How do linguists tell areal features and genetic relations apart?

Question

Languages belonging to the same family obviously share many features, most of which were inherited from their common ancestor. But, considering that languages of the same family also are usually spoken in the same geographical region, some shared features are the result of horizontal spread (areal features).

So, how do linguists tell which is which? What are the clues that linguists take into account to tell whether a certain feature was inherited or spread through an area of linguistic diffusion?

Just to cite one example that comes to mind: Indo-European ablaut. It's generally accepted that ablaut was already present in PIE and that it reflected in its daughter languages (e.g. English sing/sang/sung). How do linguists know for sure that this was not an innovation in one of the early daghter languages that all others have borrowed?

Answer 1

This problem has been in no way settled in historical linguistics and has caused much debate. However, there is some consensus that in the case of language contact there are usually some barriers to diffusion, such as geographical, social, and typological. Alan Dench (Dench 2006) argues that "where languages spring from a common source, their split must be seen as the eventual result of an accumulation of blocked diffusions - waves breaking against geographical and, most importantly, social barriers" (p. 113). Have a look at a nice collection of papers on this topic in Areal diffusion and genetic inheritance: Problems in comparative linguistics , ed. A. Y. Aikhenvald and R. M. W. Dixon, published by OUP in 2006.