The question of the relation (saṃbandha) between language (words, sentences,texts) and reality (real objects, objects of thought) received varied answers from Indian philosophers, and triggered intense debates in ancient India froman early date. While the Sanskrit grammatical tradition, beginning with Kātyāyana and Patañjali (3rd / 2nd centuries BC), generally considers that words are connected to their objects by a relationwhich is said to be “fixed” or “eternal”, this view was challenged in the first millennium AD by Buddhist philosophers like Dignāga (5th c.) and Dharmakīrti (6th c.), who hold a conventionalist position and consider that relation between language and reality is entirely man-made and arbitrary. In this lecture, I will concentrate on Bhartṛhari’s contribution to this debate, especially in the second book of his Vākyapadīya and theauto-commentary (Svavṛtti). Although a strict follower of the Pāṇinian school of Sanskrit grammar, Bhartṛhari distinguishes himself from the earliergrammatical tradition by adopting conventionalist views that announce, in many respects, those of later Buddhist philosophers. I shall argue that such aposition was suggested to Bhartṛhari by reflections on linguistic phenomena pertaining to the use of a word in more than one sense, such a polysemy, ambiguity, and metaphor. This will lead us to further thoughts on deliberate uses of ambiguity in the poetic domain, and on the possibility of multiple interpretations of a text in the Sanskrit tradition.