Phosphatidylserine Raised Acetylcholine in 12 Months — But Memory Improved While Cortisol Results Were All Over the Map

For years, PS supplements have been sold with broad “brain support” claims. A 12-month randomized, double-blind trial added a clearer clue. In 190 older adults, a PS-containing supplement increased serum acetylcholine versus placebo (β=0.441; 95% CI: 0.415–0.468) [1].

The same trial also reported better short-term memory in the PS group (β=0.600; 95% CI: 0.399–0.800) [1]. This links a measurable biomarker shift with a functional outcome in one long study.

One important nuance: the study measured acetylcholine in blood, not the brain. So the result supports a cholinergic pathway, but it does not prove a direct brain-level acetylcholine increase.

The mixed PS story gets clearer when you look at who was studied and what was given. A 16-week, double-blind, placebo-controlled trial (N=263) found no significant treatment effects of MFGM phospholipid supplementation on cognitive function. The study reported no benefits on RBANS (primary outcome) or COMPASS (secondary outcome) [2].

One practical explanation is responder vs. non-responder biology. If benefits cluster in people with weaker cholinergic signaling, then broad “healthy older adult” samples can wash out real effects.

Another key detail is the ingredient: MFGM is a milk-derived phospholipid mix, not the same thing as a PS-focused product. So a null MFGM result does not automatically cancel out PS-specific findings, but it does show that “phospholipids” as a category are not guaranteed to move cognition [2].

PS studies on cortisol also split in two directions, and formulation may be why. An 8-week double-blind RCT found a curcumin phytosome that contained PS significantly improved serum cortisol versus placebo [3]. That suggests a stress-hormone effect is possible in at least one combined formula.

But a 6-week randomized, double-blind, placebo-controlled trial found bovine milk-derived phospholipids at 2.7 g/day did not significantly attenuate the salivary cortisol stress response versus placebo [4].

These results can differ for three simple reasons: (1) serum vs. salivary testing, (2) PS-in-a-combo vs. milk phospholipid blends, and (3) baseline stress levels. Bottom line: cortisol effects are not consistent across products or studies.

PS research has also produced an unexpected metabolic signal. In a randomized, double-blind, crossover trial (n=38), PS supplementation was linked to a 13.6% reduction in HbA1c (p<0.0085) [5].

That is notable because HbA1c reflects roughly 2–3 months of average blood sugar. If the finding holds up, PS may affect more than cognition and stress.

Still, this evidence is early. It comes from one small crossover trial and needs replication in larger groups, with clear dosing details and baseline metabolic status reported.

Not all “PS” or “phospholipid” supplements are interchangeable. Older research used bovine brain-derived PS. Later products shifted to plant sources like soy or sunflower.

In the newer 12-month trial, researchers used a PS-containing supplement, not isolated PS [1]. In the cortisol-positive trial, PS appeared inside a curcumin phytosome formula [3]. These details matter because they make it hard to claim that standalone PS will deliver the same effects.

PS also interconverts with other membrane phospholipids like phosphatidylcholine and phosphatidylethanolamine [6]. That means PS works inside a larger membrane system. So the surrounding phospholipid profile of a formula may change results, even when the PS number on the label looks similar.

PS looks less like a universal “memory pill” and more like a targeted supplement that may help specific people. A long RCT found a PS-containing supplement increased serum acetylcholine and improved short-term memory. But a large trial of a different phospholipid source (MFGM) found no cognitive benefit. Cortisol findings also depend on the product and test method. If you try PS, the most evidence-backed approach is to track a simple memory measure over time and judge your own response, since non-response appears common.