Bagels around the world
132 bagel recipes in 3 languages: English 87, French 29 and German 16. Each is reduced to grams per 100 g of flour, the same as everywhere in the lab, so a French page and an American one line up. English, the largest group, is the one the others are compared with.
The language is the page's, not the cook's country: a French page may be from Quebec, a Portuguese one from Brazil, a Dutch one from Belgium. Each language's lines are read in its own words and units, and a recipe that gives “a pinch” of salt counts a pinch.
Also in the lab, too few to compare (under 8 recipes): Dutch 7, Italian 6, Spanish 4 and Portuguese 1.
Where they part from English
No language differs from English on any ingredient by more than chance would explain. Bagels read much the same in French and German.
Every ingredient, language by language
Grams per 100 g of flour. One row a language, largest first.
middle half, and the median (the number at right; when most use none, the share that uses any)middle half of every recipesmall number after a name: recipes
The typical batch in each language
The median of each language's recipes for every ingredient, scaled to 500 g of flour. No single recipe is exactly this. Under each, the recipe nearest its language's medians.
English87
- Flour
- 500 g
- Water
- 300 g
- Malt or sugar
- 23 g
- Salt
- 10 g
- Yeast
- 5 g
French29
- Flour
- 500 g
- Water
- 255 g
- Malt or sugar
- 12 g
- Salt
- 8 g
- Yeast
- 5 g
German16
- Flour
- 500 g
- Water
- 300 g
- Malt or sugar
- 23 g
- Salt
- 6 g
- Yeast
- 7 g
The medians and middle halves as a table
| Language | Water | Malt or sugar | Salt | Yeast |
|---|---|---|---|---|
| English 87 | 60%53% to 64% | 4.6%2.7% to 7.5% | 2%1.5% to 2.4% | 1%0.4% to 1.6% |
| French 29 | 51%50% to 61% | 2.4%0.3% to 6.3% | 1.6%1.3% to 2% | 1%0.4% to 1.7% |
| German 16 | 60%57% to 65% | 4.6%2.4% to 6.3% | 1.2%1% to 2% | 1.3%0.5% to 1.4% |
How the comparison works
A language gets a row with at least 8 recipes. Each is compared with the English recipes on every ingredient, and a difference is listed only when it clears two bars: the medians sit at least a third of the middle half of all recipes apart (about half a standard deviation, a medium effect), and a rank test puts the gap at 3 standard errors or more, so a language with few recipes needs a large difference. For an ingredient many recipes skip, the share using none counts too: at least 25 points apart, at the same 3 standard errors.
Every ingredient is tested for every language, a few dozen tests per lab, which is why the bar is higher than the usual two standard errors. Salt is listed last when it differs: many recipes give it as a pinch or leave it to taste. Shares and “more than” percentages are rounded down, so none is an overstatement.
These are what recipe writers in each language do, not rules of each country's cooking. A language with 8 to 20 recipes may come mostly from a few sites; each recipe page names its source.