A clause may bear a grammatical relation to another clause or depend on a nominal expression of another clause. In the latter case, it is an adnominal subordinate clause which modifies the nominal head as its attribute. There are essentially two kinds of these:
- An open adnominal clause is interpreted in such a way that its nominal head is understood to occupy the open position in the subordinate clause. Such a subordinate clause is a relative clause.
- A closed adnominal clause depending on a nominal expression is commonly understood as a hyponym to the latter. It is an adnominal substantive clause.
The following account is restricted to the structural relation between two clauses. This has two aspects:
- hierarchy
- sequential order.
Hierarchy
Grammatical relations
The grammar of parataxis vs. hypotaxis is the grammar of grammatical relations, applied to clauses. The main subdivision of grammatical relations follows from the logical possibilities provided by the concepts of grammatical relation and relationality (cf. Lehmann 1985). It is visualized in the following diagram:
loose | tight |
The primary subdivision in the diagram follows from the criterion of whether a grammatical relation is based on grammatical relationality or not. If so, it is a dependency relation; if not, it is a relation of sociation.
Sociation is a grammatical relation not based on grammatical relationality (i.e. the relata are either not relational or their relationality plays no role in the relation). It is a loose relation between two elements belonging to the same category such that, if they form a syntagma at all, the combination again belongs to the same category. The relation may therefore be symmetric, although not all sociative relations are strictly symmetric. There is, at any rate, no head.
Dependency is a grammatical relation based on the grammatical relationality of one of the relata. The relatum whose relationalality is exploited in the relation is a functor, the other one is its argument. This relation is therefore by definition asymmetric. Furthermore and independently from this distinction, one of the relata is the head of the construction, the other is the dependent. The head of a construction is that member whose category determines the category of the construction.1
The concepts of the lowest level of the above hierarchy will be taken up in the following sections. Here, a few preliminary observations have to be made. First, cross-reference is a relation between two nominal expressions at least one of which is pronominal. It cannot therefore be a relation between two clauses. It can, however, be a relation between a nominal expression in one clause and a nominal expression of another clause which together form a complex sentence. A particular form of such a relation between clauses is called (syntactic) correlation. It is treated in the section on relative constructions.
Second, apposition is a concept typically applied to nominal expressions. In the realm of verbal expressions, it does not have a clear counterpart.
Coordination vs. subordination
A word must be said on the conceptual asymmetry between ‘coordination’ and ‘subordination’, which results in the fact that the former, but not the latter appears in the above diagram.
The Latin terms coordination and subordination are calques on the Greek terms parataxis and hypotaxis. Consequently, the term pairs are originally synonymous. However, they have come to be applied differently:
- The pair parataxis and hypotaxis is reserved for clauses.
- The term subordination also tends to be reserved for clauses, although it sometimes has a wider use approaching ‘dependency’.
- The term coordination applies to all grammatical levels from the sentence down to the morpheme.
These are just terminological conventions. In particular, the third convention is evidently based on the use of the same coordinator (and, und, et, y, e etc.) at all levels. This is indeed typical for Indo-European languages, while in other languages there is less evidence of a unified grammatical relation of coordination across syntactic levels.
From these terminological conventions it follows that
- parataxis is coordination of clauses,
- hypotaxis is subordination of clauses.
The relation between two syntagmatically related sentences is by definition paratactic. In other words, only a clause, not a sentence may be subordinated.
The wide use of the terms ‘coordination’ and ‘subordination’ (which is a bit out of date for ‘subordination’) is justified whenever there are structural and functional similarities and commonalities among coordinative constructions across grammatical levels, and likewise for subordinative constructions across grammatical levels. We will here essentially deal with parataxis and hypotaxis; but other levels cannot be altogether neglected.
Sequential order
The following order possibilities obtain between clauses:
- clause 1 (obligatorily) precedes clause 2
- clause 1 (obligatorily) follows clause 2
- clause 1 is embedded in clause 2 (i.e. it takes the position appropriate for its syntactic function).
Since sequential order is a constraint associated with a certain grammatical relation, the above order statements presuppose the character of clause 1 as coordinated, subordinated or co-subordinated to clause 2.
However, subordinate clauses may have syntactic properties, and in particular properties of syntactic sequencing, that do not derive in a simple way from their syntactic function. In particular, the more sentential a subordinate clause is, the more it tends to a marginal position vis-à-vis the main clause instead of occupying the position generally occupied by simpler constituents with the same syntactic function. Here is an example from German:
- In a German subordinate clause, the (finite or non-finite) verb occupies the final position, as in .a and .a.
- German finite subordinate clauses, e.g. temporal clauses introduced by a conjunction (like nachdem), are highly sentential, thus, rather unlike simple temporal adverbials.
- While a simple adverbial in a subordinate clause, as morgen in .a and .a, obeys the verb-final constraint, i.e. precedes the verb, a clausal adverbial, as in the #b examples, freely ignores it: it is quite common for such subordinate clauses to follow the verb of their matrix clause. In fact, the #b sentences would be clumsier if the adverbial clause was in the place occupied by the adverb in the #a versions.
. | a. | Erna hofft, die Arbeit morgen zu erledigen. |
German | Erna hopes to do the work tomorrow. | |
b. | Erna hofft, die Arbeit zu erledigen, nachdem sie ausgeschlafen hat. | |
Erna hopes to do the work after having slept enough. |
. | a. | Erna hofft, daß Erwin die Arbeit morgen erledigt. |
German | Erna hopes for Irvin to do the work tomorrow. | |
b. | Erna hofft, daß Erwin die Arbeit erledigt, nachdem er ausgeschlafen hat. | |
Erna hopes for Irvin to do the work after he has slept enough. |
Quite in general, in German and in some other languages, a subordinate clause which is a constituent of a subordinate clause suspends the verb-last constraint on the latter by following it. This is probably due to an aversion against center-embedding which is the stronger the “heavier” the material to be embedded.
Sequential order properties of clauses in a complex sentence may also have a diagnostic value in distinguishing coordination from subordination. S. the chapter on syndesis vs. asyndesis for a test.
1 ‘Determines’ does not mean ‘equals’.