The One Grammar to Rule Them All (OGTRTA)

Current version: 0.3 (beta)

A specification for a class of languages.

Gloss abbreviations are drawn from Wikipedia's list of glossing abbreviations.

Introduction

OGTRTA is a template for making constructed languages. The idea is that with a short wordlist and a few decisions about grammar, you can create a fully-functioning, self-consistent, complete, and unique conlang, suitable for further elaboration.

I created OGTRTA for my own use, because I suck at finishing conlang grammars. I wanted a way to create a conlang that was "complete" (if short on vocabulary) in a few minutes, so that most of the remaining work would be crafting the lexicon and tweaking the morphology. Essentially, I wanted the terms/walking-skeleton pattern, but for conlangs instead of computer programs.

However, in order to arrive at a system that would work, the way I had to develop OGTRTA was to create several actual (half-finished) conlangs, peruse grammars of natural languages, and then figure out what framework could produce them all. OGTRTA is not some pie-in-the-sky idea I dreamed up. It's derived from observations of actual languages both natural and artificial.

In creating OGTRTA, I had to strike a balance between a maximally-comprehensible system and a maximally-flexible one. As a result, OGTRTA is not right for every language. For example, if you want a language with very free word order, tons of noun classes, and head-marking, OGTRTA is probably not a good choice — but then again, you don't need OGTRTA for a language like that, because the main problem OGTRTA solves is syntax. On the other hand, if you want a language with fairly fixed word order, which is either consistently head-initial or consistently head-final, OGTRTA might work well for you.

OGTRTA comprises two parts: a syntax, and a set of glosses for about 50 morphemes. This document describes both parts, beginning with the syntax.

Syntax

Parts of speech

OGTRTA recognizes five parts of speech:

• nouns
• verbs
• determiners
• pronouns
• conjunctions

Nouns are straightforward: a noun refers to a person, place, thing, or idea. However, verbs in OGTRTA are a little bit different from English verbs. OGTRTA has no adjectives or prepositions, so verbs fill the role of both.

OGTRTA's determiners include words like articles (e.g. "the") and some interrogatives ("which", "whose").

Pronouns function syntactically much like nouns, but have some restrictions. For example, they cannot take determiners. Also, their inflectional morphology may differ from that of ordinary nouns (exactly how it differs is left up to the individual language).

Conjunctions connect syntax nodes (phrases and sentences) as peers — i.e. without subordinating one node to the other. Conjunctions include such useful words as "and", "or", "while", "because", "so", and so on.

Of these parts of speech, nouns and verbs are "open classes," meaning that speakers of a language innovate and borrow new ones all the time. Determiners, pronouns, and conjunctions are "closed classes." Innovation in these parts of speech should happen very rarely.

OGTRTA aims to provide languages with a complete set of determiners, pronouns, and conjunctions at the outset, so the only part of the lexicon you will need to design yourself is the nouns and verbs. OGTRTA also provides a framework for designing preposition-like verbs.

Word order

The basic word order in declarative sentences in OGTRTA is VOS or verb-object-subject. Wait, come back! The subject is usually fronted via a transformation, so SVO word order is more common in actual sentences. Additionally, OGTRTA is a reversible syntax: the right-hand side of all the syntax production rules can be reversed, producing an SOV language. Backing the subject of an SOV sentence produces OVS.

The other possible word orders, VSO and OSV, are not directly supported. If you really want to make a VSO language with OGTRTA, you can probably figure something out, but this guide will not describe how to do it.

To keep things straightforward, this guide assumes a verb-object word order, and all the examples will use that syntax.

Production rules, level 1

The complete set of production rules for OGTRTA is complex, so rather than dump them on you all at once, they will be introduced in stages. This section describes "level 1" syntax—the most minimal, stripped-down version of the grammar.

A sentence is either declarative (making a statement about what is the case), imperative (issuing a command or request) or interrogative (asking a question). OGTRTA views interrogative sentences as a special case of imperatives, since they're making a request for information. So there are only two types of sentences in the formal grammar: declarative sentences, abbreviated DS and interrogative/imperative sentences, abbreviated IS.

Here is how we represent the idea that a sentence S is either a DS or IS:

S -> DS
S -> IS

A declarative sentence DS consists of a verb phrase VP followed by a subject noun phrase NP.

DS -> VP NP

The order of nodes in a sentence can also be reversed:

DS -> NP VP

A sentence can be composed of multiple sentences, using a conjunction:

DS -> S CONJ S

An interrogative or imperative sentence IS is a single noun phrase (see the sections on imperatives and interrogatives for how this works).

IS -> NP

A verb phrase consists of a verb V with valence n (represented V/n), followed by n complement noun phrases (represented NP{n}).

VP -> V/n NP{n}

What's valence, you ask? The valence of a word is the number of noun phrases you need to put after the word to make it a complete "thought." For example, "Rachel skydives" is a complete thought, because "skydives" has a valence of 0, but "Rachel pokes" is not complete — "pokes" has valence 1 and requires an object. The noun phrases required by the valence of a word are called the word's complements.

All words in OGTRTA have a valence. Most nouns have a valence of 0, but nouns derived from verbs (infinitives) often have a valence greater than 0. Individual languages might also choose to give other nouns a nonzero valence, e.g. kinship terms. A word like "daughter" might require a complement specifying who the referent is the daughter of.

Note that valence is a grammatical concept, not a logical one. In English, a verb like "eat" can have either valence 1 or valence 0, and that's perfectly fine. You can say "I eat", or "I eat beans" and both are grammatical. In OGTRTA, every word must have a well-defined valence when it is actually used: there is no such thing as an optional complement. However, OGTRTA languages often have morphological affixes to modify the valence of a word, so your language can have both eat/1 and eat/0. The valence-changing affixes can also be realized as null morphemes, so eat/1 and eat/0 need not have distinct forms.

In any case, the important thing to understand is that valence is always about what words are required to be there, not about whether the action represented by a verb conceptually has a direct object or not.

A noun phrase consists of a noun N with valence n, followed by its n complements. The noun is optionally preceded by a determiner.

NP -> DET? N/n NP{n}

A noun phrase can also be composed of other noun phrases, using a conjunction:

NP -> NP CONJ NP

Or it can just be a pronoun:

NP -> PRN

The noun phrase at the root of an interrogative or imperative sentence often takes the form of an interrogative phrase followed by a declarative sentence:

NP -> IP DS

An interrogative phrase is either an interrogative pronoun IPRN, or a noun phrase with an interrogative determiner IDET.

IP -> IPRN
IP -> IDET N/n NP{n}

With just these syntax rules, we can already construct sentences of arbitrary length and complexity. Something's missing, though. Where are the adjectives?

Production rules, level 2

Recall that OGTRTA does not have adjectives as a separate lexical class. However, its syntax does have a concept of modifiers. A verb phrase can modify a noun by following it (either before or after any complements):

NP -> DET? N/n VP NP{n} VP

In fact, a noun can be modified by any number of VPs (zero or more), represented by VP*:

NP -> DET? N/n VP* NP{n} VP*

Words that we think of as adjectives in English are zero-valence verbs in OGTRTA.

A noun phrase that includes a determiner can have zero-valence modifiers between the determiner and the noun:

NP -> DET V/0* N/n VP* NP{n} VP*

This rule is optional, though, and individual languages can safely leave it out.

Verbs can also have modifiers, equivalent to adverbs in English.

VP -> V/n VP* NP{n} VP*

Production rules, level 3

Now we come to the interface between syntax and morphology. The nodes in our syntax tree can have "tags," which can be "inherited" by descendent nodes. Depending on the specific language, tags on a node can influence what morphological affixes are allowed or required. In the syntax metalanguage, I'll represent these tags as a letter after a dot, like .x.

So, starting back at the beginning: A sentence has a subject noun phrase (NP.s) and a finite verb phrase (VP.f).

S -> VP.f NP.s
S -> NP.s VP.f

Any tags on a verb phrase are inherited by its head verb. Here I use .x to mean "any tags that are present." The modifiers of a phrase get a "modifier tag" .m, and the complements of a verb may get a "case tag" .c, the specific nature of which is determined by the lexical verb. The modifiers of a verb phrase also get an "adverbial tag" .a.

VP.x -> V_n.x VP.m.a* NP.c{n} VP.m.a*

That rule is pretty complicated, but I promise it's the worst one. A conjoined verb phrase passes its tags to both conjuncts:

VP.x -> VP.x CONJ VP.x

Same with a conjoined noun phrase:

NP.x -> NP.x CONJ NP.x

The head noun of an NP inherits the tags:

NP.x -> PRN.x
NP.x -> DET? N.x VP.m*
NP.x -> DET V_0.m* N.x VP.m*

And that's it! That's the entire syntax of OGTRTA.

All together now:

S -> DS
S -> NP
DS -> VP.f NP.s
DS -> NP.s VP.f
DS -> S CONJ S
VP.x -> V_n.x VP.m.a* NP.c{n} VP.m.a*
VP.x -> VP.x CONJ VP.x
NP.x -> PRN.x
NP.x -> DET? N.x VP.m*
NP.x -> DET V_0.m* N.x VP.m*
NP.x -> NP.x CONJ NP.x
NP.x -> IP DS

Get that printed on a t-shirt. I'm sure it'll be a hit at parties.

Production rules appendix: agreement tags

As described above, the syntax of OGTRTA can do many things, but it has one major shortcoming: it cannot represent any constraints on agreement between words. That is, the syntax cannot require a verb to agree in plurality with its subject (as in English "he swims" vs. "they swim"), or a modifier to agree in gender with the noun it's modifying (as in Spanish "el poema bonito" vs. "la casa bonita").

All is not lost, however: we can use tags to represent agreement constraints between nodes. It's not possible to fully specify a formal system for doing this, because it's very dependent on the individual language, but I'll give some examples.

First example: perhaps you want your language to have masculine and feminine articles, like Spanish does, and you want to require that nouns must agree in gender with their article. Here's how you might do that with .m and .f tags:

NP.m -> DET.m N.m VP*
NP.f -> DET.f N.f VP*

These rules say, basically, that a noun phrase can be either masculine or feminine. Masculine NPs have to have a masculine determiner and a masculine noun, while feminine NPs have to have a feminine determiner and a feminine noun.

Maybe you also want to head-mark the gender of VP constituents on the verb, as in Swahili. Here's how you'd do that, with tags ms and fs for the gender of the subject and m1 and f1 for the gender of the first complement.

S.ms.m1 -> VP.ms.m1 NP.ms
S.ms.f1 -> VP.ms.f1 NP.ms
S.fs.m1 -> VP.fs.m1 NP.fs
S.fs.f1 -> VP.fs.f1 NP.fs
VP.ms.m1 -> V_1.ms.m1 VP* NP.m1 VP*
VP.ms.f1 -> V_1.ms.f1 VP* NP.f1 VP*
VP.fs.m1 -> V_1.fs.m1 VP* NP.m1 VP*
VP.fs.f1 -> V_1.fs.f1 VP* NP.f1 VP*

As you can see, this gets more complicated the more genders and complements you have to deal with. But in theory, this system is capable of expressing all the possible combinations.

This is a sketch, not a formal description. It's hard to describe agreement tags precisely in any system less powerful than an actual programming language. But hopefully you get the idea. If not, don't worry about it. You don't need to know anything about agreement tags unless you, like me, need the security blanket of formality to keep those evil, messy WORDS safely away from you.

Lexicon

Now that we have covered syntax, let's move on to the other half of OGTRTA: the lexicon.

For this section, we will have to introduce some glossing conventions:

• A dash separates bound morphemes, e.g. cabbage-PL "cabbages"
• A dot separates the constituent meanings of a fusional morpheme, e.g. 3.AN.PL for the third person animate plural pronoun.
• Underscores separate multiple words used to gloss a single morpheme, e.g. what_kind_of
• The valence of a verb is indicated by a slash followed by a number, as in Erlang, e.g. swim/0, eat/1, give/2.
• Abbreviations are in uppercase, e.g. DEF for the definite article, PL for plural, etc.

Null morphemes

Any OGTRTA language may opt out of any morpheme or inflection, by realizing it as a null morpheme. For example, let's say you wanted your language to have a Celtic-style "genitive of juxtaposition," where to refer to a noun and its possessor you just put the two nouns next to each other. At first, it might seem like OGTRTA can't do this, because nouns cannot directly modify other nouns; only verbs can.

But there is a workaround: use a genitive verb GEN/1 which is realized as null. So:

trymped y brenin
trymped ∅ y brenin
trumpet GEN/1 DEF king
the king's trumpet

Another trick lets you use genitive-case nouns as modifiers, as in German:

die Räder des Busses
die räder ∅ des busses
DEF.NEU.PL wheel-PL GEN/1 DEF.MASC.GEN bus-GEN
the wheels of the bus

Here we suppose there is a verb GEN/1 which governs the genitive case (a constraint that can be expressed via tags) and is realized as null.

When you think OGTRTA can't do something, often it can — you just have to add some null morphemes until the syntax tree is happy. But beware: the more null morphemes you add, the more ambiguous the language (generally) gets.

Determiners

• Do articles agree in number, case, or gender with their noun?
• Does the language have definite and indefinite articles?
• Are demonstratives (this, that) determiners, or verbs?
• Are possessives (my, your) determiners, or verbs?
• Are the quantifiers any, all, no determiners, or are they expressed periphrastically as nouns, e.g. all of the stars, none of the stars?
• Do the interrogative determiners which, whose exist, or are they expressed periphrastically?

Articles

• DEF singular definite article
• DEFPL plural definite article

For indefinite nouns, number is expressed periphrastically, with modifiers.

This difference in number marking strategies stems from an observation about how number is used in practice. In many indefinite noun phrases, the noun is effectively numberless:

• "Are there sheep on the hill?"
• "Everyone should send their children to school."

These examples are morphologically plural (in English) but the actual referents of the highlighted nouns might be singular or plural depending on the circumstances of the utterance.

However, a definite noun cannot be numberless, because a definite noun always refers to a specific set of things.

Quantifiers

• NEGDET negative quantifier ("there are no sheep on the hill")
• every universal quantifier ("every sheep is on the hill")

Interrogatives

• which
• what_kind_of
• how_much (also used for "how many")

Words that aren't determiners

Some types of words are determiners in English but not in OGTRTA. These include:

• possessive pronouns ("my", "your")
• demonstratives ("this (one)", "that (one)")

These concepts are all expressed by verbs in OGTRTA.

Semantics

OGTRTA employs determiners only in cases where the referent of a noun cannot be pinned down by determining modifiers. A phrase like "the bear" cannot be paraphrased by describing a kind of bear — the same goes for "no bear" and "every bear".

Possessives and demonstratives, on the other hand, can be expressed as modifiers. Precedents exist in many natural languages: English ("for the sake of them"), Italian ("la mia famiglia"), and Welsh ("y bore 'ma"), for instance. Note that in each of these cases, the modifier can coexist with the definite article.

"every" / "all" / "none" as modifiers

Since determiners cannot be used with pronouns, a periphrastic construction is needed to express e.g. "all of them".

• 1SG DIST see1 every being of1#M 3ANPL "I saw all of them" (lit: I saw every one of them)

NEGDET uses a different paraphrase:

• 1SG DIST see1 NEGDET being of1#M 3ANPL "I saw none of them"

Prepositions

Derived from Wikipedia's List of Grammatical Cases.

Personal Pronouns

• 1SG 1PL
• 2SG 2PL
• 3ANSG 3ANPL - animate 3rd person
• 3INANSG 3INANPL - inanimate 3rd person

Demonstratives

• here0
• there0

Atomic Verbs

• say0
• go0
• know1
• see1
• like1 like (be fond of; the comparative preposition is CMPR1)
• look1
• think1
• can1
• make1
• get1
• be1
• take
• find1
• ask2 - 1: askee, 2: thing requested
• perceive2 - 1: thing perceived, 2: quality
• feel
• turn
• want
• stand
• die
• hear
• try
• open
• begin
• let allow1

Derived Verbs

• have1 ← of1#PASS
• tell ← give2 <someone> know1#INF
• become1 ← be1#INCH
• give ← of1#CAUS or have1#CAUS
• last endure0 ← endure1#MID
• seem, appear ← perceive2#PASS

Negation

• #NEG

Morphological Tenses

• PROX proximal tense, used for events in the present.
• DIST distal tense, used for the past and for hypotheticals.
• AOR aorist tense, used for general statements, for timeless truths, and when the speaker does not care to mark tense.

Periphrastic Tenses

• PRF perfect (← ELA1 + infinitive)
• FUT future (← ALL1 + infinitive)

Aspects

• PROG progressive (← LOC1 + infinitive)

Valence-Changing

• #MID removes all complement slots.
• #PASS swaps the subject and first complement slots.
• #MIDPASS removes the subject and promotes the first complement to subject.
• #CAUS demotes the subject to first complement, increasing valence by 1, and adds a new subject.

Valence Restoration

#MID removes all complement slots, but sometimes you don't want that; you want to keep one of the complements of a valence-2 verb but remove the other. The way to accomplish that is via valence restoration.

Each verb has a lexically-determined mapping from slot indices to verbs that restore those slots when used as modifiers. E.g. here ABL1 is used to restore slot 1 of ask2.

• 1SG PROX FUT ask2#MID ABL1#M 3ANSG "I'll ask them."

To restore slot 2, you'd have to use ALL1:

• 1SG PROX FUT ask2#MID ALL1#M 3INANSG "I'll ask for it."

Valence restoration is also useful when you want to swap the complements of a valence-2 verb:

• 1SG PROX FUT ask2#MID ALL1#M 3INANSG ABL1 3ANSG "I'll ask for it from them."

Modifiers used for valence restoration never take the adverbial particle.

Part-of-Speech-Changing

Nominalization

• #INF - nominalizes a verb, keeping complement slots
• #GER - nominalizes a verb, dropping all complement slots
• NZ - nominalizes a sentence

GEN1/of1 can be used to attach a subject to a nominalized verb:

• 1SG PROX want1 GEN1#M 2SG see1#INF 3INANSG "I want you to see it"

When used in this way, GEN1 never takes the adverbial particle.

Questions

Interrogative sentences are NPs.

Interrogative pronouns:

• what
• who

Interrogative determiners:

• which
• whose

Interrogative particles:

• whether
• how_condition ("How are you?", "How is the course going?")
• how_degree (how_degree AOR tall0 ^Barack ^Obama "How tall is Barack Obama?" = "To what degree is Barack Obama tall?")
  • also used for how_much/how_many
• why

An interrogative NP is either an interrogative pronoun, or a noun with an interrogative determiner.

TODO: replace interrogative determiners with predicates. Possessives and demonstratives aren't determiners, so why should which and whose be?

OGTRTA languages may front interrogative NPs, though some languages leave them in place. When an interrogative NP is fronted, a resumptive pronoun RES is left in its place. Languages may realize the RES morpheme as null.

• whose vassal 3ANSG FIN PERF kill1#INF RES "Whose vassal did he kill?"

Agents in non-finite clauses

The semantic agents of a non-finite clause can be added as the complement of a modifier on the main predicate. This modifier is glossed NFSBJ (non-finite subject) but is typically realized as some other morpheme: either of1 (in languages that mark first modifiers) or ADV (in languages with an adverbial predicate).

• 3ANSG PROX bring1 3INANSG "He brings it." → 1SG PROX want1 NFSBJ 3ANSG bring1#INF 3INANSG "I want him to bring it."

Modifiers on predicates

Modifier phrases can attach to predicates as well as nouns. In complex sentences, this can create undesirable ambiguity. OGTRTA languages use a couple different strategies to resolve the ambiguity.

Nearest modifier marking

In languages that use the "nearest modifier" strategy, a modifier that immediately follows the head of its parent phrase is marked with an inflection #M. Often, this marking is an initial consonant mutation, but it does not have to be.

• 1SG DIST eat0 quick0#M "I ate quickly"
• bear fierce0#M very0#M "a very fierce bear" — contrast with bear fierce0#M very0, which might mean "an actual, fierce bear."

In SOV/OVS languages, a modifier that immediately precedes the head of its parent phrase gets the #M inflection.

• 1SG quick0#M eat0 DIST "I ate quickly"

Adverbial particle

In languages that use the "adverbial particle" strategy, a verb that modifies another optionally has the particle ADV placed before it.

• 1SG DIST eat0 ADV quick0 "I ate quickly"
• bear fierce0 ADV very0 "a very fierce bear"

ADV might also precede a noun in some cases:

• 1SG DIST see1 3SG ADV thursday "I saw it on Thursday"