We summarize, here, the evolution of Proto-Indo-European (PIE) into Indo-European (IE) language families as formulated by T.V. Gamkrelidze and V.V. Ivanov (henceforth G&I) in their book, Indo-European and the Indo-Europeans (Berlin & New York: de Gruyter, 1995, 2 vol's), translated into English by Johanna Nichols. Our own IE Maps page sketches late language evolution within the ten major IE families, beginning ca. the second millenium B.C. Here, in contrast, we use ideas developed by G&I to sketch the earlier evolution of PIE into the IE language families. In particular, G&I use isoglosses and typological principles to propose and explain a sequence of [geographically influenced] splits/groupings of PIE and its dialects as speakers expanded from their original homeland into new territories and gradually lost regular contact with one another: G&I show IE language groups emerging and splitting, evolving into 9 [of the 10] "family proto-languages" that we start with on our IE Maps page.
The evolution of Proto-Indo-European summarized here began approximately 6,000 years ago; our summary covers 2,000 years or more of pre-history and early history, ending ca. 2,000 B.C. G&I provide very little in the way of dates for their seven (7) stages of PIE-IE language evolution, and here we hazard no further guesses of our own: we simply use their numeric Stage identifiers. It is interesting that G&I use their findings to infer a homeland in eastern Asia Minor or northern Mesopotamia, rather than in the Pontus as so many others have concluded. The reader is referred to G&I for details.
In the table below, each entry in stages 1-7 refers to an unattested proto-language that later evolved into a set of related languages. In their own diagram, G&I do not use the prefix "Proto-" (except in "Proto-Indo-European") and we likewise avoid its use here; thus each table entry may simultaneously be viewed as a language and as a group/family. In discussion, below, we may use the term "macro-family" to refer to a proto-language plus its descendants. For the reader's convenience, macro-family names may be reiterated in the Stage 7 column.
Stage 1 represents a time when Proto-Indo-European (PIE) is a single language, spoken in the homeland. Stage 2 represents a time when PIE, while still a single language, has begun to evidence signs of emerging dialects (G&I label them Area A and Area B, not shown here) that would show their effects in later macro-families.
In stage 3, (Proto-)Anatolian has emerged as a distinct macro-family, with all other PIE dialects remaining less differentiated. In stage 4 the latter group has split, one subgroup comprising Italo-Celtic-Tocharian and the other comprising all the rest (later to become Hellenic, Armenian, Aryan, Balto-Slavic and Germanic).
In stage 5, the two groups that emerged in stage 4 have themselves split, resulting in Helleno-Armeno-Aryan, Balto-Slavic-Germanic, Italo-Celtic, and the Tocharian macro-family.
In stage 6, the Hellenic macro-family has split from Armeno-Aryan, the Germanic macro-family has split from Balto-Slavic, and the Italic & Celtic macro-families have diverged. Also in stage 6, "satemization" is observed in Armeno-Aryan and [geographically adjacent] Balto-Slavic.
In stage 7, Armenian has split from Aryan (a.k.a. Indo-Iranian). G&I do not pursue the evolutionary lines any further; for example, they do not show the split of Aryan (Indo-Iranian) into the Indic & Iranian families, nor the split of Balto-Slavic into the Baltic & Slavic families. Thus they trace the emergence of 9 IE language families from PIE.
G&I pay little attention to the Balkan languages -- though, to be fair, this does not distinguish them from the vast majority of linguists: there is simply not enough data to support hard conclusions. However, most of the Balkan languages fall into the Satem category and it is easy to see places where they might possibly be located in the table above, e.g. just above Armeno-Aryan in stage 6, which allows one to infer some things about their earlier evolutionary histories.