How regular do sound changes need to be?

[Adkorr]

Hi everyone!

This is my first post. I've been on-and-off conlanging for a while now, and decided to pick it up again recently. Super happy with what I've been working on. It's an a priori protolang with ergative syntax, polypersonal agreement, and is heavily right-branching, plus a few other features I'm ironing out.

Anyway, I've come to ask about sound changes, specifically in regards to grammaticalized words. In my language, aspect (tense is not marked) is determined by a set of morphemes derived from old auxiliaries that have merged onto the verb. So far, I have them undergoing a sort of "phonetic erosion" without following any set rules but sticking to general common principles (ie certain word-medial nuclei are deleted). I justify this by saying that with these words being eroded, there is a bit more allomorphic variation, giving me more wiggle room to play with sounds.

This is in conflict with a lot of what I've been taught, but it seems like the most logical approach to me. Do any of you do this? Something similar? If not, what do you do instead?

[Sequor]

This is perfectly normal and common, and I'm surprised to hear that it could go against what you've been taught about sound change. Maybe someone inadvertently omitted it. Some natlang examples off the top of my head include:

- the reduction of Old Spanish -ades, -edes, -ides (2PL endings in conjugations) to early modern Spanish -áis -éis -ís, presenting an irregular loss of medial /d/
- the voicing of the common grammatical morphemes 的 de and 個 ge in Mandarin, so that they're pronounced with [d] and [g] (instead of the expected [t] and [k])
- the reduction of Middle English "not" to n't in many positions close to auxiliary verbs, even in contexts where they aren't auxiliary exactly (isn't, hasn't, don't, doesn't, can't, mayn't)
- the reduction of certain sequences of a subject pronoun and a following verb in English (I'm, I'd, I've, I'm(n)a, he's, you're...)
- the irregular palatalization of Proto-Semitic *ʾal to Ge'ez ʾəy- and further reduction to ʾi- as the main negator of verbs

I recall that a bunch of the verb suffixes in Japanese are also reduced forms of former auxiliary verbs.

[LinguistCat]

I think one thing to remember is that these "irregular sound changes" are going to act more often on words/word parts that are used more often. For example, in Japanese, the ends of adjectives went from having -k- to -∅- when followed by -i or -u, and verbs ending in -ki in the infinitive form lost their -k-'s before certain endings in some dialects. But other -k-'s did not elide even in similar circumstances.

(The -ku adverbial ending of adjectives was borrowed back from dialects that did not have this sound change later, but some fossilized expressions kept the form without -k-'s, like arigatou < arigatau.)

[Arayaz]

Possibly relevant:

Sound change is inherently regular, and creates irregularity, whereas
Analogy is inherently irregular, and creates regularity.

But this is not completely correct, as irregular sound changes do happen, often applying only to one common word, or a small class of common words (like "maked" becoming "made" via elision of the /k/). So you're fine eroding your grammatical endings.

[Visions1]

At one point, as I was reading a book on UG or something, I realized something.

The best rule is probably: given the circumstances, how will the kids pronounce it?
Imagining sound change (or any other change) in terms of how the next generation preserves/changes things and to what degree they will is actually a pretty useful tool (despite the fact I mostly ignore it...).

For example:
Expect sounds that are hard to pronounce to get more convenient. Note - while this can mean easier, that absolutely does not have to be the case! But that can be a rule of thumb. (See https://en.wikipedia.org/wiki/Ḍād#Pronunciation; https://www.internationalphoneticassoci ... HS0929.pdf; also see Halkomelem phonology, and note and compare which places get which forms of articulation; also note vowel development in Iatmul, which got so complicated over time that now we're not even sure how many vowels it has - https://en.wikipedia.org/wiki/Iatmul_la ... cal_system)
Expect upheavals in terms of migration, extinction, the internet, etc. to change things up (see https://en.wikipedia.org/wiki/Creole_language, https://en.wikipedia.org/wiki/Tiwi_language#Modern_Tiwi; also note how alcohol may have influenced pronunciation in Balto-Slavic languages and Australian English).
Expect surroundings to have massive effect (see https://en.wikipedia.org/wiki/Faroese_language, https://en.wikipedia.org/wiki/Sprachbund;)
Expect old changes to sort of kind of still exist in the back of the kid's head - but often be taken at face value (for example, even though in Chassidic Yiddish /ʁ/ -> /r/, the language still sounds a bit guttural - though not as much as earlier forms; another example, https://www.sfu.ca/~gerdts/papers/Gerdt ... CSNL41.pdf. Also consider how an Iatmul person might think of his own vowels compared to his neighbouring tribe's).

A good tip for phonology also - (resting) tongue position, pace of speech, and other "resting" factors actually can determine a lot in terms of phonology. For example, tongue position in Lithuanian Yiddish is fronted, which ends up fronting the vowels - in some cases so extremely that /∫, ʁ, ɔ, oj/ turn into /s, ɹ̠˔, ɜ~ɛ, ɘj~ej/ respectively. Or how Y. Frank claims in his Talmudic Grammar that Aramaic and Hebrew went through different vowel sound changes due to how fast people spoke their respective variety of West-Semitic (don't argue whether the claim is true - focus on a person would make that claim in the first place). See also that alcohol thing I mentioned.

Also, most important of all: Always try to pronounce what you create. Read it. Say it. Whisper it. Scream it. Sing it. Speak it in as many ways as you speak your native tongue, because remember - you're creating a language! Not just some silly phonemes said so stiltedly by some orcs that they are 100% acting, but like a real, in-use language.
Following this step will make sound changes (and all the other stuff I wrote) much easier to do, and make the sounds more natural.
If you feel you need to mod them, then don't be afraid, and mod the sound change rules.

Of course, you could always just ignore all this and just follow common patterns that show up everywhere, and then just throw in an Nauhan-style /tɬ/ rule (ta -> tɬa) with like no rhyme or reason - but then you might give your planet's future linguists migraines, or the language won't look consistent. But it might be fine anyways.

[Visions1]

To summarize: follow this in general.
It works for all changes,
in particular sounds changes,
in particular how regular you're wondering the sound changes should be:

1) If a kid was learning this, how would he change it/keep it the same?
2) What does the environment of the language look like?
3) How does a speaker position their mouth (I kid you not)
4) How does it sound when you say it? Is it easier than the prior form?

I think the erosion thing makes sense, but I'd make some rules for it that work inside of the words, and not just on their ends, so the change will be more consistent through the language.
Still, it seems alright good.
See Irish or Hebrew phonology for two good examples of lenition. See also here: https://en.wikipedia.org/wiki/Grammatic ... ic_erosion