Abstract The Spanish preposition para arises from fusion of older por followed by a, via usage-based chunking (Bybee, 2010) associated with the frequency of the sequence. At an early stage, semantic compositionality involving an independent meaning contribution from the preposition a is discernible in allative uses with a destination that is a person. A general decline of allative uses of para with a nominal complement, for both person and place destinations, ensues after the 14th century. In a second change beginning in the 18th century, para replaces por with infinitive complements to become the majority purposive variant. Thus, allative-to-purposive grammaticalisation (Heine & Kuteva, 2002) occurs through changes both in para’s distribution across contexts and in its variation with the older preposition.