Guess the Language
Where is the nearest station. I do not understand any of this. Ordinary sentences, written the way a speaker would actually write them, in the script that language actually uses.
The four options are always relatives, so the job is telling Danish from Swedish rather than telling Danish from Thai. Mixing families would hand the answer over: one Romance language sitting among three East Asian ones is a matter of looking at the alphabet, not of knowing anything.
Work out which language a sentence is written in.
Play Guess the LanguageDiacritics are the whole game
Within a family, the letters that are not in English are what separate the options. Icelandic has þ and ð and nothing else in the deck does. Danish and Norwegian share æ and ø; Swedish uses ä and ö instead, which is the fastest Scandinavian split available. Polish has ł and ą; Czech and Slovak-shaped languages use háčeks, the little wedge over a consonant. Romanian has ș and ț with commas under them. Hungarian has ő and ű with double acutes, and Estonian has õ.
Turkish has a dotless ı, which is the single most recognisable letter in the Turkic group, and Azerbaijani has ə. Maltese has ħ and ġ, and is the only Semitic language in the file written in the Latin alphabet, which makes it a free round once you have seen it. Vietnamese stacks tone marks on top of vowels that already have diacritics, and Yoruba and Igbo put dots under letters.
In Cyrillic, the extra letters do the same job: і, ї and є mean Ukrainian; қ, ң and ұ mean Kazakh; ө and ү mean Mongolian.
When the script answers it outright
Twenty-five of the 67 languages use a plain Latin alphabet, and those are the hard ones. The rest each carry more information in their script than in their words. The four Dravidian languages — Tamil, Telugu, Kannada and Malayalam — each use a completely different writing system, so a round from that family is decided by shape alone if you can tell the four apart, and by the script clue if you cannot.
The Arabic-script languages separate by style and by added letters: Persian adds four letters to the Arabic set, and Urdu is written in the flowing nastaliq style, which slopes down to the left in a way standard naskh does not. Japanese mixes kanji with two kana syllabaries, so a sentence with simple rounded characters interspersed among complex ones is Japanese rather than Chinese. Thai is written without spaces between words.
The clue ladder
Three clues at a rung each. The script first, described in the terms above — Latin alphabet, with ł and ą — which is the best-value clue in the game when the four options share a family but not a writing system. Then the family, which is only worth buying when the small families have been topped up from elsewhere. Then where it is spoken, which is usually decisive.
The families in the file are lopsided by design: Germanic has seven members, Romance six, Slavic six, Indo-Aryan six, and then a long tail down to Greek, Georgian, Armenian and Basque, which have no close relatives here at all. Those get three decoys from anywhere, which is the best that can be done for a language with no cousins.
How the sentences were written
Accuracy was the constraint. A wrong sentence in a language nobody in the room speaks looks fine and teaches nonsense, so nothing went in unless the wording is ordinary, idiomatic and spelled the way a speaker would write it. A dozen candidates were dropped for failing that bar rather than guessed at — mostly languages where the tone marks, diacritics or script are easy to get subtly wrong.
The sentences also avoid naming their own language. I speak a little Norwegian is not a question, it is an answer.
Scoring
Ten rounds, four rungs: 1000, 620, 380, 240. It runs entirely offline. With four options in front of you and a wrong guess costing a rung, the same arithmetic as the quote game applies — a cold pick wins outright a quarter of the time and still leaves you at 620 with three options left.