Definiteness Effect in the PP

The definiteness effect (DE), observed in the English there is construction and in different clause types across languages, has been a topic of debate ever since it was first reported by Milsark (1974, 1977). Various syntactic, semantic, and information-structural factors have been shown to interact in its emergence, but it is still an open question which of these are primary factors and which are collateral ones, and whether the DEs pointed out in different clause types (in existential, unaccusative, and transitive clauses) are independent phenomena or have the same explanation. This article brings new evidence to bear on these issues; it shows that the DE is also attested in abessive PPs (PPs expressing the lack of the complement of the adposition).¹

In such PPs, the syntactic conditions associated with the DE are absent, which argues against the syntax-based accounts of the phenomenon. On the other hand, the semantic trigger identified in DE constructions by Szabolcsi (1986a,b, 1992), an EXIST meaning component of the predicate, can be pointed out in the fine structure of these PPs. The analysis to be proposed supports Szabolcsi’s meaning-based DE theory, showing that it gives a unified explanation of DE phenomena across constructions and across languages. The analysis also has implications for the structure of the PP, supporting proposals that assign a small clause complement to a class of Ps (Beukema and Hoekstra 1984, Den Dikken and Dékány 2018).

The article is organized as follows. Section 2 provides background, discussing first various constructions displaying a definiteness restriction and then a cross-categorial account of the DE, Szabolcsi’s (1986a,b, 1992) DE theory. Section 3 analyzes the DE attested in PPs. Section 3.1 introduces abessive adpositions, section 3.2 analyzes híján/for lack of PPs, section 3.3 examines the structure of nélkül/without PPs, and section 3.4 studies the source of the DE’s optionality in these PPs. Section 4 is a conclusion, summarizing the implications of the analyses for the theory of the DE and the structure of PPs.

DE theories originally aimed to account for the “definiteness” (more precisely, presuppositionality/specificity) restriction on the postverbal subject of existential sentences such as the English there is/there are construction (Milsark 1974, 1977), exemplified in (2).

Later, the DE was also shown to arise in unaccusative, passive, and transitive clauses with a postverbal subject (e.g., Belletti 1988, Alexiadou and Anagnostopoulou 1998, Belletti and Bianchi 2016, Fischer 2016); see the Italian, French, and Dutch examples in (3a–c).

The syntactic accounts of the DE attribute it to the noncanonical, postverbal position of the subject, deriving it from an illicit expletive-subject relation (Safir 1982, Reuland 1983, Chomsky 1995:342–348) or from problems of case licensing (e.g., Belletti and Bianchi 2016). However, the DE is also attested in languages like Hungarian, where subjects do not have a dedicated vP-external position; they assume nominative case in situ (É. Kiss 2002) and are not associated with an expletive. In Hungarian, a DE subject can appear postverbally, when represented by an indefinite NP (4a), or preverbally, in a position reserved for predicative elements pseudoincorporated into the verb, labeled alternatively as Spec,PredP, Spec,AspP, or Spec,TP in the Hungarian syntactic literature (4b).² What is more, the definiteness restriction can also apply to accusative objects (4c).

The semantic accounts of the DE focus on existential sentences and derive it from the incompatibility of the existential predicate and a quantificational/presuppositional subject (see, e.g., Milsark 1974, 1977, Barwise and Cooper 1981, McNally 1997, 1998, 2011). These theories do not regard the definiteness restrictions attested in constructions like (3a–c) and (4a–c) as manifestations of the same constraint and do not aim to account for them.

The DE theory that can give a unified explanation of the DE arising in the different constructions illustrated in (2)–(4), and—as I will argue—also of the DE in the PP types illustrated in (1)—is a semantically motivated, syntactically conditioned DE theory, outlined by Szabolcsi (1986a,b, 1992) and further elaborated by Harlig (1989), Bende-Farkas (1995, 2000, 2001), É. Kiss (1995, 1998, 2002), Kálmán (1995), Maleczki (1995, 2001, 2008), Alberti (1998), Piñón (2006a,b), Peredy (2008), and Kádár (2011), among others. This theory has been mostly ignored in the discussion of the DE, and since it is practically unknown outside Hungary, I will briefly recapitulate its main points.

Szabolcsi’s theory is built on Barwise and Cooper’s (1981) assumption that predicating the existence of an individual presupposed to exist is a tautology, or in the case of negated existential sentences, a contradiction, which is not tolerated in natural language. Szabolcsi’s unique contribution to this approach is that the logical predicate of existence, EXIST, is not only encoded by a BE-type or HAVE-type existential predicate but is part of the meaning of a large, open set of verbs, including verbs of existence (5a), unaccusative verbs expressing coming into being (5b), and transitive verbs expressing creation (5c). The examples below are from Hungarian, but the theory is of crosslinguistic relevance, as I will discuss shortly.

Predicates of these types impose a so-called definiteness restriction—in fact, a specificity constraint—upon their theme argument, the subject in the case of existential and unaccusative predicates, and the object in the case of transitive verbs. Their theme can be represented by a postverbal indefinite (5a,c) or by a bare nominal in Spec,TP, the projection harboring the verb (5b). The interpretations of these two types are minimally different. An indefinite NP is referentially more independent; unlike a bare nominal, it can be referred to by an anaphoric pronoun. Crucially, however, both types of internal arguments displaying the DE are nonpresuppositional, nonspecific (Bende-Farkas 2001). In (6), for example, the indefinite object of the DE verb talál ‘find’ is necessarily disjoint from the object of the preceding clause; it can only be understood to introduce a new referent.

In Hungarian, the negative polarity equivalent of a nonspecific indefinite NP is supplied with a minimalizing particle. Accordingly, negative DE contexts only license an indefinite with a minimalizer (or a bare nominal NP, and other types of negative polarity indefinites).

Szabolcsi (1986a,b, 1992) argues that the common element responsible for the DE in the Hungarian examples above is an EXIST component in the meanings of their predicates. In sentences of type (5a), the predicate means ‘exist’; in sentences of type (4a–b) and (5b), it means ‘come to exist in a particular fashion (in the domain of discourse)’; and in sentences of type (4c), (5c), and (6), it means ‘cause to come to exist in a particular fashion (in the domain of discourse)’.

The theme argument affected by the definiteness restriction—whether represented by a bare nominal or an indefinite NP—is semantically incorporated into the verb. Szabolcsi (1986a,b, 1992:132) motivates this by claiming that the highlighting of the EXIST meaning component of a DE verb suppresses its descriptive content; therefore, the verb needs additional content to function as a natural language predicate. According to Bende-Farkas’s (2000) alternative proposal, the lexical entry for a DE verb like verb talál ‘find’ contains an event description (there is a finding event by the agent) and a consequent state, and the consequent state is contributed to a large extent by the theme argument.

Notice that the reading ‘(cause to) come to exist (in the domain of discourse)’ arises in the perfective aspect. For achievement predicates, perfective is the only aspectual possibility. However, the verbs of accomplishment predicates can also be used as activity verbs in the progressive aspect, in which case they express the change of state of their internal argument, with no DE arising. This is illustrated in (8). In (8a), where the time adverbial enforces a perfective interpretation, the definite object is ungrammatical. In the progressive (8b), by contrast, there is no definiteness restriction on the object.

Crucially, Szabolcsi’s (1986a,b, 1992) theory is not Hungarian-specific; it is just that her crosslinguistically valid generalizations happen to be more transparent in Hungarian than in many other languages. In English, the correlation between the EXIST meaning component and the DE is hard to notice in sentences other than there is/there are constructions because the verbs eliciting the DE are ambiguous: they have both a change-of-state and a coming-into-being or creation interpretation, and the DE is only triggered under the latter reading (see Piñón 2006a,b). In Hungarian, by contrast, accomplishment and achievement verbs with a change-of-state meaning are lexically distinguished from their coming-into-being and creation counterparts. Change-of-state accomplishment and achievement predicates all have a telicizing verbal particle denoting the result state of the internal argument; see the (b) examples of the minimal pairs in (9)–(10). This is absent in the case of coming-into-being and creation predicates, where the result state is the internal argument itself ((9a), (10a)). The internal argument of change-of-state predicates, undergoing the change, is presupposed to exist; hence, it cannot be represented by a nonspecific (singular or plural) bare nominal. The internal argument of coming-into-being and creation predicates, on the contrary, cannot be presupposed; that is, it cannot be definite or specific/presuppositional indefinite (É. Kiss 1998).

The DE associated with the theme argument of verbs with an EXIST meaning component has been grammaticalized in Hungarian; that is, its violation results in sharp ungrammaticality.⁵ At the same time, the choice between a particle verb and its particleless version, that is, the choice between a change-of-state meaning with the EXIST component suppressed and a creation meaning with the EXIST component highlighted, can be pragmatically controlled. Compare:

(11a–b) can be uttered in the same situation, but whereas (11a) describes it as an event changing the state of an existing shirt, (11b) represents it as the creation (or, using McNally’s (2011) term, the instantiation) of an ironed shirt.

Information structure interacts with the DE as follows. As is well-known, topicalization targets referential and specific arguments; hence, those under the force of the DE cannot be topics. More interestingly, the focusing (i.e., movement into Spec,FocP) of a constituent other than the internal argument associated with the DE neutralizes the DE.⁷ Compare:

In (12b), an adjunct identifying a circumstance of the subject’s coming into being is the focus, and the coming-into-being event itself is the presupposed background. The presuppositionality of the coming-into-being event also renders the resulting internal argument presuppositional; that is why the DE is invalidated.

The question arises whether Szabolcsi’s account of the DE is valid crosslinguistically—that is, whether it has a predictive force, for example, in English. English does not morphosyntactically distinguish the change-of-state and the coming-into-being/creation readings of accomplishment and achievement predicates, but the larger context or the pragmatics of a given sentence usually favors one interpretation or the other, and the coming-into-being/creation reading elicits the DE. In (13a), for example, the creation reading, involving the DE, is more plausible; in (13b), by contrast, the change-of-state interpretation prevails.

The perfectivity condition also holds for English;⁸ thus, progressive aspect enforces the change-of-state interpretation, no matter what the pragmatic conditions are.

As in Hungarian, focusing neutralizes the DE in English. If the subject (15a) or an adjunct (15b) is focused, the coming-into-being event becomes presupposed; hence, the resulting NP can also be presuppositional (definite or specific indefinite).

Szabolcsi’s (1986a,b, 1992) theory also accounts for Italian unaccusative sentences of type (3a), expressing coming into being in the domain of discourse. Although Szabolcsi does not discuss French passive sentences of type (3b) or Dutch transitive expletive constructions of type (3c), her approach can be extended to these sentence types as well.⁹ These are also existential sentences; il est/er are logical predicates of existence predicating the existence/occurrence of an event. Actually, all eventive clauses are assumed to involve an event variable that is bound existentially. What is specific about the sentence types in (3b–c) is that the existence of the event represents the main assertion in them; they can be paraphrased as ‘There has been an event such that . . . ’. The distinguished role of the event variable is indicated by the fact that stative predicates, which lack an event argument, are ungrammatical or marginal in these constructions. For example:

In sum: The DE has been pointed out in a wide range of sentence types across languages. Syntactic accounts derive the indefiniteness requirement on postverbal subjects from their relation to the expletive in subject position or from problems of licensing their case. Semantic explanations derive the DE from a conflict between a predicate asserting the existence of its subject and the representation of its subject as presuppositional. Szabolcsi (1986a,b, 1992) recognizes that the constituents targeted by the definiteness restriction are not postverbal subjects but internal arguments and that the predicates responsible for the DE include not only the existential predicate but any verb with an EXIST meaning component, and she proposes an explanation that extends to unaccusative sentences involving a verb of coming into being and to transitive sentences involving a verb of creation. Szabolcsi (1992:sec. 4.4.3) formalizes her DE theory in the framework of Groenendijk and Stokhof’s (1991) version of Dynamic Semantics.¹⁰ Whereas the DE has a semantic trigger in Szabolcsi’s approach, she also identifies syntactic conditions on its emergence: namely, the DE is absent in progressive sentences and in sentences with a narrow focus other than the internal argument. The pragmatic conditions may play a role in the emergence of the DE by highlighting or suppressing the EXIST meaning component of the verbs responsible for the DE.

In the following section, I test Szabolcsi’s theory against a new set of data: PPs displaying the DE.

The so-called definiteness restriction (in fact, specificity constraint) attested in PPs blocks the same types of noun phrases that are blocked in existential sentences, in unaccusative clauses expressing coming into being, and in transitive clauses expressing creation. Therefore, it is reasonable to assume that they are manifestations of the same phenomenon, and an appropriate theory of the DE should be able to account for them. The phenomenon will be introduced on the basis of Hungarian material, but examples from other languages will also be discussed.

The Hungarian Ps (postpositions) eliciting the DE are híján ‘for lack of’ and nélkül ‘without’.

In PPs displaying the DE, the complement of P cannot be represented by a presuppositional noun phrase, for example, a proper name, a personal pronoun, a definite NP, or a possessive construction with an overt possessor. Apparent exceptions to the definiteness restriction are NP types that can also occur as subjects in existential sentences (Bassaganyas-Bars and McNally 2020).¹¹

However, these intentional NPs are only formally definite; the modifiers várt ‘expected’ and ígért ‘promised’ indicate that they do not denote existing individuals. According to McNally (1997), such definite NPs are acceptable because they can be interpreted as entity correlates of properties; that is, they are nonspecific in the sense that a várt eredmények corresponds to ‘results of the expected kinds’ and az ígért segítség corresponds to ‘help of the promised kind’.

The Ps imposing a specificity restriction on their complements, híján ‘for lack of’ and nélkül ‘without’, convey a so-called abessive meaning, expressing the nonexistence/nonpresence or the nonparticipation/nonuse of the referent of their complement.¹²Híján always encodes the former of these two meanings, for example, the nonexistence/nonpresence of native speakers in (19a). Nélkül can indicate either nonexistence/nonpresence or nonparticipation/nonuse; see (19b), which allows both interpretations. The definiteness restriction can also be absent in nélkül PPs (under conditions to be discussed in section 3.4). A definite complement to nélkül enforces its non-participation/nonuse reading (19c).

(19a) expresses that there are no native speakers of the language in question (in the universe or in the domain of discourse). (19b) can be true both in a situation where there are no native speakers and in a situation where there are native speakers but they do not participate in writing the grammar. (19c) refers to an existing, familiar group of native speakers who do not participate in grammar writing.

A shared property of híján and nélkül PPs is that they create weakly negative contexts for their complements, where nonveridical indefinites can appear.

The semantic parallels and semantic differences of híján and nélkül PPs derive from their partly identical, partly different internal structures.

Hungarian Ps are either “case-like,” subcategorizing for a noun phrase, or “case-assigning,” sub-categorizing for a PP (Marácz 1986, Kenesei 1992). Híján and nélkül are case-like Ps. A typical case-like P was originally a case-marked possessum in a possessive construction, and the caseless noun phrase complementing it functioned as its possessor. The grammaticalization of the [−V,+N] head of the possessive construction into a [−V,−N] postposition usually took place after its nominal stem or its adverbial case ending or both became obsolete. This is what happened in the case of híján, as well: its stem, híj ‘void’ disappeared from the language except for the compound fog-híj ‘tooth-gap’. The original morphological makeup of híján, consisting of the noun híj, the possessive suffix -a, and the superessive suffix -n, is not transparent for present-day speakers, but the possessive construction is still present in the PP in a partially fossilized form and can be activated in certain circumstances.¹³

The possessive construction in (21) is a predication structure, a small clause, a Relator Phrase (RP) in Den Dikken’s (2006) terminology, where the predicate híj ‘lack’ predicates the nonexistence of its subject, the NP in Spec,PossP. Its subject cannot be a specific NP, for example, a personal pronoun. Present-day intuition segments the structure in (21) as [_PP [_NPtapasztalat] [_Phíján]]. Nevertheless, the underlying possessive construction becomes visible in some contexts. As Szabolcsi (1994) shows, a possessor is not always caseless; it can also be marked by dative case—optionally in some contexts, and obligatorily in others. For example, possessors represented by a wh-pronoun prefer the dative option. This is what we find in the case of híján PPs, as well: when the complement of híján is an interrogative or relative pronoun, it bears dative case. Thus, the interrogative *mi híján and the relative *ami híján ‘for lack of what’ are ungrammatical; the grammatical versions are mi-nek híján, ami-nek híján, containing a dative-marked mi/ami ‘what’.

The use of dative case on the possessor is obligatory when the possessor is extracted from the possessive construction. Certain PPs can be pseudoincorporated into the verbal predicate of their clause, forming a complex predicate with it, a condition of which is the removal of the complement from the PP. The extracted complement, to be adjoined to the VP, cannot be caseless; hence, the complement of a case-like postposition in a PP like (21) can only be extracted if it is dative-marked. This is the case also when híján is pseudoincorporated into the copula, forming an expression meaning ‘lack’ with it, as in (22).

Notice that the dative-marked NP selected by híján can also be definite or specific indefinite—but in this case it is not the logical subject of híján anymore but a complement of the complex unaccusative predicate híján van ‘lack’, whose theme argument is Pál, the grammatical subject.

For lack of, the English equivalent of híján, elicits the same DE as its Hungarian counterpart. Database searches provide examples with nonspecific indefinite complements, and native English consultants reject variants with a definite NP (23)—unless it is a type-denoting noun phrase also licensed in there is/are constructions (24).¹⁴

The structure of for lack of expressions is similar to that of híján PPs: for selects a small clause complement, an RP, containing a negative existential predicate (lack). The English RP instantiates a reverse predication relation, with the predicate c-commanding the subject.

The trigger attributed to the DE by Szabolcsi’s (1986a,b, 1992) DE theory is present in both híján PPs and for lack of PPs. Híj and lack, the predicates of the small clauses in the underlying structures of the PPs, are negative predicates of existence, instantiating a negated EXIST. As the theory predicts, (non)EXIST imposes a specificity restriction on the subject so as to avoid the contradiction that would arise between a presuppositional subject and a predicate of nonexistence. The negative component that these predicates share is responsible for the nonveridical indefinite NP in the híján PP in (20a) and in the corresponding for lack of PP in the English translation (for lack of any (kind of) evidence).

Nélkül grammaticalized into a P much earlier than híján; hence, its morphological makeup is even less transparent. That nélkül PPs have an underlying structure similar to that of híján PPs is clearest in the case of nélkül PPs containing a pronominal complement. In Hungarian possessive constructions, the possessum agrees with its pronominal possessor (it bears a morpheme complex consisting of a possessive suffix and an agreement suffix; see Szabolcsi 1994 and Dékány 2021, among others). Possessive agreement also appears on case-like postpositions taking a pronominal complement. (In the case of híján, possessive agreement is forestalled by the DE, that is, the impossibility of a pronominal possessor.) However, this is not what we find in the paradigm of personal pronoun + nélkül PPs.

Interestingly, the agreement morpheme elicited by the pronominal complement appears not on the postposition nélkül, but on a locative element in its complement, the morpheme -nál, a derivative of the reconstructed Uralic stem -na meaning ‘proximity’, surviving in Modern Hungarian as an adessive case suffix, the equivalent of the English at (Benkő 1970:996). This form suggests that the complement of nélkül is also a possessive construction functioning as a small clause, as shown in (27).

The postposition was originally kül ‘out’, expressing exclusion, which incorporated the harmonizing, front-vowel version of nál (Benkő 1970:1009), adding a negative meaning component to it. In structure (27), the incorporated locative is doubled in its original structural position. Pronouns also have a paradigm without doubling, shown in (28). In this version, the possessive agreement suffix appears on nélkül.

When the complement of nélkül is a lexical NP, there is no agreement suffix on P; P follows the NP like a case suffix. In Estonian, the abessive morpheme is not case-like but is a proper case suffix; nevertheless, Den Dikken and Dékány (2018) argue that it also takes a small clause complement. What appears to be the complement of the case suffix is selected not by the suffix but by a silent predicate, an abstract locative predicate THERE. In the Hungarian nélkül PP, the element corresponding to THERE is the locative nál in the case of pronominal complements, and an empty element (historically, the trace of the locative nél incorporated into the postposition) in the case of lexical complements. Since lexical complements elicit no possessive inflection, I label their small clause as an RP, following Den Dikken and Dékány (2018).

Den Dikken and Dékány (2018) assign a small clause complement to the English without, as well, updating Beukema and Hoekstra’s (1984) analysis of with PPs. Without selects a small clause complement with an overt or covert predicate. The predicate appears overtly in the absolute construction.

In simple, nonabsolute without PPs, the predicate of the small clause complement is an abstract locative THERE.

Notice that the morphological makeup of without is similar to that of nélkül: with was originally a locative, meaning ‘against, opposite, from, toward, by, near’, which came to be incorporated into the negative preposition out, the equivalent of the Hungarian kül. The silent THERE occupies the position vacated by the incorporated locative.

Castroviejo, Oltra-Massuet, and Pérez-Jiménez (2015) have proposed a different structure for Spanish sin ‘without’ PPs with a bare nominal complement. They discuss sin PPs that are used as postnominal modifiers in expressions like una habitación sin luz ‘a room without light’. They analyze the PP as a relative clause with a null verb HAVE and with a node encoding negation, assuming that the complementizer introducing the relative clause is the preposition (adopting Emonds’s (1985) theory identifying P and C). The relative clause is assumed to be the CP complement of a determiner, with the modified nominal (habitación) moved to Spec,CP.

The clausal analysis is supported by the fact that the PP can contain temporal and locative modifiers.

The assumption of a syntactic NEG component is based on the fact that sin licenses N-words.

The presence of a null HAVE is inferred from the fact that sin PPs display the DE, just like existential HAVE constructions.

Castroviejo, Oltra-Massuet, and Pérez-Jiménez (2015) argue that HAVE is a two-place predicate establishing a relation between habitación ‘room’ and luz ‘light’.¹⁶ When the complement of HAVE is a bare N, it pseudoincorporates into HAVE via theme suppression as a verbal modifier.

In fact, Castroviejo, Oltra-Massuet, and Pérez-Jiménez’s (2015) analysis formalizes the same intuition as structure (31). As argued by Benveniste (1966), Kayne (1993), Den Dikken (1995), and Belvin and Den Dikken (1997), among others, have is the inverse of be in possessive and existential constructions, spelling out be + P (the head of the dative or locative complement of be), taking the complement of P as its subject and the subject of be as its object. Compare the two representations in (37).

In structure (37b), luz is the theme subject of a small clause with an abstract locative predicate, expressing ‘existence at a contextually determined location’ (which is identified by the context as una habitación). The target of the definiteness restriction is the theme subject of the locative. In the HAVE construction in (37a), this relation is inverted; the complement of the locative P functions as the subject, and the theme displaying the DE is the object. Like the EXIST predicates discussed by Szabolcsi (1986a,b, 1992), HAVE needs additional content to function as a natural language predicate; therefore, it incorporates the object.

Structure (37a) is stipulated to contain a TP as well; however, there is no evidence of any finite or nonfinite T in the structure. The temporal adjunct in (33) does not necessitate a TP; noun phrases can have temporal adjuncts. T can also be dispensed with as a case assigner, as it is by no means necessary for habitación to originate in the relative clause (for arguments against the raising analysis of relatives, see for example Borsley 1997). Whereas both (37a) and (37b) encode the interpretation of sin luz appropriately, (37b) is syntactically better motivated.

In híján/for lack of PPs, the DE has been derived from an existential predicate, an instantiation of EXIST, in the small clause complement of P. The questions arise whether the DE attested in nélkül/without PPs is also a concomitant of an EXIST logical predicate of existence, and if it is, what EXIST is encoded by, and why the existential interpretation does not always obtain.

As for the locus of EXIST, recall that the predicate of the small clause complement of nélkül/without is a generally empty locative element, which is the trace of the locative nál/with incorporated into the negative P historically and is interpreted as an abstract locative synchronically. A locative is part of the canonical existential construction; it represents the optional coda in the string ‘expletive, copula, pivot, (coda)’. Kádár (2011) argues that the locative in existentials is optionally overt rather than optionally present; existence is, after all, presence at a spatiotemporal location. The covert locative, an abstract THERE, is usually understood to refer to the situation of the discourse (38a) or to have a universal meaning (38b).

There is no expletive in Hungarian existential sentences. In the case of copula + locative complexes, either the copula or the locative can be highlighted, by being moved into the position of primary stress at the left edge of the comment. Existential interpretation arises and the DE is triggered when the copula is emphasized (39a). Emphasis on the locative yields a locative sentence with no DE (39b).

In some analyses, the logical predicate of existence is instantiated by the coda, which forms a small clause with the pivot. In canonical existential sentences, this small clause is a complement to be, a raising predicate. McCloskey (2014), for example, argues that the core existential construction in Irish, occurring in a great variety of finite and nonfinite contexts, is a small clause consisting of a pivot and a coda, which is often represented by an abstract locative, the pronoun ann. According to Moro (1997), the English there, too, originates as the predicate of an existential small clause and raises to Spec,TP to satisfy the EPP. The small clause complement of the post-position nélkül, consisting of a subject and an abstract locative, is also an existential construction of the Irish type.

However, it is not the case that the locative predicate of the small clause complement of a P always functions as a logical predicate of existence. Ps taking a small clause complement with a locative predicate form a large set in Hungarian, including case-like postpositions such as alatt ‘under’, fölött ‘above’, mellé ‘near-to’, and adverbial case endings such as -ban/ben ‘in’, -ból/ ből ‘from’, as well as -val/vel ‘with’, the positive counterpart of nélkül. With PPs have been claimed to select the same type of small clause complement as without PPs both in English (Beukema and Hoekstra 1984) and in Estonian (Den Dikken and Dékány 2018). Nevertheless, these PPs display the DE much less frequently than without PPs. In a 230-million-word German corpus, the proportion of determinerless complements is 15% in the case of unter ‘under’ PPs, 21.9% in the case of mit ‘with’ PPs, and 83.6% in the case of ohne ‘without’ PPs (Kiss and Roch 2014). In a 1.4-million-word French corpus, the proportion of bare nominal complements is 15% among avec ‘with’ PPs and 39% among sans ‘without’ PPs (Le Bruyn 2017). The set of bare nominal complements is smaller than the set of nonspecific complements as the latter also includes indefinite NPs with a nonspecific determiner; this fact, however, does not invalidate the above proportions.

The frequency of a bare nominal complement significantly correlates with the interpretation of the P. Müller et al. (2012), Kiss and Roch (2014), and Kiss (2019), examining mit ‘with’, ohne ‘without’, ober ‘above’, and unter ‘under’ PPs in German, and Le Bruyn (2017), examining avec ‘with’ and sans ‘without’ PPs in French, have all found that the possibility of determiner omission is primarily determined by the sense of the preposition, with some syntactic conditions also playing a role. These studies distinguish three to five senses of with/without. Kiss and Roch (2014) assume four senses, defined as follows: (a) The sense MODAL (INSTRUMENTAL) indicates that a device, a tool, or means is (not) used for a certain purpose. (b) The sense CONDITIONAL is used when considering the (negative) condition or the prerequisite for another situation to happen. (c) PARTICIPATION (COMITATIVE) expresses that two entities (animate or inanimate) are (not) being together, (not) being involved, or (not) acting together in an activity. (d) PRESENCE indicates the presence or absence of a thing, an attribute, or a property, which is typically part of something else in a mereological relation. Kiss (2019) adds a fifth sense to Kiss and Roch’s (2014) list: RESTRICTIVE (Kiss also renames PRESENCE as MEREOLOGY). Reinterpreting CONDITIONAL as partly PRESENCE, partly PRESENCE/PARTICIPATION, Le Bruyn (2017) distinguishes three senses: PRESENCE, PRESENCE/PARTICIPATION, and MODAL—which roughly correspond to my EXISTENCE/PRESENCE, PARTICIPATION, and USE interpretations. The crucial dividing line in all three versions is between PRESENCE and the other senses: it is the PRESENCE meaning that supports determiner omission most strongly. Compare the distribution of complements with and without a determiner in ohne and mit PPs (Kiss and Roch 2014:87).

In mit PPs, determiner omission is the preferred option only with the sense PRESENCE. In ohne PPs, it is frequent with all four senses, but it is nearly obligatory with PRESENCE.

The data also indicate that the correlation between the EXISTENCE/PRESENCE interpretation of P and the nonspecificity feature of its complement is unidirectional. A P meaning EXISTENCE/PRESENCE nearly always elicits the DE, but nonspecific complements can occur with Ps of different senses, as well. This is illustrated by the minimal pairs in (41)–(42).

The definite NPs in the nélkül PP in (41a) and the without PP in (42a) elicit the PARTICIPATION and USE interpretations, respectively. The nonspecific NPs in (41b) and (42b) license both the EXISTENCE/PRESENCE and the PARTICIPATION/USE readings, but in the case of without any break, pragmatics makes the PARTICIPATION/USE reading unlikely.

The proportions in (40) suggest that what determines the frequency of the DE in PPs with a small clause complement is the frequency of the EXISTENCE/PRESENCE reading. Hence, the question to answer is why the abstract locative tends to have an EXISTENCE/PRESENCE interpretation more frequently in without PPs than in other PPs. The element that without does not share with other Ps taking a small clause complement, and that is likely to be responsible for this, is negation, represented by the exclusive out. (Le Bruyn (2017), proposing a full-fledged semantic account based on aspectual requirements, arrives at a similar conclusion.)

The role of negation in licensing existential interpretation can be accounted for in the framework of Szabolcsi’s (1986a,b, 1992) DE theory as follows. Szabolcsi describes the meaning of DE predicates as ‘((CAUSE TO) COME TO) EXIST in a particular fashion’. Existential interpretation arises when the EXIST component is highlighted, with ‘in a particular fashion’, determined by the descriptive content of the predicate, suppressed—as illustrated by the minimal pair in (11). The locative predicate of the small clause complement of P expresses ‘a particular manner of PRESENCE at a spatiotemporal location’. PARTICIPATION and USE are particular manners of presence determined by the context. When the ‘particular manner’ component is suppressed, PRESENCE, instantiating EXIST, becomes dominant, and the DE is triggered. Apparently, the meaning component ‘in a particular manner’ is less relevant and easier to suppress in the case of NONPRESENCE than in the case of PRESENCE. In other words, absence at an eventuality is interpretable without any further specification, whereas presence at an eventuality is understood as a particular kind of presence in most cases.

Abessive PPs display the DE crosslinguistically. Híján/For lack of PPs only allow nonspecific indefinite complements. In the case of nélkül/without PPs, the DE is present in a subset of cases. It is systematically elicited when the P means NONEXISTENCE/NONPRESENCE rather than NONPARTICIPATION or NONUSE. The specificity restriction that these abessive adpositions exert on their complements is similar to that exerted by predicates of existence, coming into being, and creation upon their internal arguments. The source of the specificity restriction is the same as in verbal existential constructions: an EXIST meaning component of the predicate in the small clause complement of P. In the case of for lack of, EXIST is instantiated by the nominal predicate lack. In the case of without, EXIST is a meaning component of an abstract locative predicate expressing ‘a particular manner of PRESENCE at a given spatiotemporal location’. The P with and its equivalents across languages also take a small clause complement of this type. The DE arises when the EXIST component of the locative is highlighted, and the component ‘a particular manner’ is suppressed. Its suppression is facilitated by the negative element incorporated into without; hence, the DE is triggered much more frequently in without PPs than in with PPs.

The recognition of the DE in PPs has led to new evidence in the debate on what causes the DE; which is the decisive factor among the syntactic, semantic, and pragmatic circumstances interacting in its emergence; and whether the DE attested in different syntactic contexts can be given a unified explanation.

The structural differences between the PPs and the clause types displaying the DE preclude a unified syntactic explanation. The source of the definiteness restriction cannot be either an illicit expletive-DP relation or a special way of assigning case to the complement of P.

The factor that PPs displaying the DE and existential sentences share is a predicate of existence. In canonical existential sentences, the existential predicate consists of a copula and an overt or implicit coda. The locus of existential predication in PPs displaying the DE is the small clause complement of the P. In abessive PPs involving híján/for lack of, the small clause predicate incorporated into P is a negative nominal predicate (híj/lack), the nominal equivalent of the negated copula. The coda—that is, the location of existence—remains implicit; it can be reconstructed from the context or the situation. In abessive PPs involving nélkül/without/ohne/sin, the small clause predicate incorporated into P is an abstract locative, corresponding to the coda.

The derivation of the DE from a PP-internal existential predicate supports Barwise and Cooper’s (1981) insight that the DE is a constraint blocking the predication of the existence of an individual that is shown by its determiner to exist, and it confirms Szabolcsi’s (1986a,b, 1992) DE theory, which associates this constraint with an EXIST logical predicate of existence present in a large set of natural language predicates.

The proposed account of the DE in PPs also has implications for the analysis of the PP. It provides additional support for a theory that assigns a small clause structure to the complements of a type of adposition, put forward by Beukema and Hoekstra (1984) for the preposition with, and updated and extended to other pre- and postpositions and adverbial cases by Den Dikken and Dékány (2018).

The preparation of this article was supported by grant KKP 129921 of NKFIH, the Hungarian National Research, Development, and Innovation Office. I am grateful for the useful comments of Ágnes Bende-Farkas, Marcel den Dikken, Christopher Piñón, Anna Szabolcsi, and the reviewers of LI.