Throughout my adult life, I've run into many people who talk about wanting to learn some new skill: a new language, some topic of study they didn't get to in college, a new instrument, etc. When I've encouraged them to go for it, I often hear the old adage, "You can't teach an old dog new tricks." As a social psychologist with a great deal of training in behaviorism, I know for many things, that just isn't true. It might be more difficult to learn some of these new skills in adulthood, but it's certainly not impossible. In fact, your brain is still developing even into early adulthood. And the amazing concept of brain plasticity (the brain changing itself) means that a variety of cognitive changes, including learning, can continue even past that stage.
A new study in Psychological Science examined training in different skills, comparing individuals from ages 11 to 33, and found that some skills are better learned in adulthood. They included three types of training: numerosity-discrimination (specifically in this study, "the ability to rapidly approximate and compare the number of items within two different sets of colored dots presented on a gray background"), relational reasoning (identifying abstract relationships between items, sort of like "one of these things is not like the other"), and face perception (identifying when two faces presented consecutively on a screen are of the same person or different people). They also measured verbal working memory with a backward digit-span test, which involved presentation of a series of digits that participants had to recall later in reverse order.
Participants completed no more than 1 training session per day over 20 days, through an online training platform. They were told to use an internet-enabled device other than a smartphone (so a computer or tablet). The tasks were adaptive, so that performance on the previous task determined the difficulty of the next task. To compare this training program to one you're probably familiar with, what they created is very similar to Lumosity.
They found that training improved performance on all three tasks (looking at the group overall, controlling for the number of training sessions the participants actually completed). But when they included age as a variable, they found some key differences. The improvements they saw in numerosity-discrimination were due mainly to the results of the adult age group; when adults were excluded, and they only looked at adolescents, the effects became non-significant. The same was also true in relational-reasoning performance. Though there was a trend toward more improvements among adults on face-perception, these differences were not significant. You can take a look at accuracy by skill, testing session, and age group below (asterisks indicate a significant difference):
Another key finding was that there was no evidence of transfer effects - that is, receiving ample training in one task had no impact on performance in a different task. This supports something psychologists have long argued, much to the chagrin of companies that create cognitive ability training programs (ahem, Lumosity): training in a cognitive skill improves your ability in that skill specifically, and doesn't cause any generalized improvement. That's not to say doing puzzles is bad for you - it's great, but it's not going to suddenly improve your overall cognitive abilities.
But the key finding from this study is not only that "old dogs" can learn new tricks, but that for some tricks, older really is better.
EDIT: I accidentally omitted the link to the study abstract. But the good news is, I discovered it's available full-text for free! Link is now included.