...VP.¹

Space prevents us from giving a full account of partial VP fronting and the Dutch data. These topics are covered in a longer version of this paper [2].

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

....²

We disallow the use of this rule if both L and R are instantiated as the empty list. Thus, there are no non-branching derivations.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

...VP: ³

Following most HPSG analyses of German, we do not introduce a separate HEAD-SUBJECT schema. Instead, we assume a lexical rule for finite verbs which adds the subject to COMPS.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

... constituent.⁴

Note that Hinrichs and Nakazawa use the valence feature SUBCAT instead of COMPS and assume that the subject is in general included in SUBCAT.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

... left-branching:⁵

In the examples below, VC stands for verb cluster, VP is used for (partial) verb phrases, and S is a verb phrase including a subject.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

...VP.⁶

See also [15], where an analysis along the lines of Nerbonne is proposed, and [2], where we develop a proposal in which a monotonic version of the complement extraction rule is proposed. A similar monotonic account can be found in [11].

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

... ungrammatical.⁷

Baker cannot exclude the derivation of (15b) since in her HEAD-COMPLEMENT schema she assumes that the complements of the head are the sequence union [14] of the selected complements and the (unselected) complements on the mother. This allows the derivation of the phrase das Examen können. Our HEAD-COMPLEMENT schema is more restrictive in that it requires the unselected complements to be a prefix of the complements on the head.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

... cluser.⁸

See [2] for an account of these phenomena compatible with the analysis developed below.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

... it.⁹

Note that this constraint (apart from the feature specification used to implement it) is exactly the same as the Directionality constraint.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

....¹⁰

Kathol adopts a somewhat different definition. Since he assumes a separate valence feature VCOMPS, on which verbal complements are represented and on which there is no inheritance, and since only the governors of verbal complements need to be identified, it is possible to define the governor of a verbal complement as the sign on whose VCOMPS-list this complement appears.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

... constraints:¹¹

Kathol restricts the scope of these constraints to elements that are in the verb cluster field. The reason for this restriction is that, in main clauses, finite verbs may also appear in first or second position. Kathol [6] provides an account of main clauses in which these are derived by means of the same rules used to derive subordinate clauses. The only difference is that in main clauses the finite verb occupies a different topological field. If we would adopt a similar account of verb-first and verb-second, we would have to restrict the scope of the Governor constraints to izone elements as well.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

... complement:¹²

We may in fact assume that this reentrancy is present on auxiliaries in general (and thus there is no need to distinguish between finite and nonfinite verbs in this respect) as finite verbs are never governed.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

...12¹³

Note that we assume that haben uniformly governs a past participle. This implies that there must be a lexical entry for können specified as [VFORM psp]. Other solutions are possible as well, but this is immaterial to the present issue.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.