Journey into the human mind with Dr Kristian Sandberg, Chair of COST Action The neural architecture of consciousness.
“Conscious experience is central to our existence, but we still don’t know how our brains produce it, and radically different theories are still debated within the field”, says Dr Kristian Sandberg, chair of the Action.
In 2020, we still do not know how our brain generates consciousness?
No. There are a number of theories and, of course, important advances have been made in our scientific understanding of this phenomenon, but the mystery remains unsolved. Some dominant approaches have emphasised the role of the prefrontal cortex in the generation of conscious experience, but recent studies have opened up new possibilities as they conclude that consciousness is related to a posterior cortical “hot zone” which encompasses a much larger part of the brain.
What are the major theories of consciousness?
To cut a long story short, if a consciousness scientist subscribes to an existing theory, it is typically one of three competing theories which explain how the brain works and how consciousness is produced. The first is the global workspace theory (or global neuronal workspace theory), which is often illustrated with a “theatre metaphor” in which the visible stage is what you are actually conscious of. Yet, as in a theatre, there is a lot happening backstage: your brain is still processing images, memories, and other information before they have their moment in the spotlight.
The second theory is higher-order theory. It assumes that consciousness consists of perceptions or thoughts about first-order mental states. You can say this is the ability to create an image of an image, and you are only able to do this when you are aware that what you are looking at is an image.
The third is information integration theory, which relates consciousness to a specific definition of information complexity. Essentially it posits that the more complex a system is, the more distinct states it can be in and the greater the potential for consciousness; it posits that ongoing conscious experiences are related to functionally connected groups of neurons with high information complexity according to a particular mathematical definition.
Which theory does your team find most promising?
There was a survey a couple of years ago that showed that most senior consciousness scientists have a favourite theory that they consider more promising than others; however, I am not sure I do. And to be honest, I am not sure that the theory that will eventually turn out to be true has been proposed yet. There are some ongoing efforts to pit the existing theories against each other by testing their predictions experimentally. I think this is one important way to move forward, but a lingering issue is that many theories start from the introspective perspective of the proposer and then formulate conceptually what the characteristics of conscious experience might say about the neural underpinnings. This means that the predictions of the theories are sometimes open to interpretation, and it is consequently unclear how to attempt falsification. This leads to situations in which attempts at falsification lead to only minor clarifications of the theory.
The fact that many, many excellent scientists have been debating and testing the theories for more than 20 years without resolving the matter makes me wonder if we need to take a step back and examine the context of our science: Are we asking the right questions and are our theories specific enough to be tested?
Recently, we have seen attempts to generate different models of various consciousness phenomena from different theories and then compare the evidence for these models. I think that is a very important step in moving the debate forward. I also wonder if it is perhaps not time to supplement this by building models bottom-up from the data. For either approach, we need rich datasets from a wide range of well-controlled studies. That is part of what we hope to supply in our COST Action.
You are chairing a COST Action – the neural architecture of consciousness, but what exactly is neural architecture?
Let us start with the Action itself: it is broad and inclusive, so we support researchers working on any aspect of consciousness, but the large, coordinated effort is unique in two ways. First, we create large datasets with exceptionally high statistical power for relating magnetic resonance imaging (MRI) to behavioural consciousness data, making them ideal for testing competing predictions and building models from the bottom up. Secondly, the vast majority of existing magnetic resonance imaging studies (maybe more than 90%) use functional MRI to examine ongoing activity, whereas we focus on the hardware of the machine that creates consciousness.
The term “neural architecture” mostly refers to brain structure, but I try to avoid using that word because it is my experience that many scientists associate it with a specific magnetic resonance imaging sequence, a so-called T1-weighted scan. This sequence is very common and has traditionally been referred to as a structural scan, but it really only informs us about some limited aspects of the brain’s structure. Our MRI data is much richer: we will not only be getting high-quality “structural scans”, but we will also be getting information about the brain’s microstructure, about the internal wiring of the brain, and about neurotransmitters.
What do we get as a result?
When we combine our magnetic resonance imaging data, we get what we call a neuro-architectural map – a detailed and multidimensional structure. You can think of it a bit as representing the building blocks of our mind. I would draw an analogy to a complex three-dimensional LEGO construction where each block is important in terms of its size, shape, and position. Of course, when we use MRI, the metaphor breaks down a bit as each sequence and analysis gives us different information, so you would perhaps need to think of it as many LEGO constructions superimposed on one another, or as four-dimensional.
What sets your Action apart from other ongoing efforts?
The Action is broad and inclusive, so we support researchers working on any aspect of consciousness, but the large, coordinated effort is unique in two ways. First, we create large datasets with exceptionally high statistical power for relating MRI to behavioural consciousness data, making them ideal for testing competing predictions and building models from the bottom up. Second, the vast majority of existing MRI studies (maybe more than 90%) use what is called functional MRI to examine ongoing activity, whereas we focus on the hardware of the machine that creates consciousness, i.e. what we broadly term neural architecture, which is vastly understudied.
This is a very interdisciplinary project.
Brain research has always benefited from interdisciplinarity, but for this endeavour I do not really see how else we could do it. It is now well-known that sample sizes used in previous cognitive neuroscience studies have often been too small, but when we did our statistical power calculations we were surprised that the difference in the sample size required to do just one traditional study linking one brain characteristic to one aspect of consciousness is not that different to what is required for linking 100 brain characteristics to 10 aspects of consciousness. It is maybe the difference between 150 and 250 participants. So, it basically seemed the only financially viable way to do this is large scale; double the number of planned participants to be able to do tens or hundreds of “studies” in one go. But that gave us the problem that no single scientist can possibly be qualified to work on all these different paradigms and data types, so we ended up as a group with very mixed profiles.
You mentioned characteristics of consciousness. What is that?
Well, it is typically behavioural data from healthy participants or patients. As researchers, we ask people to do very specific things and then we measure how our participants perform in those tasks. We might measure how strong an illusion is for them or how accurate they really are when presenting something so faint that that they think nothing is there. Or it could be the conscious state of a patient recovering from brain injury.
You are also going to study DNA and use machine learning? These seem like two very different methods.
Yes, these are two separate things. We will be getting DNA data because this is the “instruction manual” to what makes us human. DNA is responsible for our individual differences and this is very important data, particularly when coupled with our neuro-architectural maps. Since we’ll be collecting a lot of brain and behavioural data from hundreds or thousands of people all over the world, collecting DNA to link it to is a very low-hanging fruit that could potentially enlighten us about the imprint of nature and nurture on the brain and behaviour.
Machine learning is a way to analyse the data, and it is often a useful tool for examining how much you can predict from a data set – in our case how well we can predict conscious experiences or states from our neuro-architectural maps. But we will also use other explanatory statistical methods to test the relation between neural architecture and consciousness.
How can society benefit from your research?
Firstly, we should keep in mind that modern society is a great beneficiary of modern science, which has made our lives healthier, longer, and happier. We hope and believe that this project could substantially improve the prognosis for consciousness disorders. Accurate prognosis could have a substantial positive impact on the lives of patients and relatives, and it could facilitate clinical decisions regarding whether to escalate or stop treatment.
Ultimately, the project will be part of a great collaborative effort to uncover the answers to humanity’s biggest questions: What is consciousness? What makes us human beings? So, while others leave Earth to explore the universe, we turn our gaze inwards on an equally exciting trip: a journey into the human mind.
View the COST Action page
View the network website