Two economics students are attempting to describe a concept of language, but do not know of any formally-recognized terms or research that explain this concept.

They believe there is an underlying "law" that dictates and allows a researcher to predict what someone is going to say (or extremely unlikely to say) before they say it.


Trevor Handy Sowens and an associate were discussing a concept that Trevor has labelled The law of Symbolic Leverage, and Trevor no longer wishes to use his own custom term-of-art to describe this concept.


Trevor has noted that there is an innate behavioral trait in the way people communicate ideas. He discussed it with a close associate and the associate agrees.

The behavioral trait goes like this:

  • 1) Suppose Alice has eaten five apples over the course of one day
  • 2) Bob asks Alice to describe what she has eaten, not knowing anything about Alice's day
  • 3) It is reasonable to expect Alice to say:Today I ate five apples.

  • 4) However, it is not reasonable to expect Alice to say: Today I ate an apple, and then I ate an apple, and then I ate an apple, and then I ate an apple, and then I ate an apple.

  • 4a) The only circumstance to expect Alice to say 4) instead of 3) would be if Alice has some ulterior gain that can be realized if she uses the longer-form expression (e.g., Alice is working on a job where she gets paid more money if she spends more time talking. She gets paid per-word spoken per-day.)
  • 5) Trevor refers to the above circumstance as The law of Symbolic Leverage ... which states that people will use the least amount of symbols necessary to communicate some idea, unless the use of the symbols themselves carries an ulterior purpose, above and beyond simple communication.


  • can this hypothesis of Trevor be substantiated or repudiated, according to the knowledge of a trained Linguist?
  • what preexisting prior term of art or concept would a linguist use to describe this observation made by Trevor?
  • 2
    There are a batch of rules in English that delete repeated material, basically because it's predictable and doesn't contribute any information that's not already in the context -- or, rather, that's not already in what the speaker believes is the context; speakers can be and often are wrong in their assumptions. This is mostly due to a least-effort principle which is common to all living things -- don't waste time, attention, and resources on doing something necessary, unless you get a correspondingly better payoff by doing so. Information is like other resources except it's not fungible.
    – jlawler
    Sep 26, 2016 at 20:09
  • Right. Is there a formal term of art that explains this "least-effort principle" -- specifically in the context of linguistics and sentence construction?
    – dreftymac
    Sep 26, 2016 at 20:11
  • Now that I think about it, this may be the same principle as Tufte's Data-Ink Ratio, formulated for visual presentation of numeric data.
    – jlawler
    Sep 26, 2016 at 20:13
  • In terms of language, though, this is nothing special. Don't waste effort is like don't stop breathing -- they're automatic and built into the machinery; you have to make a special effort to do differently, in most cases.
    – jlawler
    Sep 26, 2016 at 20:15
  • 1
    Oh, yes, that's one source. But check on ecological and evolutionary uses as well; effort has a cost in energy, and energy expended without replacement is a net loss. Biology is basically economic in its concepts; or else economics is biological. Neither should be a surprise, given the etymology of economic -- laws of the home.
    – jlawler
    Sep 26, 2016 at 20:29

3 Answers 3


"Be brief" is the 3rd Gricean manner maxim of conversation. See Gricean maxims.

  • Nice. Thanks for that @Greg Lee. One underlying aspect of the question, though, is a quest to find a prescriptive or analytical basis of substantiating this principle, in addition to the normative or didactic basis embodied in these maxims. Thanks again for the link.
    – dreftymac
    Sep 26, 2016 at 21:00
  • 2
    Grice's account is very well known and much discussed by linguists and philosophers. If you want discussions of the basis for his principle, you can find plenty on the Web. I would say that the basis is that we humans are a social species, ordinarily either cooperating with one another or in conflict (which Grice's account also comprehends).
    – Greg Lee
    Sep 26, 2016 at 21:46

In computational linguistics, language modelling and perplexity measures are used to evaluate the likelihood of phrases (and continuations given a certain start of phrase) as well as the amount of information actually transmitted by an utterance.

What seems relevant to your question is research on typical information density of various languages; I won't go digging for citations now but IIRC a conclusion was that while different languages have significantly different densities of bits per letter, word or spoken syllable, the spoken languages tend to have more or less equal amount of bits per second - if a particular language requires more syllables to say the same thing, it simply gets spoken more rapidly on average.

Furthermore, there is a certain amount of redundancy in natural language, enabling listeners to correct for some amount of errors and noise; and as languages evolve, this amount seems to get preserved (to some optimal value?) - loss of syllables carrying grammatical information implies moving towards a more strict word order; when longish phrases start to get more frequently used, they get replaced by contractions to eliminate that excess redundancy; etc - your example is a sign of excess redundancy, but the observed trends show that even if it was accepted part of language, it's not stable, language users will shed such redundancy in a few generations.

  • Oustanding observations, makes perfect sense. One factor that motivated this question (and Trevor's initial observation) is that there appears to be degenerate cases that favor and stabilize instances of excess redundancy -- (for example, consider the productions of a contract programmer, whose pay and relative prestige may increase, if he can write more obfuscated or turgid code than is necessary) This was the aspect addressed by 4a) in the original post. From a game-theoretic standpoint, this would seem to have merit. Any work in computational linguistics to substantiate or refute?
    – dreftymac
    Sep 26, 2016 at 23:25
  • 1
    There are forms of communication that are less tolerant of errors and would intuitively require more redundancy - one might measure the difference between information density in normal english versus e.g. legal contracts or plane-ATC radio communications. However, hypothesis about how and why degenerate cases occur would be a different field (sociology?), for this computational linguistics can only provide some (more accurate?) measuring apparatus to assist.
    – Peteris
    Sep 27, 2016 at 9:25

A good cogsci-based/"linguistics-adjacent" place to look into is work by Florian Jaeger, Harry Tily, and Tal Linzen (inter alia, all building from work by John Hale and Roger Levy from about a decade ago) on communicative efficiency and the role of preservation/minimization of entropy (and other information theoretic measures) changes in language processing (both in production and comprehension).

Here are a couple of potentially promising papers:



Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

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