Miekko's recent observation
on his blog that conlanging suffers from the problem of <there are a ton of ""phonologies"" everywhere with nothing behind them> got me thinking.
Not because the sentiment is particularly new or anything. The entire aforementioned blog is based on this premise, and it has been a kind of quiet consensus in my circle for quite some time. Rather, it struck me that this is only one particular aspect of the problem, and the esteemed blog author has been quietly writing based on this but never said it aloud.
The problem is that a lot of conlanging is done in empty generalities.
The standard procedure, as recommended by all
the manuals, is to start with some typological framing and then work inwards; the phenomenon of the phoneme table is the most extreme end of it, but the next steps are not much better: decide dominant word order, inflectional typology &c. &c.
It is of course clear to the linguistically knowledgeable reader, which at the time of publication will presumably be all of them, that such typological identifiers are often of questionable validity and even where there is good reason to believe that they constitute useful information, from a conlanging perspective they still tell you almost nothing about the language whatsoever. There are, on conservative estimate, around 2500 SOV languages in the world. Given that these are distributed across multiple continents and a score of families, it is clear that such information is useful perhaps to the typologist, but not to the conlanger, who is usually interested in describing a single language rather than thousands.
And yet, many conlangers seem principally to be creating entries in typology databases. Is it any wonder, then, that so few conlangs are at all notable? If the information you present to me is roughly equivalent to what I might read on WALS, I will spend about as much time on it as I would on WALS. The best conlangs all share the singular feature of detail
And the interesting bit is that it is pretty irrelevant how many details you have. Having just a few interesting quirks specified pretty much immediately translates into memorability. Consider one of our more Nortorious members. Most people's conlangs, when put into one sentence, would return something like "It has 25 consonants, ten vowels, is SVO and isolating." With this person, it is "it fucking inflects nouns for the TAM of the dominating verb @#!%!". The same goes for any good conlanger.
And the best part is that it takes extremely little work to satisfy this criterion. Any of miekko's miniature conlangs does, and that blog both has very many entries and makes a point of being bite-sized. Expanding such a miniature conlang to the point that it would already rank as "very complete" by the standards of conlangdom is similarly not very hard, since general structures can be derived from details quite a bit more easily than the other way around. All that you need is to realize that it is perfectly viable to start with the details. Not a lot of people do.