1 Introduction
1.1 The phenomenon
English [Q+where] expressions do not uniformly quantify over places. The same surface family, the set of forms built on where, supports at least three referential domains: ordinary locations, place-linked institutions, and, in a restricted impersonal construction, human inhabitants of a locative domain. The central question of this paper is how one where-based paradigm can support these place-, institution-, and person-oriented readings without reducing them to simple lexical polysemy.
The baseline use is locative. In (1), everywhere ranges over places.
- (1)
- I’ve looked everywhere (in this house) for the keys. locative
A second use is institutional. In (2), everywhere can quantify over shops, businesses, or other institutions associated with the relevant locations.
- (2)
- Everywhere (in the UK) takes credit cards these days. institutional
A third use is impersonal. In some varieties of English, a restricted everywhere can function as an impersonal subject referring to a generic set of inhabitants of a given geographic location, as in (3) (cf. Sluckin & Kastner forthcoming). The apparent-time study by Sluckin & Kastner (forthcoming) demonstrates that a growing number of English speakers accept the construction; it is primarily anchored in British English, though it can be found, to a lesser extent, in North American and other varieties.
- (3)
- a.
- Everywhere in the UK eats fish and chips.
- b.
- Everywhere on the coast eats fish and chips. impersonal
The canonical and clearest impersonal use is tied to an overt locative PP. Without an appropriate locative domain, the impersonal reading is unavailable or sharply degraded; moreover, the domain must be geographically large rather than narrowly local. The impersonal reading therefore requires the locative PP, with marginal exceptions discussed in §6.3.
- (4)
- Everywhere (#in the village)/??(in the UK) eats fish and chips.
1.2 The puzzle
This three-way typology is the framing problem for the paper: one surface family quantifies over places, place-linked institutions, and human inhabitants without collapsing these readings into lexical polysemy. Within that broader frame, four more specific puzzles drive the analysis.
(1) Obligatoriness of the PP. Why is the PP-restrictor required in the canonical impersonal reading and optional for the institutional reading, when both involve the same surface form [every+where]?
(2) Restrictor size. Why is the PP additionally constrained to denote a large geographical domain (in Italy but not in Italian restaurants)?
(3) The DP–locative tension. The construction binds, controls, and antecedes pronouns like a full DP, yet its head exponent where is a locative pro-form. How can a locative element occupy A-positions otherwise reserved for full DPs?
(4) The split with everyone. Everywhere replaces everyone in precisely those contexts where a covert human argument has been claimed to be independently available and licensed by a PP (see Brody 2013), suggesting a non-accidental relation to existing mechanisms for impersonal licensing in English. What is that relation?
1.3 The proposal and roadmap
The account developed here rests on the claim that the outer syntax of [Q+where] is constant across all three readings; the impersonal reading differs from the other two in one respect, namely which silent root is merged at the base of the nominal projection. We adopt i) a Kaynian silent-noun analysis of locative pro-forms (Kayne 2005), ii) a layered extended nominal projection composed of two functionally distinct nP layers [nPea [nPcat]] (see Roberts 2019) under a P-headed extended projection (Grimshaw 2005), and iii) Brody’s (2013) observation that English locative PPs license a covert generic people argument. We adapt the Kaynian system to a DM framework while retaining its empirical basis largely unchanged (§5.3.1).1 Locative [Q+where] is built over a silent
root. Institutional [Q+where] retains this locative structure but receives a contextual institution-oriented interpretation. The impersonal reading is the most complex case: it has the same external where-morphology, but its lower nominal domain contains a silent human root.
Proposal: The impersonal use of everywhere is the spell-out of an extended nominal projection with three defining properties:
| Lexical | The unpronounced root is the silent sortal, human-descriptive , in contrast to the silent sortal locative of the locative use. |
| Categorisation |
is selected by a relational categoriser, ncat: that is, by a categoriser which introduces a relation to a place, rather than one which simply nominalises its complement. |
| Selection | The locative PP merges in Spec,nPcat, driven by a c-selectional [uPlace] feature on the relational categoriser ncat. |
These are not three independent stipulations, since the root is the single lexical choice and the relational categoriser and its [uPlace] feature are what that choice requires: the relation this categoriser introduces is a relation to a place, which a person-sorted argument cannot supply. The locative derivation needs no such structure, because there the root is already of the relevant sort. Section 5.3 sets the dependency out in full.
is itself sortal (denoting the descriptive content of being a person) but its embedding under a relational ncat yields a 2-place inhabitant relation. The relational structure is contributed by the categoriser; the root contributes only human-descriptive content. In the canonical overt-PP construction analysed here, the [uPlace] feature on ncat is valued by the locative PP, which supplies the spatial argument of the inhabitant relation.
The higher categoriser nea abstracts over an individual variable (λx) interpreted as a human referent. The resulting human property serves as the restrictor for the universal Q0 head (every).
Although we focus on English throughout, the crosslinguistic typology of impersonal pronouns is well-studied (e.g. Cinque 1988; Egerland 2003; D’Alessandro 2007; Alexiadou & Carvalho 2017). To our knowledge, no exact analogue has yet been described in the languages most often discussed in that literature, such as German man, Italian si, and French on. Indeed, brief exploration of consultant judgements indicates that the English pattern is ruled out in major continental Romance and Germanic languages (1); systematic cross-linguistic comparison is left for future work.
- (5)
- a.
- Everywhere on the coast eats fish and chips.
- b.
- German
- Überall
- everywhere
- an
- on
- der
- the
- Küste
- coast
- isst
- eats
- *(man)
- imp.pn
- Fischbrötchen.
- fish-buns
- ‘Everywhere on the coast, one eats fish buns.’
- c.
- Norwegian
- Overalt
- everywhere
- i
- in
- Norge
- Norway
- spiser
- eats
- *(man)
- imp.pn
- fårikål.
- fårikål
- ‘Everywhere in Norway, one eats mutton and cabbage.’
- d.
- Italian
- Ovunque
- everywhere
- in
- in
- Italia
- Italy
- *(si)
- imp.pn
- mangia
- eats
- la
- the
- pizza.
- pizza
- ‘Everywhere in Italy, one eats pizza.’
- e.
- French
- Partout
- everywhere
- en
- in
- France
- France
- *(on)
- imp.pn
- mange
- eats
- du
- of.the
- steak.
- steak
- ‘Everywhere in France, one eats steak.’
The impersonal derivation requires a quantified form built on an interrogative pro-form, so that the root may go unpronounced (§5.3.3), and it requires the resulting form to occur as a subject (§5.6.3). The counterparts in (5) fail one condition or both, since German überall, Norwegian overalt and French partout contain the quantifier all rather than an interrogative form, and Italian ovunque is not synchronically decomposable. These conditions on availability, however, dictate where the reading may arise, but not where it must. Indeed, in English the construction is recent, incipient and still spreading (§2.1). Absence in other languages is therefore weak evidence against the analysis, and systematic comparison is left for future work.
The paper proceeds as follows. Section 2 maps the three readings; Sections 3 and 4 develop the locative and institutional backgrounds; Section 5 gives the impersonal derivation; Section 6 turns to consequences and open questions; and Section 7 concludes the paper.
2 The empirical phenomenon
2.1 The experimental baseline
Sluckin & Kastner (forthcoming) have recently tested the impersonal interpretation of everywhere in an acceptability judgement study with 407 native English speakers (339 UK; 68 North American), using a 7-point Likert scale. They tested everywhere’s argument status, subjecthood and thematic properties through binding, control, agent-oriented adverbs, object versus subject behaviour, and compatibility with different types of transitivity and voice alternations (see e.g. Roeper 1987; Williams 1987; Koenig & Mauner 1999; Bhatt & Pancheva 2006; Landau 2010a; Williams 2015; Bhatt & Pancheva 2017).
Ordinal Bayesian regression showed that the impersonal reading depends on an overt locative restrictor (see §2.4). Bare everywhere was severely degraded across all demographics. Adding a PP restrictor denoting a very large location, which we label macro-geographic, rescues the construction by licensing full-DP syntactic behaviour of the entire everywhere complex, e.g. control via PRO and third-person plural binding (see §§2.5–2.6).
Sluckin & Kastner’s data also reveal regional variation; North American speakers rated the construction significantly lower than British speakers. The UK sample exhibits an age gradient: speakers aged 18 to 49 rate the restricted construction higher than older cohorts. However, the fact that younger speakers reject the unrestricted bare form means that the phenomenon cannot be dismissed as broad pragmatic tolerance for colloquial metonymy. Given the apparent-time design of the study, Sluckin & Kastner (forthcoming) conclude that the construction could well be an incoming structural change in the grammar of British English. While the diachronic implications of their findings call for a separate formal treatment, the present study restricts itself to a synchronic analysis of the novel impersonal everywhere construction, though the relative-clause cases in §6.3 may bear on that question.
2.2 Three readings of [Q+where]
Everywhere in British English illustrates the three-way contrast introduced above. In its locative use, the quantifier ranges over places. In the institutional use, documented experimentally by Sluckin & Kastner (forthcoming), it ranges over a set of institutions (shops, councils, universities) linked to places. In the impersonal use, it ranges over the human inhabitants of a geographic location and antecedes they/their.
- (6)
- a.
- I’ve looked everywhere (in this house) for the keys. locative
- b.
- Everywhere (in the UK) accepts credit cards these days. institutional
- c.
- Everywhere *(in the UK) loves Marmite. impersonal
Table 1 summarises the three readings; the root and structural rows preview the analysis developed in §§3–5.
Table 1: Three readings of [Q+where].
| Locative | Institutional | Impersonal | |
| Root/semantic base | ![]() |
plus contextual transfer |
under relational ncat
|
| Referential domain | places | place-linked institutions | human inhabitants |
| Pronoun behaviour | locative there/it | its or their | they/their/themselves |
| Locative domain | optional PP adjunct | optional PP helps identify the institutional set | obligatory overt macro-geographic PP saturates the place argument |
| Status of reading | structurally locative | contextual/metonymic | structurally encoded in the lower nominal projection |
The impersonal reading, akin to everyone, falls within the quasi-universal range of the impersonal-pronoun typologies of Cinque (1988) and Egerland (2003), alongside English one, German man, Italian si, and French on. We now distinguish its properties from those of institutional and canonical locative everywhere.
2.3 The universal impersonal pronoun: everyone
Everyone represents the ordinary human baseline: a quantificational head combines with a human-denoting nominal restrictor. The relevant point for the present paper is not a full decomposition of everyone, but the fact that the human component is morphosyntactically accessible. This is visible in the contrast between (7a) and (7b): ordinary everyone supports who-relatives and human pronominal binding without further machinery.
- (7)
- a.
- Everyone who knows Mary loves her.
- b.
- *Everywhere who knows Mary loves her.
In semantic terms, every combines with a property-type restrictor and yields universal quantification (see Heim & Kratzer 1998). Syntactically, we represent this as a high QP ranging over a lower human/pronominal DP, in line with i) QP approaches to quantifiers (Shlonsky 1991; Matthewson 2001) and ii) analyses of pronouns as internally structured nominal/D-domain expressions (Postal 1966; Déchaine & Wiltschko 2002).
2.4 The restrictor
The central impersonal reading is licensed by an overt locative PP, and the PP must denote a large geographic domain. Removing the PP from a sentence that supports the impersonal reading leaves only the institutional reading (8); replacing it with a small or non-geographic PP degrades the impersonal reading even where the construction is otherwise well formed (9).
- (8)
- a.
- Impersonal/Institutional:
- Everywhere in England disapproves of public displays of affection.
- b.
- *Impersonal; institutional only:
- Everywhere disapproves of public displays of affection.
- (9)
- a.
- Everywhere in Italy likes to take a nap in the afternoon.
- b.
- # Everywhere in Italian restaurants likes to take a nap.
The diacritic marks the impersonal reading; the locative reading is independently anomalous, since places do not take naps. The size restriction is specific to impersonal everywhere. The locative-licensed (pseudo-)impersonal use of they described by Brody (2013) tolerates non-geographic restrictors when the predicate fits the location:
- (10)
- a.
- In Italy they like to take a nap in the afternoon. (ex. 4, Brody 2013: 34)
- b.
- In Italian restaurants they make pasta properly. (ex. 16b, Sluckin & Kastner forthcoming)
A conceptual analogue is independently attested. Locations metonymically denote their human inhabitants only when the location can host a subdivisible population (place for inhabitants metonymy; Radden & Kövecses 1999; Littlemore 2017), as in (11).
- (11)
- a.
- London drinks more than Edinburgh.
- b.
- # The living room drinks more than the kitchen.
Impersonal everywhere uses this option but imposes explicit structural restrictions on it: in Italian restaurants can support the metonymy but cannot license impersonal everywhere. The metonymy also requires an overt subject, since *In Italian restaurants makes pasta is ungrammatical (author’s judgement); the PP cannot supply one by itself. §5.4 returns to this contrast, deriving PP-dependence from [uPlace] on a relational categorising head and treating size as an interface condition on the inhabitant relation.
The construction also appears marginally available with a that-relative in place of the PP (12); Sluckin & Kastner (forthcoming) report, in contrast, that who-relatives are generally ungrammatical. Example (12) is a fragment answer (from the television series Ted Lasso) to Where is she going? and involves existential somewhere (§2.6). Relative-clause restrictors, and associated human readings, are discussed in §6.3.
- (12)
- Somewhere that believes they deserve her. (Sluckin & Kastner forthcoming)
2.5 Nominal status and human binding
Impersonal everywhere passes the standard diagnostics for full-DP-hood, applied in Sluckin & Kastner (forthcoming) drawing on the implicit-argument literature.2 It controls (13a), antecedes pronouns across clauses (13b), and binds reflexives, reciprocals, and bound pronouns (13c)–(13e).
- (13)
- a.
- Everywhere in the UK has been trying [PRO to quit smoking].
- b.
- Everywhere in the UK is saying it these days. They don’t think it’s strange.
- c.
- Everywhere in modern Britain helps themselves first.
- d.
- Everywhere in the country helps each other.
- e.
- Everywhere in England washes their salad before eating it.
- (All examples from Sluckin & Kastner forthcoming)
Sluckin & Kastner (forthcoming) present experimental data and a detailed discussion showing that the impersonal reading of everywhere targets a generic-human variable. On the impersonal reading, the bound pronoun must be they/their or themselves; its is excluded on the impersonal reading. However, its is felicitous under the institutional reading, as shown in (14).
- (14)
- a.
- # Everywhere in the UK washes its salad before eating it. impersonal
- b.
- Everywhere (in the UK) sets its own council tax policy. institutional, its licit
The they/their-only pattern aligns impersonal everywhere with other indefinite human quantifiers, paradigmatically everyone, where singular they resolves a human-denoting variable bound by the quantifier (cf. Huddleston & Pullum 2002). The institutional reading admits both its and their, selecting between them on the basis of singular/plural construal of the metonymic referent.
2.6 Distribution
Two further properties constrain the analysis. Firstly, impersonal everywhere is restricted to subject position; object positions force a locative or institutional reading.
- (15)
- a.
- Everywhere in the UK loves Marmite. impersonal subject
- b.
- # We visited everywhere in the UK. *impersonal; locative only
Secondly, the construction extends across the quantificational paradigm, with quantifier-specific rating profiles in Sluckin & Kastner (forthcoming).
- (16)
- a.
- Somewhere in the world probably eats cat.
- b.
- Nowhere in the UK likes Andrew.
- c.
- I wonder if anywhere in the Middle East eats eagles.
The same impersonal structure must therefore be available below some-, no-, and any- as well as every-. §5 locates that structure inside the nominal projection, with the quantificational determiner merging above it.
Two questions follow from the observations in this section: what introduces the human variable inside a locative-looking QP, and why does the overt locative domain appear with it? The size restriction in (9) suggests that this domain must be able to support a subdivisible inhabitant population, while the extension across some-, no-, and any- shows that the effect is tied to where, not to every alone.
3 Locative everywhere: A layered nP analysis inside the extended nominal projection
To formalise the internal configuration of indefinite locative pronouns (e.g. everywhere, somewhere), we adopt the basic premise of Kayne (2005) that r-pronouns contain a structurally active, phonologically null spatial root that introduces a place, his NPLACE; however, we motivate here an approach making use of a layered nP analysis (see Roberts 2019) and a P-headed extended nominal projection following Grimshaw (2005). This builds on Kayne (2005), but does not implement his exact proposals.
3.1 where as an indeterminate/indefinite pronoun
Kayne (2005) notes that non-standard English dialects show patterns involving r-pronouns like here and there which systematically co-occur with overt demonstratives, e.g. this here book. Consequently, he considers r-pronouns to function as determiner-like elements embedded within a larger DP-NP structure and proposes a syntactic configuration in which an unpronounced demonstrative determiner, akin to that, selects an NP housing the r-pronoun and the covert locative noun place. To derive the surface order, the constituent containing the r-pronoun and the covert locative noun undergoes leftward movement over the covert determiner (see Kayne 2005: 69–74), as formalised in (17):
- (17)
- [there place]i that ti (ex. 30, Kayne 2005: 69)
While we accept Kayne’s premise of a silent place root, an embedded phrasal movement derivation for where is less straightforward than the one he establishes for here and there. The empirical distribution of where diverges. Firstly, where prohibits co-occurrence with overt determiners in prenominal position (e.g. *what where place, *which where place). While the ungrammaticality of an overt definite demonstrative (*that where place) could theoretically be ruled out by a simple definiteness mismatch regardless of the D- or N-status of where, substituting the indefinite use of the demonstrative this (Prince 1981) gives the same ungrammaticality: *this where place.
While the prenominal sequence *what where place is severely degraded, the postnominal sequence what place where is acceptable. In this configuration, where is an independent locative adjunct to the DP/NP, rather than part of its internal structure. This asymmetry indicates that where occupies a saturated maximal projection, which cannot act as a modifier or determiner within the nominal spine.
To capture this distributional asymmetry, we model where not as a simplex D-head, but as a compositional P+D element derived within the P-headed extended nominal projection (Grimshaw 2005).
Firstly, treating wh- as an indeterminate/indefinite variable (cf. Kratzer & Shimoyama 2002), wh- is merged in D0, selecting an nP as its complement. We develop the internal architecture of this nP in the next section; for now, we assume equivalence to Kayne’s silent NPLACE. This follows analyses in which pronouns occupy the D-layer and select an unpronounced nominal complement (Postal 1966; Abney 1987; Panagiotidis 2002). Any attempt to merge an external determiner over the projection triggers a structural violation, since D0 is lexicalised via wh-. This accounts for the ungrammaticality of determiner stacking (e.g. *this where place).
P0 hosts the locative portion of where; we retain the segmentation wh-ere, but note that Kayne (2007) treats locative-deictic -E- and -re as morphemic, while Leu (2025: 110–111) treats -re and in as allomorphic English counterparts to the Mandarin locative postposition lǐ.3 Once P0 merges above the DP, the D0 element wh- incorporates into P0 (see Baker 1988) via head movement. This incorporation is a narrow-syntactic operation rather than a purely post-syntactic or PF displacement (cf. Matushansky 2006). Incorporation via head movement thereby yields the complex D/P head realised as where. Simplified versions of the “base”-generation and spell-out derivation are given in (18a,b); these are schematic, and we expand the nominal layer in §3.2.
- (18)
- a.
- b.
Thus, the silent nominal root inside the complement of D0 is not itself pronounced by this operation; the overt form where realises the complex incorporated D/P unit.
3.2 Variable introduction by n: A VoiceP–nP parallel
We formalise the internal configuration of the locative pro-form where by adopting a split nP architecture (Roberts 2019). We follow Roberts (2019) in separating two functions that are often bundled together in the nominal domain. The lower categorising head ncat merges with the root and determines the nominal category of the expression. This assumes that acategorial roots require syntactic merger with categorising functional heads (Embick & Marantz 2008; Embick 2015). The higher head nea introduces the nominal referential argument, the R-role (Williams 1981), by abstracting over an individual variable. This division parallels the separation between verbal categorisation and argument introduction in the clausal domain: categorisation is handled low, while the variable made available for binding by higher functional structure is introduced at the edge of the nominal phase.
We motivate the semantic mechanics of nea via a structural parallel between the clausal Voice domain and the nominal n domain. Following Schäfer (2008) and Alexiadou et al. (2015), we take the introduction of an argument role to be separable from the syntactic projection of a specifier. Active Voice introduces an external-argument role and bears an edge feature that projects a specifier hosting the external argument; non-active Voice introduces the role but lacks the edge feature, leaving the argument unsaturated in narrow syntax and bound from above by existential closure in the passive case (Bruening 2013; Alexiadou et al. 2015).
We propose that nea in the locative pro-form where is the nominal counterpart of non-active Voice. The lower categoriser ncat merges with the root
to provide its nominal status, i.e. Kayne’s NPLACE, and nPcat denotes a property of locations, λy.place(y), of type ⟨e,t⟩. Immediately above this, nea identifies the open variable in its complement nPcat as the R-role of the nominal projection (Williams 1981; Roberts 2019).
Whether this variable is syntactically saturated by a DP merging in Spec,nPea is controlled by an independent edge feature on nea. In the locative pro-form, nea lacks this edge feature; no specifier is projected and no DP saturates the variable via Functional Application. The R-role variable remains open at nPea (still type ⟨e,t⟩) and is bound from above by a c-commanding operator.
- (19)
If the locative pro-form were to resolve as a referential pronoun (type e), it would clash with quantificational heads like every, which require a property-type restrictor (type ⟨e,t⟩) as their first argument (Heim & Kratzer 1998). Because nea inherits the open property of its complement without saturating the R-role, nPea remains a 1-place property of locations. Following the standard treatment of indefinites as introducing free variables (Kamp 1981; Heim 1982), the wh-feature on D0 takes nPea as its restrictor without supplying quantificational force; the resulting DP denotes an open expression of type ⟨e,t⟩.4 The free variable is bound by a c-commanding operator higher in the extended projection, e.g. the universal quantifier every in everywhere.
3.3 Adding the universal quantifier
For purely locative everywhere, we adopt an articulated nominal architecture separating quantificational heads from the lower extended nominal projection, i.e. the DP layer (Shlonsky 1991; Matthewson 2001). That is, quantifiers such as every, some, and any externally merge as the head of a dominating QP (contra DP approaches to quantifiers Abney 1987; Larson & Marušič 2004).5 When the universal quantifier every merges as the head of QP, it takes the entire open locative extended nominal projection headed by where, i.e. as its semantic restrictor. To prevent a type mismatch at this boundary, the locative P0 head acts semantically as an identity function ( ), passing the ⟨e,t⟩ type of the DP upward to the PP level unchanged, thereby providing the semantic type that Q0 requires. P0 is the locus of the locative morphological exponent -ere (§3.1) and also the target of D-incorporation. Additionally, it heads the extended nominal PP projection (cf. Grimshaw 2005), hosting D-features visible to T for probing (§5.6.2).
The spatial restriction is not obligatory, but when present it is provided by a locative PP, e.g. in the UK, which merges as an adjunct to the outer edge of the nPea projection, as shown in (20). Following Adger (2003), we assume that prepositional modifiers within the nominal domain structurally adjoin to the outermost nP projection.
- (20)
3.4 Morphological coherence at PF
The syntax in §3.3 establishes Q > P > D for binding, whereas morphological cohesion requires everywhere to resist syntactic interruption (21a). We derive where by syntactic incorporation involving leftward head-adjunction (see Baker 1988), but this operation does not derive a cohesive “word” every+where. A narrow-syntactic head-movement approach to incorporation of P0 to Q0 would incorrectly predict a linear order (*where-every) which is ruled out under the Mirror Principle (Baker 1985a; b; 1988).
- (21)
- a.
- *every exactly where
- b.
- exactly everywhere
We argue that the final operation is post-syntactic and applies to the linearised string via Local Dislocation (Embick & Noyer 2001: 563–564).6 Local Dislocation acts on linearly adjacent elements; the operation involves string-vacuous rebracketing, which Embick & Noyer take to be freely available between adjacent elements, yet intervening adjuncts cannot be ignored. In everywhere, Q0 and the complex P0 head formed by incorporation are already adjacent and in the attested order, so that nothing is reordered, and Local Dislocation thus derives the strict ban on intervention. Lowering could not derive this and would wrongly allow Q and P to form a word across the adjunct in (21a), because it applies before Vocabulary Insertion and skips intervening adjuncts (cf. Embick & Noyer 2001: 562). The restrictors -where, -one, -thing and -body are bound and require a host, which forces word formation. In contrast, in every beer the nominal is a free word without such a requirement; every other beer shows that the two need not even be adjacent.7 It is the paradigm, rather than every, that triggers the operation. The adjacency requirement excludes interruption but not modification of the quantifier, since exactly everywhere is well formed with exactly modifying every (cf. exactly everyone, exactly everything).
Overall, the analysis distinguishes two displacement operations: syntactic upward incorporation (cf. Baker 1988) of D0 to P0, which feeds post-syntactic Local Dislocation (cf. Embick & Noyer 2001) of Q0 and P0. The two operations target different elements and have separate motivations. Firstly, the incorporation analysis is motivated for the indefinite/interrogative r-pronoun where; whether it extends to other r-pronouns remains open. More generally, the idea that indefinites, pronouns, r-pronouns, determiners, and demonstratives decompose into several parts is well established (inter alia Haspelmath 1997; Déchaine & Wiltschko 2002; Kayne 2005; Leu 2015). These authors argue for decomposition, but not for head movement; the head-movement implementation is our own, and it is motivated by the determiner facts rather than by decomposition as such. D and P must form a single complex in the narrow syntax if the projection is to block an external determiner. Secondly, dislocation of a Q-head at PF may hold beyond everywhere in morphologically complex indefinite pronouns.
The operational split reflects divergent requirements of the LF and PF interfaces, following the standard Y-model architecture (Chomsky 1995; Embick & Noyer 2001). D0-to-P0 incorporation must be in the narrow syntax because it carries structural consequences, blocking external determiners such as *this where place. In contrast, the universal Q0 must remain higher than the compositional complex D+P head in the narrow syntax to c-command the extended nominal projection and bind the locative variable at LF. Since Local Dislocation applies only at PF, Q0 remains above the restrictor for semantic composition at LF, while everywhere is nonetheless assembled as a single word.
4 Institutional everywhere: Locative syntax and institutional interpretation
We turn now to the institutional use of everywhere in a sentence like (22):
- (22)
- Everywhere (in London) accepts Visa these days. (Reading: every shop/business)
Sluckin & Kastner (forthcoming) propose that the institutional reading is licensed when the discourse or the predicate supplies an Austinian topic situation (Kratzer 2021), making a contextually D-linked set of institutional referents available (Pesetsky 1987). We adopt this view: the institutional reading is a contextual interpretation of an otherwise locative nominal structure.
Institutional everywhere shares the internal syntactic structure detailed in (20). It maintains
and the structurally optional PP adjunct. It does not contain
, nor does it require a relational ncat bearing [uPlace] (see §5). The institutional interpretation arises when a predicate such as accept Visa imposes a sortal requirement that a locative expression does not meet: the predicate is naturally satisfied by shops, businesses, or other institutional entities, while the DP itself contributes a spatial property. Context can resolve this mismatch by mapping salient places to institutions located there.
This is compatible with predicate transfer in the sense of Nunberg (1995) and type coercion in the style of Asher (2011). For present purposes, the key point is structural: the transfer applies to the interpretation of the locative nominal restrictor without changing the narrow syntax.
Treating the institutional use as a contextual sortal shift over unchanged locative syntax gives rise to four contrasts with the impersonal use. Firstly, the PP remains optional. It modifies the locative base rather than satisfying [uPlace] on a relational categoriser. Secondly, institutional everywhere shows pronoun flexibility in binding. Singular its and plural their are both possible because the institutional referent may be construed either as a single institutional entity or as a plural set of institutions. Thirdly, the institutional reading is subject to cancellability. Following Sluckin & Kastner (forthcoming), acceptable sentences provide a topic situation via context or predicate entailment, as in (23a, b). Without such a situation, the institutional construal fails, as in (1c).
- (23)
- a.
- Predicate entails shops/businesses
- Everywhere takes credit cards.
- b.
- Context entails public authorities
- In the build-up to the war, everywhere built air-raid shelters.
- c.
- No reconstructable topic situation
- #/*Everywhere loves a shower.(ex. 13, Sluckin & Kastner forthcoming)
Finally, a copredication restriction (see Pustejovsky 1995; Asher 2011) obtains under the institutional reading of everywhere, which resists simultaneous locative and institutional predication.
- (24)
- #Everywhere is hilly and accepts Visa.
The failure of copredication is consistent with the transfer analysis. Once the institutional interpretation has been obtained, the QP behaves as an institutional expression and cannot act as a flexible place/institution complex. On this analysis the spatial variable is existentially closed: the locative variable serves only to identify the relevant institutional referent and is not itself projected as the matrix-level denotation. The QP can therefore satisfy an institutional predicate, but it cannot simultaneously serve as the argument of an ordinary spatial predicate.
Taken together, the diagnostics motivate two derivational routes rather than a single underspecified lexical item.
5 Impersonal everywhere: Selection of
5.1 From locative domains to human inhabitants
Our proposal builds on Brody’s (2013) observation that locatives can support covert human arguments. Sentences such as (25) permit an impersonal reading of they despite the absence of an overt plural antecedent.
- (25)
- In Italy, they like to take a nap in the afternoon. (ex. 4, Brody 2013: 34)
Brody argues that the locative expression licenses an implicit semantic human argument, conceptually a generic collective or people. The spatial expression provides an anchor for this silent human argument because geographical and institutional locations naturally host inhabitants. We take this as independent evidence that spatial and human categories can be grammatically linked.8
Brody’s mechanism is not by itself an analysis of impersonal [Q+where]. Firstly, the human variable in impersonal everywhere behaves as part of the quantified subject itself: it controls, binds pronouns, and contrasts with institutional its. Secondly, the construction is tied to the internal morphology of where-based quantifiers rather than to an ordinary locative PP in the clause. Thirdly, the contrast between impersonal and institutional readings requires a structural distinction between human and place/institutional restrictors.
The analysis below therefore preserves Brody’s insight but relocates it inside the nominal projection. The locative domain does not merely license a free-standing contextual people variable; in the canonical impersonal construction it saturates a relational nominal structure whose individual variable is human.
5.2 The core impersonal derivation
The extended nominal projection remains topped by a PP, following Grimshaw (2005), with D incorporating into P (Baker 1988), the resulting D/P complex realised as where. The QP layer hosts every, with the same Q–P dislocation at PF established in §3.4. The impersonal derivation differs from the locative derivation in three respects:
The numeration selects the unpronounced sortal root
rather than the sortal root
, embedded under a relational categorising head ncat. This extends Kayne’s (2005) and Corver’s (2008) silent Nperson, also discussed in semantic terms by Brody (2013) in his work on pseudo-impersonals.The locative PP restrictor merges in Spec,nPcat rather than as a high nPea adjunct. Merge of the PP is driven by a c-selectional feature [uPlace] on ncat.
Because the relational nPcat structure denotes a human inhabitant relation, the individual variable (λx) licensed by nea resolves to a human referent.
The formal analysis below targets the overt-PP construction, the empirical case around which the experimental generalisation is built. Marginal cases in which the locative domain is recovered from discourse or from relative-clause material are discussed separately below.
The impersonal derivation is selected for generic or existential reference to human inhabitants of a subdivisible geographic location. It is unavailable in the PP construction when the predicate is incompatible with a [+human] subject, when no overt PP can value [uPlace] on ncat, or when the PP fails the macro-geographic interface condition on the inhabitant relation. The construction is therefore structurally less complex than everyone everywhere in X (Sluckin & Kastner forthcoming): a single QP contains a silent human root embedded under a relational categoriser.9
The resulting structure is given in (26).
- (26)
The tree in (26) represents this division of labour by showing λx on nea and [uPlace] on ncat; the place argument is supplied by the PP rather than abstracted over by nea.
5.3 Building the base:
, the relational categoriser, and the [uPlace] feature
5.3.1 Syntactifying implicit person
The bridge in §5.1 motivates a connection between locative domains and human inhabitants, but it does not yet provide the internal syntax of impersonal everywhere. We syntactify Brody’s insight by locating the human-descriptive content in the nominal spine, implementing this via the unpronounced-nominal architecture of Kayne (2005) and the silent Nperson of Corver (2008).10
Kayne’s silent-noun programme allows interpretively active nominals (number, place, etc.) to go unpronounced under recoverability (Kayne 2005; 2015). Within this framework, Corver (2008) argues for a semi-lexical silent noun person (in the sense of Emonds 1985), active in hidden possessive constructions attested in Alblasserwaard Dutch, such as (27):
- (27)
- a.
- Dutch
- We
- we
- kwamen
- met
- Anna’s
- Anna-s
- tegen.
- prt
- ‘We met Anna.’
Corver’s (2008) silent noun is licensed by an interpretable [+person]/[+human] feature on the possessor. His diagnostics identify person as semi-lexical rather than a full lexical N: it resists attributive modification, plural marking, and coordination.11
We adopt Corver’s empirical motivation for a silent human-descriptive primitive but recast it in DM-style terms: Corver’s semi-lexical N person corresponds here to a silent sortal root
categorised by ncat. We recast Kaynian silent nouns generally in the same way. For Kayne (2005), silence is a property of a noun which is syntactically present but unpronounced; here it is a root’s failure to receive overt exponence in a given categorial context. Late Insertion makes these compatible rather than competing descriptions, since a terminal carries no phonological content in the narrow syntax, and the exponence-based version additionally derives where the root is silent and where it is overt (§5.3.3). We adopt Kayne’s diagnostics for unpronounced nominal structure, but not the assumption that such structure must be lexically rather than morphologically silent. Our extension to impersonal everywhere involves one distributional difference and two theoretical claims. Distributionally, it lacks overt morphological exponence (unlike Corver’s optional persoon). Theoretically, the categoriser is relational (bearing [uPlace]) and thus narrowly c-selects an obligatory spatial PP. The silent root
contributes human-descriptive content; the relational structure is located entirely on the categoriser. Keeping these contributions on distinct terminals also matters morphologically: under the assumptions set out below, ncat supplies relational structure and conditions exponence, while
supplies the sortal human content independently motivated by Corver/Kayne-style silent nominals.12
5.3.2 A multi-layered interaction
Following Borschev & Partee (2001: 6), we distinguish simple sortal nouns, which denote 1-place properties, from relational nouns, which possess an internal argument place. In baseline locative everywhere,
is sortal and projects no obligatory dependent. The root
is likewise inherently sortal (denoting the descriptive content of being a person) but in the impersonal construction, it composes with a relational categoriser ncat to yield a 2-place inhabitant relation.
By locating relationality on the functional structure rather than the root (see Adger 2013), the composition ncat(
) denotes a 2-place relation between a person and a location. We model this on Borschev & Partee’s schema for relational nouns, further restricting the variables to specific semantic sorts (cf. Borschev & Partee 2001: 12) to capture the strict structural requirements of the impersonal construction:
- (28)
The relation contributed by ncat is a relation to a place, and its value in a given derivation is fixed by the sort of its external argument rather than specified on the head. Where the external argument is person-sorted, as in the impersonal construction, the relation is one of inhabitance, anchored in the overt paradigm of relational nouns in (29); where it is place-sorted, the same head yields a containment relation, as in the boroughs of London. inhabitant-of is thus the value this relation takes when its external argument is human, so it is not a primitive stipulated on a silent head.13 The macro-geographic restriction remains an interface condition, rather than a syntactic feature, on when this relation yields a plausible distributed population. Thus, before PP saturation, ncat + (
) denotes an inhabitant relation with two argument positions, yplace and xperson. The PP then saturates yplace; nea abstracts over xperson.
The internal argument of relational nouns, overt or not, is therefore introduced by structure. The impersonal everywhere construction follows this pattern, but the relational categoriser selects a locative PP in its specifier, rather than a nominal of-phrase as the complement of a lexical N or a relator phrase (for the latter see den Dikken 2006).
We formalise the locative argument via feature-driven Merge. The relevant analogy is the clausal vP domain, where intermediate specifier positions can license internal arguments (cf. Larson 1988; Pylkkänen 2008; Alexiadou & Schäfer 2011). Specifically, to satisfy the strict Locality of Categorization (Embick 2010; 2015), ncat first merges with
as its complement. The uninterpretable [uPlace] feature on the categoriser then triggers the subsequent merge of a locative PP in Spec,nPcat.14,15
This configuration establishes a Spec-Head relation, valuing [uPlace] and providing the spatial variable (y) associated with the inhabitant relation. Nothing hinges here on whether the dependency is implemented as Spec-Head valuation or as a version of reverse Agree: the relevant point is that the selected locative PP locally values the place feature of relational ncat. At the interpretive interface, this formal dependency identifies the denotation of the PP with yplace, the open place argument of the inhabitant relation; it is not a separate syntactic projection. It also keeps the selected PP low: the PP is part of the relational nominal base, rather than an adjunct added after the human variable has already been introduced.
Overt relational nouns display a parallel need for local saturation, morphologically marked in English by of. This holds for both sorts: for inhabitant-class nouns, whose external argument is human, and for subdivision nouns, whose external argument is itself a place.
- (29)
- a.
- the inhabitants ?(of London).
- b.
- the residents ?(of this building).
- c.
- the population ?(of Italy).
- d.
- the boroughs ?(of London).
- e.
- the counties ?(of England).
- f.
- the outskirts ?(of the city).
Borschev & Partee (2001) treat such nouns as 2-place predicates, with the of-PP supplying the internal argument. The overt of-phrase is not obligatory here, since the internal argument may instead be supplied by context: the boroughs voted against it is well formed where a relevant city is salient. The argument is obligatory although its pronunciation is not, and it distinguishes relational from sortal nouns. Impersonal everywhere shares this requirement for local saturation but differs formally: the root is unpronounced, and ncat selects a spatial PP (in, on) rather than an of-phrase. Unlike the overt relational nouns, it requires an overt PP. Because the root is silent, the locative PP both saturates the place argument and supplies the local context under which that root is recoverable, so contextual supply is not available as an alternative (cf. §5.6.3).16
The derivation therefore projects a site for the locative argument, which Sluckin & Kastner (forthcoming) call a “restrictor”. In this core derivation, if no PP is merged, [uPlace] on ncat cannot be valued and the impersonal derivation crashes. The string is nonetheless well formed, because an independently convergent derivation is available: sortal, non-relational ncat selects
without a [uPlace] requirement, giving the locative or institutional parse (§3–§4). Bare everywhere is therefore grammatical but cannot receive the impersonal reading.
The lower nP thus contains two variables: the internal spatial location (y) and the external individual variable (x). The locative PP saturates y, yielding a property of individuals. Subsequently, nea merges as the nominal locus of the referential argument (the R-role) (Roberts 2019) and abstracts over x, making the resulting human property available as the restrictor for Q0 (every).
5.3.3 Morphological spell-out of where (and not who/one/person)
If the numeration selects
, why does the structure surface as where rather than who, person, or everyone? The answer, in Distributed Morphology, cannot be that the higher spatial shell overwrites the lower root. Vocabulary Insertion applies late and targets each terminal node independently (Halle & Marantz 1993; Embick 2015), and a higher P0 cannot erase, absorb, or rewrite a lower root across a cyclic boundary (Embick & Marantz 2008). The form where spells out the functional shell. The silence of
has to be explained on different grounds.
That silence is the result of local contextual allomorphy. We assume, following Embick & Marantz (2008) and Embick (2015), that roots admit contextual allomorphy conditioned by their categorising heads, including zero exponents. As argued in §5.3, the impersonal derivation embeds the human-descriptive root
as the sister of a relational categoriser ncat; this is where root exponence canonically takes place (Embick 2010). The impersonal construction thus involves a zero exponent for
conditioned by the categoriser:
- (30)
- a.
-
↔ θ /__ncat[rel] (relational categoriser)
- b.
-
↔ person/people (elsewhere)
The conditioning context is the identity of the relational categoriser, not the higher PP shell. This keeps the conditioning local and available at Spell-Out. The restriction to locative-restrictor contexts is derived independently in the syntax: ncat[REL] bears [uPlace] by specification and consequently requires Merge of the overt locative PP in Spec,nPcat. The zero exponent is confined to exactly that configuration. Outside it, the root falls under the elsewhere condition (or Subset Principle) (cf. Kiparsky 1973; Halle & Marantz 1993): under a default sortal n the root surfaces overtly, as in the person. Were this categoriser not to condition a zero exponent, Vocabulary Insertion would realise the root overtly alongside the overt PP shell, giving rise to the ungrammatical *where person.
With the root locally unpronounced, the overt morphology of the higher functional shell is realised alone. As established for the canonical locative use of where (§3), the extended nominal projection is topped by spatial P0, into which D0 incorporates, and the D/P complex is realised as where. The choice between
and
is invisible in the overt shell; the root remains syntactically and semantically active despite its locally conditioned zero exponence.
The silence of
reflects independently motivated structure. The relational categoriser ncat is required on grounds independent of exponence, i.e. it bears [uPlace] which forces the locative PP, and projects the inhabitant relation; the root
, however, is independently active, supplying the [+human] content behind they/their binding (§2.5). Zero exponence is one reflex of this configuration, not its sole motivation, and the rule is non-vacuous: under a default sortal categoriser the same root surfaces overtly (the person, people). Root allomorphy conditioned by a local category-defining head, zero exponence included, is standard in the framework (Embick & Marantz 2008; Embick 2010).
Concerning the range of available roots, we suggest that if the numeration can select
where the locative derivation selects
, nothing in the syntax excludes
,
,
or
.17 Kayne (2007) independently analyses approximative somewhere, as in she has written somewhere/*someplace between 10 and 15 papers this year, as containing an unpronounced thing rather than place.18 A [Q+where] form built over a non-locative silent root is thus already motivated outside the construction studied here.
differs in that its sort composes with the place relation on ncat, which gives rise to a spatially distributed population, and hence a domain suitable for generic quantification. Things, times, reasons and manners do not stand in an inhabitance relation to a geographical domain, so the interface condition that also yields the macro-geographic restriction (§5.4) suffices to exclude the corresponding structures.19
The choice of restrictor is not specific to everywhere: quantified pro-forms of this kind are bipartite, pairing a quantifier with a restrictor drawn from a small set of ontological categories, and that restrictor is built either on an interrogative pro-form or on a generic noun (Haspelmath 1997: 21–30). However, English forms indefinite pronouns that contain both generic-noun and interrogative bases coexisting within a single series (see Haspelmath 1997: 28, n. 7): interrogative-based somewhere and somehow versus generic-noun-based somebody, something and someplace. Our account restricts the impersonal reading to the interrogative-based type, as only here can the root go unpronounced while the word still carries an exponent, the pronounced material being supplied by the shell above it. Where the restrictor is a generic noun, that noun spells out the root, and no silent alternant is available. The exclusion of everyone, everybody and everyplace follows on the same terms. Impersonal everywhere is in this respect a hybrid; its restrictor is the generic-noun root
, while its exponence is that of the interrogative type.20
5.4 Interface-mediated constraints: Quantification, location, and predication
5.4.1 Encyclopedic knowledge and sortal restrictions
As established by Sluckin & Kastner (forthcoming), the PP in the canonical construction must denote a large-scale geographical location. We model this in two steps. The syntactic feature [uPlace] guarantees only that the PP denotes a place; it does not encode size. This division is independently motivated by the fact that small places, large places, buildings, restaurants, and countries are all syntactically locative PP domains. A simple locative feature would therefore wrongly treat them alike. While the narrow syntax forces the presence of the locative PP to value [uPlace], the macro-geographic restriction is not encoded as a formal syntactic feature. We assume a standard generate-and-filter architecture at the syntax-semantics interface (see e.g. Borer 2005a; b; Preminger 2014). That is, the restriction need not be encoded in the syntax, which builds what its features allow; structures that survive the syntax may still be rejected on interpretive grounds. Two encyclopedic conditions apply: the domain must be a geographic region, so that it has inhabitants at all, which excludes in Italian restaurants; and it must be subdivisible, so that the population counts as distributed. Micro-locations fail this post-syntactic filter because they cannot be construed as hosting a socially and geographically distributed inhabitant population suitable for generic quantification (31).
- (31)
- #Everywhere in the building helps out in the communal garden.
This contrasts with the implicit people argument associated with they in locative contexts (cf. Brody 2013). Impersonal they can be pragmatically associated with smaller domains because the locative PP remains a clausal or contextual anchor. In canonical impersonal everywhere, by contrast, the PP is inside the nominal restrictor of quantification as the local argument of relational ncat. The domain must therefore be suitable as the restrictor for a human quantifier, not merely locative in syntax. The restriction is a property of the nominal architecture of everywhere, not a general property of locative PPs.
- (32)
- In Italian restaurants, they make good pasta.
5.4.2 Genericity and quantificational variation
To capture the distribution of impersonal where-based pronouns, we pursue a DP-internal analysis of quantificational force. This avoids the tension that a clausal Gen operator would create with existential and negative forms such as somewhere and nowhere. Quantificational force comes from Q0: every introduces universal quantification (∀), binding the human variable licensed by nea. The episodic judgements in this subsection are exploratory author judgements pending systematic testing; the structural claim is only that the same lower nominal structure can occur under different Q-heads. Quantification is over individuals, not over sub-regions of the locative domain: the PP saturates the place argument and is not quantified over. Where a strict universal would be too strong, the weaker reading reflects the exception-tolerance of a characterising statement and involves no second quantifier over places.
The layered nP complex provides a human variable (λx), while the spatial restrictor (y) constrains the domain pragmatically. Following von Fintel (1994), quantificational domains are systematically constrained by contextual variables. Here the variable is interpreted through an inhabitant relation (inhabitant-of(y)(x)), so the quantifier ranges over spatially organised human individuals rather than over an unrestricted set of persons.
This spatial organisation is compatible with the construction’s generic interpretation. Because everywhere is evaluated over a geographically distributed restrictor, localised episodic events are often pragmatically disfavoured: in ordinary contexts, it is implausible for a single event to involve all relevant inhabitants across the domain. Episodic predicates are therefore not structurally banned, but they require a context supporting distributed experience across a macro-geographic domain, as in (33).21
- (33)
- a.
- Everywhere in the world watched the moon landing.
- b.
- Everywhere in the world watched the moon landing.
Without such a distributed event, the matrix predicate tends towards a characterising reading (Krifka et al. 1995), such as a cultural habit or environmental response (34). This is a preference created by the spatially distributed restrictor, rather than an independent generic operator in the clause. We use characterising for a stable or generalising property of the subject rather than a single spatio-temporally bounded event.
- (34)
- Everywhere in the Netherlands rides bicycles.
5.4.2.1 Somewhere
Replacing universal every with existential some weakens these restrictions: a localised event need not involve all inhabitants of the domain. Existential somewhere therefore appears compatible with localised episodic predicates, provided the same nominal structure is available. Because the relevant location is not identified, the resulting reading often has an epistemic flavour (1).
- (35)
- a.
- Somewhere in the world must be awake at this hour.
- b.
- Somewhere in the city must be freezing in this rain.
5.4.2.2 Nowhere
With negative no in Q0, the construction places negation over the relevant spatially organised domain. Since the absence of an event does not require a simultaneous pluralisation of a given event, i.e. there is no positive requirement that everyone in the defined domain does or is doing x, episodic nowhere seems more flexible than everywhere. The judgements are still preliminary, but examples such as (36) suggest that both habitual/generic and episodic predicates are available.
- (36)
- a.
- Nowhere in Europe eats fermented frogspawn.
- b.
- Nowhere in the country is rioting right now.
- c.
- Nowhere in Europe is experiencing power cuts today.
5.4.2.3 Anywhere
The polarity-sensitive variant anywhere instantiates the same structure with any in Q0. Following Dayal (1998), any may be treated as a universal quantifier with modal force, and its distribution as an impersonal subject is compatible with the standard restriction to non-veridical contexts (Giannakidou 1998). Under negation and in interrogatives (37a, b), the characterising-predicate preference of positive everywhere (§5.4) is plausibly weakened because these operators block an assertion of a real-world temporally simultaneous universal event.22
In positive unembedded contexts, the Free-Choice use of anywhere may avoid an episodic clash via subtrigging (see Dayal 1998), as in (37c), where the relative clause supplies the required variation over situations. This preserves the same lower nominal structure while shifting the licensing conditions to the QP and its environment.
- (37)
- a.
- I don’t think anywhere in the country is watching the playoffs right now.
- b.
- I wonder if anywhere in the UK is watching the playoffs.
- c.
- Anywhere in England that eats oysters is probably by the sea.
5.4.2.4 Summary
Across everywhere, somewhere, nowhere, and anywhere, quantificational variation is localised at Q0 rather than on a higher gen operator, while the nominal structure remains constant.
5.5 Ruling out lexical polysemy: Coercion and dot-objects
One might alternatively hypothesise that impersonal everywhere retains a spatial root (
) and receives its human interpretation through lexical polysemy or metonymy. A dot-object analysis in the style of Pustejovsky (1995), for instance, could treat [ncat +
] as a complex place∙human object whose spatial and human facets are both available. A dot object is a single item with two facets that remain simultaneously available, as book is both a physical object and a text; coercion instead shifts an item of one sort to another as the context demands. This would make the construction closer to ordinary place-for-inhabitant metonymy.
The difference between the present account and a polysemy account is as follows. Firstly, a polysemy analysis requires a separate entry for everywhere, somewhere, nowhere, and anywhere, and predicts nothing about their uniformity. Instead, our account states the option once, on a common constituent in the series, and predicts that the series patterns alike in this respect, as (16) indicates. Secondly, a polysemy analysis requires two independent stipulations: that there is a human entry, and that this entry independently demands an obligatory PP. In contrast, our structural account derives both from a single relational head, whose pairing with a human root follows from the sorts its relation imposes. Furthermore, the structural account predicts the coordination facts to be discussed in (44), the who-relative gap in (46), and the its/their asymmetry already detailed in (14), none of which follows from lexical polysemy. Finally, the PP’s obligatoriness reflects structural argument saturation, as with the relational nouns in (29), rather than an ad hoc diacritic for the impersonal reading.
Copredication is directly relevant to the dot-object version of this hypothesis (see e.g. Cruse 1986; Asher 2011). If the pronoun behaved like overt geopolitical entities that permit spatial and human predicates together, as in (38), coordination should be natural. Yet with physical or biological human predicates, simultaneous copredication is zeugmatic:
- (38)
- The city is polluted and voted Green
- (39)
- a.
- #[Everywhere in the Netherlands]i is completely flat and ti rides bicycles to work.
- b.
- #[Everywhere in Southern Spain]i gets incredibly hot in August and ti takes a siesta after lunch.
This pattern also blocks copredication in the institutional use, but for a different reason: as §4 argued, the institutional reading existentially closes the spatial variable. The impersonal pattern is instead expected if the numeration selects sortal
under relational ncat: the spatial denotation is supplied by the selected PP, not projected by the QP itself.
The contrast between (38) and (39) therefore speaks against a place ∙ human dot-object, which would keep its spatial and human facets available. It does not, however, discriminate between the present account and a metonymic one, since conventionalised place-for-inhabitant metonymies such as London votes Labour also resist simultaneous predication. That distinction rests on the considerations set out above and on the discourse-anaphora contrast to which we now turn. If the QP retained a spatial denotation, it-anaphora should be easier. Instead, the contrast in (40) suggests that the impersonal QP makes a human referent available, while spatial anaphora must target the PP restrictor or a locative use.
- (40)
- a.
- Everywhere in the Netherlands rides bicycles to work, and #it is mostly below sea level.
- b.
- Everywhere in the Netherlands rides bicycles to work, and they also eat Frikandel with chips and mayonnaise.impersonal
- c.
- Everywhere in the Netherlands is flat, and it is mostly below sea level.spatial
Taken together, the copredication and anaphora facts favour the discrete-root analysis over a dot-object or coercion analysis that leaves a spatial denotation active at the QP level. They do not by themselves prove that no metonymic route is ever available, but they combine with the binding profile of the construction to motivate a structural split.
5.6 Extensions and broader empirical coverage
5.6.1 Bound they/their
The construction obligatorily binds 3pl they/their (§2.5). Our analysis provides a human property via λ-abstraction, and quantification ranges over inhabitants of the saturating PP. Bound they/their follows from the indefinite, non-referential status of the bound variable, on a par with other human quantifiers (everyone, nobody, no one); see Huddleston & Pullum (2002) on singular they with quantified antecedents. The infelicity of it/its reflects the human-descriptive semantics of
and is therefore expected without assigning relationality to the root. A featural treatment is beyond the scope of this paper.
5.6.2 Clausal integration: Case and the EPP
The proposed nominal structure resolves the DP–locative tension raised in §1.2. The clausal subject is a QP, not a bare locative PP. All readings share this category, since the outer syntax is uniformly QP > PP > DP > nP; they differ only in the nature of ncat, in the root it selects and in where the PP attaches. The diagnostics in §2.5 demonstrate that the impersonal form behaves as a nominal argument, which is compatible with its sharing the category of locative everywhere. This Q-layer corresponds to the high quantificational layer found in ordinary expressions such as everyone. We do not assume Q = D; however, QP and DP pattern alike for subject licensing (cf. Abney 1987). This follows if T requires a nominal extended projection rather than D specifically. The QP status of the subject separates impersonal everywhere from true locative inversion. In true locative inversion, a preverbal locative PP has been claimed to satisfy the EPP on T but fails to trigger ϕ-agreement on T (see inter alia Bresnan 1994; Roberts 2010). Impersonal everywhere, by contrast, functions as a nominal QP subject that values T’s ϕ-probe.
This valuation proceeds via direct syntactic probing. In indefinite pronoun constructions in English, we assume that quantifier heads such as every, any, some, no lack a valued ϕ-set. While quantifiers like every restrict their domains and semantically select singular complements, they do not project formal ϕ-features. Therefore, T’s EPPd probe bypasses Q and directly targets the incorporated D/P complex. The upward incorporation of the D head (wh-) into the P head (-ere) places nominal D-features at the same level as P, hence there is no risk of an intervention effect. These D-features then satisfy T’s EPP requirement (cf. Chomsky 1995; 2001).
The 3sg agreement on T follows directly from this configuration. T agrees with the indefinite D/P complex (where), which we consider inherently underspecified for number as an indefinite. Therefore, the morphology inserts the 3sg affix via elsewhere realisation (Halle & Marantz 1993).23
5.6.3 The subject restriction: A Subject-of-Predication dependency
The clausal-integration account in §5.6.2 explains how impersonal everywhere satisfies T’s formal EPP and agreement requirements, but not why the impersonal reading is confined to subject-like positions. If everywhere in England were syntactically equivalent to everyone in England, it would distribute freely as an internal argument. It does not.
- (41)
- a.
- Everywhere in England loves fish and chips.
- b.
- # I met everywhere in England at the conference. *impersonal
- c.
- I met everyone in England at the conference. control
Everyone is acceptable in the same frame (41c), so the degradation is specific to everywhere. The restriction is not Case-theoretic. The impersonal reading is available on an accusative argument, provided that argument is the subject of a predication: Sluckin & Kastner (forthcoming) report it in ECM and small-clause environments, where the bound pronoun is impersonal-human their, not institutional its (§2.5).
- (42)
- Thanks to constant news coverage, the Queen watched everywhere in the UK enjoying (their) street parties for the Jubilee celebrations.(ex. 22, Sluckin & Kastner forthcoming)
The ECM subject in (42) and the direct object in (41b) are both accusative, yet only the former, the subject of a predication, supports the impersonal reading. The conditioning factor is predicational subjecthood, not Case. We capture this with the Subject-of-Predication feature [uδSoP] that Sluckin (2021; 2024) locates on C, extending the δ-feature system of Miyagawa (2017): [uδSoP] is valued under local Agree by the goal in Spec,TP. The impersonal interpretation is the SoP-licensed reading, so it is available only where everywhere itself occupies Spec,TP and values this feature.24
The object restriction then follows from the probing analysis of §5.6.2. When everywhere is merged as an internal argument below a higher argument, T’s D-probe targets that higher argument first; by minimality everywhere cannot raise to Spec,TP, [uδSoP] on C is left unvalued by it, and the impersonal derivation is unavailable. The silent
is recoverable only in this configuration, so in non-subject positions everywhere retains only its locative and institutional construals. ECM and small-clause subjects pattern with matrix subjects because there everywhere is itself the highest argument of the embedded predication; its accusative Case is irrelevant to [uδSoP].
The mechanism makes a converse prediction. Where no argument intervenes, as in passives and unaccusatives, the internal argument can itself raise to Spec,TP and value [uδSoP], so the impersonal reading should be available. This is borne out for characterising predicates, while episodic counterparts are degraded (43).
- (43)
- a.
- Everywhere in Spain is/gets paid well for their work, despite taking a two-hour nap in the afternoon. characterising
- b.
- # Everywhere in Spain was/got paid well yesterday for their work, despite taking a two-hour nap in the afternoon. episodic
Informal consultation (the author and four further speakers) found this contrast for most speakers, with some inter-speaker variation and one speaker accepting episodics more readily under get-than be-passives. Such variation is consistent with the characterising effect being a graded preference rather than a categorical ban (§5.4). The degraded passives reported by Sluckin & Kastner (forthcoming) are plausibly episodic; a controlled generic/episodic comparison is left for future work.
On this interpretation, impersonal everywhere falls within the broader class of dedicated human impersonals such as man in German and Scandinavian, French on, and Italian si; Egerland (2003) analyses the Scandinavian and Romance members as lexically underspecified for person, number and gender, carrying [+human] as their only inherent content. All supply the subject argument of the predication, though not necessarily as a DP in Spec,TP: Cinque (1988) places impersonal si in Infl, a clitic coindexed with a pleonastic pro in subject position, yet it is si that receives the external argument role, withheld at the inflectional head before it can reach that position. Impersonal everywhere differs from these in internal composition, building its human argument from a locative QP and a silent relational nominal, while sharing that defining requirement.25
One issue remains open within the synchronic account: the precise pragmatic conditions on episodic readings across the somewhere/nowhere/anywhere paradigm.
6 Predictions and open questions
Three further issues require comment: coordination, restrictor reordering, and the status of relative-clause restrictors.
6.1 Coordination with everyone
The analysis leads us to expect that impersonal [Q+where] and [Q+one] pronouns coordinate. Both denote human properties bound by Q0, and coordination requires that both conjuncts match in semantic type (⟨e,t⟩). The grammatical example in (44) supports this prediction (author’s own judgement).
- (44)
- Everyone in London and everywhere in the rest of the UK eats fish and chips.
A property-of-places analysis that obtains the impersonal reading by metonymic coercion does not naturally capture (44): it requires asymmetric type-shifting on one conjunct. This and similar coordinations require experimental testing to strengthen the argument.
6.2 Restrictor reordering
The PP restrictor appears in three positions:
- (45)
- a.
- Everywhere in the UK loves Marmite. base order
- b.
- In the UK, everywhere loves Marmite. fronted PP
- c.
- Everywhere loves Marmite in the UK. right-dislocated PP
The base order in (45a) corresponds directly to the nominal structure proposed in §5. The fronted variant (45b) preposes the PP to the left periphery over the DP and locative shell; the precise escape path remains open, as does the extraposition mechanism discussed below.26
The rightward-displaced variant (45c) surfaces with the PP separated from its selecting head. Linearising to the right of the object requires the PP to occupy a position structurally external to the subject DP at PF. We propose an extraposition-from-DP analysis: the PP merges in Spec,nPcat to value [uPlace], satisfying the local selectional requirement, and is subsequently displaced from that position by whatever extraposition mechanism the grammar makes available. The analysis is neutral between a rightward-adjunction analysis (since Rosenbaum 1967; Guéron 1980), a remnant-leftward-movement implementation (Fox & Nissenbaum 1999; Reeve & Hicks 2017), and a purely PF-side linearisation analysis compatible with the Linear Correspondence Axiom (Kayne 1994). A reviewer suggests deriving (45c) from (45b) by remnant movement, on the model of was für split (Abels 2003; Leu 2008). The analysis specifies the PP’s external merge site, leaving the direction of subsequent displacement open.
While extraposition of a selected specifier out of a DP is structurally more restricted than the extraposition of adjuncts or relative clauses (Fox & Nissenbaum 1999), PP extraposition from subject DPs is a robust operation in English (Guéron 1980). Recent theoretical accounts argue for the availability of direct rightward movement for such dependencies (Reeve & Hicks 2017; Bruening 2018). The precise constraints on moving a valued specifier are left open here.
We reject externally merging the PP as a clausal adjunct with no narrow-syntactic relation to ncat. This would require positing a null PP inside the subject DP to saturate the relational variable and discharge [uPlace]. Such a proposal weakens [uPlace] as a strict c-selectional feature and introduces a silent element with no independent motivation. The extraposition analysis preserves the c-selectional relation established in §5 regardless of which mechanism implements the displacement.
6.3 Relative clauses and locative domains
Relative-clause restrictors belong to the broader question of how the locative domain is realised. The core impersonal construction remains the overt-PP construction: [uPlace] on relational ncat is valued by a locative PP in Spec,nPcat. Relative clauses are considered here because Sluckin & Kastner (forthcoming) note that [Q + where] forms are not categorically impossible with that-relatives. The empirical pattern is split: who-relatives are sharply degraded, whereas that-relatives are marginal, context-sensitive, and untested experimentally.
The clearest contrast concerns who. Ordinary human quantifiers readily take who-relatives, but [Q + where] forms do not.
- (46)
- a.
- Everyone who knows Mary loves her.
- b.
- *Everywhere who knows Mary loves her.
This follows from the present analysis. Who requires a morphosyntactically accessible human nominal head. In the impersonal [Q+where] construction, the human variable is introduced inside the lower
–ncat complex and abstracted over by nea, but the overt shell visible to the relative system is still the where-projection. Its external morphosyntax is locative, not personal. The human component is therefore available for interpretation and binding, but not as the local nominal head required to license a who-relative.
The contrast with that is different. Since that is not specified as a human relativiser, it does not create the same immediate feature clash. Accordingly, Sluckin & Kastner (forthcoming) report marginal that-relative cases, such as (47).
- (47)
- a.
- Everyone that’s heard of Mother Teresa loves her.
- b.
- ?Everywhere that’s heard of Mother Teresa loves her.
These examples suggest that relative clauses after [Q + where] cannot be ruled out across the board. The relativiser that avoids the feature clash created by who, since it does not require an overt human nominal head. But this only removes one obstacle: it does not mean that the relative CP itself supplies the locative argument required by ncat, and it does not show that CP can replace the PP in the core syntax.
The attested Ted Lasso example in (48) is also not a clean diagnostic for the impersonal construction. It involves existential somewhere, not universal everywhere, and occurs as a fragment answer to a question resuming a directional goal.
- (48)
- Question: Where is she going?
- Answer: Somewhere that believes they deserve her.
The main analysis of such marginal cases is contextual or Brody-style locative-domain recovery. On Brody’s view, locative expressions can make an implicit people argument available: a place can support reference to the people associated with it, without the locative expression itself being syntactically converted into a human nominal. The Ted Lasso example can therefore be understood as a locative use of somewhere in which the relative clause makes this associated human variable salient. It does not show a relative CP valuing the c-selectional [uPlace] feature on ncat. The human reading remains available through Brody-style recovery, without the nominal structure of the impersonal construction.
The limits of that-relatives become clearer when the relative clause must support a fully individual-denoting impersonal subject in a matrix predicate (Sluckin & Kastner forthcoming).
- (49)
- a.
- Everyone that has a house looks after/loves their garden.
- b.
- Everywhere in England looks after/loves their garden.
- c.
- *Everywhere that has a house looks after/loves their garden.
A locative PP restrictor licenses the impersonal reading (49b), whereas a bare that-relative does not (49c). The CP relative in (49c) cannot provide the locative place argument required by the lower relational structure. This also explains why (47b) is marginal rather than ungrammatical. Nothing forces a human subject, since her refers to Mother Teresa and not the subject; thus, the sentence is available on the locative reading with Brody-style recovery, without requiring an inhabitant relation. In (49c), by contrast, bound their requires a human subject, the impersonal derivation is the only route to one, and the bare relative cannot supply its place argument. Combining a PP restrictor with the relative clause improves matters, as (50) shows, though the result remains degraded relative to the PP alone. This judgement is introspective and needs testing.
- (50)
- ?Everywhere in England that has a house looks after their garden.
Other apparent that-relative cases are likewise compatible with independently available locative or institutional readings. A rare attested example from online discourse such as (51) is too ambiguous between place → institution metonymy and quantification over inhabitants to decide the issue.
- (51)
- Everywhere that watches Netflix is part of the US cultural empire.27
Similarly, that-relatives are easier when the sentence can be interpreted over places, countries, institutions, or their representatives, as in (52).
- (52)
- Everywhere that has a royal family loves its/their queen.
Such examples do not show that that-relatives freely license the impersonal structure; they show only that relative clauses combine more easily with independently available locative or institutional uses of everywhere. Institutional readings matter here only as a source of ambiguity in particular examples. The main point is that any human interpretation in these cases can arise through contextual locative-domain recovery, rather than through CP saturation of [uPlace].
A related observation comes from embedded-question contexts tested by Sluckin & Kastner (forthcoming), where bare where appears with similar impersonal subject behaviour.
- (53)
- a.
- I knew about the first one, but I don’t know where in the world eats bird’s nest soup.
- b.
- ?I knew about the first one, but I don’t know where eats bird’s nest soup.
Although this is an embedded interrogative rather than a relative-clause restrictor, it suggests that where can appear in a CP environment while retaining an impersonal-like construal. The example is suggestive, not decisive, but it makes a P/CP-mediated analysis of marginal relative-clause cases less ad hoc.
If future studies show that that-relatives can regularly license impersonal [Q+where] readings in particular environments, the most conservative extension would be a null locative P-shell, schematically [PP P∅ [CP that …]], in line with existing missing-P, PP-like free-relative, and P/C-overlap analyses (Emonds 1985; Larson 1987; Dubinsky & Williams 1995; Grosu 1996; Caponigro & Pearl 2008). This would not be part of the core derivation, as the CP itself could not value [uPlace], while the silent P-shell would preserve PP saturation in Spec,nPcat. The null P-shell is therefore a possible extension, should future data require it, outside the core analysis of impersonal [Q+where], which retains an overt PP.
R-pronoun systems in Dutch and German show the reverse of the restriction argued for here. Noonan (2017) analyses them as decomposed functional sequences restricted to [–human] reference, with an ordinary pronoun or demonstrative required for human reference. Their restriction involves anaphoric reference to an antecedent, whereas impersonal everywhere binds a variable introduced inside its own restrictor. Since the dependencies differ, the patterns need not align; a unified account of these opposing human-reference restrictions remains outstanding.
7 Conclusion
This paper has argued that English [Q+where] expressions support three distinct quantified referent types. Locative everywhere is built over silent
inside a P-headed nominal projection. Institutional everywhere retains that locative structure, with the institutional interpretation arising through contextual predicate transfer in the sense of Nunberg (1995). Impersonal everywhere, by contrast, contains silent sortal
under a relational categoriser ncat bearing [uPlace]. In the canonical overt-PP construction, [uPlace] is valued by a macro-geographic PP in Spec,nPcat, which saturates the place argument of an inhabitant relation; the higher categoriser nea abstracts over the human individual variable, which gives the restrictor for Q0.
The account preserves the external locative morphology of where while locating the human contribution in the lower nominal domain. Relationality is contributed by functional structure rather than by the root, in line with proposals that obligatory dependents of relational nouns are introduced structurally (Adger 2013). The construction also extends the Kaynian silent-noun programme (Kayne 2005; Corver 2008; Kayne 2015): English does not need a dedicated overt impersonal pronoun here because the relevant human variable is introduced inside the nominal structure of a where-based quantifier. Brody’s locative-licensed people mechanism is therefore not abandoned, but syntactified inside the extended nominal projection.
Several issues remain open. Relative-clause restrictors and the pragmatic conditions on episodic readings across the somewhere/nowhere/anywhere paradigm await systematic testing. The subject restriction itself, however, we have analysed as a Subject-of-Predication effect (§5.6.3), situating impersonal everywhere among subject-oriented human impersonals. While the syntax fixes the presence and low position of the PP, the macro-geographic restriction is best treated as an interface condition on the inhabitant relation: its domain must support generic quantification over a distributed population.
The broader result is a compositional account of the English [Q+where] paradigm. Place, institution, and human-inhabitant readings share overt where-morphology, but they do not share the same lower nominal structure.
Notes
- In response to the concern of an anonymous reviewer, we note that Distributed Morphology and a feature-driven syntax are not competing commitments; DM is a theory of the syntax–morphology interface and presupposes a generally minimalist feature-based syntax assuming valuation/checking (see Embick 2010; Arregi & Nevins 2012). The frameworks drawn on here differ in whether morphology applies early or late. [^]
- For the diagnostic logic, Sluckin & Kastner (forthcoming) cite Koenig & Mauner (1999) on pronominal antecedence, Williams (1987), Landau (2010a), and Bhatt & Pancheva (2017) on anaphor binding by implicit arguments. [^]
- An anonymous reviewer raises the possibility that locative P is itself the exponent of the place requirement borne by ncat, so that the two would be reflexes of a single specification; this would however require a dependency across the intervening D layer, which the local Spec-head configuration adopted below does not need. We leave the question open. [^]
- An alternative tradition, originating with Kratzer & Shimoyama (2002) for Japanese indeterminate pronouns and extended to wh-fronting languages by Cable (2010), treats wh-words as denoting sets of alternatives over which higher operators (universal, existential, interrogative) quantify. Both approaches place quantificational force on a higher head; we adopt the variable-and-binding treatment for compatibility with the lambda calculus used elsewhere in the analysis. [^]
- Neither assumption is uncontroversial: Szabolcsi (2010: 189–205) surveys work on which every is not itself the universal quantifier, and on which quantifier words decompose further. Our account does not depend on every being a single morpheme. The operation in §3.4 requires only that Q0 and the complex P head be adjacent at PF, and structure inside Q does not affect that. [^]
- We thank an anonymous reviewer for suggesting Local Dislocation in place of Lowering. [^]
- The members of this paradigm are structurally deficient in the sense of Cardinaletti & Starke (1999). We state the generalisation without adopting a particular implementation of it. [^]
- Landau (2010b) argues for a syntactic link between spatial PPs and human experiencer arguments, on which experiencers are introduced under a locative P-shell. The present analysis exhibits a structurally parallel dependency in the nominal domain. [^]
- Following Sluckin & Kastner (forthcoming), we use the descriptive term PP restrictor for the overt locative PP. Formally, however, the PP is not itself the restrictor of Q0: it saturates the place argument introduced by relational ncat, and the resulting human property is abstracted over by nea to form the restrictor for Q0. [^]
- Brody (2013: 40) also touches on the comparison with Kayne’s silent-nominal framework but leaves it as a suggestive parallel rather than a structural derivation. [^]
- Kayne (2005) does not assume null person for universal indefinite pronouns such as everyone, everybody; these are overt realisations of the root. [^]
- The c-selectional feature might be decomposable into [uP] and [uLoc], in line with Adger (2014). Locative pro-forms (here, there) potentially satisfy both subfeatures under the Kayne-style PP analysis adopted here (Kayne 2005); their exclusion follows from the encyclopedic restriction, not from c-selection. [^]
- The place-sorted option raises a further question for locative everywhere, namely whether relational ncat is ever selected with
in a where-form. Nothing in what follows turns on it, and we set the question aside here.
[^]
- An anonymous reviewer asks what would go wrong if [uPlace] were borne by ncat in all three readings rather than only in the impersonal derivation. Empirically, the predictions are the same either way, since the feature would be satisfied internally wherever the root is itself place-denoting. Generalising it does, however, require that the place requirement be satisfiable by the root, and hence that roots bear sortal features visible to selection. That is a substantive commitment about roots which we cannot defend here, and we therefore retain the more conservative statement. [^]
- An alternative within contemporary DM severs argument introduction from categorising heads entirely, entrusting it to dedicated low functional argument introducers, e.g. i∗ (see Wood & Marantz 2017). Our proposals are compatible with such a model; we maintain a feature-driven ncat analysis for representational economy. [^]
- Equivalent syntactic implementations include an Adgerian c-selectional feature on the categorising n above the noun (Adger 2003; 2013) or a Relator Phrase mediating the predicative relation (den Dikken 2006; not adopted here). We remain neutral among these. [^]
- A reviewer points us to nanosyntactic work on ontological categories, in which person and place are adjacent in a universal sequence of nominal heads (see Baunaz & Lander 2018); we do not adopt that framework here. [^]
- Reprinted as Chapter 5 of Kayne (2010); the examples cited are (21)–(22) there, at p. 86. Kayne observes that this approximative use is itself restricted to some-, being degraded with every- and no- (p. 90, n. 13). We take that restriction to follow from the semantics of approximative quantity rather than from any general ban on non-locative roots in this position. [^]
- The construction studied here is anchored in British English, where -place forms are in any case not idiomatic, being chiefly American (see Kayne 2010: 86, n. 7). Everyplace appears to be degraded relative to everywhere in this variety even on the institutional reading, where no bound pronoun is at issue, so the contrast is confounded by variety rather than diagnostic of structure. These judgements are the author’s own together with those of British English consultants. [^]
- Haspelmath excludes universal quantifiers from indefinite pronouns proper, on the functional ground that universally quantified nominals are semantically definite (Haspelmath 1997: 12). The generalisation invoked here is formal rather than functional, and he extends it to the universals himself, deriving every-type quantifiers from interrogative bases in five languages (Haspelmath 1997: 156). [^]
- Author’s own judgement, supported by 4 further consultants. [^]
- Direct evidence for this prediction is currently limited: Sluckin & Kastner (forthcoming) attest anywhere only in characterising contexts, and systematic testing of episodic predicates under negation, as in (37a), remains outstanding. [^]
- For related treatments of agreement with deficient or underspecified goals, see Preminger (2014). [^]
- The feature restates the subject restriction in independently motivated terms: Subject of Predication is treated as a syntactically definable primitive in Cardinaletti (2004); Rizzi (2005); Mohr (2005); Rizzi (2018) and developed for English and Romance subject positions in Sluckin (2021); Sluckin et al. (2021), independently of DP-status, Case and agreement. Egerland (2003) also shows that Scandinavian man, on and si cannot be syntactic objects. We concede that Subject of Predication is not a primitive of standard minimalist theory; readers who reject its syntactic status will read the feature as a restatement. [^]
- This Subject-of-Predication licensing is distinct from the covert-perceiver analysis of locative inversion in Sluckin (2024), where the preposed locative is itself the subject-like element; both motivate a decomposition of English subjecthood but involve different silent elements. [^]
- Sluckin & Kastner (forthcoming) did not test reordering experimentally but remark on its acceptability, confirmed by the author’s own native-speaker judgement. [^]
- Attested example from a Hacker News discussion thread: https://news.ycombinator.com/item?id=43278694 (accessed May 27, 2026). [^]
Abbreviations
imp.pn = impersonal pronoun, prt = particle.
Ethics and consent
This theoretical study did not require formal ethics board approval. Consulted judgements are anonymised.
Funding information
The author has no funding to declare.
Acknowledgements
I thank Itamar Kastner, my collaborator on the empirical counterpart paper, for valuable discussions. I also thank Eric Fuß, Florian Schäfer, Fabienne Martin and Artemis Alexiadou, who have all given me valuable feedback or suggestions at different stages of this work.
Competing interests
The author has no competing interests to declare.
References
Abels, Klaus. 2003. Successive cyclicity, anti-locality, and adposition stranding. Storrs, CT: University of Connecticut dissertation.
Abney, Steven P. 1987. The English noun phrase in its sentential aspect. Cambridge, MA: Massachusetts Institute of Technology dissertation.
Adger, David. 2003. Core syntax: A minimalist approach. Oxford University Press. DOI: http://doi.org/10.1093/oso/9780199243709.001.0001
Adger, David. 2013. A syntax of substance (Linguistic Inquiry Monographs 64). Cambridge, MA: MIT Press. DOI: http://doi.org/10.7551/mitpress/9780262018616.001.0001
Adger, David. 2014. Variability and grammatical architecture. In Picallo, M. Carme (ed.), Linguistic variation in the minimalist framework, 179–196. Oxford: Oxford University Press. DOI: http://doi.org/10.1093/acprof:oso/9780198702894.003.0009
Alexiadou, Artemis & Anagnostopoulou, Elena & Schäfer, Florian. 2015. External arguments in transitivity alternations: A layering approach. Oxford: Oxford University Press. DOI: http://doi.org/10.1093/acprof:oso/9780199571949.001.0001
Alexiadou, Artemis & Carvalho, Janayna. 2017. The role of locatives in (partial) pro-drop languages. In Sheehan, Michelle & Bailey, Laura R. (eds.), Order and structure in syntax II: Subjecthood and argument structure (Open Generative Syntax 2), 41–67. Berlin: Language Science Press. DOI: http://doi.org/10.5281/zenodo.1116755
Alexiadou, Artemis & Schäfer, Florian. 2011. An unaccusativity diagnostic at the syntax-semantics interface: There-insertion, indefinites and restitutive again. In Reich, Ingo & Horch, Eva & Pauly, Dennis (eds.), Proceedings of Sinn und Bedeutung 15, 101–116. Saarbrücken: Universaar – Saarland University Press.
Arregi, Karlos & Nevins, Andrew. 2012. Morphotactics: Basque auxiliaries and the structure of spellout (Studies in Natural Language and Linguistic Theory 86). Dordrecht: Springer. DOI: http://doi.org/10.1007/978-94-007-3889-8
Asher, Nicholas. 2011. Lexical meaning in context: A web of words. Cambridge: Cambridge University Press. DOI: http://doi.org/10.1017/CBO9780511793936
Baker, Mark. 1985a. The mirror principle and morphosyntactic explanation. Linguistic Inquiry 16(3). 373–415.
Baker, Mark C. 1985b. Incorporation: A theory of grammatical function changing. Cambridge, MA: Massachusetts Institute of Technology dissertation.
Baker, Mark C. 1988. Incorporation: A theory of grammatical function changing. Chicago: University of Chicago Press.
Baunaz, Lena & Lander, Eric. 2018. The internal structure of ontological categories. In Caha, Pavel & de Clercq, Karen & Vanden Wyngaerd, Guido (eds.), The unpublished manuscript: A collection of lingbuzz papers to celebrate Michal Starke’s 50th birthday, 1–18. https://lingbuzz.net/lingbuzz/003993.
Bhatt, Rajesh & Pancheva, Roumyana. 2006. Implicit arguments. In The Blackwell companion to syntax, vol. 2, 558–588. Blackwell. DOI: http://doi.org/10.1002/9780470996591.ch34
Bhatt, Rajesh & Pancheva, Roumyana. 2017. Implicit arguments. In Everaert, Martin & van Riemsdijk, Henk C. (eds.), The Wiley Blackwell companion to syntax, second edition. Blackwell. DOI: http://doi.org/10.1002/9781118358733.wbsyncom118
Borer, Hagit. 2005a. Structuring sense, vol. 1: In name only. Oxford: Oxford University Press. DOI: http://doi.org/10.1093/acprof:oso/9780199263905.001.0001
Borer, Hagit. 2005b. Structuring sense, vol. 2: The normal course of events. Oxford: Oxford University Press. DOI: http://doi.org/10.1093/acprof:oso/9780199263929.001.0001
Borschev, Vladimir & Partee, Barbara H. 2001. Genitive modifiers, sorts, and metonymy. Nordic Journal of Linguistics 24(2). 140–160. DOI: http://doi.org/10.1080/033258601753358597
Bresnan, Joan. 1994. Locative inversion and the architecture of universal grammar. Language 70(1). 72–131. DOI: http://doi.org/10.2307/416741
Brody, Michael. 2013. Silent people: The pseudo-impersonal. In Brandtler, Johan & Molnár, Valéria & Platzack, Christer (eds.), Approaches to Hungarian: Papers from the 2011 Lund conference (Approaches to Hungarian 13), 33–44. Amsterdam/Philadelphia: John Benjamins. DOI: http://doi.org/10.1075/atoh.13.03bro
Bruening, Benjamin. 2013. By phrases in passives and nominals. Syntax 16(1). 1–41. DOI: http://doi.org/10.1111/j.1467-9612.2012.00171.x
Bruening, Benjamin. 2018. CPs move rightward, not leftward. Syntax 21(4). 362–401. DOI: http://doi.org/10.1111/synt.12164
Cable, Seth. 2010. The grammar of Q: Q-particles, wh-movement and pied-piping. Oxford University Press. DOI: http://doi.org/10.1093/acprof:oso/9780195392265.001.0001
Caponigro, Ivano & Pearl, Lisa. 2008. Silent prepositions: Evidence from free relatives. In Asbury, Anna & Dotlačil, Jakub & Gehrke, Berit & Nouwen, Rick (eds.), Syntax and semantics of spatial P, 365–385. Amsterdam: John Benjamins. DOI: http://doi.org/10.1075/la.120.18cap
Cardinaletti, Anna. 2004. Toward a cartography of subject positions. In Rizzi, Luigi (ed.), The structure of CP and IP: The cartography of syntactic structures, vol. 2 (Oxford Studies in Comparative Syntax), 115–165. Oxford: Oxford University Press. DOI: http://doi.org/10.1093/oso/9780195159486.003.0005
Cardinaletti, Anna & Starke, Michal. 1999. The typology of structural deficiency: A case study of three classes of pronouns. In van Riemsdijk, Henk (ed.), Clitics in the languages of Europe, 145–234. Berlin: Mouton de Gruyter. DOI: http://doi.org/10.1515/9783110804010.145
Chomsky, Noam. 1995. The minimalist program. Cambridge, MA: MIT Press.
Chomsky, Noam. 2001. Derivation by phase. In Kenstowicz, Michael (ed.), Ken Hale: A life in language, 1–52. Cambridge, MA: MIT Press. DOI: http://doi.org/10.7551/mitpress/4056.003.0004
Cinque, Guglielmo. 1988. On si constructions and the theory of arb. Linguistic Inquiry 19(4). 521–581.
Corver, Norbert. 2008. On silent semi-lexical Person. Lingue e Linguaggio 7(1). 5–24. DOI: http://doi.org/10.1418/26707
Cruse, David Allen. 1986. Lexical semantics. Cambridge: Cambridge University Press. DOI: http://doi.org/10.1017/CBO9780511753467
D’Alessandro, Roberta. 2007. Impersonal si constructions: Agreement and interpretation (Studies in Generative Grammar). Berlin: Mouton de Gruyter. DOI: http://doi.org/10.1515/9783110207514
Dayal, Veneeta. 1998. Any as inherently modal. Linguistics and Philosophy 21(5). 433–476. DOI: http://doi.org/10.1023/a:1005494000753
Déchaine, Rose-Marie & Wiltschko, Martina. 2002. Decomposing pronouns. Linguistic Inquiry 33(3). 409–442. DOI: http://doi.org/10.1162/002438902760168554
den Dikken, Marcel. 2006. Relators and linkers: The syntax of predication, predicate inversion, and copulas. MIT Press. DOI: http://doi.org/10.7551/mitpress/5873.001.0001
Dubinsky, Stanley & Williams, Kemp. 1995. Recategorization of prepositions as complementizers: The case of temporal prepositions in English. Linguistic Inquiry 26(1). 125–137.
Egerland, Verner. 2003. Impersonal pronouns in Scandinavian and Romance. Working Papers in Scandinavian Syntax 71. 75–102.
Embick, David. 2010. Localism versus globalism in morphology and phonology (Linguistic Inquiry Monographs 60). Cambridge, MA: MIT Press. DOI: http://doi.org/10.7551/mitpress/9780262014229.001.0001
Embick, David. 2015. The morpheme: A theoretical introduction. Mouton de Gruyter. DOI: http://doi.org/10.1515/9781501502569
Embick, David & Marantz, Alec. 2008. Architecture and blocking. Linguistic Inquiry 39(1). 1–53. DOI: http://doi.org/10.1162/ling.2008.39.1.1
Embick, David & Noyer, Rolf. 2001. Movement operations after syntax. Linguistic Inquiry 32(4). 555–595. DOI: http://doi.org/10.1162/002438901753373005
Emonds, Joseph E. 1985. A unified theory of syntactic categories. Dordrecht: Foris. DOI: http://doi.org/10.1515/9783110808513
Fox, Danny & Nissenbaum, Jon. 1999. Extraposition and scope: A case for overt QR. In Bird, Sonya & Carnie, Andrew & Haugen, Jason D. & Norquest, Peter (eds.), Proceedings of the 18th West Coast Conference on Formal Linguistics (WCCFL 18), 132–144. Somerville, MA: Cascadilla Press.
Giannakidou, Anastasia. 1998. Polarity sensitivity as (non)veridical dependency (Linguistik Aktuell/Linguistics Today 23). Amsterdam/Philadelphia: John Benjamins. DOI: http://doi.org/10.1075/la.23
Grimshaw, Jane. 2005. Words and structure. Stanford, CA: CSLI.
Grosu, Alexander. 1996. Remarks and replies: The proper analysis of “Missing-P” free relative constructions. Linguistic Inquiry 27(2). 257–293.
Guéron, Jacqueline. 1980. On the syntax and semantics of PP extraposition. Linguistic Inquiry 11(4). 637–678.
Halle, Morris & Marantz, Alec. 1993. Distributed Morphology and the pieces of inflection. In Hale, Ken & Keyser, Samuel Jay (eds.), The view from Building 20: Essays in linguistics in honor of Sylvain Bromberger, 111–176. Cambridge, MA: MIT Press.
Haspelmath, Martin. 1997. Indefinite pronouns (Oxford Studies in Typology and Linguistic Theory). Oxford: Clarendon Press.
Heim, Irene. 1982. The semantics of definite and indefinite noun phrases. Amherst, MA: University of Massachusetts dissertation.
Heim, Irene & Kratzer, Angelika. 1998. Semantics in generative grammar. Oxford: Blackwell.
Huddleston, Rodney & Pullum, Geoffrey K. 2002. The Cambridge grammar of the English language. Cambridge: Cambridge University Press. DOI: http://doi.org/10.1017/9781316423530
Kamp, Hans. 1981. A theory of truth and semantic representation. In Groenendijk, Jeroen & Janssen, Theo & Stokhof, Martin (eds.), Formal methods in the study of language, 277–322. Amsterdam: Mathematical Centre.
Kayne, Richard S. 1994. The antisymmetry of syntax (Linguistic Inquiry Monographs 25). Cambridge, MA: MIT Press.
Kayne, Richard S. 2005. Here and there. In Movement and silence. Oxford University Press. DOI: http://doi.org/10.1093/acprof:oso/9780195179163.003.0004
Kayne, Richard S. 2007. A short note on where vs. place. In Maschi, Roberta & Penello, Nicoletta & Rizzolatti, Piera (eds.), Miscellanea di studi linguistici offerti a Laura Vanelli da amici e allievi padovani, 245–257. Udine: Forum.
Kayne, Richard S. 2010. Comparisons and contrasts. Oxford: Oxford University Press.
Kayne, Richard S. 2015. English one and ones as complex determiners. In Gallego, Ángel J. & Ott, Dennis (eds.), 50 years later: Reflections on Chomsky’s Aspects. MIT Working Papers in Linguistics. Available at LingBuzz: lingbuzz/002542.
Kiparsky, Paul. 1973. “Elsewhere” in phonology. In Anderson, Stephen R. & Kiparsky, Paul (eds.), A festschrift for Morris Halle, 93–106. Holt, Rinehart, and Winston.
Koenig, Jean-Pierre & Mauner, Gail. 1999. A-definites and the discourse status of implicit arguments. Journal of Semantics 16(3). 207–236. DOI: http://doi.org/10.1093/jos/16.3.207
Kratzer, Angelika. 2021. Situations in natural language semantics. In Zalta, Edward N. (ed.), The Stanford Encyclopedia of Philosophy. Stanford, CA: Stanford Encyclopedia of Philosophy. https://plato.stanford.edu/entries/situations-semantics/. Winter 2021 edn.
Kratzer, Angelika & Shimoyama, Junko. 2002. Indeterminate pronouns: The view from Japanese. In Otsu, Yukio (ed.), The proceedings of the Third Tokyo Conference on Psycholinguistics, vol. 3, 1–25. Tokyo: Hituzi Shyobo.
Krifka, Manfred & Pelletier, Francis Jeffry & Carlson, Gregory N. & ter Meulen, Alice & Chierchia, Gennaro & Link, Godehard. 1995. Genericity: An introduction. In Carlson, Gregory N. & Pelletier, Francis Jeffry (eds.), The generic book, 1–124. Chicago: University of Chicago Press.
Landau, Idan. 2010a. The explicit syntax of implicit arguments. Linguistic Inquiry 41(3). 357–388. DOI: http://doi.org/10.1162/ling_a_00001
Landau, Idan. 2010b. The locative syntax of experiencers (Linguistic Inquiry Monographs 53). Cambridge, MA: MIT Press. DOI: http://doi.org/10.7551/mitpress/8387.001.0001
Larson, Richard K. 1987. “Missing Prepositions” and the analysis of English free relative clauses. Linguistic Inquiry 18(2). 239–266.
Larson, Richard K. 1988. On the double object construction. Linguistic Inquiry 19(3). 335–391.
Larson, Richard K. & Marušič, Franc. 2004. On indefinite pronoun structures with APs: Reply to Kishimoto. Linguistic Inquiry 35(2). 268–287. DOI: http://doi.org/10.1162/002438904323019075
Leu, Thomas. 2008. What for internally. Syntax 11(1). 1–25. DOI: http://doi.org/10.1111/j.1467-9612.2008.00113.x
Leu, Thomas. 2015. The architecture of determiners. New York: Oxford University Press. DOI: http://doi.org/10.1093/acprof:oso/9780199945238.001.0001
Leu, Thomas. 2025. Nǎlǐ nàlǐ — ça, c’est quoi? A comparative note on that and where. In Baunaz, Lena & Bocci, Giuliano & Nevins, Andrew (eds.), The ziggurat of grammar: In honor of Ur Shlonsky (Language Faculty and Beyond 20), 101–117. Amsterdam: John Benjamins. DOI: http://doi.org/10.1075/lfab.20.06leu
Littlemore, Jeannette. 2017. Metonymy. In Dancygier, Barbara (ed.), The Cambridge handbook of cognitive linguistics (Cambridge Handbooks in Language and Linguistics), 407–422. Cambridge: Cambridge University Press. DOI: http://doi.org/10.1017/9781316339732.026
Matthewson, Lisa. 2001. Quantification and the nature of crosslinguistic variation. Natural Language Semantics 9(2). 145–189. DOI: http://doi.org/10.1023/A:1012492911285
Matushansky, Ora. 2006. Head movement in linguistic theory. Linguistic Inquiry 37(1). 69–109. DOI: http://doi.org/10.1162/002438906775321184
Miyagawa, Shigeru. 2017. Agreement beyond phi. Cambridge, MA: MIT Press. DOI: http://doi.org/10.7551/mitpress/10958.001.0001
Mohr, Sabine. 2005. Clausal architecture and subject positions: Impersonal constructions in the Germanic languages (Linguistik Aktuell/Linguistics Today 88). Amsterdam: John Benjamins. DOI: http://doi.org/10.1075/la.88
Noonan, Máire B. 2017. Dutch and German R-pronouns and P-stranding: R you sure it’s P-stranding? In Newell, Heather & Noonan, Máire & Piggott, Glyne & Travis, Lisa deMena (eds.), The structure of words at the interfaces. Oxford: Oxford University Press. DOI: http://doi.org/10.1093/oso/9780198778264.003.0010
Nunberg, Geoffrey. 1995. Transfers of meaning. Journal of Semantics 12(2). 109–132. DOI: http://doi.org/10.1093/jos/12.2.109
Panagiotidis, Phoevos. 2002. Pronouns, clitics and empty nouns: ‘Pronominality’ and licensing in syntax (Linguistik Aktuell/Linguistics Today 46). Amsterdam: John Benjamins. DOI: http://doi.org/10.1075/la.46
Pesetsky, David. 1987. Wh-in-situ: Movement and unselective binding. In Reuland, Eric J. & ter Meulen, Alice G. B. (eds.), The representation of (in)definiteness (Current Studies in Linguistics 14), 98–129. Cambridge, MA: MIT Press.
Postal, Paul M. 1966. On so-called “pronouns” in English. In Dinneen, Francis P. (ed.), Report of the seventeenth annual round table meeting on linguistics and language studies, 177–206. Washington, D.C.: Georgetown University Press.
Preminger, Omer. 2014. Agreement and its failures (Linguistic Inquiry Monographs 68). MIT Press. DOI: http://doi.org/10.7551/mitpress/9780262027403.001.0001
Prince, Ellen F. 1981. On the inferencing of indefinite-this NPs. In Joshi, Aravind K. & Webber, Bonnie L. & Sag, Ivan A. (eds.), Elements of discourse understanding, 231–250. Cambridge: Cambridge University Press.
Pustejovsky, James. 1995. The generative lexicon. Cambridge, MA: MIT Press. DOI: http://doi.org/10.7551/mitpress/3225.001.0001
Pylkkänen, Liina. 2008. Introducing arguments (Linguistic Inquiry Monographs 49). Cambridge, MA: MIT Press.
Radden, Günter & Kövecses, Zoltán. 1999. Towards a theory of metonymy. In Panther, Klaus-Uwe & Radden, Günter (eds.), Metonymy in language and thought (Human Cognitive Processing 4), 17–59. Amsterdam: John Benjamins. DOI: http://doi.org/10.1075/hcp.4.03rad
Reeve, Matthew & Hicks, Glyn. 2017. Adjunct extraposition: Base-generation or movement? Syntax 20(3). 215–248. DOI: http://doi.org/10.1111/synt.12136
Rizzi, Luigi. 2005. On some properties of subjects and topics. In Brugè, Laura & Giusti, Giuliana & Munaro, Nicola & Schweikert, Walter & Turano, Giuseppina (eds.), Contributions to the XXX Incontro di Grammatica Generativa, 203–224. Venice: Cafoscarina.
Rizzi, Luigi. 2018. Subjects, topics and the interpretation of pro. In Petrosino, Roberto & Cerrone, Pietro & van der Hulst, Harry (eds.), From sounds to structures: Beyond the veil of Maya, 510–529. Berlin: Mouton de Gruyter. DOI: http://doi.org/10.1515/9781501506734-019
Roberts, Ian. 2010. Agreement and head movement: Clitics, incorporation, and defective goals. Cambridge, MA: MIT Press. DOI: http://doi.org/10.7551/mitpress/9780262014304.001.0001
Roberts, Ian. 2019. Parameter hierarchies and universal grammar. Oxford: Oxford University Press. DOI: http://doi.org/10.1093/oso/9780198804635.001.0001
Roeper, Thomas. 1987. Implicit arguments and the head-complement relation. Linguistic Inquiry 18(2). 267–310.
Rosenbaum, Peter. 1967. The grammar of English predicate complement constructions. Cambridge, MA: MIT Press.
Schäfer, Florian. 2008. The syntax of (anti-)causatives: External arguments in change-of-state contexts (Linguistik Aktuell/Linguistics Today 126). Amsterdam/Philadelphia: John Benjamins. DOI: http://doi.org/10.1075/la.126
Shlonsky, Ur. 1991. Quantifiers as functional heads: A study of quantifier float in Hebrew. Lingua 84(2–3). 159–180. DOI: http://doi.org/10.1016/0024-3841(91)90069-H
Sluckin, Benjamin L. 2021. Non-canonical subjects and subject positions: Locative inversion, V2-violations, and feature inheritance. Berlin: Humboldt-Universität zu Berlin, Sprach- und literaturwissenschaftliche Fakultät dissertation. DOI: http://doi.org/10.18452/23715
Sluckin, Benjamin L. 2024. The covert perceiver in English locative inversion: An alternative to expletive pro. Evolutionary Linguistic Theory 6(1–2). 192–243. DOI: http://doi.org/10.1075/elt.00058.slu
Sluckin, Benjamin L. & Cruschina, Silvio & Martin, Fabienne. 2021. Locative inversion in Germanic and Romance: A conspiracy theory. In Wolfe, Sam & Meklenborg, Christine (eds.), Continuity and variation in Germanic and Romance, 165–192. Oxford: Oxford University Press. DOI: http://doi.org/10.1093/oso/9780198841166.003.0007
Sluckin, Benjamin L. & Kastner, Itamar. forthcoming. ‘Everywhere here can say this’: The English locative impersonal. English Language and Linguistics. Preprint: http://ling.auf.net/lingbuzz/006708.
Szabolcsi, Anna. 2010. Quantification. Cambridge University Press. DOI: http://doi.org/10.1017/CBO9780511781681
von Fintel, Kai. 1994. Restrictions on quantifier domains. Amherst, MA: University of Massachusetts dissertation.
Williams, Alexander. 2015. Arguments in syntax and semantics. Cambridge: Cambridge University Press. DOI: http://doi.org/10.1017/CBO9781139042864
Williams, Edwin. 1981. Argument structure and morphology. The Linguistic Review 1(1). 81–114. DOI: http://doi.org/10.1515/tlir.1981.1.1.81
Williams, Edwin. 1987. Implicit arguments, the binding theory, and control. Natural Language and Linguistic Theory 5(2). 151–180. DOI: http://doi.org/10.1007/BF00166581
Wood, Jim & Marantz, Alec. 2017. The interpretation of external arguments. In D’Alessandro, Roberta & Franco, Irene & Gallego, Ángel (eds.), The verbal domain. Oxford: Oxford University Press. DOI: http://doi.org/10.1093/oso/9780198767886.003.0011





