N-gram
A run of n adjacent characters or words, used to measure how language-like a text is.
Bigrams (n=2), trigrams (n=3), quadgrams (n=4) and so on. Languages have stable patterns: "TH" and "HE" are common in English, while "QZ" almost never appears. Counting n-gram frequencies over a large corpus gives a scoring function for candidate plaintexts. Text that looks like language scores high; random letters score low.
The same counting sits underneath statistical language models, which simply swap characters for words or tokens. Cryptanalysts were counting letter sequences for decades before anyone called it language modelling.