Sensory and Association Cortex Make Different Kinds of Inferences

Sensory neurons belonging to early sensory cortex in the back of the brain are responsible for creating mental imagery but only hold information for very brief periods. They stop firing rather quickly after being stimulated and thus quickly lose the many detailed relationships that they have the capacity to hold. Only some of these relationships between lower-order features of the environment are preserved and only because they are shuttled up to association neurons (like those in the PFC) in the front of the brain where neurons can fire for sustained periods. These association neurons can then send this information back to earlier sensory cortex allowing them to rerepresent what they have forgotten. This is done constantly during the thinking process. However, what the sensory areas are directed to rerepresent is rarely identical to what they represented earlier. This is because association areas don’t just hand back the same information organized in the same way. Association areas actually hold information from multiple, successive instances of sensory imagery, recombine the elements that they think are important and then ask sensory areas to build an unique image using the new set of elements. This process allows conscious thought, but only really evolved in mammals because it is capable of making mistakes, unless it has the luxury of a good amount of early trial and error. Mammals, and especially humans, with their protracted childhoods and doting mothers, have this opportunity. Thus mammals are capable of carefully programming their association cortices to understand relationship that are spread apart in time, whereas other animals (from reptiles to invertebrates) really only have a chance to program their sensory cortices to understand relationships between stimuli that present at the same time. This allows lower-order sensory areas to build imagery that combines features that have never been actually witnessed before, and to simultaneously represent subjects that are always removed in time, within the same image.

Sensory neurons generally hold veridical information, true relationships between lower-order features in the environment. In other words, they are less susceptible to certain kinds of mistakes. They might combine simultaneously presented features in the wrong way, and thus misrepresent what they are perceiving, and they do this from time to time resulting in illusions or hallucinations. They are much less likely than higher-order areas though, to make faulty generalizations across time. In other words, their activity alone could not be responsible for a delusion. Lower order sensory areas hold snapshots of environmental imagery and because these neurons do not fire for very long they can usually safely infer that they are representing existing relationships between coocurring features. When you look down and see a glass of milk spilled across the floor your early sensory areas safely know exactly what they are looking at. There is no ambiguity. When you notice that your glass of milk has been moved to the sink and you see wet towels in the trash, you can only infer that your milk was spilled and cleaned up. This kind of inference requires the persistent firing seen in association area neurons that are capable of bridging across multiple early sensory images. How is this inference generated? The PFC unconsciously picks representations that have proven to be reinforced in similar scenarios in the past. Neurons in association and sensory areas do something very similar. They are experts at determining which of their neurons should be converged on by their inputs. In other words, when we find ourselves thrust into a new and unexpected situation, association areas, taking in to consideration all of their inputs (as well as their current activity) choose for us what aspect should be taken through time. This is based on prior probability and is intended to arm us with the microrepresentations necessary for us to act adeptly in our current scenario.

Association neurons are apprised of sensory information but because they fire for longer periods, they generate inferences based on multiple appraisals that have arrived at different times. In some ways these time-delayed inferences are more tenuous, but on the other hand they allow complex and abstract thought. Association neurons create subjective inferences about relationships between objective topics. Because they fire for extended periods, they have the capacity to misrepresent the environment in misleading ways. For instance, we could infer that the milk cleaned up itself or that the milk and the paper towels fell directly into the trash. We don’t make these kinds of absurd inferences though, because we have sufficient early experiences to have a sense for how improbable such occurrences would be. Our previous experiences, and common sense gained in our childhood keeps us from activating the neurons that would be responsible for these kinds of inferences. Sensory areas make records saying, I know that the sequence of scenes that I just saw are possible in my world but I don’t know how or why they might be related to each other. The association areas, on the other hand, are actively trying to make sense of these scenes by perpetuating the salient features, across many scenes, in the mind’s eye. They do this hoping that the features that they selected as salient will converge on the sensory circuits that will result in imagery that represent the proper inference. It is surprising how well this works for us in today’s world especially considering that this process was engineered to help humans make inferences about hunting and gathering.

Read an article that I published on this issue here:


http://www.sciencedirect.com/science/article/pii/S0031938416308289