I recently finished Mark Solms’ new book, The Hidden Spring: A Journey to the Source of Consciousness. There were a few surprises in the book, and it had what I thought were strong and weak points.
My first surprise was Solms’ embrace of the theories of Sigmund Freud, including psychoanalysis. Freud’s reputation has suffered a lot over the decades. Solms thinks this is undeserved. He also complains about the behaviorist movement, which largely arose as a reaction to Freudian type theories, an attempt to put psychology on a more empirical footing. In Solms’ view, this came at the cost of removing the first person perspective from the field, something he doesn’t think has been adequately corrected for even in the post-behaviorist era.
What wasn’t a surprise is Solms’ belief that consciousness is rooted in the brainstem, something he’s well known for. In particular, he sees it centered on three regions in the midbrain region, which he refers to as the “decision triangle”: the PAG (periaqueductal gray), the superior colliculi, and the midbrain locomotor region. His reasoning is similar to Bjorn Merker’s final integration for action idea of consciousness, which Solms regards as the final decision point. The PAG in particular is the center of action.
However, Solms’ views are nuanced. He sees consciousness rooted in affect, in particular the conscious feelings of emotions and drives, which in his view originate from and terminate back to these regions. But the consciousness he sees here is affect consciousness. He acknowledges that perceptual consciousness is a cortical phenomenon, albeit one that only exists when the cortex is suitably aroused by the RAS (reticular activating system), another brainstem region.
This is a minority view in neuroscience, although the differences with the mainstream are also nuanced. All neuroscientists agree that the cortex is aroused by the RAS. They also agree that the most basal drives originate from the brainstem. And even that final integration of action takes place in the midbrain. But most see conscious experience as a cortical phenomenon, with subcortical forebrain regions like the amygdala, nucleus accumbens, hypothalamus, as well as the insular, cingulate, and orbitofrontal cortices, as playing major roles in affective feelings.
In general, I didn’t feel like Solms adequately engaged with the reasons for these mainstream views. He largely went the route of strawmanning them, saying that they only exist due to “theoretical inertia”. Anil Seth in his review of the book linked to some of the reasons. I think Solms passed up an opportunity by not engaging with the broader literature.
It’s worth noting that whether or not consciousness requires a forebrain has little bearing on animal consciousness. The evidence from ontogeny, fossil records, and model organisms all show the forebrain-midbrain-hindbrain architecture arose very early in vertebrate evolution. Any vertebrate fish you commonly think of has a forebrain, including a pallium, the covering over the forebrain that’s the precursor to the mammalian cortex. The idea that the brainstem is the most ancient structure is a common misconception, one even some scientists appear to buy into.
Anyway, Solms also sees Karl Friston’s free energy principle as a major part of his theory. He gives one of the best descriptions of that principle that I’ve seen. Unfortunately my understanding of it remains somewhat blurry, but my takeaway is that it’s about how self organizing systems arise and work. He identifies four principles of such systems:
- They are ergodic, meaning they only permit themselves to be in a limited number of states.
- They have a Markov blanket, a boundary between themselves and their environment.
- They have active inference, that is, they make predictions about their own states and the environment from that environment’s effects on their Markov blanket.
- They are self preservative, which means minimizing their internal entropy, maintaining homeostasis, etc.
3 is understood to involve active Bayesian processing, in other words, predictions. All of which leads us to the predictive theory of the brain, which I think is where Solms is at his strongest. Perception involves the more central regions making predictions, which propagate toward the peripheral (sensory) regions. But the peripheral regions compare the prediction with incoming sensory information and send back prediction error signals. This happens across numerous layers. We see what we expect to see, with the incoming information forcing corrections.
Solms notes that a self evidencing system receiving information that violates its expectations can react in one of three ways.
- It can act to change conditions to bring the signals more in line with its expectations.
- It can change which representation(s) it’s currently using to make better predictions. This is perception.
- It can adjust the precision of the predictions to more optimally match the incoming signal.
Solms identifies the last as consciousness. This strikes me as essentially learning, which is in the same ballpark as Simona Ginsburg and Eva Jablonka’s theory. Although Ginsburg and Jablonka require a more sophisticated form of learning (unlimited associative learning) than what Solms appears to be focusing on. The main point here for Solms, is that this is the origin of a conscious feeling, an affect, which again is what he considers the root of consciousness.
Toward the end of the book, Solms provides an extended critique of David Chalmers’ description of the hard problem of consciousness. His main point is that Chalmers overlooked affects in his deliberations. If he hadn’t, maybe he wouldn’t see the problem quite as daunting. Given that the hard problem is often phrased along the lines of, “Why does it feel like something to experience X?”, I think Solms has a point. Although in my experience, talking about affects usually isn’t seen as sufficient by those troubled by the hard problem.
Solms finishes up by noting we won’t know whether we’ve solved the problem of consciousness until we can build an artificial consciousness. So he’s working on a project to do just that, incorporating the free energy principle model. What he describes sounds like it will be a sort of artificial life. He emphasizes that intelligence isn’t the goal, just a self evidencing system concerned with its own survival. The problem will be finding a way to conclusively demonstrate success despite the problem of other minds.
If he succeeds, it sounds like he will immediately turn it off and try to patent the process to prevent it from falling into commercial hands. Aside from the ethical issues, he notes the danger in building self concerned systems. I usually think fears about artificial intelligence are overblown, but in this case, I agree with him on the danger. The good news is I don’t know how useful such systems would be for most commercial purposes anyway. Do we really want self driving cars or mining robots being worried about their own survival?
There’s a lot of interesting stuff in this book. I do think Solms makes a good point that affects, conscious feelings, are often overlooked in theories of consciousness. And I agree with him on their crucial role. But that role gains its power by the reactions of the perceptual and executive systems. Without those reactions, affects are little more than automatic action programs. They only become affects, conscious feelings, in a conscious system, which means making them the foundation of consciousness is a bit circular.
All of which brings us to a point I often return to, that consciousness is a complex phenomenon, one that can’t be reduced to any one particular property. Unless of course, I’m missing something?
This brief summary of Solms’ views omits a lot. If you’re interested in knowing more, aside from reading the book, he has his own blog, well worth checking out.