Semantic prosody, also discourse prosody, describes the way in which certain seemingly neutral words can be perceived with positive or negative associations through frequent occurrences with particular collocations.
An example given by John Sinclair is the combination, set in, which has a negative prosody: rot is a prime example for what is going to set in. Other well-known examples are cause, which is also used mostly in a negative context (accident, catastrophe, etc.), though one can also say that something "caused happiness" (Stefanowitsch and Gries 2003).
In recent years, linguists have found many hidden associations affecting the neutrality of language, through the use of corpus linguistics and concordancing software. The software is used to arrange Key Words in Context from a corpus of several million words of naturally-occurring text. The collocates can then be arranged alphabetically according to first or second word to the right or to the left. Using such a method, Elena Tognini-Bonelli (2001) found that the word largely occurred more frequently with negative words or expressions, while broadly appeared more frequently with positive ones. Lexicographers have often failed to allow for semantic prosody when defining a word, although with the recent development and increasing use of computers, the field of corpus linguistics is now being combined with that of lexicography.
Prosody has also been used to analyze discourse structure. Discourse is not a mere concatenation of utterances; talk is organized in sections through relations between discourse segments, topicality, or other ways. Prosody has been found to correlate with these structures of discourse, notably via key (the pitch of a first prominent syllable in an utterance).