"A decent proportion of women we see who are having more memory changes than other women will go on to get dementia," said senior study author Francine Grodstein, an associate professor of medicine at Brigham and Women's Hospital and Harvard Medical School.

"If alcohol at moderate levels is helping to prevent some changes in memory today, that most likely will translate 10 years from now into them being less likely to develop dementia," she added.

The findings appear in the Jan. 20 issue of the New England Journal of Medicine.

Dr. Denis A. Evans, author of an accompanying editorial in the journal, noted that Grodstein's work was an observational study, and so deserves caution.

"This is a step down the road, rather than something that leads to absolutely definite conclusions," said Evans, director of the Rush Institute for Healthy Aging in Chicago. "But that doesn't detract from its [the study's] importance in this context."

While experts know that excessive alcohol intake on a regular basis can damage the brain, it has been unclear what the effects of moderate consumption might be. Light drinking has been linked to several positive health results, including a decreased risk of cardiovascular disease. Given that cardiovascular disease and cognitive impairment share risk factors, it makes sense that light drinking might also benefit memory and other aspects of cognitive functioning, Grodstein said.

"There's increasing evidence that the same things that are good or bad for the heart are also good or bad for the brain," she explained.

Grodstein and her colleagues collected information on alcohol consumption from 12,480 women, aged 70 to 81, who participated in the Nurse's Health Study. Moderate consumption was defined as less than 15 grams (one glass) per day of beer, wine or liquor. The study authors then assessed the participants' cognitive functioning.

Women who drank one-half to one glass of alcohol per day had better cognitive scores than teetotalers. Such moderate consumption appeared to produce a 20 percent reduced risk of cognitive impairment. There seemed to be no difference between those who had more than one drink per day and those who abstained. And there didn't appear to be any difference in the effects of different alcohol beverages.

Because the study was an observational one, there are certain caveats. For one thing, the editorial pointed out, people who consume small amounts of alcohol seem to have better health than people who don't drink at all. It's therefore possible that the changes in cognition could be attributable to something other than alcohol intake.

"In general, people who are older who indulge in very light alcohol consumption, less than one drink per day, are probably a little bit healthier than those who don't, on average," Evans said.

There is also the issue of how often the researchers measured cognition -- just twice during the seven-year study. Measuring three or four times would have been better, Evans said.

Grodstein said her group would continue tracking the women to see if the same patterns hold true over time. "That would certainly make us feel more confident, that this wasn't just a one-time thing," she said.

Given the difficulty of testing this type of hypothesis in a randomized, double-blind study, this new information is about as good as researchers are going to get, Grodstein said.

"Alcohol has important individual and societal risks, so you want to be pretty cautious," she added. "But for those people who are comfortable that they can stay at pretty low levels of drinking, there did seem to be some benefits in terms of memory."