Breaking out of the 'derive everything' trap?

Question

Conlang guides often recommend you derive words wherever you can, only generating new roots when no form of compounding can work.

Problem is, as the 'natural meta language' demonstrates, most words can be 'derived' from just a few dozen words. When I try to think of derivations, I keep falling into an endless train of deriving whole chains of words. Even if that doesn't happen though, I still find I keep deriving words too much, creating a conlang that's more repetitive than even Esperanto.

English maybe gives me the wrong instincts, since it relies more on borrowing loanwords from other languages that coining its own. However, even in German, which relies far more on internal derivation, somehow its not nearly as bad as conlangs tend to be.

Another problem is I find that deriving words through morphology results in ultra long words (the same problem that Volapük suffers from).

What I think I need is some more esoteric derivational morphology that keeps things from getting too repetitive. Being able to shorten compounds would also help to avoid the overly long word problem. Or maybe I should just rely more on a random word generator and not care so much about the forms of my words. I know that tends to generate bad results, like common words being too long, no coherent sound symbolism of any sort, or sometimes too many words due to the user preferring to use the program over coining their own words. Still, I think it may be the best way to avoid making yet another oligosynthic language essentially.

Is there any better way to deal with this problem?

Answer 1

Well, concatenation isn't the way words get derived is it? That leaves semantic shift out of the equation, for instance - instead of just "X could turn into Y if only given more morphemes", it could also be true that "X could turn into Y if only given enough time", or that "X could turn into Y if only given enough metaphorical extension".

If you can't get out of the headspace of simply deriving everything, try forcing yourself to not use derivational morphology or compounding, and ask yourself how you can reuse an existing root as-is instead of always needing to pile more affixes on top.

I mean, the real, sustainable long-term answer is to get comfortable with not deriving everything. This advice to only coin new roots when no form of compounding will work - I've heard it too, and it's bullshit. There's a grain of truth* to it, but it's so poorly articulated that it causes far more harm to beginning conlangers than it averts. I have seen one derive a word for "farmer" as "one-who-causes-to-become-large". You know how actual, real languages, like English, derive it? As "farm-er". "farm" is just a root in and of itself. Could it theoretically be broken down further? Sure. But "it could be broken down further" does not imply "it must be broken down further", which you can tell because English demonstrably doesn't break it down further.

*The grain of truth is that these roots are often, in turn, derivatives of earlier forms, even earlier compounds, that simply got reduced so much by sound change or diverged so much in semantic drift, that they got reanalyzed as a new thing in and of themselves. "lord" < hlafweard "loaf-ward; keeper of the bread", evolving into a new root that in turn gets used in compounds like "lordship" or "edgelord", is one of my favorite examples. So it is, sort of, broadly true that it's derivation all the way down. Natural languages though spread this derivation out in time and semantic space. It's certainly not necessary for everything to be transparently derived synchronically.

I don't know if this analogy will help you, but - when studying a foreign language, you're typically taught to translate the meaning of a sentence, not the words of a sentence. Well, what if that meaning is encapsulated in a single word? If there's a single word that maps to that idea, then fixating on how, instead, you could break that idea into component parts is - mistranslating? It's perfectly fine to have single words for concepts that English does not have a single word for; just because English circumlocutes, doesn't mean you have to. One of my languages has a single devoted word for "to have one's heart beat" (lak’salos). This derives straightforwardly - with some verbalizing morphology but no compounding - from a proto-root *læqʃ- meaning "heart" (nothing about "beating", actually), that yields a verb for "to encourage; to hearten; to rally troops wavering or in retreat back to line of battle; to dispel fear or angst" (lax̌šʷanla) in another descendant of the same proto. Notice how I'm giving multiple possible English meanings - that's because the word (lax̌šʷanla) fundamentally maps to a concept, an image I have in my head. It's a concept that could be decomposed into component parts multiple different ways, but the thing I'm trying to translate isn't the component parts of the concept, but the concept itself. Another one my languages has not just a single word, but a single monosyllablic word (ki) for "the divine right to rule".

Or maybe I should just rely more on a random word generator and not care so much about the forms of my words. I know that tends to generate bad results, like common words being too long, no coherent sound symbolism of any sort, or sometimes too many words due to the user preferring to use the program over coining their own words.

These problems all strike me as... very self-imposed? I use a word generator when coining new words (originally Awkwords, now a word generator of my own creation), and nothing compels me to use the first word it spits out. Quite from it - I'll have it generate hundreds of words at a time and ones I like, I'll copy and paste in a "word bank" - an Excel file of words that have pre-vetted for sounding good. Then when I need to coin a new word, I'll peruse the word bank for inspiration, for which one has the right, ineffable vibe for the meaning I'm looking for. I can get quite picky about which words I'll accept from the generator - for every 100 words generated at a time, usually 1-2, maybe 3, end up in the word bank. (It's trivial to generate another 100, after all - I just click "Generate" again. The words are not exactly in short supply.) Even then some of the words in the bank just sit there never getting used, because it never ends up feeling like they have the matching "vibe", and I'll just go and generate thousands more words instead.

So the suggestion to me that random word generators confined you to words that are too long or not having the right symbolism, sounds nonsensical to me, when I am exercising discretion at every step. Ultimately I'm the one choosing the words - arguably I'm way too choosy, but at any rate, nothing about the use of a word generator is somehow forcing my hand into using words I don't want to use.