Against deficiency-based typologies: Manner-alternation parameters in Italian and EnglishGoing Romance 28th, Universidade Nova de Lisboa and

Universidade de LisboaDecember 4-6 2014

Raffaella Folli, University of Ulster & Heidi Harley,University of Arizona r.folli@ulster.ac.uk

hharley@email.arizona.edu

1. IntroductionTalmy (1975, 1985) observes that the verbal inventories of languages can differ systematically in the semantic content they encode.

This observation has given rise to a number of analyses of this variation which broadly fall into two general families.

One approach focuses on the lexical availability of a prepositional element encoding Path semantics (see Higginbotham 2000 and references therein), unavailable in languages of the Romance type;

Another posits the possibility of a special type of structure-building operation, often referred to as ‘manner incorporation’, which allows verbs to express Manner semantics(see Harley 2005, McIntyre 2004, Mateu 2002, 2008 and references therein), similarly unavailable in languages of theRomance type.

Both such approaches ascribe the difference to a deficiency in either the lexicon or the structure-building operations of one typological group.

In this paper, we propose that instead the parameter is of a very standard type, parallel to other well-understood varieties of parametric difference in the verbal domain.

1

We propose that languages of the Romance variety are subject to a head-movement requirement in the lowest domain of the clause, the first-phase syntax of the vP.

English-type languages are not subject to this requirement.

This simple difference accounts for the observed variation intypological behavior without resorting to the invention of a novel type of parameter resting on the notion of grammatical or lexical deficiency.

Crucially, we argue that the two language families are not different in any broad sense with respect to the quality of verbal elasticity.

Rather, they differ in the source of the lexical content of the v head: in Romance languages, it has to be supplied via head-movement, while in English-type languages, it need not be.

In examining the behavior of Italian and English verb frameswe will consider two different domains:

- manner-of-motion constructions- the material/product alternation.

2. Verb-framed vs Satellite-framed

Talmy showed us that languages differ with respect to the ability of a motion verb to express the manner and path of motion: Boats float towards and under bridges in English but not in Italian; to express this in Italian an adjunct is required: (1) a.         The boat floated into the cave.        b.        *La barca galleggió alla grotta.

(no motion possible)                 The boat floated at.the cave.

"The boat floated into the cave"         c. La barca galleggió nella grotta.

(Locative interpretation only,"The boat floated in the cave"         motion

interpretation

2

impossible)         d.         La barca entró nella grotta galleggiando.                   The boat entered into the cave, floating                 "The boat entered the cave, floating." Talmy correlated the availability of the manner-of-motion structure with other patterns:

-resultative adjectives-verb-particle constructions.

Subsequent work has argued that two other structures also correlate with the availability of manner-of-motion:

-ditransitive ('double object') constructions (e.g. Harley 2011)

-productive noun-noun compounding (e.g. Snyder 1995)

(2) Resultatives:  English speakers knock themselves silly, but Italian speakers don't.Particles: English speakers lock themselves out, but Italian

speakers don't.Double objects: English speakers show people things, but

Italian speakers don't.Compounds: English speakers book [hotel rooms], but

Italian speakers don't.

The robustness (and cross-linguistic variability) of this contrast cries out for explanation, and has been the focus of intensive investigation for the past few decades, in formalist, functionalist and psycholinguistic literature (see Beavers, Levin and Wei, 2010 for one recent summary and synthesis).

Our understanding of the contrast has been greatly enhanced by the rapid development of theories of event structure composition and the syntax-semantics interface (Borer 1994, 2002, Ramchand 2008, Folli 2002, Harley 1995, 2005, Folli and Harley 2004, 2007 and many many others).

However, consensus has not yet been reached on the nature of the parametric variation which underlies the contrast between the two languages.

3

In the next section we sketch the two major families of proposals concerning this variation.

3. Lexical parameter vs. Syntactic parameter

Family of proposals A: Verb–framed languages are claimed to lack non-verbal lexical items encoding Path semantics—only verbs can encode Paths. Satellite-framed languages have nonverbal elements which lexicalize Path.

We outline below the details of one such approach developedin Folli (2002).

Verb-framed languages lack nonverbal lexical items encoding Path semantics (a literal implementation of Talmy's vision of 'verb-framed' lexicalization patterns). The lexicons of satellite-framed languages, in contrast, contain items of other categories that lexicalize Path.

For example, the preposition to in English includes a Path component, while the preposition a in Italian is purely locative, and hence is more accurately translated as one of the English locative prepositions at or in.

The impossibility of a motion interpretation in the English sentence in (3b) is then exactly parallel to the absence of a motion interpretation in an analogous Italian sentence with a,in (3c): (3) Italian a = English at/in; Italian lacks an equivalent of

English toa.          The car shuddered to a stop

(motion)b.         The car shuddered at a stop

(location) c. La macchina ha scricchiolato alla fermata dell’autobus

the car         has screeched to.the stop ofthe bus

(location/ no motion)

4

Even with Italian verbs that lexically refer to a motion event, such as navigare ‘sail’ or viaggiare 'travel', it is impossibleto include a Goal argument expressing the endpoint of motion using these simple locative prepositions:

(4) Gianni ha/*è viaggiato a Roma (only locative—hence be auxiliary impossible)Gianni has/is travelled at Rome

The framing idea of the approach is that it is the lack of anequivalent of to (= 'Path') which prevents Italian from expressing a motion meaning with a non-motion verb.

Italian motion verbs like  andare 'go' encode Path themselves, so their Goal can be introduced as a simple location PP.

(5) Gianni è andato nel negozioJohn went in the shop.

In summary, the key idea of this type of approach is that Italian-type languages have a lexical deficiency, which prevents them from expressing a motion event via the combination of a manner verb and a prepositional phrase, sinceItalian prepositions do not encode Path semantics. English propositions in contrast are not so limited.

Family of proposals B: The grammars of verb-framed languageslack a particular kind of structural operation which is used in other languages to create manner+motion verbs. This operation might be (morpho)lexical (Compounding, Zero-derivation (Snyder 1995, 2001), (morpho)syntactic (Manner Incorporation, Renumeration (Harley 2005, McIntyre 2004, Mateu2002, 2008) or semantic (Rule R, Generalized Modification, telic-pair formation1 (Beck and Snyder 2001, Higginbotham 2000). English-like languages are able to execute this operation, while Italian-like languages lack it.

We outline below the details of one such approach developed

1 Since Higginbotham links the lack of telic-pair formation to the absence of an accomplishment preposition, this is also a kind of variant of proposal A; it is then subject to the same problem of lack of generality of proposals of the A type.

5

in Harley (2005).

This approach assumes a small clause structure for change-of-state and change-of-location verbs.(6) a. Mary [vP make [SC John sick]] b. [vP become [SC John sick]]

In manner of motion constructions, the light verb ('cause' or'become') is modifIed by another verb root which describes a manner—a verb whose meaning by itself doesn't have any directed motion component. (7) a. Jill [vP CAUSE (by floating) [SC the canoe up

to the dock ]] b. [vP BECOME (by floating) [SC the bottle up to the dock]] The particular structural operation which permits thismodification is Manner Incorporation (see also Mateu 2002), aprocess which is not fully spelled out (in Harley’s trees thisis represented with thoughts ballosn) , but the central ideais that English can do it, but Italian can't.

In essence here we see that the presence of an(unincorporated) small clause complement to a transitionallight verb allows a language to add lexical 'manner'information to the verb via a 'manner incorporation'operation.

6

A comparison between the two families of approaches revealsstrengths and weaknesses in each.

Problems with family B proposals: A truly structural parameter, according to which a

particular structure-building operation is globallyunavailable in some languages does not comport well withstandard views of parametric variation, in which thesyntactic combinatorial operation Merge is a fundamentaluniversal, and cannot vary across languages.

Semantic parameters are even more problematic: Byassumption, the interface with the interpretive componentworks identically across languages, since the generalcognitive system is presumably identical across speakersof different languages, and the same interpretiveoperations therefore must apply across the board.

Problems with family A proposals: Why couldn't some Paths be lexicalized as verbs and

others as Ps (e.g. why couldn't a language have manner-of-motion constructions for movement to but not movement from)? Finally, why couldn't a language simply borrow a Path-referring preposition?  

Most crucially, such approaches have difficulty in dealing with the combinatorial parametric quality of the effect across constructions: Why should the (un)availability of an independent preposition lexicalizing Path semantics affect the availability of adjectival resultatives, or double object constructions, or verb-particle constructions, or compounds?

Analyses which take seriously the insight of Borer (1984)that parameterization should be a morpho-lexical phenomenonare most compatible with the Family A approaches: the notionthat languages differ in their inventory of quasi-functionallexical items is a very natural one, and syntactic variationdepending on such differences is predicted by such analyses.

In what follows, we will present a Minimalist implementation of the manner-incorporation parameter which brings it into thefold of well-established syntactic parameters—those having to

7

do with verb movement.

In essence, we propose that: Italian verbs require head movement in change-of-state

constructions, including in directed motion constructions, while English-type verbs do not.

Unlike global morphological constraints against verb flexibility, the account correctly predicts that some kinds of verb frame alternations will be permitted in Italian-like languages while others will not.

Finally, we note that morphological filters on well-formedness, defined with respect to particular lexical items, are still useful within the account in deriving the behaviors of different morphological classes of verbswithin English.

4. Change-of-state constructions: flavors of v

Well-established syntactic parameters tend to ascribe different-but-equal status to the relevant lexical features which drive variation: weak vs. strong, early vs. late, [±EPP]etc.

In general, languages are all treated as arriving at equivalent LF representations (given equivalent ingredients), and varying only in the locus of visibility.

Parameters which ascribe a broader inventory of LF representations to one language and a narrower inventory to another are atypical.

In thinking about the nature of the verb-framed/satellite-framed parameter, it is important to establish whether or not the fundamental structures available in both classes of languages are similar (as for other parametric variation casesmentioned above) or different (as in the case of variation in lexical and featural content between languages, in gender, honorific systems, etc.).

Italian (like other verb-framed languages) has a rich inventory of change-of-state and caused change-of-state

8

constructions, including change-of-location constructions, allwith satellite-framed-language analogues.

We believe that this suggests that the fundamental ingredients in change-of-state structures are identical between verb-framed and satellite-framed languages, here specifically between Italian and English.

In addition, it is not the case that Italian verbs are generally morphosyntactically inflexible—there are verb-frame alternations in Italian.

This suggests that the inventory of syntactic operations is also comparable between the two languages: whatever operationsgo into verb-frame flexibility exist in both kinds of languages.

Consider the carve-sculpt alternation in English and Italian.

In English, such verbs allow for three alternating structures with three distinct readings:

(8) a. Maria carved a doll (from the wood).(Product)

b. Maria carved the wood. (Material)c. Maria caved the wood into a doll. (Created

Result)

Creation (Product) reading: Direct object refers to an item which is created by the activity denoted by the verb (6a). Optional PP specifies material from which the object is created.

Resultative (Material) reading: Direct object refers to the material affected by the action denoted by the verb (6b).

In addition, in English, the direct object refers to the affected material and a Goal PP is added specifying the item which is created as the result of the action (6c); we'll call this the Created Result reading.

9

Italian equivalents of creation verbs like carve and sculpt lack the Created Result reading.

Note: Resultative (Material) readings for these verbs are good, but only without the Goal PP specifying the object. (9) is perfectly well formed, and the verb names the result—the effect upon the wood.

(9) a. Maria ha intagliato una bambola (da un pezzo di legno). (Product)

Maria has carved a doll from a piece of wood.

b. Maria ha intagliato un pezzo di legno.(Material)

Maria has carved a piece of wood

c. *Maria ha intagliato un pezzo di legno in una bambola. (Created Result)

*Maria has carved a piece of wood in a doll.

Participles of intagliare can modify either created items or affected items, confirming the alternation:

(10) a. La Matrioska è una bambola intagliata nel legno. (Product)

The Matroshka is a doll carved from wood.b. Cavallo in legno intagliato (Material)Horse in carved wood (Describing an

item for sale)

When the Goal PP is present however, specifying the created result of the change of state (9) the construction is ungrammatical—presumably because at that point, intagliare 'carve', is a manner, not a result, and manner-of-change-of-state verbs are impossible in verb-framed languages.

What is crucial here is that we have a case of verb flexibility: The verb is altering between a manner-of-creationverb (9) and a result verb (9); these alternations are also

10

possible in English. But the third alternation, where the verbis a manner-of-change-of-state, is impossible.

Italian is, therefore, not rigidly prohibited from participating in these kinds of verb frame alternations by anyblanket constraint. The prohibition is specific to manner-of-change-of-state constructions.

This point is underscored by the fact that Italian also allows another famous verb-frame alternation from Levin's collection: There are plenty of spray-load alternations in the language; indeed, in Italian there's an extra frame with theseverbs which is unavailable in English (Damonte 2005):

These events all involve a Theme moving into or out of a Location (hence this is usually termed the 'locative' alternation).

(11) a. Gianni ha caricato la paglia sul camionGianni has loaded the hay on.the truck

b. Gianni ha caricato il camion con la paglia.Gianni has loaded the truck with the hay.

c. Gianni ha caricato il camion di paglia.Gianni has loaded the truck of hay.

Pour-variant: Direct object is the Theme (11), PP specifies locational Goal;

Fill-variant: Direct object is the Goal, PP specifies material Theme (11), the moving material.

Also: of-variant: Direct object is the Goal, while the Themeis contained in of-PP (11).

In sum: Change of state/location structures, complete with Path

semantics, are clearly well-formed in Italian.

11

Verb frame alternations, equally, are well-formed, as shown by the spray/load alternations2—so there's no blanket prohibition on verb frame alternations.

Rather they behave in a productive, predictable way, as expected if they are the result of syntactic operations.

What is the problem with the Created Result reading of carve/sculpt, then?

Change-of-state structures which involve encoding the resultin the verb are good. Structures in which a creation occurs in the manner encoded in the verb are good In particular, alternations in which the verb in Italian encodes a couple of different Results (as in the locative alternation), or a Manner of Creation (as in the material/product alternation), are fine.

What is not good are change-of-state structures where the result is not encoded in the verb.

Moral: It really is Manner-of-Change-of-State alternations that are impossible in verb-framed languages. It's not a problem with verb flexibility in general.

When a change-of-state event is described, the resulting state must be encoded in the verb—this is what it means to be verb-framed.

Note: what sets this view apart from the two previous approaches is the fact that verb-framed languages have an extra requirement, not a prohibition on the construction.

5. Head-movement parameters

In most of the accounts fitting the consensus pattern described above, morphological ill-formedness is appealed to

2 Indeed, they behave in a productive, predictable way, as expected if they are theresult of first-phase syntactic operations. See Alexiadou and Anagnostopoulou 2011 for documentation and discussion of the productivity of this same, similarly productive class of locative alternations in Greek, another verb-framed language.

12

to restrict the application of the operation which creates manner-of-motion verbs in verb-framed languages.

However, the fact that the constraint is general to all kinds of nonverbal potential Result categories including Adjectives makes it difficult to think that the restriction isspecific to a category like P, and suggests the need for a more general account.

We take our model for a general account of the relevant typefrom a more well-understood domain of verbal syntax: The V-to-T parameter that is set differently in French (and other Romance languages) than in English, or the T-to-C parameter that is set differently in French than e.g. German.

The intuition is that in verb-framed languages, there is a "Result-to-v" parameter which is set to 'on': Feature checkingbetween (change-of-state) v and Result always requires overt head movement in these languages, while in satellite-framed languages, the same parameter is set to 'off'—that is, in satellite-framed languages, checking between a change-of-statev° head and the Result in its complement can occur without triggering movement. In satellite-framed languages, v-Result feature checking can occur with Result in situ, just as feature checking between T and V (in the traditional account of the verb-raising parameter) occurs with V in situ in English.

Crucially, we need to be able to distinguish between Change of State v, which selects for a Result, and Creation/Activity v, which does not—we need to ascribe different head-movement properties to different flavors of v much in the same way that auxiliary verbs in English have different head-movement properties than main verbs.

13

5.1 Technical implementation I: V to T movement

We need a formal feature-checking account of the V to T parameter in place, on which to model the account of the Res-to-v parameter.

To implement head-movement in terms of feature-checking requirements, we will adopt (for argument's sake) Matushansky (2006)'s theory of head-movement in conjunction with Adger (2003)'s theory of selection by category-feature checking.

Moved heads re-merge at the root of the derivation, and immediately undergo an m-merger ('bundling') operation to adjoin to the head. ('M-merger' = Noyer and Embick's 'lowering' operation).3

… X'

Y° X'

X° YP

tY° ….

We assume a simplified version of Adger's feature system in which all complementation involves categorial feature checking. Movement is triggered by EPP features associated with particular unvalued features. We annotate this EPP feature as *, following Adger 2003.

Head-movement from the complement is thus triggered by an EPP feature linked to the uF which triggers complementation onthe selecting head.

So, for example, in English, T° bears uv, rather than uv*: that is T° which selects v° lacks the additional movement-triggering EPP feature. Lexical verbs thus do not raise higher

3 Manner modifiers are attached to v° by the same series of operations, modulo being adjoined to the root by External merge, rather than Internal merge. They subsequently undergo m-merger in exactly the same way, however.

Y° undergoes lowering /m-merger with X°

14

than v° in English. T°'s uv feature is checked against v° in situ. In French, T° bears uv*, triggering v-to-T°.

On the other hand English auxiliary verbs do raise to T°. Auxiliaries, then, are a distinct subcategory (or 'flavor') ofv°; call it vAux; the T° which selects them bears uv*

(12) Ingredients for the analysis:a) uF triggering complementationb) uF* triggering complementation + head movementc) Different 'flavors' of the same category can have

different * properties

15

5.2 Technical implementation II: Res to v° movement

Assumption: change-of-state v° heads like vCAUSE and vBECOME share a selectional feature which ensures they take a result-denoting (Small Clause) complement, which we will term Res(ult)P here, following Ramchand (2008).

vP

DP v'

John vCAUSE ResP =SC

DP Res'

the door Res°open

Proposal: in verb-framed languages, the complement-selectingfeature on change-of-state v° is uRes*, requiring head-movement from Res° to v°.

The requirement that Res move to v will ensure that a verb in a change-of-state construction must 'lexicalize' the resultof the change-of-state.

Remember: with respect to other verb flexibility metrics, verb-framed languages are just as free to reinterpret/coerce verbal meanings as any other language.

The paradigm illustrating this effect is given again below:

(13) a. Maria ha intagliato una bambolaCreation

Maria has carved a doll from a piece of wood. intagli-√ = Manner

b. Maria ha intagliato un pezzo di legno Change of State

16

Maria has carved a piece of wood intagli-√ = Result

c. *Maria ha intagliato un pezzo di legno in una bambola.Change of State

Maria has carved a piece of wood in a doll. * intagli-√ = Manner

Crucially, when the construction entails a result, the verb must denote the result state- the result cannot be 'stranded' independently of the verb, as in (13).

The derivation of intagliare on the result-naming reading thenis the following:

(14) v'

v°uRes* ResP

Res° v° uRes* DP Res' (√intagli-) (a-) (un

pezzo di …tRes°....legno)

No Res° head can be left stranded in verb-framed languages. This entails that there can be no adjectival resultatives, no verb-particle constructions, and no manner-of-motion constructions—essentially, no structures like (13), illustrated below, where Res does not incorporate into v°.

(15) * v' Ill-formed due to unchecked uRes* feature.

v°uRes* ResP

√ v°uRes*DP Res' ( intagli-√ ) (a-) (un carve pezzo di Res° DP

legno) (in) (una bambola)a piece of into a dollwood

17

Simply remove the * from uRes, however, and you have the appropriate structure for the English equivalent, carve a piece of wood into a doll. The uRes complementation feature on English CoS v° is able to check against Res° in situ, no movement needing to be triggered.

Extending the proposal: the spray/load verbs like caricare, as proposed by Rappoport & Levin, Tenny, and others, are flexiblein that they can be understood as naming either one of two different result states:

-the state of a container that has been loaded/sprayed or

-the state of a (specific amount) of a Theme that has been loaded/sprayed.

Both involve interpretation of the verb as a result, so bothare fine in Italian.

5.3 Technical implementation III: Manner Incorporation

Manner modification is the e-Merge equivalent of Matushansky's i-Merge head-movement structure: If instead of selecting an element from within the complement for re-Merge (i.e. instead of i-Merge), a verbal element from the Numeration is selected, that element can be e-Merged at the root of v' and subsequently lowered/m-merged to adjoin to v° This gives a manner modification structure. Given Matushansky's technology, it would be surprising if such a manner modification operation did not exist; e-Merge and i-Merge are equally predicted to be possible, from a Minimalist viewpoint.

The interpretation of the adjoined √ element as a manner adverbial also follows naturally; the sister-to-v' position isthe classic locus for adverbial modification in general.

The fact that the root subsequently undergoes m-merger with the v° head does not affect its LF interpretation, which proceeds as for any adverbial; carve √ is interpreted as an eventmodifier, contributing its content exactly as expected within a neo-Davidsonian event semantics.

18

Note that this operation is just as available in Italian as in English. Nothing about the syntax or morphology of Italian prevents e-Merge of a verb root at the v' level. The explanation for the unavailability of manner-of-result clausesrelies on the need for EPP-feature checking in Italian v, not on banning e-Merge as an operation.

5.4 Manner/Result complementarity

Why can’t an e-Merged Manner element from the numeration undergo m-merger/lowering with a change-of-state v° that has already checked its uRes* feature, by moving and m-merging with a Res° element? That is, why Manner/Result complementarity—Why can't both ani-Merged and an e-Merged Root both adjoin to the v°? This would produce a kind of V-V compound, where one V element encoded Manner and the other Result, something like Gianni ha scolpi-intagliato il pezzo di legno, 'He sculpt-carved the piece of wood.'

The complementary must derive from a restriction imposed by the morphology/syntax interface: the syntactic operations to generate such a structure are available. Instead, a well-formedness constraint is violated when a second √ element is m-merged with a v° head that has already had a √ element m-merged with it.

Several approaches are possible here: Perhaps the m-merge cycle can be calculated only once

for a given projection; Perhaps there is a morphological constraint in non-

incorporating languages against having two √s in a singlev°;

Instead we adapt a version of Embick's categorization approach which is already fleshed out in the literature.

Embick 's categorization restriction: Verb roots in verb-framed languages are ill-formed without

19

incorporation/conflation with v°, because they would remain uncategorized. 4

If a Manner root adjoins to v, stranding an uncategorized Result (=SC predicate) root downstairs, the result is ill-formedness.

We will adopt Embick's key insight that a single v° head cancategorize only one element.

The idea is that little x° categories can determine/check the category of the element they m-merge with only once.

Let us assume that the m-merger/lowering operation is drivenby this category-determining relation, say a morphological uCategory feature on little x heads like v. This feature can be checked by any non-v category, including √. This will derive Manner/Result complementarity: a Manner element adjoined to a v° which has already checked its uCategory feature against a Result cannot m-merge with it.5

This approach also explains the optional availability of Res-to-v movement in English-type languages, that is the availability side by side of John opened the door and John pushed the door open.Although uRes does not force i-Merge of the Res head, the possibility of i-Merge is still available, and can be exploited to satisfy the uCategory feature of v°, that is e-Merge or i-Merge are freely available, as we would expect given Minimalist assumptions.

4 Notice that Embick exploits his categorization constraint in attempting to account for the Talmian parameter. His notion is that the problem with manner-of-motion structure in verb-framed languages is that their result-denoting verb roots would be ill-formed if stranded, since without incorporation/conflation with v° they would remain uncategorized. For Embick, then, the Talmian pattern is due to root stranding, not (as in the current proposal) any requirement of the v° head. Unlike the current proposal, however, Embick's approach fails to explain why adjectives, PPs, or particles cannot occur in Result position downstairs in verb-framed languages. They're categorized, certainly, so the root-categorization constraint would not be violated by such structures. In essence, Embick's proposal is another variety of deficiency account, albeit a morphological one: Italian lacks√ elements which can occur on their own in the Result position.5 It may be able to cliticize to it, if it's an independently well-formed item; cf complex phrasal manners-of-motion in head-final languages cross-linguistically.

20

This is what it means to be 'satellite-framed': able, but notrequired, to express Result independently from v°.

5.5 Other flavors of v

On the creation reading (14a), intagliare is a Manner element,modifying vDO. The vDO head, however, does not take a ResP complement; rather, it selects for a DP.

Its uD feature allows checking in situ; hence Manner elements may adjoin to v' and m-merge with vDO, as in intagliare una bambola, 'carve a doll'

(16) v'

vDO°uD DP

√ vDO°uD (una( intagli-√ ) (a-) bambola)

Confirmation of the view that these properties are dependenton the particular flavor of v comes from the interaction of animacy with the creation vs result interpretations of verbs like intagliare or modellare (see Folli and Harley (2005).

Applying this diagnostic to our Italian alternating verbs, we discover that the creation interpretation is infelicitous with inanimate subject (17), while felicitous with the change-of-state reading (17).

(17) a. #il vento ha modellato una farfalla (sulla spiaggia)

The wind has modelled a butterfly (on.the beach)

b. il vento ha modellato la spiaggiaThe wind has modelled the beach.

We take this as confirmation that the two readings of these verbs involve distinct flavors of v: The creation reading (which involves a true Incremental Theme in the sense of Dowty(1991)) requires vDO, and hence is infelicitous with an

21

inanimate subject, which cannot serve as a teleologically capable Agent (Folli and Harley 2008).

5.6 Manner leakage in verb-framed languages? Decomposing Res As it has frequently been noted (Folli 2001 among others), certain manner of motion verbs can co-occur with a goal-denoting PP complement in Italian:

(18) a.Gianni è corso al negozioGianni is ran to.the storeb. l’aereo è volato sotto il pontethe plane is flown under the bridge

How can the head-movement parameter we propose be adapted to allow for such lexeme-specific 'leakage'?

The feature-based checking account we propose can be adaptedto allow specific verbs to permit stranding of goal PPs, in the same way that English allows the class of auxiliary verbs to head-move to T°.

Folli (2001) showed that Italian the class of verbs which permit the expression of a Goal PP have Path semantics contained within them, so in that sense, they are not exceptions.The question is why these verbs allow a prepositional expression of the location serving as the endpoint of the path.

Following Svenonius, Ramchand & others: ResP is probably made up minimally of a scalar projection (PathP, in the spatial domain) and an optional terminus point specification (LocP, in the spatial domain).6

v then has a uPath* feature in verb-framed languages.

Pure Result verbs (intagliare 'carve' on Result reading, caricare 'load' on either reading, all other normal change-of-state verbs) are the result of head-movement from Loc into 6 Again, as others have noted, the necessary presence of the scalar projection within the ResP complex is consistent with Bobaljik's genearlization concerning thescalar structure that appears to be necessary in deadjectival verbs to account for their comparative forms: to better, not *to gooden; to worsen, not *to bad(den).

22

Path and thence to v. By hypothesis those verbs form a class; Path selecting a LocP of that class will have a uLoc* feature,mandatorily triggering head-movement.

A second type of Path head (realized by verbs meaning go, run, perhaps jump etc.), however, will allow Loc to be stranded; this class of Path heads (like the T selecting for vAuxP instead of vP) have a uLoc feature, not a uLoc* feature. For these verbs and these verbs only, the LocP may be stranded. The u Path* feature on CoS v, however, may never be left unchecked.

These apparent 'manner' verbs are not manner verbs at all; their lexical content is related to the Path° component of thestructure (hence they are not then true exceptions to the generalization about verb-framed languages). We can see this in the bleaching of the verbal entailments related to manner in sentences with and without the Goal PP:

(19) a. Gianni ha corso. # Ha preso la macchinaGianni has ran. # He took the car.b. Gianni è corso al supermercato. Ha preso la macchinaGianni is ran to the supermarket. He took the car.

6. Satellite-framed languages: Optionality vs Requirement

In satellite-framed languages, CoS v's uRes feature may be checked without movement.

This allows for wholesale Manner incorporation in change-of-state structures: external merge of a Manner element, followedby m-merger.

(20) Babe Ruth homered his way into the hearts of America.etc.

But satellite-framed languages also have Res movement— sometimes mandatorily, as with verbs like compose, which rejectresultative and particle constructions, sometimes seemingly optionally as with open.

23

(21) a. John opened the window up. Res-stranding, √open=Mannerb. John opened the window. Res-moving,

√open=Result

The difference between the two structures can be diagnosed by re-affixation, which requires movement from the lower domain (see Harley 2011).

(22) a. John reopened the window. (Res-moving,re-affixation possible)

b. *John reopened the window up.(Manner-adjunction, no re-affixation)

How can we allow both options within a single language?

Optional bundling of * property with change-of-state v's selectional features.

Something similar is needed to differentiate different structures with vDO: dance vs do a dance, etc. In fact we see things like this in v-framed languages all the time, too: camminare 'walk' vs. fare una camminata 'do a walking', etc. This is the difference between a 'heavy verb' and a 'light verb' structure with the same interpretation (see Folli, Harley and Karimi 2005).

But within English the optionality can be overridden by specific selectional/morphological needs:

Derivations of verbs with bound roots in them will fail in the morphological component if they happened to be selected by a v bundled with uRes (*compose up an idea vs write up an idea);

Similarly, derivations of verbs with resultative or particle complements will fail in the morphological component if they happened to be selected by a v bundled with uRes*.

Crucially for us, the morphological component acts as a filter on derivations, not as a driver of them .

24

We discuss this briefly in the next section, in addressing the failure of (certain) Latinate verbs to occur in the ditransitive frame.

7. Extending the proposal: dative shift and verb-particle constructions.

We have shown that Talmy's parameter falls within a well-understood class of head-movement parameter, for manner of motion, resultative and particle constructions.

In this section, we show that the approach can explain why dative shift is also part of this typology, and thus has the properties of a combinatorial parameter in the sense of Higginbotham. We also examine why dative shift is subject to exceptions in English-style languages, showing that interface filters also have a role to play in a complete account.

English ditransitive verbs can often productively ‘dative-shift’.

(23) a. Mary showed the procedure to the classa'. Mary showed the class the procedure

b. Mary gave a book to her sonb.' Mary gave her son a book

In Italian-type languages, dative-shift is impossible.

On the present account, this must be because dative-shift involves stranding the Res° head, which in Italian, but not English, would result in an unchecked uRes* feature.

The approach to dative shift proposed in Harley (2007, 2009)predicts exactly this.

The to-dative involves movement from within the ResP/SC, while the double object construction strands a Result-denotingPHAVE element, with e-Merge and m-merger of the verbal element, which describes the manner of change-of-possession.

(24) vP

25

DP v'

Mary vCAUSE ResP

DP Res'

her Res° PPhat lent

to Bill

(25) vP

DP v'

Mary v° ResP

V° v° DP Res' lent

Bill Res° DPPHAVE her

hat

In sum, double-object constructions are a species of change-of-location structure, requiring a null preposition PHAVE in theRes° position. The verb root in a double-object construction,then, occurs in v° via e-Merger, not by raising from the Res° head. Thus, double object constructions can exist in English-type languages, but not Italian-type languages.7

But what about the inverse problem? A robust class of English ditransitive verbs cannot dative shift. Diachronicallyspeaking, they’re usually verbs of Latinate origin:

(26) a. Mary demonstrated the procedure to the classa'. *Mary demonstrated the class the procedure

b. Mary presented the award to the author.b'. *Mary presented the author the award.

7 Harley 2002 argues that in fact double-object structures, albeit concealed ones, do occur in Italian-type languages. That is mutually incompatible with the present account, which assumes the more widely-held view that Italian-type languages lack double-object structures entirely, hidden or not.

26

Similarly, verb-particle and resultative constructions in English don't occur with Latinate verbs (see Harley 2007):

(27) Verb-particle constructions write it up *compose it

up/*arrange it upeat it up *consume it upfinish it up *complete it up

(28) Resultatives cut it apart *divide it apartfill it full *inflate it fullwork yourself ragged *decide yourself

ragged

How to prevent this type of verb from stranding Res°, while allowing others to, optionally?

Latinate verbs have their Res° head built in—they are well-formed only if the Res° head—the unstressed prefixal element preceding the root—is present and prefixed.

ex-hibit, com-pose, in-flate—the prefix is a Res° head, and the whole is only interpretable if it appears in its appropriate place. Removing the prefix and substituting PHAVE in Res° produces ill-formedness (see Harley 2009), hence no dative-shift.

ALSO, derivations of these verbs will fail at the morphological component if they happen to be selected by a v bundled with uRes; there is no interpretation listed for just √hibit, pose, flate√ √ as main verbs, with their prefix stranded: *hibit the paintings ex, *pose the story com, *flate the balloon in…

Similarly, derivations of verbs with resultative or particlecomplements will fail at the morphological component if they happened to be selected by a v bundled with uRes*: *upopen the door, *outthrow the garbage, etc.

27

8. Conclusions

o The lack of flexibility in CoS constructions in verb-framed languages is simply a reflex of a head-movement requirement in those languages, modeled in a Minimalist framework as a feature-checking requirement on CoS v° flavors—these languages have uRes*, not uRes.

o Satellite-framed languages (at least English) have both uRes* and uRes available. Interface requirements on interpreting individual lexical items determine their well-formedness in a given structure or not.

Selected References

Adger, D. 2003 Core syntax, OUP.

Alexiadou, A. and Anagnostopoulou, E. 2011. "Decomposing the Greek Verb." Talk presented at 'Approaches to the lexicon: Roots III', June 16, 2011, The Hebrew University of Jerusalem,Jerusalem, Israel.

Beavers, John, Beth Levin, and Tham, Shiao Wei. (2010). "The typology of motion expressions revisited". Journal of Linguistics 46:331-377

Beck, S. and Snyder, W. (2001a) "Complex predicates and goal PP's: Evidence for a semantic parameter." In A. H.-J. Do, L. Dominguez, and A. Johansen (eds.) Proceedings of the 25th Boston University Conference on Language Development. Somerville, MA: Cascadilla Press.

Borer, H. (1984). Parametric Syntax: Case Studies in Semitic and Romance Languages. Dordrecht: Foris Publications.

Borer, H. (1994) The projection of arguments. University of Massachusetts Occasional Papers in Linguistics, 17, E. Benedicto & J. Runner, eds. Amerst: GLSA, University of Massachusetts.

Borer, H. (2005b). The Normal Course of Events. Structuring Sense, Volume II. Oxford: Oxford University Press.

28

Folli, Raffaella. 2002. Constructing Telicity in English and Italian. Ph.D. dissertation, University of Oxford.

Folli, Raffaella. and Heidi Harley. 2007. Causation, obligation and argument structure: On the nature of little v. Linguistic Inquiry 38.2, 197-238.

Folli, Raffaella, Heidi Harley and Simin Karimi. 2005. Determinants of event structure in Persian complex predicates.Lingua 115.10, 1365-1401.

Folli, Raffaella. and Harley, H. (2004) "Flavors of v: Consuming results in Italian and

English," in Roumyana Slabakova and Paula Kempchinsky, eds., Aspectual Inquiries, 95

120. Dordrecht: Kluwer

Harley, Heidi. 1995. Subjects, Events and Licensing. PhD dissertation, Cambridge, MA: MIT.

Harley, Heidi. 2002. Possession and the double object construction . In The linguistic variation yearbook, vol. 2, Pierre Pica and Johan Rooryck (eds.), 29–68. Amsterdam: John Benjamins.

Harley, Heidi. 2005. How do verbs get their names? Denominal verbs, Manner Incorporation and the ontology of verb roots in English. In Nomi Erteschik-Shir and Tova Rapoport, (eds.), The Syntax of Aspect, 42-64. Oxford: Oxford University Press.

Harley H. 2007. The bipartite structure of verbs cross-linguistically, or Why Mary can't "exhibit John her paintings", Ms University of Arizona (lingbuzz).

Harley, H. 2009. "A morphosyntactic account of the 'Latinate' ban on dative shift in English" Handout from talk given at UCSC, May 15, 2009.

Higginbotham 2000 "Accomplishments." Ms. University of Oxford.

Mateu, J. 2002. Argument Structure. Relational Construal at the Syntax- Semantics

Interface. PhD dissertation, UAB.

29

Mateu J. and G. Rigau, 2008 Romance Paths as Cognate Complements,Dins P. Masullo Erin, O'Rourke & Chia-Hui Huang (eds.). Romance Linguistics: Structures, Interfaces, and Microparametric Variation. 221-235. Amsterdam / Philadelphia: John Benjamins.

Matushansky, O. 2006. Head movement in linguistic theory. Linguistic Inquiry 37.1, 69-109.

McIntyre A. 2004, Event paths, conflation, argument structure,and VP shells, Linguistics 42–3 (2004), 523–571.

Ramchand G. 2008. Verb meaning and the lexicon. Cambridge: Cambridge University Press.

Snyder W., 1995 Language acquisition and language variation:The role of morphology. Cambridge, MA: MIT dissertation.

Snyder, W. (2001) "On the nature of syntactic variation: Evidence from complex

predicates and complex word-formation." Language 77:324-342

Talmy, Leonard. 1985 ‘Force Dynamics in language and thought’ In Papers from the

Regional Meetings, Chicago Linguistic Society, 21, 293-337.

30