I'm looking for laws (hypotheses, presumed laws) of semantic change such as the following:

(a) law of differentiation: nearby synonyms tend to diverge in meaning over time

(b) parallel change: words with related meanings tend to change in similar ways over time

I wonder if someone could name me similar laws/hypotheses, or provide some literature links. I'm also interested in quantitative linguistic laws relating to semantics, if any exist.

  • I can live with long and with opinion based answers.
    – mathse
    Nov 13 '15 at 16:18
  • 1
    Um, sorry. Please try to be helpful for someone else, not me. There is a well-defined inventory of principles of sound change. I'm asking for principles of semantic change, not arbitrary opinions.
    – mathse
    Nov 13 '15 at 16:23
  • 2
    Experts don't theorize without data. It's easy enough to come up with Greek names for any kind of meaning extension you can think of, but they're not general. I can tell you that metaphors, especially metaphors based on the human body, human body parts, and human activities, form the vast majority of terms and change in understandable but not predictable ways. For more, check out Lakoff and Johnson.
    – jlawler
    Nov 13 '15 at 19:57
  • 1
    @mathse take a look at most recent research by Anna Zalizniak, e.g. Zalizniak et al. 2012
    – Alex B.
    Nov 13 '15 at 20:57
  • 1
    Especially the reference "The catalogue of semantic shifts as a database for lexical semantic typology" by Zalizniak et al. seems quite helpful. I'll have a closer look into it.
    – mathse
    Nov 14 '15 at 18:48

This is a good question but only in the sense that it opens a possibility for rejecting the very premise on which it is based.

The short answer is, there are no laws formulated for linguistics that could be analogous to some of the laws of physics that would be constant through time and space. The last attempt at this were the sound change laws in the 19th century and those do not take us very far.

The closest linguistics (in its broadest sense) comes to this is a search for universals. There are many debates about what is and can be universal in linguistics - most of them happening in the realm of syntax (with some phonetics). But even the most formal of these universals (as represented by Universal Grammar) do not claim the status of a 'law' with the same predictive modeling power as laws of thermodynamics.

In semantics, universals are viewed with suspicion. Perhaps the most stark statement of universality would be the semantic primes theory which posits certain core meanings that are shared by all languages - based on empirical research of a large (but still limited) sample.

Leonard Talmy outlines some universals in his chapter on Universals of Semantics but they are very much linked to broader universals of cognition or conversely more constrained combinatory universals from morphology. There are also many broad but not universal tendencies such as the use of space to model time, etc. Some of these were proposed by Emmon Bach in his reaction to Everett.

The 'laws' you propose have been formulated but they could never achieve the status of invariant principles let alone something which you could build predictive models on (despite computational attempts at such a thing). The principal problem with them is that they are formulated with the assumption that 'meaning' is something that 'words' have in the same way you see in the dictionary. But meaning cannot be described statically in this way. It is a process that includes a complex networks of usage patterns that are constantly shifting. They are sufficiently stable to allow for communication but not stable or discrete enough to enable prediction or even a foundation for some sort of a retrospective model.

However, some basic universal semantic processes could be formulated relatively easily:

  • Categorization (cat refers to a class)
  • Hierarchical organization of semantic networks (cat is an animal)
  • Use of figurative processes (analogy, similarity)
  • Presupposition and speech acts (traditionally pragmatics)
  • Polysemy
  • Use of human bodily experience for conceptualization of abstract concepts (at least some)
  • Metalinguistic awareness (with huge variation across speakers)
  • Ability to express basic logical relationships (not, and, or - but not necessarily lexicalized in the same way)
  • Presence of certain semantic classes (e.g. kin terminology, words for food)

However, most of these are universal by virtue of being axiomatic because without them communication or reasoning would be impossible. For instance, if we found a language that did not use categories - we would find ourselves in an alien world beyond anything imaginable.

While these do not relate to change, you could expect change to happen within these constraints.

  • Thanks for your answer, Dominik. I disagree with your outlook, however. First, there are laws in linguistics (or linguists claim there are). The relevant subfield of linguistics is called "quantitative linguistics" (see "Journal of Quantitative Linguistics"). Prominent names are Altmann, Köhler, Best, etc. These people do claim that their laws hold across space and time. I also feel that most researchers from the subfield of computational linguistics (CL; and which I am quite familiar with) would disagree with your statements.
    – mathse
    Nov 16 '15 at 9:07
  • Secondly, that meaning is embodied in networks of linguistic units is, on the one hand, a standard approach to semantics in CL and does, on the other hand, not at all prevent semantics from being predictable. However, thanks a lot for your insights.
    – mathse
    Nov 16 '15 at 9:07
  • Hi @Dominik Lukes. There are now (at least) two high-profile papers on laws of semantic change, rejecting the very premise on which your answer is based. See arxiv.org/pdf/1605.09096v3.pdf and aclweb.org/anthology/P/P16/P16-2009.pdf
    – mathse
    Aug 24 '16 at 10:19

Not the answer you're looking for? Browse other questions tagged or ask your own question.