Pre-D non-possessor positions in Hungarian

The paper investigates pre-D non-possessor positions in Hungarian. In Hungarian, non-deverbal nominal constructions containing pre-D non-possessor positions are acceptable only if they contain a demonstrative pronoun and also an adjective, and the appearance of a pre-D possessor does not impact the acceptability of the sentence. The paper also gives a brief discussion of similar constructions with pre-D non-possessors in German, mainly to shed light on the Hungarian data. Although German also allows for pre-D non-possessors, it does so under different conditions. A short topicalized element can readily appear in German sentences as a non-possessor dependent, but in this language a possessor can never appear in the same noun phrase. The paper also discusses deverbal nominal constructions with pre-D non-possessor dependents in Hungarian. In these constructions the presence of a possessor argument is indispensable. This is due to the fact that the placement of the non-possessor argument in a position preceding the possessor is legitimized by the fact that the former takes scope over the latter within the internal information structure of the matrix noun phrase. The paper also deals with the syntactic structure of said deverbal nominals.


Introduction
This paper investigates pre-D non-possessor positions in Hungarian, first in the light of findings currently proposed for German. In both languages it is possible for a non-possessor dependent to appear before the definite article. The conditions, however, are different across the two languages.
The phenomenon is all the more interesting considering the general agreement amongst Hungarian mainstream linguists that nothing can appear before the pre-D possessor position, not at least as an element forming one constituent with the noun head (see Bartos (2000), for instance). However, as is shown in (1), this is not the case. In the Hungarian examples we use a constituency test (offered by Alberti, Farkas, & Szabó (2015)) based on the for instance-construction. This contrastive topic construction 2 can be completed with a resumptive pronoun (such as az 'that'), 3 which signals the end of the nominal constituent tested, while the sequence na például 'well for instance' signals the beginning of the tested construction.
In section 1, German constructions containing pre-D non-possessor dependents will be analyzed, followed by the investigation of similar Hungarian constructions. Section 2 deals with Hungarian deverbal nominal constructions, specifically, factors conditioning the occurrence of pre-D non-possessor dependents. In section 3, a syntactic structure is offered for the constructions discussed in section 2. The paper concludes with a short summary (section 4).

Pre-D non-possessors in German and in Hungarian
Pre-D non-possessors do not exclusively appear in Hungarian, they can also be found in German; see Roehrs's (2014) examples in (2): 1 The following abbreviations are used in the glosses: (i) case suffixes: acc(usative), dat(ive), ela(tive), gen(itive), ill(ative), ine(ssive), sub(lative); (ii) gender suffixes: msc (masculine), fem(inine), neu(tral); (iii) other suffixes on nouns: poss (possessedness suffix), poss.1sg/poss.2sg (possessive agreement suffixes); (iv) affixes on verbs: 1sg (agreement suffix), perf (a perfectivizing verbal prefix (preverb)); (v) derivational suffixes: adj(evtivalizer), inf(initive), part(iciple); Throughout the whole paper, the following six-degree scale of grammaticality judgments, given in Broekhuis, Keizer, & Dikken (2012, viii), is used: *: unacceptable; *? : relatively acceptable compared to *; ?? : intermediate or unclear status; ? : marked: not completely unacceptable or disfavored; (?) : slightly marked, but probably acceptable. We also follow Broekhuis, Keizer, & Dikken (2012, xiv) in using introspective judgments of the two authors (both native speakers of Hungarian) as the criterion of what word orders are part of the language associated with what readings (cf. Featherston, 2007, section 5.4). 2 The contrastive topic status of the construction in question can be verified by the fact that it is "pronounced with a particular, contrastive intonation comprised of a brief fall and a long rise" (É. Kiss, 2002, p. 22). 3 Note that in Hungarian the form of the resumptive pronoun az 'that' is identical with that of the definite article (which can also appear as a if the next word in the noun phrase begins with a consonant). Note also that the noun phrase with a pre-D non-possessor can also contain the demonstrative pronoun az 'that' (as in (1b), for instance), the form of which is identical with the resumptive pronoun. The former is the part of the noun phrase under investigation but the latter is not. For the sake of clarity, the borders of the noun phrase construction with a pre-D non-possessor are marked by square brackets.
( The construction is fully acceptable in Hungarian if the pre-D non-possessor is not a simple proper name but a noun phrase with a more complex structure containing a demonstrative pronoun and attributes (5a-b). The reason for this may lie with the higher degree of specificity 4 of the matrix noun phrase in the latter case, which can be regarded as a legitimizing factor to fill in the position in question (cf. Alberti, 1997;Farkas, & Alberti, 2018, pp. 11-13 It is worth investigating if and how the appearance of a possessor impacts the acceptability of these constructions. In German, a possessor cannot appear in such constructions either in a pre-D position (6a), or postnominally (6b). To sum up the requirements, a short topicalized element can appear as a non-possessor dependent in German while a possessor cannot appear in the construction. In Hungarian, however, the construction in question is acceptable only if it contains a demonstrative pronoun and an attributive adjective, while the appearance of a pre-D possessor has no impact on the acceptability of the construction.

Hungarian deverbal nominal constructions
This section discusses deverbal nominal constructions in Hungarian. A pre-D non-possessor dependent can appear in Hungarian in certain deverbal nominal constructions (8a). As is shown in (8b), however, a simple topicalized proper noun cannot appear here (cf. the German examples in (2) We propose that the placement of the non-possessor argument in a position preceding the possessor is legitimized by the former taking scope over the latter within the internal information structure of the matrix noun phrase, and, at least within the prenominal zone, there is no other possibility to express this scope by word order. In what follows, the principal grounds for this assumption will be provided.
In Hungarian, word order always corresponds with the scope hierarchy of the arguments, as is shown by É. Kiss (2002, pp. 113-115). What can be observed in examples (9a-b) is that if we reverse the order of the two quantifiers, the meaning changes parallel to the change in scope hierarchy. É. Kiss (2002, p. 47) refers to this phenomenon as Scope Principle, "according to which operators have scope over the domain they c-command" (É. Kiss 2002, p. 47)) and in Hungarian this scope order corresponds with the surface word order. Karácsonyi (2017) pointed out that this Scope Principle is not only valid for finite verbs but is a general pre-head principle. As is demonstrated in the series of examples in (10), the principle also holds for infinitives and participles. If the order of the quantifiers in these constructions is reversed, the meaning also changes parallel to the change in scope hierarchy, just as in the case of finite verbs. Next, let us consider the question of scope hierarchy in the case of noun phrases. In Hungarian, both the dative case-marked possessor and the unmarked possessor precede the prenominal position capable of hosting non-possessor dependents (11a). In a case like this, the following question arises: How can the reverse scope order be obtained. The first option is to place the possessor dependent after the noun head, as in (11b). In this case, however, the resulting construction is ambiguous: it can be associated both with the meaning where the quantified possessor takes scope over the quantified non-possessor dependent (see meaning1) and with the meaning where the non-possessor takes scope over the possessor (see meaning2).
The ambiguity of the constructions containing post-head operators can also be observed in the case of finite verbs (12), as is shown by É. Kiss (1992, pp. 163-164 As is shown in (13b-b'), the presence of an explicit possessor is indispensable in such noun phrases: the construction containing a pre-D non-possessor dependent is unacceptable if it "only" contains a possessor which is pro-dropped (13b) or if it contains no possessor (13b'). The explicitness of the possessor, while obligatory, is not the exclusive condition. As is shown in (14), only a possessor with a relevant-set based operator function can appear in such constructions, that is, noun phrases with focus or quantifier functions (see also (8b)). As was mentioned in connection with example (8), the same holds for non-possessor dependents. As for the number of pre-D non-possessor dependents, Karácsonyi (2017) points out that what hosts such dependents is not a single pre-D position but a zone which can host more than one non-possessor dependent (15) To sum up the requirements concerning the pre-D non-possessor position in the case of deverbal nominals in Hungarian: a position in question is acceptable only when its filler serves as a quantifier or focus and is followed by a(n overt) possessor also functioning as a quantifier or focus. This is due to the fact that the pre-D non-possessor position can only be used if a nonpossessor dependent takes scope over the possessor (which otherwise precedes the nonpossessor element(s)). Filling in the pre-D non-possessor position is the only way to unambiguously express the [non-possessor > possessor] scope order. 5 This is in total harmony with the characteristic feature of Hungarian according to which word order always reflects scope order.

The syntactic structure of deverbal nominals containing a pre-D non-possessor dependent in Hungarian
This section discusses the structure of Hungarian deverbal nominals containing a pre-D nonpossessor dependent. Figure 1 below represents the structure of (the relevant part of) the noun phrase in (13a), repeated here as (16).
(16) többször is mindkét kollégának a meghívása several_times also both colleague.Dat the invitation.Poss 'inviting both colleagues (together) several times' In order to capture the phenomenon of internal-scope taking within nominal expressions, we propose a general syntactic representation in which the essentially morphology-based Hungarian traditions are integrated with Giusti's (1996)   . The result is a tripartite nominal structure consisting of thematic domains (ΘVΔ, ΘNΔ), agreement domains (ΦVΔ, ΦNΔ) and discourse domains (ΩVΔ, ΩNΔ) according to Grohmann's (2003, p. 211 (37b)) theory of Prolific Domains.
We constructed the layer hierarchy of the structure of the nominal expression as is shown in Figure 1 essentially on the basis of the proposals by Bartos (2000) and É. Kiss (2002, pp. 151-180). However, it required immense notational and some derivational modification to adapt it to Grohmann's (2003, pp. 227-228) two basic generalizations over movement or derivational dependencies within tripartite clause-like cycles: (i) cycle-internal movement always targets the next higher domain (according to this order: ΩΔ←ΦΔ←ΘΔ), and (ii) movement across cycles targets a position within the same type of domain in the next higher cycle (i.e., ΩΔ←ΩΔ, ΦΔ←ΦΔ, ΘΔ←ΘΔ). Thus we accept Grohmann's Anti-Locality Hypothesis (2003, p. 26), according to which "movement must not be too local". Bartos (2000, pp. 678-683), by referring to Baker's (1985) Mirror Principle, proposed layers between D and N essentially on the basis of the assumption that morphology is frozen syntax, that is, "today's morphology is yesterday's syntax" (Givón, 1971, p. 413). The morphology of the noun head in (16) suggests that (i) a PossP-layer builds upon the NP-layer, reflecting the change resulting in a "possessed noun", and (ii) then comes a NumP-layer for numeral information, and (iii) then an Agr N P-layer, given the agreement between possessor and possessee in number and person. In Figure 1, (i) PossP is referred to as nP on the analogy between the (non-thematic) argument generating function of Poss/n and the Agent "generating" function of v, 6 (ii) there is no NumP-layer as the singular number does not require its creation, (iii) there is no agreement layer because this kind of agreement is asymmetrical, or defective, in the sense that in possessive structures with non-pronominal possessors there is no agreement (Bartos, 2000, pp. 678-683).
Our tree building method is completed with the consistent consideration of a principle of D-visibility. 7 The principle declares that either the specifier or the head of the DP must be spelled out. In Figure 1, the default filler of the DP appears, namely the definite article a(z) 'the' in the head of D.
As the given noun phrase is a highly verbal nominal expression, namely a complex-event denoting deverbal nominal construction, an appropriately extended VP-structure based upon 6 The use of nP in Hungarian was proposed by Giuliana Giusti (p.c., 25 May 2016). 7 Its application to Hungarian on the basis of a proposal by Alexiadou (2004, p. 47) is convincingly argued for by Egedi (2015, p. 6), among others. Something similar, namely that economy forces in some languages to have a zero D when (Spec,DP) is occupied by an overt element and to have a filled D when (Spec,DP) is non-overt or not filled at all, is proposed in different works by Giusti on Romanian (e.g., Giusti, 2005, p. 37) as an Economy Principle.
an embedded verb hív 'invite' as its head is assumed to be taken by the nominalizer -Ás in the noun head. The idea that a deverbal nominalizer occupies the position of the N head in the center of the deverbal nominal construction and takes a projection containing a VP is essentially based on Fu, Roeper, & Borer (2001). We consider this embedded verbal construction in the depth of the noun phrase to be the "scope-semantic" source of the internal scope. The reason why it is possible in Hungarian for an internal information structure to be hosted in noun phrases is that the Hungarian noun-phrase structure is (even) more flexible than was hypothesized earlier.
The preverb is base-generated in the complement position of the verb, from which position it moves to the specifier of AspP in the Φ-domain as the element responsible for the given (perfective) aspect. The head of the AspP is occupied by the verb stem. 8 In the particular nominal expression in Figure 1, the embedded verb has one argument (besides the preverb), which is base-generated in the specifier of the embedded VP. From this position, it is raised into the specifier of the (Spec,CentP) as a central argument of the verb. Then, from this Φ-position, it moves to the nominal hemisphere, where it appears as a possessor. The movement from a Φ-position (belonging to the verb) to another Φ-position in the nominal domain is in harmony with Grohmann's (2003, pp. 227-228) generalization over movement across cycles (the moving element targets a position within the same type of domain in the next higher cycle). In the case of nominals with the deverbal nominalizer -ás, prototypically the Theme argument is designated for this role. The possessor is thus first raised into (Spec,nP), which layer is responsible for checking (the mere fact of) possessedness. In harmony with this, the n head is occupied by the possessedness suffix -a. Due to the suffix status of the morpheme, this phrase is right-headed just as the N head occupied by the deverbal nominalizer. After the Theme argument moves to the specifier of the little nP, which belongs to the Φ-zone of the noun, it rises further to a pre-D layer, since it is a NAK possessor. The target position is in the Ω-domain of the nominal hemisphere, namely the specifier of the lower quantifier, the Q Pos P, 9 due to the operator character of the phrase in question. 10 The last phrase to be discussed is the free adjunct többször is lit. 'also several times'. Being a non-central dependent of the verb, this phrase is base-generated in the specifier of a NonCentP as a free adjunct-in total harmony with Grohmann's (2003, p. 313 (O1b)) theory. From this Φ-position it moves to another Φ-position in the nominal domain (named ΦP in Figure 1). 11 As the free adjunct in question is also a quantifier, it rises to an Ω-position in the nominal hemisphere, namely to the specifier of the upper quantifier, that is, the specifier of the Q NonPos P. Indices 'Pos' and 'NonPos' are needed to ensure the order of the operator projections, because, as is shown in (17a), the non-possessor dependent must precede the possessor in the pre-D zone, while the reversed order is unacceptable. 12 The reason for this may lie with the fact that reversed scope order can be expressed by using an attributive construction (17b) and there is no reason to use a marked structure.

Summary
The paper discussed pre-D non-possessor positions in Hungarian. These positions can be found in two types of constructions in the language.
On the one hand, a pre-D non-possessor can appear in non-deverbal nominals in the case of which the construction is acceptable only if it contains a demonstrative pronoun and an attribute, and the appearance of a pre-D possessor has no impact on the acceptability of the construction. The acceptability of this type of construction increases in correlation with the higher degree of specificity of the matrix noun phrase, which can be regarded as a legitimizing factor to fill in the position in question. It was briefly discussed in connection with this type that similar constructions can also be found in German. Nevertheless, the conditions are different in the two languages. In German, it is possible for a short topicalized element to appear as a non-possessor dependent, while a possessor cannot appear in the construction.
On the other hand, pre-D non-possessor dependents can appear in certain deverbal nominal constructions in Hungarian. In this case the presence of a possessor argument is indispensable, and the placement of the non-possessor argument in a position preceding the possessor is legitimized by scope visibility, prevalent in Hungarian: it is only in this way that the non-possessor argument can unambiguously take scope over the possessor within the internal information structure of the matrix noun phrase. The paper also proposed a syntactic structure for complex-event denoting deverbal nominals containing pre-D non-possessor dependents, which is in total harmony with Grohmann's (2003) theory of Anti-Locality.
As for future research, it would be worthwhile to investigate other Germanic languages or different language families to see if they have pre-D non-possessors. In case of a positive answer, it should be explored whether the conditions in those languages are similar to the ones that can be detected in German or in Hungarian, or if they are different from those discussed in this paper.