Caffeine and memory consolidation


Caffeine is a psychostimulant with many known cognitive-enhancing effects such as increased alertness, arousal, enhanced attention and focus, and reduced distraction. We were curious if it also had an effect on long-term memory storage and forgetting. As avid coffee drinkers (collectively), we thought we would do this the most rigorous way we could. We borrowed a post-training design from animal literature pioneered by Jim McGaugh and conducted a study in humans where we identified an enhancing effect of caffeine on memory consolidation. The effect was specific to consolidation and not retrieval. The paper was published in Nature Neuroscience in January of 2014.

Press releases by Johns Hopkins and UC Irvine. See more on our media page. Johns Hopkins also produced a short video about this work:

The Science behind it:

The task and overall experimental design are shown in the figure below (Fig 1a in the paper). During encoding, subjects saw pictures of every day items and indicated whether each picture was an “indoor” or “outdoor” item. They were then administered either caffeine or placebo. After 24 hours they returned to take a surprise test. During the test they saw completely new items, old items and items that were similar but not identical to the ones they saw the first day (i.e. lures). They were asked to indicate whether the items were “old”, “similar” or “new”. We observed that those who were administered caffeine (200 mg) were better than those were administered placebo on lure discrimination (i.e. calling a lure item “similar” rather than “old). This was quantified several ways. Fig 1b-d show the raw performance data. Differences were detected only in responses to the lures.


However, since raw response proportions are notoriously known to be affected by subject response bias, we calculated a lure discrimination index (LDI) by calculating the difference of probabilities of calling a lure item “similar” and calling a foil item “similar” (to correct for the general tendency to call any item “similar, i.e. response bias). Fig 2a shows a comparison of LDI in the immediate vs. placebo conditions (p<.05 two-tailed). We also ran another condition where subjects received caffeine 1 hr before test and found no difference from placebo (p>.05 two-tailed). The delayed group was also significantly different from the immediate (p<.05 two-tailed). We also ran a 2-way ANOVA with caffeine and similarity and found main effects for both (p<.001 two-tailed) with no interaction. Finally, we ran a dose ranging study (Fig 2c) with placebo, 100, 200, and 300 mg. We found that 100 mg was no different from placebo and 300 mg was no different from 200 (however, at 300 mg subjects started to report side effects like headaches and jitteriness). We also ran a similar analysis but using caffeine concentrations in saliva instead of canonical doses and found evidence for U-shaped curve (Fig 2d).


For questions about this work, contact the Principal Investigator. Access the paper here.