Both Śabara’s and Veṅkaṭanātha’s commentary on the Mīmāṃsā Sūtra insist that mantras are not important only insofar as they are pronounced, but rather that they convey a meaning (technically: they are vivakṣitārtha `they have intended meanings’).
One of the evidences for the meaningfulness of mantras is the fact that mantras are modified (ūh-) in the ectype rituals. If, for instance, the archetype ritual is for Agni and the ectype ritual is offered to Indra, the mantra will be accordingly changed (e.g., from Agnaye juṣṭam to Indrāya juṣṭam). If the mantras had no meaning, there would be no scope for modifying them. If the pronunciation were enough to achieve some unseen potency (apūrva), one would just repeat the mantras in the same form.
In his commentary, Śabara focuses on a problematic Vedic passage which appears to forbid modification (that of na mātā vardhate na pitā). Veṅkaṭanātha explains the same instance, but first spells out the argument in full. At the end of the explanation, which I have summarised above, he also adds a further reason. Since the reason is not spelt out, I assume Veṅkaṭanātha imagined his readers to recognise it immediately, which is interesting, since it may reveal something about his target readers or about how conversant they were with Mīmāṃsā devices. The text passage runs as follows:
The modification is purposeful because, if there were no expression [of the meaning], there would be no application of the Aindrī rule.
vacanābhāve caindrīnyāyānavatārād ūho’rthavān (ad PMS 1.2.52)
I have encountered the aindrī-rule only in ŚBh ad PMS 3.2.3. It points to the fact that in case of conflict between prescription and mantra, the latter should be interpreted as yielding a secondary meaning. Here the point is: If the mantras expressed no meaning at all, then there would be no scope for the application of such a rule.