I met Columbia neurologist John Krakauer (who has a very interesting brother at the Santa Fe Institute, by the way) at the 2006 SfN meeting in Atlanta after his PLoS Biology paper on how motor learning transfers differently from easy to difficult tasks and vice versa. He's a very smart guy with an interesting background. John has now written a short article for Scientific American entitled: "Why do we like to dance--And move to the beat?". He could be onto something, but I think he's not taking the idea far enough: When I read this I was a little confused as to what is actually stimulating our reward centers, moving ourselves or watching movements of others, but he clarfiied this later in the article with a reference to mirror neurons (which is not what I'm going to discuss now).
Anyway, not only we get a kick out of moving. Controlling one's own movements in space can be shown to be inherently rewarding in many animals, including insects. For instance, the fruit fly Drosophila prefers to fly in flight directions where there is no artifically added "wobble" of its visual environment. This preference is so strong, that we can use it as reward in a conditioning experiment. Visual patterns (say, upright and inverted Ts) denote different flight directions in a cylindrical environment where the fly has been fixed in the center (see video). Whenever the fly is flying towards, for instance, the upright T, the computer makes the environment oscillate around the flight path such that the fly can still see the target and is able to steer away from it. If it flies towards the other, inverted T, no such oscillations occur. After a few minutes of such training, the flies will show a preference for the inverted T, even if there are no more oscillations. Thus, having full behavioral control over its movements in the environment is a strong reward for the fly, rivalling that of primary rewards such as food and water. The tranmitter involved in processing reward in insects is octopamine. Octopamine is also involved in behavior initiation and control. For example, firing octopaminergic neurons in locusts triggers flight behavior. Interestingly, the neurons triggering flight in locusts are homologous to the neurons thought to be involved in reward in honeybees.
Thus, at least on the surface, there is a tight connection between behavioral initiation and control and reward in insects.

In mammals, the transmitter involved in processing reward is dopamine. John writes in his article: "some reward-related areas in the brain are connected with motor areas". He is talking about the striatum which is involved in action selection as well as operant learning in mammals. The dopaminergic neurons are the ones degenerating in Parkinson's disease. Parkinon's patients show the typical tremors and inabilities to initiate and control movements and also have impairments in operant learning. Music also stimulates the striatum, John writes in the article. Thus, also in mammals, the transmitter system involved in movement initiation and control is the same as (and morphologically related to) the system processing rewards.

Is this all coincidence? Superficial similarities, like human faces in clouds found by human brains looking for patterns? Or is there an underlying evolutionary reason, maybe that one of the first control system early ambulatory animals needed was one for behavioral control? The solution to controlling behavior was to invent a teaching signal which was derived from continuously monitoring the animal's output and compared it with its sensory input. Such a comparison yields a difference signal between stimuli generated externally and stimuli generated by self-mtion. If minimizing externally generated stimulation were rewarding, animals would always strive to control their behavior. Reinforcement learning and optimal control theory provide the theoretical background for such simple systems. Could it be that all other reward systems stem from this initial reward system? Or why is motor control so tightly coupled with reward?

Now, in the postgenomic age, I wonder if one could look at what is known about the genes involved in setting up the reward centers in mammals and insects. Maybe there is a simple master gene (or gene cassette) which organizes the system with dopamine in one variant and with octopamine in one variant. Any bioinformaticians out there who'd like to study this?

UPDATE: John has replied via email and said "I agree with you more than you realize: I am attaching a paper we published in J.Neurosci last year which argues very much for your idea." Here's the reference and link to the paper: The Journal of Neuroscience, July 4, 2007 • 27(27):7105–7116: Why Don’t We Move Faster? Parkinson’s Disease, Movement Vigor, and Implicit Motivation. Pietro Mazzoni, Anna Hristova, and John W. Krakauer.