Though the training didn't directly relate to driving itself, but the study found it extremely effective in reducing the crashes involving elders.
"There are no other cognitive training programs, or 'brain games', that have been demonstrated by published, peer-reviewed studies to enhance driving performance," said Jerri Edwards of the University of South Florida, a co-leader of the study.
The results contradict a study of 11,000 people earlier this year, carried out by Adrian Owen at the University of Cambridge and colleagues, which found that brain training didn't help improve cognitive skills outside the game itself.
"Overall, people need to know that not all brain training is equal. Some programs work and some don't," Edwards added.
With an average age of 73, the 908 participants in the latest study were assigned to one of three different computer-training programs or to no training at all.
One program focused on improving reaction speed, another on reasoning skills and the third on memory.
After 10 sessions the participants were tracked for six years to see how many times they had road crashes for which they were personally responsible.
The reaction speed and reasoning skills programs helped reduce accidents by 50 per cent, but the memory training made no difference.
Of the participants with no training, 18 per cent had at least one crash, just slightly ahead of the 16 per cent of memory course participants who had accidents.
By contrast, only 10 per cent of the speed-training group had crashes, and 12 per cent of those on the reasoning course.
"On the road, the brain needs to process a lot of visual information quickly. So the visual speed-of-processing training directly improves brain functions involved in driving safely, making them faster and more accurate," said Steven Aldrich, chief executive of Posit Science, the company in San Francisco, California, that developed the programs.
Other benefits of brain gym are that trained brains were 38 per cent less likely to develop depression up to a year afterwards and 68 per cent of those who took the reaction-speed course retained their increased reaction times at a two-year follow up.
The research is published in the journal of American Geriatrics Society.