An Introduction to Working Memory Capacity and rs fMRI Mapping

Using Connectomics to Visualize the Storage on the Human Memory Chip

Anastasija Petrovic
8 min readNov 11, 2021

Quick, you have a test tomorrow! What should you do?

The obvious, but very discouraged cramming?

The consequential, yet at the time rewarding art form, winging it?

Or will you consult a researcher to expand your memory?

It sounds like it is just the thing we need, right? Well, before you try out the Connectomics technique of the future, we have to understand the current memory processes and the connectomes that contribute to their function.

What is memory?

Memory is usually defined as the processes that are used to acquire, store, retain and later retrieve information from the brain. Memory is a crucial part of human cognition as it allows one not just to checklist items off at the grocery store, but also to through past experiences frame their current understanding and behaviour. There are three major processes that characterize how memory works: encoding, storing and retrieving (recall).

For a more extensive overview of memory check out the article I wrote for this project!

Working Memory Capacity

Working memory capacity (WMC) is the maximum amount of context information that an individual can retain in the absence of external stimulation. Essentially, without any outside help or pressure, WMC is the maximum one can retain (store and then retrieve) the information.

Functional connectivity, the collective set of functional connections in the brain is usually measured with the functional magnetic resonance imaging technology (fMRI). Check out my article here to learn more about fMRIs and functional connectivity!

With this technique, as proclaimed first by Sebastian Markett et al. on their work with connecting WMC and the functional connectome, future researchers are able to gain a better understanding of memory retention and manipulating the capacity.

Working memory is the ability to retain such information in the absence of external stimuli. This is one of the most fundamental portions of cognitive neuroscience for its connection to clinical phenotypes, intelligence and academic success.

As I highlighted in the image above, specified regions within the lateral prefrontal cortex are the most prominent supporters of working memory processes. During the storing portion of the memory processes, PFC regions show sustained memory-based activity throughout.

However, neuroscientists are still debating whether the PFC stores any memory itself. Is the region merely carrying out the process but now storing anything? Current neuroscience points us to understand that PFC merely provides attention control signals for the extrastriate area (the area which processes visual information). So, with this lead, we have come to another dead end.

Other than extrastriate area-based areas, there is another neural correlate for visuospatial information in the posterior parietal regions along the dorsal visual stream.


This region exemplifies the idea that neural activity is always sensitive to the amount of information retained. Evidence from electroencephalographic (EEG) recordings and functional magnetic resonance imaging has demonstrated a linear relationship between activity and working memory capacity in terms of identifying individual memory capacities. This was observed in the intraparietal sulcus (IPS) in the right hemisphere.

The key idea to takeaway from these dead ends is that implicated regions exchange memory. So far, many things seem very messy and way too connected. If our brain had a LinkedIn account it would definitely have the 500+ connections symbol!

One last thing in terms of what has been understood of WMC is that cognitive neuroscience sees a greater need for understanding first the aspect that guides human behaviour the most, vision. The visual aspect of memory is focused on even though there are other transporters of memory, like auditory for instance.

Baseline Understanding of Anticipated Results

The field of connectomics is currently emerging as the science of brain networks and connectomes. Although the Human Connectome Project is hard at work, we are still looking into Drosophila larvae from Katharina Eichler et al. mapping of the learning and memory centres.

To understand the best-case scenario of working memory capacity I will lay out the units and examples of what results should look like.

The connections focused around (KC) Kenyon cells (intrinsic neurons) and the random combinations of inputs presented. These neurons in the insect mushroom bodies (MB) serve to form and retain associations between stimuli.

If we were to look above in the image of the results MBONs are in the compartments innervated by their dendrites. Most connections among MBONs are axo-axonic, and fewer are axo-dendritic. Additionally, it has been found that MBONs have direct feedback with MBINs.

Above the researchers exemplified their findings of the MBONs and modulator neurons (MBINs) associated with every MB compartment. Each compartment is innervated by 1–5 MBONs and 1–3 MBINs. Above we can see that the three MBINs are associated with the neurotransmitters GABA (gamma-Aminobutyric acid), Glutamate and Acetylcholine.

As is shown above some inputs led to axo-dendritic connections some to dendo-dendtritic, but overall this random combination can be brought to a conclusion differently in terms of the behaviours presented.

Method of Visualizing Working Memory Capacity

Now that we have understood the results of mapping similar centres in insect brains, it’s time to understand the complicated yet fantastical mess in the human head.

In order to study the functional connectome, cognitive neuroscience studies have extensively capitalized on resting-state functional magnetic resonance imaging (rsfMRI).

Those two added letters take us beyond what I was talking about in my previous article. We are taking this to the next level. The resting-state brain level.

Working memory capacity cannot be directly observed but can only be inferred across a set of instances. It is almost like a cognitive trait with network properties of the resting brain. To map this centre we must use two types of mapping:

Degree Centrality (DC)

Through centrality mapping, a weight is assigned to each voxel based on its connectivity pattern with all the other voxels (value on a grid). We calculate degree centrality (DC) which is a voxel-wise summary unit of connectivity strength. In the essence of understanding “first step connectivity.”

Eigenvector Centrality (EC)

This type of mapping takes the results from the DC (centrality scores) and is added weights in accordance with those results. During EC mapping, centrality scores are weighted by the centrality of connected voxels.

So you know understand the theory behind this work? Let’s do it now!

My Turn: Preprocessing

Ah, the preprocessing portion. As much as it is difficult it is useful. But still, it is crucial.

The goal is to turn data into information, and information into insight. — Carly Fiorina

Data preprocessing is crucial in computational biology. If there is much irrelevant and redundant information present or noisy and unreliable data, then knowledge discovery during the training phase is more difficult.

The preprocessing of the fMRI data was carried out with CONN toolbox, which differentiated from Sebastian Makett’s work with the same WMC test. My preprocessing contained the following steps:

  1. Removal of first ten volumes
  2. Slice timing
  3. Realignment
  4. Voxel movement control
  5. Bandpass filtering
  6. Alignment of images with high-resolution structural scan
  7. Spatial normalization
  8. Spatial smoothing

Some of those preprocessing steps have been debated before but as I am looking into the newer field of rs fMRI I tested all the common steps with the help of Sebastian Markett et al. ‘s paper.

My Turn: Graphical Analysis

Let’s bring ourselves back to the beginning of my connectomics journey. I had just started learning about the connectivity matrix. When looking into the graphical results I applied the adjacency matrix once more.


This symmetric N-by-N matrix where N is the number of voxels is in graph theory used to create a square matrix to represent a finite graph. The elements of the matrix indicate whether pairs of vertices are adjacent or not in the graph. In the special case of a finite simple graph, the adjacency matrix is a (0,1)-matrix with zeros on its diagonal. With the presented project method, the N-by-N connectivity matrix is defined by Cij = (1 + Cij).

My Turn: Results

For the results, I will display the graphical results in this article, but feel free to watch my video demonstrating the entire process and results here.

In this graph, one can see the participants’ working memory capacity depending on the working memory load. For example the quantity of information or “items” on the memory at the time of data capture.

Here, one can see the results from the whole brain individual differences on the relationship between working memory capacity and the functional connectome.

If you look closely, particularly at the cluster in the right intraparietal sulcus (x = 42, y = -57, z = 48) it shows an inverse relationship between working memory capacity.

The top scatter plot shows extracted DC values, whilst the scatter plot on the bottom shows extracted EC values.

Throughout the work within the WMC idea, it is evident that working memory capacity relates positively to general cognitive ability. This neurobiological foundation highlights the idea that the frontal and parietal sites of the human brain show great differences. For example, in the resting state, the intraparietal sulcus is the central hub of the fronto-parietal network.

In the end, it must come down to understanding the intricate differences. We need further work, further referencing, further building, to truly apply the idea of this limit of working memory capacity to a future where we can maybe even consider expansion?

Ciao, I’m Anastasija, a 15 y/o innovator at TKS interested in the intersection between bioinformatics and biotechnology! Check out my website, connect with me on LinkedIn and follow me on Twitter!



Anastasija Petrovic

I’m Anastasija, a 17-year-old interested in the intersection between biotechnology and bioinformatics. I also write about mindsets and emerging technologies!