Feature 16. Partial VP fronting

found in question(s): 41a, 41b, 42a, 42b, 43a, 43b

Definition and illustration

Some languages allow the preposing of a constituent containing a usually non-finite verb and some of its arguments. In English, this involves the verb and its object:

(1)

Two important properties characterizing the VO- vs. OV-dichotomy are thought to impact this fronting: Since the constituent is usually identified with the VP (more precisely, the vP), the presence/absence of the EPP affects the possibility of the fronting to also include the subject. While in English, inclusion of the subject is impossible because it has moved out of VP prior to VP-fronting, in languages without an EPP-requirement (or an optional EPP), the subject may remain within VP and hence can be affected by VP-fronting. German has been argued to allow the inclusion of subjects (Haider 2010; the facts are somewhat contested, though; see also feature 17 on VP-internal subjects):

(2)

There are two theoretical accounts for an OV-VO difference in the availability of partial VP fronting in ditransitives. Early on (Thiersch 1985), partial VP-fronting was linked to the availability of A-scrambling since this way, the VP-constituent could be partially or fully emptied before VP-fronting ("remnant movement", see Müller 1998 for an overview). Since OV-languages are suppossed to exhibit A-scrambling, OV languages are predicted to feature partial VP-fronting as well.

In another account agnostic regarding remnant movement (see Fanselow 2002 for a critique), partial VP-movement is directly constrained by a structural difference between ditransitive VPs in OV and VO, with VO languages obligatorily requiring verb movement to form a VP-shell quite generally (Haider 2010) or only in double object constructions / ditransitives with neutral alignment (Janke & Neeleman 2012). This results in a structure such as [V [IO [< V > DO]]] where the verb does not form a subconstituent with any of its arguments (but only with all of them), while OV languages trivially feature proper subconstituents containing V and its arguments [S [IO [DO V]]]. Under such a perspective, abstracting away from scrambling, only proper VP subconstituents can be fronted (Landau 2007), which has different consequences in the two types of languages, i.e, V+DO, V+DO+IO in OV-languages or V+IO+DO in VO-languages (but not only V+DO or V+IO). If in addition a language has A-scrambling, it should be possible to front partial VPs of all kinds (regardless of whether A-scrambling is achieved via movement or base-generation).

Furthermore, the availability of VP fronting in general may be constrained by V-to-T movement (Holmberg & Platzack 1995) in that movement of the verb out of VP will bleed VP-fronting. One may expect TP-fronting in such languages instead, but that seems to be much more constrained across languages, e.g., English TP-fronting is ungrammatical (a fact often linked to anti-locality as it involves movement of the complement of a phase head):

(1)

The lack of VP fronting in a language can, then, be explained as the result of obligatory V-to-T movement even for non-finite verbs, which would then necessitate TP fronting. This explains why French does not generally allow for VP fronting (Authier 2011 combined with insights from Hacquard 2010) and why finite verbs can not be fronted in general (i.e., because TP-fronting is generally very restricted). As a result, lack of any kind of VP fronting should coincide with the availability of neutral adverbial intervention (features 2,3,4) due to verb movement.

Finally, VP-fronting can serve as a diagnostic of underlying word order. Should a language allow for both OV and VO orders with non-finite verbs but only allows for one of these orders under VP fronting, that order can be argued to be the underlying order. This test was applied in Mócheno (Germanic) which allows for both OV and VO orders, but only allows for VO order under VP-fronting (Cognola 2013).

In general, the predictions only apply for clearly verbal VPs, not nominalisations (fronted VPs are nominalized in some languages, thereby often leading to word orders characteristic of the noun phrase).

Questions 41-43 test different aspects of these predictions. The information-structural component is employed in order to help elicit (partial) VP frontings since language consultants could have problems recognizing the availability of such structures without a proper context. Building on prior experience, givenness and contrastive focalization would license VP fronting.

Correlations

Question 41 investigates whether the verb can be fronted by itself, without its objects. This question can falsify theories of partial VP fronting. If partial VP-fronting is only linked to A-scrambling, any language with A-scrambling should exhibit V fronting. If instead A-scrambling is limited to OV-languages, we expect partial VP-fronting only in such languages.

• Prediction 1: A-scrambling → partial VP-fronting
• Prediction 1a: V-O → ¬A-scrambling → *[VP fronted verb] ... OBJ1 OBJ2
• Prediction 1b: O-V → A-scrambling → [VP fronted verb] ... OBJ1 OBJ2
• Prediction 2: V-to-T → *[VP fronted verb] ... OBJ1 OBJ2

Question 42 investigates whether partial VP-fronting is possible, i.e., fronting of the verb together with a subset of its objects, concretely the direct object. As in 41, this question can differentiate between theories of partial VP fronting. If a language with A-scrambling disallows partial VP fronting, the remnant-movement account is falsified. If a language with VP-shell formation allows partial VP fronting (e.g., V+DO), the subconstituent approach is falsified.

• Prediction 1: A-scrambling → partial VP-fronting
• Prediction 1a: V-O → ¬A-scrambling → *[VP V DO] .... IO
• Prediction 1b: O-V → A-scrambling → [VP DO V ] .... IO
• Prediction 2: VP-shell → *[VP V DO] .... IO
• Prediction 2a: V-O → VP-shell → *[VP V DO ] .... IO
• Prediction 2b: O-V → ¬VP-shell → [VP DO V ] .... IO

Question 43 investigates the availability of VP-fronting in general. Absence of VP fronting could be a sign of V-to-T movement. Under the assumption that there is no rightward head-movement (Abels & Neeleman 2012) and hence never [TP [ [VP O ] ... ] V ], all OV languages should allow for VP fronting.

Question 42-43 also investigate whether the word order in a fronted VP is rigid even if it is flexible without fronting. If only one order is available in a fronted VP, that order is likely the underlying order.

• Prediction 1: (preservation of basic word order)
• Prediction 1a: V-O → [VP V-O] …
• Prediction 1b: O-V → [VP O-V] …

See also

This feature is closely connected to several others that explore the consequences of the presence/absence of A-scrambling, i.e., 10, 12, 14.

References