There have always been statistical outliers, but the average life expectancy dropped precipitously after the Agricultural Revolution, when Neolithic humans began subsisting on grains instead of meat and their civilized descendants became even less carnivorous. Skeletons dug up of the last of the hunter-gatherers show that the life expectancy of an adult who had survived childhood (no mean feat of course) was in the low fifties. By the Roman era it had dropped to the low twenties. Prosperity increased the meat in the diet and by the end of the 19th century life expectancy (of adults) in the USA was back up into the thirties. Antibiotics, asepsis and vaccinations reduced infant mortality, the discovery of vitamins and minerals gave everyone a healthier diet even if they insisted on eating too many vegetables, hospitals allowed more women to survive childbirth and automation allowed men to take jobs that were less life-threatening, so the life expectancy of adults has doubled, and the life expectancy at birth has more than tripled.
That's nice Fraggle, but despite all those advances in health care and medical science, the
maximum life span hasn't changed in recorded history. The
biological and
physioloigcal factors for human lifespan have remained the same, despite changes in environmental factors.
Of course, when it comes to explaining selection, average life span is a much better metric than maxima.
And of course dentistry and fluoride have helped us all keep more of our teeth.
Natural selection only takes place if the trait in question
is actually selected for or against, either by survival or mating practices. No one dies because they have wisdom teeth, and I have to say I have never heard of anyone counting a prospective mate's teeth before deciding to procreate with them.
However, they do that with horses.
Count the teeth, not procreate.
I know you didn't say this, but if selection isn't acting on a trait, evolution can still occur there, and very often does. This is called drift, and is a product of a finite population. Hardy-Weinberg equilibrium exists if there is random mating and an infinite number of alleles to draw from, and no selection. However, neither of the former two assumptions are true outside of the model, especially for the patchy distribution of macro-fauna, like most mammals and birds. Because of this, you can still have changes in gene frequency (evolution) without any selection occurring.
This drift is taken advantage of with genetic fingerprinting, since there are regions of our genome that, as far as we know, aren't being used for anything. These regions used to be called junk DNA because they didn't code for a functional product, but much of that "junk" is actually active in other ways. These genomic regions that don't appear to be doing anything have high mutation rates, which, because the DNA doesn't code for anything or function in any way, don't affect the organism. Since these mutations occur at random, if you look at enough loci where random mutations occur, it is very unlikely that two people will have the exact same mutations.