To start with, three rules will be introduced that are the guts of the grammar of Tyl-Sjok. Because of their importance and frequence of mentioning in this document, they will be called the Basic Rules or simply the Rules.
The definition about who is agent and patient is given in detail in Section 5.1.1.
This chapter gives a detailed overview of the syntax. Some of the following chapters will specialise in word categories found in other languages, e.g. English, and show how their function is modelled in Tyl-Sjok.
Tyl-Sjok has only two syntactic word categories: governors and particles. Phrases and governors may require other phrases to be attached to them to make them complete. This is called saturation. The phrases that must be added will be called sub-ordinates or dependents, while their governor is called head. Usually, it is easiest to think of governors as verbs and of sub-ordinates as agent or patient.
Please note that the term `saturation' is used a little bit off the standard in Tyl-Sjok. It is not used for adding agents to verbs. This is done, because adding agents and forming genitive constructions is the same thing in Tyl-Sjok. And because you can add another genitive modification to virtually anything, it is not a process of saturation, but of extension. However, the valence of verb may still require that an agent be added via genitive.
By this naming convention, a saturated verb is one that behaves like an intransitive one, i.e., all the objects are added.
Sentences are formed by three principles.
Deciding whether a genitive or an agent binding is used in another language will depend on whether you regard the sub-ordinate phrase as a verb or a noun: `my drinking of coffee' and `I drink coffee' will be the same phrase in Tyl-Sjok.
Three categories of particles will be distinguished: nullary (at the end of a phrase), unary (directly before a phrase) and binary (between two phrases).
FIXME: Topic marker? Composite nouns (like `xiōng dì jiě mèi' in Mandarin)?.
Tyl-Sjok is a high context language. Sentences are often so ambiguous that translating them into another language requires a lot of understanding of the underlying text. Although Tyl-Sjok has the means of being precise, the usual way of forming sentences is to rely on the context and leave out everything that is clear.
The Tyl-Sjok sentence structure makes ambiguity problems very overt, as there are no signals that indicate how sentences are embedded. The boundaries of auxiliary clauses must be inferred from the context.
Take the sentence
| good | man | see | me |
The above sentence can be translated in several ways, because the order of composition is unclear:
This is possible, because the action of perceiving the man can be the patient to the verb `good' (Note that man see me good is also a correct translation for `It is good that a man sees me', with a slightly different meaning, see Section 5.1.1 for details).
To specify what precisely is meant in the ambiguous sentence above, Tyl-Sjok offers the possibility to clearly disambiguate the meaning, though the meaning is usually assumed to be understood from the context.
Tyl-Sjok's grammar should fulfil the following constraints.
Furthermore, it was decided that all parts that are clear from the context may be simply left out and that the context plays an important role.
Many possible ways of composition were tried and checked for the constraints. The following word orders were considered to be used, because they were the only ones, apart from total reversal of the words, that fulfilled all needs. The decision for this and not the reversed order was made, because the author liked it this way.
To make things clear, the author has not renamed the standard word order schemata to match the active case marking in Tyl-Sjok. We still say write SVO (subject verb object) instead of AVP (agent verb patient), which would be more precise.
Although Tyl-Sjok verbs never have two patients directly, the curried version (a verb that takes a patient and yields a verb that takes a patient) does exist and, of course, is isomorphic. So this problem will occur in Tyl-Sjok.
The sentence i drink tea good has 5 analysis trees when including the possibility of an ellipsis.
FIXME: draw the trees.
.
Both saturation and genitive attach to the right.
The sentence `good drink tea i' has 9 analysis trees (6 of them because of an ellipsis and the combination of `tea i' as the single patient/agent).
The sentence `drink good tea i' has FIXME: count analysis trees.
The author chooses SVO as the word order for Tyl-Sjok, because it has fewest problems and nice properties.
The problem of double saturation, e.g. of verbs like `to give' or `to tell' that usually take two patients, has to be taken into account. Often, additional verbs are used. Other verbs, like `to be like', which have two patients, but no agent, are allowed to be used without additional verbs.
When trying to translate I drive car closely into another language, usually genitive constructions will be used: `My car driving.' This is, because in Tyl-Sjok, adding an agent is often the same as genitive constructions in other languages.
We will investigate what other languages use a genitive for to see what the difference is compared to Tyl-Sjok.
Genitive, in general, adds a specification of a class or kind. This is a vague defition, so the following list shows possible relations a genitive can express in other languages. Maybe there are more.
In Tyl-Sjok, there are agents and patients rather than subjects and objects. So these shades of the genitive will be called agentive and patientive.
Languages vary in what they allow in genitive constructions. To allow
comparisons, in the following, the term `genitive' will be used even
in languages that do not have a genitive case. It seems sensible to
use it for `of' constructions in English, `de' in French, `de' in
Mandarin, compound nouns in German and English, etc. Further note
that the term `partitive' is used for a special case in Finno-Ugric
languages which is not used for (exactly) the semantic
relations given for the `partitive' item above.
The above classifications must be distinguished to understand Tyl-Sjok's system. Tyl-Sjok's system is totally different: genitive, in general, is only possible of the modifier is the controller. The reason for this is the agent-patient structure in Tyl-Sjok.
Recall that we will call the attachment of an entity to the right of a phrase a genitive construction in Tyl-Sjok.
Obviously, agentive usage is possible in Tyl-Sjok and used for the literal translation of the sentence at the beginning of this section. Agentive genetive is used to attach an agent to verbs. So this is the basic form.
Because of the structure, we can immediately exclude
patientive usage, since that is patient saturation in Tyl-Sjok.
In Tyl-Sjok, patientive usage is not possible, because patients
must be put behind the word being saturated.
In Tyl-Sjok, appositional usage is not possible, but can be replaced by a patient saturation: using a predicative verb with a patient is possible so no copula is needed. E.g. in German, appositional isn't possible either (`city of New York' in German: `die Stadt New York' which is a normal apposition without genitive marking).
| City | Nju-jok | ||||||||||||||||||
| `New york is a city. (This can be used for embedding.) | |||||||||||||||||||
So Tyl-Sjok does not really have appositions, because they are
immediately verb-patient constructions. E.g. `Brutus caesarem
imperatorem interficet' 
Ty-lu-tus kill imperator
Se-sal. Here, `imperator' is a predicate, thus, a
verb.
For numbers, multiplicative usage is additionally possible in Tyl-Sjok. This could be interpreted as selectative genitive of an infinite amount (if Tyl-Sjok allowed selectative genitives, which it doesn't). There is no plural marker in Tyl-Sjok, so it is not distinguishable whether to analyse partitive or multiplicative.
The analysis would be something like : `three apple: `apple'=patient of `to be three'.
| three | apple | ||||||||||||||||||
| Three apples | |||||||||||||||||||
| `apples are threed.' | |||||||||||||||||||
Multiplicative words like `once', `twice', etc., are expressed in Tyl-Sjok by simply using a number before a verb.
| I | three | hit. | |||||||||||||||||
| I hit three times. | |||||||||||||||||||
| `I (have) three hits'. | |||||||||||||||||||
In Tyl-Sjok, quantitative usage is possible in the same way as multiplicative, but for uncountable things. An analysis would be that in `cup tea', `tea'=patient of `to be a cup' (although the tea occupies the cup, the cup controls the tea by defining its amount). See the system by the similarity between `three tea', `cup tea' and `three cup tea'.
Final usage is possible in Tyl-Sjok: `tea cup' can be analysed as `the cup is teaed' or `the cup has tea' or, because there are no marked moods, `the cup may/shall have tea'.
An alternative analysis would be that, if you want to stress the final
meaning, the verb `for/to serve' may be used: `for tea cup': `the cup
is for tea'. Then `tea cup' is a shortened form..
The ancestor is always in control.
Sometimes, the sentence keeps quite clear by this:
| tall | father | peter | |||||||||||||||||
| Peter's father is tall. | |||||||||||||||||||
This is, because no descriptive meaning is possible. It would have to be rendered with a coordination:
| tall | peter | SEQ | father | (he). | |||||||||||||||
| Peter, who has a father, is tall. | |||||||||||||||||||
However, sometimes, the genitive construction is ambiguous:
| tall | father | son | |||||||||||||||||
| The son's father is tall. | |||||||||||||||||||
| The father's son is tall. | |||||||||||||||||||
You would use a REF particle then to change the focus. This happens at top-level, too: even very short sentences might be ambiguous:
| father | son | ||||||||||||||||||
| The father has a son. | |||||||||||||||||||
| The son has a father. | |||||||||||||||||||
When the relation is on the same level, one of the participants acts like a static verb, the other as a patient. The modifier then comes first:
| brother | sister | ||||||||||||||||||
| The sister has a brother. | |||||||||||||||||||
| the brother's sister | |||||||||||||||||||
In Tyl-Sjok, selectative genitive is directly possible, all other's aren't, but are expressed by selectative.
For groups or parts, the group/part is in control of entities being part of the group.
Note again: it cannot be seen which one is modified if control defines the order. Only if both parties have equal control, the second entity is modified by the first.
To disambiguate the referent in a matrix clause, use the REF particle:
| family | REF | father | |||||||||||||||||
| father of the family (selectative) | |||||||||||||||||||
| REF | family | father | |||||||||||||||||
| family of the father (aggregative) | |||||||||||||||||||
| group | REF | person | |||||||||||||||||
| a person of the group (selectative) | |||||||||||||||||||
| REF | group | person | |||||||||||||||||
| a group of people (congregative) | |||||||||||||||||||
The same holds for quantifiers: phrases like `one of us', `all of us', `some of us' etc., are translated in the same way:
| We | some | eat. | |||||||||||||||||
| Some of us eat. | |||||||||||||||||||
| WHICH | they | two | |||||||||||||||||
| who of the two | |||||||||||||||||||
Please note: the verb `to have' may be used for possesive usage, if the possession is to be stressed.
For selectative, aggregative and congregative usage, `be part of' has to be used. Note that `to have' can often be left out if the context makes the situation clear.
Caution: Do not mix up the usage! In Tyl-Sjok, `to have' can neither express selectative, nor aggregative, nor congregative. You have to be more precise if you use a verb. If you don't want to be precise, leave out the verb.
Probably wrong:
| have | leg | I | |||||||||||||||||
| I have a leg. (But this sentence is not about mine.) | |||||||||||||||||||
Instead:
| be-part-of | person | leg | |||||||||||||||||
| person | leg | ||||||||||||||||||
| the person's leg | |||||||||||||||||||
| (but also: the leg's person) | |||||||||||||||||||
Possession immediately implies control.
| He | like | you | car. | ||||||||||||||||
| He likes your car. | |||||||||||||||||||
(Note: as an experiencer, `he' is in potential control, and is therefore assigned the agent slot.)
The control for a possessive is also selected if the possessed is also in control, too:
| He | like | you | friend. | ||||||||||||||||
| He likes your friend. | |||||||||||||||||||
But not the contrast to the following paragraph.
In the phrase `the slaves's master', the master is definitely in control and possession is vice versa. In this case, the slave cannot get a controlling agent slot.
In Tyl-Sjok, pure locative usage is not possible. A location verb is usually used. However, a genitive can imply some kind of partitive meaning, so in the above example, street festivals not only take place in New York, but have become part of it. In this case, locative meaning can be expressed by using a genitive in Tyl-Sjok. This is quite similar to German and English usage.
Very closely related to the agentive genitive is the initiative genitive.
In Tyl-Sjok, initiative is expressed by using the verb and possibly a NULL particle: `the driver of the car' = NULL drive car. The NULL particle usually implies that it is referred to, so no REF particle is needed. Of no reference is wanted, even NULL is often dropped.
So drive car may translate to the following sentences.
| there's car-driving | have drive car | |
| someone drives a car | SKIP drive car | |
| driver of the car | NULL drive car | |
| the car-driving | REF drive car | |
| the driven car | drive REF car | |
| to drive a car | drive car REF |
Cf. Chapter 8 about particle usage. In such a short sentence, usually the particles are used unless the context does not allow that.
Further note that `to drive' is translated to `to move' in Tyl-Sjok.
And because there is a mechanism called default verbs (see
Section 
), the verb may even be dropped
completely.
A special genitive is used for ordinals: `fourth event' = `event four'. An analysis would be: the event controls the fourth position. See the system by the similarity between `imperator Se-sal' = `Cesar, the imperator' and `imperator one' `the first imperator'.
Note that the ordinal genitive may be used for adjectives, too. It then indicates superlative or order: red one = `first red' = `the reddest'. burn one = `burn first'.
FIXME: Check that this is feasable.
Conceptionally uncountable things like `coffee' are directly countable in Tyl-Sjok by assuming an implicit partitive before the counting takes place. This is like in English (or German): `two coffees' is interpreted as `two portions of coffee'.
Tyl-Sjok distinguishes the level of saturation for every governor. Unsaturated words are often referred to as verbs. Verbs may subcategorise to have their agent and/or patient slot filled in in order to be saturated.
Verbs that have only one slot to saturate are called intransitive verbs. Those that have two are called transitive verbs.
In Tyl-Sjok there are no verbs that have more slots that agent and patient. However, valence sometimes needs more slots (e.g. the verb `to give' or `to be like' (which would need two patients)). Auxiliary constructions will be described in the following section.
Saturated composed phrases behave exactly like saturated single word lexicon entries.
Many `nouns' can be used as stative verbs in Tyl-Sjok. E.g. to express that there is fire, simple say `fire'. Usually, sentences will use a bit decoration around these monosyllabic utterances, but the principle is just like that. Static verbs of valence 0 have no agent or patient slot, so they are already saturated.
Verbs that only have an agent slot are e.g. `to walk'. Agents are attached to the word to the left, so `the man walks' would be rendered as follows:
| man | walk | ||||||||||||||||||
| The man walks. | |||||||||||||||||||
This is a saturated phrase, so it can be used in other phrases. E.g.
| man | walk | talk | |||||||||||||||||
| The man, who walks, talks. | |||||||||||||||||||
| The man walks and talks. | |||||||||||||||||||
In such a construction, one of the constituents of the embedded phrase is `exported' into the matrix clause. It the case above, the only logical export would be `man'. In ambiguous cases, there is a special particle (REF) to mark the exported entity in the sub-ordinated clause. Note that all sub-phrases may potentially be exported. During the following sections, this will probably become clear.
Examples are `to be red'. The sub-ordinate usually has no control over being red, so it is assigned the patient slot. The patient is attached to the right of the governor.
| red | book | ||||||||||||||||||
| The book is red. | |||||||||||||||||||
Again, this is a saturated phrase and can be used in a matrix clause.
| man | read | red | book. | ||||||||||||||||
| The man read a red book. | |||||||||||||||||||
Tyl-Sjok has something that maybe called promotion. It is the metamorphosis from valence 0 to something else, e.g. valence 1p in Tyl-Sjok. This means that `the man' can be used as `being a man'. However, using the verb `be equal to' is also possible. This verb is often dropped in Tyl-Sjok in almost all cases even in embedded phrases, so it is more likely to assume valence 1p for `the man' then. Another example would be `the taste' which is often used for `to taste' as if it had valence 1p.
Other valence promotinos are also possible: `car' may be used as a verb with an agent. It then means `to drive a car', denoting the `default action to do with a car'. This is similar to English (especially American): `to party' is the default for `to have a party'.
Another examples is `food' that can be promoted to `eat'. There are not many other things to do with food, so eating can be inferred from the context.
Actually, the whole phenomenon may be called demotion when viewed from the other end of the scale (`food' is the patient of the verb `to eat'). This may be a better explanation for why NULL eat means `eater' and eat NULL means `the eaten' or `food'.
The group of valence 2 verbs that have both agent and patient is long. An examples is `to read'.
| man | read | book | |||||||||||||||||
| The man reads a book. | |||||||||||||||||||
In this example, promotion of `book' may be used, too:
| man | book | ||||||||||||||||||
| The man reads a book. | |||||||||||||||||||
Some verbs naturally have two patients but no agent. For clarification, an auxiliary is often used to attach the second patient, usually either `to harass', `to please' or `to control'. In most phrases, however, it is common to drop `to harass/please/control'. Examples are `to be like', `to be similiar to', `to be equal to'.
| (equal-to) | imperator | (control) | Cesar | ||||||||||||||||
| Cesar is imperator. | |||||||||||||||||||
Both `to control' and `equal-to' are usually left out. Then, it looks as if the valence 0 word `imperator' has been promoted to valence 1 with the patient Cesar. (The precise analysis is really irrelevant in Tyl-Sjok).
An example for a valence 3 verb is `to give'. The recipient is is rendered either using `to take' or `to arrive-at' in Tyl-Sjok.
| you | take | I | give | book. | |||||||||||||||
| I | give | book | go-to | you. | |||||||||||||||
| I give the book to you. | |||||||||||||||||||
The meaning of these sentences is slightly different. The former expressed active reception, while the second does not.
Some governors can modify others and yield something that is not saturated. Often, valence is preserved. This is called modification. An example is `good' which can modify `to taste' or `to speak'.
| man | good | speak | Tyl | Sjok | |||||||||||||||
| The man can speak Tyl-Sjok well. | |||||||||||||||||||
The auxiliary `can' may be left out in Tyl-Sjok, as it is clear from the context.
This modification can be explained easily for promoted valence 1p verbs like `to taste'. The following sentence would be no surprise:
| good | taste | meal | |||||||||||||||||
| The meal tastes good. | |||||||||||||||||||
Because actually, `taste' is a valence 0 word which was promoted to valence 1p. So we could also have used `to have' here:
| have | good | taste | meal | ||||||||||||||||
| The meal tastes good. | |||||||||||||||||||
| `Good taste is with the meal. | |||||||||||||||||||
This explanation, however, it not applicable to verbs like `to speak', where a modification of the verb is clear. Of course, in Tyl-Sjok, `with' can still be used (see Section 5.1.1).
Other examples of modification verbs are `very' or `extremely', etc.
Particles are different from governors, because they change the structure of the sentence or the way the sentence must be analysed. They usually do not have a meaning on their own and can, therefore, not be used without a sentence.
The following categories of particles might exist:
This categorisation of particles into these groups is not very important, it is only given for completeness.
It was not said yet how and when the Rules shall be applied to form a good sentence. To explain this, the following paragraph introduces functional categories.
Although Tyl-Sjok does not have concepts like verb, nouns, adjective, etc. at its syntax level, in actual sentences, functions can be identified. The language has concepts like entities, actions, properties etc. These word categories are called functional categories.
The standard way of thinking is to assume there are two basic functional categories: nouns and verbs. Verbs often require some of the Rules to be applied to form a correct sentence. This is called sub-categorisation.
Intransitive verbs like `to run' require at least one genitive construction in order to bind the agent. Transitive ones like `to write' may require one or more applications of the saturation rule.
Verbs require the application of Rules to a degree that can be anywhere on a scale between `always' and `never'. Additionally, because Tyl-Sjok is a high context language, if the context can provide the necessary clarifying information, most things may be left out of the sentence.
Adjectives in Tyl-Sjok are verbs that require either a genitive binding for the agent, or a patient binding, whose property is described. This depends on the type of adjective.
E.g. you could think of the word `red' to mean `to be
red'. The thing being red (usually) does not influence this, so
it is patient. But being cruel, for instance, would require the
person to be cruel to be the agent of that adjective in Tyl-Sjok.
Equally saturated phrases in Tyl-Sjok may be used again to compose
larger phrases by the Rules.
Note that the order of the words never defines which one is the referent. Modification is possible from both sides, left and right. The reason for this is the rigid rules of control (see Section 5.1.1) that determine the order of the words. This is especially meaningful for expressions that translate to genitives as described in Section 3.2.
There are several ways of interpreting the sub-ordinate phrase:
| I | think | red | car | ||||||||||||||||
| I think the car is red | |||||||||||||||||||
The translation is often an auxiliary clause (e.g. a `that' sentence) or a gerund. An infinitive may be appropriate if the subject is missing.
If this kind of embedding is analysed, no things talked about in the auxiliary clause are restricted:
| I | assume | you | book | ||||||||||||||||
| I assume you have the/a book. | |||||||||||||||||||
This is interesting, since `you book' is interpreted in a very original way: as a propositional sentence expressing possessive: `you have a/the book'.
In constrast to the meaning if of the full embedded sentence is referenced, all the other forms of interpretation restrict the referenced thing from the auxiliary clause in the matrix clause, i.e., they are attributive (see below).
| I | repair | red | car | ||||||||||||||||
| I repair a/the red car | |||||||||||||||||||
Translations may be nouns with an attributive adjective (`a red car') or a resticting relative clauses or the like. Participles may also be appropriate.
Here, the original statement `the car is red' is used to restrict `car' in the matrix clause by the attribute `red'. This is in contrast to the full embedding interpretation above, where nothing is restricted.
Note that a referential meaning is not possible here (dt.: `Ich repariere das Auto, das ja rot ist.'), the auxialiary clause always has attributive meaning (dt.: `Ich repariere jenes Auto, das rot ist'). To express referential meaning, two sentences may be used in Tyl-Sjok. E.g.:
| I | repair | car | SEQ | red | it. | ||||||||||||||
| I repair a/the car, which is red. | |||||||||||||||||||
Here, `it' is used explicitly, because a) otherwise the prosody would be too important to mark the end of the first sentence, b) `red' alone is too short. It could only be used alone in an elliptic answer to a specific question.
| I | repair | you | destroy | car | |||||||||||||||
| I repair the car you destroyed. | |||||||||||||||||||
The sub-ordinate clause need not have any subject, in which case a passive construction is most likely used in the translation:
| I | repair | destroy | car | ||||||||||||||||
| I repair a destroyed car. | |||||||||||||||||||
Again, the auxiliary clause has always attributive meaning here, never referential meaning (see above).
There are examples that are ambiguous, because valence restrictions do not exclude enough insensible interpretations:
| I | see | you | destroy | car. | |||||||||||||||
| I see you, who destroyed the car. | |||||||||||||||||||
| I see the car, you destroyed. | |||||||||||||||||||
| I, who sees you, destroy the car. | |||||||||||||||||||
| I see that you destroy the car. | |||||||||||||||||||
Sentences like these have to be disambiguated by using particles. Usually, the reference particle REF is helpful:
| I | see | REF | you | destroy | car. | ||||||||||||||
| I see you, who destroyed the car. | |||||||||||||||||||
| REF | I | see | you | destroy | car. | ||||||||||||||
| I, who sees you, destroy the car. | |||||||||||||||||||
| I | see | you | destroy | REF | car. | ||||||||||||||
| I see the car that you destroyed. | |||||||||||||||||||
| I | see | you | destroy | car | REF. | ||||||||||||||
| I see the car that you destroyed. | |||||||||||||||||||
Of course, the interpretation of this construction can be ambiguous. There are several ways of making clear which part of the sub-ordinate clause is referenced in the matrix clause. The most important particle is REF. Its usage and other particles are described in Chapter 8.3.
These are tendencies only. Context may change them.
However, this is only the preferred strategy of interpretation if there is no context. If the context is clear, other meanings may be intended. In contrast to many other languages, this principle is a very weak preference in Tyl-Sjok.
Again, the preference is weak in Tyl-Sjok.
There are particles for changing these tendencies. See Chapter 8.4.
The items used in the Basic Rules to compose larger items need not be atomic. Composed phrases behave just like words. The last word of a phrase is stressed.
FIXME: continue.