The neural mechanics of sensory consciousness

This is the fourth in a series of posts inspired by Todd Feinberg and Jon Mallatt’s new book, ‘The Ancient Origins of Consciousness: How the Brain Created Experience‘.  The first three were:

• What counts as consciousness?
• Predators and rise of sensory consciousness
• Types of sensory consciousness

So, at this point in the series, we’ve talked about what is primary or sensory conscious, how it evolved, and the three types: exteroceptive consciousness of the outside world, interoceptive consciousness of internal body states, and affective consciousness of emotional reactions.  But how does it actually happen?  How is it actually implemented in an animal’s brain?

A quick note before we get into this: while there’s undoubtedly enormous value in looking at how consciousness arises in biological systems, I think we have to keep in mind that the way it is specifically implemented there may only be only one of many particular ways to achieve it. Indeed, F&M (Feinberg and Mallatt) note that within the realm of biology itself, there appear to be a wide range of architectures among differing species.

There’s a tendency in these types of discussions to assume that there’s a type of magic involved in the precise mechanisms.  But what I think is far more important are the capabilities, capabilities which may be achievable in other ways.  But maybe this is just my own bias toward functionalism showing.

The fundamental unit of the nervous system is the neuron, a cell that specializes in communication.  Neurons typically have thousands of branching tendrils with connections, called synapses, to the tendrils of other neurons.  Synapses get stronger with more use and weaker with non-use.  (Neurons that fire together, stay together.)  Some synapses excite the receiving neuron, and some inhibit it.  A neuron essentially sums up all the incoming signals from its inbound excitatory and inhibitory synapses and, if excited to a certain threshold, fires off a signal to its outbound synapses, to other neurons which are summing their inputs and deciding if they’ll fire.

The perceptual senses start off with a layer of neurons that fire because of some physical excitation.  For vision, these are photoreceptor neurons on the retina that are sensitive to light in various ways.  The pattern of signals they receive from incoming photons create a neural firing pattern, which triggers signals that go up the optic nerve to the brain, where the overall current mental image of the world is constructed.

The earliest layers of neurons in visual processing are principally excited by things like lines, edges, colors, and other primal aspects.  Subsequent layers become more selective in what excites them, perhaps being triggered by certain shapes or a certain kind of movement.  As the signals propagate up the layers, the neurons become progressively more selective about what triggers them, until we get to clusters of neurons that may only become excited when a certain person’s face is seen, or maybe a certain kind of animal, or any specific object.

F&M identify these neural hierarchies as a crucial element in biological consciousness.  Each sense has its own set of hierarchies, but there is substantial crosstalk between them, and it’s probably reasonable to surmise that there are effectively hierarchies of the hierarchies.

Each concept in the brain has its own twisting winding hierarchy that culminates in clusters of neurons which only light up for that concept.  (In reality, anything but the simplest concepts probably have innumerable hierarchies spread over different senses.)  At the lower levels, the conceptual hierarchies overlap substantially, but the pattern of the neurons firing at those levels determine which neurons at the higher levels light up.  As we go higher up in the hierarchies, they overlap less and less.

This concept is sometimes derisively referred to as the idea of the “grandmother neuron”.  But it’s important to understand that there is no image of grandmother in any one neuron (or of whatever the specific concept is).  That image exists throughout the entire hierarchy.  It’s also unlikely that the top of a hierarchy ever converges to only one neuron.  It may converge on clusters of neurons that fire in specific patterns.  It may be that the same clusters firing in different patterns is the top of a different concept hierarchy.

Neurologist Antonio Damasio calls these culmination points convergence-divergence zones.  In both his and F&M’s views, the hierarchies can be activated from the bottom up, from incoming sensory activation, or from the top down, when crosstalk from some other hierarchy activates a higher point, leading to downward propagation of the hierarchy.  This is likely part of what happens when we remember or imagine an object we’re not currently sensing.

My way of understanding the overall significance of this framework is that the breadth of the sensory layers provide the resolution of what can be perceived in the lower levels, as well as the overall number of concepts that can be perceived at the higher levels.  But the depth of the hierarchies, how deep they can go, probably determines just how much abstracted meaning a system can extract from the sensory information, in other words, how much understanding it can achieve.

F&M investigate how deep these hierarchies need to be for what we’re calling sensory consciousness to exist.  They tentatively conclude that the minimum number of layers is five for vision, but may only be three for the other senses.  They end up splitting the difference and using four as the average minimum.  But they fully admit that this is not at all an area that is fully understood, and that the actual minimums may be much higher.  My own intuition here is that there’s no fact of the matter distinction, just increasing effectiveness as the layers increase.

Obviously brains with more neural substrate will have an advantage, although we always have to remember that absolute size isn’t the crucial factor.  (A substantial portion of an elephant’s or whale’s sensory breadth is taken up with processing all of the interoceptive information coming from their vast body.)  But all else being equal, broader and deeper neural hierarchies will enable broader and deeper sensory consciousness.  Human consciousness and intelligence, with its symbolic and abstract thought, likely requires very deep, integrated, and nested hierarchies of hierarchies.

One quick note about the conceptual hierarchies.  The description above may make it sound like their formation is entirely contingent on incoming sensory information.  But it’s important to remember that minds don’t start as blank slates.  They start with a genetic predisposition to recognize certain concepts.  For example, healthy human babies have a built in capacity to recognize human faces.  Many hierarchies come pre-wired or primed for action, although possibly refined by later events.

Anyone in the know will recognize that this description is woefully oversimplified.  It’s a heavily summarized version inspired by what F&M describe in their book, with insights that I picked up from Antonio Damasio’s book, ‘Self Comes to Mind‘.  There is an enormous amount of important detail I’m leaving out.  The only goal here was to give you a taste of it.  If you want more, I highly recommend their respective books.

In the next and final post of this series, we’ll swing back to the hard problem of consciousness, and discuss to what extent we may have made any progress on it.