Word Repetitions
Words that appear two or more times in direct succession within a sentence — what linguists call epizeuxis or immediate repetition. Examples: „ich will immer immer mehr“, „du bist so so so schön!“. Only truly consecutive occurrences count; a punctuation token or any other word in between breaks the sequence and is not counted. The top 500 most-frequently-repeated types per archive are shown, ranked by the number of sentences containing such a sequence.
Linguistic Insights
Epizeuxis: from sound to intensification
Immediate repetition in songs serves two main functions. Most frequent is the sonic one: syllables and interjections such as na, la, bum, oh, ja top the list and belong to the same paralinguistic layer (interjections and onomatopoeia) seen in the part-of-speech distribution. (A repeated die is a frequent but non-paralinguistic outlier: a syntactic doubling — as in "…auf die die Stars sich nicht drauftrau'n" — rather than genuine epizeuxis.) The second function is genuinely rhetorical: intensification. Repeating a gradable adverb or particle raises its degree: immer immer, so so, sehr sehr, bitte bitte bitte. Here the doubling adds emphasis rather than sound.
A signature of hook-driven genres
Archives differ sharply in how much they rely on immediate repetition. HipHop and chart pop show it by the hundreds, where chanted hooks and ad-libs are structural; narrative and lyric-driven repertoires use it far less. Immediate repetition seems thus a performative intensifier characteristic of rhythm- and hook-based music, while storytelling traditions build emphasis across lines instead of stacking a word in place. Note that counts are per sentence: a single repeated hook recurring throughout a song inflates the figure, so the numbers track refrain structure as much as vocabulary.
| # | Word | Sentences | Example |
|---|---|---|---|
| Run generate_word_repetitions.R to populate this table. | |||
Immediate (consecutive) repetitions only. Top 500 types per archive, ranked by number of sentences. Case-insensitive.
These analyses are based on corpus data as of July 29, 2026.