Embracing uncertainty in neuroscience

As 2025 ended, you may have seen the latest scientific breaking news – adolescence lasts until your 32 years old! Can you believe it? It is our project so yes, we believe it… maybe? Interpreting neuroscience research accurately requires: 1) an understanding of what the researchers measured and 2) an awareness of that measurement’s limitations.  With just a little background in both areas, you can understand the scientific claims and their limitations enough to appreciate the exciting progress while recognizing how much we still don't know.

In our study, we explored how the brain rewires (i.e., strengths and weakens connections) across the lifespan. We found that across a sample of healthy people ranging from infants to 90-year-olds, the brain rewires in 5 distinct phases which are separated by four major turning points around 9, 32, 66, and 83 years old. Of particular interest was that a single phase of rewiring – what we called the adolescent phase – extends from around age 9 to 32. 

There are a lot of questions that pop up from these findings. How are 30-year-olds in the same phase as pre-teens? What do these brain changes mean for behavioural development? Does the 32-year-old turning point relate to lifestyle changes, such as having a family or progressing in a career? The short answer to these types of questions is we don’t know. We didn’t measure any of that. Research progress happens in tiny steps, and we hope others will build from our work and answer the questions we did not.

Still, even when it comes to what we did measure, there are many reasons to look our findings through a critical lens.

Neuroimaging is informative yet imperfect and tracing methods are flawed

Figure 1. A connectome of an adult's brain showing the complex map of structural connections. Red represents connections spanning left-right, green represents connections stretching front-back, and blue represents connections extending top-bottom.

Models of brain networks, called connectomes (Figure 1), are created from magnetic resonance imaging (MRI; Figure 2). MRIs provide us with brain scans, from which we can map or trace the connections in the brain to create connectomes (Figure 2). Our connectomes can tell us a lot about who we are – from our genetics, personality, and cognitive traits to mental health and environmental experiences.

MRIs require the person being scanned to stay very still. Motion distorts the images, blurring the boundaries of different tissues and connections. Even if someone is perfectly still, blood flow causes the brain to wobble. We take these imperfect images and do our best to correct for these distortions and artifacts – essentially ‘cleaning it up’ the way a photographer might edit a picture to improve its quality.

Then, we need to take these brain images and convert them into 3-dimensional connectomes (Figure 1). Each pixel (what we call a ‘voxel’), tells us which direction a connection is pointing in that tiny area, thanks to the basic principle of water diffusion. Water molecules in the brain will flow along the connection, not through the wall. Therefore, you can imagine each voxel of our brain scans having a tiny arrow showing what direction water is moving, which conveniently will also be direction of the connection.

Figure 2. MRI scans can capture the direction of water movement. Fibre tracking then can be performed to trace path of water movement across the whole brain, creating a complex map of connections – a connectome.

Thanks to this feature of water diffusion, we can perform a method called fibre tracking (Figure 2), which allows us to trace a connection from voxel to voxel like you could use your finger to trace the route between two points on a road map.

Now, while this connection-tracing method is cool (at least to me), it is flawed. It’s hard to detect connections when multiple paths cross through the same area, which happens often. It’s well-known that these methods sometimes identify connections that aren’t actually there (due to noise from our imperfect images) or miss connections that are present (due to our flawed tracing methods). There are methods to control for these mistakes, such as removing very weak connections from the network as they may be not actually present, but these corrections are not fool proof. Luckily, there are many incredible researchers working on a variety of ways to improve the accuracy of fibre tracking and in the coming years we hope to continue to improve our understanding of complex brain connectivity.

Participants typically don’t represent the real world

It is thanks to generous volunteers all over the world that we can conduct such large-scale research projects. However, participant samples typically do not reflect the world around us. Marginalized groups are underrepresented, such as queer people, people of colour, and those from low socioeconomic backgrounds. This severely limits how research findings can apply to these communities. This critique applies to the datasets we used in our studies as well, which included mostly White participants from the U.S.A. and U.K. As with other limitations, many researchers are focused on bridging this gap by seeking out those from underrepresented communities.

Development is a moving target

Our project and many others are cross-sectional, not longitudinal. This means we have one brain scan per person from people at different ages. Longitudinal studies, on the other hand, use multiple brain scans from the same person at different points in their life. Cross-sectional studies are more feasible because they don’t need the same person to come back over and over, but the studies have important limitations. If we compare a 5-year-old to a 10-year-old, any observed differences may be due to their age difference (suggesting a developmental factor) or it may simply reflect that they’re two different people.

What does this all mean for our project?

Now, this may seem negative, but limitations are a reality in science and understanding them is essential for accurate interpretation. We need to understand what studies did not do in order to extract real, meaningful insights from research. When we interpret our results, we need to know that models of brain networks are flawed, participants may not be representative, and developmental changes across different people have been inferred.

It is helpful to approach our project and many others from a ‘cup half full’ perspective. What neuroscience research can’t tell us is frustrating, but it also means there’s significant room for growth. Technology is improving rapidly (see portable MRIs), our algorithms are getting more accurate (see how neuroimaging can help predict patient outcomes), and researchers are actively working to recruit more representative samples. We know relatively little about the brain, so it is a very exciting area of research to explore.

So, does the adolescent phase of brain rewiring really extend into your 30s? Given the data and resources available now, yes, it looks like it. But despite some of the headlines, this study really was just a small step forward. The real question now is whether decades of improvements in neuroimaging, algorithms, and datasets will one day reveal what we missed.

Embrace the uncertainty!