At the beginning of his chapter on sentence meaning, Kumārila sets the problem of what is the meaning-bearer in the case of a sentence (see this post). Later in the chapter, he will discuss sphoṭa, apoha and then present his abhihitānvayavāda, but first he discusses in general the possibility of a sentence-meaning. There can be no sentence-meaning out of the sum of the word-meanings, since those are instantaneous and cannot connect (kā 6–8). The same applies to their cognitions (kā 9). Further, neither words (pada) nor the concepts evoked by them (tadbuddhi) can really connect, so that a sentence-meaning is stricto sensu impossible.
Now, it might seem obvious that words or that the concepts of them can connect, since they, e.g., expect each other through syntactical expectancy (ākāṅkṣā), but this is again impossible, given that they do not exist at the same time (kā 11–12). For the same reason, they cannot be connected insofar as they are part of a single cognition (kā 18). Nor can one accept that the sentence-meaning is unitary and part-less, as with the theory of sphoṭa (Pārthasārathi ad kā 18).
The next discussion (departing from kā 27) regards the possibility of connecting words into a sentence meaning insofar as the one affects the other. The typical example is that of śuklaḥ gauḥ `the white ox’. Isn’t it the case that śukla is the viśeṣaṇa ‘qualifier’ of gauḥ, the viśeṣya ‘qualifiand’?
The problem seems akin to the one discussed also in the Vākyapadīya (see the discussion of Helārāja ad VP 3.14.97 in Ogawa in Saṁskṛta-Sādhutā) and, more importantly, in the Mahābhāṣya ad 2.1.1, where the topic under discussion is the relation between rājan and puruṣa in the sequence rājñaḥ puruṣaḥ (=rājapuruṣa) ‘the king’s servant’. Notwithstanding this difference, Patañjali, like Kumārila, employs the terms bheda ‘difference’ and saṃsarga ‘connection’ and speaks of terms that are vyavacchinna ‘determined’.