(Conlangs) Q&A Thread - Quick questions go here

A forum for all topics related to constructed languages
User avatar
Vlürch
sinic
sinic
Posts: 334
Joined: 09 Mar 2016 21:19
Location: Finland
Contact:

Re: (Conlangs) Q&A Thread - Quick questions go here

Post by Vlürch » 26 Jan 2020 19:54

Is it too unnaturalistic for a language to just straight up lose tone, creating tons of homophones? Especially if the same language has lots of mergers, like /pʰ/ and /kʰ/ both becoming /h/ word-initially and stuff like that?

User avatar
Creyeditor
mongolian
mongolian
Posts: 4545
Joined: 14 Aug 2012 19:32

Re: (Conlangs) Q&A Thread - Quick questions go here

Post by Creyeditor » 26 Jan 2020 21:24

I think tons of homophones are unrealistic, but there are repair strategies like compounding to disambiguate. Losing tonal contrasts and consonantal contrasts must have happend in some more natlangs.
Creyeditor
"Thoughts are free."
Produce, Analyze, Manipulate
1 :deu: 2 :eng: 3 :idn: 4 :fra: 4 :esp:
:con: Ook & Omlűt & Nautli languages & Sperenjas
[<3] Papuan languages, Morphophonology, Lexical Semantics [<3]

Ælfwine
roman
roman
Posts: 970
Joined: 21 Sep 2015 01:28
Location: New Jersey

Re: (Conlangs) Q&A Thread - Quick questions go here

Post by Ælfwine » 27 Jan 2020 00:10

I forget who said this, but generally languages only deal with the mess afterwards caused by sound change. It's reasonable to assume that any loss of tone would create lots of homophones, which would be dealt with the speakers of the language (for example, compounding like in Chinese.)
My Blog
Current Projects:
Crimean Gothic — A Gothic language spoken in Crimea (duh)
Pelsodian — A Romance language spoken around Lake Balaton
Jezik Panoski — A Slavic language spoken in the same area
An unnamed Semitic language spoken in the Caucus.

Salmoneus
MVP
MVP
Posts: 1788
Joined: 19 Sep 2011 19:37

Re: (Conlangs) Q&A Thread - Quick questions go here

Post by Salmoneus » 27 Jan 2020 01:45

Ælfwine wrote:
27 Jan 2020 00:10
I forget who said this, but generally languages only deal with the mess afterwards caused by sound change. It's reasonable to assume that any loss of tone would create lots of homophones, which would be dealt with the speakers of the language (for example, compounding like in Chinese.)
It seems that way, but we should also bear in mind that it has to seem that way.

I suspect, and I think studies of incipient changes support, that languages do resist changes that would cause too much 'damage'. But of course, we can't see that historically, because all the changes that don't happen are invisible...

That said, sometimes this 'preservation' mechanism fails (perhaps because compounding or other strategies are already underway?), and massively damaging changes do take place.

User avatar
LinguoFranco
greek
greek
Posts: 473
Joined: 20 Jul 2016 17:49
Location: U.S.

Re: (Conlangs) Q&A Thread - Quick questions go here

Post by LinguoFranco » 29 Jan 2020 18:47

Okay, I hit a roadblock with my conlang because of vocabulary generation. I found out that many of my words tend to have a one on one correspondence with English, with the exception of a few words.

What are some tips for avoiding this while generating vocabulary?

User avatar
Jackk
roman
roman
Posts: 1132
Joined: 04 Aug 2012 13:08
Location: Damborn, Isr Boral

Re: (Conlangs) Q&A Thread - Quick questions go here

Post by Jackk » 29 Jan 2020 20:00

LinguoFranco wrote:
29 Jan 2020 18:47
Okay, I hit a roadblock with my conlang because of vocabulary generation. I found out that many of my words tend to have a one on one correspondence with English, with the exception of a few words.

What are some tips for avoiding this while generating vocabulary?
- when you create a conlang word X, give it 3 or so different English translations
- when you want a word for some English word Y, create 3 or so conlang words for the differing uses of Y
- translate texts and create words as you need them; this can lead to conlang words with fun ranges of meaning
mundum impūrum incolāmus
hamteu n'un mont sug
let us live in a dirty world

User avatar
Dormouse559
moderator
moderator
Posts: 2885
Joined: 10 Nov 2012 20:52
Location: California

Re: (Conlangs) Q&A Thread - Quick questions go here

Post by Dormouse559 » 29 Jan 2020 20:04

LinguoFranco wrote:
29 Jan 2020 18:47
What are some tips for avoiding this while generating vocabulary?
A big thing is to look at how non-English languages have divided up the semantic space. A Conlanger's Thesaurus gives a lot of crosslinguistically common polysemies. With a quick scroll-through, I found a few listed as common that aren't found in English, like using the same word for "arm" and "wing"; "slow" and "cold"; or "bowl" and "gourd". (Note: The thesaurus has weird encoding, so it's not searchable unless you use the replacement scheme in the spoiler at the end of this post.)

In some cases, your language can also make more distinctions than English does. The first examples I think of are from Romance languages. Spanish distinguishes between ser and estar, which are both translated as "be" in English. Many Romance languages distinguish a reflex of Latin sapio from a reflex of cognosco (Fr. savoir vs. connaître; It. sapere vs. conoscere; etc.); in English, they can both be translated as "know".

Then, there are less binary changes one can make. Kinship terms are a great place for variance, since languages can make more distinctions, fewer, or just go off in different directions. English follows what's called the Eskimo kinship pattern with its terms (no distinction between maternal/paternal relatives; distinction between nuclear family and extended family). The Crow kinship pattern has about the same number of distinctions, but divvied up differently. For example, the same term refers to one's father, one's paternal uncle and the son of one's paternal aunt. Meanwhile, the same words are used for one's siblings, the children of one's maternal aunt, and the children of one's paternal uncle.

Spoiler:
plain letter → encoded letter (e.g. To search <a>, type <D> in the search bar.)
a→D
b→E
c→F
d→G
e→H
f→I
g→J
h→K
i→L
j→M
k→N
l→O
m→P
n→Q
o→R
p→S
q→T
r→U
s→V
t→W
u→X
v→Y
w→Z
x→[
y→\
z→]

Post Reply