Human behavior is labile. This is true evolutionarily, developmentally, and immediately. Hines and Bishop (1983) showed this at the evolutionary level, demonstrating that the stable solution for an evolutionary game with learning could differ from that for the same game without learning. I went further in 1990, showing that the solution could have qualitatively different dynamics--whereas the population strategy for the game without learning evolved to a fixed point, with learning it would evolve chaotically.

At the developmental level, we are finding that vertebrate neurons make their connections in response to stimulation, rather than having 'pre-canned' connections that are later trimmed back. (See Churchland and Sejnowski, 1994, for a general discussion.) That implies the behavior of an animal reflects its experience and also reflects a process of reaching out to that experience ("intentionality" &emdash; Freeman, 1995).

At the immediate level, we find that the brain interacts with the world in an experimental fashion, involving the same processes that a scientist uses more formally in hypothesis definition and testing.

The human brain does not operate digitally or syntactically, and so does not contain symbolic representations of data. Data are represented in distributed patterns of neuroactivity, which can be measured as local spiking frequencies (Churchland and Sejnowski, 1994). The term `percept' will be used here to designate a pattern of neuroactivity, with raw sensory data being treated as a specific example. Those patterns (forming a `vector coding') are manipulated to produce the large number of objects that the brain can learn. The creation of those objects involves at least three types of neural processing:

1. hebbian or causal learning, where the brain learns causal relationships based on a specific sequence of percept arrivals (Churchland and Sejnowski, 1994). In this case, when two neurons fire in a specific order, a third neuron may fire indicating the related firings have occured. This can be implemented as a vector coding, either through amplification or by parallel processing to become a `causes' relationship.
2. classification, where the brain learns relationships among related percepts without a close time ordering, and produces an indication that an event in a given class has been recognized (Alkon,1995). Here, a single neuron may respond to multiple patterns, creating a generalization. (This is called the `sigma-pi' neuron model.) The indication is signaled by vector coding, with many neurons firing in parallel and might be called the `classifies' relationship.
3. association, where the brain learns relationships among an arbitrary set of percepts, and modulates the signals designating those percepts in synchrony so that their relatedness can be detected by classification neurons (Douglas and Martin, 1995). In the brain, groups of hypothetical associations often compete to classify the afferent signal with one winning and the remainder silent (Yao and Freeman, 1990; Skarda and Freeman, 1987). Doyon et al. (1994, 1995) have shown that this process is a bifurcation cascade, equivalent in both function and performance to a binary search. Interestingly, there is even a distinct `failure to match' signal that can be generated (Yao and Freeman, 1990; Erwin 1993). No one currently understands how the associations are sorted to support this search. This is the `associates' relationship.

There is recent evidence that there are two memory systems involved in this. The first (motor memory) involves cholinergic neurons and is unconscious. This system does not seem to forget. The second (conscious memory) involves glutamergic neurons and is capable of forgetting. Whether behavior is mediated by one or the other system seems genetically programmed.

The implication of these results is that if we expect to find 'genetically programmed' behavior in some area, we must define our hypotheses very carefully. The key indicator is probably the dynamics of the behavior. As shown in my 1990 paper, 'genetically programmed' behavior will usually converge on a fixed point (becoming stereotyped) even when the game being played by the behavior is intrinsically chaotic and learned strategies for the game can be expected to change dynamically. Hence behavior that is initially stereotyped or that becomes stereotyped after critical life experiences is likely to be 'genetically programmed.' If it is cross-cultural in nature, the genes controlling that behavior almost certainly became fixed prior about 30,000 years before the present.

Mostly, sexual behavior is variable. (See Diamond, 1993, for example.) There are, however, some areas where we find stereotyped behavior, sometimes after an initial learning experience. These seem to involve sexual orientation, sexual fidelity, the definition of acceptable sexual partners, the process of sexual maturation, and the response to sexual abuse in childhood or early adolescence. Sexual orientation is a complex subject, but it is interesting that evidence now indicates it has a strong genetic component. The implication is that homosexuality was an element of middle paleolithic cultures. Patterns of sexual fidelity appear to be cross-culturally consistent--serial polygamy (one wife at a time) appears to be the norm, with 10-40% of the children born to the wife not fathered by the current husband (Diamond, 1993). The degree of sexual dimorphism seen in populations of H. sapiens and H. erectus is consistent with this pattern. Earlier hominids appear to have been much more polygamous, based on the very high degree of sexual dimorphism seen in the fossil record. Acceptable sexual partners appear to be defined based on appearence and familiarity. Individuals with whom there was extended contact during childhood appear to be categorized as probable relatives and thus uninteresting sexually. The process of sexual maturation appears to involve neuron growth during adolescence and early adulthood that changes attitudes and behavior in the following areas:

• musky smells cease being offensive and become attractive. Note that beer (which contains some of the same chemicals) becomes attractive at the same time.
• exposure to sexual situations and material ceases being 'yucky' and becomes (highly) interesting.
• social interests and bonding cease being same-sex and include the opposite sex. Note that a significant, although sometimes sublimated, homoerotic element persists in all-male groups involved in hunting and fighting. This was noted in elite troops in WWII, and Jared Diamond (1993) discusses penis sheaths in this context. This can involve rape of low status males by high status males, intercourse between status peers, and sharing of sexual partners by status peers.
• innervation of the sexual areas of the body (pudenda and penis) increases, supporting orgasm. Note that this innervation is constructivist in nature so that portions of the skin stimulated as part of sexual behavior become sexually sensitive.

There is even a stereotypical response to sexual (and sometimes other) abuse prior to readiness for sexual contact. In girls, this type of experience seems causally linked to early maturation. In both sexes, there are behavioral and attitudinal changes including a strong tendency to express the same or related behavior in turn.

The implications are interesting. Many of the `non-normative' sexual behaviors we see appear to be sufficiently stereotyped that they are probably based on genetics. Even more interesting is that some of these behaviors remain latent and only appear in response to specific life experiences. For example sexual abuse of children may be `self-sustaining' in this sense, with most individuals carrying the behavioral geneset. The possibility of latent behavioral genesets becoming expressed in the general population may underlie much of the current American concern about sexuality.

References

Alkon, D., 1995, From memory molecules to Alzheimer's disease. Presented at the Krasnow Institute, December 4, 1995.

Churchland, P., and T. J. Sejnowski, 1994, The Computational Brain, MIT Press, Cambridge, Massachusetts.

Diamond, J. M., 1993, The Third Chimpanzee: The Evolution and Future of the Human Animal, HarperPerennial.

Douglas, R. J. and K. A. C. Martin, 1995, Neuronal networks: vibrations in the memory, Nature 373:561-564.

Doyon, B., B. Cessac, M. Quoy, and M. Samuelides, 1994, On bifurcations and chaos in random neural networks, Acta Biotheoretica, 42:215-225.

Doyon, B., B. Cessac, M. Quoy, and M. Samuelides, 1995, Mean-field equations, bifurcation map and chaos in discrete time, continuous state, random neural networks, Acta Biotheoretica, 43:69-175.

Erwin, H. R., 1990, The dynamics of peer polities, Presented at the Cambridge Conference on Dynamic Modelling and Human Systems, December 10-13, 1990.

Erwin, H. R., 1994, The application of `Katchalsky Networks' to radar pattern recognition. In K. H. Pribram (Ed.), 1994, Origins: Brain and Self-Organization, Hillsdale, New Jersey. LEA, INNS Press, Proceedings of the Second Appalachian Conference on Behavioral Neurodynamics, October 1993.

Freeman, W. J., 1995, Societies of Brains: a Study in the Neuroscience of Love and Hate, Lawrence Erlbaum Associates, Inc.

Hines, W. G. S. and D. T. Bishop, 1983, On learning and the evolutionarily stable strategy, Journal of Applied Probability, 20:689-695

Skarda, C. A. and W. J. Freeman (1987), How brains make chaos in order to make sense of the world. Behavioral and Brain Sciences, 10:161-195.

Yao, Y. and W. J. Freeman (1990), Model of biological pattern recognition with spatially chaotic dynamics. Neural Networks, 3(2):153-170.