ZZ DELETE - Kali ISA v1 (superseded)

The question this article is about

It is still commonly said that no kava-drinking cultures existed in Papua. The chemistry says otherwise, and so does the agronomy. Thirty-three laboratory assays of Papua New Guinea and Merauke material, run by one accredited laboratory across eight years, put numbers on eight Papuan cultivars and on Kali ISA, the ISA-derived line now farmed on the American mainland. The material spans the entire chemotype series that Lebot and Lévesque associated with kava domestication, from wild forms at one end to one position short of the kavain-first endpoint at the other.

Two threads run through it. The first is historical: the 254 and 256 chemotypes that define ISA and much of the Papuan material are the same chemotypes that name 105 two-day cultivars in the schedules to the Vanuatu Kava Act. In Vanuatu that chemistry is read as the product of centuries of deliberate cultivar selection. In Papua the identical chemistry is read as evidence that domestication never really happened. Two territories, one chemistry, two incompatible stories. The difference is not in the plants.

The second thread runs from those gardens to a Florida farm, and it is the practical one. Kali ISA descends from ISA material carried out of Papua New Guinea in the 1990s and selected in Hawai'i on one criterion — cold and heat tolerance — with no attention paid to composition. The line now returns some of the lowest flavokavain figures in the dataset: flavokavain-to-kavalactone ratios of 0.017 to 0.025 in the farm lots, five to eight times below the noble group mean of 0.13. Two separate things produced that. Selection fixed the ratio, and it did so before anyone was looking at flavokavains at all. Post-harvest handling — which fraction is lifted, and when — decides what that ratio is multiplied by, and therefore where a given lot lands against an absolute specification. The rest of this article is the evidence for both halves.

The absence that is an absence of survey

Papua New Guinea holds the widest kava germplasm in the Pacific and the only populations where the wild form and the domesticated form share a landscape. It is also the least characterised. Vanuatu germplasm has been described morphologically, chemically and genetically for four decades and the results are written into national law. The most substantial chemical survey of Papua New Guinea cultivars, 47 samples of root, stem and peeling from Lae, Madang and the Keravat germplasm collection, was conducted by the National Agricultural Research Institute and remains an institutional report rather than a journal article. It also predates the general availability of flavokavain assay, so it carries none.

That asymmetry does most of the work in the received account. When a territory has been surveyed cultivar by cultivar for forty years, its dihydrokavain-led kavas become named varieties with schedules and legal categories. When a territory has had one unpublished survey, the same kavas become an undifferentiated mass, and the absence of documentation reads as an absence of practice.

The material assayed here bears on one specific piece of that account. The Papua New Guinea survey reported a mean of 3.82% total kavalactones across 15 root samples, low by Pacific standards, and that figure has circulated as a property of Papuan kava. Madang Short and Madang Tall grown outside Papua New Guinea returned 16.0% and 14.1%. The cultivars conserved their chemotype rank order across two decades and two hemispheres while their potency quadrupled. The low figures describe the material as sampled — its age, handling, plant part and growing conditions — rather than a ceiling on what the cultivars can do.

One garden, two plants, two uses

The clearest single result in the dataset comes from Merauke, in the Indonesian half of New Guinea, where kava is wati and is reserved for ancestral rites, peacemaking and initiation rather than taken as a daily drink. Two cultivars were collected from the same gardens: the wati itself, and MISA, reported as put to a different use.

They are chemically unrelated in the way that matters. The wati was desmethoxyyangonin-dominant with kavain in last place among the six kavalactones, in both plant fractions and on retest. MISA returned 256431, with desmethoxyyangonin its smallest kavalactone. The desmethoxyyangonin-to-kavain ratio was 2.82 in the wati and 0.31 in the MISA.

Two plants, one garden, same soil, same water, same grower, ninefold apart on that ratio and opposite in the rank position of kavain. Nothing about the site explains it. Someone selected and maintained two lines for two purposes and kept them distinct. That is what cultivar selection looks like when it is happening rather than when it is being reconstructed.

The agronomic record from Indonesian Papua points the same way, and it is current rather than residual: five named Marind cultivars and some seventy local names across 32 villages in Kameubun's Göttingen dissertation, and nine cultivation sites across four Merauke sub-districts, propagated exclusively by stem cutting under organised farmer groups, in Suharno and colleagues' survey.

Rai Coast Pink, and a code that was being read backwards

Rai Coast Pink is reported from the Rai Coast of Madang Province as a wild-type kava with a pink stem, used for toothache, fever and initiation rather than as a beverage. The chemotype code circulating with it is 153, usually glossed as meaning high dihydrokavain and dihydromethysticin.

Under the standard numbering that gloss cannot be right. Position 1 is desmethoxyyangonin. The assay confirms the code and refutes the gloss: Rai Coast Pink came back desmethoxyyangonin-dominant at 39.5% of total kavalactones, chemotype 153624, with the highest flavokavain-to-kavalactone ratio in the whole set at 0.063 and a kavain-to-flavokavain-B ratio of 1.60, sitting between the published two-day and wild group means. The plant is what the code says it is. The number had simply been travelling with the wrong caption.

Kavain in last place

One pattern in the dataset was not expected. Across all 36 readings, kavain came last among the six kavalactones in every sample from a plant reserved for ceremonial or medicinal use, and in none of the other 31. The desmethoxyyangonin-to-kavain ratio ran from 2.42 to 5.23 in that group and from 0.28 to 0.56 in everything else, with an empty band between 0.56 and 2.42 and no overlap.

This is worth stating plainly for what it is. The pattern was found in the data after the fact, it rests on five readings from two cultivars, and a clean separation across 36 samples is the kind of result that narrows when the sample grows. It is offered as a hypothesis with a preliminary fit: that where a culture set a plant aside for ritual or medicine, kavain is at the bottom of the profile, and that this is more informative than the leading digits everyone reads first. It is not a validated test, and the companion manuscript says so in its own limitations.

It took work to get to 254

The chemotype series Lebot and Lévesque associated with domestication runs from 521634 at the wild end through 256431 and 265431 to 246531, and ends at 426135 with kavain first. The Papuan material in this dataset covers the whole of it: Rai Coast Pink at 153624 and Merauke wati at 125634 with kavain sixth, the trade chips and summer Kali ISA at 256431, and Madang lateral roots at 245631 with kavain second, one position short of the endpoint.

That is a gradient, and gradients of this kind are not what unmanaged populations look like. Getting from a desmethoxyyangonin-dominant wild form to a kavain-second cultivar means generations of people choosing which plants to cut and replant, on criteria they could taste and feel. The same schedules that name 105 two-day cultivars in Vanuatu record 79 medicinal varieties alongside 12 nobles: 184 of 196 named Vanuatu cultivars are not noble. The dihydrokavain-led chemistry is not a Papuan anomaly to be explained. In Vanuatu it is the majority of the named germplasm, and there nobody describes it as arrested.

What differs is the documentary record. Vanuatu has forty years of cultivar-by-cultivar survey and a legal framework built on it. Papua has one unpublished report, a dissertation, a handful of ethnobotanical papers, and oral accounts.

What the genetics do and do not say

ISA is frequently described as genetically distinct, and it is. Vandenbroucke and colleagues genotyped 103 accessions with two marker systems and recovered 30 genotypes, confirming that the cultivated Pacific kavas are a narrow clonal set. Two accessions stood outside every cluster: Isa and Iwi, both Papua New Guinean, distant from each other and from the Vanuatu and Polynesian groups, with Isa the cultivar placed closest to the Vanuatu wild-type accessions.

Two things follow, and one does not. It follows that Isa is an old and independently domesticated line rather than a recent derivative of Vanuatu material, which is what you would expect at the centre of origin. It follows that Papuan germplasm carries diversity the rest of the Pacific does not. It does not follow that Isa is a wild type: position in a distance matrix is not a use-category assignment, and that same study classifies its cultivars, Isa included, as var. methysticum.

The dataset then supplies the mirror image of the point. Iwi, the cultivar those markers place furthest from Isa, returned chemotype 254631 — the same code as Kali ISA, with a relative composition matching it to within about 3 percentage points at every position. The two most genetically distant Papuan domesticates are chemically near-identical. A chemotype code is not an identity test, and genotype is not a use category. Both statements are demonstrated inside this one dataset.

Selection set the ratio, and it did so by accident

Kali ISA is the practical reason the dataset exists. The line was selected in 2015 on the Hamakua Coast of Hawai'i Island from ISA material introduced from Papua New Guinea in the 1990s, on one criterion: cold and heat tolerance. Not potency, not chemotype, and certainly not flavokavains. The most cold-tolerant individuals were tissue-cultured and the trait held through the first generation grown out in 2017. Plants of the line later survived a week-long freeze reaching 29 °F in Sarasota, Florida, in poor soil and full sun, and the same plant yielded the most potent lateral-root material in the study at 19.9% w/w.

Across seven root-fraction assays spanning three growing seasons, two states, three pre-harvest temperature regimes and a 2.8-fold range of potency, the chemotype never moved: 254631 every time. Chips read 254631 in the January harvest and 256431 in the two summer harvests; stems read 253461. Every one of those variations is a swap in the third and fourth positions. The dihydrokavain–dihydromethysticin pair at the top was invariant in every fraction assayed.

The flavokavain result is the one that matters commercially, and it is the one nobody was aiming at. The farm lots returned kavain-to-flavokavain-B ratios from 8.98 to 18.06, against a noble group mean of 7.31, and flavokavain-to-kavalactone ratios of 0.017 to 0.025 — five to eight times below the noble mean of 0.13. Selection for climate tolerance, carried out with no reference to composition, did not degrade the chemistry; on the flavokavain axis it delivered material that sits below the nobles it is usually compared against. That ratio travels with the line. It is the half of the result that a grower inherits rather than earns.

Handling decides what the ratio gets multiplied by

The ratio is fixed. The absolute figure is not, and the absolute figure is what a specification reads. Flavokavain content w/w is the ratio multiplied by potency, and potency is set after the plant is in the ground — by which fraction is lifted, and when.

Fraction first. Across the Kali ISA material the root fractions carry the lowest flavokavain-to-kavalactone ratios in the set, 0.017 to 0.025 in the farm lots; chips and stems run higher and shift their third and fourth chemotype positions while doing it. Choosing lateral root rather than chips is a post-harvest decision, made at the wash table, and it moves the number that ends up on a certificate.

Timing second, and it is measurable. Lateral roots from one plot gave 7.00% total kavalactones at a 62 °F pre-harvest average, 8.10% at 72 °F, and 6.38% at 80 °F in a week with 113 °F highs. The 72 °F harvest was the best of the three on potency and on flavokavain fraction together. The January harvest, at the coldest average in the series, still produced commercially usable lateral root — which is the finding that matters to anyone trying to grow kava outside the tropics. One plant per date and no replication, so these are three observations rather than a response curve.

Put the two together and the practical position is this. A line carrying a 0.02 ratio will pass an absolute ceiling comfortably at 8% kavalactones and fail it at 20%, without one molecule of flavokavain having been added. The Sarasota lot is exactly that case: it is the single Kali ISA sample over the 0.4% w/w ceiling, at 0.697%, and its ratio of 0.035 is no worse than lots that pass. Handling did not raise its flavokavains. Handling raised everything, and the ceiling reads only the numerator.

The ceiling problem, in six subsamples of one lot

The most useful accident in the dataset is a single lot of traditional unpeeled Papua chips, sampled six times. All six read 256431. Their total kavalactones ranged from 4.18% to 23.1% w/w, a coefficient of variation of 47.2% inside one lot.

Set that against the 0.4% w/w absolute total flavokavain ceiling in the American Kava Quality Standard. Four of the six subsamples failed it and two passed. Meanwhile the flavokavain-to-kavalactone ratio across the same six held between 0.026 and 0.033, a coefficient of variation of 7.6%. The subsample that passed most comfortably returned the highest ratio in the series; the one that failed by the widest margin returned among the lowest.

The reason is arithmetic rather than chemistry. An absolute weight-for-weight ceiling scales with potency, so strong root is harder to qualify than weak root, and a failing lot can be brought into compliance by cutting it with low-potency chips — without removing any flavokavain at all. A drink, though, is dosed by effect and not by mass. At equal kavalactone delivery the 19.9% Sarasota lot that fails the ceiling supplied 8.8 mg of flavokavain per serving, against 8.9 mg for a lot that passes it, from 45% of the mass. Against the parent material it supplied 8.8 mg where Madang Short lateral roots supplied 12.1 mg.

None of that argues for abandoning the ceiling, which is simple, auditable and correct in most cases. It argues for reporting a ratio next to it. A specification of total flavokavains at or below 0.4% w/w or a flavokavain-to-kavalactone ratio at or below 0.05, with flavokavain B reported either way, would pass the Sarasota lot, would still fail all four Papua New Guinea samples above the ceiling, whose ratios run from 0.040 to 0.063, and would fail Rai Coast Pink on both tests. A formal proposal on this point goes to the Kava Coalition framework with the underlying data.

What this does not show

1. No genetic authentication was performed on any sample. Cultivar identity rests on chemotype consistency and on the designations supplied with each collection, and the dataset itself demonstrates that a chemotype code is not an identity test.
2. Several cultivars are represented by single lots. The comparison between Kali ISA and its Madang parent material on flavokavain content is one of those, and it is the finding most exposed to the author's commercial interest.
3. The kavain-last pattern is post hoc and rests on five ceremonial readings from two cultivars.
4. The values here are not method-matched to the historical Papua New Guinea figures, so the potency comparison carries a methodological component that cannot be separated from a material one.
5. The selection history of the Kali ISA line is the account of the growers who hold it, corroborated where the genetic literature records the same route and identified as uncorroborated where it does not.
6. The harvest-timing series is one plant per date on one plot, with no replication, so it establishes that timing moves potency without establishing by how much.
7. The domestication argument is an interpretation of chemotype distribution, not a demonstration. Testing it needs cultivar-level survey in Papua on the scale Vanuatu has had: named cultivars, documented use categories, kavalactones and flavokavains by the same method, and genotyping on the same accessions.

Disclosures

The author has a commercial interest in Kali ISA and in kava retail and kava-bar operations, and holds a position in the American Kava Association. The Merauke material was contributed by John Sanday of Melanesian Trustee Services Limited, who declined payment, asked to be cited, and had no role in the analysis or in the decision to publish. Where his field account presses against the framework used here, it is reported in his terms rather than resolved in the association's favour.