I've always found so-called innate behaviour to be much more puzzling than learned behaviour. I can somewhat fathom how a neural network might learn a complex task through training but how does a gene do it? My research program over the past ten years has basically been to show that just knowing the connections of a neural network is not enough to specify what it does. Details like time scales and synaptic strengths matter crucially. Although the neural circuit of the worm C. elegans has been mapped out for quite some time, we still don't know how the creature functions. Since most of the animal world functions just fine on genetically programmed traits, we must be missing something.
An advanced online paper in Nature (Manoli et al., June 15, 2005) has found that the fruitless (fru) gene in Drosophila is responsible for male courtship behaviour. This gene encodes a set of male-specific transcription factors that are expressed in about 2% of neurons in the central nervous system. Inactivating this gene completely wipes out all male courtship behaviour but seemingly preserves all other functions. Expressing the gene in females induces courtship behaviour. Yes, females with fru will attempt to mate with other females even though they lack the apparatus to do so.
Most interestingly is when the gene is selectively inhibited in specific systems like olfaction. Naive males will typically court all drosophila they encounter, male or female, but they quickly learn to not try to mate with other males. However, when fruitless is inhibited in the olfactory receptor neurons, the flies will persist in courting males. This fact has been played up in the press as evidence of a gay gene but it actually indicates an inability to distinguish between males and females. What is amazing to me is that a single gene (although it does encode for a number of proteins) has a nonmonotonic action. Knock it out everywhere and the fly won't mate; knock it out in a specific location and the fly won't stop mating.
We still have no idea what the gene does but this result seems to imply that male courtship behaviour is pre-programmed into the neural circuitry and is activated by priming a subset of neurons or neural connections. Now we need to do the electrophysiology on the neurons expressing fru and try to untangle this mystery.