A large brain-imaging study suggests that different kinds of mental health symptoms line up with different patterns in how the brain’s networks connect at rest. The research was led by Yueyue Lydia Qu and Avram J. Holmes and published in Nature Mental Health. In simple terms, the brain features linked to “inward” symptoms like anxiety tended to resemble each other and they also differed from the features linked to “outward” behaviors like aggression. You can read the original study.
That idea matters because people’s struggles rarely fit into one neat box. Many symptoms overlap, especially in kids and teens. If researchers can map broad symptom patterns to broad brain patterns, it may help explain why certain problems often travel together and why other problems tend to show up in a different cluster.
Still, brain scans are only one piece of the story. Life context, stress, sleep, relationships and school or work demands also shape mental health. This study adds a clearer picture of how resting-state brain connectivity relates to two big symptom families across development.
Two Symptom Clusters That Show Up Across Mental Health
Psychologists often sort symptoms into two big groups called internalizing symptoms and externalizing symptoms. Internalizing tends to include experiences such as anxiety, sadness, withdrawal and physical complaints that show up with stress. Externalizing tends to include rule-breaking, aggression and other disruptive behaviors.
These categories show up in research on children, teens and adults. They also show up in everyday talk. A child who “keeps it all inside” often looks different from a child who “acts it out,” even when both kids are having a hard time.
At the same time, people can have a mix. Someone might feel worried and also get into conflicts. Because symptoms can blend, researchers have wanted stronger tools for spotting shared roots, including brain-based patterns.
One reason scientists care about these clusters is that they guide how studies are built. They also shape how mental health questionnaires are scored. In this study, the researchers asked a focused question: do brain network features that predict internalizing look more like each other than they look like predictors of externalizing?
How The Study Tested Brain Connectivity Across Age Groups
To explore that question, the team used functional MRI scans collected while people rested quietly. This method is called resting-state fMRI. It captures natural, ongoing activity patterns, rather than responses to a specific task.
Instead of looking for a single “spot” in the brain, the researchers focused on connections. They estimated functional connectivity, which reflects how strongly activity in different brain regions rises and falls together over time.
Importantly, they tested the idea across age. The largest group came from the Adolescent Brain Cognitive Development project, often called the ABCD study. It included 5,260 children around age 10. Two other datasets helped test whether patterns carried over: 229 adolescents aged 12 to 18 from the Healthy Brain Network and 423 young adults from the Human Connectome Project.
Each participant had symptom ratings from questionnaires. Those ratings captured internalizing and externalizing traits. Everyone also had high-quality resting scans, with the brain divided into 419 regions for connectivity estimates.
Next came the prediction step. The researchers used a machine learning approach called kernel ridge regression. The goal was to predict symptom levels from each person’s connectivity pattern and then compare how similar the “predictive maps” were within and across symptom categories.
Brain-Based Predictors Clustered By Symptom Category
The key result was about resemblance. Predictors for internalizing symptoms tended to look more similar to other internalizing predictors. Predictors for externalizing symptoms tended to look more similar to other externalizing predictors.
Put another way, the brain pattern that helped predict anxiety looked closer to the pattern that helped predict withdrawal than to the pattern that helped predict aggression. The same kind of within-group similarity showed up for externalizing measures.
What makes this interesting is that the result showed up across children, adolescents and adults. The strength of prediction varied, but the “family resemblance” of the predictive patterns held up. That supports the idea that internalizing and externalizing reflect partly distinct brain connectivity signatures.
There was also evidence of overlap. Some connectivity features seemed relevant across symptom types. That fits with what many families and clinicians already see, which is that different symptoms can share underlying risk factors.
Which Brain Networks Mattered And How That Shifted With Age
Beyond the broad split, the study also asked which networks carried more weight. In children and adolescents, externalizing symptoms were more strongly linked to connectivity involving the brain’s visual network. That might sound surprising at first, yet visual and attention systems work closely with how kids take in social cues and respond in real time.
Internalizing symptoms in younger participants leaned more toward connections involving subcortical regions. These are deeper brain structures that play roles in emotion, motivation and learning from reward and threat. Many researchers connect these systems to how people detect stress and how strongly they react.
Adults showed a somewhat different picture. In the young adult sample, prediction relied more on connectivity within large-scale networks, including limbic and temporal-parietal systems. These networks are often discussed in relation to emotion, memory and social understanding.
One of the more thought-provoking findings was that the same connection could “matter” in different ways depending on age. A link between subcortical regions and a temporal-parietal network related to externalizing in youth and internalizing in adults. Development changes the brain’s wiring and the brain’s job demands change too, so the meaning of a connection may shift over time.
If you’re a parent, teacher, or mentor, this age angle can feel familiar. The same child behavior can signal different things at different stages. Brain research cannot read minds, yet it can highlight a simple idea: the developing brain reorganizes and symptom patterns may ride on that reorganization.
How Accurate Were The Predictions?
The models predicted symptom levels with modest accuracy in the large child sample. The results were statistically stronger than chance, yet they were far from perfect.
When the team tested generalization in the adolescent and adult datasets, performance dropped. Sample size likely played a role. Smaller datasets make it harder to get stable estimates, especially for complex brain-wide patterns.
This is a common reality in brain prediction studies. Human brains vary a lot. Symptoms also vary across settings, reporters and time. So a scan-based model may pick up some signal, while leaving plenty unexplained.
What The Findings Suggest And What They Do Not Yet Provide
These results suggest that internalizing and externalizing are linked to partly distinct patterns of brain connectivity. The patterns showed up across development. That gives researchers a clearer target for future work that tries to map symptom structure onto brain structure and function.
At a lifestyle level, the findings also offer a calmer framing for why different struggles can feel so different. Some people mainly deal with inner distress. Others mainly deal with outward conflicts. Those differences may reflect different mixes of brain network organization plus life experiences.
Callout: A helpful way to read this study is “patterns across groups.” It supports broad categories. It does not label any individual person based on a scan.
It’s also worth remembering what the study design can support. The research used prediction and similarity analyses. It helps show that symptom clusters align with different predictive brain features. It does not tell us which life events caused those brain patterns, or whether changing symptoms would reliably change the connectivity pattern.
Limits To Keep In Mind
One limitation is time. The datasets were cross-sectional, which means different people were measured at different ages. This design can compare age groups, yet it cannot show how the same person’s connectivity and symptoms move together across childhood, adolescence and adulthood.
Another issue is real-world complexity. Questionnaires are useful, yet symptom scores depend on who is answering and what they remember. Stressful weeks can look different from calm weeks. That adds noise to any model that tries to predict symptoms from brain data.
Generalization is also a challenge. The child sample was very large. The adolescent and adult samples were smaller. Differences in recruitment, scanning procedures and demographics can also affect transfer from one dataset to another.
Callout: The takeaway is best described as “distinct group-level brain patterns.” It sets up future research, including studies that follow the same individuals over time and test broader samples.