In this chapter, Pessoa aims to build a hypothetical 'minimal brain'. Why might this be important? For starters, the brain is an immensely complex organ, and this is even before we talk about human brains! Before we increase the level of complexity and look at more advanced capabilities, we begin by establishing the parts of the brain that contribute to basic animal functions like defending oneself and seeking rewards, that is, survival!
An aside
Before we move on, I'd like to point out an implicit assumption that runs throughout the chapter -- that the brain is an information-processing computer that processes and transforms input signals to generate some form of output. While commonly conflated, the reductionist view (i.e., that the brain can be reduced to isolated regions of interest) and computational theory of mind are separate. My understanding so far is that while Pessoa rejects the reductionist view of the brain in favour of one that is more integrated, complex, and network-based, he hasn't thus far said much about the brain-as-a-computer analogy.
Coming from an ecological perspective, I naturally reject the computational theory of mind. An alternative exists, though -- ecological neuroscience argues that the brain resonates instead of computes. That being said, I don't yet have the requisite knowledge or vocabulary to provide a comprehensive and convincing overview of this alternative. At the moment, I'll continue using the language as used by Pessoa, and refrain from interjecting with too many anti-computation comments (if only because this might disrupt the flow of the blog).
The second visual system
What do we first need to build a minimal brain adequate for survival? In other words, what's the first step to detecting danger and rewards? In most animals, perception is most reliably and efficiently performed using vision. So, we'll need to start with a brain that allows for robust visual perception. In a previous blog, it was mentioned that one of the well-known functions of the occipital lobe is that it houses the primary visual cortex (see Fig. 1), responsible for visual perception. Early evidence for this came from studying dogs and monkeys, with subsequent studies in humans showing that the extent of damage to the visual cortex is correlated to the extent of one's blindness.
Fig. 1 The primary visual cortex (source: link)
While this seems like a pretty straightforward case of function-to-structure mapping, there was a problem. Here, some studies showed that even with extensive damage to their visual cortices, blind individuals could somehow detect motion at above chance levels. This led to the discovery of the superior colliculus, which Pessoa terms the second visual system of the brain and also the main focus of our chapter on the minimal brain.
The superior colliculus is located on either side of the midbrain, situated at the top of the brainstem. Structurally, it is made up of multiple layers (see Fig. 2), with the topmost layer receiving signals from the retinas in our eyes. Here, the signals received by the superior colliculus are said to be topographical, that is, the spatial layout of light hitting the retina is preserved in the superior colliculus. In other words, the cells activated in the superior colliculus are thought to form a map of the external world, hence enabling an animal to know the spatial layout of their surroundings.
Fig. 2 Layered structure of the superior colliculus (Pessoa, 2022)
Meanwhile, the deep layers in the superior colliculus propagate signals that control muscular action. Combining the superficial and deep layers, the superior colliculus acts as a middleman that transforms visual inputs to motor outputs that help direct body, head, and eye movements (see Fig. 3).
Fig. 3 Superior colliculus as a sensorimotor interface (Pessoa, 2022)
Linking this to the minimal brain, Pessoa imagines a simple schematic diagram (see Fig. 4B) where sensory input is always transformed to produce motor output. So far, so good -- we want a minimal brain to be able to respond to external stimuli by producing the appropriate movements. The problem here is that such simplicity does not allow for much flexibility. In this case, the same inputs will always produce the corresponding output (in Pessoa's words, input and output are always coupled). This isn't what we see in reality, where animals can perform a wide array of actions despite being provided with the same stimulus or sensory input. The solution to this is to decouple inputs and outputs to increase the flexibility of the system (see Fig. 4C). This provides some first hints as to why brains are often so complex -- it is this complexity that increases the flexibility of our behaviours!
Fig. 4 Schematic diagrams of input-output coupling and decoupling
Another aside
At this point, I'd like to note the input-output computation-esque language used here. The meaningless input is transformed and processed by some brain structure to produce meaningful and useful outputs. Coincidentally (but not unsurprisingly), this is a main reason why psychology heavily relies on the concept of representations. The world itself provides ambiguous information -- we must therefore possess mental representations that transform this meaningless information into something that makes sense to us.
As briefly mentioned in another blog, this is based on the flawed assumption that the world is inherently ambiguous. But what if it isn't? What if the world is full of ecological information, that is, kinematic information that uniquely specifies the dynamic causal properties that created them? Then, by detecting this information, we can perceive the world without the need for any superfluous mental gymnastics (i.e., information processing, computations, transformations).
How might this work in neuroscience? Once again, I'll be the first to admit that I am still learning and am not 100% sure about the exact mechanisms involved. But from my knowledge, an alternative to computation is resonance. Think about two tuning forks that are next to each other (see Fig. 5). When one of the tuning forks is hit, it starts vibrating at a specific frequency. Given the right conditions, the other tuning fork starts to vibrate at that same frequency (and with a larger amplitude). This is resonance -- the information from one tuning fork is transmitted to the other tuning fork, where the latter resonates with the former by acting in a way that is specific to the properties of the former.
Fig. 5 Resonance between tuning forks (source: link)
Note how there is no processing or computation going on here. The vibrations of the first tuning fork aren't being acted upon to produce those of the second tuning fork. In a similar fashion, could brains work the same way? Instead of receiving sensory inputs that are processed to give motor outputs, perhaps neurons and signals in the brain are merely resonating to the unique structure of ecological information picked up by our perceptual systems.
At the moment, I don't have enough to convince you that resonance is the mechanism by which the brain works. All I am doing is to point out how rampant and implicit the computation analogy of the brain is, and that there is, in fact, an ecological alternative out there! But enough with my aside, let's return to the chapter.
Adding complexity to our minimal brain
The rest of the chapter outlines other important functions the minimal brain would need to perform for survival, as well as brain structures that might be responsible for them. Firstly, Pessoa points out 3 typical behaviours animals engage in when faced with danger, namely: fighting, fleeing, and freezing (i.e., the absence of either fight or flight responses). Here, experiments with rodents show that rats tend to freeze more often when faced with a novel stimulus in an unfamiliar environment. This suggests that animals show context sensitivity, where their action outputs are a result of combining sensory inputs with information about their context. Briefly, Pessoa claims that the hypothalamus provides contextual information about the animal's internal states to the middle layers of the superior colliculus (see Fig. 2). Hence, focusing on the superior colliculus, visual inputs from the top layer combine with contextual inputs from middle layers to produce motor outputs by the deep layers!
Moving on, areas such as the periaqueductal gray, or PAG (pronounced 'pee-eyh,-gee'), receive motor outputs from the superior colliculus and determine whether an animal flees or freezes. Meanwhile, the superior colliculus's connections with the PAG and substantia nigra (responsible for producing dopamine) help facilitate reward-seeking behaviours.
Extending the circuit
I've left out lots of details, but I think this general overview of the potential contributors to a minimal brain suffices for now. Just to recap, we started with the superior colliculus, which allowed us to generate motor responses from visual sensory input. Then, we added context sensitivity with the hypothalamus and the ability to respond to danger with the PAG. Finally, reward-seeking was facilitated by the PAG and substantia nigra. Importantly, all of these areas are connected and centrally linked back to the superior colliculus (see Fig. 6).
Fig. 6 Extending the minimal brain (Pessoa, 2022)
So, we've identified what it takes to build a minimal brain. At this point, Pessoa poses an interesting question: can we investigate basic functions for survival with respect to the minimal brain and the minimal brain alone? If you subscribe to a function-to-structure mapping approach to understanding the brain, you might have said yes. However, Pessoa reminds us of the central hypothesis of the book: that we can't understand behaviour and function with respect to isolated brain regions. Instead, our goal should be to find out how function results from complex distributed circuits in the brain.
That being said, Pessoa admits that, for pedagogical purposes, it is still useful to start by learning what individual brain regions might be doing. This is because it is typically easier to understand and learn things in a sequential manner, especially when encountering them for the first time. This approach might also make sense given the history of neuroscience, which tends to (and still does) ascribe functions to anatomically isolated brain structures. Here, while we advocate for an alternative approach, there is no reason to throw the baby out with the bathwater -- we can still learn from past research and try to incorporate it within this alternative framework.
Concluding remarks
To end off, Pessoa advocates that while we might start with an isolated brain region, we then need to move on to understanding function and behaviour with respect to that brain region and a gradually expanding circle of other areas connected to it. This then begs a question: what are brain regions, and how should we approach them in neuroscience? This forms the basis of the next chapter, which I will cover soon in a separate post.
References
Pessoa, L. (2022). The entangled brain. Journal of Cognitive Neuroscience, 35(3), 349–360. https://doi.org/10.1162/jocn_a_01908
No comments:
Post a Comment