Caffeine content and decaffeination, explained

It starts with the species

Sourced

Robusta carries more caffeine than arabica. One study of 74 green coffee samples measured an average of 1.8 per cent of dry weight in arabica and 2.9 per cent in robusta by its own method, and quotes two earlier sets of figures: 1.3 and 2.3 per cent in its introduction, and in its discussion another study's averages of 1.3 and 1.8 per cent, with ranges (0.8 to 1.8 and 1.2 to 2.5 per cent) that overlap. The measured pair puts robusta at about 1.6 times arabica, and the two quoted pairs at about 1.8 and about 1.4. Roughly one and a half times is the fair summary of those averages, a flat twice is above all three, and the authors' own conclusion is that caffeine alone cannot reliably tell the species apart.

An average is not a rule for one coffee. The same study's own samples ran from 1.20 to 2.37 per cent among its 32 arabicas and from 1.95 to 3.53 per cent among its 42 robustas, as its chart plots them, so its lowest robusta held less than its highest arabica. A Brazilian study roasted one lot of each species six ways and measured from 5.6 to 10.2 milligrams of caffeine per gram of roasted arabica and from 9.3 to 33.3 per gram of roasted robusta; set roast against roast, those two lots stand anywhere from level to six to one. Its authors call caffeine stable to heat and do not explain why their robusta read so much lower after the harshest roasts, so part of that spread is unexplained. Two lots say nothing about a species. What they do show is that no single ratio turns an arabica figure into a robusta one.

A blend containing robusta therefore usually starts higher before anything about roasting or brewing is considered, and a bag rarely states the proportion. Nobody has published the caffeine in a cup of brewed robusta with the dose stated, in anything CoffeeHQ has read. The caffeine estimator therefore prints no single range for robusta. It makes two separate statements: at least as an arabica brewed the same way, and not more than all the caffeine in the richest robusta measured, as if every milligram reached the cup. That ceiling is a margin of caution and not what a cup is expected to hold, and where between the two a given robusta falls is not known.

What was found in real servings

Sourced

Researchers in Poland measured the caffeine in 299 servings of coffee, some bought from coffee shops, petrol stations and bakeries and some prepared in a laboratory in the way most Polish households do, by pouring hot water over a spoonful of ground or instant coffee.

Americano had the most on average, 143 milligrams a serving. Espresso averaged 75 milligrams, which the authors put down mainly to its small volume, 34 millilitres on average. The home-style brew of one teaspoon of ground coffee averaged 23 milligrams in a 160-millilitre cup. Coffee bought ready-made carried about three times as much as the home-made coffees. Four per cent of all samples exceeded 200 milligrams in a single serving, and among Americanos nearly one in five did.

The spread within each kind of drink is the more useful finding. The largest serving of a given type held between three and almost seven times the caffeine of the smallest. The authors conclude that advice counted in cups should also say how large the serving is and how strong the coffee.

Published reference figures

Sourced

These are a public body's general figures, not measurements of any particular coffee, and they are given here as that body states them.

A range this wide is the point. The amount in a given cup depends on the coffee and the recipe, and the measured servings above fall outside it in both directions.

What roasting does

Sourced

The belief that dark roasts are stronger in caffeine, and the opposite belief that roasting burns caffeine away, have both been tested. One study roasted two coffees across a wide range, brewed them in an AeroPress at a fixed ratio and measured the caffeine. Under identical brewing, the caffeine in the cup was highest from light-to-medium roasts and lower from darker ones. The authors read that as indirect evidence that some caffeine is lost late in a roast.

The same study found that when a dark roast and a lighter one were brewed to the same extraction yield, the dark roast's brew held more caffeine. The two results are not in conflict. Darker roasts in that study gave up less material overall, and caffeine made up a larger share of what they did give up.

The differences were modest. Across every roast and brew time the study reports 134 to 165 milligrams in an 8-fluid-ounce cup for one coffee and 145 to 165 for the other, and its authors describe roast's effect on caffeine as evident but less pronounced than its effect on other compounds. The study covered two coffees, one brewer and one grinder, and its authors say more work is needed before the result is extended.

A second study measured the roasted coffee itself and not a brew: one lot of arabica and one of robusta from Brazil, each roasted once under six profiles, three of them named for roast defects. The six arabica figures, 5.6 to 10.2 milligrams per gram, did not differ from one another by the study's own test. The robusta read 30 to 33 milligrams per gram after the three mildest profiles, 19 after a dark roast, and 12 and 9 after the two the authors call baked and scorched. They describe caffeine as stable to heat and give no account of the fall. It is one lot, roasted once at each setting, so it is reported here as a result that does not fit the first study, not as a measure of how much caffeine a dark roast loses.

Why concentration misleads

Sourced

A review of ten years of studies found that espresso had the most caffeine per litre of drink, for arabica, robusta and blends alike, and linked that to how much coffee is used for how little water. Per litre is not per serving. The same review notes that the studies it compared used different amounts of coffee and water, that some did not report caffeine per serving at all, and that comparing them properly would need the other variables held constant.

So espresso is the most concentrated common coffee and, served as a single small drink, often not the largest dose. A drink built on two shots is a different matter, and so is a large cup of anything.

Cold brew follows the same arithmetic: it is often made with a large amount of coffee and not always diluted much. Whether a long cold steep takes more or less caffeine from each gram than a short hot brew has been measured directly in two studies read for this page, and they do not agree. One steeped an arabica coffee from Hawaii at 35 grams to 350 millilitres for 24 hours at room temperature and brewed the same coffee in a French press at the same ratio for six minutes. The cold brews measured 990 to 1,230 milligrams of caffeine a litre and the hot ones 840 to 1,060; the difference was significant for the coarser grind and not for the finer one. The other steeped three coffees at 7 grams to 100 millilitres for nine hours in a refrigerator and brewed the same coffees in a drip machine. One came out lower cold than hot, and for the other two the study's own statistics do not separate cold from hot: 630 to 800 milligrams a litre cold against 740 to 770 hot.

Read together, they say that a long steep can dissolve as much caffeine as a hot brew and sometimes more, that a cold, shorter one may not, and that the amount of coffee used matters more than either.

Caffeine for each gram of coffee used

Sourced

Only a few studies say how much coffee went into the drink they measured. Dividing what they found by that dose gives caffeine per gram of ground coffee, which is the figure a recipe can be compared on. Every coffee in these studies was arabica, or of unstated species.

These are the spans of a handful of studies, not limits. The caffeine estimator's ranges are these spans rounded outward, each held below the caffeine an arabica bean contains, and the page on its method lists each study and where the figure was read.

How caffeine is removed

Widely documented

Solvent processes

Green beans are steamed and then rinsed with a solvent that takes up caffeine, commonly methylene chloride or ethyl acetate, after which the solvent is removed and the beans are dried. Ethyl acetate made from sugar cane is the reason some coffees are labelled sugar-cane decaf.

Water processes

Beans are soaked in water already loaded with coffee's other soluble compounds, so that caffeine moves out of the bean and most of the rest stays. The caffeine is filtered out of the water with carbon and no solvent is added.

Carbon dioxide process

Carbon dioxide held at high pressure acts as a solvent for caffeine. It is passed through moistened beans and the caffeine is recovered from it. The equipment suits large volumes.

How much caffeine decaf still has

Sourced

Decaffeination removes most of the caffeine and not all of it. The Food and Drug Administration's guidance gives 2 to 15 milligrams for an 8-fluid-ounce cup. Several cups add up, which matters to anyone advised to avoid caffeine altogether.

Rules differ by product and place. In European law, instant coffee and other coffee extracts may be labelled decaffeinated only if their caffeine does not exceed 0.3 per cent of the coffee-based dry matter. That directive covers extracts; it does not regulate roasted beans, for which no rule was opened here.

The descriptions of the three processes above are general knowledge, written without a source open, and no comparison of how they affect taste was read.

What this page does not tell you about your own cup

Sourced

It does not say how much caffeine is right for anyone. The Food and Drug Administration cites 400 milligrams a day as an amount not generally associated with negative effects in most adults, and says in the same place that people vary widely in how sensitive they are to caffeine and how fast they clear it, and that anyone pregnant, trying to become pregnant or breastfeeding should ask a health professional. That figure is for a day's caffeine from everything, not from coffee alone: the same guidance names tea, chocolate, energy drinks and dietary supplements among its sources. That is a public body's general statement and the whole of what is repeated here.

It also gives no figure for a named drink, brand or brewing method as a fact about that drink; the measured servings above describe one country's cups. It does not compare decaffeination methods for taste or safety, and it gives no figures for energy drinks, tea or other sources of caffeine.

If you are adding up a day's caffeine, a coffee estimate is one part of it. Where another product states its caffeine per serving, that stated figure is the one to count for it, separately. A plant ingredient on a label, such as guarana, kola nut or green tea extract, is weighed as the ingredient and not as caffeine, so its weight is not a caffeine figure. In the United States a label need not state how much caffeine a product holds, and caffeine that comes from such an ingredient does not appear in the ingredient list, so where a product does not say, its caffeine is unknown rather than zero.

What to do next

• Caffeine estimator — Works from the dose of coffee in a recipe and the share of the brew you drink, and shows the spread.
• How CoffeeHQ handles caffeine figures — Where the estimator's numbers come from and why they are ranges.
• Decaf coffee — Choosing and brewing decaf, in more depth.
• Robusta cultivars — The species with more caffeine is not one plant; what its varieties are bred for.