In his Tantrarahasya, Rāmānujācārya discusses various interpretations of Kumārila's theory of linguistic signification as regards prescriptive sentences (e.g. "The one who desires heaven should sacrifice"). Why do these sentences impel one to act? –asks Kumārila. It is noteworthy that the Mīmāṃsā account does not presuppose a speaker as a necessary condition and focuses instead on the listener's point of view. In the passage below, Rāmānujācārya investigates about one of the possible ways of constructing Kumārila's theory, namely, that the prescriptive force (the force expressed by prescriptive sentences and inducing one to act, in Sanskrit śabdabhāvanā, see here) is a cognition. But if it so, asks an objector, how can a linguistic unit (śabda), which is usually the object of a cognition, be at the same time its instrument? Rāmānujācārya's answer is that one should not conflate the way one expresses things (language) and the way things are (reality). One can express the same action in different ways, but this does not affect the nature of the action. Hence, that the word appears as object and as instrument of the cognition in different sentences is no hindrance. Similarly, the axe can be an instrument (if one says "I fell the tree with an axe") or an object ("I lower the axe").
This point leads him to an interesting excursus on kārakas (linguistic functions, cf. Chomsky) from the Mīmāṃsā point of view. Rāmānujācārya either slightly modifies or re-interprets Pāṇini's definitions (see, especially, his understanding of svatantrakartā which in Pāṇini identifies the "subject" as independent of any semantic burden, whereas in Rāmānujācārya stresses its independence as for the initiation of the action).
Unfortunately, Mīmāṃsakas do not share Pāṇini's concern in distinguishing kārakas from vibhaktis (i.e., logical functions from the case-endings expressing them).
[UP:] «But if a cognition (jñāna) were the denotative (abhidhā) function (vyāpāra), then how could the word (śabda) –which is the syntactical object (karman) in regard to a cognition (e.g., “Devadatta knows a word”)– be [at the same time] the instrument (karaṇa) with respect to this [denotative] function (vyāpāra)? Indeed, the usage (prayoga) “Devadatta understands a meaning by means of a word (śabda)” is [commonly] seen [so, the word is commonly known to be the cognition's instrument, and how can it be both instrument and object?]».
[PP/Bhāṭṭa:] «It [must] be said [in reply]: The distinct settlement (vyavasthā) of the factors of action (kāraka) like object, instrument, agent [is not made] with respect to the general (mātra) form (rūpa) of the activity (vyāpāra), rather it is made with respect to the delimitation (avacchid-) of the activity (vyāpāra) by means of this or that result,
even though the own nature (svarūpa) of the activity is the same. To elaborate, when precisely (eva) that cognition which [has been previously expressed as] having the word (śabda) as syntactical object (karman), is delimited (avacchid-) by a result characterised as the apprehension (pratipatti) of the meaning, and includes (vyāp-) the word (śabda) (in sentences such as “she knows that meaning through that word”), then the word (śabda) is an instrument because it is included in the activity (vyāpāra) initiated (pravṛt-) for another purpose (i.e., it is an instrument because it is necessarily part –vyāp– of the action though not being the most desired element of it, the object). The meaning is, on the other hand, the syntactical object (karman) because it is the substratum of the action’s result. And then that cognition receives the title (vyapadeśa) of “designation” (abhidhā) because it (the title “designation”) has been comprehensively learnt (vyutpatti) in this regard (that is, in regard to what has as its syntactical object the apprehension of a meaning). When, on the other hand, one wishes to express (vivakṣā) the autonomy (svātantrya) [of the word] with regard to the function (vyāpāra) [of denoting], then the word (śabda) is the agent which designates (abhidhā) the meaning (and in fact the agent is defined in the Aṣṭādhyāyī as svatantra, “autonomous”). But when just this word (śabda) is made by the cognition into [its] content (viṣaya), then the word, partaking (bhaj-) of the result (because the word is the substratum of the result, through the connection of word and meaning), i.e., the displaying of the meaning, is the syntactical object (karman), like in “she knows [p.46] the word (śabda)”. Then, indeed, the cognition does not partake of (bhaj-) the title (vyapadeśa) “designation” (abhidhā). Rather, it must be simply called “cognition”, because the word (śabda) “designation” has not been learnt (vyutpatti) with regard to that (knowledge of the word, not of the meaning). Like an axe: like an axe is the syntactical object (karman) in regard to raising (udyamana) and sinking. When raising and sinking are delimited (avacchid-) by the result of splitting into two, and the axe is included (vyāp-) by them, then the axe is an instrument (karaṇa) because of being included in an activity (vyāpāra) initiated (pravṛt-) for another purpose (artha) (that is, chopping the tree). The wood-logs, instead, are the syntactical object (karman), since they are the substratum (āśraya) of the action’s (kriyā) result. And raising and sinking, then, receive the title of “cutting”, since [the word “cutting”] has been learnt (vyutpatti) in their regard. However, when in regard to this same function (vyāpāra) one wishes to express [the axe's] autonomy then [the axe] appears as the agent, as in “the axe cuts the wood-logs”. When on the other hand precisely those raising and sinking are designated (abhidhā) by “he raises” and “he sinks”, expressing (vācin) [an action] delimited by a result which is the conjunction (saṃyoga) [of the axe] with the upper or lower space-region, then the axe is the syntactical object (karman), partaking of (bhaj-) the result, i.e., [its] union with this or that (region), like in “he raises [and] sinks the axe”. Therefore, it must be considered (dṛś) that the activities (vyāpāra) expressed (vac-) by this or that verbal root (dhātu) (“to raise”, “to sink”, and “to cut”) –though sharing a single (prātisvika) own nature (svarūpa), (i.e., chopping a tree)– are the cause of a distinct settlement (vyavasthā) of these factors of action (kāraka) because of a different delimitation (avacchid-) by means of this or that action's result, and that they (activities) are expressed (vac-) by this or that word (śabda) [again, according to the different delimitation through this or that result, and not because of the activity expressed by the verbal roots themselves, which remains the same]. If, for a distinct settlement (vyavasthā) of the factors of action (kāraka), merely (mātra) the own nature (svarūpa) of the activity (vyāpāra) would be required (apekṣ-), the action (kriyā) designated as “he walks” could become transitive (sakarmika) like the one designated as “he goes (which can have a prāpya karman, and hence be transitive.)”,
or the action “he goes” may become intransitive (akarmaka), and there would not be any distinction (vyavasthā) in use (prayoga) (that is, “he walks” and “he goes” describe the same activity, so if it were just up to the activity itself, they would be precisely the same and there would not be any difference in their employment). [The above explained procedure] can be applied (yuj-) accordingly (yathāsambhavam) also when the denotative (abhidhā) function (vyāpāra) is characterised (lakṣana) as a mnestic trace (saṃskāra). Among the [action factors], agent (kartṛ), object (karman) and instrument (karaṇa) are factors of action (kāraka) reciprocally distinguished (pravibhakta). The agent is autonomous (svatantra) in regard to the action (kriyā), the syntactical object is endowed with the result of the action, the instrument is [necessarily] included in an action (kriyā) which has been initiated (pravṛt-) for another purpose (artha). The intrinsic characters (svabhāva) of dative, ablative and locative, on the contrary, are intermixed with [those of] the [two] factors of action (kāraka) agent, and [syntactical object]. For instance, the locative is the the substratum of either the agent or the object, like “Devadatta sits on the mat (kaṭa), he cooks rice in the saucepan”. The dative is what is held in view through the object, e.g. “he gives a cow to [his] teacher”. [Finally,] the ablative is the general limit (avadhi) [of the action] of the agent (kartṛ) factor (kāraka), e.g., “a leave falls from the tree”. [The distinct settlement of the action factors] must be considered in this way, according to what is suitable in each case (yathāsambhavam)».