Jörg Rhiemeier's Conlang Page

Introduction

There are many different types of conlangs, and how to classify them has been subject to debate for many years. There is yet no general agreement on which kind of classification is the best. I shall discuss the various kinds of classification here.

Obviously, there are many possible "classifications" that most people would consider meaningless. Things such as the first letter of the conlang's name, the background colour of the official flag or emblem, or the country of birth of the author may be interesting to some people, but they do not really say much about the structure or the purpose of the language. Two languages may share the initial letter "Q", or a green flag, or being both made by a Swede, and still be utterly different from each other. This shows that more meaningful distinctions are in order here.

Classification by purpose

Constructed languages have been made for very different purposes. Some are meant as a means of communication between people of different native languages; others are languages of fictional peoples; yet others serve to demonstrate some kind of "ideal" languages; and many more. It is now generally agreed that classification by purpose is a meaningful thing to do. Usually, three major purposes are distinguished:

1. The auxiliary language or auxlang: a language meant to bridge communication barriers between people of different native languages. One can further distinguish between:
   1. Global auxiliary languages meant for worldwide use. Examples: Esperanto, Novial.
   2. Regional auxiliary languages meant for use in a particular region. Examples: Afrihili, Folkspraak.
2. The artistic language or artlang: a language constructed as a piece of art. These can be broken up into:
   1. Fictional languages: languages of imagined races or nations, such as Quenya, Sindarin, Klingon or Verdurian.
   2. Personal languages: languages that embody a personal linguistic ideal. Example: Vorlin.
   3. Religious and magical languages are also best placed here, better at any rate than among the auxlangs or engelangs. After all, much art has the purpose of celebrating the higher powers. Examples: Bala-i-balan, Enochian.
   These classes, however, overlap: Quenya was Tolkien's ideal language (and can thus be considered a personal language) as much as it is a fictional language.
3. The engineered language or engelang: a language designed to meet objectively testable goals. Three main groups can be distinguished:
   1. Philosophical languages: languages which embody a particular philosophy. The term is used most often for taxonomic languages which derive their vocabulary from a general taxonomy of ideas. Examples: Ars signorum, Ro (taxonomic languages), Toki Pona.
   2. Logical languages: languages which implement a system of formal logic. Example: Lojban.
   3. Experimental languages: languages which serve to test a hypothesis or idea. Example: Ithkuil.

The Gnoli Triangle

However, these three major categories (and their subcategories) are not three boxes with fixed boundaries into which all conlangs can be neatly sorted. There are many overlaps and "in-betweens". For instance, the boundary between artlangs and engelangs is not easy to draw. There are many fictional languages and other artlangs that embody elements of engineering. The conlangs of Henrik Theiling may be mentioned here. Likewise, there are intersections between auxlangs and engelangs, as the examples of the 17-th century taxonomic languages and, more recently, Lojban demonstrate. Even the categories of auxlangs and artlangs overlap, since fictional languages from some imaginative fiction settings have been suggested as international auxiliary languages. Hence, there are intermediates between auxlangs and engelangs, between artlangs and engelangs, between auxlangs and artlangs and even between all three.

This leads to a space in which conlangs can be located - the Gnoli Triangle (named after conlanger Claudio Gnoli, author of Liva, itself a language that doesn't neatly fall into one of the three classes). The original Gnoli Triangle had a "loglang" vertex rather than an "engelang" vertex, but And Rosta suggested changing this, and it has caught on. The triangle looks like this (the colour scheme was added by Raymond A. Brown):

The Gnoli Triangle. Image by Raymond A. Brown.

One could say, for instance, that a language is "70% artlang, 20% engelang and 10% auxlang", and accordingly position it on the triangle.

Classification by structure

Besides purpose, conlangs can be classified by their structure. Generally, the whole raster of linguistic typology can be applied to conlangs, distinguishing analytic and synthetic, fusional and agglutinating, head-marking and dependent-marking, head-initial and head-final, accusative, ergative, active/stative etc. languages, and so on.

There are, however, some structural classificational parameters specific to conlangs. Perhaps the oldest of these schemes is the a priori/a posteriori distinction; another is the degree of naturalism, which, however, means different things in auxlang and artlang circles.

A priori vs. a posteriori

This classification was developed in the early 20th century for auxlangs, but it can in principle be used for any kind of conlang. The key point here is the origin of the lexicon. An a priori language consists of words that were coined afresh and do not originate in any other language. An a posteriori language uses words from one or more existing languages (usually natlangs). The 17th-century philosophical languages were a priori; their words did not originate in natlangs, but were derived from a taxonomy of ideas. Esperanto and most of the later auxlangs, in contrast, are a posteriori: they use words from various European languages.

Not all conlangs, however, can easily be placed in one or the other of these two classes. There are many mixed languages, which use partly a priori and partly a posteriori vocabulary; in some languages, such as Volapük and Lojban, the vocabulary is technically a posteriori but disfigured beyond recognition such that the language appears to be a priori.

This classification is more common with auxlangs than with artlangs or engelangs, and auxlangs are where it is most useful, as the origin of the vocabulary is considered a major factor in assessing the cultural neutrality of the language. Yet, artlangs and engelangs can of course also be classified according to the origin of the vocabulary. (Some types of engelangs, however, practically require an a priori vocabulary.)

Naturalistic vs. schematic

This classification also originates in the auxlang scene, but it is also used for artlangs (rarely, as will be seen, for engelangs) - and means something different there.

When auxlangers speak of a naturalistic language, they mean one that is radically a posteriori: a language which does not only use lexical roots extracted from source languages and subjects them to an artificial word-formation mechanism (as does Esperanto), but which extracts complete words from the source languages, thereby admitting irregular, unproductive derivations for the sake of recognizability of the vocabulary.

In the artlang scene, the definition of naturalistic has nothing to do with the origin of the words. A naturalistic artlang can be a priori as well as a posteriori. A naturalistic artlang is defined by its resemblance to natural languages: it could be mistaken for a natlang. Naturalism is a goal often pursued in fictional languages; examples range from Tolkien's conlangs to most of the creations of the Internet artlang scene.

A subgroup of naturalistic artlangs are diachronic artlangs. These do not only attempt to resemble a natlang at a particular point of time, but emulate the diachronic development of the language. J. R. R. Tolkien was the eminent pioneer of the diachronic method in artlang creation. Quenya and Sindarin are not just similar - they are derived from a common ancestor, Common Eldarin, by an imagined history of sound change and other diachronic processes.

Neither of these two meanings of naturalistic is usually applied to engelangs, as these usually require a divergence from patterns found in natlangs by virtue of their design goals. In fact, a "naturalistic engelang" is often considered a contradiction in terms.

Miscellaneous

There are other possible structural distinctions that cut across the Gnoli Triangle:

• Taxonomic languages are languages in which the lexemes are constructed along a taxonomy of idea (somewhat like the decimal classifications used in libraries, only with phonemes instead of digits).
• Arithmographic languages are a related concept, first envisioned by G. W. Leibniz (who, however, did not coin the word "arithmographic" - this is my own invention). In an arithmographic language, prime numbers are assigned to semantic primes. By multiplying the prime numbers associated with the semantic primes contributing in a concept and encoding the resulting number in speech sounds, you get the word for that concept. (For example, if 'life' is assigned the prime number 2, all living beings receive even numbers and all non-living things receive odd numbers.)
• Oligosynthetic languages are ones with no open lexical class. All words are built from a closed set of morphemes. This gets close to the taxonomic languages as most words are compounds which betray the semantic field to which the word belongs.
• Self-segregating morphology: The morphemes of the language are shaped such that one can always tell for sure where morpheme boundaries are.
• Speedtalk is the idea of a language in which each morpheme is exactly one segment long - there is a 1:1 correspondence between phonemes and morphemes. Obviously, you need a very large phoneme inventory and a very small morpheme inventory, i.e. the language must be oligosynthetic. The concept dates back to Robert A. Heinlein 1949 science fiction story, "Gulf".
• Briefscript is a similar concept as speedtalk: a conlang that is especially concise in written form. This need not be a speedtalk; this may indeed be a language with a small phoneme inventory that uses the Latin (or some other) alphabet as a syllabary (see Ray Brown's Briefscript project).

All of these work best with a priori vocabularies, and are considered unnaturalistic.

© 2012 Jörg Rhiemeier
Last update: 2012-09-17