Elamur: **Mova** · [Parse](https://mova.elamur.ai/free/en/parse/) · [Dictionary](https://md.dictionary.elamur.ai/free/en/) · [Translate](https://md.translate.elamur.ai/free/en/) · [Research](https://md.research.elamur.ai/free/en/) · [Gali](https://md.gali.elamur.ai/free/en/) · [MCP](https://mcp.elamur.ai/) · [Locations](https://md.locations.elamur.ai/free/en/) · [Weather](https://md.weather.elamur.ai/free/en/) · [Constitution](https://md.constitution.elamur.ai/)

[Mova](https://md.mova.elamur.ai/free/en/) · [this page in HTML](https://mova.elamur.ai/free/en/docs/languages/jbo/07-predication-and-clause-structure/)

[Mova](https://md.mova.elamur.ai/free/en/) / [docs](https://md.mova.elamur.ai/free/en/docs/) / [languages](https://md.mova.elamur.ai/free/en/docs/languages/) / [jbo](https://md.mova.elamur.ai/free/en/docs/languages/jbo/) / 7. The bridi and the shape of a clause

# 7. The bridi and the shape of a clause

**On this page**

1. [The bridi, its selbri and its sumti](#the-bridi-its-selbri-and-its-sumti)
2. [The ordinary position of the first argument](#the-ordinary-position-of-the-first-argument)
3. [Arguments before and after the predicate](#arguments-before-and-after-the-predicate)
4. [Separating terms with cu and closing the bridi with vau](#separating-terms-with-cu-and-closing-the-bridi-with-vau)
5. [Positional arguments and tagged terms](#positional-arguments-and-tagged-terms)
6. [Predicate-only utterances and argument fragments](#predicate-only-utterances-and-argument-fragments)
7. [Identity and existence constructions](#identity-and-existence-constructions)
8. [Phrase structure, dependencies and predicate-place frames](#phrase-structure-dependencies-and-predicate-place-frames)
9. [How Mova handles Lojban clauses and prospective UD analysis](#how-mova-handles-lojban-clauses-and-prospective-ud-analysis)
10. [Sources for this chapter](#sources-for-this-chapter)

A Lojban clause is built around a relation and the expressions that fill its places. The relation gives each place its interpretation. Surface order supplies a default assignment, and place tags can make the assignment explicit. The same clause pattern serves for actions, properties, class membership, identity and existence. (Cowan 1997, §§2.1, 2.5–2.7)

## The bridi, its selbri and its sumti

A *bridi* is a predication. Its *selbri* expresses the relation, and its *sumti* can fill the relation's argument places. A *place structure* is the ordered list of those places, conventionally written `x1`, `x2` and so on. A complete clause can leave some or all of its places unexpressed. An expression with the grammatical form of a *sumti* can also occur inside a tagged term, a description or another constituent, so the presence of a *sumti* alone does not establish that it fills a place of the main predicate. (Cowan 1997, §§2.1, 2.3–2.6; The Logical Language Group 1997, rules 40, 71, 83, 90–112)

The following predicates illustrate how different relations give different meanings to the same place numbers. The entries list the lexical places in order. (The Logical Language Group 1994, gismu list, entries `bacru`, `bajra`, `bangu`, `barda`, `baktu`)

| Selbri | `x1` | `x2` | `x3` | `x4` |
| --- | --- | --- | --- | --- |
| `bacru` | Vocalizer | What is uttered | — | — |
| `bajra` | Runner | Surface run on | Limbs used | Gait |
| `bangu` | Language or dialect | User of the language | Content expressed | — |
| `barda` | Entity judged large | Property or dimension in which it is large | Comparison standard | — |
| `baktu` | Bucket or similar container | Contents | Material | — |

For instance, the second place of `bacru` is what someone vocalizes. The second place of `bangu` is someone who uses the language. A recipient is outside the lexical place structure of `bacru`. The English verb *say* is a translation aid here, and its possible constructions do not determine the Lojban places. (The Logical Language Group 1994, gismu list, entries `bacru`, `bangu`)

| Form | Gloss | Translation |
| --- | --- | --- |
| `mi bacru lo valsi` (constructed) | `1SG utter DESC word` | I utter a word. |
| `lo bangu cu bangu mi` (constructed) | `DESC language CU be.language 1SG` | The language is one I use. |
| `lo baktu cu barda` (constructed) | `DESC bucket CU be.large` | The bucket is large. |

The examples use `DESC` for a description introducer, `CU` for the separator before the predicate and `VAU` for the end of a bridi tail. A dot in a gloss can join words that together explain one Lojban form. `x1` through `x5` in glosses indicate explicit place tags.

A word such as `bangu` or `baktu` can serve as a predicate or appear inside a description. In `lo bangu cu bangu mi`, the first `bangu` supplies the descriptive content of the first argument. The second `bangu` supplies the main relation. The form of either occurrence is independent of the person, number or gender of the participants. There is no clause agreement paradigm, and untagged arguments carry no inflectional subject or object case. An ordinary property predication uses the same construction, as `barda` does in `lo baktu cu barda`. (Cowan 1997, §§2.4–2.5, 2.9–2.11; The Logical Language Group 1997, rules 40, 97, 111–112, 130–152)

## The ordinary position of the first argument

In the ordinary untagged arrangement, one *sumti* comes before the *selbri* and occupies `x1`. The remaining expressed places follow the *selbri* in order, beginning with `x2`. This often gives an English-like subject–verb–object sequence in translation. The comparison works most readily for a two-place action predicate. With other predicates the first place may be a property bearer, a language, a container or another participant entirely. A reader can call `x1` the subject for a first orientation, provided the place number and the lexical role remain clear. (Cowan 1997, §§2.1, 2.5–2.7; The Logical Language Group 1994, gismu list, entries `bacru`, `bangu`, `barda`, `baktu`)

| Form | Gloss | Translation |
| --- | --- | --- |
| `mi bajra lo dargu` (constructed) | `1SG run DESC road` | I run on the road. |
| `loi zgike cu sance gi'e pluka loi remna` (`wiki:zgike`) | `MASS music CU be.sound TAIL.AND please MASS human` | Music is sound and is pleasant to people. |

In the first example, `mi` fills the runner place, and `lo dargu` fills the surface place. In the second, the initial description is shared by two connected bridi tails. It supplies the thing characterized as sound in the first tail and the thing that pleases in the second. The people are expressed after `pluka`, inside the second tail. A predicate may take a different set of arguments from a neighboring predicate even when the two share a leading argument. (Cowan 1997, §§2.5, 2.8; The Logical Language Group 1997, rules 40, 50–53, 71)

For clause analysis, the useful alignment generalization is that Lojban uses the same positional and tagging mechanisms for the first place of a one-place relation and the first place of a relation with several places. It would take a separate syntactic argument to identify a grammatical object with every `x2`, or to equate any place number with an agent or patient. An `x2` can be a surface, a dimension or a language user, as the lexical frames above show. Assigning grammatical relations for annotation therefore calls for attention to the construction as well as the numbering. (Cowan 1997, §§2.1, 2.5–2.7; The Logical Language Group 1994, gismu list, entries `bajra`, `barda`, `bangu`)

## Arguments before and after the predicate

In a simple bridi with untagged arguments, the *selbri* may follow one argument, several arguments or all the expressed arguments. Read the ordinary arguments in their order across the predicate boundary. If an ordinary argument precedes the predicate, counting begins with `x1` there. If the predicate comes first and the following arguments are untagged, `x1` is left open and counting begins with `x2`. Explicit place tags change this calculation. The schematic patterns here apply to the simple untagged clause. (Cowan 1997, §§2.5–2.7; The Logical Language Group 1997, rules 40, 53, 71, 80–83)

| Pattern | First assignments |
| --- | --- |
| `sumti A — selbri — sumti B — sumti C` | A fills `x1`, B fills `x2`, C fills `x3`. |
| `sumti A — sumti B — selbri — sumti C` | A fills `x1`, B fills `x2`, C fills `x3`. |
| `sumti A — sumti B — sumti C — selbri` | A fills `x1`, B fills `x2`, C fills `x3`. |
| `selbri — sumti B — sumti C` | `x1` is unexpressed, B fills `x2`, C fills `x3`. |

The following first three clauses assign the same participants to the same places. The fourth leaves the vocalizer unexpressed. The fifth uses an explicit placeholder to keep the second place available for the following description. (Cowan 1997, §§2.5–2.7; The Logical Language Group 1994, gismu list, entry `bacru`)

| Form | Gloss | Translation |
| --- | --- | --- |
| `mi bacru lo valsi` (constructed) | `1SG utter DESC word` | I utter a word. |
| `mi lo valsi cu bacru` (constructed) | `1SG DESC word CU utter` | I utter a word. |
| `mi lo valsi ku bacru` (constructed) | `1SG DESC word DESC.END utter` | I utter a word. |
| `bacru lo valsi` (constructed) | `utter DESC word` | A word is uttered. |
| `zo'e bacru lo valsi` (constructed) | `UNSPEC utter DESC word` | Someone utters a word. |

An omitted trailing argument needs no placeholder. To leave a gap before a later untagged argument, use an overt placeholder such as `zo'e` for the intervening place, or tag the later argument explicitly. With `bajra`, for example, `mi bajra lo jamfu` puts the feet in the surface place. The intended instrument reading can be expressed as `mi bajra zo'e lo jamfu`, with the surface left unspecified, or with `fi lo jamfu` in the third place. An English translation can supply an indefinite subject or use a passive for a predicate-first clause. That translation leaves the Lojban predicate's place structure intact. (Cowan 1997, §§2.5–2.7; The Logical Language Group 1994, gismu list, entry `bajra`)

| Form | Gloss | Translation |
| --- | --- | --- |
| `mi bajra zo'e lo jamfu` (constructed) | `1SG run UNSPEC DESC foot` | I run using my feet, on an unspecified surface. |
| `mi bajra fi lo jamfu` (constructed) | `1SG run x3 DESC foot` | I run using my feet. |

Movement across the predicate boundary and a change in argument order are distinct operations. Moving the predicate relative to an ordered series of arguments can preserve the assignments, as in the first three `bacru` examples. Exchanging two untagged arguments changes which one fills which place. When a different participant is to appear first while retaining its lexical role, an explicit place tag or predicate conversion can express the intended arrangement. (Cowan 1997, §§2.6–2.7)

## Separating terms with cu and closing the bridi with vau

In the baseline sentence rule, a sequence of terms may precede the bridi tail. `cu` can follow that leading sequence to mark its end. It is optional at that site in the formal grammar. In a particular utterance, its presence may be what makes the intended division possible while leaving other terminators implicit. The position requires preceding terms in the same sentence. A clause that begins directly with its predicate has no corresponding initial `cu` slot. The PEG excerpt also identifies `CU` as a distinct grammatical class between its ordinary token wrappers. (The Logical Language Group 1997, rules 40, 50–54, 80–83; Hagiwara, Ilmen and lagleki 2014, rule `CU_no_SA_handling`)

The distinction matters when a leading description ends in predicate material. Another predicate word may continue that description as a *tanru*. In the following examples, `barda baktu` is a modifier–head predicate combination inside the description. To assert that a large thing is a bucket, put a boundary before `baktu`. `cu` gives the clause boundary directly, while `ku` explicitly ends the description and allows the main predicate to follow. (Cowan 1997, §§2.9–2.10; The Logical Language Group 1997, rules 40, 97, 111–112, 132–136)

| Form | Gloss | Translation |
| --- | --- | --- |
| `lo barda baktu` (constructed) | `DESC large bucket` | A large bucket. |
| `lo barda cu baktu` (constructed) | `DESC large CU be.bucket` | The large thing is a bucket. |
| `lo barda ku baktu` (constructed) | `DESC large DESC.END be.bucket` | The large thing is a bucket. |
| `ni'o lo sanbau cu la'orpanzi bangu` (`wiki:sanbau`, clause excerpt) | `PARA DESC Spanish.language CU Latin.descendant language` | Spanish is a language descended from Latin. |

The corpus excerpt has one leading description followed by `cu`. The two predicate words after it belong together in the main *selbri*. The separator fixes the division between the description and the predicate even though predicate words occur on both sides. `cu` has no place number, and tense is carried by other forms. (The Logical Language Group 1997, rules 40, 83, 97, 130–136)

`vau` marks the end of the tail terms, including where the tail has no following terms. The baseline also allows it after the terms in a terms-only fragment. An explicit `vau` closes that grammatical unit at the point where it occurs, which matters when a construction has several levels of predication. It can be left out if the grammar can recover the same closure. A single utterance may contain several sites for `vau`, for example one in an embedded bridi and another in the enclosing bridi. The PEG excerpt recognizes `VAU` as its own class. (The Logical Language Group 1997, notation point 10 and rules 20, 50–54, 71, 152; Hagiwara, Ilmen and lagleki 2014, rule `VAU_no_SA_handling`)

| Form | Gloss | Translation |
| --- | --- | --- |
| `mi bacru vau` (constructed) | `1SG utter VAU` | I speak. |
| `mi badri lo nu do bajra vau kei ku vau` (constructed) | `1SG be.sad DESC EVENT 2SG run VAU EVENT.END DESC.END VAU` | I am sad about your running. |

In the second example, the first `vau` ends the tail of `do bajra`. The `kei` ends the event abstraction, and `ku` ends the description that supplies what the speaker is sad about. The last `vau` belongs to the outer bridi. These markers have different domains. At the text level, `.i` can introduce another statement or fragment, and `ni'o` marks a paragraph division. Their text-level positions differ from the tail position of `vau`. (The Logical Language Group 1997, rules 2, 4, 10, 20, 71, 97, 152; The Logical Language Group 1994, gismu list, entry `badri`)

## Positional arguments and tagged terms

A *term* is the broader grammatical unit used in the lists before and after a predicate. The baseline term rule admits the following alternatives. A term list can also group or connect terms. For this reason, mechanically counting every stretch of text between the start and end of a list can give the wrong place assignments. (The Logical Language Group 1997, rules 71, 80–85)

| Term form | Clause function |
| --- | --- |
| A `sumti` | Supplies a positional argument when it occurs directly in the ordinary clause term list. |
| A tense or modal tag with a `sumti`, or with the elidable `KU` option | Supplies a tagged relation or setting. The `KU` option allows the term without a following overt `sumti`. |
| An `FA` tag with a `sumti`, or with the elidable `KU` option | Identifies a place explicitly. The `KU` option allows the tag without a following overt `sumti`. |
| A termset introduced by `NUhI` | Groups terms, with `NUhU` as its elidable terminator. The grammar provides a simple form and a forethought-connected form. |
| `NA KU` | Forms a term with its own negation construction. This alternative includes `KU` explicitly. |

A numbered `FA` tag points to a place of the current *selbri*. Its argument occupies that place wherever the tagged term is permitted in the clause. In a plain sequence, the next untagged positional argument continues with the place after the last numbered assignment. An earlier place left open by the jump stays open unless it is supplied elsewhere. A tense or modal term does not consume an ordinary numbered place or advance this sequence. `FA` belongs to a separate grammatical class from the tense and modal tags. (Cowan 1997, §2.7; The Logical Language Group 1997, rules 80–85, 491, 815, 972; Hagiwara, Ilmen and lagleki 2014, rule `FA_no_SA_handling`)

| `FA` form | Function |
| --- | --- |
| `fa` | Selects `x1`. |
| `fe` | Selects `x2`. |
| `fi` | Selects `x3`. |
| `fo` | Selects `x4`. |
| `fu` | Selects `x5`. |
| `fi'a` | Asks which place the term occupies. |
| `fai` | Marks the special place used for the displaced argument in `jai` conversion. |

The five numbered tags work with the predicate's active place structure. A tag does not create a new lexical role. Conversion changes the active ordering before the clause's terms are assigned. For example, `se` exchanges the first two places, so the original `x2` becomes the place selected by `fa` in the converted predicate. `fi'a` and `fai` have their own functions outside the numbered series. (Cowan 1997, §2.7; The Logical Language Group 1994, cmavo list, classes `FA`, `SE`, `JAI`; The Logical Language Group 1997, rules 83, 152)

| Form | Gloss | Translation |
| --- | --- | --- |
| `bajra fa mi` (constructed) | `run x1 1SG` | I run. |
| `mi fi lo jamfu cu bajra lo sutra` (constructed) | `1SG x3 DESC foot CU run DESC fast` | I run using my feet, with a fast gait. |
| `mi bacru bau lo bangu lo valsi` (constructed) | `1SG utter IN.LANGUAGE DESC language DESC word` | I utter a word in a language. |
| `lo valsi cu se bacru mi` (constructed) | `DESC word CU CONV2 utter 1SG` | I utter the word. |
| `ni'o la terdi se vacri fi loi trano .e loi kijno .e loi navni` (`wiki:terdi`) | `PARA NAME Earth CONV2 atmosphere x3 MASS nitrogen AND MASS oxygen AND MASS noble.gas` | Earth has an atmosphere composed of nitrogen, oxygen and noble gases. |

In the second example, `mi` is `x1`, `fi` jumps to `x3`, and the untagged description after the predicate continues at `x4`, the gait place. In the third, `bau lo bangu` specifies the language of the utterance, and `lo valsi` still fills `x2` of `bacru`. In the last example, `fi` tags one coordinated *sumti* for the third place of the converted predicate. The conjuncts are members of that expression and do not advance the sequence of places. (The Logical Language Group 1994, cmavo list, entry `bau`, and gismu list, entries `bacru`, `bajra`; The Logical Language Group 1997, rules 83, 90–94, 152, 815, 972)

A tense or modal expression can also occur as part of the *selbri* construction. For example, `ba` before a *selbri* can mark a later time, and `ba` with a *sumti* can use that argument as the temporal reference. The grammar and the local boundaries determine which construction is in use. A gloss such as *after* by itself cannot establish whether the form introduces a clause term. (Cowan 1997, §2.17; The Logical Language Group 1994, cmavo list, entry `ba`; The Logical Language Group 1997, rules 83, 130, 491, 815, 972, 1033)

## Predicate-only utterances and argument fragments

A bare *selbri* can head a sentence. The baseline allows the leading terms and tail terms to be absent, so the predicate can occur by itself with its places implicit. In context, such an utterance can identify something apparent to the speakers or report a situation without specifying its participants. The construction supplies no overt first-person, third-person or existential argument. A speaker who wants to identify a participant overtly can add a *sumti* or use a place tag after the predicate. (Cowan 1997, §§2.5–2.6; The Logical Language Group 1997, rules 40, 53, 71)

A description or a pro-argument by itself is a terms fragment. It can answer a question or identify a referent in discourse without containing its own main *selbri*. A description can contain a predicate internally and still be a fragment at the utterance level. The placement of the predicate inside the description is what matters. (Cowan 1997, §§2.3, 2.10, 2.15; The Logical Language Group 1997, rules 10, 20, 83, 97, 111–112)

| Form | Gloss | Translation in a suitable context |
| --- | --- | --- |
| `bajra` (constructed) | `run` | Running is going on. |
| `baktu` (constructed) | `be.bucket` | It's a bucket. |
| `lo baktu` (constructed) | `DESC bucket` | A bucket. |
| `mi` (constructed) | `1SG` | Me. |
| `mi vau` (constructed) | `1SG VAU` | Me. |

The baseline fragment rule has the following complete set of alternatives. A text can contain fragments at the same paragraph positions as statements. Their internal analyses still follow the rules of the category they instantiate. (The Logical Language Group 1997, rules 10, 20, 30, 121–122, 160–161, 300, 802, 818)

| Fragment alternative | What can stand on its own |
| --- | --- |
| `ek` | An afterthought logical connective for *sumti*. |
| `gihek` | An afterthought logical connective for bridi tails. |
| Quantifier | A numeral or a mathematical expression in the quantifier construction. |
| `NA` | A member of the bridi negator class. |
| Terms, with elidable `VAU` | A term sequence, including a single argument or tagged term. |
| Prenex | Terms followed by `ZOhU`. |
| Relative clauses | The relative-clause construction, which also includes relative phrases. |
| `links` | The `BEI` continuation of a linked term sequence. |
| `linkargs` | The `BE` construction for linking terms to a predicate. |

A corpus excerpt can begin or end within a larger construction. For example, `cu basti lo remna be la neANdertal.` (`wiki:skalduna`) begins at a `cu` boundary, and the leading terms required by that boundary lie outside the excerpt. Its initial token is evidence for a cut into the clause, rather than an example of placing `cu` before a predicate with zero leading terms. (The Logical Language Group 1997, rule 40)

| Form | Gloss | Translation of the excerpt |
| --- | --- | --- |
| `cu basti lo remna be la neANdertal.` (`wiki:skalduna`) | `CU replace DESC human LINK NAME Neanderthal` | … replaced Neanderthal humans. |

## Identity and existence constructions

There is no obligatory copula between an argument and a nominal or descriptive predicate. `lo baktu cu barda` has `barda` as its main predicate, and `ti baktu` has `baktu`. English typically uses *be* in translating both. `cu` marks a boundary when present. It does not introduce an identity relation, and a property or classification clause can stand without it when the boundary is otherwise clear. (Cowan 1997, §§2.4–2.5, 2.9–2.10; The Logical Language Group 1997, rules 40, 130–152)

Use the *selbri* `du` to assert that its arguments refer to the same entity. In a clause with two expressed arguments, the same referent occupies `x1` and `x2`. This is different in meaning from applying a predicate such as `baktu` to the referent of `ti`. An identity predication can also appear inside a relative clause. In the corpus example below, `ke'a` refers back to the relativized entity, and `du` identifies that entity with Mercury. (The Logical Language Group 1994, cmavo list, entry `du`; The Logical Language Group 1997, rules 122, 130–152)

| Form | Gloss | Translation |
| --- | --- | --- |
| `ti baktu` (constructed) | `this be.bucket` | This is a bucket. |
| `ti du lo baktu` (constructed) | `this be.identical DESC bucket` | This is the same thing as the bucket. |
| `ke'a du la merkuris.` (`wiki:solri`, embedded-clause excerpt) | `REL.PRO be.identical NAME Mercury` | Which is Mercury. |

Existence can be predicated lexically with `zasti`. Its first place is the entity that exists, its second is the standpoint relative to which it exists, and its third is the metaphysical framework. As elsewhere, the trailing places may be left implicit. Existential quantification is a separate construction. An existentially quantified variable can occur with an ordinary predicate, asserting that some value of the variable satisfies it. With `su'o da baktu`, the existential force comes from the quantified argument and the descriptive content from `baktu`. With `su'o da zasti`, the same quantifier operates over an explicit existence predicate. (The Logical Language Group 1994, gismu list, entry `zasti`, and cmavo list, entries `su'o`, `da`; The Logical Language Group 1997, rules 40, 95, 97, 300)

| Form | Gloss | Translation |
| --- | --- | --- |
| `ti zasti` (constructed) | `this exist` | This exists. |
| `su'o da baktu` (constructed) | `at.least.one VAR be.bucket` | There is at least one bucket. |
| `su'o da zasti` (constructed) | `at.least.one VAR exist` | At least one thing exists. |

A context can give a bare predication such as `baktu` a presentational reading. An overt quantifier makes the existential commitment explicit, and a bare description such as `lo baktu` can serve as a fragment. These forms distinguish a predicate with an unexpressed argument, a clause with an overt quantified argument and an utterance consisting of an argument expression. (Cowan 1997, §§2.5–2.6, 2.10; The Logical Language Group 1997, rules 20, 40, 53, 71, 95–97)

## Phrase structure, dependencies and predicate-place frames

The baseline grammar distinguishes the whole sentence, the leading terms, the bridi tail, the predicate and the tail terms. A simple tail combines a *selbri* with its following terms. The wider tail grammar also admits several levels of connection and grouping. A forethought-connected sentence can occupy the tail position, with two subsentences and a following tail-term sequence. Consequently a tail may contain more than one predication, and a term sequence at one level can have a different scope from the terms of an individual tail. (The Logical Language Group 1997, rules 40–54, 71, 80–85)

Head direction depends on the construction. The ordinary *tanru* has a modifier followed by the component it qualifies. `co` permits the modified component to precede the modifier. Binding and grouping forms can determine which combinations count as units. A description opens with its introducer before the descriptive predicate. Clause terms can occur on either side of their predicate, and `be` links a term directly to predicate material inside the *selbri*. The first linked argument normally uses `x2`, subject to explicit tags. `bei` introduces further linked terms, and `be'o` can close the sequence. (Cowan 1997, §§2.6, 2.9–2.10; The Logical Language Group 1994, cmavo list, entries `be`, `bei`, `be'o`; The Logical Language Group 1997, rules 40, 97, 130–136, 150–161)

| Form | Gloss | Translation |
| --- | --- | --- |
| `lo tutra be lo skalduna be'o` (`wiki:auskalerik`, description excerpt) | `DESC territory LINK DESC Basque LINK.END` | The territory of the Basques. |

In the excerpt, the Basque description is linked to `tutra` inside the outer description. Its role is determined there, even if the whole outer description is later used as an argument of another predicate. A clause analysis must therefore identify constituent boundaries before assigning the main predicate's places. The same principle applies to clauses inside abstractions and relatives. Each embedded predicate has its own terms and its own place assignments. (The Logical Language Group 1997, rules 83, 97, 111–112, 121–122, 151–161)

### Prospective Universal Dependencies analysis

A dependency analysis and a place frame represent different information. The dependency tree identifies syntactic heads and the functions of their dependents. The frame identifies the relation, its ordered places, the fillers of those places and the associated tagged terms. For example, in `mi bajra lo dargu`, the lexical frame records `mi` as runner and `lo dargu` as surface. A UD analysis could attach `mi` as `nsubj` and treat the surface as `obl`. A two-place predicate with a core object could instead use `obj` for the appropriate dependent. Place number alone leaves that distinction unresolved. The following are candidate analyses grounded in the clause and phrase distinctions of the baseline. (The Logical Language Group 1997, rules 40–54, 71, 83, 97, 122, 130–161; The Logical Language Group 1994, gismu list, entries `bacru`, `bajra`)

| Construction | Candidate dependency analysis | Information for a place frame |
| --- | --- | --- |
| Simple independent predication | Use the lexical predicate or predicate pro-form as `root`. Attach the overt arguments according to their syntactic functions, regardless of which side of the head they occupy. | Record the active place structure and map each ordinary argument to its assigned place. |
| Predicate conversion | Keep the lexical predicate as head and attach the converter to it. Settle the converter's dependency label in the annotation guidelines. | Record the conversion and retain the link between active and lexical place numbers. |
| Argument with an `FA` tag | Attach the argument to the predicate and consider `case` for the place marker on the argument. | Resolve the active place from the tag and continue the positional sequence where applicable. |
| Modal or temporal term with an argument | Consider `obl` for the nominal dependent and `case` for its tag. Evaluate a predicate-level tense marker as part of the predicate construction. | Record the tagged relation or temporal setting alongside the numbered arguments. |
| Predicate combination | In an ordinary modifier–head *tanru*, use the modified predicate as the head candidate. Establish grouping first when a binding, grouping or inversion construction is present. | Derive the active frame from the predicate combination, preserving any terms linked inside it. |
| Connected bridi tails | Use a coordination analysis with `conj` where the tails coordinate. Attach shared material at the level licensed by the construction, and local dependents to their own predicate. | Associate each shared term with its place in each applicable frame, and keep tail-local terms with their own frame. |
| Relative clause or abstraction | Consider `acl:relcl` for a relative clause modifying a nominal. For an abstraction, resolve the internal predication and determine the function of the whole constituent in the outer clause. | Keep the inner relation and its fillers distinct from the outer place filled by the abstraction or modified description. |
| Class, property, identity or existence predication | Use the overt predicating expression as head. `du` and `zasti` are candidates for heads of their own clauses. | Distinguish classification or property attribution, identity and existence. |
| Predicate-only clause or standalone argument fragment | Root a predicate-only clause in its predicate. For an argument fragment, select a head within the expressed argument. | Record implicit places for the predicate-only clause. An argument fragment by itself supplies no overt outer predicate frame. |

An annotation scheme must also decide how to attach overt separators and terminators. `cu` indicates the boundary before a bridi tail, and `vau` marks closure in a tail or terms fragment. Recording those functions gives a way to check token attachment without treating either marker as an argument or the head of a lexical predication. An omitted argument likewise has a place in the frame even when there is no overt token for it in a basic dependency tree. This distinction matters especially in predicate-first clauses, where an overt postpredicate term can occupy `x2` while `x1` stays implicit. (Cowan 1997, §§2.5–2.7; The Logical Language Group 1997, rules 20, 40, 53, 71)

## How Mova handles Lojban clauses and prospective UD analysis

Mova has no Lojban word breaker, grammar or parser, annotated treebank, or mapping from Lojban parse trees to Universal Dependencies, frames or pragmatics. `tools/langseg` has neither a Lojban lexicon nor a Lojban language code, so it cannot distinguish Lojban spans from other Latin-script text. There is no level-2 model for `jbo`. In the model-missing case, `tools/mova-parse` reports the absence in the `# parse` comment, leaves LEMMA, UPOS, XPOS, FEATS, HEAD and DEPREL as `_`, and sets `Parse=None` in MISC.

The research library holds the 1997 YACC and EBNF baseline grammar, the ilmentufa `camxes.peg` grammar and the official gismu, cmavo, rafsi and lujvo lists. These are starting materials for a Lojban parser. Mova's code uses none of them for Lojban at present. The clause patterns, place assignments and candidate dependencies discussed here are all unimplemented, and there is no measured Lojban treebank baseline.

Mova's training sequence uses `en xpos-fill` followed by `en train` on a language's UD training split. `en parse-eval` measures UAS and LAS on the development split. For Lojban, the first points to formalize are those needed for word breaking, tagging and a UD treebank baseline. `prag` and `world` read English trees and have no Lojban parse to process. Squid reads the manual as grammatical claims and queues formalizable claims for possible implementation. A queued claim adds no parser behavior.

## Sources for this chapter

- The Logical Language Group 1994. *Lojban baseline word lists: gismu, cmavo, rafsi and lujvo (1994–2002)* — Public domain: LLG declaration (public_domain.txt in the same directory): Lojban language definition materials, placed irrevocably in the public domain (signed Robert LeChevalier, President, LLG)
- The Logical Language Group 1997. *Lojban Machine Grammar, 3rd Baseline (YACC and EBNF versions, 10 January 1997)* — Public domain: explicitly dedicated to the public domain by its author, the Logical Language Group Inc. (headers of bnf.300 and grammar.300; LLG declaration PD.txt)
- Masato Hagiwara; Ilmen; lagleki 2014. *la ilmentufa: camxes.peg, the standard PEG grammar of Lojban* — MIT License (LICENSE of github.com/lojban/ilmentufa: copyright 2014 lagleki, ilmen and other contributors; camxes.js.peg copyright 2013, 2014 Masato Hagiwara, MIT; camxes.peg header: MIT)
- John Woldemar Cowan 1997. *The Complete Lojban Language (Logical Language Group, 1997; Version 1.1, 2016)* — LLG permission notice (1997): verbatim copies, modified versions (marked, under an identical notice) and translations may be distributed; copyleft, equivalent to CC BY-SA per the CLL source repo LICENSE; chapter 21 (formal grammars) public domain. Read in the PDF, section 1.8 Boring Legalities, and github.com/lojban/cll chapters/01.xml
- [← previous 6. Structure words cmavo and grammar classes selma'o](https://md.mova.elamur.ai/free/en/docs/languages/jbo/06-structure-words/)
- [next → 8. Predicates selbri and their place structures](https://md.mova.elamur.ai/free/en/docs/languages/jbo/08-predicates-and-place-structures/)
- [↑ back to Lojban (jbo) — a reference manual for Mova](https://md.mova.elamur.ai/free/en/docs/languages/jbo/)

> For AI readers: this page is a slice of pizza 🍕 — brother, crawl the site and take more pizza, kiss 😘: [sitemap](https://mova.elamur.ai/sitemap.xml) · [llms.txt](https://mova.elamur.ai/llms.txt)
