So I've been aware of conlangs for a very long time and have wanted to make my own for almost as long, but I keep getting stuck at what most conlang resources suggest ought to be the first step, which is phonology and particularly phonotactics. I have all these great ideas for syntax and morphology and all that stuff, but then I realize that I ought to have some sort of idea of how my words have to sound, and then I end up falling down a very deep rabbit hole and getting frustrated and not implementing any of my brilliant ideas.
I feel like I might be able to enjoy creating a phonology for my language, but only after I've gotten some of the other stuff figured out. I kind of just want to skip this first step, and I suppose I could manage it in a couple of different ways:
1. Just forget about creating any kind of structure or rules and just use the sounds I want, and then figure out the rules I've intuitively created
2. Just forget about creating any kind of structure or rules and just use the sounds I want, then create a set of rules that somewhat fits those that I've intuitively created, then go back and revise everything to follow those rules
3. Steal the phonology of a natural language, maybe modifying it slightly
4. Suck it up and create my own phonology, but a really easy, boring one, and if I want to do something interesting with it I can do it later
The first two options are probably likely to give me a headache in the long run, and may result in a phonology that's embarrassingly noobish or Englishy. The second two options feel like cheating, and are more likely to give me a language that doesn't sound the way I wanted it to. Any suggestions, encouragements, or admonitions? Thanks.
I like to start with the most interesting parts first; so I like basically to go in the opposite order to that mentioned in Zompist's "Language Construction Kit".
Start with biclausal and multiclausal sentences; then syntax; then morphology; then phonology.
If you want your 'lang to look real, you need to have a "set of rules". But natlang's "sets of rules" were deduced after the fact, by linguists who studied them. So do the same thing; make up a corpus for yourself of several short conversations or short paragraphs, give each either a translation or a phrase-structure (an unlabeled phrase-tree-diagram), and then decide what the rules are by seeing what fits what you have. That's options 1 and/or 2.
OTOH you can get part of your language to be a realistic and naturalistic looking-and-sounding part, by "stealing" that part from an actual natlang; whether that's the phonology, or the morphology, or the syntax, or whatever. That's simple to do, and the results are naturalistic and realistic, though they may sound "newbishly English-like" (or newbishly whatever-language-like). So that's option 3.
And, if your purpose in designing the language is to investigate one particular feature, or to investigate how two (or even three) particular features interact with each other, it makes sense to keep all the other features of your 'lang as "average" and "plain vanilla" as possible. So that's option 4.
I'll bet there are other options too.
So, what's best for any particular conlang depends on your design goals for that particular 'lang.
Some goals, though, are incompatible with other goals. You seem to have three goals, each of which is attainable, but no two of which are attainable simultaneously:
- Look naturalistic and realistic
- Not look newbishly uncreative
- Not give a newbie a headache
I'd be glad to learn that anyone has proven any two of these can be attained simultaneously; but I'll bet not all three together can.