The past few chapters have provided a general overview of the structure of the brain and have also briefly mentioned a few important functions associated with those brain regions. But before we delve deeper into conventionally explored brain regions, one question remains: how should we be thinking about these brain regions? Are we taking a modular view, that is, each region is defined as a spatially isolated area of the brain that carries out its own independent function? Or should we subscribe to Pessoa's view that modularity is an oversimplification, and we should instead be interested in the multifunctionality of regions? This is what I'll be attempting to answer in this post.
Discovering the function of brain areas
Historically, one of the most important methods of uncovering brain function is the use of lesion studies. Briefly, lesion studies explore how different behaviours and mental capacities might be impaired after naturally occurring damage in human brains, or surgically induced damage in animals. One of the most important findings from lesion studies came from Paul Broca in 1961, who concluded that speech impediments in his patient (nicknamed 'Tan') were caused by lesions in the frontal lobe. Meanwhile, other researchers like Hitzig and Fritsch carefully removed different brain areas from dogs and saw movement impairments in their contralateral side (i.e., lesions in left brain areas caused movement impairments on the right side, and vice versa). Importantly, other functions and behaviours were left unaffected in the cases of Tan and the dogs.
An important thing to note here is that there is a gap between the results we see and the conclusions we make. Observing that a lesion in a given brain area leads to a functional impairment is NOT the same as saying that that brain area is causally responsible for that function. This is basically a more specific case of correlation is not causation. To get to this conclusion, a causal inference must be made. And in the fields of neuroscience and neuropsychology, this form of causal inference is hugely influenced by the logic of dissociations.
Single dissociations
Let A and B refer to 2 separate tasks, with m being a manipulation of a specific brain region, BR1. Then, a single dissociation occurs when m of BR1 affects the performance on A, but not on B. Let's make this abstract formulation more concrete. Let task A refer to face recognition (i.e., the ability to distinguish between and recognise faces), task B refer to object recognition, manipulation m refer to a lesion, and brain region BR1 refer to the fusiform face area (FFA). Then, a single dissociation is observed when a lesion to the FFA leads to a loss of face recognition ability, but leaves object recognition unaffected. And this was exactly what Wada & Yamamoto (2001) found in a patient (FYI, this condition is known as prosopagnosia).
From single dissociations, it might seem intuitive to conclude that we have found the underlying causal basis for a function. In the above example, one might readily conclude that face recognition is a function performed specifically and independently by the FFA. However, there is a problem -- an alternative account exists. What if the FFA is actually involved in visual recognition more generally, that is, it is involved in BOTH face and object recognition? And when damaged, the first functions that are affected are those that are more difficult to accomplish? In this case, we are making the sensible assumption that recognising faces is more challenging than recognising objects. Consequently, it is not hard to see how this account can explain why damage to the FFA might lead to impairments in face, and not object, recognition.
Double dissociations
So, we can't exactly make a strong case that the FFA is uniquely involved in face recognition from single dissociations alone. How has neuroscience attempted to get around this? The answer lies in double dissociations, which is the additional observation that when m of another brain region, BR2, performance on B, but not of A, is affected. Applied to our prosopagnosia example, a double dissociation is observed if, in addition to what we have observed with the FFA, damage to another brain region impairs object recognition, but leaves face recognition intact. And this is what we find in patients with agnosia. For instance, Moscovitch et al. (1997) showed how patient CK (presumably with damage to brain regions distinct from the FFA) had object recognition deficits and not issues with face recognition.
The strength of double dissociations lies in their ability to rule out the more generic difficulty account outlined above. If this difficulty hypothesis were true, we would not expect a case where an 'easier' function (i.e., object recognition) is impaired while a 'harder' function (i.e., face recognition) remains unaffected. But because this is what we observe, we can more confidently infer that the brain region, independent of other areas, is in fact uniquely involved in that function. Another famous example of a double dissociation is that of Broca's and Wernicke's areas in the brain, damage to which impairs speech production and comprehension, respectively, while leaving the other function intact.
The modularity assumption
Double dissociations are typically seen as the 'gold standard' evidence that neuroscientists should be striving for, but conclusions inferred from them still hinge on an important assumption, specifically that of modularity. What does it mean for a system to be modular? Pessoa provides different ways to conceptualise modularity, such as a system having parts that can be altered separately, or a system whose subcomponents carry out specific subfunctions that contribute to an overall function. But he points out that the most important feature of the brain as a modular system is that it has to be physically modular, that is, it can be physically decomposed into different modules, or brain regions, that don't spatially overlap. In other words, saying that the brain is modular is to say that not only are there separate processes and functions that are carried out by the brain, but that these processes and functions are independently implemented by physically separated and isolated parts of the brain.
Another way to put this is to say that functions arise from the inherent properties of a specific brain region, rather than from the interactions between regions. This melds really well with the traditional reductionistic approach to doing science, which involves breaking down a larger, complex system into its subcomponents, studying the properties of those simpler elements, and using those properties to explain the behaviour of the whole system. Think about how we typically talk about the brain. Roughly, the brain can be split into its cortical lobes. Then, we study each lobe individually. From there, we conclude that visual processing is a property of the occipital lobe, executive functioning is done by the frontal cortex, auditory processing by the temporal lobes, and so on. Finally, we bring them back together and explain the whole system (i.e., the brain) in terms of the properties of each individual element (e.g., the cortical lobes).
We have talked about issues with this approach plenty of times thus far, but just to reiterate, assuming modularity in biological systems is problematic because it assumes that conflates complexity with complicatedness. Physics, and more specifically engineering, often deals with complicated systems. This is why reductionistic analyses work here! But for unpredictable complex systems characterised by non-linearity and self-organisation, the same approach doesn't suffice.
And we have empirical evidence of the limitations of modularity, too! In a modular view of the brain, you would expect to see a one-to-one mapping (see Fig. 1A) between brain area and function; that is, each brain region is uniquely involved in one and only one function. However, we see evidence suggesting otherwise. For instance, initial suggestions that the prefrontal cortex was uniquely involved in working memory were challenged by findings that this area is also linked with other executive functions, such as attention, inhibition, and even emotions and motivation. This suggests a one-to-many mapping (see Fig. 1B). Meanwhile, research shows that both the frontal and parietal cortices are involved in attentional control, indicating a many-to-one mapping (see Fig. 1C). All in all, this suggests that the relationship between brain regions and functions involves a many-to-many mapping (see Fig. 1D), which is completely at odds with the modularity assumption.
Fig. 1 Possible area-to-function mappings in the brain (Pessoa, 2022)
Brain areas are multifaceted
So, if describing brain area function via a one-to-one area-to-function mapping is limited, how then should we conceptualise brain regions to emphasise the alternate many-to-many mapping? Here, Pessoa suggests describing the functional repertoire of brain regions, that is, to describe each brain region based on its relative involvement in various functions.
To illustrate this, Pessoa considers 2 brain areas in the visual cortex, namely visual area 4 (V4) and MT (found in the middle temporal lobe), as well as 3 distinct functions, namely perceiving stimulus direction (i.e., which direction is it moving?), wavelength (i.e., what colour is it?), and orientation (i.e., how is the object positioned). Results from neuroimaging studies show that, given a direction the stimulus is moving, 85% of MT cells respond, compared to 5% of cells in V4. When it comes to stimulus wavelength, this changes to 50% for area MT and 0% for area V4. Finally, the percentages change again to 75% and 50% for cells in MT and V4, respectively. Using just 2 brain regions and 3 functions, we can now describe each area using the format (% of cells activated for direction, wavelength, orientation). Consequently, this gives us two 3-dimensional vectors of (0.85, 0, 0.75) and (0.05, 0.50, 0.50) for areas MT and V4, respectively (see Fig. 2).
Fig. 2 A multifunctional description of brain regions (Pessoa, 2022)
Note here that the number and type of dimensions we use are completely arbitrary. On the one hand, we want to introduce enough functional dimensions to comprehensively cover all possibilities of psychological and physical capacities humans are capable of, and to ensure that each combination of dimensions uniquely distinguishes one brain area from the next. On the other hand, we want to be parsimonious and not introduce more functional dimensions than necessary.
This new issue aside, Pessoa and his colleagues scoured through the entire database of fMRI studies on isolated brain regions and attempted this functional repertoire analysis of brain regions using 20 different dimensions that captured key aspects of perception, action, cognition, and emotion (see Fig. 3). Doing so also allowed them to summarise the functional diversity of each brain region, that is, the extent to which each brain region is involved across multiple domains.
Fig. 3 The 20 functional dimensions used (Pessoa, 2022)
Concluding remarks
This approach is not without its limitations. Other than arbitrary determinations of functional dimensions, it currently relies on a bedrock of studies that likely suffer from confirmation bias. For example, if researchers believe that the amygdala is linked to emotional response, they are likely to actively seek statistical significance for this association. Combined with publishing bias for positive results, it won't be surprising if results for this area-to-function mapping are overrepresented in the literature. This has obvious impacts on the functional repertoire analysis conducted by Pessoa and colleagues (e.g., the emotion dimension of the vector summarising the amygdala's functional repertoire would be artificially inflated).
References
Moscovitch, M., Winocur, G., & Behrmann, M. (1997). What Is Special about Face Recognition? Nineteen Experiments on a Person with Visual Object Agnosia and Dyslexia but Normal Face Recognition. Journal of Cognitive Neuroscience, 9(5), 555–604. https://doi.org/10.1162/jocn.1997.9.5.555
Pessoa, L. (2022). The entangled brain. Journal of Cognitive Neuroscience, 35(3), 349–360. https://doi.org/10.1162/jocn_a_01908
Wada, Y., & Yamamoto, T. (2001). Selective impairment of facial recognition due to a haematoma restricted to the right fusiform and lateral occipital region. Journal of Neurology Neurosurgery & Psychiatry, 71(2), 254–257. https://doi.org/10.1136/jnnp.71.2.254
No comments:
Post a Comment