Chocoholics take heart—cocoa beans not only make your favorite treat, they may improve memory as well. Researchers led by Scott Small, Columbia University, New York, report in the October 26 Nature Neuroscience that cocoa flavanols increase blood volume to an area of the human brain called the dentate gyrus, and improve recall of the memories stored there. The study for the first time causally links age-related changes in the dentate gyrus with deterioration of memory, Small said. However, don’t reach for the Godiva yet. The doses used in this study far exceed those in a box of chocolates. Mainstream media from The New York Times to The Los Angeles Times featured these findings.

Scientists have been trying to figure out why memories deteriorate as people age. Previous findings suggested that the dentate gyrus was at least partly to blame (see Small et al., 2002Yassa et al., 2011). However, those studies only correlated changes in the dentate gyrus with memory impairment; they did not prove causation. Small reasoned that if he improved dentate gyrus function and that strengthened cognitive performance, then he would have proof.

Researchers had already demonstrated this in mice. In 2007, scientists at Fred Gage's lab at the Salk Institute for Biological Studies in La Jolla, California, reported that epicatechin—a flavanol found in cocoa—raised the density of blood capillaries and dendritic spines in the mouse dentate gyrus. It also improved their spatial memory, especially when the animals also got a dose of physical exercise (see van Praag et al., 2007).

Measuring changes in the dentate gyrus in conscious humans, however, is more difficult. Co-first authors Adam Brickman, Usman Khan, and Frank Provenzano first had to reconfigure a magnetic resonance imaging protocol to pick up subtle changes in dentate gyrus function, and improve the sensitivity of a memory test. The researchers refined their methods of analyzing functional MRI data to view a single three-dimensional snapshot of the cerebral blood volume in the hippocampus, rather than looking at it slice by slice. This allowed them to more accurately pinpoint differences between subjects. They applied this analysis to scans from 35 healthy people aged 21-65 and saw that of all the hippocampal regions, the blood volume of the dentate gyrus changed most with age.

Next they modified the Benton Visual Retention test, which they had previously found depended on the dentate gyrus (see Brickman et al., 2011). The new variation, called the ModBent, is more challenging, wiping out a ceiling effect in healthy adults and making it more suitable for repeated testing. Subjects view an image built from sinusoidal curves. After the image is removed, they see two more, from which they select the one that matches the previously studied image. Response times in this test are sensitive enough to detect age-related declines in performance in healthy people, they found. In 149 volunteers aged 21-69, ModBent reaction time averaged 220 milliseconds slower every decade.

With these two tests in hand, the researchers enrolled 37 cognitively healthy, sedentary adults, aged 50-69, for the study.  Once a day for three months, 19 volunteers drank a mixture containing 900 mg of cocoa flavanols and 138 mg epicatechin. Candy-maker and study co-sponsor Mars Inc. supplied the mix, which it prepared for research purposes. Eighteen others got a low-flavanol mix with 10 mg of cocoa flavanols and 2 mg epicatechin. To see if exercise enhanced any flavanol effect, the researchers further subdivided both groups so that about half exercised one hour per day, four days a week, and the rest added no activity. At baseline and at the end of the study, Small and colleagues assessed cerebral blood volume with fMRI and dentate gyrus-dependent memory with the ModBent. The researchers also gave everyone a delayed-retention task, which depended on the entorhinal cortex, to test if the intervention affected other brain regions.

At the end of three months, those on the high-flavanol diet had greater cerebral blood volume in the dentate gyrus as measured by fMRI. In the ModBent, they reacted, on average, 630 milliseconds faster than those on the low-flavanol supplement. This translates to roughly three decades worth of improvement, the authors wrote. Cerebral blood volume did not change in the entorhinal cortex, nor did performance on the EC-specific delayed-retention task, indicating that the improved performance was due specifically to changes in the dentate gyrus.

Surprisingly, exercise had no impact on cerebral blood volume or memory performance. However, neither did it improve respiratory oxygen capacity (VO2 max), which suggests that the intervention itself was ineffective, said Small. Though the exercise regimen used boosts VO2max in younger people (see Pereira et al., 2007), older adults may require more rigorous activity, suggested Small. “I do believe that when exercise is effective in increasing aerobic fitness, it will improve cognition in older subjects,” he told Alzforum.

The authors concluded that the dentate gyrus is one of the brain structures underlying age-related memory decline. Boosting blood volume to this area may stave off memory loss, Small said, though he noted that it would be impractical to achieve the flavanol levels used in this study just by eating chocolate—people would have to scarf around 25 chocolate bars per day. However, if future research reveals that lower doses are effective, then a combination of flavanol sources, such as tea, fruits, or vegetables, could help, said Small. He and his co-authors hypothesize that the flavanols enhance cerebral blood flow by increasing capillary density in the area, as suggested by the results from Gage’s lab.

The study says nothing about how cocoa flavanols might affect people with, or at risk for, Alzheimer’s disease. “I think that these dietary interventions have an effect on the brain, but I’m not sure if it’s going to outstrip the effects of Alzheimer’s,” Small told Alzforum. While he has no plans to test that hypothesis, he said it is a plausible idea that others might investigate.

Small plans to replicate this experiment in a larger cohort to see how long it takes the cognitive benefits to kick in, and to peter out after consumption stops, and whether a lower dose of cocoa flavanols might be equally effective.

“In general, this is a very sophisticated and admirable study,” wrote Greg Cole, University of California, Los Angeles. The ModBent test and precise cerebral blood volume measures allowed them to detect improvements even with a small sample size, he said. Still, Cole cautioned that the study needs to be replicated in a larger cohort. The reaction time in the ModBent worsened in the low-flavanol group, and this may have magnified any flavanol effects in the treatment group, he said.

Orly Lazarov, University of Illinois at Chicago, wondered if cocoa flavanols ultimately benefit through hippocampal neurogenesis, which occurs in the subventricular zone of the dentate gyrus. “Neural stem cells rely on the vasculature,” she said. If the flavonoid increases cerebral blood volume, that boost may be a boon to newborn neurons there, she said. Scientists previously showed that a pattern separation task similar to the ModBent depended on neurogenesis in rodents (see Sahay et al., 2011). Until researchers come up with a way to monitor human neurogenesis in vivo, however, it will be difficult to tell, she said.— Gwyneth Dickey Zakaib



  1. In general, this is a very sophisticated and admirable study. Small and colleagues have developed a functional novel-object-recognition test (ModBent) that maps to a dependence on a specific anatomical compartment, the dentate gyrus of the hippocampus. They show that the ModBent correlates with age-related functional declines quantitatively reflected in reduced blood flow that they detected in precise cerebral blood-volume imaging with fMRI. This has permitted a study with small numbers (roughly 10 per group) that still detected significant improvement in ModBent and dentate gyrus (DG) cerebral blood volume with their three-month cocoa-flavanol intervention. The low vs. high flavanol cocoa packets provide a nice placebo-controlled design for this type of intervention. It is also interesting that their aerobic-exercise intervention failed to produce improvements and that their high-flavanol intervention impacted DG but not the entorhinal cortex and related tasks. The ability to distinguish focal improvements will be critical for development of complementary combinations.

    There are a few caveats.

    1) The sample sizes are very small. The improvement in ModBent reaction time (RT) of the high-flavanol group (in supplementary figure 3a) seems as much due to a rise in the low-flavanol group RT at three-month follow-up as to the drop in the high-flavanol group. If the low-flavanol group had been stable, I am not sure that the high-flavanol group would have shown a significant improvement. On this point, it would be helpful if the investigators could tell us whether the low-flavanol group’s baseline RT significantly differed from the high-flavanol group’s RT at three months. The 630-millisecond (ms) improvement of the high-flavanol group compared to low-flavanol over the three-month follow-up appears to drop to a few hundred ms or so if you compare baseline high vs. follow-up high. So while the study needs to be repeated with a larger sample size to be really convincing, at the very least, it provides a new approach for clinical research on brain aging.

    2) The specific improvement in speed in this novel-object task is nice to see in an elegant human study with normal aging subjects, but this type of age-related decline has to be distinguished from the debilitating declines seen with AD. That said, if the effect involves inhibition of neuroinflammation-induced deficits in neurogenesis, or the increased angiogenesis and spine density that some have described in aging rodents, the intervention might become a component of a nutritional program aimed at helping those developing neurodegenerative diseases of aging and associated clinical problems.

    View all comments by Gregory Cole

Make a Comment

To make a comment you must login or register.


Paper Citations

  1. . Imaging hippocampal function across the human life span: is memory decline normal or not?. Ann Neurol. 2002 Mar;51(3):290-5. PubMed.
  2. . Age-related memory deficits linked to circuit-specific disruptions in the hippocampus. Proc Natl Acad Sci U S A. 2011 May 24;108(21):8873-8. Epub 2011 May 9 PubMed.
  3. . Plant-derived flavanol (-)epicatechin enhances angiogenesis and retention of spatial memory in mice. J Neurosci. 2007 May 30;27(22):5869-78. PubMed.
  4. . Hippocampal subregions differentially associate with standardized memory tests. Hippocampus. 2011 Sep;21(9):923-8. Epub 2010 Sep 7 PubMed.
  5. . An in vivo correlate of exercise-induced neurogenesis in the adult dentate gyrus. Proc Natl Acad Sci U S A. 2007 Mar 27;104(13):5638-43. PubMed.
  6. . Increasing adult hippocampal neurogenesis is sufficient to improve pattern separation. Nature. 2011 Apr 28;472(7344):466-70. PubMed.

External Citations

  1. The New York Times
  2. The Los Angeles Times

Further Reading


  1. . Effects of cocoa flavanols on risk factors for cardiovascular disease. Asia Pac J Clin Nutr. 2008;17 Suppl 1:284-7. PubMed.

Primary Papers

  1. . Enhancing dentate gyrus function with dietary flavanols improves cognition in older adults. Nat Neurosci. 2014 Dec;17(12):1798-803. Epub 2014 Oct 26 PubMed.