The ongoing debate on what constitutes a “healthy” level of alcohol consumption may have gotten more complicated, thanks to the results of a newly published study in BMJ.

Previous research has demonstrated that heavy drinking can contribute to brain degeneration and related conditions such as dementia, while other evidence has shown that small doses of consumption can actually protect the brain against prolonged impairment.

But moderate drinking, which was classified as eight to 12 drinks per week for this study, was linked to cognitive decline.

Researchers at the University of Oxford analyzed responses from 550 government employees in London who were already participating in UK’s Whitehall II study, also known as the Stress and Health study, which asked a variety of questions including ones regarding how much alcohol the participants’ regularly drank. The participants responded to the survey questions every five years since 1985.

In addition to the 30+ years of data, the researchers conducted MRI scans on the participants to see how their reported alcohol habits correlated with effects seen in their brain images. They also periodically performed memory tests on the participants.

As expected, heavier drinkers showed faster decline in language skills and had poor white matter integrity, which helps process thoughts.

But moderate alcohol intake did more damage than expected, and was associated with degeneration and shrinking of the hippocampus, as well as degeneration of the brain’s white matter.

More specifically, 65 percent of those who drank on average between 14 and 21 units of alcohol a week had shrinkage on the right side of the hippocampus, while that figure rose to 77 percent for those who drank 30 or more units a week.

The researchers did note that the observed changes were only statistically significant for the right hippocampus, but not the left, and they are unsure why this is the case.

Contradictory to previous research, abstinence or small amounts of alcohol consumption showed no protective effect on the brain over the course of the 30-year study period.

But, like many similar studies conducted over the years, there are limitations that may not make the results properly representative of a wider population.

The team did account for age, sex, social activity and education, but some other lifestyle factors such as nutrition were not included. All of the participants were ruled out as being “alcohol dependent” but self-reported drinking habits can also be inaccurate.

While additional research needs to be conducted to prove a definitive cause-and-effect, the study does raise the question of what can be considered a healthy or even beneficial level of drinking.