This final post addresses an obvious puzzle: why is reflective thinking sensory based? We can, after all, think about all sorts of abstract nonsensory topics. We think about God, the size of the universe, the mental states of other people, the validity of arguments, arithmetical facts and other mathematical entities, and so on and so forth. Why is our conscious thinking about such matters constrained to take place in a sensory-based imagistic format? After all, when artificial intelligence designers create systems that can reason about such matters, they do not use a sensory-based system to do so. On the contrary, the representations they design to be involved in such processes are just as amodal as the subject-matters those representations are about. So why are we so different?
The answer I propose is an evolutionary one. We are constrained to use a sensory-based working-memory system for conscious thinking for the same basic reason that we are constrained to be four-limbed creatures. Both properties are too deeply built into the body plans of mammals and birds (and perhaps all vertebrates) for evolution to have been capable of providing any other solution. Suppose that human life-ways had evolved to require the use of three hands, forcing cooperation between two or more people for day-to-day survival. It would be natural to wonder in such circumstances why humans had not evolved a third arm. But in this case the answer would be plain: such a radical change in body-plan could not appear suddenly in the human lineage. The answer is the same, I suggest, for sensory-based thinking.
We know that basic brain architectures, just like skeletal structures, are highly conserved across species. And we know that all mammals have attentional networks homologous with those of humans. Indeed, even birds have attentional networks that are partly homologous to ours. This is likely so for good reason. All creatures face a version of the same problem, which is to select, from among the deluge of information available to them, some smaller set of items for deeper processing. The upshot is the “centered mind”, in which attended items are globally broadcast for a wide range of different systems to process and respond to, thereby coordinating the activity of those systems (and hence the organism as a whole) around a single focus.
We also know that many mammals and birds not only have the capacity to sustain and manipulate representations in working memory, but that many species make use of these capacities for planning and problem solving. The same attentional network that was initially designed to select perceptual items for global broadcasting was adapted to sustain those representations in the absence of a stimulus (as when an object disappears behind an occluder), and to activate representations from memory into the global workspace as well. Moreover, many species seem capable of mentally rehearsing potential actions, attending to the sensory forward models that result, and evaluating the likely consequences in advance of acting. In most cases what are manipulated in working memory are representations of perceptually-available objects and properties, however, such as how the water level in a tube would change if one dropped a sequence of stones into it. When humans (and to some degree other primates) started to use their working-memory systems to reflect on more abstract matters, then, they were constrained to use sensory-based representations (visual images or inner speech, for examples) in order to do it.
These points tie nicely back to the topic of my last post. It has often been claimed that System 2 reflective reasoning is uniquely human. But this is mistaken. Prospective reasoning, in particular (in which one mentally rehearses the actions open to one and responds affectively to the results), is employed by many different creatures besides ourselves. What is different about humans is that we have vastly more concepts that can be bound into the sensory-based contents of the global workspace, and that (being uniquely capable of speech) we can rehearse speech actions to guide and control our reflections. We may also be unique in making chronic use of the working-memory system in mindwanding, which for us is our “default mode” (though this has yet to be established). Moreover, humans are unique in acquiring norms of reasoning from their culture, and constraining their working-memory processes accordingly.