An article in LiveScience entitled "Evolution Can't Go Backward" has highlighted an experiment done to determine whether a protein could be "unmutated" (i.e. returned to its original state) and regain its ancestral functions.

The results indicated that this reversibility wasn't possible because of other mutational factors that occurred during the entire time interval rendering a return to a primordial form impossible¹.

This does raise an interesting question and suggests some possible misunderstandings about what this actually means.

In the first place, reversibility implies a "memory" of some type, or at least an orderliness that allows a process to be undone.  This seems implausible since there is no reason why a gene should need to retain information about it's past. While the gene is certainly the product of everything that occurred in its past, any modifications that would have occurred would not have been "layered" with the idea that the changes might need to be backed out.

In order for past information to be usable, it would require the ability to correlate a particular adaptation to a particular set of circumstances against which it first emerged.  This is an analytical process and wouldn't make sense within the genetic context.  

This suggests that evolution is definitely operating in one direction only, which is to move forward from whatever the current circumstances are.  This doesn't mean that the genes are "evolving" with some purpose, but basically it means that only the present circumstances make a difference in determining what the survivability of a gene is into the next generation.  Just like the past can't help, there is no ability to project into the future either.  There is only the present and whether a gene is successfully reproduced.

Therefore even if the environment were to revert back to some previous level, there would be no way for the gene to know that this has occurred before.  It would still be a new selection pressure that needed to be responded to, from which some genes might survive and be passed on. From a genetic perspective there can't be any past or future.  There is only the present.  So if the environment were to change to some previous level, it is most likely that the gene may well "find" new and novel solutions for adaptation rather than reuse previously existing ones².

Once again, it is important to not view the gene as having some intent, or even of problem solving.  There is a very simple principle at work here that simply requires that there be fewer steps to move forward than to attempt a reversal.  In short, we can assume that unless a gene is adapted in very recent history there will always be fewer steps to adapt from a known working point, than to revert back to an adaptation that was successful in the past.  If a reversal seemed to occur, I would suggest that some of the older genetic forms were still present in the population and were probably simply re-emerging rather than that an a gene reversal had actually occurred.

Using the example from the article, it would be easier to adapt whale's metabolisms and breath-holding using lungs than to envision some reversal of processes to "redevelop" gills³. In other words, the genes wouldn't have "known" that they had a gill design that might work, they could only move forward by the process of an animal with lungs adapting to water.  However, conserved processes such as the growth of nerves, blood vessels and even the internal bone structure would not have required any significant adaptive changes, since they already set to respond to whatever circumstances they are activated in.

Similarly even if we took scenarios like those in Jurassic Park and recovered DNA from dinosaurs, they would simply represent genes that may or may not still be functional in today's environment.  Whatever course those animals took for reproduction, it would NOT be a reversal of evolutionary processes.  It would simply introduce an older form into new circumstances.

This is the distinction between reversibility and simply dealing with older forms.  The older forms would not have evolved, whereas the idea of reversibility is to take an adapted gene and attempting to reverse the process to some older form.  

-------------------------------------------------------------------------------------------------------------------
¹ Of course, it also suggests that the reversing process was incomplete and something was actually overlooked, so that the recovered gene was not truly its ancestral form.

² This should not be confused with conserved processes that have a long history of reuse and retain the plasticity necessary to respond successfully to variations.

³ Consider the idea of an animal reverting back to the water.  It would hold its breath, maybe to escape predators or to obtain food.  Each generation that was able to hold its breath longer might be more successful in finding mates (and surviving to reproduce), so that eventually we have an animal that can hold its breath for extended periods.  However, during this process there would never have been some magic moment where the gene would've "realized" that this was an attempt to live in the water, so therefore gills could be a good choice.