AbstractIn PIE, quality modifiers were expressed by stative verbs and nominal epithets, rather than by special adjectival lexemes. Adjectives did not form a separate lexical class. This made the encoding of the NP constituency less explicit. If we consider what I suggest calling “second-generation IE languages” we can observe a general tendency to create new, more explicit morphological means of dependency marking within a NP. The exact outcomes of this diachronic process vary from one language to another. However, if we parametrise the variation, a common pattern becomes clearly observable. In all the languages analysed in the present paper, there is a pronoun undergoing grammaticalisation as a dependency marker. What varies is (1) the position of this element with respect to the nominal base (pre- vs. postposed); (2) the degree of agglutination (bound morpheme vs. clitic vs. free morpheme); and (3) the locus of marking (head vs. modifier vs. double or alternant marking); (4) the source morpheme that undergoes grammaticalisation (relative vs. demonstrative pronoun).