In historical linguistics, it is obvious that the general principles of phonetic changes (which some, following Haudricourt, call ‘Panchronic Phonology’) are much better understood than those operating in morphology, and that historical morphology is generally better understood than historical syntax (a field which is however undergoing important developments in recent times with articles such as Eythorsson and Barðdal 2005).
Reconstructing the semantics of proto-etyma (or proto-roots), proto-morphemes or proto-constructions is to some extent even more difficult, especially in the case of families without ancient written sources. The comparative method is poorly equipped to handle semantics, and other methods need to be developped. Even in the Indo-European family, reconstructing the semantics of proto-etyma is often partly subjective for unattested stages. Even if the phonetic shape and the morphological structure of an etymon can be reconstructed through the comparative method, its exact meaning is difficult to ascertain if the daughter languages disagree (and even if they agree, we have no guarantee that they did not undergo parallel semantic shifts).
In the case of language families with ancient sources, it is possible to follow the written record of one language through centuries and even millenia, and these precious examples provide (almost) unambiguous evidence concerning possible semantic changes and their directionality.
One way to establish general principles on the directionality of semantic change is to collect and classify all attested examples from families with ancient written record. Unfortunately, this method has a severe limitation: for the great majority of the world’s languages and language families, there is no reliable documentation before the 20th century, and almost all languages with ancient record belong to a few families from Eurasia. Such an endeavour is indispensable, but on the other hand too restricted to have a general applicability.
I recently found a very useful article by Matthias Urban in the first issue of the new Journal of Historical Linguistics, which is freely available on Benjamins’ website:
This author proposes a new method to use data from families without ancient written attestation in a general typology of semantic change. This method is interesting, because it avoids circularity and subjectivity, and is nevertheless relatively easy to apply on any well-described language.
Urban proposes that the “asymmetry in overt marking is a clue to the directionality of semantic change”. For expressing two related concepts A and B (for instance “skin” and “bark”), languages have the following logical possibilities, but in many cases one possibility is not attested:
1) Distinct unrelated etyma for A and B > English “skin” and “bark”
2) Using only one word for both A and B > languages where a single word refers to both “bark” and “skin”
3) The concept A is expressed by a simple word, while the concept B is expressed by a compound word including the root which refers to A. > languages with a word “skin” (which cannot be applied to bark) where the concept “bark” is expressed by a compound word “tree-skin”.
4) The reverse of 3) > languages where the concept “skin” would be expressed by a compound word such as “man-bark” or “body-bark” (unattested)
Urban observes that cases such as “bark” and “skin” are quite widespread, and concludes that, the ability or inability of a concept to be expressed by a complex form are indicative of the directionality of metaphorical mapping, and therefore of semantic change.
He collected a database of including 50 concepts accross a large language sample, and established an impressive series of “networks of asymmetries in overt marking”, in other words directed graphs in which the direction of the arrows indicate that the concept right of the arrow can be expressed by a compound including the concept left of the arrow. (for “skin” and “bark”, the minimal graph would be skin > bark).
Then, he compared his findings with data for families with a written record such as Indo-Aryan, and found congruent results.
Urban’s non-subjective method to determine the directionality of semantic change is a welcome addition to existing methodologies, and it is hoped that it will be applied on a broader set of concepts in the future.
Urban’s test case was on concepts that are generally expressed by nouns in the great majority of languages (natural phenomena and body parts). I think that applying this method to verbs (I mean, to concept generally expressed by verbs in most languages) will be equally successful, but may need to be adapted to some extent, as languages tend to have fewer compound verbs (except those which allow incorporation). On the other hand, many languages in diverse families have for instance directional, instrumental affixes or lexical affixes, which often change the meaning of the verb root in unpredictable ways. Thus, from the German example machen > aufmachen /ausmachen, using Urban’s method we can build the graph:
close < make, do > open
I surmise that examples of verbs meaning ‘make, do’ build out of a compound including a root meaning “close” or “open” are not attested, and that the semantic changes in question are unidirectional (as a parallel example for ‘do’ > ‘close’, in both Tibetan and Japhug Rgyalrong the verb meaning ‘close’ (the door) originates from ‘to do’).
Comparing meanings such as “to take” cross-linguistically is however much trickier than body parts, especially if Athabaskan languages are taken into account. These meanings are much more abstract, and it is by no means obvious how many basic meanings need to be postulated in the first place.
Eythórsson, Thórhallur and Jóhanna Barðdal 2005. Oblique Subjects: A Common Germanic Inheritance. Language 81–4:824–81.
Haudricourt, André-Georges, 1940, Méthode pour obtenir des lois concrètes en
linguistique générale, Bulletin de la société linguistique de Paris. 41.1:70-7.
Urban, Matthias 2010. Asymmetries in overt marking and directionality in semantic change. Journal of Historical Linguistics 1.1:3-47.