The games are designed to address a specific category of brain function such as attention, working memory, or problem solving capabilities. Progress in a particular field is measured and tracked over time based on performance changes in these tasks (typically by finishing tasks quicker, or finishing more of them in a set time.)
There's little doubt that these measurements can track tangible improvements in that particular task-but do they translate to general improvements in brain function? In other words, does your increasing ability to recall random phrases in a game make it more likely you'll get home from the grocery store without forgetting the milk and eggs, or that you'll remember the names of everyone you met at your sister's wedding?
Groups like Lumosity, one of the most well-known sites offering brain-training exercises, can point to a significant amount of scientific research to back up their claims. However, at least in one area, a recent study by researchers at Oregon University suggests that transferring game improvements to general cognitive improvement can't necessarily be generalized. These findings were published in the Journal of Neuroscience.
The team studied inhibitory control through a task known as a "stop signal task." Participants from 18-30 years in age were instructed to push an arrow key as fast as possible upon receiving a "go" signal. During the trials, one-quarter of the participants received a "stop" signal instructing them to stop pressing the arrow key; a control group was given an alternate task unrelated to inhibitory control. Researchers adjusted the difficulty to individual skills, allowing participants to find their own level of training.
Activity changes in the brain were tracked using fMRI (Functional Magnetic Resonance Imaging). The research team found that activity in two areas of the brain known to be associated with inhibitory control was lowered during the exercise but raised just before the inhibition control period. During inhibition control, further changes were shown, but these changes were relatively small.
According to the research team, the training linked the performance improvements to specific cues predicting that inhibitory control was going to be needed-basically, for that specific task under those specific conditions, the brain learned how to better predict when to stop the task. This implies the improvement may not necessarily be generalized. For example, an Olympic sprinter who used this style of training to get a better jump off of the blocks may not see any improvement unless the same cues are available in the training and the Olympic event.
This work doesn't suggest that brain-training programs aren't useful, but it does suggest that transfer into real-world improvements may not always be straightforward.