In all cases, the ancestral protolanguage is not known directly and it may be reconstructed by comparing different members of the language family via a technique called the comparative method, by internal reconstruction or other methods. Through this process only a part of the proto-language's structure and vocabulary can be reconstructed; the reconstruction remains the more fragmentary the more ancient the proto-language in question relative to the number of its descendants. Examples of unattested but (partially) reconstructed proto-languages include Proto-Indo-European, Proto-Uralic, Proto-Bantu and Proto-Paman.
The Proto-Indo-European language (PIE) is the most elaborated example of a proposed language. Although there is no direct evidence that this language ever existed, there is copious evidence for its existence in the many similarities of the Indo-European languages. A great amount of work has been put into the reconstruction of Proto-Indo-European, but there are no means of determining its success.
Sometimes, however, the proto-language is a language which is known from inscriptions (perhaps due to misunderstanding), an example being the Proto-Norse language attested in the Elder Futhark runic inscriptions. For more examples of proto-languages, see the category "proto-languages" (below).