This paper focuses on the prosodic realization of Urdu/Hindi kya ‘what’ in polar and wh-constituent questions. The wh-word kya ‘what’ is polyfunctional in that it is used in wh-constituent questions to mean ‘what’, but also serves as a marker of polar questions. The distribution of kya is relatively free in both types of questions, which can lead to syntactically (and therefore semantically) ambiguous structures involving kya ‘what’.We show that prosodic information is crucial for the disambiguation of such sentences. We report on a production experiment which establishes that the wh-constituent kya is prosodically focused while polar kya is accentless. Moreover, the nouns following wh-constituent kya have shorter duration as compared with the nouns following polar kya, which have longer duration and an LH contour. We show that speakers of Urdu/Hindi are perceptually sensitive to the prosodic properties of wh-constituent and polar kya and the following nouns. We take the information established about kya ‘what’ and show how the prosodic differences guide syntactic disambiguation at the prosody-syntax interface, which in turn results in the activation of the appropriate semantic information (polar vs. wh-constituent readings of kya). We model our analysis within Lexical-Functional Grammar (LFG) and work with Bogel’s framework of the prosody-syntax interface (Bogel 2015, this volume).