According to Mīmāṃsā authors, and unlike Nyāya ones, Vedic sentences do not convey the existence of something, but rather that something should be done. This means that the entire Veda is an instrument of knowledge only as regards duties and cannot be falsified through sense-perception, inference, etc. No Mīmāṃsā author, for instance, could ever blame a scientist for reaching a conclusion that clashes with data found in the Veda.
In contrast, sense perception, inference, etc. only convey knowledge about what exists. Therefore, since duties cannot be known through any other instruments of knowledge but the Veda, the Veda remains unfalsifiable. This unfalsifiability by default would not be enough for thinkers like the Buddhist epistemologists, who claim that we need to find solid reasons in favour of something or someone in order to believe it or her. However, Mīmāṃsā authors claim that all cognitions are intrinsically valid unless and until they are falsified. Therefore, the fact that the Veda conveys knowledge which cannot be overturned makes it into a valid instrument of knowledge.
But how can one claim that the whole Veda only conveys duties? What about Vedic sentences which look like descriptions of states of affairs? Historically, the primary focus of Mīmāṃsā authors has always been the Brāhmaṇa portion of the Veda, which entails sacrificial prescriptions. Mīmāṃsā authors have been busy for centuries trying to make sense of the seeming confused sum of statements relative to the various sacrifices, and ranging from direct prescriptions to aetiologies. They learnt to put at the centre a originative prescription (utpattividhi, the one which first mentions that a given sacrifice has to be performed) and to link to it all the other prescriptive and seemingly descriptive statements. In this way, they could find out that many statements of the latter group are in fact to be read either as prescriptions or as prescriptions’ supplements. Among the former are statements such as “The ladle is made of palāśa wood”, which in fact means “Carve a ladle out of palāśa wood” (a contemporary reader may recall descriptive statements with the same purport in recipe books). As for the latter, statements such as “Vāyu is the swiftest deity” are re-conceived as eulogising the prescription or one of its elements, so that they convey a meaning only together with it. In the example given, the final meaning would be “Sacrifice to Vāyu, since you will get your result promptly”. Passages out of the Vedic saṃhitās are labelled mantras and have a role only within a sacrifice. Thus, they too do not aim at conveying a descriptive meaning.
Prābhākara authors extend this approach to language in general, which they explain on the basis of their explanation of Vedic language. Accordingly, they claim that also worldly language only conveys things to be done. This theory has been opposed by Bhāṭṭa Mīmāṃsā and Vedānta authors, and it may seem to be counter-intuitive, but in fact it probably only opposes 21st c. European common-sense assumptions more than general human intuitions. Some common examples may help to get what is at stake with it. When, for instance, a mother tells her children “It is nine o’clock”, her children all immediately understand that this descriptive sentence is in fact only a supplement to an unspoken prescription, namely “Go to bed!”. In other words, it is not at all clear that we are really so interested in sheer facts. By contrast, many (most?) descriptive statements seem to be mostly just accessories to an underlying prescriptive one (“X said…/Y did…” are in fact just an accessory to the unspoken prescription “Stop talking to X!/Do like Y!”).
The opposition from Bhāṭṭa authors centers around counter-examples such as descriptive statements which appear to convey a meaning on their own. The stock examples are “Your wife gave birth to a son” and “Your unmarried daughter is pregnant”. One might imagine that, like in the examples I made up in the paragraph above, also in those cases there are unspoken prescriptions these statements supplement (e.g., “Come home to celebrate!” and “Find a solution!” respectively). However, Prābhākara authors retort that in these cases the sentences do not actually convey a meaning alone, but through the fact that their speakers’ mimicry helps us understanding that we have to feel happy or desperate.
By contrast, Vedānta authors have deeper reasons to oppose the Prābhākara (and also the Bhāṭṭa) theory. Vedānta authors, unlike Pūrva Mīmāṃsā ones, historically started their inquiry by focusing on the Upaniṣads, which contain much less prescriptive statements than the Brāhmaṇas and many more narrative passages with seemingly descriptive statements. Moreover, their ontology is based on the content of such Upaniṣadic statements, since, they claim, the ultimate reality, which they call brahman, cannot be known through perception, inference, etc., and can only be known through these Upaniṣadic statements. Thus, if the Upaniṣadic statements can only be valid insofar as they supplement a prescription, they would no longer be able to convey information about the brahman. Vedānta authors, therefore, have to ensure that Upaniṣadic statements can convey a descriptive meaning. Śaṅkara, the main author of Advaita Vedānta, tries to achieve this goal by separating the destinies of Upaniṣads and Brāhmaṇas, just like he separated Pūrva and Uttara Mīmāṃsā (i.e., Vedānta). Rāmānuja, the main author of Viśiṣṭādvaita Vedānta, by contrast, aimed at considering the whole Veda as a whole and therefore tried to construct all Vedic sentences as conveying a descriptive meaning, which can eventually lead to action in order to achieve the purpose they describe.