The previous chapter described how Tyl-Sjok syntax works. There is an alternative way of analysis that is equally feasible and directly mappable to the analysis used throughout this document.
In the previous analysis, there were three syntactic categories in a sentence:
As an example, take the sentence `I drink water' jo ljot ak. It has the following structure:
Because of its configurationality, the verb+patient part is a completed phrase, so we could split that part off into a seperated sub-tree, depicting the saturation operation, while the attachment of the agent would be done via a genitive construction:
An alternative is to only have two categories of edges:
And to have only one category of nodes (i.e., one grammar rule):
The resulting tree has the same structure as the one before:
This approach is sometimes more elegant than the previous. Take the following sentence:
|I drive a car.|
With the new approach, there is only one analysis making things clearer: `I.controller car.controlled'.
Throughout this document, we will use the agent-patient analysis, because it usually seems to be easier to think that way. However, we can always analyse an agent-patient structure to the controller-controlled structure:
[A.Agent B.Verb C.Patient] = [A.Controller [B.Controller C.Controlled].Controlled]
Tyl-Sjok's word order is defined by control together with a -relation.
Let a, b be phrases and be the amount of control a has over b (in whatever measure). For the phrase sequence , it must hold that .
If and only if the amounts of control are equal, then the modifying phrase comes first.
As shown in the previous section, Tyl-Sjok word order is totally dominated by control. Many other language find it more important to mark modification. This shows up very clearly in the difference how Tyl-Sjok handles genitives and how other language do it.
It was said before that agent assignement and genitive constructions are the same in Tyl-Sjok. Now the alternative syntax analysis shows that genitive and saturation are also exactly the same thing.
This stresses that Tyl-Sjok does not distinguish between verbs, nouns, adjectives, and whatever other word categories. A possessive genitive `my car' is exactly the same structure as patient-binding in `to eat an apple'. You might think if it as `the process of eatings's apple': the eating is in control. Adding another controller, you get a full sentence: `I eat an apple.'
It can also be seen more clearly now, why attributive and predicative sentences are no different: because which one is modified is not a major criterion for the syntax. The reference particle REF handles the difference.