Comments on: All languages equally complex

By: Alex

Alex — Wed, 08 Jul 2009 23:16:16 +0000

The idea that all natural languages are equally complex cannot really be called a theory, since linguists only make this claim in passing (usually in the context of making some point in support of a strongly nativist view of language), but if pressed could not say what they mean by “complexity”. John McWhorter received a lot of attention for saying that creoles were “simple”, by which he only meant that they systematically lack certain kinds of morphology. The only attempt I know of to really articulate what complexity would mean is a proceedings paper by Max Bane from 2007, “Quantifying and Measuring Morphological Complexity”, where he defines complexity in terms of an information theoretic measure, Kolmogorov complexity. And it is definitely not known whether languages lose or gain complexity over time, though it is generally assumed that language change is atelic.

I realize the main point of this post is to say something about programming languages, but a lot of the claims about human language are misleading, in the sense that much less is known about this question of complexity than certain comments here suggest.

By: John Venier

John Venier — Wed, 13 May 2009 20:01:59 +0000

More odd than the language in some ways is the culture. Here are a couple of links to periodical articles about them:

http://www.newyorker.com/reporting/2007/04/16/070416fa_fact_colapinto

http://www.spiegel.de/international/spiegel/0,1518,414291,00.html

By: John

John — Wed, 13 May 2009 00:07:50 +0000

John, I looked at the Pirahã link. That is really odd.

By: John Venier

John Venier — Wed, 13 May 2009 00:02:38 +0000

Don't overlook the value of redundancy in natural languages. I believe many seemingly useless language features persist because they add redundancy. I believe this is also true of programming languages, just not in regards to their functional aspect. I seriously doubt that all natural languages are equally complex, but it is clear that they do have differences in types of complexity. And what about "untranslatable" terms, usually inseparable from culture? Obligatory programming reference, then a truly bizarre and controversial language: A common question from C users learning Fortran used to be, "What's the Fortran command to clear the screen and put the cursor in the top left-hand corner?" My favorite answer was, "Surely you mean to ask how do you punch the top left-hand corner of the punch card." The most bizarre natural language I have heard of is Pirahã spoken by the Pirahã people.

By: John

John — Tue, 12 May 2009 04:34:12 +0000

Nick, I think that grammatical gender contributes to complexity, but it has compensating benefits. For example, antecedents may sometimes be disambiguated by gender. Maybe languages with grammatical gender are simpler in some other way. This may be an analog in programming languages. C++ has three operators (->, ::, and .) that all map to just . in C#. Most would say C# is simpler in this regard, but others would argue that C++ is more clear by, for example, making a symbolic distinction between namespace scope and object membership.

By: Nick Barrowman

Nick Barrowman — Tue, 12 May 2009 03:26:42 +0000

As far as natural languages are concerned, what about essentially useless characteristics like gender of nouns? Perhaps linguists don’t consider them as contributing to complexity, but any language learner certainly does!

I wonder if there are analogous “features” in programming languages.

By: sambiomato

sambiomato — Mon, 11 May 2009 20:26:02 +0000

There is many “definitions” of complexity for programming systems/languages :

- The complexity you discussed in this post, we can define it by difficulty to understand syntax, grammar, etc. ex : Perl > (harder) Python, C ~= Java. You can add usability of standard library, etc.

- The computational complexity which said that every programming language (i.e type-1 Chomsky language) are equally complex.

- Kolmogorov complexity is a middle on the intuitive definition and the computational one. The idea is to define complexity of a message (aka source code) by its length on a certain Turing machine M. So complexity of a message depends of the “environment”. However, there isn’t any total order.

Programming language are “simple”. I don’t trust the intuition to define if one or an other is more complex. I think there is two questions to ask :
- Firstly, when you are in the real world, it’s clear that ASM isn’t the best choice for programming Rich Interface Application. Conversely, using ActionScript to numerical simulation for jetplane design is dangerous. So, complexity of a language depends of goals.
- Secondly, the theorical question : Can I write anything in any language (again, chomsky type-1) ? have an answer : yes, you can.