What pH Does, and Does Not, Do

The question the trade argues about and nobody had measured

Water chemistry is the variable kava preparers argue about most and have the least evidence for. Acidify the water with lemon or citric acid and you will be told the drink comes out stronger; use alkaline water and you will be told the same thing by someone else. Nobody has published a measured comparison on the same material.

White Paper No. 2 closed the accounting on where kavalactones go during a short warm-water extraction, but it did so under a single fixed condition — pH 8.0, 38 °C — and said explicitly that nothing in it identified an optimum. This paper takes the same framework and the same three American-grown cultivars and moves one variable: the initial pH of the extraction water, at 5.0, 6.0 and 8.0, adjusted with USP-grade citric acid. Every fraction was weighed and assayed by the same accredited laboratory. Nine preparations were made in total, one per cultivar per pH.

What this paper is — One run per condition is not a study that can prove anything, and this paper does not claim to. It is a version 1.0a exploratory release. What it can do is put numbers where there were none, and it produces one clean negative result and one modest positive one — plus a correction to a claim the authors' own earlier draft of this dataset had made.

pH is the natural first variable to move, for three reasons. It is the one preparers change most readily and most cheaply — a squeeze of lemon, a spoon of citric acid, a jug of alkaline water. It has a plausible physicochemical mechanism, since the kavalactones are α-pyrones whose enol and lactone functionality could in principle respond to protonation state, and since matrix components that govern particle behaviour — pectins, starch, cell-wall polysaccharides — are strongly pH-sensitive in ways that would change how much solid crosses a strainer. And it is the subject of confident, contradictory and entirely unevidenced claims in a trade that now serves a large and rapidly broadening consumer base.

The published literature offers little help. Reported aqueous kavalactone recoveries span from under 5% to over 50%, and the classic Fijian work of Duve and Prasad reported figures from 22–29% to 81–83% depending on conditions. Much of that variance is now understood to reflect whether particle-associated material was filtered out before analysis rather than any genuine difference in extraction. Aqueous solubility data for the individual kavalactones exist and vary by more than an order of magnitude between compounds, which is the basis for expecting differential rather than uniform behaviour. But we are not aware of a published study that varies extraction pH on a common material and measures the outcome by full mass balance.

How it was done

Three cultivars — Kali-ISA, Kali-Rogu and Kali-Hiwa — were grown in the American Kava Association research greenhouse in Sacramento, California, and harvested at three years of age. Roots and rhizomes were washed, homogenized on a Robot Coupe CL60 with a grating disc, flash-frozen at −20 °C and stored until use. Botanical identity of all three was confirmed by HPTLC against authenticated Piper methysticum references and reported in White Paper No. 2; no new authentication work was performed for this study and none is claimed.

Nine preparations were made: three cultivars × three initial water pH values. Each cell of that design is a single run. There are no replicates at any condition. This is the governing limitation of the paper and it is not recoverable by analysis. A single run cannot separate a pH effect from run-to-run process variation, and No. 2 measured that process variation directly: coefficients of variation on marc dry weight of 4.9% for Kali-ISA, 4.6% for Kali-Rogu and 20.2% for Kali-Hiwa. Differences of the size reported here are comfortably inside the Kali-Hiwa process-variation envelope and only marginally outside the Kali-ISA one.

Filtered municipal tap water from Rancho Cordova, California (native pH 8.0) was used throughout, warmed to 38 °C. For the pH 6.0 and pH 5.0 conditions the water was adjusted with USP-grade citric acid before the plant material was added, and pH was measured on each batch with an Apera PH20 meter calibrated against NIST-traceable buffers.

Two things to be plain about — pH was set on the water, not controlled thereafter. The plant material has its own buffering capacity, so the pH of the slurry during maceration was not the nominal figure and was not recorded. The conditions are therefore initial water pH — which is what a preparer actually controls — and not extraction pH. And citric acid is not an inert proton source. It is a chelator and a polar co-solute, so any difference attributed here to acidity could in part be an effect of citrate. Separating the two would require a mineral-acid comparator, which was not run.

The extraction protocol followed No. 2. Frozen homogenate was thawed at 23 °C until fully malleable, combined with pH-adjusted water at 38 °C, macerated in a Vitamix 5200 on high for 30 seconds, transferred to a 150 µm nylon straining bag and kneaded by hand until flow slowed to roughly one drop per second. The cycle was repeated three times with fresh water and the filtrates combined. Kali-ISA and Kali-Rogu were run at 114 g of frozen homogenate to 1000 mL of finished beverage; Kali-Hiwa at 454 g to 4000 mL. The root-to-beverage ratio is equivalent and all per-serving figures are normalized to 125 mL, but the two series are not pooled anywhere in this paper.

Spent root — the marc — was dried at 38 °C for 48 hours to ≤ 3.0% residual moisture, weighed and submitted for assay. All assays were performed by Flora Research Laboratories by HPLC with ultraviolet detection, using the single-laboratory validated procedure of Liu and colleagues, with calibration across eight points from 5 to 250 µg/mL at R² ≥ 0.999. Reported values are the mean of duplicate injections.

Table 1. Starting materials, as reported on the signed Flora Research Laboratories analytical reports. All three are first-material lots. DW₀ is the dry weight of starting material per run; beverage volume is the measured finished volume.

• Kali-ISA (LOT#KI-R) — Total KL (%): 8.98 · Total FK (%): 0.242 · Chemotype: 256431 · DW₀ (g): 24.3 · Bev. (mL): 1000 · FRL job: J25-0930-H
• Kali-Rogu (LOT#KR-R) — Total KL (%): 10.5 · Total FK (%): 0.175 · Chemotype: 423561 · DW₀ (g): 20.6 · Bev. (mL): 1000 · FRL job: J25-0930-H
• Kali-Hiwa (LOT#KH-RM, laterals) — Total KL (%): 10.7 · Total FK (%): 0.234 · Chemotype: 463251 · DW₀ (g): 127.6 · Bev. (mL): 4000 · FRL job: J26-0223-L

The three lots differ substantially in strength and profile. All are beverage grade with low flavokavain content.

A correction, stated up front

The particle/non-particulate split depends entirely on one measured quantity: marc dry mass. That quantity was mis-recorded in the working analysis of this dataset, and the correction changes results, so the audit trail belongs here rather than buried in an appendix.

The working spreadsheet's Kali-Rogu sheet carried marc weights of 9.5, 9.9 and 9.7 g at pH 8, 6 and 5 — identical, to three significant figures, to those on the Kali-ISA sheet, despite the two cultivars having different starting dry weights. Independent weighings of different materials do not agree that way. The bench master weights table records the actual Kali-Rogu marcs as 6.8, 8.1 and 7.0 g; the values on the cultivar sheet were a transcription of the Kali-ISA column. The effect is not cosmetic: Kali-Rogu's extraction efficiency rises from an apparent 78.1 / 80.2 / 76.9% to an actual 84.3 / 83.8 / 83.3%, and its apparent pH-6 optimum disappears into a flat line.

One anomaly remains. Kali-Hiwa records the same marc mass (32.68 g) at pH 8 and at pH 5. Unlike the Kali-Rogu case this one has an independent check, because Kali-Hiwa is the cultivar whose finished beverage was also dried and weighed: marc plus recovered beverage solids can be compared against the starting dry weight. That check closes to −2.9% at pH 8 and −1.1% at pH 5, which is as good as gravimetric closure gets in this work and is not what a stale duplicated figure would produce. We therefore report 32.68 g at both conditions as recorded. The pH 6 run, whose recorded marc is 20% heavier, closes at −9.8% and is the weakest run in the study.

What the correction protects — Extraction efficiency and chemotype are much less exposed to marc mass than the partition estimates are. Chemotype rank strings are computed from relative abundances within a fraction and are essentially unaffected by marc mass — which is why the chemotype result, unchanged by this correction, is the one we are most confident in.

One further note on terminology. The non-particulate fraction (NPF) reported below is a residual, not a measurement: it is what is left after modelled particle-associated kavalactone is subtracted from measured beverage kavalactone, and it therefore absorbs genuinely dissolved kavalactone, colloidal and emulsified material fine enough to pass 150 µm, and accumulated error. Calling it "water-soluble" would be wrong. An earlier working draft of this dataset called it the "water-solubilized fraction"; that name has been retired here, in line with No. 2.

Chemotype: nine runs, nine matches

This is the central result of the paper and the one least vulnerable to its weaknesses. The bulk beverage chemotype matched the raw-root chemotype in every one of the nine preparations, at every pH. Kali-ISA gave 256431 in the root and 256431 in the beverage at pH 8, pH 6 and pH 5. Kali-Rogu gave 423561 throughout. Kali-Hiwa gave 463251 throughout. Not one rank position moved.

An earlier working draft of this dataset reported the opposite — Kali-Rogu shifting from 423561 to 432615 at pH 6 and 436251 at pH 5, and Kali-Hiwa shifting from 462351 to 643251 — and drew from those shifts the conclusion that pH could be used to tune beverage chemotype. Those strings were computed against raw-material values that do not appear on any signed report for these lots: a Kali-Rogu profile totalling 10.185% used for the pH 6 and pH 5 arms, where the signed report for LOT#KR-R gives 10.5% with a materially different distribution, and a Kali-Hiwa profile totalling 2.825%, where the signed report for LOT#KH-RM gives 10.7%. Recomputed against the signed values, the shifts vanish. pH does not tune beverage chemotype, and the claim that it does should not be repeated.

The marc and non-particulate fractions did diverge from the bulk, in the directions No. 2 found. Marcs were enriched in the less mobile compounds. The non-particulate fraction moved the other way: in eight of the nine runs it placed kavain and dihydrokavain in its top two positions regardless of the cultivar's own profile, reproducing under three water conditions what No. 2 observed under one. The single exception is Kali-Rogu at pH 6 — which is also the run whose non-particulate fraction is largest and most exposed to gravimetric uncertainty.

Table 2. Chemotype rank strings by cultivar, fraction and initial water pH. Digits: 1 desmethoxyyangonin, 2 dihydrokavain, 3 yangonin, 4 kavain, 5 dihydromethysticin, 6 methysticin, descending by abundance.

• Kali-ISA — Raw root: 256431 · pH: 8.0 · Bulk beverage: 256431 · Marc: 254631 · Non-particulate fraction: 256431
• Kali-ISA — Raw root: 256431 · pH: 6.0 · Bulk beverage: 256431 · Marc: 254631 · Non-particulate fraction: 256431
• Kali-ISA — Raw root: 256431 · pH: 5.0 · Bulk beverage: 256431 · Marc: 256431 · Non-particulate fraction: 256431
• Kali-Rogu — Raw root: 423561 · pH: 8.0 · Bulk beverage: 423561 · Marc: 423651 · Non-particulate fraction: 243561
• Kali-Rogu — Raw root: 423561 · pH: 6.0 · Bulk beverage: 423561 · Marc: 423651 · Non-particulate fraction: 423561
• Kali-Rogu — Raw root: 423561 · pH: 5.0 · Bulk beverage: 423561 · Marc: 423651 · Non-particulate fraction: 243561
• Kali-Hiwa — Raw root: 463251 · pH: 8.0 · Bulk beverage: 463251 · Marc: 463521 · Non-particulate fraction: 426531
• Kali-Hiwa — Raw root: 463251 · pH: 6.0 · Bulk beverage: 463251 · Marc: 463251 · Non-particulate fraction: 426351
• Kali-Hiwa — Raw root: 463251 · pH: 5.0 · Bulk beverage: 463251 · Marc: 463251 · Non-particulate fraction: 426351

The bulk beverage column reproduces the raw-root column in all nine rows.

Yield: a real effect, on one cultivar out of three

Extraction efficiency by cultivar and initial water pH

Kali-ISA is the only cultivar with a monotonic pH response. Kali-Rogu spans one percentage point across the entire range. The composite bars average over three materials that behaved differently and should not be read as an optimum.

• Kali-ISA — pH 8.0: 82.1 % · pH 6.0: 85.1 % · pH 5.0: 88.3 %
• Kali-Rogu — pH 8.0: 84.3 % · pH 6.0: 83.8 % · pH 5.0: 83.3 %
• Kali-Hiwa — pH 8.0: 78.6 % · pH 6.0: 74.8 % · pH 5.0: 80.3 %
• Composite — pH 8.0: 81.7 % · pH 6.0: 81.2 % · pH 5.0: 84 %

Percentage of the kavalactone mass in the starting material that reached the beverage. Every bar is a single run.

Total extraction efficiency across all nine runs spanned 74.8% to 88.3%. Every run recovered the majority of available kavalactone — consistent with No. 2, and far above the low aqueous recoveries that still circulate as rules of thumb.

pH moved the figure in one cultivar and not in the others. Kali-ISA rose steadily as the water was acidified, from 82.1% at pH 8 to 88.3% at pH 5, a gain of just over six percentage points and the only monotonic response in the set. Kali-Rogu was flat: a total span of one percentage point, with a slight decline toward acidity that is well inside anything this design could resolve. Kali-Hiwa was non-monotonic, giving its lowest figure at pH 6 and its highest at pH 5.

Averaged across the three cultivars the composite means were 81.7%, 81.2% and 84.0% — a total span of 2.8 percentage points, and not monotonic. The spread between cultivars at a single pH runs from 5.7 points at pH 8 to 10.3 points at pH 6. Both the mild pH 6 maximum and the smooth acid-favouring gradient reported in earlier drafts of this dataset are artifacts of erroneous inputs rather than features of the data.

How much of this could be a weighing error? — Reversing Kali-ISA's ordering — making its pH 5 run no better than its pH 8 run — would require the pH 5 marc to have weighed 14.77 g rather than the recorded 9.70 g, an error of 52%. Reversing Kali-Hiwa's would require a 28% error. Reversing Kali-Rogu's nominal ordering would require only 6%, which is to say Kali-Rogu has no resolvable ordering at all: its three runs are indistinguishable, and that is the correct reading of them. The Kali-ISA trend is the one result here robust to any plausible gravimetric mistake.

Table 3. Extraction efficiency, marc retention, dry-matter pass-through and per-serving potency by cultivar and initial water pH. Every cell is a single run.

• Kali-ISA — pH: 8.0 · Marc dry mass (g): 9.50 · Pass-through fₚ (%): 60.9 · Extraction efficiency (%): 82.1 · Marc retention (%): 17.9 · Total KL per 125 mL (mg): 224
• Kali-ISA — pH: 6.0 · Marc dry mass (g): 9.90 · Pass-through fₚ (%): 59.3 · Extraction efficiency (%): 85.1 · Marc retention (%): 14.9 · Total KL per 125 mL (mg): 232
• Kali-ISA — pH: 5.0 · Marc dry mass (g): 9.70 · Pass-through fₚ (%): 60.1 · Extraction efficiency (%): 88.3 · Marc retention (%): 11.7 · Total KL per 125 mL (mg): 241
• Kali-Rogu — pH: 8.0 · Marc dry mass (g): 6.80 · Pass-through fₚ (%): 67.0 · Extraction efficiency (%): 84.3 · Marc retention (%): 15.7 · Total KL per 125 mL (mg): 228
• Kali-Rogu — pH: 6.0 · Marc dry mass (g): 8.10 · Pass-through fₚ (%): 60.7 · Extraction efficiency (%): 83.8 · Marc retention (%): 16.2 · Total KL per 125 mL (mg): 226
• Kali-Rogu — pH: 5.0 · Marc dry mass (g): 7.00 · Pass-through fₚ (%): 66.0 · Extraction efficiency (%): 83.3 · Marc retention (%): 16.7 · Total KL per 125 mL (mg): 225
• Kali-Hiwa — pH: 8.0 · Marc dry mass (g): 32.68 · Pass-through fₚ (%): 74.4 · Extraction efficiency (%): 78.6 · Marc retention (%): 21.4 · Total KL per 125 mL (mg): 334
• Kali-Hiwa — pH: 6.0 · Marc dry mass (g): 39.28 · Pass-through fₚ (%): 69.2 · Extraction efficiency (%): 74.8 · Marc retention (%): 25.2 · Total KL per 125 mL (mg): 317
• Kali-Hiwa — pH: 5.0 · Marc dry mass (g): 32.68 · Pass-through fₚ (%): 74.4 · Extraction efficiency (%): 80.3 · Marc retention (%): 19.7 · Total KL per 125 mL (mg): 341
• Composite — pH: 8.0 · Marc dry mass (g): — · Pass-through fₚ (%): — · Extraction efficiency (%): 81.7 · Marc retention (%): 18.3 · Total KL per 125 mL (mg): —
• Composite — pH: 6.0 · Marc dry mass (g): — · Pass-through fₚ (%): — · Extraction efficiency (%): 81.2 · Marc retention (%): 18.8 · Total KL per 125 mL (mg): —
• Composite — pH: 5.0 · Marc dry mass (g): — · Pass-through fₚ (%): — · Extraction efficiency (%): 84.0 · Marc retention (%): 16.0 · Total KL per 125 mL (mg): —

Marc retention is the mirror image and remains substantial. Between 11.7% and 25.2% of the starting kavalactone stayed in the spent root. Acidification reduced retention in Kali-ISA, moved it marginally the wrong way in Kali-Rogu, and did not act consistently in Kali-Hiwa.

pH does not rearrange the individual kavalactones

Per-analyte extraction efficiencies moved with pH in step with the totals rather than independently. In every one of the nine runs, dihydrokavain was among the two most completely extracted compounds and desmethoxyyangonin was the least — an ordering that tracks published aqueous solubility and that held at pH 5, pH 6 and pH 8 alike. The spread between best- and worst-extracted compound within a run was narrow throughout, from 2.4 to 5.7 points. Nothing in this dataset shows pH selectively mobilizing one kavalactone against another. Whatever pH is doing, it is acting on the whole matrix roughly uniformly.

Table 4. Extraction efficiency by individual kavalactone (%). M methysticin, DHM dihydromethysticin, K kavain, DK dihydrokavain, Y yangonin, DMY desmethoxyyangonin.

• Kali-ISA — pH: 8.0 · M: 82.4 · DHM: 82.4 · K: 81.1 · DK: 81.9 · Y: 83.5 · DMY: 81.1 · Total: 82.1
• Kali-ISA — pH: 6.0 · M: 85.5 · DHM: 85.4 · K: 83.8 · DK: 85.4 · Y: 85.9 · DMY: 82.7 · Total: 85.1
• Kali-ISA — pH: 5.0 · M: 87.7 · DHM: 88.5 · K: 87.8 · DK: 89.6 · Y: 87.6 · DMY: 85.1 · Total: 88.3
• Kali-Rogu — pH: 8.0 · M: 83.5 · DHM: 84.1 · K: 84.0 · DK: 85.6 · Y: 84.5 · DMY: 83.0 · Total: 84.3
• Kali-Rogu — pH: 6.0 · M: 83.0 · DHM: 83.6 · K: 83.4 · DK: 84.8 · Y: 84.5 · DMY: 82.6 · Total: 83.8
• Kali-Rogu — pH: 5.0 · M: 82.8 · DHM: 83.3 · K: 83.0 · DK: 84.5 · Y: 83.8 · DMY: 80.6 · Total: 83.3
• Kali-Hiwa — pH: 8.0 · M: 76.8 · DHM: 78.1 · K: 79.8 · DK: 82.5 · Y: 76.9 · DMY: 77.3 · Total: 78.6
• Kali-Hiwa — pH: 6.0 · M: 74.1 · DHM: 75.2 · K: 74.8 · DK: 77.5 · Y: 73.8 · DMY: 73.3 · Total: 74.8
• Kali-Hiwa — pH: 5.0 · M: 79.0 · DHM: 80.5 · K: 81.0 · DK: 83.3 · Y: 78.7 · DMY: 79.3 · Total: 80.3

What actually reaches the cup

Total kavalactone per 125 mL serving

The practical scale of the argument. Acidifying moved Kali-ISA by 17 mg per serving and Kali-Rogu by 3 mg. The gap between cultivars is roughly a hundred milligrams.

• Kali-ISA — pH 8.0: 224 mg · pH 6.0: 232 mg · pH 5.0: 241 mg
• Kali-Rogu — pH 8.0: 228 mg · pH 6.0: 226 mg · pH 5.0: 225 mg
• Kali-Hiwa — pH 8.0: 334 mg · pH 6.0: 317 mg · pH 5.0: 341 mg

All values normalized to a 125 mL serving. Single run per bar.

A gain of just over six percentage points from alkaline to acidified water in Kali-ISA is worth having — on that lot it is the difference between 224 mg and 241 mg in a 125 mL serving — but it is not the difference between a weak drink and a strong one. Kali-Rogu, prepared identically, returned 228, 226 and 225 mg across the entire pH range: three numbers that are for practical purposes the same number. A preparer choosing between cultivars is making a far larger decision than a preparer choosing between waters.

Why would one cultivar respond and another not at all? The honest answer is that this dataset cannot say, and that the plausible mechanism is not about the kavalactones. The per-analyte results show pH acting on the whole matrix roughly uniformly rather than selectively mobilizing particular compounds — in Kali-Rogu, five of the six analytes shift by no more than 1.1 percentage points across the entire pH range. That points away from a direct effect on kavalactone solubility and toward an effect on the plant matrix: how readily cell-wall material fragments, how fine solids aggregate, how much passes a 150 µm mesh. Cultivars differ in starch content, root-to-rhizome ratio and fibre structure, so a matrix-mediated pH effect would reasonably be present in some materials and absent in others. That is a hypothesis this study generates and cannot test.

Phase partitioning: pH moves it, but not out of the suspension regime

Non-particulate fraction of delivered kavalactone

Cultivar dominates pH by a wide margin. The largest within-cultivar swing is 7.0 points; the gap between Kali-Hiwa and Kali-ISA at a common pH reaches 24 points. At every condition tested, the majority of delivered kavalactone remained particle-associated.

• Kali-ISA — pH 8.0: 25.8 % · pH 6.0: 30.4 % · pH 5.0: 31.9 %
• Kali-Rogu — pH 8.0: 20.6 % · pH 6.0: 27.6 % · pH 5.0: 20.8 %
• Kali-Hiwa — pH 8.0: 5.4 % · pH 6.0: 7.5 % · pH 5.0: 7.4 %

NPF is a modelled residual, not a direct measurement. See the sensitivity band in Table 6.

The direction of the pH effect is what the matrix chemistry would predict. Acidification of a plant slurry tends to increase pectin solubilization and reduce the aggregation of fine solids, both of which would push more material into the non-particulate residual. But only Kali-ISA showed a clean monotonic response, and the effect measured is in any case a shift within a suspension rather than a transition out of one. At every pH tested, in every cultivar, the majority of delivered kavalactone remained particle-associated — 68.1% at the lowest, 94.6% at the highest.

The practical consequence — Acidifying the water does not turn kava into a solution. Any method that filters or centrifuges a kava beverage and assays the clear phase will substantially understate delivered dose, and acidifying the preparation does not rescue it. The conclusion of No. 2 — that mesh aperture and grind, not water chemistry, are the dominant controls on beverage strength — is unchanged by this dataset and is reinforced by it.

Table 5. Partitioning of delivered kavalactone between the particle-associated fraction (KLₚ) and the non-particulate residual (NPF), as a percentage of the beverage total. Single run per cell.

• Kali-ISA — pH: 8.0 · KLₚ (%): 74.2 · NPF (%): 25.8 · NPF, M: 26.0 · NPF, K: 24.9 · NPF, DK: 25.6 · NPF, Y: 27.1
• Kali-ISA — pH: 6.0 · KLₚ (%): 69.6 · NPF (%): 30.4 · NPF, M: 30.7 · NPF, K: 29.3 · NPF, DK: 30.6 · NPF, Y: 31.0
• Kali-ISA — pH: 5.0 · KLₚ (%): 68.1 · NPF (%): 31.9 · NPF, M: 31.5 · NPF, K: 31.6 · NPF, DK: 32.9 · NPF, Y: 31.4
• Kali-Rogu — pH: 8.0 · KLₚ (%): 79.4 · NPF (%): 20.6 · NPF, M: 19.8 · NPF, K: 20.2 · NPF, DK: 21.7 · NPF, Y: 20.7
• Kali-Rogu — pH: 6.0 · KLₚ (%): 72.4 · NPF (%): 27.6 · NPF, M: 26.9 · NPF, K: 27.2 · NPF, DK: 28.4 · NPF, Y: 28.2
• Kali-Rogu — pH: 5.0 · KLₚ (%): 79.2 · NPF (%): 20.8 · NPF, M: 20.2 · NPF, K: 20.5 · NPF, DK: 21.9 · NPF, Y: 21.2
• Kali-Hiwa — pH: 8.0 · KLₚ (%): 94.6 · NPF (%): 5.4 · NPF, M: 3.1 · NPF, K: 6.8 · NPF, DK: 9.8 · NPF, Y: 3.3
• Kali-Hiwa — pH: 6.0 · KLₚ (%): 92.5 · NPF (%): 7.5 · NPF, M: 6.6 · NPF, K: 7.5 · NPF, DK: 10.7 · NPF, Y: 6.2
• Kali-Hiwa — pH: 5.0 · KLₚ (%): 92.6 · NPF (%): 7.4 · NPF, M: 5.9 · NPF, K: 8.2 · NPF, DK: 10.7 · NPF, Y: 5.5

How much of the partition result survives a weighing error

Marc mass is the sole driver of the split, and one error in the marc record has already been found and corrected. Table 6 therefore recomputes the non-particulate fraction with marc mass scaled to 85% and 115% of the recorded value, so the partition estimates can be read with an explicit tolerance rather than on trust.

Table 6. Sensitivity of the non-particulate fraction (%) and extraction efficiency (%) to a ±15% error in recorded marc dry mass.

• Kali-ISA — pH: 8.0 · NPF, marc −15%: 21.3 · NPF, as recorded: 25.8 · NPF, marc +15%: 30.7 · Eff., marc −15%: 84.8 · Eff., marc +15%: 79.4
• Kali-ISA — pH: 6.0 · NPF, marc −15%: 25.2 · NPF, as recorded: 30.4 · NPF, marc +15%: 35.9 · Eff., marc −15%: 87.3 · Eff., marc +15%: 82.9
• Kali-ISA — pH: 5.0 · NPF, marc −15%: 26.6 · NPF, as recorded: 31.9 · NPF, marc +15%: 37.5 · Eff., marc −15%: 90.0 · Eff., marc +15%: 86.5
• Kali-Rogu — pH: 8.0 · NPF, marc −15%: 17.0 · NPF, as recorded: 20.6 · NPF, marc +15%: 24.3 · Eff., marc −15%: 86.7 · Eff., marc +15%: 82.0
• Kali-Rogu — pH: 6.0 · NPF, marc −15%: 22.8 · NPF, as recorded: 27.6 · NPF, marc +15%: 32.7 · Eff., marc −15%: 86.3 · Eff., marc +15%: 81.4
• Kali-Rogu — pH: 5.0 · NPF, marc −15%: 17.1 · NPF, as recorded: 20.8 · NPF, marc +15%: 24.6 · Eff., marc −15%: 85.8 · Eff., marc +15%: 80.8
• Kali-Hiwa — pH: 8.0 · NPF, marc −15%: 4.4 · NPF, as recorded: 5.4 · NPF, marc +15%: 6.5 · Eff., marc −15%: 81.8 · Eff., marc +15%: 75.4
• Kali-Hiwa — pH: 6.0 · NPF, marc −15%: 6.0 · NPF, as recorded: 7.5 · NPF, marc +15%: 9.1 · Eff., marc −15%: 78.6 · Eff., marc +15%: 71.0
• Kali-Hiwa — pH: 5.0 · NPF, marc −15%: 6.1 · NPF, as recorded: 7.4 · NPF, marc +15%: 8.8 · Eff., marc −15%: 83.3 · Eff., marc +15%: 77.4

Three things survive the band: Kali-Hiwa is the lowest cultivar at every scaling; the within-cultivar direction of the pH response is unchanged at every scaling; and particle-associated delivery dominates across the entire band. What does not survive is any specific percentage. The partition columns in Table 5 should be quoted with this band, or not quoted at all.

Does the mass-balance estimator still work when the water is acidified?

White Paper No. 2 validated the estimator against direct HPLC of the finished beverage at pH 8.0 only, and noted that as a limitation. Kali-Hiwa was assayed directly at all three pH conditions here, which tests the estimator outside the condition it was built on. It holds. Calculated and measured total kavalactone per 125 mL agreed to a mean absolute 2.3% (SD 0.9%, n = 3), sitting comfortably inside the 2.8% reported across ten runs at pH 8 in No. 2.

Table 7. Kali-Hiwa: mass-balance-calculated against directly HPLC-measured total kavalactone per 125 mL, with gravimetric closure, at three initial water pH values.

• 8.0 — Beverage solids (g/125 mL): 3.08 · HPLC measured (mg): 338.2 · Mass balance (mg): 333.6 · Absolute difference (%): 1.4 · Closure gap (%): −2.9
• 6.0 — Beverage solids (g/125 mL): 3.15 · HPLC measured (mg): 307.8 · Mass balance (mg): 317.4 · Absolute difference (%): 3.1 · Closure gap (%): −9.8
• 5.0 — Beverage solids (g/125 mL): 3.01 · HPLC measured (mg): 332.2 · Mass balance (mg): 340.7 · Absolute difference (%): 2.6 · Closure gap (%): −1.1
• Mean — Beverage solids (g/125 mL): — · HPLC measured (mg): — · Mass balance (mg): — · Absolute difference (%): 2.3 ± 0.9 · Closure gap (%): —

Two caveats. The pH 8 run is the run already reported in No. 2, so only two of these three comparisons are new evidence. And gravimetric closure was much poorer at pH 6 (−9.8%); the Kali-Hiwa pH 6 result should be treated as the least reliable run in the study.

What a preparer should actually do with this

These consequences follow from the measurements rather than from any particular interpretation of them, and they are offered at the confidence level of a version 1.0a paper: enough to inform a decision, not enough to settle an argument.

For preparers and kava bars

• Adjusting pH will not change the profile of what you are drinking. Nine out of nine preparations delivered the cultivar's own chemotype regardless of water pH. If you are acidifying in the belief that you are shifting the drink toward kavain, this dataset says you are not.
• Acidification may buy a little yield on some material and nothing on the rest. On one of three cultivars it was worth about six percentage points. On Kali-Rogu it was worth nothing measurable — three servings within 3 mg of each other across the whole range. If you want to know whether it helps with your material, the only way to find out is to test your material.
• It is still a suspension at every pH. Nothing tested here moved the drink out of the particle-dominated regime. Agitation before pouring remains a dose-control step, and grind and mesh remain the strength controls.
• The bag still holds a lot. Between about 12% and 25% of the starting kavalactone stayed in the marc across all nine runs, and acidifying the water is not a reliable way to get it out.

For producers and standards work

• A pH specification is not a potency specification. Fixing extraction pH will not by itself make servings comparable between batches, because the pH effect is smaller than the cultivar and lot effects around it. A specification that fixes pH without also fixing grind, mesh, cycle count and agitation has fixed the least important variable.
• Chemotype claims survive preparation. For a producer selecting on chemotype this is good news and worth having measured: the profile reaching the drinker is the profile of the root, across the pH range a commercial kitchen would use.
• Liquid-phase-only assays remain wrong across the pH range. A method that assays the clear phase of a kava beverage captured between 5% and 35% of delivered kavalactone in these runs. Acidification does not fix that and mildly worsens the variability of it.

What this paper does not establish

It does not establish that pH has no effect on yield; it establishes that on these nine preparations the effect was confined to one cultivar, small where it appeared, and inconsistent in direction elsewhere. It does not establish an optimum pH, and the composite curve should not be used as one. It does not establish that the partition percentages are accurate — they should be read with their sensitivity band. It does not establish that the differences between cultivars are genetic rather than lot-level, since each cultivar is one lot. It does not separate acidity from citrate. It does not address taste, absorption, bioavailability or safety. And with one run per condition, it does not establish that any single number in it would reproduce.

Limitations

• There is one run per condition. This is not a limitation that careful analysis can work around. Nine preparations across nine cells give no estimate of run-to-run variation, and the process variation measured in No. 2 — up to 20.2% CV on marc dry weight for Kali-Hiwa — is large enough to account for most of the differences reported here.
• The gravimetric record needed correction, and the correction mattered. The Kali-Rogu marc weights in the working analysis had been transcribed from the Kali-ISA column. One error of that kind found is a reason to assume others may remain.
• pH was set, not controlled. The reported values are initial water pH. Slurry pH during maceration and final beverage pH were not measured, and plant material buffers. Without those measurements the dose-response axis of this study is only approximately defined.
• Citrate is confounded with acidity. Citric acid was the sole acidulant. It chelates and it is a polar co-solute, so any effect attributed here to pH may be partly a citrate effect.
• Only three pH values were tested, and none between 6.0 and 8.0, none below 5.0 and none above 8.0.
• Two of the three cultivars are different material from White Paper No. 2. This study used first-material lots throughout; No. 2 built its triplicates on second-material lots for Kali-ISA and Kali-Rogu. Absolute values should not be compared across the two papers for those cultivars, and the pH 8 Kali-Hiwa arm is a run already reported in No. 2 rather than new evidence.
• The partition model remains indirect. KLₚ is inferred from gravimetric pass-through rather than measured, and NPF is a residual absorbing all upstream error.
• Nothing here addresses palatability, absorption or effect. pH changes taste, and taste governs what people actually drink. That was not measured, and neither was bioavailability.

Where this goes next

The replicated study needs a minimum of three independent runs per cultivar per pH, prepared in randomized order rather than in cultivar blocks. Marc mass should be recorded in duplicate, entered directly into the analysis sheet rather than transcribed, with the balance reading photographed. Slurry pH and final beverage pH should be measured rather than inferred, and a hydrochloric or phosphoric acid comparator at matched pH would separate proton effects from citrate effects.

Beyond that, direct fractionation — centrifuging the beverage and assaying pellet and supernatant separately at each pH — remains the single most valuable next experiment, exactly as No. 2 concluded. Testing the matrix hypothesis would mean measuring particle size distribution directly.

Conflicts of interest

All three authors are commercially engaged in the kava trade and are owners and operators of kava nurseries and kava businesses in California and Florida, which produce the American cultivars evaluated here. T.B. holds an ownership interest in Root of Happiness, which was the named client on the analytical reports commissioned for this study and which supported the work. J.B. is affiliated with Nakamal at Home and M.M. with Kavafied. The plant material was grown by T.B. in the American Kava Association research greenhouse. This paper is published by the American Kava Association and the authors are members of that organization.

The authors stand to benefit commercially from favourable findings about American-grown kava and about the SWWE protocol. The principal finding of this paper is a negative one that withdraws a commercially attractive claim previously drawn from the same dataset. The analytical data were produced by an independent accredited laboratory from samples submitted without any of the interpretations offered here. No comparison against any competing preparation method or apparatus was performed, and no claim of superiority over any other preparation method or device is made or supported here.

This work received no external grant funding. All materials, analytical services and laboratory costs were borne by the authors, with support from Root of Happiness Kava. T.B. conceived the study, performed the extractions, and prepared this paper; T.B., J.B. and M.M. contributed to conceptualization, and J.B. and M.M. to review and editing. The study involved no human or animal subjects.

Use of generative artificial intelligence — During preparation of this paper the authors used Claude Opus 5 (Anthropic) to recompute the mass balance from the signed laboratory reports, construct the tables, run the sensitivity analysis, and audit the working dataset against source documents — an audit that identified the raw-material, gravimetric and statistical errors corrected here. ChatGPT 5.2 (OpenAI) was used for grammar and language editing. All source values were taken from signed Flora Research Laboratories reports and verified by the authors against those reports. No data were generated, imputed or interpreted by generative artificial intelligence, and no substitute values were created for missing or doubtful measurements. The authors take full responsibility for the content.