Porecelain

Flawless skin starts with the fine print.

Preservatives

Fulton's 1989 table, the source of most comedogenic numbers in circulation, prints nine rows together near its end, and three of them are unambiguously preservatives. All three are parabens, all three read 0 of 5. A fourth row in the same run is printed with its score cells blank and cannot be identified, let alone classified, so no exact count of preservatives in that run is available to us. The one number among the nine that Fulton's own key calls borderline belongs to a sulfur. Two of the five ingredients we file under this family have a row there. The other three have no row we can find, in that paper or in the three other assay inventories we can read line by line.

5 rated 0 Clogging 0 Caution 5 Clear 3 never assayed

What they are

A preservative is defined by what it does, not by what it is. Everything in this family is here because it stops a water-containing product spoiling, and that is a FUNCTIONAL grouping, not a chemical one — which is worth saying out loud, because it is the same kind of category that caused the trouble on our oil hub. The five materials are chemically unlike each other. METHYLPARABEN and PROPYLPARABEN are esters of 4-hydroxybenzoic acid, differing only in the length of the ester chain. PHENOXYETHANOL is a glycol ether — a benzene ring joined by an oxygen to a two-carbon chain ending in an alcohol. SODIUM BENZOATE is the sodium salt of an aromatic acid, and it only works in an acidic formula, because the antimicrobial work is done by undissociated benzoic acid rather than by the benzoate ion. CAPRYLYL GLYCOL is 1,2-octanediol, a short diol whose functions in a formula are not only preservative — it is an emollient and a skin-conditioning agent as well, which is one reason it turns up alongside other preservatives rather than in place of them. Two of the five are the same chemistry. The other three are three different chemistries. Nothing below should be read as a property of the group.

What the evidence actually says

NINE ROWS RUN TOGETHER IN FULTON'S 1989 TABLE, in the order printed below, in the run that follows his vitamin and herb rows. They are the only place in the primary comedogenicity literature we hold where anything is scored under a preservative name at all, and they are where every preservative number in circulation that we were able to trace comes from. We locate them by their place in the run of rows rather than by a section title, and the reason is worth stating before the numbers arrive: our 400dpi transcription of that table was made row by row, without headings, so we hold no heading over these nine. An earlier draft of this page rested its whole argument on one. A heading we have not read cannot carry a page. Here are the nine, as printed, on Fulton's 0 to 5 scale. Methyl paraben 0. Propylparaben 0. Phenoxyethyl paraben 0. Allantoin 0. Hydantoin — no score printed. Sodium hyaluronate 0. Chondroitin sulfate 0. Precipitated sulfur 0. Water-soluble sulfur 3. THREE OF THE NINE ARE UNAMBIGUOUSLY PRESERVATIVES, and all three are parabens: methyl paraben, propylparaben and phenoxyethyl paraben, each printed 0 of 5. A FOURTH ROW CANNOT BE COUNTED EITHER WAY, for the reason set out below — its score cells are empty and we cannot say what material it denotes, which means we cannot say whether it is a preservative. So the honest arithmetic is three and a question mark, not three of nine. Fulton's own key, printed once in his methods and almost never quoted by the charts that reprint his numbers, reads as follows. "As reported in our previous studies, a minimal grade of 0 to 1 is not considered significant. Grade 2 to 3 is borderline. However, a grade of 4 to 5 is uniformally reproduceable and considered positive." So a 0 of 5 is the floor of the scale, and it is the reading printed against all three paraben rows. THE REST OF THE RUN IS NOT PRESERVATIVES, whatever it may have been filed under. Allantoin is a skin-conditioning agent, and we file it as a BOTANICAL. Sodium hyaluronate is a humectant, and we file it as one. Chondroitin sulfate we do not rate at all. The two sulfurs are acne actives. Not one of those five rows is a preservative, and the member table on this page does not contain any of them. That matters for one specific reason. THE ONLY NUMBER IN THIS RUN ABOVE THE FLOOR OF HIS SCALE IS WATER-SOLUBLE SULFUR AT 3 OF 5 — borderline, by his 1989 key — AND IT IS NOT A PRESERVATIVE. A reader who takes a neighbouring row, or a heading, as a verdict on a class ends up carrying a borderline grade that no preservative in the table earned. AND OUR OWN FILING NEEDS SAYING PLAINLY HERE, because this is exactly the junction where a read-across happens. We do not file a material called sulfur. What we file is COLLOIDAL SULFUR, and its printed 3 is not the 3 above. Ours is Fulton 1984's colloidal sulfur row, read under the 1984 key, where 3 or above is positive. The 3 above is Fulton 1989's water-soluble sulfur, and his 1989 key calls a 3 borderline. Different row, different paper, different key, same digit. Fulton 1989 prints no colloidal sulfur row. Sliding that number between the two papers is an error this project has already made once and retracted, which is why the distinction is spelled out at this length rather than left to the reader. HERE IS THE ROW WITH NOTHING IN IT. "Hydantoin" is printed among the nine and its score cells are empty. Our own transcription of Table I, made visually off a 400dpi render, records the cell as blank in print. We cannot tell you what material that row denotes. Hydantoin is the ring that gives DMDM hydantoin its name, and DMDM hydantoin is a common cosmetic preservative; it is also a compound in its own right, and Fulton prints no CAS number, no supplier and no grade for any row in the table. A blank cell is not a zero, and we are not going to convert it into one. It is also the reason the count above stops at three. ONE THING WE CANNOT TELL YOU IS WHETHER A SOURCE-DEPENDENCE MARK HANGS OVER THESE ROWS. Fulton's footnote "results depend on source of raw material" sits on the heading of his oils, "VI. Oils*", where it governs every row beneath it, and it appears against individual rows elsewhere in the table. None of these nine rows carries it. Whether a heading above them does, we do not know: the heading is not something we hold, and an absence in our own notes is a fact about the transcription rather than about the paper. An earlier draft of this page turned that gap into a finding and built a paragraph on it, which is the failure this project keeps having to catch in itself. HIS ROWS AND OUR FAMILY ARE DIFFERENT SETS, in both directions, and this is the second hub on which that has turned out to be the story. Two of our five members have a row among the nine — methylparaben, printed as "Methyl paraben", and propylparaben. Both are the material itself rather than a neighbour or a derivative, and both read 0 of 5. The other three do not appear among those rows or anywhere else in the paper. We looked for a second reading and did not find one. Morris and Kwan's 1983 rabbit-ear assay, an independent laboratory on the same 0 to 5 scale that contradicts Fulton in print on other materials, tested 32 materials — seven lanolins, nine vegetable oils, eleven esters and five surfactants — and no preservative is among them. Fulton's own earlier 1984 paper, whose table we hold in full, carries no preservative row. The fourteen materials named in the abstract of Nguyen, Dang and Maibach's 2007 rabbit-ear study include no preservative. Those are the four inventories we can read line by line, and three of them are silent. For the Kligman-lineage papers and for Lanzet 1986 we do not hold complete material lists, so our statement about those is narrower still — we found no report of a preservative in them. Finally, the near-name, and a correction we owe our own readers. Fulton's third paraben row is PHENOXYETHYL PARABEN, which is not phenoxyethanol. A paraben is an ester of 4-hydroxybenzoic acid; phenoxyethyl paraben is the one whose alcohol component is phenoxyethanol. The names are one word apart, the circulating number for phenoxyethanol is the same 0, and the temptation to declare that the charts misread the row is considerable. Our phenoxyethanol page declined to make that claim, and it was right to: a matching number is not a traced citation. We did not follow the trail from that row to any published chart, and until someone does, "they copied it" is a story that fits the evidence rather than a finding. Two of the pages under this hub — methylparaben and propylparaben — currently state the stronger version, that the 0 following phenoxyethanol is borrowed from the phenoxyethyl paraben row. That is more than we can support, and this hub is the page that can see both at once. What is documented is narrower and still worth the space: the primary table has no phenoxyethanol row, and it has a row for a near-namesake that is a different molecule.

What the evidence doesn’t tell you

What those nine rows are is a fact about a table. It is not a finding about pores, and it does not establish that preservatives are safe for anyone's skin. Three zeros are three zeros in three rabbits. Everything above comes from one 1989 survey whose method was an ingredient diluted to roughly 10% in propylene glycol, 1 ml applied to the inner ear of three New Zealand albino rabbits, five days a week for two weeks, follicles then measured for keratosis with a micrometer. Fulton wrote of the survey that it "is not at all definitive but simply designed to stimulate research", and the warning that travels with this model is his own: not everything that irritates this model will also irritate human skin. A 0 in that model is the absence of an effect in three animals at one concentration, and we identified no human comedogenicity study of any of the five materials in this family. The concentration gap is real in both directions and we are not going to reason across it. The assay ran at roughly 10% in propylene glycol, which is a laboratory dose chosen to provoke a response rather than a formulating level. An earlier draft of this page went further and printed EU caps for three of these materials to two decimal places, with no source entry behind them. We have not opened the regulation that sets those caps, so the figures are gone and the shape of the point is what survives: where a rabbit reading exists at all it was taken at an assay dose, no paper we hold repeats any of these materials at a second concentration, and where no reading exists there is no dose-response curve to read. The assay and the product were never in the same place. A family-level generalisation is exactly the kind of claim this site exists to distrust, INCLUDING WHEN WE ARE THE ONES MAKING IT, and this page is making one. So here is what it does not cover. It does not cover the three members with no row — phenoxyethanol, caprylyl glycol and sodium benzoate — beyond recording that we found no reading for them; their ratings are inferences from an absence, which is what an inference from an absence is worth, and untested is not the same as tested and cleared. It does not cover any preservative we do not rate, and there are many. It does not license reading two paraben zeros across to a third paraben, let alone to a glycol ether or a benzoate salt. And it does not establish a direction: three readings at the floor of one scale, from one laboratory, in one paper, with no replication, is not a property of a class. What each of these materials DOES have is human skin data, and every piece of it measures something other than a comedone. Methyl- and propylparaben were patch-tested at up to 20% in petrolatum in 397 volunteers, with contact sensitisation as the endpoint. Phenoxyethanol appears in patch-test surveillance across Germany, Switzerland and Austria, again for allergy. Sodium benzoate has a 1984 occupational report of transient contact urticaria in three factory workers. FOR CAPRYLYL GLYCOL WE HAVE NOTHING WE HAVE READ. An earlier draft of this page described a Cosmetic Ingredient Review safety dossier for it, covering irritation, sensitisation, penetration and systemic toxicity. No such document is cited below, and we have no record of opening one, so the description is withdrawn rather than sourced after the fact. Allergy is not comedogenicity. Irritation is not comedogenicity. A safety conclusion is not evidence of noncomedogenicity in either direction. These are well-studied materials that have been asked almost every question except this one. And the ceiling over all of it. A 2025 review in JAAD Reviews surveying comedogenicity testing from 1972 onward found that no standardised comedogenicity test exists and that the term "non-comedogenic" is unregulated. There was no settled procedure any of these five could have been put through, even if someone had set out to test them.

All 5 preservatives we rate

The legacy score is printed exactly as it appears in the paper, including the ranges, the asterisks and the parentheses. Those marks are not decoration. They are the paper telling you what its own number is worth.

Sources

  1. Fulton JE. "Comedogenicity and irritancy of commonly used ingredients in skin care products." J Soc Cosmet Chem 40:321–333 (1989) The rabbit-ear survey most circulating comedogenic numbers descend from. Method, in his own methods section, ingredients mixed in propylene glycol at a 9 to 1 dilution, three New Zealand albino rabbits per assay, 1 ml applied daily to the inner surface of one ear five days a week for two weeks with the opposite ear as untreated control, follicular keratosis measured macroscopically and microscopically with a micrometer and combined on a 0 to 5 scale. He also scored a second column, surface irritancy, on a parallel 0 to 5 scale, but our transcription holds one score per row, so nothing on this page describes what a second column says. THIS PAGE RESTS ON NINE CONSECUTIVE ROWS AND ON NO HEADING. Our transcription of Table I, made visually at 400dpi, records rows without their section titles — the same gap that caused us to publish an unqualified 4 for cocoa butter — and the two headings we have since read off the scan and entered with their provenance are "VI. Oils*" and "X. Vitamins and herbs". No heading over these nine has been entered that way, so an earlier draft of this page quoted one it could not produce, and the claim is withdrawn until the scan is read and the heading is recorded with its date. What we do hold is the nine row values and their order, including the blank score cell against "Hydantoin", and that they follow the vitamin and herb rows. Fulton called the survey "not at all definitive but simply designed to stimulate research".
  2. Morris WE, Kwan SC. "Use of the rabbit ear model in evaluating the comedogenic potential of cosmetic ingredients." J Soc Cosmet Chem 34:215–225 (1983) An independent laboratory's rabbit-ear assay on the same 0 to 5 scale, six years BEFORE Fulton 1989, and not a republisher of him — it contradicts him in print on several materials. Its 32 test materials are seven lanolins, nine vegetable oils, eleven esters and five surfactants. Cited here for a verified absence rather than for a result: no preservative is among them, so it offers the three paraben zeros neither corroboration nor contradiction. Its oils table is printed rotated and returns nothing to OCR, which is why anyone who scraped this paper by machine lost part of it.
  3. Fulton JE Jr, Pay SR, Fulton JE III. "Comedogenicity of current therapeutic products, cosmetics, and ingredients in the rabbit ear." J Am Acad Dermatol 10(1):96–105 (1984) Fulton's earlier rabbit-ear survey, on a 0 to 5 scale with a DIFFERENT KEY that cannot be read alongside the 1989 one — 1984 reads 1 to 2 as not significant and 3 or above as positive, where 1989 reads 0 to 1 as not significant and 2 to 3 as borderline. We hold its Table I in full. Cited only for an absence: its five ingredient blocks are lanolins, surfactants and detergents, D and C pigments, miscellaneous pigments and miscellaneous ingredients, and it carries no preservative row. It therefore adds no second reading of any material in this family.
  4. Nguyen SH, Dang TP, Maibach HI. "Comedogenicity in rabbit: some cosmetic ingredients/vehicles." Cutan Ocul Toxicol 26(4):287–292 (2007) A later rabbit-ear study on a 0 to 4 scale — a third scale, easy to misread as Fulton's. Only its abstract has been read; the abstract names its fourteen test materials, which are esters, fatty alcohols, hydrocarbons, one butter and one surfactant. Cited for that verified absence only: no preservative is among them.
  5. Cosmetic Ingredient Review Expert Panel. "Final Amended Report on the Safety Assessment of Methylparaben, Ethylparaben, Propylparaben, Isopropylparaben, Butylparaben, Isobutylparaben, and Benzylparaben as Used in Cosmetic Products." Int J Toxicol 27(Suppl 4):1–82 (2008) The industry expert-panel SAFETY dossier for the parabens. WHAT WE READ AND WHAT WE WILL NOT SAY ABOUT IT: we read it as published in the journal, where the article's own title begins "Final Amended Report on the Safety Assessment of". We have not opened a CIR-hosted PDF of it, so we quote no cover-page status line and we make no claim about its status beyond that printed title — this project has three times cited a comment-stage CIR document as though it were settled, and the cheap fix is to say which document was in front of us. Cited only for identity and function — that methylparaben and propylparaben are esters of 4-hydroxybenzoic acid used as preservatives, and that they are among the most widely used cosmetic preservatives. It assesses systemic exposure, irritation and sensitisation. IT COUNTS NO COMEDONES, and a safety conclusion is not evidence of noncomedogenicity in either direction. It is on this page as a source for what these materials are, not for how they behave in a follicle.
  6. "Comedogenicity in cosmeceuticals — a review of clinical relevance, regulatory gaps, and future directions." JAAD Reviews (2025) A recent peer-reviewed review of the whole field, covering the literature from 1972 onward. It is a synthesis of the same legacy papers rather than a new measurement, and independent of Fulton only in authorship. Cited for one field-wide caveat that bears hard on a family whose evidence is three rows in one table: no standardised comedogenicity test exists, and "non-comedogenic" is an unregulated label.

‹ All ingredients