Comparative method

The purpose of comparative method is to detect historical relationships between languages and to establish a consistent relationship hypothesis by reconstructing:

• the common ancestor of the languages in question,
• a plausible sequence of regular changes by which the historically known languages can be derived from that common ancestor.

The essential steps are as follows:

• Relationship between two (or more) languages can be suspected if they show a number of regular correspondences in lexicon, which means that there is a regularly recurring match between the phonetic structure of words with similar meanings (one usually begins with characteristic sets like family terms, numerals, body parts, etc.). The notion of regular correspondence is very important here: mere phonetic similarity, as between English day and Latin dies (same meaning), has no probative value. English initial [d-] does not regularly match Latin [d-], and whatever sporadic matches can be observed are due either to chance (as in the above example) or to borrowing (e.g. Latin diabolus, English devil, both ultimately of Greek origin).

There is, however, a regular correspondence between Latin [d-] and English [t-]:

decem | ten

duo | two

duco | tow

Old Latin dingua | tongue

Closer analysis reveals that the correspondence is both regular and pervasive, and that it is part of a more general regular pattern (Grimm's law)

More trivial equations also hold between Latin and English:

mater | mother

ment- | mind

mus | mouse

They demonstrate that Latin word-initial [m] corresponds to English [m]. However, it is the regularity of the matches, not the identity of sound, that counts here.

• A really systematic correspondence can hardly be accidental. If we can rule out alternative possibilities like massive borrowing, the correspondence can be attributed to common descent. If there are many regular correspondence sets of this kind (the more the better), and if they add up to a sensible pattern (one that could have been produced by known types of sound change), common origin becomes a virtual certainty.

• On the basis of regular correspondence sets formulate a relationship hypothesis, involving an attempt to reconstruct the hypothetical ancestor of the languages being compared. Without going into detail, Latin [d] and English [t] are both derived from primitive *d (the asterisk means that the sound is inferred rather than historically documented) in the reconstructible common ancestor of both languages (called Proto-Indo-European or PIE for short). We also attempt to recover the past sound changes responsible for the historically known reflexes of the reconstructed protoform. For example (the symbol > should be read as "became"):

PIE *dek^{^}m > Proto-Germanic *texun > Old English teon (attested, yielding Modern English ten)

PIE *dek^{^}m > Proto-Italic *dekem > Latin decem (c = /k[?]/ in Classical Latin)

PIE *dek^{^}m > Proto-Indo-Iranian *daCa > Sanskrit das′a

PIE *dek^{^}m > Greek deka

Each step must be justified, e.g. *k^{^} > *x (the sound of German ch) is part of a regular pattern seen also in Latin cord- | Germanic *xert- 'heart' (> English heart, German Herz) and many similar equations. The weakening and loss of this *x between vowels in the history of English (*-x- > *-h- > zero) is also regular. So are other changes visible in these word histories, e.g. the development of the syllabic nasal at the end of the word into Greek and Indo-Iranian [a], the change *e > *a (or rather the falling together of *e, *o and *a) in Indo-Iranian, or the so-called Satem development of *k^{^} in the same group (giving a Sanskrit palatal fricative via an Indo-Iranian palatal affricate).

Regular sound changes form historical sequences and often "feed" one another (an older change creates an environment in which more recent changes apply).

• The reconstruction of proto-sounds and their historical transformations enables us to proceed futher: we can compare grammatical morphemes (word-forming affixes and inflectional endings), patterns of declension and conjugation, and so on. The full reconstruction of an unrecorded protolanguage can never be complete (for example, proto-syntax is far more elusive than phonology or morphology, and all elements of linguistic structure undergo inevitable erosion and gradual loss or replacement over time), but a consistent partial reconstruction can and must be attempted as proof of genetic relationship.

(See also lexicostatistics[?] and Morris Swadesh)