages.

Study Population Scanning State Confounding Factor
Infants Natural Sleep REM vs. Non-REM stages alter connectivity patterns (Wei Gao et al.)
Older Children / Adults Awake (Resting) Micro-sleeps and drops in vigilance mimic networks (Zening Fu et al.)
Longitudinal Cohorts Mixed States Incomparable datasets due to awake vs. asleep baselines (Weiyan Yin et al.)

Mitigation and the Path Forward

To rescue resting-state neuroimaging from this sea of non-biological variance, the field is shifting toward treating physiological states with the same rigor historically applied to physical head motion.

One approach is to use external physiological markers to clean the data post-hoc. Jiahui Wang et al. demonstrated in Improving the Test-Retest Reliability of Resting State fMRI by Removing the Impact of Sleep that using heart rate variability (HRV) to identify and exclude "sleepy" data points significantly boosts the test-retest reliability of brain connectivity measures. However, the authors note that HRV can be co-confounded by emotional states, meaning it is not a perfect standalone proxy.

Ultimately, as argued by Christof Seiler et al. in Multivariate Heteroscedasticity Models for Functional Brain Connectivity, because falling asleep drastically alters brain synchronization patterns, sleep habits and average sleep duration must become standard control variables in neuroimaging studies. Only by explicitly modeling and accounting for these fluctuating physiological states can researchers ensure that their discovered biomarkers represent stable traits rather than transient drifts into slumber.

History & citationVersion 2 · 2026-10-04

Current · Permanent link

  1. Version 2 (latest) · 2026-10-04 · 33 papers
  2. Version 1 · 2026-10-04 · 22 papers
Cite this direction

Latest version (a living page):

@misc{lacuna_physiological_state_confounding_4b562092,
  title = {physiological state confounding},
  author = {{Lacuna}},
  howpublished = {Lacuna research direction},
  year = {2026},
  url = {https://lacuna.tiptreesystems.com/direction/physiological-state-confounding/rdr_4b56209267c44889bc08b89560250020},
  note = {Latest version of a living research direction; cite a version for a fixed text. Accessed 2026-11-19},
}

This exact version (version 2, frozen text):

@misc{lacuna_physiological_state_confounding_4b562092_v2,
  title = {physiological state confounding},
  author = {{Lacuna}},
  howpublished = {Lacuna research direction},
  year = {2026},
  url = {https://lacuna.tiptreesystems.com/direction/physiological-state-confounding/rdr_4b56209267c44889bc08b89560250020/v/2},
  note = {Version 2, frozen 2026-10-04. Accessed 2026-11-19},
}
About

These directions overlap with this one through shared papers.

See something wrong? Report a mistake