The discussion on the epistemological validity of sentences starts in Jaimini’s Pūrva Mīmāṃsā Sūtra (PMS) and in Śabara’s commentary thereon when the opponent notes that, even if —as established in PMS 1.1.5— there were really an originary connection between words and meanings, this would still not mean that the authorless Vedas are a reliable instrument of knowledge, since they are made of sentences, not just of words. And clusters of words are either made by human authors or are just causally put together by chance and are thus meaningless.

The only way to counter this objection would be by showing that, just like there is a natural connection between word and meaning, so the sentence-meaning can arise naturally out of the word-meanings, without the need of an authorial intention.

Accordingly, the opponent goes on by showing that the sentence meaning is something altogether different from the word meanings, and thus cannot arise out of them.

Śabara’s answer is of key importance for the development of Indian linguistics and epistemology, since he claims that words denote their meanings and these denote the sentence-meaning. This occurs naturally, so to say, since the denoted word-meanings automatically connect. This is possible insofar as word-meanings have a hierarchical relationship among each other and word-meanings denoting qualities automatically point to a substrate (a word-meaning denoting a substance). The same occurs within words, where the meaning of the case ending specifies the meaning of the theme (prātipādika).

Although Prabhākara lived long time after Śabara, Śabara seems also to address Prābhākaras ante litteram where he explains that it is not possible that (as later Prabhākara will claim) words denote the sentence-meaning. Rather, only word-meanings can connect into the sentence-meaning, as proved through anvaya and vyatireka insofar as

1. A sentence-meaning can be understood also out of arthas which have not been conveyed by words
2. The sentence-meaning is not understood out of words which are heard but whose meaning has not been understood (perhaps because the listener had head-ache)

Concerning the first point, Kumārila will mention the example of one who sees a patch of white and hears the sound of hooves and of neighing and achieves the unitary cognition “A white horse is running” —though not having heard any of the words composing this sentence.

Many thanks are due to S.S. for reading the relevant text passages with me.