Katapharteo: an engelang with only one type of syntactic relation and roots that conjugate in base four

[dɮ the phoneme]

I've had this idea for a non-naturalistic, engineered language floating around in my head for a while, and I recently thought it was time to finally hammer out some details. I've never really attempted an engelang before, and I don't see too many of them posted around here, so this should be a fun experiment. I think working on something that is explicitly, intentionally, wildly un-naturalistic will be a nice change of pace from my usual conlanging workflow. Thus, I give you Katapharteo: a very strange language. This thread will a be a scratchpad of sorts, as the details are not all completed, so expect updates as well (I mean, if I actually get around to them...).

Idea & Design Goals

I've had a four main goals in mind when designing Katapharteo. The first, and most fundamental, is to create a language that has only one type of syntactic relation. That is to say, if you were to draw a syntax tree for a Katapharteo sentence, you wouldn't need to make use of VPs, NPs, PPs and so on, you'd only need a single phrase type: XP. As a corollary, there are no distinct parts of speech in Katapharteo. I'm going to be calling this syntactic relation "adjunction", because I think that dependents in Katapharteo are best understood as analogous to adjuncts in natural language.

The second design goal is to create something that is, in theory, unambiguously parsable by a computer. This is pretty self-explanatory, so I won't say much more about it. The third goal was to create something that, unlike a lot of engelangs I've seen, is actually phonologically interesting, and uses that phonology in interesting ways as part of its grammar. I ultimately settled on what basically amounts to, uh, using consonant gradation to inflect words for numerals in base four, which... well, you'll see what I mean when I start explaining the grammar.

The final design goal was to not worry to much about design goals, or coherence or anything like that, and just kinda go wild. Throw in every half-interesting feature that I'd never dream of putting in one of my ordinary, naturalistic conlangs and just see what happens. I guess in some sense, I'd just like to buck the trend of engelangs designed for "elegance" or """logic""" or whatever, and make something completely alien just for the sake of it.

Hopefully I've accomplished all of those to a reasonable degree, but you'll have to be the judge of that.

Phonology

First, we have to talk about phonology. There are three classes of phonemes in Katapharteo: initials, pre-initials, and vowels. Syllables have the structure (P)IV, where I is an initial and P is a pre-initial. All initials and pre-initials are phonetically consonants, but that's not so much a relevant phonological class here. The initials are all stops, laid out in a regular 4x4 grid. These are the most important sounds in the language, and they undergo gradation to indicate all grammatical relations between words. The pre-initials are basically just there to bump up the number of possible syllables, with the exception of /l/, which has a very special function.

Initials:
/p t tʃ k/ <p t c k>
/pʰ tʰ tʃʰ kʰ/ <ph th ch kh>
/p' t' tʃ' k'/ <p' t' c' k'>
/b d dʒ g/ <b d j g>

Pre-initials:
/N s x r l/ <n s x r l>
/N/ is a nasal unspecified for place, which assimilates to the following stop. /s x/ assimilate in voicing to a following stop

Vowels:
/i ɨ u/ <i eu u>
/e ə o/ <e eo o>
/ɛ ɔ/ <ea oa>
/æ ɑ/ <ae a>
+length, nasalization, and high/low tones <VV, Vm, V́/V>

Roots & Root Structure

The notion of a "word" is less important for the syntax of Katapharteo than that of a root. Roots are strictly bisyllabic, and underlyingly all have the structure (P)SV(P)SV, where P is a pre-initial other than /l/, and S is one of /p t tʃ k/. The initials in a root undergo gradation to indicate grammatical relations, and thus, on the surface, all initials can occur in a root. There are 2560 total possible roots, of which I intend 2000 or so to actually be used. More complex lexemes are formed out of multi-root constructions, which I'll be calling "compounds" even though they really aren't. More on that in the syntax section.

Morphophonology

Ok, now we're getting to the interesting bits. Roots in Katapharteo inflect for a single category, which I'll call "level". In essence, "level" is just an integer expressed in base four. The digits 0, 1, 2, and 3 are denoted by the phonation of the initials in a root: a plain stop denotes 0, an aspirate denotes 1, an ejective denotes 2, and a voiced stop denotes 3. A bisyllabic root, containing two initials, thus has the potential to express any two-digit base four number, or any number from 0-15. A root's "level" is simply the number encoded in its initials. An important note is that these numbers are expressed with the least significant digit first. That is, the first initial encodes the one's place, and the second initials encodes the four's place.

For example, take the root kata - "language, speech, word". In its unmarked form, it is in level zero: its initials correspond to the base four digits 00. Inflecting it for level 1 corresponds to aspirating the first stop in the word: khata. Inflecting it into level two would look like ejectivizing the first stop: k'ata. Level five would require aspirating both stops: khatha.

If you ever need to inflect a root beyond level 15, the solution is simple: extend the root via the suffix -lta. This suffix can be added to a root as many times as is necessary, in order to make the root long enough to express an arbitrary integer level. This suffix is also the sole occurrence of the pre-initial /l/, the intent of which is to make root extensions unambiguously identifiable.

I'll be glossing level using quaternary numbers in the same order as the digits are encoded. Thus, kata would be glossed "language.00", khata would be "language.10", and so on.

Syntax

As mentioned, there is only one syntactic relation in Katapharteo, which I'm calling "adjunction". This is a relation between roots, and on a semantic level, a root which is adjunct to another root should be taken to be modifying it in some way, or otherwise contributing extra information. Context obviously plays a huge role here, as what precisely that extra information means is dependent on the root in question. However, this is par for the course in natural languages as well (think about how general a notion of "modification" is encoded by the genitive particle in Japanese, for example), so I don't think it needs too much justification.

Adjunction is indicated by level. In particular: a root of level n is adjunct to the closest root of level (n-1) to its left. This might be a bit counter-intuitive, so hopefully an example will clarify. Take the following several sentences:

kata xthóamske
language.00 difficult(y).10
"(a) difficult language"

Here, the root xthóamske is in level one, and thus modifies kata, the most recent word of level zero.

kata xthóamske t'urpo cheanteoo
language.00 difficult(y).10 very.20 strange.10
"(a) very difficult and strange language"

Here, t'urpo is in level two, modifying xthóamske, while cheanteoo is in level one, meaning that it again modifies the previous level zero root, kata.

Ok, so this is all well and good for things that are already adjuncts in most natural languages, but how does this handle verbs? Especially transitive verbs? Well, the answer is that, in general, "verb phrases" (again, really just general XPs with a semantically "verb-like" head root) are adjunct to their subject, and objects are adjunct to "verbs". Take the following example:

tenta sthiiki k'apo
man.00 see/sight.10 dog.20
"(a) man sees (a) dog"

This might be best translated in a very literal way as "man with sight of dog", or some such. Notice that this means that relativizing on subjects is identical to simply describing a subject with multiple adjuncts, i.e

tenta sthiiki k'apo thapo st'opa
man.00 see/sight.10 dog.20 directed_movement.10 store.20
"(a) man sees (a) dog and goes to (the) store" or "the man who sees the dog goes to the store" or "the man who goes to the store sees a dog"

Something like "man with sight of dog and movement to store ".

When there are multiple semantic roles involved, these are communicated through something akin to converbs: "verb-like" roots that modify the main "verb" and add extra information. For example:

stété phunkee nk'apá thaku rt'exkiduke
exchange.00 give.10 money.20 receive.10 banana.20
"to buy a banana"

teke sthété p'unkee ngapá t'aku rdexkitukhe
1sg.00 exchange.10 give.20 money.30 receive.20 banana.30
"I buy a banana"

One final thing to not is about multi-root lexemes. Look at the word rtexkituke "banana". It's actually just two roots, rtexki - "fruit" and tuke - yellow, which stand in adjunction to one another. That's why the internal structure of the "word" shifts from sentence to sentence: its constituent roots have to take on different levels. The fact that it's written without a space simply indicates that it's a conventionalized lexical item.

Anyway, that's all I got for now. Hope this is something a bit new and refreshing for everyone, as making it was for me.

[Pabappa]

Very well made. I'm glad to see you were able to succeed with this idea where so many people have had one good idea but then got stuck halfway with the implementation.

Little-endian base 4 is an interesting idea .... is there any reason particular you decided to make it little-endian?

Could you explain in particular what the levels mean? Do their syntactic meanings change for different roots?

Lastly, the number 2560 stands out to me ... is it because you multiplied 256 by 10, and if so, what made you choose these numbers?

[aporaporimos]

This is a neat way of making the syntax tree unambiguous—feels like something a programmer would come up with for serializing data. Are there third person pronouns, or some other way of referring back to a word/phrase earlier in the discourse other than by the adjunct relation? Like if you have the sentence:

man.00 see/sight.10 dog.20 directed_movement.10 store.20
"A man sees a dog and goes to the store."

and then you want to say something further about the dog, you can't do it using the adjunct relation as you've described it because dog.20 has been "shadowed" by store.20, and any further predicate with level 3 will modify the store. (If I understand things right.) Do you just start a new sentence with dog.00 and infer from context that it has the same referent as dog.20 in the previous sentence?

Side note, but what combinations of vowel length, tone, and nasalization are contrastive? If they all contrast independently then I count 2,560,000 possible syllables! But, even not worrying about naturalism, it'd be pretty tricky to contrast things like tenta, temta, and temnta.

Even more of a side note, but no matter how many times I read the name "Katapharteo" I'm convinced it must be a transliterated Greek verb...

[Torco]

naturalism necessarily aside, it seems like an extremely impractical thing for a human speaker to have no way of disambiguating the relationships between these constituents (adjuncts?). If i've understood yo well this language literally only grammaticalizes dependency, which means every other nuance of meaning has to be either implicit or explicitly communicated: like, if I say me0 dog10 small13 or sth, does that mean my dog is small, my small dog, I am a small dog? I think that it could function, i.e. such a language is probably -speakable. I'd imagine after any amount of time by people, though, and things like grammar will start to emerge, like just conventions right? Just instinctively, say, to mark tense you get used to (and learn that it makes you much more understood) if you use very close... scores? how do you call the quantity of the number? you know what I mean, me0 future9 go10 or me0 go10 future9 standing for I will go vs. me0 future10 go20 denoting that time i time-traveled into the future. Or, say, sentences without any zero score word coming to feel like a passive voice or something.

That'd just be my first instinct if instructed to use such a language. I do like how alien it is, though. I'm imagining a creature with, say, genetic perfect pitch using some system of 15 pitches per octave to encode the score part of a word or something. "we hear the melody of the words and we just know how the parts of speech relate to each other. ah, yes, we also use subtle lexical differences and slight nuances in synonyms to make ourselves clearer, complex rules of glisandi to suggest our emotions about what we're saying, and vibrato to encode saliency, but I wouldn't know where to start explaining them to you, earthling. none of it is grammar, though, it's all optional and highly idiosyncratic".

[dɮ the phoneme]

Are there third person pronouns, or some other way of referring back to a word/phrase earlier in the discourse other than by the adjunct relation?

At present, no, but that's just because I haven't gotten around to them yet. I'm hoping to find some other, slightly more interesting solution to the problem you mention, but barring that I'll just bite the bullet and add in third person pronouns. I was actually hoping to avoid using any pronouns at all; the word I glossed as "1sg" in one of my examples is actually just a root that means "self".

Side note, but what combinations of vowel length, tone, and nasalization are contrastive? If they all contrast independently then I count 2,560,000 possible syllables!

Well, there are 16 initials, 6 pre-initials (counting null), 11 vowel qualities, 2 lengths, 2 options for oral/nasal, and 2 tones. Just multiplying that out, it's 8448 possible syllables. But the underlying structure of roots is much more restricted: only the initials /p t tʃ k/ show up uninflected, and the pre-initial /l/ is not present. Factoring in that roots are bisyllabic, this gives 2*4*5*11*2*2*2 = 3520 possible roots. Hm, it looks like my initial calculation was actually off by a bit, but still in the same ballpark.

[Vilike]

Very intriguing, I wonder if given a text of reasonable size and without knowing the principle behind the thing, someone could come up with a different analysis of the phenomenon. The phonology is nice.

kata xthóamske t'urpo cheanteoo
language.00 difficult(y).10 very.20 strange.10
"(a) very difficult and strange language"

Here, t'urpo is in level two, modifying xthóamske, while cheanteoo is in level one, meaning that it again modifies the previous level zero root, kata.

Would kata xthóamske cheanteoo t'urpo language.00 difficult(y).10 strange.10 very.20 mean "a very difficult and very strange language"?
And how is xthóamske parsed? I count two preinitials in front of the second syllable...

[dɮ the phoneme]

Would kata xthóamske cheanteoo t'urpo language.00 difficult(y).10 strange.10 very.20 mean "a very difficult and very strange language"?
And how is xthóamske parsed? I count two preinitials in front of the second syllable...

kata xthóamske cheanteoo t'urpo would mean "a difficult and very strange language". Roots of level n are only adjunct to the closest root of level (n-1) to their left, so t'urpo here could only refer back to cheanteoo.

As for xthóamske, the <m> indicates a nasal vowel, it isn't a pre-initial.

[dɮ the phoneme]

Miscellaneous Additions


Grammaticalized Roots

So, I should clarify that just because Katapharteo only has one type of syntactic relation, that doesn't mean it only has one grammatical rule. Just as English uses words which are syntactically prepositions, or Japanese uses words which are syntactically nouns, for a variety of conventionalized grammatical functions, so too are there grammaticalized roots in Katapharteo.

One of these roots is scímkoa, meaning "being subject to, undergoing". It's used more-or-less like a passive, especially when one needs to introduce a "relative clause" whose head is also a patient. For example, compare

tenta sthiiki k'apo thapo st'opa
man.00 see/sight.10 dog.20 directed_movement.10 store.20
"(a) man sees (a) dog and goes to (the) store" or "the man who sees the dog goes to the store" or "the man who goes to the store sees a dog"

with

tenta schímkoa st'iiki gapo thapo st'opa
man.00 undergoing.10 see/sight.20 dog.30 directed_movement.10 store.20
"man undergoing sight of dog, going to store", ie "the man was seen by the dog and went to the store"

There are also grammaticalized (thought optional) tense markers, including:

keapéa - "past"
cuntu - "future"
rpítíi - "present"
topóa - "today"
xkeukém - "yesterday"

And maybe some others. Note that these are also just the ordinary lexemes with these meanings, are there isn't a firm line between them and what might be considered regular temporal "adjectives" other than frequency of use.

Phonology Changes

aporaporimos has pointed out that distinguishing between all of tenta, temta, and temnta might be a challenge. I think the first two are distinct enough, but indeed I don't really want to force people to tell the difference between nasal and oral vowels before a nasal, so we'll say that nasality of vowels is non-contrastive and allophonic before a nasal pre-initial.

NUMBERS

For Janko, the numbers. Numbers in Katapharteo use a mixed base 4/base 16 system. Each numeral root consists of two syllables, each representing a base four digit. This leaves 16 numeral roots, which are then used to compose larger numbers in base sixteen. The base syllables are pa - 0, rteo - 1, tó - 2, and stam - 3. Numerals are then composed of these base syllables, again in a little-endian fasion. Thus, the numerals 0 - 15 are:

papa - 0
rteopa - 1
tópa - 2
stampa - 3
parteo - 4
rteorteo - 5
tórteo - 6
stamteo - 7
pató - 8
rteotó - 9
tótó - 10
stamtó - 11
pastam - 12
rteostam - 13
tóstam - 14
stamstam - 15

By the way, this is the source of the language's name: Katapharteo is language.00 four.10, or "language of four".

[aporaporimos]

Side note, but what combinations of vowel length, tone, and nasalization are contrastive? If they all contrast independently then I count 2,560,000 possible syllables!

Well, there are 16 initials, 6 pre-initials (counting null), 11 vowel qualities, 2 lengths, 2 options for oral/nasal, and 2 tones. Just multiplying that out, it's 8448 possible syllables. But the underlying structure of roots is much more restricted: only the initials /p t tʃ k/ show up uninflected, and the pre-initial /l/ is not present. Factoring in that roots are bisyllabic, this gives 2*4*5*11*2*2*2 = 3520 possible roots. Hm, it looks like my initial calculation was actually off by a bit, but still in the same ballpark.

Sorry, I meant possible roots, not possible syllables. (And I missed one of the vowel qualities.) The roots being bisyllabic doesn't double the number of possible roots, it squares it, so the number is (4*5*11*2*2*2)^2, or 3,097,600 possible roots.

Edit: posted this before seeing the note in your latest post about not distinguishing nasalization before /N/, so the actual number's a little lower but in the same order of magnitude.

[dɮ the phoneme]

The roots being bisyllabic doesn't double the number of possible roots, it squares it, so the number is (4*5*11*2*2*2)^2, or 3,097,600 possible roots.

You are right, I feel dumb for making that mistake. Perhaps I will decrease the number of vowel qualities then, or maybe eliminate length. I like both tone and nasalization, but will that many possible roots I don't think a distinction between e.g. /ɛ/ and /æ/ is really called for.

[malloc]

This looks amazing. I really like the idea of merging parts of speech and syntactic relations into one.

[Qwynegold]

That's a really interesting conlang idea. I just think the phonology is really difficult. You've got initial clusters that begin with /r/ and /x/, and /s/ followed by affricates.

For removing a vowel quality I would suggest nasalization, because you already decided to not distinguish it in every environment. If you removed it all together things would be"cleaner". Though I understand the want to keep in features you like.

[WeepingElf]

This is a very interesting engelang idea which vaguely reminds me of this.

[missals]

This is quite interesting. It's kind of similar to a project I have in which only five syntactic relations exist (two of them kinda marginal) but more ambitious. I think it would be interesting to see what kind of grammatical constructions would emerge in a framework like this. I agree about the phonology being difficult, though.

It's a bit forward of me, but if I were to redesign it, I might do trisyllabic roots, and use the following realizations for the four place-values:

/p t tʃ k/
/m n ɲ ŋ/
/p' t' tʃ' k'/
/b d dʒ g/

With only one preinitial, /s/, and five vowels /i e a o u/, which gives 2*3*5*2*3*5*2*3*5 = 64000 possible roots. Or perhaps adding an additional preinitial, maybe /f/, and maybe one more vowel, maybe /ə/, and making it disyllabic, giving 3*4*6*3*4*6 = 5184 roots, which is probably sufficient.

[dɮ the phoneme]

...

Yes, I was thinking of overhauling the phonology anyway, as I agree that's it's quite impractical. Creating a fairly unnaturalistic phonology was part of the design, so I'm going to keep that aspect. I'll just reduce the number of distinctions somewhat. With that in mind:

Phonology, v.2:

As before, there are three classes of phonemes: initials, pre-initials, and vowels. Syllables structure is the same, (P)IV. However, elements of the former aspirate series are now most frequently realized as fricatives. Similarly, the voiced stops /b d dʒ g/ are realized as nasals [m n ɲ ŋ] before nasal vowels

Initials:
/p t tʃ k/ <p t c k>
/f θ ʃ x/ <ph th ch kh>
/p' t' tʃ' k'/ <p' t' c' k'>
/b~m d~n dʒ~ɲ g~ŋ/ <b d j g>

The class of pre-initials is somewhat reduced (though not much, because xC and rC clusters have incomparable phonaesthetics). Thus, we're left with:

Pre-initials:
/s x r/ <s x r>

Again, /s x/ assimilate in voicing to a following initial. The fricative series /f θ ʃ x/ may be realized realized as strong aspirates, often with affrication, after the fricative pre-initials. Thus /sf sθ sʃ sx/ and /xf xθ xʃ xx/ may be [sp(f)ʰ st(θ)ʰ stʃʰ sk(x)ʰ] and [xp(f)ʰ xt(θ)ʰ xtʃʰ xk(x)ʰ], respectively. This is most frequent in clusters wit where the elements have the same or similar POA; thus /sθ sʃ/ are virtually always [st(θ)ʰ stʃʰ], and /xx/ is virtually always [xk(x)ʰ]. As mentioned, the aspiration is quite strong, with VOT in the range of 100-180ms (depending on POA).

The vowels have been majorly reduced, and the shape of the vowel system changed. It's now a skew-parallelogram, whereby /o/ is quite high and alternates with \[u\] and [ʊ], and [ɛ] varies considerably are the mid-front region, from [æ] all the way up to [ɪ]. Nasalization and tone are maintained as distinctive, again for phonaesthetic concerns.

Vowels:
/i/ <i u>
/o/ <o>
/ə/ <eo>
/ɛ ɔ/ <e oa>
/a/ <a>
+nasalization, and high/low tones <Vm, V́/V>

Root structure remains the unchanged.

The extenton suffix -lta is now -rka, and in order to maintain unambiguous parsing that syllable does not occur elsewhere in the language.

[Qwynegold]

The extenton suffix -lta is now -rka, and in order to maintain unambiguous parsing that syllable does not occur elsewhere in the language.

I think the L was a good idea.

[Elancholia]

Alright, I know this is very much a necropost at this point, but I have a longstanding sorta-naturalistic conlang that does something like this: lexical items are marked as the heads of phrases by lengthening their primary-stressed vowel. Secondary-stressed vowels in subordinated words also become de-stressed. For instance, from the roots

/tʃəˈɹan/
"to manufacture"

/tə̞ˌmaˈɹa.jə̞m/
"to observe"

A speaker could form

/tʃəˈɹaːn.tə̞.mə̞ˌɹaˌjə̞m/
"to manufacture observation"

or

/tʃəˌɹan.tə̞ˌmaˈɹaː.jə̞m/
"to observe manufacturing"

The language is omnipredicative, polysynthetic, noun-incorporating, and extensively zero- (or affixal-)copula. Relating roots in this way is one of the most basic elements of its grammar, as well its main strategy for derivation. Word order encodes very little other than saliency and topicality. Now, you can only "promote" something by one level, and recursion is done through the incorporation of pronouns and classifiers into the predicate (there is a classifier for events and actions, which is useful for complex statements with multiple verb-y elements), not by making primary vowels ever-longer -- but I thought it fit the topic of this thread fairly well, as corroboration on the thought experiment, as it were. Caveat that I haven't thought about it very much in a while, so the grammar isn't as developed as I'd like, but I think it's a fun idea.