In neuroimaging studies using functional or diffusion Magnetic Resonance Imaging (fMRI/dMRI), head motion has long been treated as the ultimate nuisance. When a participant moves inside the scanner, it blurs the images, distorts magnetic fields, and introduces artificial correlations that can easily be mistaken for real neural activity.
To combat this, the standard operating procedure has been simple: compress a participant's movement over time into a single aggregate scalar metric—such as average Framewise Displacement (FD)—and use it to either throw away high-motion data ("scrubbing") or statistically control for motion during preprocessing. This approach relies on a foundational assumption: that residual head motion is merely random, unstructured noise.
However, a growing body of research turns this assumption on its head. Instead of meaningless artifactual noise, an individual's movement profile within the scanner exhibits a highly consistent, reproducible spatiotemporal structure. Head motion, it turns out, is a behavioral trait in its own right.
Beyond Aggregate Metrics: Motion as a Trait
When we reduce an entire scanning session to a single average motion score, we lose the rich dynamics of how an individual moves. In Correlates of spatio-temporal motion characteristics during fMRI, Thomas William Arthur Bolton et al. investigated the detailed time courses of translation and rotation across individuals. They discovered that even after applying standard data-cleaning techniques, the remaining "clean" data contains highly structured, individual-specific patterns of movement.
To understand these patterns, the authors applied spectral clustering to group subjects based on their spatiotemporal motion profiles. Rather than separating people simply into "high movers" and "low movers," this approach revealed distinct "mover subgroups"—clusters of individuals who move in fundamentally similar ways over time and space.
Crucially, these subgroups were not random. They significantly correlated with a wide array of non-imaging variables, including:
- Anthropometric factors: Such as body mass index (BMI) and physical build.
- Behavioral traits: Including cognitive performance and compliance.
- Clinical factors: Reflecting underlying physiological or psychological variations.
As detailed in their parallel work, Agito ergo sum: Correlates of spatio-temporal motion characteristics during fMRI, Thomas William Arthur Bolton et al. argue that head motion acts as an endophenotype—a measurable, stable biological characteristic that bridges the gap between high-level behavior and underlying biology. For instance, a participant's respiration patterns, anxiety levels, or inability to inhibit minor physical adjustments manifest as a unique spatiotemporal signature in the motion data.
The Lifespan Trajectory of Movement
If head motion is an expression of individual biology and behavior, we should expect it to evolve systematically across the human lifespan. To map this trajectory, Kurt G. Schilling et al. conducted a comprehensive, large-scale study of head motion in diffusion MRI across thousands of subjects in Head Motion in Diffusion Magnetic Resonance Imaging: Quantification, Mitigation, and Structural Associations in Large, Cross‐Sectional Datasets Across the Lifespan.
The authors demonstrated that motion follows a highly predictable, non-linear U-shaped curve over a person's life:
- Infancy and Childhood: Motion levels are at their peak, driven by low behavioral compliance and developing motor control.
- Young Adulthood: Motion reaches its lowest, most stable baseline.
- Older Age (60+): Motion begins to rise significantly again, accompanied by a sharp increase in statistical "motion outliers."
Typical Lifespan Motion Trajectory
High | \ /
| \ / <- Outliers increase after 60
| \ /
Low | \___________________________/
+-----------------------------------------
Infants Young Adults Older Age
This predictable variation across age groups reinforces the idea that motion is deeply intertwined with the biological state of the participant, rather than being an unpredictable, exogenous accident.
The Paradox of Motion Correction
This shift in perspective—from viewing motion as random noise to viewing it as a structured behavioral correlate—introduces a profound paradox for neuroimaging analysis.
If standard preprocessing pipelines handle motion by blindly scrubbing out high-motion frames or regressing out motion parameters, they may be doing more than just cleaning the data. As Thomas William Arthur Bolton et al. point out, because spatiotemporal motion structures are systematically linked to the very behavioral, cognitive, or clinical traits under study, aggressive motion correction risks inadvertently discarding valuable neural information or introducing severe statistical biases. For example, if a clinical population naturally moves in a specific rhythmic pattern due to anxiety or respiration, erasing that pattern might erase the neural signatures of the condition itself.
Conversely, we must ensure that motion does not create artificial illusions of structural differences. Addressing this side of the coin, Kurt G. Schilling et al. categorized subjects into percentile-based groups of "movers" and "non-movers." By comparing these groups, the authors demonstrated that modern, advanced preprocessing pipelines are remarkably effective at decoupling artifactual motion biases from true underlying anatomy. When these pipelines are applied correctly, individuals with high motion do not show significant artificial differences in brain microstructure or connectomics compared to low-motion individuals.
Why This Matters
Understanding that head motion is structured and behaviorally relevant changes how we design and interpret neuroimaging experiments:
- From Exclusion to Modeling: Instead of treating high-motion subjects as failed experiments to be thrown away, researchers can begin to treat spatiotemporal motion profiles as a feature. The way a patient moves in the scanner is data.
- Flaws of Simplification: Relying solely on aggregate metrics like average Framewise Displacement obscures the nuanced dynamics of human movement, potentially hiding systematic differences between clinical cohorts.
- Informed Pipeline Design: Knowing that motion correction can strip away behaviorally relevant variance forces researchers to be more deliberate, balancing the removal of physical artifacts against the preservation of individual traits.
Key Foundational Work
Agito ergo sum: Correlates of spatio-temporal motion characteristics during fMRI (2020) — contributed 11 concept elements to this direction.
: Correlates of spatio-temporal motion characteristics during fMRI (2020) — contributed 8 concept elements to this direction.
Head Motion in Diffusion Magnetic Resonance Imaging: Quantification, Mitigation, and Structural Associations in Large, Cross‐Sectional Datasets Across the Lifespan (2025) — contributed 2 concept elements to this direction.