This is a 2-part series of articles. The first one is a tour of brain anatomy whilst the second is about two major non-invasive approaches we can use to record brain activity. Click here for part 2.
I like to start by stating the obvious:
Brains were not engineered or designed, they evolved over time.
First, Evolution is conservative – it tends to work via a “if it ain’t broke, don’t fix it” principle. Every new module which arose in the nervous system was built on top of older ones and had to plug into them to function. This means that “new” brain structures mainly affect the world by controlling “old” structures and that they see the world after it’s filtered by those older structures as well. Knowing this helps see sense in a lot of the convoluted pathways of the brain.
Second, Evolution acts through mutations in the genetic programs regulating embryonic development. This makes development a great clarifying principle from which to study brain anatomy, since structures developing from the same embryonic compartment tend to be functionally related and to have evolved at the same time.
Today we are touring 4 great sub-systems which correspond to the compartments from which different brain parts arise during embryonic development: hindbrain, midbrain, diencephalon and forebrain.
The Hindbrain: Life-support, Sensorimotor Reflexes and Learning
The first problem nervous systems had to solve was simply to stay alive. This involved coordinating vital organs, moving the body in response to stimuli and learning from experience. The Spinal Cord, Medulla, Pons and Cerebellum contain the minimal toolkit to achieve this.
Spinal Cord, Medulla & Pons
The Spinal Cord has a sensory branch on its back-facing side and a motor branch on its belly-facing side. The sensory branch senses information about touch, muscle stretch and pain throughout the body below the neck, and the motor branch contains neurons whose terminals innervate muscle fibres and cause them to contract. From the neck down, everything we feel enters the brain via the Spinal Cord. Everything we do exits the brain via the Spinal Cord. It is the earliest processing stage for bodily sensations and the final common pathway for every motor command. The circuits of the Spinal Cord control reflexes which allow us to withdraw from pain and noxious stimuli and maintain our balance. They are surprisingly complex and sufficient to enable us to walk.
The Medulla and Pons can be seen as extensions of the Spinal Cord. The Medulla controls ventilation, regulates blood pressure, can override the heart’s pacemaker circuits in “fight-or-flight” scenarios, and is responsible for vital head reflexes such as vomiting, coughing, sneezing, and swallowing. The Pons serves a similar earliest sensory relay and final common pathway as the Spinal Cord but for nerves of the head and neck. In addition, it contains two key neuromodulatory structures, the Locus Coeruleous and the Raphe nuclei, which regulate arousal, mood, aggression and sleep by releasing norepinephrine and serotonin widely throughout the brain.
The Cerebellum has almost 4x as many neurons as the Cerebral Cortex, and a similar outward appearance. It is organised into independent modules, each receiving, processing and outputting different streams of data: there are motor, visual and auditory modules, amongst others. However, the internal structure of cerebellar modules is the same, which suggests that the same computation is being run in parallel on the different data streams. Earlier theories restricted the Cerebellum to a role in fine motor coordination. More recent views suggest that its simple, modular, massively parallel circuits are very well suited for supervised learning, because they can compare expected versus actual associations between stimuli. If a new association is met or an old one violated, the Cerebellum can alert the rest of the brain to this via its connections to the Pons.
The Midbrain: Instinct and Dopamine
The Midbrain is a small collection of structures typically bundled together with the Hindbrain. I highlighted 2 of its roles: extracting sensory features important for survival and using them to unleash instinctive actions, and signalling reward expectations to the rest of the brain using Dopamine.
The Tectum receives lightly-processed visual and auditory information and looks for survival-relevant features in order to activate instinctive behaviours. These are more complex than simple reflexes and usually involve chains of actions. For example, in rodents the tectum can detect overhead expanding shadows – signs of approaching aerial predators – and provoke escape responses by activating Hindbrain motor circuits. The neurons of the Tectum also receive inputs from the Cerebral Cortex, which adapt their activity to the current context. Their outputs are other structures in the Midbrain, Brainstem and Spinal Cord, whose circuits encode basic motor programs that can be chained together into instinctive actions. Within the Midbrain, a key output is the Peri-Aqueductal Gray. Activating different sites in the PAG can release fleeing, freezing or avoidance behaviours and modulate the cardiovascular system accordingly.
Dopamine: the Substantia Nigra and Ventral Tegmental Area
Midbrain Dopaminergic neurons usually fire around 5 spikes per second. When a reward is encountered, they signal to the rest of the brain how “surprising” that reward was. If an unexpected reward is met, their firing rate transiently goes up. If a reward was expected but wasn’t delivered, their firing rate temporarily decreases. As an animal develops an association between a stimulus and a reward, they cease firing to the reward itself (it is now expected, so “surprise” = 0) and begin firing to the stimulus or action preceding it.
Midbrain Dopaminergic neurons keep the brain informed about the reward value of stimuli and actions through different pathways. The Substantia Nigra supplies Dopamine to the Basal Ganglia, helping select valuable actions. The Ventral Tegmental Area provides it to the Prefrontal Cortex and Nucleus Accumbens, influencing motivation and decision-making, and regulating the desire for reward.
The Diencephalon: Interfacing Between Worlds
The Diencephalon has 2 key structures: the Thalamus and Hypothalamus. The Thalamus connects the sensory organs and the Cerebral Cortex, and the hypothalamus connects the brain with the body’s internal organs.
The Thalamus contains predominantly excitatory glutamatergic neurons and its best-studied subdivisions each receive input from a different sense organ and send output to a primary sensory area in the Cerebral Cortex, suggesting a role as a sensory relay. However, it also contains inhibitory circuits which are critical for attention. If you’re trying to follow a conversation in a noisy environment, the Cortex can activate these circuits to focus your attention by inhibiting competing stimuli in the environment, “dialling down the volume” of distractors. The thalamus may also influence how areas of the Cortex talk to each other. Most of its subdivisions actually receive input from the Cortex instead of the sense organs, and in turn output to multiple cortical areas, allowing the thalamus to amplify or quieten select messages in cortex. Finally, as part of this “information dam” function, the Thalamus facilitates sleep by generating slow rhythms which decouple the Cortex from sensory input, allowing it to “switch off” from the outside world.
The Hypothalamus receives information about the time of day, ongoing motor and sensory activity, and cognitive and arousal state, and uses this to control hormone release by the Pituitary, a master gland which influences every organ in the body. The hypothalamic-pituitary axis ensures the internal organs are working in sync with the demands of the environment and ongoing actions.
The Hypothalamus also controls food and water intake by sensing hormones which signal how full the stomach is or how much salt is in the blood, for example. Furthermore, it is involved in releasing defensive, predatory, sexual and parenting behaviour – all of which are highly influenced by circulating hormones – through connections to downstream motor centres in the Midbrain and Hindbrain.
The Forebrain: Memory, Cognition, Unsupervised and Reinforcement Learning
The cortex is the wrinkly structure covering the brain’s entire surface. Its outer gray matter contains all its neurons’ cell bodies, whereas the underlying white matter contains the “cabling” of its inputs and outputs. The cortex is organised in layers, each populated at different density with neurons of different shapes and sizes. Around 80% of cortical neurons use glutamate as their neurotransmitter and are excitatory (i.e. their outputs make other neurons more likely to fire). The remaining ~20% use GABA as a neurotransmitter and are inhibitory (i.e. i.e. their outputs make other neurons less likely to fire). There are dozens of subtypes within each of these broad classes, and they intermix within each layer, forming orderly circuits connecting within and between layers.
The cortex is not a homogeneous tissue and can be divided in 40-50 regions on the basis of anatomy. Each region receives input from different areas, processes it through subtly different internal circuits spanning its 6 layers, and sends the resulting outputs to different brain regions. The interplay of these 3 factors is what gives each area a unique function.
Primary Sensory areas – of which there are 5, one for each sense – each receive information from one sense organ via the thalamus and extract simple features (e.g. location of edges in a scene, for vision). Higher Order sensory areas organise these simple features into more complex ones, which they detect in the sensory field (e.g. a collection of edges into a square) and relay for more advanced processing. This simple combinatorial process goes on through multiple rounds and allows the brain to detect very sophisticated features. For example, the Fusiform Face Area is selective for human faces, which it detects from complex combinations of the features which make up a person’s eyes, lips, nose etc.
Associative Areas sit downstream of higher-order sensory areas and, as the name implies, associate together the complex elements detected in the world through each of the senses in parallel. An associative area may respond to all of the following: seeing an object, hearing it mentioned in conversation, reading about it, etc. Eventually information reaches Executive Areas in the frontal lobe in highly abstract and location-invariant form. The frontal lobe is where information about perception, reward, emotion and memory comes together and is shaped into plans and decisions. These are communicated to the Motor and Pre-Motor cortices, where they are transformed into sequences of instructions sent to the Midbrain, Brainstem and Spinal Cord, which finally turn these instructions into muscle contractions and behaviour. This is an extremely simplified view; in reality, the Cortex is in a constant feedback loop with the structures below it and it functions not just by reacting to events in the environment, but also by anticipating and trying to predict them. Nonetheless, it’s useful to highlight how cortical signals can flow hierarchically.
What does Cortex enable us to do?
The Cerebral Cortex underlies our fine sensory perception, planning, reasoning, fine motor skills, language and other cognitive processes. But most important of all, Cortex enables mammals to learn “models of the world” from experience.
The world we inhabit does not come with labels for concepts which are useful. Nor does every action or decision we make come with a reward or punishment attached. Intelligence thus requires an additional type of learning, beyond supervision and reinforcement. There is a lot of structure and regularity to the world around us. Learning what is likely to happen in a situation and which things tend to happen together is a major advantage because it enables an organism to predict and anticipate, rather than simply react, to the world around them, which is what the Hindbrain and Midbrain are for. Furthermore, events in the world have causes and consequences. The Cortex and Forebrain enable mammals to “simulate” the consequences of different possible actions, hypothesise causes for events in the world (e.g. given that I’m in the forest, what is most likely to have caused the noise I just heard?), and learn continuously in an uncertain and ever-changing environment. Animals such as invertebrates, reptiles, fish and most birds rely on automatic stimulus-response strategies baked by evolution into neural circuits for recognising sensory patterns and triggering behavioural responses. They learn about their environment via habituation and sensitization, classical and operant conditioning, but they lack the mental and neural structure for the deeper models of the world which mammals can learn.
The hippocampus is located deep in the medial Temporal Lobe. It receives highly processed sensory input from associative cortical areas and connects reciprocally with the Prefrontal Cortex. Both spatial and memory theories have been proposed for it, which are not necessarily incompatible.
The spatial theory of the hippocampus is based on place cells and grid cells. Place cells fire bursts of action potentials when an animal traverses specific locations of space, with each cell having a preferred “place field” it fires to. Grid cells – found in the entorhinal cortex, which supplies information to the hippocampus – fire at multiple, equally spaced places arranged in a grid. Through these complementary systems for different scales, the hippocampal system maps layouts of the environment, like a GPS system.
The memory theory came from patients who lost the ability to form new long-term memories following hippocampal lesions. Long-term memories are initially stored in the Hippocampus, which then uses its ability to self-generate repeating patterns of activity (like a student memorising a fact) to consolidate these memories and gradually transfer them to distributed sites in the Cerebral Cortex. Once consolidated, because of these widespread connections, the Hippocampus also functions as a search engine for retrieving cortical memories and as an “auto-complete” to fill in patterns, such as when a smell reminds you of a particular place and time in your memories.
The Basal Ganglia (BG) are a collection of tightly-connected structures deep inside the white matter underlying the Cerebral Cortex: the Striatum, Globus Pallidus, Subthalamic Nucleus and Substantia Nigra (which is actually in the Midbrain). The BG have 4 functional territories, each connected with different cortical and thalamic areas: limbic, associative, sensory and motor. Information from Cortex enters the BG via the Striatum and Subthalamic Nucleus. It is processed by internal inhibitory and dopaminergic circuits, and sent out as an inhibitory input to the Thalamus, which projects back to the Cortex.
This loop between Cortex, BG and Thalamus enables the brain to learn from rewards and select actions with the highest reward expectations. To achieve this, the parallel cortical input channels to the BG carry “bids” for candidate actions. Depending on internal BG processing and Midbrain dopaminergic inputs encoding reward expectation, channels carrying bids for a high value action will be disinhibited, whereas the remaining ones will be inhibited and prevented from accessing motor pathways. In this way, the selected bid’s corresponding circuits will be amplified by the now disinhibited thalamic input, allowing that action to unfold.
This ends our tour of brain anatomy and the major subsystems of the brain. In Part 2, we will review two affordable yet powerful techniques for recording human brain activity without surgery.
About the Author
André Marques-Smith is a Scottish-Portuguese Neuroscientist, with a PhD from Oxford University and a postdoctoral career spanning King’s College London and the Sainsbury-Wellcome Centre for Neural Circuits and Behaviour, where he held a Henry-Wellcome fellowship. After early research in human neuropsychology, André worked on mapping developing cortical circuits in mice using electrophysiology and optogenetics. He went on to work on neurotechnology (electrocorticogram and Neuropixels invasive probes) and Systems Neuroscience (connections of the thalamus and behaviour), before joining a venture-backed non-invasive BCI startup in London. He is now CTO and co-founder of a new startup applying state-of-the-art neuroscience outside the lab.