Mīmāṃsā authors refute the Nyāya and Buddhist theory of a conventional relation and try to prove that nobody would ever be able to establish a linguistic convention without words, since any convention-maker would in turn need words to explain that a certain word X is to be connected with a certain meaning. It follows that, in order to avoid a circular regress, at some point one necessarily needs words whose relation with their meanings is not conventional. Later Nyāya authors introduce here the idea of a God who creates words with an embedded conventional relation, but this thesis implies, according to Mīmāṃsā authors, far too many unwarranted assumptions. Mīmāṃsakas rather stick to common experience, in which language is a given.
Mīmāṃsā authors also dedicate much energy to the explanation of the process through which one learns a language, first understanding the meaning of basic sentences and then the meaning of their constituent words.

This process, called vyutpatti, is debated at length in the two subschools of Mīmāṃsā. It is first described by the Prābhākara Śālikanātha, who explains through it that the Prābhākara anvitābhidhāna theory must be right. In fact, the atomistic theory according to which one learns one by one the meaning of each single word is refuted by our common experience, in which we see that children or less experienced speakers learn how to speak by observing experienced speakers interact among each other through sentences, not words. Similarly, the holistic theory of Bhartṛhari leads to unacceptably anti-economic consequences, since one would need to postulate that each single sentence needs to be learnt separately, whereas we commonly see that speakers who master a given set of words can easily understand new sentences, if only they are composed of these same words. What remains open to debate, however, is how one moves from the experience of sentences to the understanding of the single word-meanings (see freschi01 for the different Bhāṭṭa and Prābhākara solutions).