The abstract says "non-coffee drinkers demonstrated better memory performance." The results section, describing the same episodic memory test (a modified Rey word list), says "at the baseline visit, performance was similar between CD and NCD." The figure shows a medium sized difference on the total score and nothing in the text explains the gap between the two sentences. That is a real inconsistency, and it sits in the one finding the post built its headline on.

Nothing was proved. This is a cross sectional comparison of 31 self selected coffee drinkers and 31 self selected abstainers on a secondary outcome. The authors write, in the limitations, that the study "may not have been sufficiently powered to detect small to medium effect sizes for outcomes beyond our primary hypothesis." Memory is one of those outcomes. The nondrinkers were also less emotionally reactive and less impulsive at baseline, which is exactly the sort of difference you would expect between people who chose coffee and people who chose not to, before any coffee is involved.

Accurate. Two weeks off all caffeine, including tea and dark chocolate, then three weeks of blinded caffeinated or decaffeinated instant coffee, block randomized and stratified by sex. Withdrawal symptoms (headache, drowsiness) rose on days two and four and had largely settled by day 14. Cravings, interestingly, never budged.

The blinding covers caffeine, not coffee. Every participant in the second phase knew they were back on four cups a day. There is no arm that stayed off coffee, and the never drinkers were only measured on day zero. So any change that appears in both the caffeinated and the decaf groups is consistent with coffee's nonstimulant chemistry, and equally consistent with the return of a ritual people had been missing for two weeks. The paper cannot tell those apart, and does not claim to.

The paper's own subheading is "coffee intake drives targeted shifts in gut microbial populations without changing overall diversity." Alpha diversity did differ between drinkers and never drinkers, but in the drinkers it did not change across the washout or the reintroduction, and beta diversity did not move with reintroduction either. Seven strains moved. Cryptobacterium curtum and two Eggerthella species were what habitual drinkers had more of at baseline, and C. curtum fell steadily during the washout. What bloomed when coffee came back, on day 21, in both arms, was Veillonella. Veillonella parvula rose significantly only in the decaf group.

"Fundamentally reshaped" describes a handful of strains shifting inside an otherwise stable community. And the post has the direction backwards on its named bacteria: Cryptobacterium and Eggerthella are the ones coffee drinkers carry, the ones that drifted toward the nondrinker profile when coffee stopped, and the ones that had not recovered within three weeks of restarting. The species the reintroduction actually boosted was the one the post did not name.

Comparing the start and end of the reintroduction, both groups improved on perceived stress, the Beck depression inventory, and the impulsivity scale. That part is in the paper. The "driving" is not. The link from bacteria to mood is a Sankey diagram of pairwise statistical associations among species, metabolites, and questionnaire scores. It is a map of what correlated with what, in 31 people, and the authors present it as exactly that.

Two things. First, the same scores had also improved during the two weeks off coffee (impulsivity and emotional reactivity both fell during the washout), so "coffee made people calmer" and "a five week study with repeated questionnaires made people report being calmer" are hard to separate. Second, the stress physiology did not move at all. Salivary cortisol, the cortisol awakening response, and the response to the cold pressor test were the same in drinkers and nondrinkers, unchanged by washout, and unchanged by either coffee. The calm is on the questionnaires, not in the saliva.

The caffeinated group improved on the paced auditory serial addition test (attention and vigilance) and reported lower trait anxiety and lower psychological distress at the end of the intervention. The decaf group did not. The decaf group instead reported better sleep, more physical activity, and less emotional reactivity. Caffeine also did something to blood pressure worth noting: two weeks of abstinence lowered systolic and diastolic pressure in the habitual drinkers, and then the caffeinated group's systolic pressure fell again relative to decaf during the reintroduction. That is the AHA's tolerance story from August, observed in miniature.

"Slashed" is doing a lot of work for a within group change in sixteen people over three weeks. The authors describe "distinct benefits based on coffee type," which is a fair reading. Also worth knowing: the attention test had improved after the washout too, which the authors attribute, plainly, to people getting better at the test.

On the word list test, significant post intervention improvement was seen only in the decaf group. On the CANTAB paired associates task, errors fell in the decaf group and not the caffeinated group. The authors suggest "non-caffeine components of coffee may improve learning and memory," and then, in the next clause, that "there is also a chance of learning effect." The task was given at baseline and at the end. People do better on memory tests the second time.

The decaf group is fifteen people. The mechanism (polyphenols, digestive tract, neural pathways) was not tested anywhere in the paper; the metabolomics show that phenolic metabolites rise with either coffee and that individuals vary widely in how they process them. That is a hypothesis the authors offer for future work. The post reported it as the finding. And the decaf group, for what it is worth, also reported sleeping better, which is the oldest known memory intervention.