A recent publication in Nature suggests that in many ways, birds are baby dinosaurs. The finding is less unusual than it might seem, afterall it is well established that humans have many traits of baby apes and dogs are in some ways baby wolves. The process is known as paedomorphosis or neoteny - the retention of juvenile traits in the adult form. This can take the form of enlarged eyes (birds), larger brains (humans) or retention of juvenile behaviour (dogs).
The reason why paedomorphosis works is that the basic body plan has much deeper evolutionary roots than the species-specific add-ons. Think of it this way, all mammals pretty much have the genetic program to make a nose, but only the elephant has evolved an additional genetic plan to turn that nose into a trunk. Deep in the genetic code of the elephant there is still the "standard nose" code (and indeed, the foetus has a relatively normal nose), it just has added lines of code that upgrade the standard nose into a trunk. This means that in theory, the elephant could evolve away from the trunk by just ditching the upgrade code, letting it default into the standard nose code. This is true for most of development - new code is never optimally created for the organ, rather it is always adding a bit of extra code to change the outcome. For a software engineer it would be the hight of laziness, creating bloated useless code, with every problem solved by kludge.
Despite being inefficient and inelegant, the system of "generic code" plus "species specific" is very useful for evolution. This is because species evolve to be adapted to a specific environment. The flamingo beak is fantastic for a filter-feeder, but it has lost the generic functions that a sparrow could use its beak for.
Imagine an island with brine lakes that is populated only by flamingoes. If those brine lakes dried up, the flamingoes would go extinct. But what if new niches opened up? The ordinary "forward" process of evolving a generalist beak is quite slow, because you need to generate new code, but the "backwards" process of paedomorphosis could be quite fast, because it is just the process of deleting the species-specific code, defaulting back to the generic beak (as in anything else, destruction is faster and easier than generation). It is not difficult to imagine a relatively small set of genetic deletions that would mean the adult flamingo retained the juvenile generic beak, and then these "de-evolved" generalist birds could take advantage of the new habitat, and indeed start to evolve specific changes to specialise towards that new habitat.
As a general rule, following a large change in the environment, the generalised (juvenile) body plan is probably going to be more successful than the specialised (adult) body plan. Paedomorphosis in effect provides a default option to revert to in case of catastrophic change, allowing a species to shed its specialised features and start again. One possibility that interests me is that an open niche may drive paedomorphosis by selecting for rapid population growth. Consider the drying up of Africa that occured 5 million years ago. All of the apes that were specialised to live in rainforest would have seen dramatic contraction of their habitat, leaving just a few thousand gorillas left today. But the drying also created a new niche, the savanna, which could be exploited by any ape that was able to adapt. Paedomorphosis probably played a role in human evolution, by shedding the arboreal features required to swing in trees, allowing the pre-humans to venture onto the savana. Now consider the first pre-humans that were suitable for the savana - they has a continent to spread across, with the only limitation being the reproduction rate. We already know that a truly open niche creates an evolutionary pressure to fill it - such as the natural selection of cane toads in Australia with longer legs simply because they can move faster into virgin territory. What if this put selection on humans to reproduce at a younger age? Any variants that became fertile younger (and thus, while still carrying juvenile features) would outcompete the others, creating a population shift. In effect, there would be selection for paedomorphosis simply to increase the reproduction rate, with the retention of other juvenile traits (such as a larger brain) being a side-effect.
If this model it correct, it would mean that open niches would drive paedomorphosis via two mechanisms - by selecting for the retention of juvenile traits to give a more generalist body plan, and by selecting for sexual maturity at a younger age to give more rapid reproduction. This dual selection force would drive much more rapid evolution, and may be responsible for some of the most remarkable evolutionary shifts, including the evolution of humans.