The science of historical linguistics is both fascinating and inexact. The notion of uncovering, branch by branch, a family tree of the world's languages, showing how each is related to every other, and how they all emerged from one common ancestral language, is fascinating. Given the lack of precise evidence - much of this familial development occurred either in an era without writing, or in an era whose writings did not survive for archeologists to find - it is inexact. Linguists work hard to retain the status of a science; some less than rigorous linguists have given ammunition to those who would question whether or not linguistics is truly to be counted among the sciences.
Robert Wright details the methods of linguists as they analyze the systematic changes made as sounds shift within languages to yield new languages, or at least to yield new versions of old languages:
Resurrecting dead and unfossilized languages would be easier if the laws of phonetic change were simple and universal - if, for example, p's always turned into f's, given a millennium or two, and f's were always the residue of former p's. It doesn't work that way. Still, there are vaguer things that can be said about phonetic drift generally. For example, p's often turn into f's - and, moreover, f's almost never become p's. (Try uttering pf, then try fp. Doesn't the second effort feel more like a struggle against biomechanical forces?)
For example, the linguist Jacob Grimm articulated a pattern, noted by earlier linguists, called Grimm's Law or die Erste Lautverschiebung. In this pattern, p's do indeed turn into f's, and the word for foot, which in the most primitive languages begins with a p as in the case of the Sanskrit word for foot, begins with an f in developed languages. Likewise the word for fish. Grimm's Law also reveals patterns in which b's turn into p's, d's turn into t's, and g's turn into k's. Many of these changes can be documented, because they are recorded in some of the earliest samples of writing to survive. But such shifts do not apply with the consistency of algebraic properties. The shifts may occur at different times for different regions or for different sets of vocabulary; there are exceptions of various types.
This complex algebra of sounds gives rise to an intricate science. It is sounds which primarily interest the historical linguist; the written symbols must be understood as merely the conveyors of sound. Writing can be misleading in this regard, for the same th that represents the sounds in thick and thistle also represents the different sounds in lithe and tithe. And, of course, the different letter of the Greek and Cyrillic and Runic alphabets can present the same sounds.
While not strictly universal, shifts such as those described in Grimm's Law provide clues to linguists, because they are directional. In large numbers, p's turn into f's, but not the other way around.
An example of how this directionality comes in handy is the reconstruction of of the proto-Indo-European word for "birch." Here is the word in four Indo-European languages, each drawn from a different main stock: German (Germanic) birke, Lithuanian (Baltic) berzas, Ossetian (Iranian) barz, and Sanskrit (Indic) bhurja. To judge sheerly by numbers, you might guess that the third consonant in the proto-Indo-European word was a relatively soft sound - like the z in berzas or the j in bhurja - and that through a freak mutation it got hardened in the German birke. But it turns out that such a mutation would be freakish indeed. It is common for a "velors" (k or a hard g) to shift into "affricates" (ch or j, respectively), and even to slip further, winding up as "fricatives" (s or z), but the reverse is almost unheard of. So it looks as if the third consonant in the proto-Indo-European "birch" was a hard sound, a velor.
The exact reconstruction is one part of historical linguistics; that reconstruction yields clues, and deciphering the historical meanings of those clues is another part. We can learn about conditions and societies which left no written record, because when we reconstruct their languages, we find that their vocabularies reveal their lifestyles. Societies which lived in the desert or in tropical regions have no word for "ice" or "snow," and civilizations located far from coastlines have no word for "whale" or "jellyfish." William Bennet writes:
Common Indo-European words indicating seasons, flora, and fauna, together with ethnic and geographic data, suggest that the home of the Indo-Europeans was a district connecting southeastern Europe with Asia, probably southern Russia. As the tribes expanded over an increasingly wider area, they became separated into numerous smaller groups, which absorbed varying proportions of other populations. Whether the Indo-Europeans were already of mixed origin is a matter for conjecture: their possession of a common language indicates only that they had been affiliated by social and cultural ties. In the course of the expansion and ethnic mixture, extending over many centuries, the speech of the separate Indo-European groups became progressively divergent, though within each community some degree of linguistic reintegration must have taken place as certain dialects became predominant and others became extinct.
It is a commonplace that the landscape of Russia contains more birch forests than, say, the landscape of North America. And so the fact that the original Indo-Europeans had a word for "birch" is historically significant, even if our reconstruction of their word is only approximate. As Robert Wright notes,
At any rate, the letter-for-letter accuracy of each reconstructed word is in some ways moot. For historians and archaeologists, how the Indo-Europeans pronounced the word for "birch" is less important than that they did pronounce it: apparently they lived somewhere in the vicinity of birch trees. And "birch" is just the tip of the iceberg. There is a whole book called Proto-Indo-European Trees, and countless hours have been spent trying to infer the Indo-Europeans' homeland from their flora and fauna. Maps of Europe and southwestern Asia have been drawn with birch zones, beech zones, beaver zones, and so forth, in hopes of finding a region common to all. It hasn't worked. Argument about the proto-Indo-European homeland persists.
From that homeland, wherever it was, a few groups emerged, wandering hundreds and then thousands of miles in various directions. Those few groups - Germanic, Slavic, Baltic, Indo-Iranian, Romance, Greek, etc. - multiplied as each of them subdivided: the Germanic group yielded Dutch, Swedish, Flemish, Norwegian, Danish, Icelandic, Gothic, German, etc., while the Slavic group yielded Polish, Russian, Czech, Slovakian, Ukrainian, Serbian, Bulgarian, etc.
Historical linguists, in effect, are researching the family tree of civilization and culture.