Every oil recipe here specifies drops. Eight of mandrake, four of frankincense, two of dragon's blood. Five of lavender, five of camphor, five of eucalyptus, five of orange, four of sandalwood, three of rosemary, three of arnica, two of pine.
In a system that argues at length that objects infused with the quality of Five react with the universe as Five would react, those numbers ought to mean something. A grid of seven stones broadcasts seven. Does a recipe of five drops broadcast five?
This page is the census that asks. It was run against the twelve recipes directly, and the answer is more interesting than a yes.
Twelve compound recipes are given: Protection, Merlin's, Ieroventus Pocket's, Oceanus, Dragon's Fog, Healing, Employment, Wizards', Saint Augustine of Canterbury's, Longinus', Mordekai's Vanquishing, and Wizard's Tongue.1 Across them, forty-five drop specifications were legible in this edition's conversion. The counts used are:
1 appears 3 times. 2 appears 5 times. 3 appears 6 times. 4 appears 12 times. 5 appears 12 times. 6 appears twice. 8 appears 3 times. 10 appears twice.
⚠ Two recipes came out of the conversion with fewer entries than they visibly have, so this is a census of what is legible rather than of what is printed. Re-check the plate before treating any total as final.
Now the result, and it is a negative one. The counts do not sort by what the oil is for. Take the twelve fives, which are the largest group along with the fours. Eight of those twelve sit inside just two recipes: four fives in Healing Oil — lavender, camphor, eucalyptus, orange — and four fives in Mordekai's Vanquishing Oil — betony, benzoin, anise, asafetida.
One of those recipes heals and the other vanquishes the forces of darkness. If five were carrying its own quality into the mixture, those two would be the last pair to share it. The same holds for four. Twelve fours, of which four are in Dragon's Fog Oil — vervain, oak, witch hazel and pine, all four at four. That is not four being selected for its properties. That is one recipe written at a single value throughout.
So the counts cluster by recipe rather than by function. A given oil tends to be written in one or two numbers, and the number is a property of how that recipe was composed rather than a code for what it does. Reported as what the census shows, against the expectation the numbers material creates.
One result is clean and it is the most striking thing in the census. Seven does not appear anywhere in the twelve recipes. Neither does nine.
Every other value from one to six is used, plus eight and ten. The two gaps are exactly the two numbers this system gives the most elaborate treatment to elsewhere — seven with its holy name, its eleven-fold list and its seven-pointed star, and nine with its three-times-three and its nine-night candle. ⚠ Read that as an observation, not a doctrine. Forty-five specifications with two missing values is a small sample, and their absence may be nothing but the arithmetic of small recipes. It is recorded because it is checkable and because the two absentees are not random ones.
The recipes are one dataset. The separate list of individual oils and their uses is another, and there the counts behave differently — because each entry is a complete instruction rather than one line of a mixture. Bergamot: five drops in the back of each palm, in the form of a five pointed star, and you will be protected against the forces of darkness. That is a five doing exactly what five does elsewhere — the elemental pentagram, the shield, drawn in oil.
Asafetida: seven drops in the left palm when exorcising evil spirits. Seven for the Sword of Heaven's number, on the expelling hand. Almond: four drops upon the left hand every day for ten days at the fifth hour, with a prayer for money. Three numbers in one instruction — a count, a course and an hour.
Anise: one drop on each temple and then one on the forehead, for clairvoyance. Three ones, placed to make a triangle on a face. ⚠ A full census of that list was not obtainable from this conversion — the entries run together inline and a parse breaks down partway through. The named examples above are legible and are reported individually rather than as a distribution.
Which gives the honest shape of the finding. In the recipes the numbers are quantities. In the individual instructions some of them are clearly meanings. The same numeral is doing different work in the two lists, and nothing anywhere says so.
A practitioner has to decide what to do when a recipe is short an ingredient, and the two readings give opposite answers. If the counts are coded, substituting is impossible. Four drops of sandalwood is four-plus-sandalwood, and three drops is a different operation entirely — the same logic that forbids substituting one metal for another, where identity is absolute and quantity is not.
If the counts are proportions, the recipe is a ratio and could in principle be scaled. Nothing in the material says whether a half-batch is a valid batch.
The census supports the second reading for the recipes and the first for the individual instructions. So: follow the recipe drop for drop because you have no basis for doing otherwise, and take a single-oil instruction's number seriously, because bergamot's five is drawing a pentagram.
One more figure is worth having. Four and five together account for twenty-four of the forty-five specifications — more than half. Add three, and three values cover thirty of forty-five, or two thirds.
That is what a hand-written recipe collection looks like. A composer reaching for a plausible small number lands on three, four or five most of the time, and the larger values — eight and ten — appear only where a single ingredient is meant to dominate: eight of mandrake heading Protection Oil, ten of oak in Merlin's, ten of bergamot as the whole of Oceanus.
That last observation is the useful one for a practitioner. The largest count in a recipe is the ingredient the recipe is really about. Mandrake for protection. Oak for Merlin's. Bergamot for Oceanus. Frankincense at eight for the Wizards' Oil.
Four things to do with the numbers in front of you. Follow them exactly. Whatever the counts encode, you have no basis for varying them, and the material gives no permission to scale.
Read the largest count as the subject. The dominant ingredient tells you what the mixture is built around, which is useful when a recipe's name does not. Treat a single-oil instruction's number as meaningful. Bergamot's five drops make a five-pointed star; asafetida's seven go in the expelling hand. Those are not measurements.
Use a separate dropper for each ingredient, which is the section's own rule and the only way the counts stay accurate anyway.
The recipes and the individual instructions are printed as given; the census above is this build's own count and is stated as such. ⚠ The census is of what is legible in this edition's conversion. Two recipes yielded fewer entries than they visibly contain, and the individual-uses list could not be parsed cleanly. Totals are a floor, not a final figure. Re-check the plate.
⚠ An earlier pass over this material recorded that six of eight fives are warding and four of six eights are magical power or spirit contact. That does not reproduce against the recipe list, where the fives run twelve rather than eight and cluster inside Healing Oil and Mordekai's Vanquishing Oil — one healing and one warding. The claim is corrected here rather than carried forward. It may hold against the individual-uses list, which is the dataset a complete census would need.
🚩 The absence of seven and nine is real in the sample counted and is reported as an observation at n=45 rather than as a structural finding. Unresolved: nothing anywhere states whether recipe drop counts are quantities, ratios or coded values, and the material's own numbers chapter would predict the third.
Follow the numbers chapter and every count is active. Objects infused with the quality of a number react with the universe as that number would react, and a grid of five broadcasts five. On that reading a recipe with four fives in it is broadcasting five, four times.
Follow the recipes and the counts are measures. Dragon's Fog is four, four, four and four; Healing is five, five, five, five, four, three, three, two. Nothing in either pattern tracks what the oil does, and the eights and tens fall where one ingredient dominates.
Both are in the same material, one chapter apart. And the practitioner mixing an oil has to decide which they are doing — dosing a mixture, or assembling a numeric broadcast — with no guidance either way.
Afa, Geomancy, Binary Code — In both traditions a specific count is the operative content rather than a convenience, and a pattern of definite size is what carries the message. Neither treats the number as bookkeeping.
Afa's numbers are addresses in a fixed space. A cast gives a sixteen-position pattern of open and closed, the pattern maps to a named figure, and there is a canonical list to check it against. A dibia can be wrong, and being wrong is detectable, because the space of valid figures is closed and known.
Here there is no such list, and neither corpus notices what that costs. A recipe says four drops and nothing anywhere records what four drops of sandalwood in a healing oil is supposed to be an instance of. Which means an error is undetectable in principle — miscount by one and you have made a different mixture with no name, no entry and no way to discover it. A closed system of figures makes numeracy checkable. An open list of quantities makes it a matter of care alone.
Sharpest implication. The counts cluster by recipe rather than by function. Four of the twelve fives sit in Healing Oil and another four in Mordekai's Vanquishing Oil — one for curing and one for destroying the forces of darkness. If the number carried its own quality into the mixture, those two would not share it.
Generative questions.