Category: worldbuilding

So I was playing around with the language family when I realized that I really liked the sound of Language 1b partway through its evolution. If you allow the language to develop back harmony and vowel coalescence, but stop it before it develops rounding harmony, the result has a fun and distinctive phonaesthetic. I picture this ancestor language being kept around primarily as a liturgical language, while the later language 1b is used in day-to-day speech. I’ll call the ancestor language Ancient 1b, and its descendant Common 1b. (Better names coming soon.) That by itself would not be a full blog post. However, there are a few changes that I want to make to alter (and hopefully improve) the phonaesthetic.

The primary change I made was putting velar-uvular harmony in this language too. I like the uvulars for this language, especially combined with /ɯ/. It gives it a very Turkic sound that reminds me especially of languages of the Kipchak family, such as Kazakh and Kyrgyz. We can say that the uvularization of emphatic velars was an allophonic feature in protolanguage 1 before language 1a and 1b diverged. The uvulars will still be ditched by Common 1b – I don’t like them as much there aesthetically, and it provides another point of divergence between 1a and 1b.

I also changed around the ordering of sound changes. I wanted this language to keep the voiced stops, so I moved voiced stop changes to happen after rounding harmony, which I hadn’t initially planned for. This means that the voiced uvular stop, /ɢ/, will be present in this language. I’ve already talked about /ɢ/ and its rarity compared to other uvulars. However, I like it for this stage of the language – it gives a very distinctive sound, and separates it from the Turkic languages. This gives us this final list of sound changes:

Protolanguage 1 to Ancient 1b

kˤ, xˤ, gˤ, ɣˤ → q, χ, ɢ, ʁ

short i, u → j, w / V _

short i, u → j, w / _ V

a → ∅ / _ V(+long)

uj, aj → yː, ɜː

a → ɜ

iˤ, ɜˤ, yˤ → ɯ, ɜ, u

iw, ɯw, ɜw, ɑw → yː, uː, ɞː, ɒː

Emphasis distinction lost

Ancient 1b to Common 1b

q, χ, ɢ, ʁ → k, g, x, ɣ

Rounding harmony: i, ɯ, ɜ, ɑ → y, u, ɞ, ɒ

C(+voiced)(+obstruent) → C(-voiced) / _ C(-voiced), C(-voiced) _, _ #

b, d, dʑ, g → β, ð, ʑ, ɣ

tɕ, ɕ, ʑ → tʃ, ʃ, ʒ

Here is the phonology and romanization for Ancient 1b:

Ancient 1b Consonants		Bilabial	Alveolar	Avleolopalatal	Velar	Uvular
Stop/Affricate	Voiceless	p ⟨p⟩	t ⟨t⟩	tɕ ⟨ch⟩	k ⟨k⟩	q ⟨q⟩
	Voiced	b ⟨b⟩	d ⟨d⟩	dʑ ⟨dzhj⟩	g ⟨g⟩	ɢ ⟨gq⟩
Fricative	Voiceless	f ⟨f⟩	s ⟨s⟩	ɕ ⟨sh⟩	x ⟨kh⟩	χ ⟨qh⟩
	Voiced	v ⟨v⟩	z ⟨z⟩	ʑ ⟨zh⟩	ɣ ⟨gh⟩	ʁ ⟨rh⟩
Liquids		–	ɾ l ⟨r l⟩	j ⟨i⟩	w ⟨u⟩	–
Nasals		m ⟨m⟩	n ⟨n⟩	–	–	–

Ancient 1b Vowels	Front		Back
	Unrounded	Rounded	Unrounded	Rounded
Close	i iː ⟨i ii⟩	yː ⟨y⟩	ɯ ɯː ⟨ı ıı⟩	u uː ⟨u uu⟩
Open	ɜ ɜː ⟨e ê⟩	ɞː ⟨eu⟩	ɑ ɑː ⟨a aa⟩	ɒː ⟨au⟩

Some things to note:

• For this language, I opted for vowel doubling to indicate length. I used a circumflex instead on ⟨ê⟩, because ⟨ee⟩ looks like /i/ to English speakers. Besides, I like the circumflex.
• As vowel hiatus isn’t a thing, I can use vowel letters for glides.
• I opted for historical spellings for ⟨eu⟩ and ⟨au⟩. They look cool and aren’t that uncommon for open front and back rounded vowels respectively.

That’s pretty much all the phonological information for Ancient 1b! Next time, I stop going on tangents and return to making morphology, setting templates for some morphological innovations in language 1a.

We’ve already established that protolanguage 1 had a direct-inverse system of alignment. (See the seventh post in this series, “A Sketch of Verb Agreement”). However, I don’t want that to be the case in protolanguage 0. Instead, I want protolanguage 0 to only have focus agreement in the verb, using the inverse as a kind of “patient trigger” to specify that the object’s being marked rather than the subject. I use the terminology “patient trigger” because this looks fairly similar to a rudimentary system of symmetrical voice, also known as Austronesian alignment. While symmetrical voice is extremely complicated, here’s the brief run-down:

• One noun is marked as the focus.
  • In Austronesian languages, this is typically done by placing the noun in the direct case. In protolang 1, it’d be done by subject agreement.
• The verb takes a marker to show how the focus relates to it. For instance, the “patient trigger” specifies the focus as the verb’s object.

In language family 2, my plan is for this to develop into a more robust symmetrical voice system. The patient trigger will fuse with various applicatives to form into other agreement markers. In protolanguage 1, on the other hand, the non-focus pronoun is incorporated into the verb as a suffix. This means that, when the patient trigger is absent, the suffix pronoun is the patient, but when the patient trigger is present, the suffix pronoun is the agent. Now, our patient trigger is starting to look a lot like an inverse marker, especially as protolanguage 0 preferred to make the higher-animacy argument the focus. The applicatives will then stick around as true applicatives.

Now is probably a good time to note that protolanguage 0 uses verb-initial word order. I have the vague idea that there was a pre-protolanguage 0 which was SVO, which explains why subjects were marked as prefixes. However, as the language shifted to a more symmetric-voice-like system, the subject moved after the verb, a better configuration for this type of alignment. Verb-initial word order isn’t something that conlangs often play around with, but I think it makes sense for this language. 

Now that we know how protolanguage 0 operated, we can start to figure out more about how protolanguage 1 will operate. Specifically, we can now begin to evolve some tense and aspect. The tense system in protolanguage 0 was very simple. A past tense suffix, /-ɬaː-/, was affixed, directly after the verb stem. Any tense other than the past was inferred by context or a time-specifying adverb or clause.

The past/non-past distinction from protolanguage 0 is kept in protolanguage 1. However, aspectual distinctions will begin to emerge. Two verbs begin to semantically weaken, becoming light verbs that serve to convey aspectual information. These light verbs take focus agreement, but the patient trigger still occurs on the main verb. I think this makes sense – after all, in English, we say “It has been broken,” not “It is had broken.” The past tense affix also goes on the main verb. Because of this, it’s easy for these light verbs to agglutinate to the main verb, with the weak verb root essentially becoming an aspect marker that goes between the focus marker and the main verb. Because this occurs before most of the impactful sound changes between protolanguage 0 and protolanguage 1, they’ll behave like normal verb prefixes. 

I’ve decided to keep the aspect system itself fairly simple, with two affixed aspects, perfect and prospective:

/-hu/ (own, v. protolanguage 0) → 

/-u-/ (ᴘᴇʀꜰ., protolanguage 1)

/-ri/ (go, v. protolanguage 0) → 

/-ri-/ (ᴘʀᴏsᴘ., protolanguage 1)

In language 1a, these affixes will stick around alongside the past tense affix. This provides a 3-aspect 2-tense system, kind of similar to latin’s 3-tense 2-aspect system, except with a future-in-the past and no future perfect:

	Past /-aː/	Non-past /-∅/
Perfect /u-/	Pluperfect	Perfect
Simple /∅-/	Past	Present
Prospective /ri-/	Future-in-the-past	Future

The future-in-the-past is very volatile, and may gain some kind of modal or conditional meaning. I also need to figure out the distinction between past and perfect. I’m thinking that the simple past either becomes a past imperfective (like in Latin & the romance languages) or a discontinuous past. We’ll see what’s necessary once I start developing more morphology and periphrastic tense/aspect constructions in language 1a. 

In language 1b, however, I want the past/non-past distinction to dissolve, and have it be supplanted by the aspect system (with the perfect turning into the past and the prospective turning into the future). Aspectual morphology can be conveyed by periphrasis. A fun thing about language 1b is that the perfect aspect marker, /u-/, will take part in vowel coalescence. This means that you have to learn a separate set of personal prefixes for the past tense:

	Singular	Plural
1st	sYː-	sYː-…-k
2nd	ʃuː-	ʃUː-…-k
3rd proximate	u-	xYː-
3rd obviative	ðOː-	ðYː-
3rd inanimate	ʒUː-
Indefinite	nOː-

This ended up being much longer than I had intended, and I honestly have no clue how to finish it up neatly. But I think that this is all the work we need to do on protolanguage 0 for now. Next up, I go on another completely random tangent because this is how my brain works I guess.

Let’s finish the sound changes between protolanguage 0 and 1 by adding some vowel loss. This will help with some synchronic features – justifying the forms of the personal suffixes and adding some interest to verb conjugation. 

I’ve already established that word-final vowel loss happens between protolanguage 0 and protolanguage 1. This explains why an affix that shows up as prefixed /si-/ shows up as suffixed /-s/ and not /-si/. However, there’s a bit of a problem with that. Protolanguage 0 permitted words to end in consonants. That means, that if you have a root like /xat-/, and you suffix the ending /-da/, after vowel loss it’ll become /xatd/, which is extremely noncompliant with the phonotactics of protolanguage 1. We can insert the epenthetic vowel /a/ to clear up this issue, making /xat-da/ become /xatad/. This change will make long vowels short, rather than deleting them, and it will also happen before the disappearance of /h/ and /ʔ/, giving a contrast between word-final consonants, short vowels and long vowels.

However, that’s not the only vowel loss I want to implement. I really like verb roots that start with CCV in this language. However, to do that, right now you need a verb that starts with VCCV by default, which won’t be too common. However, we can get more CCV roots by generating them from CVCV. I’ll therefore make this rule:

V[-long] → ∅ / _ CVV // $ _, $ C _, CC _

In other words, short vowels are deleted in syllables directly before long vowels, if it wouldn’t cause an illegal syllable structure. I’ll also say that when an /i/ dissappears this way after an alveolar or velar obstruent, it’ll palatalize it:

{t,k}, {d,g}, {s,x}, {z,ɣ} i → tɕ, dʑ, ɕ, ʑ / _ CVV // $ _, $ C _, CC _

Combine these changes with the voiced obstruent shifts in languages 1a and 1b, and this produces some very interesting divergent forms:

/gitaːb/ → /gitaːb/ → /kitɑːv/, /ɣitɜːp/

/si-gitaːb/ → /sidʑkaːb/ → /siʝtɑːv/, /sitʃtɜːp/

I’ll also add a similar rule for vowels in syllables after long vowels, though there won’t be any palatalization occurring here:

V[-long,-stress] → ∅ / VVC _ // _ CC, _ C $

(I added the stress exception so that the prefix diː- and other prefixes with long vowels can’t trigger this in the stem.)

In practical terms, this means that every verb will have a maximum of three principal parts – one for when they only take a prefix, one for when they only take a suffix, and one for when they take both. (There’s no situation in this language where a verb takes no prefix or suffix.) Not every verb will have all of them, so I picture an English-like situation, where only as many as are necessary will be listed in the dictionary.

This may not truly be the end to the sound changes between protolanguage 0 and protolanguage 1, as they are looking far too similar for my liking still. But this is enough to go off of right now – we can back-implement other sound changes later. Next time, I’ll return to verb agreement. But this time, we’re looking at protolanguage 0 again, and seeing how its system of verb agreement evolved into the one in protolanguage 1.

Hello! This time, we interrupt our regularly scheduled programming of morphology for romanization! As a warning, this post is going to be fairly chart-heavy. I’m just going to show the completed IPA tables with their romanizations, and briefly discuss some of the choices I made in making them. Let’s get into it!

Lang 1a Consonants		Bilabial	Alveolar	Palatal	Velar	Uvular
Stop/Affricate		p ⟨p⟩	t ts ⟨t ts⟩		k ⟨k⟩	q ⟨q⟩
Fricative	Voiceless	f ⟨f⟩	s ⟨s⟩	ç ⟨sh⟩	x ⟨kh⟩	χ ⟨qh⟩
	Voiced	v ⟨v⟩	z ⟨z⟩	ʝ ⟨j⟩	ɣ ⟨gh⟩	ʁ ⟨rh⟩
Liquids		–	ɾ l ⟨r l⟩	–	–	–
Nasals		m ⟨m⟩	n ⟨n⟩

 

Lang 1a Vowels	Front	Back
Close	i iː ⟨i ī⟩	u uː ⟨u ū⟩
Mid	e eː ⟨e ē⟩	o oː ⟨o ō⟩
Open	ɐ ⟨ă⟩	ɑ ɑː ⟨a ā⟩

Some things to note:

• The use of ⟨ă⟩ for /ɐ/, inspired by the Romanian use of ⟨ă⟩ for /ə/.
• The use of ⟨rh⟩ for /ʁ/. As far as I know, this isn’t precedented. However, in the languages that have /ʁ/ as a dedicated phoneme, the prevailing standard seems to be ⟨ğ⟩, which I am not a fan of.
• The use of ⟨j⟩ instead of ⟨zh⟩. I just like it better for this language.

 

Lang 1b Consonants		Labial	Dental/Alveolar	Palatal	Velar
Stop/Affricate		p ⟨p⟩	t ⟨t⟩	tʃ ⟨ch⟩	k ⟨k⟩
Fricative	Voiceless	f ⟨f⟩	s ⟨s⟩	ʃ ⟨sh⟩	x ⟨kh⟩
	Voiced	v ⟨v⟩	z ⟨s⟩	ʒ ⟨zh⟩	ɣ̞ ⟨gh⟩
Approximant		β̞ ⟨bh⟩	ð̞ l ⟨dh l⟩	j ⟨y⟩
Taps		–	ɾ ⟨r⟩	–	–
Nasals		m ⟨m⟩	n ⟨n⟩

 

Lang 1b Vowels	Front		Back
	Unrounded	Rounded	Unrounded	Rounded
Close	i iː ⟨i í⟩	y yː ⟨ü ű⟩	ɯ ɯː ⟨ı ıı⟩	u uː ⟨u ú⟩
Open	ɜ ɜː ⟨e é⟩	ɞ ɞː ⟨ö ő⟩	ɑ ɑː ⟨a á⟩	ɒ ɒː ⟨o ó⟩

The vowel system’s romanization is very wacky, probably because the vowels themselves are pretty weird. That’s one of the shortcomings of romanization – it has issues when dealing with extremely non-latin-like phonologies. This may not be the final version, but it’s good enough for now. Some things to note:

• Acute accents for long vowels, including double acutes for long umlaut vowels – think Hungarian. I like this better than double letters, because ⟨oo⟩ and ⟨ee⟩ are weird for English speakers, and I just don’t like the look of ⟨öö⟩. 
• A digraph for /ɯː/. I just couldn’t figure out how to distinguish ı with an acute accent from í. Besides, ıı looks cool. However, it’s not great, and I may change it later.
• The very English-y ⟨o⟩ for /ɒ/.
• The digraphs for the approximants. Yeah, I could have used the single letters /b d g/, but the language is already non-intuitive enough.

So that’s the romanization! With that out of the way, we can get back to our next order of business: more morphology!

…Wait, we’re going to be making protolanguage 0 instead?

…Why are we doing that now?

…OK, I guess we’re doing it. Next order of business: making protolanguage 0!

Hello! Here, I’m going to show off what the verb agreement will look like in the descendant languages (which both keep the same basic system.) But to do that, we need a cleaner notation system. Therefore, let’s set a standard for these languages by talking about underspecified phonemes!

Underspecification is not something that I understand fully. But, when talking about harmony, an underspecified phooneme is a symbol used to refer to all potential versions of a phoneme in words with different harmonies. For instance, in language 1a, the underspecified phoneme /I/ corresponds to the phonemes [i] and [e]. These are helpful when writing out morphology – because instead of having to write each form of the affix, you can write it using the archiphoneme. To do that, however, we’ll need to set what these underspecified phonemes are:

Lang 1a
Symbol	Close/Velar	Open/Uvular
A	ɐ	ɑ
U	u	o
I	i	e
K	k	q
X	x	χ
G	ɣ	ʁ

(When one of the underspecified vowel phonemes is followed by a long mark, ː, that means that it’s a long version of whichever vowel the underspecified phoneme becomes.)

Lang 1b
Symbol	Front Unrounded	Front Rounded	Back Unrounded	Back Rounded
A	ɜ	ɞ	ɑ	ɒ
I	i	y	ɯ	u
Yː	yː	yː	uː	uː
Oː	ɞː	ɞː	ɒː	ɒː

Same deal applies here. Note the underspecified phonemes Yː and Oː, which were formed when diphthongs with an off-glide [u̯] coalesced. We won’t see these here, but we’ll be dealing with them later. With that out of the way, let’s get to the actual affixes:

Language 1a
	Singular	Plural
1st	sI-, -s	sI-…-K, -sIK
2nd	çU-, -ç	çU-…-K, -çuk*
3rd proximate	∅-, -∅	XI-, -X
3rd obviative	tA-, -z	tIː-, -zI
3rd inananimate	tsU-, -ʝ
Indefinite	nA-, -n

Relative singular	vA-
Relative plural	vAX-
Relative inanimate	vAʝ-

*(Note that the suffix form of the second person plural will always be close/velar, because the [ç] blocks backing harmony.)

Language 1b
	Singular	Plural
1st	sI-, -s	sI-…-k, -sIk
2nd	ʃu-, -ʃ	ʃu-…-k, -ʃuk
3rd proximate	∅-, -∅	xI-, -x
3rd obviative	ðA-, -t	ðIː-, -ðI
3rd inanimate	ʒU-, -tʃ
indefinite	nA-, -n

Relative singular	vA-
Relative plural	vAx-
Relative inanimate	vAʒ-

And with that, the verb agreement affixes are pretty much done! Next time, I’ll show the romanization system of these languages, and discuss some of the choices I made.

Language 1b’s vowel harmony system was looking pretty boring, so let’s change that by introducing an additional axis of harmony: rounding!

The approach I took to introducing front/back harmony in this language was very non-traditional. The rounding harmony will occur in a much more standard fashion. The roundedness of the vowels [u], [y], [ɞ], and [ɒ] will transmit their roundedness to other vowels in the word. This is, by the way, part of why I introduced the rounded vowels in the last part: I imagine that rounding harmony will happen more easily if rounded/unrounded pairs already exist. Vowels affected by rounding will transform in this way:

i, ɯ, ɜ, ɑ → y, u, ɞ, ɒ

There’s a catch, though – unlike emphasis harmony, which spreads bidirectionally by default, rounding harmony often will only spread in one direction. In fact, emphasis harmony itself may be the result of two different processes, one spreading backwards (regressively) from the emphatic consonant and one spreading forwards (progressively) – which is why it often has different spreading patterns progressively than regressively. 

I want to keep vowel harmony bidirectional in this language; the language makes extensive use of both prefixes and suffixes, and I think it would get confusing if they worked too differently. However, I’ll have the progressive spread be halted by the vowel [ɑ] – it’s such an open vowel that you need to move your lips significantly to make the rounded or unrounded version, and since you’re already moving them, you might as well unround them. This wouldn’t happen regressively, because the regressive harmony is based on anticipation of the rounding, not lingering on it.

I also don’t want this harmony to spread from affixes – that would cause all kinds of  stuff that, while interesting, is not what I’m trying to do with this language. So instead, I’ll make this stress-dependent. Stress in language 1b always falls on a syllable in the stem (for reasons we’ll get into later), and so the vowel harmony will always be triggered by a stem vowel.

Our final table looks like this:

Front		Back
Unrounded	Rounded	Unrounded	Rounded
i iː*	y yː†	ɯ ɯː	u† uː†
ɜ ɜː	ɞ ɞː	ɑ§ ɑː§	ɒ ɒː

* opaque progressively to backing harmony

† sometimes transparent to rounding harmony

§ opaque progressively to rounding harmony

(Note that the long rounded vowels will be transparent to rounding harmony outside the stem if they came from a diphthong. This is weird, and I might change it later, but it was the simplest solution.)

This overall system of vowel harmony might look too clean – there are two basic vowel forms, each one with four potential forms for the four potential combinations of backness and roundedness that correspond pretty exactly on the vowel chart. But I’m fine with this system for this language; the complex spreading rules and multiple hybrid vowels (vowels that are sometimes neutral) make this system weird enough for me. Besides, we can mess it up more in descendant languages. 

Again, I have nothing to recommend for further reading, at least apart from what I’ve already recommended – Rose and Walker’s “Harmony Systems” is an excellent overview, and one of the only scholarly works on harmony that I’ve found that was concise and comprehensible to someone without a linguistics degree. That’s all for now! In the next installment, I’ll be giving the consonants some love. (And by “giving them love”, I mean killing a few of them in the name of phonaesthetics.)