The top of the list is twelve insurance policies

I sorted a 1,647,269-entry dump of German word forms by length. Here is the summit: maschinenbetriebsunterbrechungsversicherungen, forty-five letters, "machinery breakdown business interruption insurance policies". Then unternehmensberatungsdienstleistungsangeboten. Then unternehmensbeteiligungsgesellschaftsgesetzes.

The first twelve are insurance products, consulting services or acts of parliament. Not one scientific term, not one poetic oddity. The longest words in German are about paperwork.

None of that will surprise anyone. What stopped me was the second result, out of the same corpus a few minutes later: of the five languages I measured, German is the one that speaks in three-letter words the most. By a wide margin.

Three letters, 31% of the time

Here is how spoken German breaks down by word length, weighted by how often each word actually occurs:

Nearly one word in three is exactly three letters long. That is comfortably the highest of the five: English sits at 24%, Italian at 16.9%, French at 16.8%, Spanish at 15.5%.

The culprits are few and you know them already: ich (3.89% on its own), sie, das, ist, die, und, der, wir, was, ein, mir, mit, wie, den. Fourteen words carrying a large share of every German conversation. The declined articles do much of it. Where English has one the, German has der, die, das, den, dem, des, six three-letter forms that turn up in front of every noun phrase.

The widest gap I have measured

Put the two measurements side by side and German becomes a language with two faces:

The average German dictionary word is more than three times the length of the average German word in a sentence. None of the other four comes close to that ratio.

And spoken German is not even the shortest of the five in absolute terms. At 4.58 letters per word it is the longest. So the entire gap comes from the dictionary, not from the conversation. What that 3.16 measures is not a thrifty language, it is an inflated inventory.

What that inflated inventory actually holds

I counted the tail, language by language:

Three German entries in four pass twelve letters. One in eight passes twenty. In the other four languages, the twenty-letter mark is reached by roughly one entry per thousand at best.

In raw counts: 33,226 entries of at least twenty-five letters, 3,890 of at least thirty, 407 of at least thirty-five, 49 of at least forty.

The cause is well known and I will not re-explain it in detail, the article on why German is hard covers it: German writes its compounds with no space and no hyphen. Kraftfahrzeughaftpflichtversicherung is not a word the way horse is a word, it is a glued-together phrase. English says the same thing in four words, "motor vehicle third-party liability insurance", and nobody finds the English remarkable.

Which means those 12.3% are not measuring the length of German words. They are measuring a typographic convention.

The records, and why they are not records

The unofficial champion is still Rindfleischetikettierungsüberwachungsaufgabenübertragungsgesetz, sixty-three letters, a law from Mecklenburg-Vorpommern on delegating the supervision of beef labelling. Adopted in January 2000, repealed in May 2013, which prompted the German press to announce that the longest word in the language had just died.

It never held the official record anyway. That went in 2003 to a Grundstücksverkehrsgenehmigungszuständigkeitsübertragungsverordnung, sixty-seven letters, a regulation on transferring jurisdiction over land transaction permits. Nobody has ever said that one out loud either.

The famous one is neither: Donaudampfschifffahrtsgesellschaftskapitän, forty-two letters, the captain of the Danube Steamboat Shipping Company. It has served as a tongue-twister for generations and has never referred to any actual person.

Duden sticks to what it actually prints, and its ranking is a good deal less spectacular. Its longest headword is Aufmerksamkeitsdefizit-Hyperaktivitätsstörung, forty-four letters, attention deficit hyperactivity disorder. Second comes Kraftfahrzeug-Haftpflichtversicherung, thirty-six. Two words people genuinely use, in a sentence, without meaning it as a joke. Which is the whole point.

That caution is well placed, because there is no such thing as the longest German word. Compounding is productive: any speaker can stick one more noun on the end and produce a longer word than the last one, and it will be perfectly grammatical. Asking for the longest word in German is asking for the largest whole number.

At the other end, almost nothing

German has no one-letter words. The spoken corpus assigns 0.30% of running text to single-letter tokens, and most of that is an apostrophised s: wie geht's, gibt's, an elision of es that subtitles sometimes split off. For comparison, one-letter words make up 6.5% of spoken Italian, 6.2% of English and 4.3% of Spanish.

So the shortest real German words run to two letters, and there are not many: du, er, es, in, zu, ja, so, an, da, um, im, wo, oh, ei. Together they account for 9.9% of running text, the lowest score of the five and a long way from French at 26.9%.

That completes a consistent picture. German skips the tiny-word stage, parks itself heavily on three letters, then compounds upward with no ceiling at all.

Sources

For the records and the Duden ranking:

• Duden, "Die längsten Wörter im Duden", which gives the full order of the longest headwords in the Rechtschreibduden
• The Duden entry for Kraftfahrzeug-Haftpflichtversicherung
• German Wikipedia for the history of the 1999-2000 law and the successive Guinness records: Rindfleischetikettierungsüberwachungsaufgabenübertragungsgesetz

For the measurements, two corpora, as in the rest of this series. Flat length counts come from the public lists at lorenbrichter/Words, from which I kept 1,647,269 German entries after normalising (lowercase, ß to s, umlauts preserved, anything that is not a letter dropped). Usage-weighted figures come from FrequencyWords, fifty thousand words with counts, built from subtitles.

Three caveats, and the first one matters. A list of German words cannot really be a list: since compounding is unbounded, any inventory is an arbitrary sample of what someone decided to include. So the 12.3% of entries past twenty letters describes a compilation policy about as much as it describes the language. The comparison against the other four languages still tells you something; the absolute value much less.

Second, the subtitle corpus is film dialogue. It under-represents exactly the registers where the monster compounds live, administration and law, which mechanically widens the gap I am measuring. And third, German subtitles sometimes write elisions detached, which is where those single-letter tokens come from.

One last thing

I started this article expecting to write about forty-five-letter words. I ended up writing about ich, das, ist and und.

There is something fitting in that. German's reputation rests on a handful of administrative monsters nobody ever says out loud, while the actual language moves along at three letters a word, faster than any of its neighbours.

Elsewhere in this series: the shortest and longest words in Spanish, which language is hardest at Wordle, and letter frequency in four languages.