A corpus language is a language that has no living speakers but for which numerous records produced by its native speakers survive.^[1] Examples of corpus languages are Ancient Greek, Latin, the Egyptian Language, Old English and Elamite.

Some corpus languages left a very large corpus, such as Ancient Greek and Latin, and therefore can be fully reconstructed, even though some details of the pronunciation may be unclear. Such languages can be used even today, as is the case with Sanskrit and Latin. Others have such a limited corpus that some important words, e.g. some pronouns, are not found in the corpus. Examples for this are Ugaritic and Gothic. Languages that are only attested by a few words, often names, and a few phrases (called Trümmersprachen in German linguistics, literally "rubble languages") can only be reconstructed in a very limited way and often their genetic relationship to other languages remains unclear. Examples are the Lombardic language and Dadanitic, a Semitic language that may be close to classical Arabic.

Corpus languages are studied using the methods of corpus linguistics, but corpus linguistics can be used (and is commonly used) for the study of the writings and other records of living languages.

Not all extinct languages are "corpus languages," since there are many extinct languages in which few or no writings or other records survive.