Garum – The Roman Ketchup : Part 1 – The Evidence

An army may have, as Frederick the Great intimated, “marched on its stomach”, but that didn’t necessarily mean that they liked what they ate. When a plebeian joined up with the intention of becoming a citizen and marched across the empire during his four years of service, he had to pay for everything, including his own uniform, all except for one thing…

A small phial which was issued to every soldier and its contents were topped up once a month without cost. This phial was secured to the soldier’s belt with a leather thong and contained the most important thing known to him, a condiment called ‘Garum’.

For food, the army could live off the land and the local cuisine, but to render it palatable, the soldier would smother the ‘weird foreign flavours’ with Garum… and you thought that smothering local cuisine with ketchup was a uniquely British trait – we got it from the Romans!

Garum is a fermented sauce which was prepared from the blood and intestines of fish through a reaction called autolysis. Autolysis is a process of self-digestion, whereby the cells of an organism are dissolved by the digestive fluids that they produce. Other decomposition processes are hindered in the manufacture of garum through salting the fish. The combination of salt and the enzymes of digestion produce a unique reaction however, the result of which is a high concentration in the product of monosodium glutamate.

Monosodium glutamate (MSG) is used in the food industry as a flavour enhancer, it doesn’t alter the flavour itself, but increases the intensity of the signals from the taste buds in the same way  turning up sound volume amplifies the sensation but doesn’t change the sound. While other flavour complexes provide the taste experience, the MSG amplifies the sensation, resulting in a condiment which can effectively smother any unwanted flavour and, as an additive to any recipe, is a natural flavour enhancer.

According to Alexis Soyer’s ‘Pantropheon’ (printed in 1853), the word Garum came from the classical Greek garos, meaning shrimp, indicating that garum was originally made from this crustacean. There appears to be no confirmation of this in classical Greek however and I have certainly not found this as a word for shrimp in any of the lexicons that I possess. The general consensus is that garos (γάρον) was classical Greek for a fish, the identity of which, no one has yet managed to ascertain.

There is another possibility…


A work on farming, called Geoponika was produced in the 10th century and although the book post-dates the fall of Rome by 500 years, it summarises a wealth of information from earlier sources which date back to the time of the Roman empire. The 46th chapter of the 20th volume is dedicated entirely to the production of garum. Since the book is 1000 years old, and the translation that I have used (not being able to get the original Latin text) was published in 1805, both are out of copyright and so I will quote the chapter in its entirety:


Geoponika Liber XX:XLVI – The Composition of Garum
What is called Liquamen is thus made: The intestines of fish are thrown into a vessel, and are salted; and small fish, especially atherinæ, or small mullets or mænæ, or lycostomi, or any small fish, are salted in the same manner; and they are seasoned in the sun, and frequently turned; and when they have been seasoned in the heat, the garum is thus taken from them. A small basket of close mixture is laid in a vessel filled with the small fish already mentioned, and the garum will flow into the basket; and they take up what has been percolated through the basket, which is called Liquamen; and the remainder of the floculence is made into alec. But the Bythinians prepare it in this manner: they indeed take small, or large mænæ, which are more eligible; but if they cannot get them, lycostomi or sauri, or scombri, or alec, and a mixture of all; and they throw them into a baking trough, in which they have been used to mix their meal; and having applied two Italian sextarii of salt to a modius of fish, they work them, that they may be mixed with the salt; and having suffered them to lie during one night, they put them into an earthen vessel; and they set this in the sun during two or three months, stirring them with a stick at stated periods; they then take and stop them and lay them by. Some indeed pour two sextarii of old wine on a sextarius of fish. But if you wish to use the garum immediately, that is not to insolate it, but boil it, you are to do it in this manner: take some strong brine that is proved, so that an egg being put into it may swim (but if it sinks, it has not a sufficient quantity of salt); then throw the fish into the brine in a new pot, and adding some origanum, set it over a good fire, until it boils, that is, until it begins to be a little diminished (some also add sapa to it); then when it is cool, pour it into a strainer a second and a third time, until it comes out clear; and having stopped it, lay it by. But the best garum, which is called Aimation, is thus made: the intestines of the tunny, with the gills and the ichor, and the blood, are taken, and they are sprinkled with a sufficient quantity of salt; and they are left in the vessel during two months in general; the vessel being tapped, the garum called aimation is drawn.


What we can see from this chapter is that there are several fish from which garum is made:


Atherinæ = Mediterranean Sand Smelt
Mænæ = Anchovy
Lycostomi = Sprat
Sauri = Infant Mackerel
Scombri = European Pilchard
The guts, gills and ichor of Tunny = Tuna
and as it says, ‘or any small fish’.
The similarity between all of these fish as that they are small, either small fish or infant fish, typically between 25mm and 50mm long, what we would term today as whitebait or fry. Generally, the young of many different species of fish tend to gather together into shoals for protection, from where they can be caught using fine meshed nets. The whole of the fish, caught as fry is edible, including the head, fins, bones and guts and since the fish is not fully formed, any autolysis process would dissolve the organism entirely to a liquid. Whereas of larger fish, only the guts, ichor and gills are used in garum production. It is consequently a possibility that the term garos is classical Greek, not for an unidentified species of fish, but for the fry of many species.This seems to be confirmed by the distribution of garum factory sites across the Mediterranean, and their processing capacities, which coincide with the migratory routes of such shoals, with an especially increased presence between Valencia, the straits of Gibraltar and round the Spanish coast to the Bay of Douarnenez in Brittany, where particularly large shoals are known to swim. Indeed according to Strabo, the most renowned and highest quality garum that was imported into Rome, was known as garum sociorum (garum of the allies) and came from the factories of Cartegena in southern Spain, which was directly located on a large migratory shoal route.
It appears that the manufacture of garum was exclusively an industrial process, as generally described above, but there was also a ‘mock-garum’ which could be made in very little time, in the culina. This was called muriæ and we can assume that the cooked garum recipe also given in the Geoponika is a form of muriæ. The main difference between garum and mock-garum is that there is no process of autolysis in the making of muriæ and as a consequence, the production of MSG is minimal. As a consequence, although muriæ may possess the flavour of garum, it does not provide the flavour-enhancement that appears to be the addictive component of it.


There were several different qualities of Garum but it is not known what the variations were between them. Some have postulated, and it is quite probable, that the differences were based on the exclusivity of the source fish, for example Apicius considered that the most refined garum was to be made from the Red Mullet (Mullus surmelutes), but this consideration was probably made based on the rarity of the fish and its concomitant cost; one existing document gives the cost of one Red Mullet as 60 sestercii (15 denarii), whereas a Mackerel would cost one semis, or in other words you could get 32 Mackerel for 1 denarii. This probably indicates more to us about the gourmet Apicius’s’ snobbery than the composition of Garum; having dined his way through a personal fortune of 100 million sestercii and finding that he had a mere 10 million sestercii left and afraid of dying in poverty, he took his own life. Juvenal stated in his writings that he could live quite happily on the income he could generate from 400,000 sestercii.
Liquamen is sometimes described as being different from garum, and sometimes as the same. From reading the Geoponika, it seems that a sauce which has been strained into a clear liquid is liquamen, the solid residue of which is alec whereas a sauce which has not been strained, is garum.


It is also most likely that the lowest quality ordinary garum was made from whatever fish by-products were available and it is most likely to have been this quality which was issued as the free ration to the army; if they used it up too quickly, then they had to purchase their own top-up and garum was usually available at every vicus along a Roman road.

The qualities in between are anyone’s guess although Geoponika states that the best garum is aimation.

The principal factor in the making of garum is autolysis, however there are other factors that have a bearing on its production. To understand these factors, it is important to examine all of the available evidence. The most prolific of which is not the literary descriptions of the product but the findings from the archæological excavations of the numerous garum factories around the Roman world.

Evidence indicates that the garum factories used concrete basins of decreasing sizes, which indicate a process of reducing volume. It can be assumed that the larger vessels take the newer product and that during autolysis, there is a reduction in volume as the material becomes more concentrated. When this reduction is achieved, the floculence is transferred into the next vessel (in order for the first to be refilled with the next batch) and again and again until the material ends up in the smallest vat, from where the liquor is drawn off from the alec.

On his page Garum oder Liquamen der Alten, Heinrich Wunderlich demonstrates the manufacture of garum using a yoghurt maker and jam-jars, but this may not produce an accurate garum since the concentrating of the liquid through evaporation would not be occurring in the sealed containers.

According to Athena Trakadas’s article ‘The Evidence for Fish Processing in the Western Mediterranean’, garum factories were roofed over in order to prevent rapid evaporation from direct sunlight, which also eliminates ultra violet light as a factor in the process. On the other hand, Trakadas indicates that workshops were either open at the sides (very similar to monastic cloisters) or had windows which allowed significant ventilation. This appears to be a consistent feature of all garum workshops excavated. It can be concluded therefore that the action of breezes, assisted the process of evaporation and is a significant part of the garum making process.

The presence of a hypocaust is also noted, which was used to ensure an even vat temperature. An 83mcistern which was constructed beneath the workshop, must also be considered as part of the process.

The quantity and shapes of the the garum making vats remains consistent across most of the workshops that have been excavated. Take for example the vats in the workshop at Cotta in Tangiers: There are two large vats (nos 1 & 12), four smaller vats which are ranged next to each of the large vats; 2 – 5 for large vat 1 and 13 – 16 for large vat 12. There are then four even smaller, vats next to large vat 1 (6 – 9) and two smaller vats which are next to large vat 12 (10 & 11).

What is interesting is that in each case, the vats maintain specific proportions in terms of the surface area of the liquid. Vats 2 – 5 & 13 – 16 have a surface area which is roughly ⅔ the surface area of vats 1 & 12, while vats 6 – 9 have a surface area which is roughly ¼ the surface area of vat 1. Vat 10 is roughly ⅓ the surface area of vat 12, while vat 11 is roughly ⅔ the surface area of vats 13 – 16.

This pattern is repeated through most of the excavated workshops, which indicates that the garum makers were controlling the speed at which the floculence was concentrating, by regulating the evaporation rate of liquid from the vats.

It is fairly easy to calculate average evaporation rates, providing you have some specific meteorolgical data for the locality such as the average temperature, the average relative humidity and the mean dew point. To save space here, I will post the calculations I have used in a separate blog.

Tanks 1 & 12 are 3.5m x 3.5m, giving an evaporation area of 12.25m2 and an evaporation rate of 358.707 litres per day

On the other hand, tanks 6 – 9 are 1.2m x 1.2m, giving an evaporation area of 1.44m2 and an evaporation rate of 42.166 litres per day.

Therefore it can be postulated that decanting between tanks was necessary because regulating the speed of evaporation at different stages was an essential part of the process.

The problem that we still have is working out the process. We can be fairly sure that the layout of all the excavated workshops indicated that each tank was related to its neighbour in terms of process-flow. And, in addition to this, we can also assume that the large vats would be used in starting the process. We can therefore assume that there were two batch processes starting simultaneously in vats 1 & 12 and that once the first process was complete, the contents were decanted into their adjacent vats (2-5 & 13-16).

Vat Length (m) Width (m) Depth (m) Capacity (litres) Evaporation Area (m2) Evaporation Rate (litres/day)
1 & 12 3.5 3.5 2 24,500 12.25 358.707
2-5 & 13-16 3.5 2.6 2 18,200 9.1 266.468
6-9 1.2 1.2 2 2,880 1.44 42.166
10 3.5 1.75 2 12,250 6.125 179.353
11 3.5 1.3 2 9,100 4.55 133.234

Vats 2-5 & 13-16 have a capacity of 18,200 litres (25% that of the larger vats).

With an evaporation rate of 358.707 litres per day, it would require 18 days for the content of vat 1 to reduce sufficiently to be able to be decanted into one of the secondary vats (2-5).

Since two of the secondary vats would exceed the capacity of the primary vat, we can be sure that the primary vat was not decanted into two of the smaller vats, therefore we have to conclude that each primary vat batch was rendered down to the capacity of one secondary vat. Consequently the first ferment would require approximately 18 days to complete.

It is also fairly certain that if vats 6-9 related to the secondary vats 2-5, they would be in their proximity. Instead, they are ranged along another wall between vat 1 and vat 12, so it would seem probable that they were filled directly from vat 1, rather than as part of a tertiary process from vats 2-5.

The total capacity of vats 6-9 is 11,520 litres (2880 litres each) and it would take 24 days for vat 1 to render down sufficiently to fit into them.

We can assume then that vat 1 sequentially fed vats 2-5 and 6-9 over a period of 94 days, comprising of four 18 day ferments and one 24 day ferment.

We can assume that whatever happened in vats 2-6, the process was in excess of 54 days, otherwise there wouldn’t be a requirement for four secondary vats. At an evaporation rate of 266.468 litres per day from the surface area of these vats, 54 days would constitute 14,389 litres of liquid evaporated off (80% of the capacity), which would be unfeasible considering the quantity of solids which would be present in the tank. It is likely therefore that as part of the secondary process, the vats received a regular top-up from the 86,000 litre cistern which was constructed beneath the workshop.

It is also probable therefore that the output of the secondary tanks constituted the workshop’s output of ordinary garum; 116,480 litres (allowing for a 20% reduction in capacity per tank during the secondary process). This would constitute  213,333 sextarii in Roman measurements, which according to Diocletian’s “Edict of Maximum Prices” issued in 301 AD, which specified that a sextarius of ordinary garum cost 12 sestercii (3 denarii), means that over a 94 day period (¼ of a year) the income for the workshop was 640,000 sestercii.

It is probable that tanks 6 -11 were used for the processing of smaller quantities of specialist sauces, such as hydrogarum, œnogarum, oxygarum etc..

Tradakas’s article mentions that in the Cotta complex:

“An olive press was also installed in the building, probably producing olive oil for consumption by the workers of the site.”

Although it is likely that this was used for consumption by the workers, other uses cannot be ruled out, such as use of the oil for lighting and as a component of the garum production.

Many books and websites tend to categorically state that the smell of garum production was so awful that the factories were banished from within city walls. There is neither archaeological nor literary evidence for this.

Rather, the reason that garum factories were sited away from cities was that they were required to be as close to their source material as possible. As martial stated:

Take this costly gift, the proud gift
That’s made from the first blood
Of a still breathing Mackerel

— Martial : Book 13, Epigram 102

The fish had to be alive when it was thrown into the garum vat and salted. Although recent archaeological evidence has discovered that there was a small amount of garum making within the city of Pompeii, the factory which we know to have existed in the area has not yet been found. This is unsurprising when you examine the location of the city,

which is within the bay of Naples and hardly within easy access to a significant shoal migratory route. The garum of Pompeii was considered to be one of the best available, but it is entirely likely that the factory was a significant distance from the city. In all likelihood it will be sited near Sorrento, Positano or Amalfi.

Baelo Claudia

Tradakas’s article mentions that the Cotta complex possessed a watch-tower for sighting shoals in order to direct the fishing boats at sea towards the catch. All of the examined garum workshops possessed similar proximity to the shore.

The garum factory at Bælo Claudia is positioned directly between the town and the shore. With the wind most likely flowing from the sea and across the land, Bælo Claudia would be disadvantageously down-wind of the factory if the smell was as disgusting as is frequently described.

It is certain that the production of garum generated a smell, but for a society that was used to digesting the product of the process, it is unlikely that they would have been so hostile to the smell of the factory. Indeed, when you are acclimatised to a smell for long enough, you cease to notice it.

Vat Evaporation Calculations

Coming next…

Experiment 1: Testing Heinrich Wunderlich’s process


Leave a Reply

Your email address will not be published. Required fields are marked *