Both comparative and internal methods of reconstruction are used to establish earlier stages of language development where no records of that language exist. A proto language is a parent or ancestor language which is presumed to have existed and given rise to daughter languages. The concept of language evolution from a common parent can be shown through the application of both methods and this forms a central tenet of historical linguistics.

Comparative reconstruction

The comparative method establishes genealogical relationships of languages by recognising the regularity of sound changes which lead to the production of daughter languages. This chronological restructuring, which in the case of proto Indo-European (proto IE), marked the development of Germanic and other language groups, is shown in Grimm's Law. Here, the corresponding consonants of the Germanic languages show the regularity of change, which, when followed in reverse, can recreate the parent language.

Words and morphemes of sister languages having similar forms and meanings are cognates and their comparison shows the patterns of sound correspondences. The different sounds of the daughter languages are reflexes of, and give clues about, the parent from which they are descended. Although some relationships are fairly obvious, a majority rule and other methods are necessary to decide with any degree of certainty what elements of the parent would give these reflexes. Regard must be given also to known phonological behaviour to exclude unlikely sound changes. Jeffers and Lehiste (p18) describe the correspondences between Sanskrit (Skt), Ancient Greek (Gk) and Old Church Slavic (OCS) in their words for 'cloud'.

The usual phoneme symbols / / are ommitted. The asterix * signifies a reconstructed word form.

Following the rules mentioned above, the suggested proto language elements are:

So reconstructed proto IE would be *nebhos. Summarised from Jeffers and Lehiste (p18).

The situation is often more complex, however, and several reflexes may result from one element in the parent (a split). Verner's Law shows how different phonological environments cause this and explains the exceptions to Grimm's Law. The example below helps to show how the correspondences of Sanskrit and Gothic dental consonant reflexes derive from fewer elements of proto IE.

Suggested proto language elements (the reasoning is ommitted) are :

Summarised from Jeffers and Lehiste (p20).

The elements *t, *d, *dh, in proto IE have six correspondences in Sanskrit and Gothic. Here *t splits to give three reflexes in Gothic and *dh splits to give two reflexes in Skt. A further complication arises when the opposite process to splitting has occurred, i.e. merging. Here features of the parent language are represented by fewer reflexes, therefore some evidence is lost.

The comparative method is, however, able to refine the tree structure by identifiying languages sharing common features and thus forming a sub-group. Of say, four languages, three may show certain innovations, demonstrating a closer relationship, in the same way that West Germanic languages are more closely related to each other than to North Germanic ones.

Internal reconstruction

In contrast, internal reconstruction begins with synchronic language study and examines discrepancies in word forms and morphemes. These inconsistencies are often remnants of redundant grammatical structures from previous stages in that language's development. Internal reconstruction attempts to recover the earlier condition and demonstrate the process of innovation or change that occurred. This method can be used when no sister languges exist and can even be applied to proto languages.

Differences of pronunciation in the English words serene/serenity, and profane/profanity, are explained by Aitchison (p29) as a result of sound changes. In this case, records confirm that the addition of the 'ity' as an abstract noun ending did not originally require a vowel change and that this occurred only later.

More interesting are examples of different suffixes (allomorphs) added to transitive verbs in Samoan, and to nouns in Ancient Greek described by Crowley, and Jeffers and Lehiste respectively. Internal reconstruction shows that parts of the present suffixes are in fact final elements of the original stems which have disappeared in uninflected forms. Thus Ancient Greek appears to have some irregular genitive endings, but with other noun stems, the usual genitive ending (os) is retained.

Summarised from Jeffers and Lehiste (p40).

So, irregular 't' and irregular 'k' of the genitive endings can be shown to be previous final stem consonants.

Besides the paradigm allomorphy above, discrepancies in statistical occurrence of sounds, asymetry or gaps in correspondence sets and changes in stem vowels are all starting points for internal reconstruction. Saussure studied proto IE verb roots and suggested older features (laryngeals) to account for these. After the discovery and translation of the Hittite language, written evidence confirmed his theory in 1927. This shows the effectiveness of the internal method. However, there are limitations to the internal method, especially when evidence has been lost. Partial mergers and splits leave clues for the internal method but total mergers are irrecoverable.

The two methods of reconstruction examined are important in historical linguistics; the comparative method builds proto languages with some accuracy, while the internal method identifies specific changes in the structure which have contributed to an overall change in the phonological system.

Bibliography:

Aitchison, Jean, Cambridge University Press, 1991, Language Change: Progress or Decay?

Crowley, Terry, University of Papua New Guinea Press, 1987, An Introduction to Historical Linguistics.

Jeffers, RJ and Lehiste, I, MIT Press, 1979, Principles and Mehtods for Historical Linguistics.