Every so often, a thread appears asking for help with creating a triconsonantal conlang. And indeed, there is a severe shortage of a priori triconsonantal conlangs.
As someone who is in frequent use of 4 or 5 Semitic languages (depending on how you count) and acquainted with others, as well as well-read on Semitic linguistics, I feel the need to step in and write an accessible tutorial on how to create an a priori triconsonantal conlang that is realistic and follows current understanding of the history of triconsonantal languages.
This tutorial will guide you through setting up and beginning your first a priori triconsonantal conlang, with illustration. I will be creating a conlang along the way. This initial post is intended to build a foundation to springboard off of. At the moment, it is a draft, and is rather rough in my opinion. It may be edited in the future based on feedback and resulting questions, to clean up its sloppiness, and to be expanded with more demonstrations from actual triconsonantal languages. It will be followed by a post on finalizing your triconsonantal system. Depending on the demand and apparent usefulness, I may write future posts that will address more specific and more advanced topics.
By the end of this two-part tutorial, you should have your own a priori triconsonantal conlang, like the one I developed for this tutorial and opened the post with:
Paren pindās ma paned bi-yakus asen pasneddā!
paren pind-ās ma paned bi-yakus asen pasneddā
I post-ACC DEF.ACC write.PAST in.order.to-you.PL it subscribe.SUBJ
I wrote this post, that you might subscribe to it. (Three words with root √p-n-d).
Introduction and Guiding Principles
What follows in this section is a brief explanation of how to conceptualize triconsonantal languages and your conlang. It is dense, but useful. We do not start developing a conlang until the next section. Feel free to skip this section if you are antsy. I recommend skipping it if you are a more hands-on learner.
The phenomena behind triconsonantal languages are not unique to them. They are widespread and exist worldwide. Despite the Semitic languages having a reputation of being unique in having a triconsonantal system, there are probably many other languages that have similar systems but the ones whose systems have been studied have been analyzed through a more informed linguistic lens. (That is, triconsonantalism is not really a scientific linguistic feature). I always intend to look into this more, but have never quite gotten around to anything more than a quick glance.
Many people, especially (but not only) people venturing on snobbishness regarding triconsonantal systems, prefer to refer to nonconcatenativity. This relates the triconsonantal system to a wider range of linguistic phenomena characterized by adding, deleting, or altering internal parts of a word or stem in order to change it morphologically. This is very common in synthetic, non-agglutinating, usually fusional, languages. Nonconcatenativity exists even in English. However, triconsonantal systems are a specific and identifiable subset of languages with nonconcatenative morphology, and I do not believe there should be any qualms about identifying them as a specific subset.
However, knowing that triconsonantal languages are a subset of languages with nonconcatenative morphology, you can refer to non-triconsonantal languages with nonconcatenativity in order to inform your triconsonantal conlang. Moreover, this allows us to begin setting up guiding principles.
1. Be fusional
It is helpful to know that nonconcatenativity necessarily requires a language to be synthetic, and that triconsonantal languages necessarily must be fusional. This is helpful in determining how morphology will even work in your conlang. But if you aren’t well aware of what fusional morphology is, it’s not too important that you grasp what they are – just helpful.
Note that it is not necessary that every bit of morphology be fusional. Fusional languages, including triconsonantal languages, also typically have both analytic and agglutinating features as well.
Fusional morphology will be the result of our triconsonantal conlanging, not necessarily the beginning of our conlanging.
2. Be systematic
One of the most common problems with a priori triconsonantal conlangs is that they are not systematic. The conlanger just changes vowels here and there without rhyme or reason. That’s not only unrealistic, but virtually impossible.
It may surprise many people to know that the relationships between triconsonantal vowel structures are complex and intricate. It may even surprise some people to know that the vowels are morphemes in and of themselves, and carry multiple bits of grammatical information that should be repeated in similar grammatical contexts.
3. Use umlaut and vowel assimilation (or vowel harmony)
Triconsonantal languages are sometimes described as “ablaut on steroids.” This is oversimplifying, of course, but it correctly highlights one of the most important historical processes that create triconsonantal systems. I highly advise you have a basic understanding about ablaut, umlaut and vowel assimilation and how they come about before endeavoring on your triconsonantal language. The Indo-European languages are a great place to look for these.
4. Use vowel reduction, vocalic epenthesis, and metathesis
Vowel reduction is the shortening or deletion of vowels. Vocalic epenthesis is the insertion of a vowel where it didn’t exist previously. Metathesis is the switching of the order of phonemes. The dialectal variant of “aks” for the English word “ask” is an example of metathesis.
Like vowel assimilation, these processes are key. This is what causes vowels to appear between some consonants, and not others, depending on the grammatical context. In tandem with vowel assimilation, it is also how you get shifting vowel structures in what are similar or identical syntactic contexts at first glance.
Take for example, Hebrew mɛlɛk shɛlo and malko, both phrases meaning “his king,” but one using an independent pronoun and the other using an affixed pronoun. The base word “king” significantly changes morphologically, but has no change in meaning. Moreover, these two vocalic structures for “king” actually originated from the same historical vowel structure.
4a. Have a well-defined stress regime from the onset
As I understand it, vowel reduction and vocalic epenthesis in triconsonantal systems are entirely dependent on stress patterns and shifts in stress due to both word- and phrase-level syntax. I am open to other phenomena influencing vowel reduction and vocalic epenthesis, but every situation I have seen can be explained by shifting stress patterns.
You don’t have to worry too much now about how this works, it will simply suffice to say that stress is a phonological feature you should work out early on.
5. Use extreme leveling
This historical phenomenon is very commonly undercounted as key to developing triconsonantal systems. Leveling is a complex aspect of historical linguistics comprising of a whole set of processes that produce a similar result, but a nutshell explanation is enough for now. Leveling is the extension of a vowel structure from one grammatical context to another by analogy. An illustrative example of nonconcatenative levelling by English speakers is the application of goose – geese and mouse – mice to make moose – meese. (Note that leveling is paradigmatic, which is not clear in this example).
Leveling is a necessary historical simplification that allows triconsonantal systems to function and prevent them from becoming overloaded.
Arabic is particularly productive with leveling. Its extreme regularity has led many observers to postulate that it is extremely conservative in terms of Semitic morphology. In reality, Arabic is nearly as innovative as most other Semitic languages, but part of its innovation was leveling to produce regularity.
6. Don’t go crazy on the vowel phonology
The more vowels you have, the harder it is for a triconsonantal system to develop. Paucity of vowels enables the processes of analogy and leveling necessary for nonconcatenativity to reach the extreme of a triconsonantal system.
Until you are experienced and have a good grasp on what you’re doing in conlanging a triconsonantal language, I would recommend starting off with 6-8 vowels, with the same short vowel/long vowel symmetry of Proto-Semitic.
7. Be primarily--but not exclusively--triconsonantal!
Similar to the previous point, too few or too many consonants in your system will make it statistically more difficult for a triconsonantal system to develop. From what I can tell – though, again, this is an area I need to research more about – the only consonantal root system possible is triconsonantal. There are no biconsonantal systems and there are no quadriliteral systems.
This is an appropriate time to squash the pervasive rumor among hobbyists that Semitic languages developed out of a biconsonantal root system. This is a theory floating around in academia, but fringe and rejected by most Semitists. The theory originates in the fact that some Semitic roots were originally biconsonantal. The vast majority of roots do not reflect this, however, and it is statistically unlikely (if not impossible) and the would-be scenario would require a substantive cognitive load unnatural to language development. The explanation for biconsonantal roots is very simple – they were words with only two consonants in an ancestor of Semitic! That is hardly an argument that triconsonantal roots were also words with only two consonants in an ancestor of Semitic.
What these historically biconsonantal words do tell us, though, is that triconsonantal systems have a strong tendency to force words into the triconsonantal paradigms. Likewise, Semitic languages commonly adapt quadriliteral roots to its triconsonantal system too. Dealing with and conforming non-triconsonantal words is an advanced topic to be explained and tutorialized in a future post.
Let’s Make A Conlang!
Above, we established the principles “Be systematic,” “Use umlaut and vowel assimilation,” “Use vowel reduction and vocalic epenthesis,” and “Use leveling.”
This is going to force us beyond the usual scope of a priori conlanging. These are historical processes that are dynamic and not easily incorporated synchronically in a conlang. The easiest approach is to begin by creating a proto-language.
Note that the proto-language is a means to an end, however. You will not be creating a full-fledged or functional conlang as your proto-language. You’ll be creating just enough to sort out the triconsonantal system systematically.
So let’s begin. In this tutorial, I will be creating my own conlang in order to guide you through illustration. What I demonstrate is not the only way to do it, but will provide a structure for you to work off of. I will attempt to suggest alternatives to what I do here and there, without overwhelming you by being too abstract.
We will begin, as typical in conlanging, with the phonology of the proto-language. It’s not necessary to work out the consonant inventory at this point, but you may if you want. Instead, we want to focus on vowels and stress patterns.
In accordance with principle #6 (“Don’t go crazy with vowel phonology”), we won’t go crazy with vowel phonology. I recommend 6-8 vowels, with symmetrical distinction of long/short vowels. I suggest one of these three systems:
1. 6 vowels: Three vowel qualities, two lengths. /a i u a: i: u:/. The exact realization of these vowels is unimportant, but they should conform to the extreme phonemes /a i u/. (That is, they might be, e.g., [æ eɪ ʊ] in quality. But since this is the proto-language, it makes no difference to our project if it is [a i u] or [æ eɪ ʊ]).
2. 7 vowels: The 6 vowel system, but with a “neutral” centralized vowel. /a i u ə a: i: u:/.
3. 8 vowels: Four vowel qualities, two lengths. You have a little more flexibility with this, but it should approach the four extremes of the mouth. Example: /a ɔ i u a: ɔ: i: u:/.
For my proto-language, I will stick with the 6 vowel system.
In accordance with principle #4a, we should have a well-defined stress regime from the get go. You should avoid static stress systems that are strictly “ultimate”, “penultimate”, on the first syllable of the stem, etc. You should have a dynamic system wherein the stress of a word can shift depending on morphemes affixed to a word.
Let’s take a cue from Semitic. Semitic stress patterns are universally derived from the following system:
Syllables consist of two possible constituents, C(onsonant) and V(owel). Long vowels are two constituents: VV. Every syllable must begin with a consonant, and consonant clusters are forbidden within a syllable. This allows Semitic to categorize two types of syllables:
1. Heavy syllables: Syllables that are CVC, CVV.
2. Light syllables: CV syllables.
The last heavy syllable in a word is stressed. If there is no heavy syllable, the antepenultimate (third-to-last) syllable is stressed. This creates a dynamic stress system because sometimes an affix can cause syllables to open up or close, allowing syllables to shift from heavy syllables to light syllables and vice versa.
This is a good system to stick with. If you do something different, you might find it’s not as dynamic as you intended it to be.
My proto-language is going to have a different system, just because the Semitic system is an easy one for you to grasp on your own, and so I can introduce an alternative to you. Note that if you devise your own system, you should be prepared to go back and revise it to be more favorable to shift stress.
The syllabification is going to be more diverse. A syllable can be V, VC, VVC, CV, CVC, CVV, or CCV. I’ll restrict the types of syllables in my triconsonantal conlang later on. I’m also going to consider how these syllables will interact with surrounding syllables, and define the following rules:
- VC.CV > CCV
- CVC.V > CV.CV
- CV.V > CVV
- CV.VC > C.VC > CVC
- CV.VVC > C.VVC > CVC
- CV.CV.VVC > CVC.VVC
- (C)VC.CV.VC > (C)V.CCV.VC
Stress will fall on the final syllable, or on the last syllable that is VVC, CVV, CCV, or syllables that were combined from two syllables. CVC syllables are not automatically stressed.
Just Ignore: Allophones, Sandhi, etc.
While consideration of phonemic consonants in the proto-language is up to you, it is not advisable to sort out details about allophones or phonemic realization or sandhi in the proto-language. The proto-language might have it, but these phonological features probably aren’t going to carry over into a triconsonantal system, which tends to block those types of considerations. One exception is with sandhi of syllabic structures and stress, as alluded to above.
The next step is to make an outline of the morphological features of our proto-language. As a rule of thumb, I make proto-languages relatively simple. Along with taking sides in theoretical debates of what a proto-language is, I just find this easier to manipulate complex features in the resulting conlang.
First, I start with broad syntactic considerations, such as word order.
I’m going to go with SOV, on the simple fact that I find SOV word order the most aesthetically pleasing.
Just for fun, I’m going to add that my proto-language is going to have head-marking tendencies. Don’t worry too much about what that means if you don’t already know. I’m adding it just to give me more direction, and because triconsonantal systems are easier to develop from head-marking.
Because my proto-language will be more head-marking, I’ll start outlining the verbal system.
I want three marked tenses: Past, Present, Future.
I want two marked moods: Indicative and Subjunctive.
I am also going to conjugate verbs according to their subjects, as head-marking languages are apt to do. I will do this with prefixes that vary by person and number (singular and plural).
I now have an idea of what I need to conjugate verbs. I can now make a chart to come back to later which contains my verbal system. Because I already know I want subject-marking to be an invariable prefix, I can just leave subject marking off the chart, and make a chart for tense and mood:
Head-marking languages don’t have a tendency towards noun cases. But I imagine a lot of you will want to play around with noun cases, so my proto-language will have noun cases. As an admonishment: the more complex your noun declension system is, the more difficult it will be to develop a triconsonantal system.
I will have Nominative, Accusative, Dative, and Genitive cases.
I will also suffix definite articles to the case in order to mark definite nouns.
Finally, just to keep things simplistic in the proto-language, I am going to treat adjectives identically to nouns.
We can now build a chart for nouns:
What is the order of direct objects and indirect objects; of nouns and their adjectives? How will I deal with prepositional phrases?
The possessed noun will precede the possessor: house-NOM-the dog-GEN-the.
I want indirect objects to come after direct objects.
I want adjectives to come after their nouns.
I’m going to have prepositions follow the head noun in a prepositional phrase: dog-NOM-the house-DAT-the in blue-DAT-the is. This syntax is a bit odd for prepositions that are not enclitic, but I could explain it somehow. This is not important for this tutorial.
The next thing we need to consider is how we are going to derive words from each other. At this point, we are going to begin creating dummy words and stems in our proto-language.
English puts us at a disadvantage for this part, because our derivational morphology is all over the place. If you don’t have much experience with derivational morphology, it might be helpful to read up on how different languages treat derivational morphology. You might also look into productive derivational morphemes in English such as -ify, -ate, -ation, re-, -ity, co-/con-, etc., as well as less productive derivational morphemes like de- (intensifying), pro-, circum-, etc. I may do an advanced post on derivational morphology later, but it proves too messy to try to sort out here.
Below is a chart of words we can use for our dummy stems. We can derive them all from a single dummy root, meaning “scribble.”
Your proto-language does not require this natural phenomenon. However, in a case like this where I use a proto-language to derive one single conlang, I just go ahead and plug in the desired meaning of the word in the resulting conlang. After all, if you were reconstructing a proto-language using attestations from just one language, you would probably just plug the attested meaning into the proto-language anyway, instead of trying to imagine up previous meanings of the word with no other data.
I went even a little further here and added “listserv” – I intend this to be a modern coinage in my conlang. That means this word did not actually exist in the proto-language, even with some other meaning. There was no semantic shift leading to it. Native speakers just used the derivative morpheme of “place” productively to derive the word!
Note also that I did not fill out every possible square. This is also natural in derivation. Rarely do you find roots in any natlang that employs every possible derivative morpheme to create a new word.
Filling in the morphology
Next, we are going to take those charts we made, and fill them out like you would with a more typical conlang: take the dummy stem and add affixes.
But first, we need to know how you’re going to inflect verbs and nouns. Fill in those charts.
Plug this stem into each derivation on the chart. Then add some derivative morphemes! In most cases, you should use one morpheme per derivative column or row. This isn’t fusional, but remember that fusional morphology is simply the desired result.
I recommend marking the stressed syllable, too, to get an idea when or if it will change based on your derivation. It may also help to break up syllables with a period (.), if types of syllables determine stress, as is the case in my proto-language. Remember that the stress will not be static, and your stem may not reflect this stress pattern when it is actually inflected!
Below is my table, followed by some commentary about what I did.
Now we have the stems. Now let’s create words by adding inflections. You can choose which inflections to plug in. I’ll conjugate verbs for 3rd person indicative past tense and decline my nouns in the definite accusative.
Notice in the reflexive stems, the prefix s- metathesizes. This is because my proto-language does not like a word to begin with the cluster sp-. It deals with this phonological constraint by metathesizing the s and p, a phenomenon frequent in the world’s languages. This is where infixes (morphemes inserted into a word rather than added to the beginning or end) come from. We will talk about this more in the next post on leveling.
Diachronics: Simplify with Principles #3 and #4
To begin creating your triconsonantal system, it is now necessary to work out some diachronics that include umlaut, vowel assimilation, vowel reduction, and vowel epenthesis. You do not need all four processes, but I would recommend mixing at least 3 of them. Optionally, you may also consider sound changes, resyllabification, and/or a slight and logical shift in stress patterns.
You will probably want to pay attention to the following types of syllables vis à vis stress:
1) Tonic: A stressed syllable
2) Pretonic: The syllable before a stressed syllable
3) Propretonic: Two syllables before a stressed syllable. It may also refer to any syllable that is two or more syllables before stress. If you want to treat such syllables differently, you may distinguish between propretonic and pro-propretonic syllables.
4) Post-tonic: The syllable after a stressed syllable
5) I don’t actually know how to refer to a syllable that is two syllables after a stressed syllable. I pretty much always use propretonic syllable rules for this category.
These distinctions tend to be key motivations in vowel reduction.
Because the diachronics get pretty messy – as you will see shortly – it is very important that you set and apply processes in the order in which they occur over time. You can adjust this order as you go, but be sure you apply the adjustment uniformly in your chart.
Your diachronics can create new phonemic vowels. But be wary about creating more than 1 or 2; this will make it more difficult to become a triconsonantal system. I will create 2 phonemic vowels, one long and one short, in order to demonstrate how this might happen, and how this becomes phonemic.
Here are my diachronics, in order of occurrence and application. Don’t worry too much about what I do. It’s just an example for you to see and mimic if you’d like. You probably will get a better idea of what I’m doing from the actual chart below.
1. Sound change: VnC > ṼC > VC; Vn#, VsC > VʰC > VC (i.e., preconsonantal n and s are elided; n at the end of a word is elided)
2. Progressive vowel assimilation:
- Initial i: #i(C)CV, a,u > i / V # a,u (i.e., initial vowel i causes subsequent vowels a and u to become i)
3. Umlaut and Vowel epenthesis:
- Final short vowels are elided if unstressed.
- VCu#, a,i > u / V # a,i (i.e., final short u causes previous a and i to become u)
- VCi#, a > e / V # a; u > I / V # i (i.e., final short i causes a to become new phonemic vowel e and u to become i)
- CCV# > CC# > CeC# (i.e., resulting CC# cluster takes epenthetic vowel e)
4. Vowel reduction and Vowel epenthesis:
- CV.CV > CVC (i.e. two open short syllables in a row collapses the final vowel)
- CV.CV́ > CCV́ (i.e. two open short syllables in a row with the second vowel stressed collapses the first vowel)
- CV.CV.CCV > CCV.CCV (i.e. two open short syllables in a row followed by a consonant cluster collapses the first vowel)
- Pretonic CV syllable before an open syllable metathesizes: (CV > VC).
- Propretonic syllable V is elided.
- Short vowels (a,u) in a propretonic open syllable becomes e.
- Pro-propretonic vowels are unaffected.
- Short vowel (i) is unaffected, but assimilates a pro-pretonic vowel.
- This change is blocked if the pretonic vowel is e.
- Resulting CCC clusters from vowel reduction result in CeC.C
5. Resyllabification: The resulting conlang now strongly prefers syllables that begin C. General tendency towards CVC syllabification.
- #VCC > CVC (i.e., initial vowels are metathesized in between consonant clusters)
- #VCV > CVV (i.e., initial vowels are metathesized and lengthen the following vowel)
- #aCV > #Cā / V # a,u (i.e. initial vowel is metathesized, then the following vowel becomes long ā, when vowel V is a or u)
- #aCi > #Cē / V # i (i.e. initial vowel is metathesized, then the following vowel becomes long ē, when vowel V is i)
- #uCV > #Cū (i.e. initial vowel is metathesized, then the following vowel becomes long ū)
- CV.CCV > CVC.CV (i.e., consonant clusters are split at syllabic barriers when preceded by a vowel)
- CV́ > CV́V (i.e., stressed CV syllables are compensatorily lengthened due to stress)
As mentioned earlier in this tutorial, do not apply sound changes to the consonants of the stem that depend on any surrounding phonological feature. These phonemic sound changes are blocked in the development of triconsonantal systems through leveling.
Below is what happens in my table diachronically. Although it’s a bit messy, I show what happens to these words step-by-step, and bold the outcome:
Now let’s take a look at the chart, cleaned up a bit more, with the stem underlined.
Observations and Outlook
Some things to notice:
- Notice correlations in vowel changes and stems. There is rhyme, reason, and pattern!
- For instance: most of the difference between the base stems and the associative stems is a prefix added to the stem – not changes in the vowels! The reflexive and the causative-reflexive mostly differ only in reduplication – the morphemic marker for the causative. Notice also a tendancy for the doer of the verb (the agent) to have the same stem as the place where they do the action. In fact, in the base stems, the stem is almost the same in agent, profession, instrument, and place, depending simply on whether the vowel in the open syllable collapses or not.
So: Triconsonantal systems don’t just stick vowels in willy-nilly! And this is why people get snobbish about referring to triconsonantal languages as nonconcatenative languages instead.
- Though I have rules for vowel assimilation and umlaut, I actually apply it very little here. This was because my vowel assimilation rules are mostly related to the vowels i and u, which are in very few of these words. If these words had more u’s and i’s in them, or if I have vowel assimilation rules related to the vowel a, the stems would see even more variation.
- Morphemes must also be involved in this diachronic process! Notice how the subject marker on verbs as- has entirely disappeared. The Tense-Aspect-Mood morpheme -an also diverges into two separate morphemes, depending on stem. The next post will address stabilizing such morphemes, so that they evolve naturally, but don’t get out of control.
- These words are quite long for a system that should be highly fusional (principle #1). But as the stems become more diverse from one another, the more information they carry – and thus the more fusional they become! As this happens, we can start dropping morphemes as they become redundant. Earlier in the post, I already showed you how I reduced the words “scribble” and “post.ACC” from panedná and pindārā́zen to paned and pindās respectively. Doing this in an effective way will also be addressed in the next post.
- Finally, look at what a mess the associative is! This sort of mess is entirely possible in a realistic triconsonantal system. However, it does create a cognitive burden. In the next post, we will look at how leveling plays a key role in reducing cognitive burdens, regularizing our paradigms, and yet reinforce the triconsonantal system we’re developing.
- Congratulations! You now have the beginnings of a working triconsonantal language. We’re not done yet. So far these patterns we have developed only work if all root stems in our proto-language are modelled CaCCa. This is quite unlikely. Moreover, as mentioned just above, the language has too many morphemes to be stable. The next post will address these problems and smooth out our triconsonantal language into a functional and sustainable one!