On the distinction between ‘primary’ and ‘secondary’ Sanskrit compounds

This paper critically evaluates the relevance of the linguistic distinction between so-called ‘primary’ (dvandva, tatpuruṣa and karmadhāraya) and ‘secondary’ (bahuvrīhi) compounds, as well as the validity of the generative conception of the former as ‘bases’ for the latter. These notions were used, to varying degrees, in the Sanskrit grammars of Whitney, Macdonell and Renou. As we shall show, their systematisation in Scharpé’s grammar, on the basis of examples of homonymous compounds, alone demonstrates their intrinsic inadequacy, which may explain their subsequent disappearance. This is not the case with the (slightly later) distinction between ‘exocentric’ and ‘endocentric’ inherited from the neogrammarians (especially Brugmann), taken up by Bloomfield, Carnoy, Renou, Coulson etc. and still in use (Gillon, Lowe) for distinguishing bahuvrīhis from the other types of compounds (Wackernagel and other scholars preferred to stick to the indigenous classification in their grammatical description and avoid introducing such ‘modern’ linguistic notions). Homonymous Vedic compounds distinguished by accent, such as the famous mythological example of indraśatru discussed by Patañjali, Bhartṛhari and Kumārila, and the case of the compound proper names of the Mitanni Āryas, will help us to confirm at least the linguistic irrelevance (and thus the grammatical uselessness) of the idea that the bahuvrīhi is something ‘secondarily’ generated from a ‘primary’ tatpuruṣa or karmadhāraya.