The vast interior of Eurasia is a linguistic spread zone—a genetic and typological bottleneck where many genetic lines go extinct, structural types t… - Johanna Nichols

Several kinds of evidence for the PIE locus have been presented here. Ancient loanwords point to a locus along the desert trajectory, not particularly close to Mesopotamia and probably far out in the eastern hinterlands. The structure of the family tree, the accumulation of genetic diversity at the western periphery of the range, the location of Tocharian and its implications for early dialect geography, the early attestation of Anatolian in Asia Minor, and the geography of the centum-satem split all point in the same direction: a locus in western central Asia. Evidence presented in Volume II supports the same conclusion: the long-standing westward trajectories of languages point to an eastward locus, and the spread of IE along all three trajectories points to a locus well to the east of the Caspian Sea. The satem shift also spread from a locus to the south-east of the Caspian, with satem languages showing up as later entrants along all three trajectory terminals. (The satem shift is a post-PIE but very early IE development.) The locus of the IE spread was therefore somewhere in the vicinity of ancient Bactria-Sogdiana. This locus resembles those of the three known post-IE spreads: those of Indo-Iranian (from a locus close to that of PIE), Turkic (from a locus near north-western Mongolia), and Mongolian (from north-eastern Mongolia)... Thus in regard to its locus, as in other respects, the PIE spread was no singularity but was absolutely ordinary for its geography and its time-frame. ... The reason that dialect divisions arising in the locus show up along more than one trajectory is that the Caspian Sea divides westward spreads into steppe versus desert trajectories quite close to the locus and hence quite early in the spread. In contrast, developments that occurred farther west, as the split of Slavic from Baltic in the middle Volga may have, continue to spread along only one trajectory. This is why the Pontic steppe is an unlikely locus for the PIE spread. ...Thus the structure of the IE family tree, the distribution of IE genetic diversity over the map, and what can be inferred of the geographical distribution of dialectal diversity in early IE all point to a locus in western central Asia

The bifurcation of Indo-Iranian is well known to be evident not only in South Asia, where all three of Indic, Nuristani and Iranian sub-branches are found, but also in ancient eastern Anatolia, where either an Indie language or undifferentiated Indo-Aryan or Indo-Iranian is evident in early Mitannian vocabulary (e.g. aika-wartanna ‘one course’, where aika is cognate to Sanskrit eka ‘one’, an Indie word) while Old Persian and Avestan are Iranian...

A consequence of the reconstruction offered here is that the attested distribution of IE then turns out to be no singularity but just one regular episode in a standing pattern of spreads.