On the Same Wavelength: How Brain Similarity Predicts Who Becomes Friends

This post presents fascinating findings by Shen and colleagues (2025) on the neuroscience of friendship.

What is the Science?

Picture the first week of a new community. People are polite, a little guarded, and quietly trying to answer the same question: Who feels like "my kind of person" here? Friendships often get explained as luck plus proximity. You sit near someone, your schedules overlap, you run into each other enough times that a greeting turns into a conversation, and then into something sturdier.

Social scientists have a name for the other force that shapes these outcomes: homophily, the tendency for people to connect with others who resemble them. Sometimes that resemblance is obvious, like age, language, or background. Sometimes it is subtle, like humor, conversational tempo, or what kinds of stories you find compelling. The challenge is that subtle similarities are hard to measure without asking people to describe themselves, and self-descriptions can miss the very things that make two people click.

This is where Shen and colleagues bring neuroscience to bear on a very human puzzle. They ask whether friendship is partly built on a deeper kind of alignment: not just liking the same things, but processing the world in similar ways. In practical terms, they test whether two strangers who show similar brain responses to the same movie clips tend to end up closer in the friendship network months later.

The core scientific idea is simple enough to be intuitive.

If two people reliably notice the same moments, react emotionally in similar beats, and interpret ambiguous scenes in comparable ways, it may feel easier to connect. Conversation flows. Jokes land. Misunderstandings are fewer.

What this paper adds is a careful attempt to measure that alignment before friendship forms, which helps separate selection from the more obvious alternative: that friends become similar because they spend time together.

What Did They Study?

The study follows a real group of people as their social world takes shape: an incoming MBA cohort of 288 graduate students. Early in their arrival on campus, before most friendships had time to form, a subset of students took part in a functional MRI (fMRI) session. A functional MRI does not record thoughts, and it does not read minds. It measures changes in blood oxygenation that track, indirectly, the activity of nearby neurons. It remains one of the best tools we have for asking how the brain responds over time to complex experiences.

In total, 41 students were included in the fMRI analyses. The researchers then mapped the cohort's friendship network twice: once about two months into the program and again about eight months in.

That gave them two ways to ask their core question. First, does early neural similarity predict who ends up being direct friends or several steps apart in the network? Second, does early neural similarity predict who grows closer over time, as the network changes through friendships forming, holding, or fading?

This second question is a quiet strength of the paper. Friendship is not just a snapshot. Early connections can be opportunistic and temporary. Later connections may reflect more enduring compatibility. The authors set out to see whether brain-based similarity relates not only to where people end up, but to the direction their relationships travel.

How Did They Do It?

The scanning task was intentionally "life-like." Instead of pressing buttons or doing simple laboratory puzzles, participants watched a sequence of short movie clips chosen to be engaging and varied. The goal was to hold attention while also inviting differences in interpretation and emotion. Two people can watch the same scene and focus on different details, make different inferences about motives, or feel very different things. Those differences are not noise here. They are the signal.

To quantify neural similarity, the researchers used a method commonly called intersubject correlation. Here is the idea in plain terms. Take one brain region, such as a patch of cortex involved in visual processing or a region often linked to social understanding. For each participant, track that region's activity moment by moment across the movie clips. You get a time series, a line that rises and falls over time. Now compare two participants' lines. If their lines rise and fall together, the correlation is high, and those two people count as more neurally similar in that region while watching the same clips.

They did this not for one region but across the brain, using a standardized "parcellation" that divides the cortex into 200 regions, plus 14 subcortical regions like the amygdala and thalamus. In total, that made 214 regions for analysis.

Friendship was defined conservatively. A tie counted only if it was reciprocal, meaning both people named each other. The researchers also computed social distance within the network, defined as how many friendship links connect two people. Friends have a distance of 1. Friends-of-friends are typically distance 2. The paper focuses on distances up to 3 because beyond that the meaning of closeness becomes less stable.

Because brain data and network data create lots of dependencies, the statistics were built to respect that structure. The team used permutation tests, repeatedly shuffling participant labels to estimate what patterns would look like by chance while preserving the shape of the social network. They also corrected for multiple comparisons across the many brain regions tested.

They added checks to keep the interpretation honest. After the scan, participants rated how interesting and enjoyable they found each clip. The researchers were then able to test whether neural similarity was simply a fancy version of "we liked the same videos."

They also statistically controlled an extensive set of sociodemographic similarities, including age, gender, nationality, and background features, to see whether the brain effects were merely reflections of demographic sorting.

What Did They Find About Friends?

The results are best understood as two stories: one about who ends up as friends, and another about who grows closer over time.

When the authors compared future friends to all non-friends lumped together, they did not find a broad, clear difference in pre-existing neural similarity. That matters, because it prevents the study from collapsing into a simplistic "brain scan predicts your friends" headline. Friendship is complex, and a lot of it remains contingent on circumstances.

But when they sharpened the comparison, a specific effect appeared.

Future friends showed greater pre-existing neural similarity than pairs who ended up three steps apart in the network, and that difference localized to a portion of the left orbitofrontal cortex, a region often linked to valuation and subjective preference.

The authors interpret this as consistent with shared tastes and preferences: what feels rewarding, funny, or appealing may be aligned for some pairs even before they become friends. That said, this particular effect was sensitive to demographic controls, an important caution we will return to.

The more striking and broader result concerns change over time. Between the 2-month and 8-month surveys, the friendship network reorganized. Some relationships intensified, others cooled, and many indirect connections shifted as different friendships formed and dissolved around them. The authors calculated whether each dyad grew closer, stayed the same, or grew farther apart.

Here, pre-existing neural similarity mattered most when comparing dyads that grew closer to dyads that grew apart. Dyads that moved toward each other in the social network showed higher neural similarity as strangers across many regions, including the thalamus, left amygdala, and a wide spread of cortical regions. The regions implicated span systems involved in visual processing, attention, and higher-order meaning-making.

This pattern supports the authors' central psychological interpretation.

The brain similarities that predicted social convergence were not limited to "liking" in a simple sense. They seem to involve alignment in how people interpret unfolding events, where attention tends to land, and how emotional responses develop across time.

Watching movies is a useful probe for this because it demands continuous integration. You are tracking characters and motives, updating expectations, noticing social cues, and reacting emotionally while the story moves.

The control analyses deepen the takeaway. Similarity in self-reported enjoyment or interest did not fully explain the neural effects. In other words, neural similarity captured something beyond a few post-scan ratings.

Meanwhile, sociodemographic similarities did help explain part of the orbitofrontal finding distinguishing future friends from distance-3 dyads, and gender similarity was singled out as especially influential in reducing that effect when controlled. Yet the broader links between pre-existing neural similarity and growing closer over time were less easily dismissed as demographic sorting.

The study suggests that early, circumstance-driven ties may come and go, but later-emerging closeness may be shaped by deeper compatibilities visible in how similarly two brains respond to the same rich, real-world stimuli.

What Were the Strengths and Limitations?

The biggest strength is timing. The fMRI data were collected at the start of the social story, not after it had already unfolded. That makes the findings harder to shrug off as "friends become similar because they are friends." It does not prove causality, but it strengthens the inference that similarity can come first.

Another strength is ecological validity. Naturalistic movie clips are not neat laboratory tasks, but they resemble the complexity of everyday experience. If the question is about shared reality and social compatibility, it makes sense to use stimuli that force people to build meaning in real time.

The paper also benefits from measuring the entire cohort's network with extremely high response rates at both follow-ups, and from using conservative network definitions like reciprocal friendship.

At the same time, there are real limitations. The fMRI sample was 41 participants, which is respectable for an intensive neuroimaging study but still small and context-specific. These were MBA students in a particular program with structured group assignment and housing arrangements. The social environment is unusually dense and immersive, which may shape how friendships form.

It was also an observational study. Even with early brain measures, unmeasured factors could drive both neural similarity and later friendship. Cultural background, formative experiences, and values could create similar ways of processing the world and also increase the likelihood of friendship, without neural similarity being the causal mechanism itself.

A practical limitation is that a planned longer-term follow-up fMRI was disrupted by the COVID-19 pandemic. That matters because it would have helped test social influence more directly, examining the possibility that friendships further increase neural similarity over time.

Finally, the result pattern itself gives a healthy warning against overstatement. The clearest and broadest neural signal appeared for who grew closer versus who grew apart, not for the simple friend versus non-friend split. That is more subtle than fortune-telling, and more faithful to real social life.

What Was the Impact?

This paper reframes a familiar idea in a measurable way. People often describe close friends as sharing a worldview, being "on the same wavelength," or understanding the same moment without needing much explanation. Shen and colleagues offer evidence that this alignment can be detected, at least in part, before friendship solidifies.

Methodologically, the study demonstrates how to link neuroimaging to real social networks over time, rather than studying isolated individuals. Conceptually, it suggests that compatibility may not only live in personality adjectives or stated preferences, but in the continuous, often unspoken ways people attend to and interpret the world.

The impact is not that we should scan people to match them with friends. The impact is that it gives researchers a new handle on what "shared reality" might mean at a mechanistic level, and it provides a rigorous design for separating selection from influence. It also opens practical questions for future work: which kinds of stimuli reveal the most socially meaningful similarity, how cultural context shapes these patterns, and how neural homophily interacts with the push and pull of social environments.

Key Takeaways

1. Neural similarity measured when people were still near-strangers predicted who ended up socially closer in a real friendship network months later.

2. A narrow effect in orbitofrontal cortex distinguished future friends from pairs who ended up three steps apart, consistent with aligned valuation or taste.

3. The strongest finding was about trajectories: people who grew closer over time were more neurally similar at the start than people who drifted apart, across many brain regions.

4. Self-reported liking of the clips did not fully explain the neural effects, suggesting the brain measures captured richer alignment than simple preference ratings.

5. Some effects were sensitive to demographic controls and the context is specific, so the work supports neural homophily while also highlighting the need for replication in broader communities.

Glossary

blood-oxygenation-level-dependent (BOLD) signal: a change in MRI signal tied to changes in blood oxygenation, used as an indirect marker of neural activity.

default mode network (DMN): a set of brain regions that tend to co-activate during internally oriented processing, such as constructing meaning and social understanding.

false discovery rate (FDR): a procedure for controlling the expected proportion of false positives among statistically significant results when many tests are run.

functional MRI (fMRI): a neuroimaging method that estimates brain activity by tracking changes in blood oxygenation over time.

geodesic distance: a shortest-path measure in a network, defined as the smallest number of links needed to connect two nodes.

homophily: a tendency for people to form social ties with others who are similar to them.

intersubject correlation (ISC): a correlation-based measure of how similarly two people’s brain activity time courses respond to the same stimulus.

naturalistic stimuli: a set of complex, lifelike materials, such as movies, that evoke rich cognitive and emotional processing.

orbitofrontal cortex (OFC): a frontal brain region often implicated in representing subjective value, preferences, and reward-related meaning.

parcellation: a division of the brain into predefined regions to support consistent analysis across participants.

Pearson correlation: a statistic that quantifies the degree to which two variables or time series vary together.

permutation test: a statistical approach that estimates chance effects by repeatedly shuffling labels to generate a null distribution.

reciprocal friendship tie: a social connection counted only when both people name each other as friends.

social distance: a network measure describing how many relationship steps separate two individuals.

time series: a sequence of measurements taken over time, such as brain activity sampled repeatedly while someone watches a video.

Open-Access Reference

Shen, Y. L., Hyon, R., Wheatley, T., Kleinbaum, A. M., Welker, C. L., & Parkinson, C. (2025). Neural similarity predicts whether strangers become friends. Nature Human Behaviour, 9, 2285–2298. https://doi.org/10.1038/s41562-025-02266-7

About the Author

Fred Shaffer earned his PhD in Psychology from Oklahoma State University. He earned BCIA certifications in Biofeedback and HRV Biofeedback. Fred is an Allen Fellow and Professor of Psychology at Truman State University, where has has taught for 50 years. He is a Biological Psychologist who consults and lectures in heart rate variability biofeedback, Physiological Psychology, and Psychopharmacology. Fred helped to edit Evidence-Based Practice in Biofeedback and Neurofeedback (3rd and 4th eds.) and helped to maintain BCIA's certification programs.

Support Our Friends

On the Same Wavelength_ How Brain Similarity Predicts Who Becomes Friends.pdf

Download PDF • 2.07MB