A Natural Science of Culture; Or, a Neurological Model
of the Meme and of Meme Replication. Version 3.2
F. T. Cloak 02/01/2015
Only by providing a physical model of meme replication can memetics take its rightful position in the list of replicators covered by what Hull terms "general selection theory." Until then, they remain simply an analogy to the better-known case of genes. (Aunger 2000a, p.213)
Part One. A model of behavior.
The meme is the unit of culture and of cultural transmission/acquisition. Memes are copied from brain to brain and, when activated, mediate behaviors which, in turn, result in social interactions, artifact creation, world views, and everything else we customarily refer to as “cultural”.
But how are memes copied from brain to brain? How are they stored? Once stored, how do they mediate actions? In other words, what, exactly, is a meme, neurologically speaking? How does a meme work?
To answer these questions, we need first to adopt a neurological model of behavior-in-general. Surprisingly, there is no generally accepted model of behavior per se; that is, there is no generally accepted model
To understand the meme, I propose to use the behavioral model laid out by William T. Powers in 1973, and since refined by him and numerous co-workers. It’s called Hierarchical Perceptual Control Theory (hPCT).
hPCT asserts that when an organism acts, it is invariably comparing its perceptions to perceptual reference standards stored in its nervous system, and that it will continue to act until its perceptions approximate those standards. The neuroanatomy and physiology of perceptual control* is a product of hundreds of millions of years of evolution *.
The reference standards, too, are products of blind variation and selective retention: genetic evolution, learning by individual carriers, and, in some species, cultural evolution. Thus the perceptual control apparatus is adapted to early and recent past environmental conditions -- both within and outside of the carrying organism.
The perceptions which the apparatus controls are, however, perceptions of the organism's current environmental conditions, as detected by its sensory cells and inferred by its neural input machinery.
So, unlike the “standard model” assumed by most psychologists, and people in general, hPCT does not view behavior as programmed or reflexive responses to immediate stimuli. Instead, hPCT understands behavior as the maintenance of previously established states of the nervous system in the face of variations in its current surroundings.
The elemental neural mechanisms by which activity is mediated, the units of behavioral organization (Powers 1973: 221), are perceptual control systems, orCSes. CSes operate in functional hierarchies in the organism's central nervous system, each CS-in-a-hierarchy evoking reference standards in CSes below it. The following film clip illustrates the structure and operation of a CS.
In an exactly similar way, the address signal, and therefore the reference standard, may remain constant while the input perception changes, creating a disturbance to the CS; again, the Comparator, recognizing the discrepancy between input and reference signal, sends an error signal to the Output Function, which works to restore equilibrium.
For example, CS"x", in a motorist, controls a perception of Hood Ornament of Car at Certain Horizontal Distance from Edge of Road ("HOCCHDER"). As the car approaches a curve, the Input Function of CSx perceives a change in that horizontal distance and, accordingly, changes the signal it sends to the Comparator. Detecting a discrepancy between that signal and HOCCHDER, the reference signal, CSx’s Comparator sends an error signal to its Output Function. The Output Function, therefore, sends one or more signals down to address lower CSes, thereby causing a cascade of CS output signals down to the motorist’s muscles (and a responding cascade of CS input perceptual signals up to CSx) which turn the steering wheel just enough to restore the ornament-edge distance perceived by CSx to its reference standard; i.e., to correct the disturbance.
This is not a one-shot action, of course; on the contrary, it is a circular negative-feedback process that iterates steadily, smoothly, as the car negotiates the curve and, indeed, until the motorist reaches his destination. Moreover, it is one of perhaps hundreds of parallel processes, voluntary and involuntary, simultaneously going on in that motorist’s nervous system.* The next clip illustrates that phenomenon.
The CSes involved in Mr. Percy Neblett's creating a spoked wood wheel (Cloak 1968/2002) might serve as an example of empirically researching a PCT hierarchy from the Program level down to the Intensity level. So might Don Crabtree's (1966) meticulous attempt to reconstruct the steps involved in producing a Folsom projectile point, if interpreted in hPCT terms.
Reference Standards. The level of a CS is determined by the perception it controls; in other words, by its reference standards. For an Intensity-level CS, the reference standard could be as simple as the average firing-frequency of a bundle of neurons. For a mid-level or higher CS, it can be a much more elaborate representation. It suffices to say that such representations are somehow encoded in the nervous system, along with the mechanism(s) for comparing them with current input perceptions.* hPCT simply shows how these capabilities are organized to yield goal-directed actions. [To anticipate, a meme is a reference standard, of a CS at any level in the hierarchy,* which has been acquired by observational learning or through verbal tuition.]
This next clip shows an example of how the perceptual control model accounts for an accomplishment as simple as focusing on one item in the immediate environment, in support of a program of acquiring that item to eat. It also shows that a PCT hierarchy can be more complicated and interesting than the linear model presented so far.
To summarize our presentation of hPCT so far:
Part Two. Evolution of the Input Function, from Object Recognition to Culture Acquisition: The Origin of the Meme
We've said that Input Functions are, in effect, Inference Engines. So far the inferences we've discussed have been empirical, such as a scientist or other careful objective observer might make.
For example, from a set of perceptions from 2nd Level (Sensation) modules, a 3rd Level (Configuration) Input Function infers the presence of a bowl of fruit; or an Event Level Input Function infers that a dog is chasing a cat, from Transition Level perceptual inferences of a moving dog, a moving cat, and so forth. These empirical inference engines no doubt evolved from simpler neural networks half a billion years ago, and have been refined by natural selection ever since.
A similar example would be a prey animal (e.g. an antelope) inferring that a predator (e.g. a lioness) is nearby.
But successful inferences may be intuitive. As a very simple and early example, the antelope might infer/intuit whether or not the lioness has detected him.
Continuing in that vein, the next evolutionary stage might have been such that he infers whether or not she intends to attack: From input perceptions of her demeanor, he is actually inferring the reference standard to which she is controlling her perceptions.*
It would be interesting to survey the literature of ethology to learn what species are capable of making inferences at each of these stages, and trying to reconstruct their phylogeny. Certainly we can say that the ability to guess, with increasing degrees of accuracy, what the other fellow was trying to do, what he was "up to", has had great survival value.
Besides predator/prey relations, such ability has been especially useful in interactions with conspecifics* involving courtship, sex, fighting, dominance, and other social activities. We can also say that it was a long time coming, and that it has been around for several dozen million years.
As it evolved, reference standard inference in the social realm eventually required and enabled the recognition of individuals, empathy, "theory of mind", and the emergence of the sense of self, beginning perhaps 10 million years ago. Along about then, and more to the point of our discussion here, the capability of imitation evolved.
Imitation entails perceiving the actions of another, inferring the reference standards being controlled to, recognizing those standards in one's own repertory, and evoking them to control one's own perceptions. Recent research on “mirror ‘neurons’” suggests how imitation works all the way down to the level of proprioception; that is, to reference standards of skeletal positions, gestures, gaits, grips, and so forth. *
About two million years ago, there began the evolution of genes modifying input functions and the CS storage apparatus to enable our ancestors not only to recognize each other's reference standards but actually to adopt them, if they were not already in repertory. A reference standard so acquired/adopted is the very definition of a meme.
Since the ability to store inferences from perceptions was present long before, it's rather puzzling that the emergence of culture acquisition took so long. The ability to store a perceptual inference of a satisfactory behavior of one's own as a reference standard, during trial-and-error learning, goes way back in our phylogeny.* Apparently storing a perceptual inference of a behavior of another animal as one's own reference standard is a very different matter, but that's the essence of culture acquisition, as we shall see.
Finally, and parenthetically, the language instinct may well have evolved as a spectacular elaboration of the culture acquisition machinery.
Individuals acquire culture largely through observational learning. Descriptively, that means simply that an observer animal O observes the behavior of a demonstrator animal D and later exhibits similar behavior, of which he has never before been capable.
Without PCT, one is hard put to understand how culture-acquisition works, because it's difficult to conceptualize how O can copy an entire behavior of D, especially its motor elements, into his own nervous system. This difficulty is exacerbated because O may simultaneously modify the behavior to suit his own capabilities and circumstances.
PCT nicely provides the mechanism for culture-acquisition, as per this definition:
Observing the behavior of a demonstrator animal D, an observer animal O infers the reference standards to which D is controlling her perceptions, and stores the inferred standards in the memory of one or more (perhaps new) CS Modules in his own brain.*
Because they are so acquired, and because they are new to O’s repertory, these reference standards are elements of culture; i.e., memes. Then O's newly acquired or augmented CS Modules control their input perceptions by using existing, and/or other newly-acquired, lower-level Modules.
To the extent that O's movements now mimic D's, it is because they now have a common modular CS hierarchy. In the vernacular, O observes D's behavior and figures out what she is trying to do. Then, if necessary, he works out the means to that goal for himself.*
Recall that the above discussion assumes that O does not already possess all of the reference standards in question; if he does, he is not acquiring culture from D. When he controls his perception to those standards, he is simply imitating her.
To illustrate this formulation, here is a clip interpreting a recent experiment (Range, et al., 2007) involving imitation, and perhaps culture acquisition, by dogs:
As I said at the beginning of Part One, CSes are the units of behavioral organization. The hPCT-derived understanding of culture acquisition therefore solves, in my opinion, the problem of the elemental units of culture. Control modules or reference standards so acquired are what I once called “cultural instructions” ("I-culture", Cloak 1975) and Richard Dawkins (1976, 1982) named “memes”. In fact, the only difference between a meme and any other control module reference standard is the means by which the carrying organism acquired it.
So far in Part Two, we have shown how Input Functions of Control Modules may have evolved as inference engines. Initially they were simply interpreters of observable conditions. Later they became interpreters of others' intentions, and finally they became acquirers of others' intentions; in other words, culture acquisition mechanisms – meme copiers at every hPCT level, from Intensities to Systems.
Each of these capabilities was built on the preceding ones, and each capability was refined constantly, always by the natural selection of genes. That means simply that genes which made more accurate and otherwise useful perceptual input functions and storage mechanisms thereby enabled themselves to propagate at the expense of competing genes.
As (some) Input Functions have evolved to become culture acquisition engines, genetic selection has clearly favored those which made more faithful copies of the demonstrator's reference standards, for exactly the same reasons that, starting at least 3.6 billion years before, selection has favored mechanisms which made more faithful copies of genes (cf. Dennett 2008). If having a cultural repertory is valuable, then being able to propagate that repertory accurately must also be valuable; so one should not be terribly surprised that humans, particularly the young, can rapidly and accurately acquire cultural features; i.e., memes.*
Part Three. The Evolution of Cultural Features; i.e., of Memes, by means of Natural and Cultural Selection
As meme copying becomes more accurate, selection more powerfully operates on that which is copied; so memes which are better adapted to their environment will succeed at the expense of those less well adapted, where "adapted" refers to the ability to cooperate and compete with other memes in that environment. If a meme makes its carrier better able than his competitors to acquire resources, avoid predation and other dangers, form social bonds, enjoy sexual opportunities, and so forth, then that carrier, and that meme, will differentially succeed.
Memes (and for that matter genes) are subject not only to natural selection but also to cultural selection. If the memes carried by a group have set up a prestige system, for example, then novel memes (or genes) may succeed only to the extent that they enhance, or at least don’t detract from, their carriers’ culturally defined prestige.*
The operative word there is "better". We don't expect perfection from gene selection, far from it, and we shouldn't expect it from meme selection. But selection for the better can, and has, produced some remarkable results: Equipped with a full complement of memes, encompassing every level of the CS hierarchy,* a people can build a whole culture, complete with its characteristic ethos, rituals, social organizations and structures, child-rearing patterns, artifacts, gaits, gestures, and facial expressions.*
Copyright 2015. Ted Cloak. All rights reserved.