Reply to Baars

Reply to Baars

by Teed Rockwell

     My claim that Skinner believed in psychological atoms is actually strengthened by Baars' remark that Skinner's behaviorist atoms could take a variety of physical forms. ( "A rat in a box could depress the bar by sitting on it, by using its paws, or biting it: these physically different responses were functionally equivalent operant behaviors.") Baars is correct that Pavlov, unlike Skinner, thought that psychological atoms were identical to certain physiological items. But Skinner, as a non-reductive atomist, thought he could permit his psychological atoms to have a variety of physical forms. He still believed that even though each S-R connection was not really physical, it could nevertheless be understood as being independent of all other S-R connections ,and without reference to the laws of physics. It doesn't really make sense to speak of something as being both ontologically determined by its function, and ontologically independent, but philosophical clarity was not Skinner's strong point.

     In fact, Skinner's brand of atomism actually forced his science to have far less explanatory power than the physical sciences he was trying to emulate. One of the things that caused the fall of behaviorism was the gradual awareness that this simplicity made real scientific laws impossible. Behaviorism was collecting lists of S-R connections, but couldn't make the items on those lists interact with the complexity of chemistry equations. Titchener did not make this mistake. His frequent use of chemistry metaphors indicates that he hoped his psychological atoms would eventually interact in ways that would give rise to properties that modern cognitivists would call "innocently emergent" or "network" properties. But because Titchener never got past the gathering stage, his atomism resembled Skinner's list-making in fact, even if it did not in theory. For this reason, I think Baars is mistaken in saying my critique is only applicable to Titchener. On the contrary it is more applicable to Skinner, for he made a virtue out what Titchener thought would be only a temporary necessity. When the behaviorists attacked Titchener, they did not reject his atomism: a point also made by Danziger in the quotation I cited. On the contrary, they gave it's list-making simplicity a permanent legitimacy which even Titchener, let alone Wundt, would never have accepted.

     This kind of "atom-gathering" is nicely described by Baars' phrase "superstitious scientism". But I certainly never meant to accuse contemporary cognitive psychology of that charge. The functional analysis which is essential to modern cognitive science is capable in principle of creating real laws, not just laundry lists, even if it does not always succeed. But there are two different attitudes towards functional analysis: one of which I was defending, and the other I was attacking even while declaring it to be superior to Skinnerian-Titchenerian atomism. The atomism of chemistry does not deny the existence of network properties, it merely declares that all emergent properties can be inferred from the intrinsic properties of the elements in the network. (this is what supposedly makes them "innocently emergent".) But Chemistry is not the only way of doing science, and neither is the kind of atomism which is inspired by it.

     To some degree, Baars' commentary makes the same mistake as Titchener and Skinner: he assumes that breaking things into atoms is the only possible way of comprehending them. But to some degree I have made a mistake that aids and abets his. I have not sufficiently clarified two very different kinds of atomism, both of which I am attacking for different reasons. Besides Skinnerian Atomism, there is a more sophisticated form of functionalist atomism which assumes that a functional analysis must necessarily be describing physical modules, each of which has a concrete location in physical space (usually the brain.) This could be true, and today a lot of intelligent people believe it. But I think it is important to remember that it is not the only possible position.

     When Baars and Titchener both say that it is impossible to even think without referring to elements and their relations, they are ignoring the possibility that mental activity could be a process all the way down, and that the functions being analyzed are mere events, like eddies in a stream, rather than modules with concrete material forms. It is this possibility which I believe partly inspires the critiques of cognitive psychology made by Nesser and Jenkins, and by Lakoff's critique of Chomsky. Each of these critiques claims that it is impossible to completely separate one mental function from the other, or to separate mental functions in general from the environment in which they take place. This fact could be accounted for if mental functions are the result of fluid interpenetrating processes, rather than physically independent modules.

     The process view of cognitive functions has already made inroads into what Baars calls the neuron doctrine. The neuron doctrine could be seen as a form of atomism back when people believed that my memory of my grandmother was stored in a "grandmother cell". But now that neuroscientists recognize that cognition is distributed throughout the brain, the fundamental cognitive units are transformation processes between computational weight spaces, not neurons. And there are other neuroscientists, like Walter Freeman, who stress that the brain should be seen as a dynamic system rather than a computer with neuronal modules. I probably stressed the fallibility of process based analysis more than most other advocates of Dynamic Systems Theory (DST). But almost every advocate of Cognitive DST does acknowledge that if we see an organism as what Dewey calls a systems of tensions, rather than a modular computer, it will not be possible to separate the organism from it's environment when studying it. Processes, unlike atoms, are not separated by sharp borders. Thus, if cognition is a process all the way down, it will be impossible to have the completely closed systems which make predictions certain.