3.1 The identifying function of (proper) noun modifiers

Noun modifiers are not noun phrases but nouns or nominals.Footnote ¹¹ They typically have classifying meaning, i.e. they designate a subtype of the head. As such, they are positioned in prehead position. A standard example would be this expensive [ cat food] where the noun modifier cat specifies the type of food. In contrast, determiner genitives help to specify the referent of the noun phrase. Rather than restricting the denotational scope of the head noun, they restrict the reference of the noun phrase, as in John's house, where John specifies whose house it is. As such, determiner genitives have identifying function, defined as ‘reference restriction’ (see above). Determiner genitives are regarded – from different theoretical angles – as ‘(referential) anchors’ (e.g. Löbner Reference Löbner1985, Reference Löbner and Botley1998, Reference Löbner2011; Rijkhoff Reference Rijkhoff, Connolly and Dik1989; Hawkins Reference Hawkins1991; Barker Reference Barker1995; Haspelmath Reference Haspelmath1999; Rosenbach & Vezzosi Reference Rosenbach, Vezzosi, Sornicola, Poppe and Shisha-Halevy2000; Koptjevskaja-Tamm Reference Koptjevskaja-Tamm, Haspelmath, König, Oesterreicher and Raible2001, Reference Koptjevskaja-Tamm2002) or ‘reference point entities’ (Langacker Reference Langacker, Taylor and MacLaury1995; Taylor Reference Taylor1996; Keizer Reference Keizer2007; Willemse Reference Willemse2007; Willemse et al. Reference Willemse, Davidse, Heyvaert and McGregor2009) that help to identify the referent of the noun phrase by reference to this anchor (or reference point). It is by reference to John that the hearer/addressee can identify whose house it is, i.e. John's.

In Rosenbach (Reference Rosenbach2007a, Reference Rosenbach2009, Reference Rosenbach, Traugott and Trousdale2010) I argue that some (proper) noun modifiers may share the identifying function of determiner genitives and can be considered referential anchors, at least in certain contexts. In a similar vein, Schlücker (Reference Schlücker2013) analyses identifying proper noun modifiers in GermanFootnote ¹² as ‘anchoring modifiers’ within a version of Rijkhoff's (Reference Rijkhoff2002, Reference Rijkhoff and McGregor2009) model of the English noun phrase adapted by Zifonun (Reference Zifonun2010) for German.Footnote ¹³ In the expression the FBI director the proper noun modifier FBI helps to restrict the reference of the noun phrase, specifying whose director it is, i.e. the FBI's, and in this respect functions like a corresponding determiner genitive (the FBI's director). As referential anchors, however, proper noun modifiers must themselves be referential. Theoretically, this is a problematic assumption, as a noun modifier – unlike a determiner genitive (or a corresponding of-genitive) – is not a noun phrase and should thus constitute an ‘anaphoric island’, in the sense of Postal (Reference Postal1969) or Sproat (Reference Sproat, Hammond and Noonan1988), and thus be non-referential.Footnote ¹⁴ However, proper nouns constitute a special case as they have unique reference and may retain their inherent referentiality as modifiers. This is demonstrated in (2) by Schlücker (Reference Schlücker2013: 466), applying the classic test for referentiality, i.e. anaphoric reference (see Ward et al. Reference Ward, Sproat and McKoon1991).

1. (2) (a) the search of the Wulff_i residence he_i was not informed of

   (b) the beginning of the Berlusconi_i trial where he_i however did not appear

Further evidence for the referentiality of proper noun modifiers comes from translations. Rosenbach (Reference Rosenbach, Traugott and Trousdale2010: 165) gives an example where the English proper noun modifiers in the Grieg Piano Concerto and the Sibelius Violin Concerto get translated in German by a prenominal genitive (Griegs Klavierkonzert) and postnominal von-construction (das Violinkonzert von Sibelius) respectively. That is, in both cases the translator chose a construction with a clearly referential possessor in German, although German has the structural means for a one-to-one translation (das Grieg-Klavierkonzert, das Sibelius-Violinkonzert). To illustrate this further, English often uses a (proper) noun modifier where Afrikaans uses a determiner genitive (Rosenbach Reference Rosenbach2017: 7), as in (3), where the Afrikaans se-genitive in Taalejare se dorpsdam in (3a), a prenominal determiner genitive, gets translated into English as the proper noun modifier the Yearsonend dam in (3b) (Taalejare/Yearsonend is the name of a village).

1. (3) (a) GrootKarel droom van ’n opgaardam vir daardie nat streek, en ’n

   “geut” – so noem hy dit – wat die water na Tallejare se dorpsdam

   bring. (Etienne van Heerden, Die swye van Mario Salviati, p. 43,

   emphasis mine)

   (b) Big Karel had dreamt of a collecting dam in that area and an aqueduct

   – as he called it – bringing the water to the Yearsonend dam. (Etienne van Heerden, The Long Silence of Mario Salviati, p. 38, emphasis mine)

And in the Afrikaans translation of the first Harry Potter novel we find the expression the Gryffindor common room translated either one-to-one as die Griffindor-geselskamer (a proper noun modifier/compound) or as Griffindor se geselskamer (a determiner genitive) throughout the novel. These examples illustrate that the translators considered constructions with determiner genitives and proper noun modifiers to be equivalent in meaning in these contexts.Footnote ¹⁵

This is not to say that the two constructions, the one with a determiner genitive and the one with a (proper) noun modifier, referentially anchor in the same way. In fact they don't. In English, determiner genitives take determiner position and are tied to the expression of definiteness within the possessive noun phrase, narrowing down the referent of the noun phrase to a (typically) uniquely identifiable referent.Footnote ¹⁶ In contrast, noun phrases with (proper) noun modifiers are determined by the initial determiner and as such are open to the expression of definiteness. In Rosenbach (Reference Rosenbach2009) I argue that the initial definite article works together with the (proper) noun modifier when referentially anchoring the NP (see also Koptjevskaja-Tamm & Rosenbach Reference Koptjevskaja-Tamm and Rosenbach2005: §4.2). Schlücker (Reference Schlücker2013) attributes more importance to the definite article in the process of identifying the referent of the noun phrase, assigning the identifying proper noun modifier only ‘assisting’ status.

Schlücker's analysis ties in well with Löbner's (Reference Löbner1985, Reference Löbner and Botley1998, Reference Löbner2011) theory of definites, where the definite article is taken to indicate that the head noun is to be interpreted as a functional noun that needs to be anchored. Under this view referential anchoring is only possible within definite noun phrases (see, however, sections 3.2 and 3.4 for a critical assessment of this view).

Breban (Reference Breban2018) argues that proper noun modifiers are not referential anchors but mostly epithets and that although they may have identifying function, they identify the referent in a different way than a determiner genitive does. According to Breban (Reference Breban2018: 391),

Analysing proper noun modifiers as epithets allows Breban (Reference Breban2018) to present a uniform and thus elegant analysis of (most) proper noun modifiers, as epithets may convey meanings compatible with both qualifying and identifying uses. Her main argument is that qualifiers can also identify. Thus in the red car the adjective red helps to identify the referent of the NP. This analysis has an instant appeal as it covers the wide range of relations that can be expressed by a proper noun modifier as well as their positional restriction – they precede classifying elements and the head in the noun phrase. Evidence for the adjectival character of proper noun modifiers comes from translations, in particular the fact that expressions such as the Brussels landscape get translated into Dutch as het Brusselse landschap (denominal adjective) or het landschap in Brussel (prepositional phrase) but not as *het Brussel landschap (proper noun modifier); see Breban (Reference Breban2018: 393). Breban argues that due to the absence of denominal adjectives for locations, English resorts to proper noun modifiers (the Brussels landscape). Note, incidentally, that English does have productive morphological processes for forming denominal adjectives from proper nouns, as e.g. in the Hallidayan framework. They do not necessarily block the formation of a corresponding proper-noun modifier (the Halliday framework), nor that of a determiner genitive (Halliday's framework). The fact that these proper noun modifiers may alternate in English with a corresponding adjectival form and the evidence from translations above indeed support Breban's epithet analysis. However, Breban needs to widen Halliday's (Reference Halliday1994) framework by allowing nouns into the epithet function, originally reserved for adjectives only. And the analysis of proper noun modifiers as epithets leaves us with the mystery of a referential nominal modifier in a position usually not allowing referential elements.

Proper noun modifiers that are internal arguments of a deverbal head noun, as in a Kerry supporter, form a special case in Breban's account, because they are usually treated as complements in the grammars of English (Payne & Huddleston Reference Payne, Huddleston, Huddleston and Pullum2002: 452–3), which are distinct from modifiers, so a modifier or epithet analysis cannot apply to them. Breban analyses them as part of ‘Thing’ in Halliday's model, i.e. as part of the head and thus structurally as a compound. The examples chosen by Breban (e.g. a Kerry supporter) clearly exemplify a deverbal head noun and internal relation. In practice, however, it is often difficult to clearly identify internal arguments.Footnote ¹⁷ For example, how do we classify head nouns like fan, which are not deverbal but still require an argument, as in example (4) below?

1. (4) I have become an Anita Shreve fan and plan to read more of her books. Her weave of characters, crime and passion …Footnote ¹⁸

Note that in this example the proper noun modifier Anita Shreve is referential as it is anaphorically referred to twice in the following context, which is a problem for a compound analysis. Even with clearly deverbal head nouns a unified compound analysis strikes me as problematic, as examplified by (5) below:

1. (5) The key point that blunts the Gould and Lewontin critique of adaptionism is that … (Pinker & Bloom 1990: 709, emphasis mine)

This cannot be a compound, not even within approaches allowing for phrasal compounds (e.g. Lieber Reference Lieber1988), as there is a postverbal complement of the deverbal head noun critique. Note, incidentally, that the proper noun modifier in (5) easily alternates with a corresponding determiner genitive (6) and in fact does so in the same source.Footnote ¹⁹

1. (6) Dennett … argues that Gould and Lewontin's critique is remarkably similar in logic to … (Pinker & Bloom 1990: 727, fn. 2, emphasis mine)

Schlücker's analysis (Reference Schlücker2013, Reference Schlücker, Ackermann, Simon and Zimmer2018) of proper noun modifiers as anchoring modifiers captures their (potential) syntactic nature well and can account for the referentiality of identifying proper noun modifiers, but, as correctly pointed out by Breban (Reference Breban2018: 388), this analysis does not take account of the ordering restriction whereby identifying (proper) noun modifiers strictly follow any qualifying adjectives, because in Rijkhoff's (Reference Rijkhoff2002, Reference Rijkhoff and McGregor2009) model anchoring modifiers precede quantifying, qualifying and classifying modifiers. That is, Schlücker's analysis cannot account for the fact that it is the new FBI director and not *the FBI new director in English.

So, from a theoretical point of view identifying (proper) noun modifiers either constitute unusual referential anchors (occurring in the wrong position in the noun phrase) or – when fitted into the right position as epithets or, for deverbal head nouns, compounds – exhibit properties not wholly consistent with epithets or compounds either. I prefer to analyse (proper) noun modifiers, or, more precisely, those (proper) noun modifiers interchangeable with determiner genitives, as unusual referential anchors rather than unusual epithets because this stresses their ability to alternate with determiner genitives.Footnote ²⁰ In Rosenbach (Reference Rosenbach, Traugott and Trousdale2010) I argue that they constitute a mismatch in the usual correspondence of semantic function (determination, qualification/premodification, classification) and word order within the noun phrase, where elements contributing to the reference of the noun phrase are typically situated at the left edge of the noun phrase, while elements contributing to the denotation (i.e. properties of the head noun) are typically positioned close to the head (see, e.g., Teyssier Reference Teyssier1968; Seiler Reference Seiler and Seiler1978; Quirk et al. Reference Quirk, Greenbaum, Leech and Svartvik1985; Adamson Reference Adamson, Fischer, Rosenbach and Stein2000). The analysis of identifying proper noun modifiers as epithets cannot resolve this violation of the usual iconicity in the noun phrase, either, though Breban (Reference Breban2018: 399) explicitly stresses the fact that her version of Halliday's (Reference Halliday1994) framework ‘more radically detaches form and function’, which is one way of incorporating such mismatches into a model of the noun phrase. None of the grammars of English such as Quirk et al. (Reference Quirk, Greenbaum, Leech and Svartvik1985), Biber et al. (Reference Biber, Johansson, Leech, Conrad and Finegan1999) or Payne & Huddleston (Reference Payne, Huddleston, Huddleston and Pullum2002) nor treatments of the English noun phrase such as Rijkhoff's (Reference Rijkhoff2002) or Keizer's (Reference Keizer2007) mention (proper) noun modifiers with identifying function. They have only recently received attention in the linguistic literature and, as the preceding discussion shows, their theoretical status is tricky and still open to discussion.

While we have seen that it is not clear how best to treat identifying (proper) noun modifiers theoretically, the analyses both as reference points and as identifying epithets capture the fact that they restrict the reference of the noun phrase, i.e. have identifying function, so descriptive synonymy with determiner genitives is actually given in either case. Does it matter by which mental process precisely the reference of the noun phrase gets restricted? Not from a variationist point of view. To illustrate this point further, consider again classic genitive variation (the FBI's director vs the director of the FBI). Here the dependent in the determiner genitive has the structural position and function of a determiner, while the dependent in the of-genitive is usually considered a restrictive postmodifier (Quirk et al. Reference Quirk, Greenbaum, Leech and Svartvik1985: § 17.38; Biber et al. Reference Biber, Johansson, Leech, Conrad and Finegan1999: §8.1) or complement (Payne & Huddleston Reference Payne, Huddleston, Huddleston and Pullum2002: § 14.1), without conferring definiteness on the possessive noun phrase. Like constructions with an identifying (proper) noun modifier, postnominal possessives with an of-genitive receive their (in)definiteness from the initial determiner, which is precisely the reason why such constructions (like constructions with noun modifiers) can be either definite or indefinite, e.g. a/the/this director of the FBI. Functionally, they can be considered referential anchors that help to identify the referent of the possessive noun phrase just like determiner genitives (e.g. Hawkins Reference Hawkins1991; Keizer Reference Keizer2007; Rijkhoff Reference Rijkhoff and McGregor2009), though, from a cognitive point of view, the conceptual path for identification in constructions with determiner genitives and of-genitives differs (see, e.g., Taylor Reference Taylor1996; Keizer Reference Keizer2007). Under a variationist definition of equivalence as given in section 2 above, however, the mental route by which anchoring possessives identify lies outside the realm of their descriptive meaning. Thus, what matters from a variationist point of view is the fact that determiner genitives and (proper) noun modifiers (and of-genitives, for that matter) help to identify the referent of the noun phrase and not how this identification process works in detail, though the latter is certainly an important question for an in-depth semantic–pragmatic study of these constructions and helps us to identify the factors that govern the alternation.

3.2 (In)definiteness of the matrix NP

Breban (Reference Breban2018) makes the important observation that proper noun modifiers may have identifying function even within indefinite noun phrases, pace Schlücker Reference Schlücker2013, who argues that identifying function is restricted to definite noun phrases and that noun modifiers in indefinite noun phrases trigger a classifying interpretation.Footnote ²¹ Breban argues, to my mind convincingly, that in expressions like a Christmas day tragedy the proper noun modifier identifies an instance of the type tragedy (one that happened on a specific day, i.e. Christmas day) and thus has identifying function, rather than denoting a type of tragedy. The indefinite article merely signals that the referent cannot be sufficiently identified via the proper noun modifier. In (7) there are two more examples I found that further illustrate the identifying function of proper noun modifiers within an indefinite noun phrase:

1. (7) (a) Although Professor Simpson's comments referred to a 1943 case, …

   (Minette Walters, Disordered Minds, p. 19, emphasis mine)

   (b) Her hair was nearly as blond as when Gary had first met her, at a Bob Seger concert at the Spectrum (Jonathan Franzen, The Corrections, p. 163, emphasis mine)

In both examples the proper noun modifier is part of an indefinite noun phrase which refers to a specific referent. It helps to identify this referent by reference to a specific point in time (7a) or via a specific person (7b), but while the referents of (7a) and (7b) are specific, they are not identifiable by the addressee. We can paraphrase these expressions as ‘a (certain) case from 1943’ and ‘a (certain) concert of/by Bob Seger’, respectively. Breban (Reference Breban2018) analyses such proper noun modifiers as identifying epithets where the proper noun modifier ‘describes a property that is not sufficient to identify the referent’ (p. 394). Epithets in general can be freely found in indefinite noun phrases (Breban Reference Breban2018: 393), but note that Halliday (Reference Halliday and Matthiessen2004: 319) explicitly restricts defining (i.e. identifying) epithets to definite noun phrases, so Halliday's model needs further widening to include such identifying epithets.

Can these proper noun modifiers possibly be referential anchors? Referential anchors are usually treated within the context of definite noun phrases (e.g. Löbner Reference Löbner1985, Reference Löbner and Botley1998, Reference Löbner2011; Taylor Reference Taylor1996; Rijkhoff Reference Rijkhoff and McGregor2009), because they are closely tied to the expression of definiteness (cf. also Haspelmath Reference Haspelmath1999). Von Heusinger (Reference von Heusinger2002), however, presents an account of referentially anchored specific indefinites in the sense that ‘the referent of the specific NP is functionally dependent on the referent of another expression’ (p. 268). Under this view, ‘a specific indefinite is an underspecified representation that needs an anchor in the context’ (von Heusinger Reference von Heusinger, Comorovski and von Heusinger2007: 290). In this account the anchor must be familiar to both speaker and hearer (see also von Heusinger Reference von Heusinger, von Heusinger, Maienborn and Portner2011: 1047). The examples a Christmas day tragedy, a 1943 case and a Bob Seger concert all represent specific indefinite noun phrases where the anchor is a proper noun modifier, which by definition has unique reference and thus is familiar. It remains to be seen, though, how precisely identifying proper noun modifiers can be incorporated into von Heusinger's (Reference von Heusinger2002) formal-semantic definition of specific indefinites as indexed epsilon terms.

The important question in the context of the present article is whether proper noun modifiers as in a Christmas day tragedy can alternate with a corresponding determiner genitive. Note that prenominal possessives consist of two noun phrases, the possessor noun phrase and the (matrix) possessive noun phrase. Only the former is explicitly marked for definiteness, while the definiteness of the latter must be inferred and thus leaves room for interpretation. While the status of possessives with indefinite determiner genitives (a friend's house) is highly debated (see section 3.4 below), it is generally assumed that definite determiner genitives always render the possessive noun phrase definite (e.g. Huddleston Reference Huddleston1984: 253; Löbner Reference Löbner1985, Reference Löbner and Botley1998, Reference Löbner2011; Quirk et al. Reference Quirk, Greenbaum, Leech and Svartvik1985: 326; Taylor Reference Taylor1996; Biber et al. Reference Biber, Johansson, Leech, Conrad and Finegan1999: 271; Payne & Huddleston Reference Payne, Huddleston, Huddleston and Pullum2002: 467–8).Footnote ²² Accordingly, the FBI's director translates into ‘the director of the FBI’ or ‘the FBI director’ and not ‘a director of the FBI’/ ‘a FBI director’.

So, while Christmas day's tragedy, 1943's case or Bob Seger's concert are all perfectly formed possessives, they designate uniquely identifiable referents or events and as such are not equivalent to a Christmas day tragedy, a 1943 case or a Bob Seger concert, where the proper noun modifiers do not uniquely identify the noun phrase's referent, as flagged by the indefinite article. Thus, proper noun modifiers within an indefinite noun phrase cannot alternate with a corresponding determiner genitive, at least not determiner genitives as we typically know them.Footnote ²³

There is, however, a very marginal prenominal genitive construction that allows an indefinite reading of a proper noun possessor. The following example (8) is taken from the American English news section of the ARCHER corpus.

1. (8) a Panama's main military force (ARCHER corpus, American English: 1955.wash1.n8a)

The initial indefinite article clearly determines the head (force) and not the proper noun possessor (Panama), suggesting the structure of a classifying genitive. However, the fact that the head is directly premodified by main military is not consistent with a classifying genitive, which typically is adjacent to the head.Footnote ²⁴ What is more, the possessor Panama helps to identify the referent of the noun phrase, suggesting the reading ‘a main military force of Panama’ rather than a type of main military force. However, it doesn't single out a uniquely identifiable referent as a typical determiner genitive would, as indicated by the initial indefinite article. So what we see here is the production, in a published newspaper, of a determiner genitive construction that is explicitly marked for indefiniteness and that violates a core property of determiner genitives in English, i.e. their co-occurrence restriction with other central determiners (*the/this John's book).

What looks like a really odd construction and might strike most people as an error, is actually not an isolated example, though it is not frequent either. The examples below, all taken from the Web, illustrate the same type of construction with a locative proper noun possessor in (9a), a temporal possessor in (9b) and even a human proper noun possessor in (9c).

1. (9) (a) One visit to Via Quadronno is all it takes to determine why it has

   become a New York's dining critics’ favoriteFootnote ²⁵

   (b) Local youths during a 2016's birding inter-schools excursion hosted

   by the Camdeboo Honourary Rangers.Footnote ²⁶

   (c) This outstanding showing paves the way for a Merkel's third term.

   Germans voters have indeed handed Merkel her third term in a wave of appreciation for her role in handling the economy during the Eurozone crisisFootnote ²⁷

This construction has not received any attention in the grammars of English or the literature on English possessives so far (to the best of my knowledge). Clearly, the genitive in this case does not occupy the central determiner position as is typically the case with determiner genitives, as this position is already taken by the preceding indefinite article. On the other hand, it doesn't take the typically prehead position of a classifying genitive, either. Rather, it follows the central determiner a and precedes the postdeterminer third in (9c) and the restrictive adjective main in (8). Interestingly, this ‘quasi’-determiner genitive can also be found with definite possessive noun phrases where the definiteness marking usually would be redundant; see the following examples with a locative possessor (10a), temporal possessor (10b) and human proper noun possessor (10c and 10d), which all have identifying meaning.

1. (10) (a) By the 1980s, increased police surveillance and implementation of

   increased security measures (razor wire, guard dogs) combined with

   continuous efforts to clean it up led to the weakening of the New

   York's graffiti subculture.Footnote ²⁸

   (b) She participated in the 2016's edition of Aristoteles Workshop and

   came back to lend a hand to this year's students.Footnote ²⁹

   (c) The chances of the Merkel's coalition surviving are “very open,” Nils

   Diederich from Berlin's Otto Suhr Institute for Political Science told

   dpa, a comment that raises the spectre of early elections.Footnote ³⁰

   (d) Nationally, support for the Merkel's bloc fell to 28 percent in an

   Infratest Dimap poll for broadcaster ARD, …Footnote ³¹

There is evidence suggesting that this construction is not a recent innovation. Example (8) above dates from 1955; example (11a) below comes from T. E. Lawrence's (1888–1935) memoirs; and example (11b) is from a 2011 reproduction of a book published in 1848.

1. (11) (a) Our today's fatigue corporal was a literalist, so we obeyed literally.Footnote ³²

   (b) I don't mean the schoolboys, who lose a good deal of fun by a

   January's restriction on sport. (The Sporting Review, by John William Carlton, p. 210, emphasis mine)

More examples of genitives with temporal dependents alternating between a clear determiner-genitive structure versus the quasi-determiner genitive described above can be found in Pepys’ seventeenth-century Diary, e.g. last night's drinking versus my last night's drinking, briefly discussed in Rosenbach (Reference Rosenbach2007a: 182–3).Footnote ³³

Either the usual article–possessor complementarity does not hold in this specific case, with English having not fully grammaticalised it (yet?),Footnote ³⁴ or the possessor must be considered a postdeterminer or restrictive adjectiveFootnote ³⁵ rather than a central determiner in these examples.

What is important in the context of the present article is that this quasi-determiner genitive, like (proper) noun modifiers, has identifying function and is neutral to the expression of definiteness. As such it may serve as a structural genitive correlate to a proper noun modifier within an indefinite noun phrase, and we can envisage the following alternation (the attested uses from the examples above are marked in bold in (12)).

1. (12) (a) a Christmas day tragedy versus a Christmas day's tragedy

   (b) a New York dining critics’ favourite versus a New York ’s dining

   critics’ favorite

   (c) a 1943 case versus a 1943's case

   (d) a third Merkel term versus a Merkel's third term

However, this quasi-determiner construction is so marginal that we lack clear intuitions about it (at least I do). It is therefore somewhat difficult to determine for individual expressions whether there is potential alternation and what it would look like. A Merkel's third term should alternate with a third Merkel term and does so (see (13a)), but a Merkel third term (13b) is also attested (if rare), which indicates a further blurring of the boundaries between determiner genitives and (proper) noun modifiers.

1. (13) (a) Will a third Merkel term bring a softening of austerity?Footnote ³⁶

   (b) If voters have decided to reward her with an absolute majority, a

   Merkel third term is likely to be marked by the same pragmatism as the first two.Footnote ³⁷

To conclude, the potential variation between proper noun modifiers and determiner genitives is typically – though not necessarily – restricted to the context of a definite noun phrase. Note that indefinite possessive noun phrases in general are extremely uncommon. According to Haspelmath (Reference Haspelmath1999), about 95 per cent of possessive noun phrases are definite,Footnote ³⁸ which reflects their function as referential anchors which usually is tied to the expression of definiteness in the noun phrase, though, as argued, referentially anchoring is also possible within specific indefinites (von Heusinger Reference von Heusinger2002).Footnote ³⁹ Potential alternation is limited within this marked context between a marginal type of quasi-determiner genitive and proper noun modifiers (a Christmas day's tragedy vs a Christmas day tragedy). For the potential of common noun modifiers within indefinites (a barn door) to alternate with a corresponding determiner genitive (a barn's door) see section 3.4 below.

Note, finally, that the fact that referential anchoring is possible within an indefinite noun phrase indicates that the definite article is not necessary to induce the identifying function of a proper noun modifier, pace Schlücker (Reference Schlücker2013) (see section 3.1 above).

3.3 The type of semantic relation

In general, it is not at all clear what type of taxonomy we should adopt to compare the semantic relations that allow both determiner genitives and (proper) noun modifiers as there is no uniform terminology for describing either, let alone both. Rather than solving this problem and providing an exhaustive analysis of the semantic relations that can be expressed by both a determiner genitive and a (proper) noun modifier, in this section I will address a few general issues that bear on the question of how equivalence is defined from a variationist perspective.

Breban (Reference Breban2018) argues that the range of semantic relations that can be expressed by both proper noun modifiers and determiners does not overlap completely, as proper noun modifiers can express relations that cannot be expressed by a corresponding determiner genitive (e.g. the Abu Graibh prison), and vice versa, and that this, in turn, challenges the assumption of equivalence between the two constructions. There certainly is evidence, from various languages, that proper noun modifiers can express a wider range of semantic relations than determiner genitives; see Breban (Reference Breban2018) for English, Koptjevskaja-Tamm (Reference Koptjevskaja-Tamm, Kersti, Denison and Scott2013) for Swedish and Schlücker (2013, 2018) for German. Drawing on collaborative work with Julia Kolkmann and John Payne, Breban (Reference Breban2018) also argues that there are some core relations typically expressed by determiner genitives, such as kinship (John's sister) or ownership (John's bike), that do not occur with proper noun modifiers (actually I put forward the same claim in Rosenbach Reference Rosenbach2007a: 171 and was likewise wrong). Their claim is based on the observation that there is no single occurrence with either relation in a corpus study containing a dataset of 179 proper noun modifiers. Indeed, kin relations and legal ownership are very rarely used with (proper) noun modifiers, but they do exist, so this claim – based on negative evidence from a small dataset – can be discarded.Footnote ⁴⁰ The examples in (14) all illustrate proper noun modifiers conveying an ownership relation.Footnote ⁴¹

1. (14) (a) “Mr. Lindenbaum had two bags, a small one and a larger canvas

   duffel …” […] “The Lindenbaum bag and its contents, and these

   objects alone, of all the artefacts recovered from the crash”

   (Kathy Reichs, Fatal Voyage, p. 350, emphasis mine)

   (b) Chuck nodded, looking past Alfred at the Lambert house.

   (Jonathan Franzen, The Corrections, p. 286, emphasis mine)

   (c) I told Zamzow about the findings at the Foote farm, …

   (Kathy Reichs, Bare Bones, p. 225, emphasis mine)

   (d) I found no reference to Prentice Dashwood, to the Arthur property, or to the officers of the H&F Investment Group.

   (Kathy Reichs, Fatal Voyage, p. 261, emphasis mine)

Proper noun modifiers with kin relations are more difficult to find, but they do exist. I googled for the expression the Obama daughter and found the following occurrences on the Web (all retrieved 23 December 2017):

1. (15) (a) Is The Obama Daughter, Malia, Going To Go On Birth Control?Footnote ⁴²

   (b) See photos of the Obama daughter's trip to Morocco …Footnote ⁴³

   (c) Malia Obama was spotted in the Hamptons this past weekend

   celebrating her 19th birthday with friends at one of Montauk's hot spots. She's the Obama daughter that came into this world with fireworks, as Malia was born on the holiday, the Fourth of July …Footnote ⁴⁴

   (d) RTE News presenter Bryan Dobson later picked up on the Obama daughter's apparent boredom, tweeting: …Footnote ⁴⁵

Schlücker (Reference Schlücker, Ackermann, Simon and Zimmer2018: 294) shows that out of a collection of 115,000 German proper noun compounds there are 47 with the head lemma Sohn (‘son’), 139 with the head lemma Tochter (‘daughter’), 60 with the head lemma Kind (‘child’), 17 with the head lemma Vater (‘father’) and 20 with the head lemma Mutter (‘mother’). So while they are comparatively rare given the size of this huge database, especially the ones with head nouns denoting unique reference, like father or mother, there is apparently no categorical ban against them, at least not in German, which seems to pattern largely like English with respect to proper noun modifiers. In other words, the type of semantic relation expressed certainly plays an important role when choosing between a determiner genitive and a proper noun modifier and should therefore be studied as a factor. Prototypical, core possessive relations, however, do not per se define an area of non-equivalence (or categorical context).

This finding is in fact further supported by Breban et al. (Reference Breban, Kolkmann and Payne2019), who tested the acceptability of constructions with determiner genitives and proper noun modifiers in an experimental study with authentic examples from the British National Corpus (BNC) in three conditions exhibiting different semantic relations.

1. 1. Only the determiner genitive should be possible:

   Northern Ireland's experience in Spain and Mexico in the last two World Cups taught them that heat can pose more problems than the opposition for British-based players.

2. 2. Only the proper noun modifier should be possible:

   President Bhutto of Pakistan was seeking to adopt a more conciliatory approach to the resolution of the Kashmir problem.

3. 3. Both determiner genitive and noun-modifier should be possible:

   Jesslyn Parkes, the England goalkeeper/England's goalkeeper, will be hoping to guide her new team, Middlesex, to a winning start.

Subjects had to judge the acceptability of both a determiner genitive (e.g. Northern Ireland's experience) and a corresponding proper noun modifier (e.g. the Northern Ireland experience) on a scale of 1–10. The results largely confirm the authors’ hypotheses about which semantic relations are favoured by which type of dependent (determiner genitive vs proper noun modifier): determiner genitives typically denote a possessive relation (defined here narrowly as ownership), while proper noun modifiers have a preference for a name relation (paraphrasable as, e.g., ‘called’, ‘known as’, as, e.g., in the Sainsbury family). However, their results also show that subjects hardly ever completely rejected a construction that should not have been possible according to the researchers’ hypotheses. The expression Kashmir's problem (6.2) was even judged almost as good as the Kashmir problem (6.7), though according to the researchers the intended goal (or involvement) interpretation should only allow the latter (i.e. the proper noun modifier). Breban et al. construe most deviations from their hypotheses as cases where subjects have come up with a different interpretation. In the rating task subjects had to give paraphrases of any variant rated with 4 or higher, which gave the researchers a good idea what kind of interpretation subjects had in mind when rating a variant. So, for example, subjects giving Kashmir's problem a fair rating would often paraphrase it in a way indicative of an undergoer interpretation (‘the problem that Kashmir has’), which is consistent with a determiner genitive, but is not the correct interpretation in this context. Note that subjects could also give proper noun modifier constructions a possessive interpretation, contrary to the researchers’ hypotheses. Breban et al. interpret the space for interpretation and variability they encounter in their study as evidence that the two constructions are semantically underspecified and may be associated with various interpretations that only get finally specified/interpreted in context, in line with Barker (Reference Barker1995), Vikner & Jensen (Reference Vikner and Jensen2002) and Peters & Westerståhl (Reference Peters and Westerståhl2013). The probabilistic nature of the mapping of semantic role to construction type observed by Breban et al. is precisely what leaves room for semantic overlap and variation. As such, the idea of semantic underspecification ties in well with a variationist approach.Footnote ⁴⁶

Note that Breban et al. (Reference Breban, Kolkmann and Payne2019) only focus on some of the semantic relations determiner genitives and proper noun modifiers can express, i.e. those represented in their dataset. While all seven semantic relations tested in their study can be expressed by both a determiner genitive or proper noun modifier (if not equally likely), further research needs to specify further which semantic relations are categorically banned from either construction. Also, some of the semantic roles used in Breban et al.’s study still need to be fine-tuned. For example, the location role can express a variety of relations, some alternating with a determiner genitive, as in (16a), others not, as in (16b).

1. (16) (a) …, but she's also interested in Quebec history.

   (Kathy Reichs, Death du Jour, p. 72, emphasis mine)

   (b) There must be paperwork on it somewhere. At the Kensington house,

   I should guess? (Elizabeth George, Playing for the Ashes, p. 162,

   emphasis mine)

An exclusively locative meaning, paraphrasable with ‘in’ (16b) or ‘from’ cannot be expressed by a corresponding determiner genitive (*Kensington's house), while a location relation allowing a have interpretation as in (16a) allows alternation (Quebec history vs Quebec's history). Note that in both (16a) and (16b) the proper noun modifier has identifying function and can be considered to be a referential anchor; however, only the former is compatible with a possessive interpretation. This shows that being a referential anchor is not a sufficient condition for potential alternation with a determiner genitive; the semantic relation expressed must be a possessive one (in a broad sense of the term as in figure 2 above).

3.4 Beyond proper noun modifiers: identifying common noun modifiers

The referentiality of the dependent is an important precondition for the equivalence of a noun modifier and a determiner genitive. Only if the dependent is referential, referring to a specific person or object, can it have the identifying function typical of referentially anchoring possessives. As argued above, proper noun dependents may retain their inherent referentiality as modifiers. However, proper noun modifiers do not represent a special case but simply the clearest case of an identifying noun modifier as they are high in saliency and topicality and thus the best candidate to alternate with a determiner genitive. The examples in (17) and (18) below all illustrate common noun modifiers with identifying meaning.Footnote ⁴⁷

1. (17) (a) She is the convent archivist, you know. (Kathy Reichs, Death du Jour,

   p. 6, emphasis mine)

   (b) Robin's father, Nick, was Fazio's middle child and the only Passafaro

   of his generation who never got with the Teamster program. Nick was

   the family brain and a committed Socialist; … (Jonathan Franzen, The Corrections, p. 394, emphasis mine)

   (c) The first of the cinema occupiers emerged, to the relief and consternation

   of the tense crowd, … (Timothy Pears, In a Land of Plenty, p. 643, emphasis mine)

2. (18) (a) To one side of this building stood a barn, and music issued from inside:

   Frank Sinatra by the sound of it, crooning a pop tune in Italian. St James

   headed in this direction. The barn door hung partially open, and he

   could see that its interior was whitewashed and lit by rows of

   fluorescent tubes that dangled from the ceiling. (Elizabeth George, A

   Place of Hiding, p. 311, emphasis mine)

   (b) Chip's doorman, Zoroaster, hurried out to help with the luggage and

   installed the Lamberts in the building's balky elevator. […] “Dad and I

   were at the housewarming in June,” Enid said. “It was spectacular.

   They'd had it catered, and they had pyramids of shrimp. It was solid

   shrimp, in pyramids. I've never seen anything like it.” “Pyramids of

   shrimp,” Chip said. The elevator door had finally closed. (Jonathan

   Franzen, The Corrections, p. 24, emphasis mine)

   (c) Fondly, she rubbed her cheek against his fur before replacing him in the

   cage. The piece of toast was nearly his size, but he managed to drag it

   industriously towards his nest. Elena smiled, tapped her fingers on the

   cage top, grabbed the rest of the toast, and left the room. (Elizabeth

   George, For the Sake of Elena, p. 14, emphasis mine)

In the examples in (17) the common noun modifiers are topical in context. In (17a) the whole context is about the convent, in (17b) the passage is explicitly about the family and in (17c) the larger context in which this example is embedded is situated within the cinema. In the examples in (18) the noun modifiers (barn, elevator, cage) are all topical in context, too. In addition, they also receive their specificity/referentiality from the fact that they denote a part/whole relation. In a part/whole relation the modifier so to speak inherits the referentiality of the noun phrase. If there is a specific door, then also the barn (18a) or the elevator (18b) it belongs to must be specific; likewise the presence of a specific instance of a top implies the presence of a specific cage (18c). Further evidence for the referentiality of these noun modifiers comes from the fact that they can be anaphorically referred to, as is explicitly done in (18a), where barn gets referred to later by its.

Schlücker (Reference Schlücker2013: §4.5) argues that barn in (18a) above receives its alleged referentiality by implicit or explicit reference from the context but in fact is not referential, although she does accept anaphoric accessibility as a test for the referentiality of proper noun modifiers. Schlücker treats proper and common noun modifiers differently for theoretical reasons: not being full noun phrases, noun modifiers should be anaphoric islands (Postal Reference Postal1969) and thus should not be accessible for anaphor, i.e. they shouldn't be referential. Proper noun modifiers are an exception because ‘they evade this restriction as they are inherently definite, and, for that reason, do not need an extra determiner’ (Schlücker Reference Schlücker2013: 470).Footnote ⁴⁸ It is actually impossible to decide empirically if the anaphoric reference to barn in (18a) is possible because barn is inherently referential or if the referentiality is pragmatically conditioned. Does it even matter for the question of equivalence (in the variationist sense)? Not really. Note that the existence of an entity can be evoked by ‘bridging inferences’ (Clark Reference Clark, LaBerge and Samuels1977; Clark & Haviland Reference Clark, Haviland and Freedle1977), which also go under the names of ‘inferrables’ (Prince Reference Prince and Cole1981) and ‘associative anaphor’ (Hawkins Reference Hawkins1978) in the literature. Two examples for bridging inferences are given in (19) (cited from Birner & Ward Reference Birner and Ward1994: 94):

1. (19) (a) I had dinner at that new Italian restaurant last night. It was a nice place,

   but the appetizer was far too spicy for my tastes.

   (b)  I hated that book. The author is an idiot.

Bridging inferences typically work via common-sense knowledge, e.g. the knowledge that dinners include an appetizer (19a) and that books have an author (19b), or by other reference to recently activated entities. The latter account for the examples in (17) above, where the referents of the proper noun modifiers have all been activated in context. According to Clark (Reference Clark, LaBerge and Samuels1977), bridging inferences help the listener to identify a referent as part of the problem-solving process in communication. Unique identifiability, i.e. the cognitive status Gundel et al. (Reference Gundel, Hedberg and Zacharski1993) assume for definite noun phrases of the type {the N} in the Givenness Hierarchy, can indeed be achieved by such bridging inferences (Gundel et al. Reference Gundel, Hedberg and Zacharski2001). So, what distinguishes determiner genitives from identifying noun modifiers is that the former are noun phrases and as such are referential, while a possible referential interpretation of the latter needs to be inferred from the context and thus is context dependent. Part/whole relations in specific noun phrases as in (18) all allow the inference of a specific noun modifier, as it is part of our world knowledge that the existence of a specific barn door/elevator door/cage top entails the existence of a specific barn/elevator/top.Footnote ⁴⁹ There is therefore no reason to assume that the barn door shouldn't alternate with the barn's door when barn allows a referential interpretation, even if it's ‘just’ pragmatically inferred. In fact, the examples of non-animate common noun modifiers in (17) and (18) above can all alternate easily with a corresponding determiner genitive: the convent archivist vs the convent's archivist, the family brain vs the family's brain, the cinema occupiers vs the cinema's occupiers, the barn door vs the barn's door, the elevator door vs the elevator's door, the cage top vs the cage's top. Example (20), discussed in detail in Rosenbach (Reference Rosenbach2007a: 177–9), illustrates this variation in discourse:

1. (20) Cut an ordinary photograph snipping a leg off the chair pictured. Then the

   chair leg is no longer visible. It is no longer part of the photographic image.

   

   Now snip off a comparably sized piece from a diffraction image hologram containing the same chair information. When this mutilated hologram is illuminated by the reference beam the whole real space image appears – albeit dimmer and fuzzier. The chair's leg is preserved. In fact it can't be removed from the hologram by cutting!

   That is because any part of the hologram relates to the whole of the real space image.Footnote ⁵⁰

As argued in Rosenbach (Reference Rosenbach2007a), the presence of the possessive ’s individuates and foregrounds the referent of the dependent, so the two constructions differ in their discourse-pragmatic functions.Footnote ⁵¹ Referentiality, usually treated as a binary feature, appears to be a graded semantic feature and different syntactic positions go with different degrees of referentiality (see Rosenbach Reference Rosenbach2006: 106–7 and references cited therein). Accordingly, a noun modifier can be referential (if by inference), but in this syntactic position is less so than a corresponding determiner genitive.Footnote ⁵² What matters for the question of equivalence, however, is that (proper) noun modifiers can be sufficiently referential to exert identifying function rather than the degree of their referentiality.

Schlücker (Reference Schlücker2013) rightly points out that in expressions like barn door the noun modifier has considerable classifying force and analyses it as an instance of the type barn door rather than the door of a specific barn. Common noun modifiers, like proper noun modifiers, often waver between an identifying and a classifying interpretation.Footnote ⁵³ So indeed a specific barn door is always (and necessarily) a particular instance of the type barn door, and the Major plan could potentially also refer to a plan typical of the former British prime minister John Major if we can conceptualise it as a type (Rosenbach Reference Rosenbach2007a: 149). What is crucial for the present argumentation is that the construction as such is compatible with both an identifying and a classifying interpretation in this context. For common noun modifiers to be equivalent to determiner genitives it is sufficient that the identifying meaning is available, no matter what other meaning(s) are implicitly present as well. There only needs to be a sufficient semantic overlap to allow alternation, not complete identity in the descriptive meaning(s) that can be expressed by either construction (see section 2 above).

Further research is needed to spell out the precise contexts where common noun modifiers can have identifying function and where such a function is banned. For example, a part/whole relation does not seem to be a sufficient criterion for the dependent to have identifying function, as e.g. in the barn door, the elevator door or the chair leg. An expression like *the building door, although expressing a part/whole relation, seems to be ruled out and it is not quite clear to me yet why. Another puzzle is the observation that identifying human common noun modifiers barely occur; diachronically human noun modifiers only surface rather late in Rosenbach's (Reference Rosenbach2007a) diachronic corpus (see also Breban & De Smet Reference Breban and De Smet2019) and constitute only about 1 per cent of all noun modifiers in its most recent time interval (1950–99), the majority of which are proper nouns.

Yet another question is whether common noun modifiers can also have identifying function within an indefinite noun phrase. Under a specific interpretation of a barn door we can assume that barn, as in the barn door, may get a referential interpretation. The question, however, is whether it can alternate with a corresponding determiner genitive (a barn's door). This is a tricky issue and depends on the theoretical stance taken on the definiteness of indefinite determiner genitives. Under a morphosyntactic definition of definiteness, defined configurationally (C. Lyons Reference Lyons1986, Reference Lyons1999), determiner genitives occupy determiner position and render the noun phrase invariably definite, which is the standard view expressed in the grammars of English (cf. Huddleston Reference Huddleston1984: 253; Quirk et al. Reference Quirk, Greenbaum, Leech and Svartvik1985: 326; Biber et al. Reference Biber, Johansson, Leech, Conrad and Finegan1999: 271; Payne & Huddleston Reference Payne, Huddleston, Huddleston and Pullum2002: 467–8). Under this view, alternation is ruled out as a barn's door translates into the barn door and not a barn door. However, the definiteness status of possessive noun phrases with indefinite determiner genitives (a barn's door) is highly disputed. They have been analysed as invariably definite (Woisetschlaeger Reference Woisetschlaeger1983; C. Lyons Reference Lyons1999, and see the references above), as semantically indefinite (Löbner Reference Löbner1985, Reference Löbner and Botley1998, Reference Löbner2011; Barker Reference Barker, Maienborn, von Heusinger and Portner2011), as weakly definite (Rosenbach Reference Rosenbach2006: 107–9), as being compatible with definiteness (Taylor Reference Taylor1996: §7) or as functionally indefinite (Willemse et al. Reference Willemse, Davidse, Heyvaert and McGregor2009).Footnote ⁵⁴ If we allow for the possibility that indefinite determiner genitives are (somehow) indefinite, a barn door can, in principle, alternate with a barn's door, under a specific reading of the matrix noun phrase.Footnote ⁵⁵

3.5 Lexicalisations

Constructions with determiner genitives can undergo conventionalisation and become part of the lexicon just like any other syntactic phrase. Some typical examples of lexicalised genitive constructions are given in (21) below. They can either derive historically from productive determiner genitive constructions or are conventionalised classifying genitives (see Taylor Reference Taylor1996: 311 or Rosenbach Reference Rosenbach2007a: 180–1), often getting further reduced by omitting the ’s. As fixed expressions they come to denote concepts. The examples in (22) represent a special case, where the proper name dependent is used neither to identify nor to classify the head, but helps to denote a concept by connection to a certain individual. These genitives have become known as ‘onomastic genitives’ in the literature (see Taylor Reference Taylor1996: 295–7; Koptjevskaja-Tamm & Rosenbach Reference Koptjevskaja-Tamm and Rosenbach2005; Rosenbach Reference Rosenbach2007a: 176–7). The particular relation between a proper noun modifier and the head noun has been called ‘commemorative’ (Koptjevskaja-Tamm Reference Koptjevskaja-Tamm2009, Reference Koptjevskaja-Tamm, Kersti, Denison and Scott2013; Schlücker Reference Schlücker2013).

1. (21) (a) [a/the driver]’s licence > a/the [driver's licence] > a/the [driver

   licence]

   (b) [a/the butcher]’s knife > a/the [butcher's knife] > a/the [butcher knife]

2. (22) (a) [St Valentine]’s Day > [St Valentine's Day] > [St Valentine

   Day], [Valentine Day]

   (b) [Planck]’s constant > the [Planck's constant] > the [Planck constant]

Naturally, there is variation between a genitive and a noun modifier for such expressions as there is typically a period during lexicalisation where both the older and the newer expression co-exist, a situation known as ‘layering’ within the grammaticalisation framework (Hopper & Traugott Reference Hopper and Traugott1993). Meaning equivalence is generally given because it is the same concept/meaning that undergoes lexicalisation. Yet the issue of equivalence is a tricky one with such lexicalising expressions, as the nominal dependents lose their referentiality in the process and may become semantically specialised. For example, under a determiner reading a butcher's knife designates the knife of a specific butcher, which could be any kind of knife, while under a classifying reading the expression designates a type of knife as typically used by butchers, with a non-specific reading of butcher. Such lexicalising expressions can be ambiguous in interpretation in contexts that allow a specific reading of the matrix noun phrase, as in the case of a/the barn door discussed in section 3.4 above. The fine and subtle linguistic distinction between the particular instance of a knife of a specific butcher (determiner reading) and the particular instance of the type butcher(’s) knife (classifying reading) is often irrelevant in the actual context, especially in contexts where there is a specific butcher, as exemplified in (23).Footnote ⁵⁶

1. (23) Naturally, the “uptown” butchers are kept pretty busy. Although there were a

   few more fat pigs than usual lying hog-tied awaiting the butcher's knife on the

   seventeenth, Chencuo's day of observance, they weren't lined up.Footnote ⁵⁷

A similar type of layering of a determiner and a classifying reading holds for onomastic genitives; see the following examples from Rosenbach (Reference Rosenbach2007a: 177):

1. (24) (a) Wave Structure of Matter (WSM) explains The Famous Scientist Max

   Planck's Constant and Quantum energy states of Light and …Footnote ⁵⁸

   (b) This constant came to be known as Planck's constant …Footnote ⁵⁹

   (c) For example, the simple harmonic oscillator has energy levels with a

   constant spacing proportional to a fundamental constant known as the

   Planck's constant …Footnote ⁶⁰

In (24a) we see a determiner genitive, referring to the scientist Max Planck ([the famous scientist Max Planck]’s constant), while (24b) is structurally ambiguous between a determiner and a lexicalised (onomastic) reading, though the context is suggestive of the latter. Example (24c) shows an unambiguous onomastic genitive, where the whole expression names this constant. In Wikipedia this constant is listed as Planck Constant, though it explicitly acknowledges the alternative expression Planck's constant, which indicates their meaning equivalence; see (25):

1. (25) The Planck constant (denoted h, also called Planck's constant) is a physical

   constant that is the quantum of action, central in quantum mechanics.Footnote ⁶¹

In onomastic genitives both the determiner and onomastic/commemorative reading may be available (though not necessarily so for all expressions and in all contexts).Footnote ⁶² Note again that in usage the difference in interpretation, which for linguists goes together with different structures, is not really important, as under both readings Planck's constant is a constant invented by Planck, no matter whether we more strongly individuate the inventor (determiner reading) or the name deriving from him (onomastic reading).

All this demonstrates that the question of meaning equivalence is a difficult one for lexicalising expressions. Strictly speaking, meaning equivalence should only be given between classifying (or onomastic) genitives and a corresponding noun modifier, as they are both non-referential and express a name relation, while the determiner genitive is referential and conveys a possessive relation. However, as also argued in section 3.4, the borderline between determination and classification (i.e. referential and non-referential anchoring) can be very fuzzy to the extent that in some contexts it is irrelevant in actual communication. My position is to include lexicalising expressions in a variation study of determiner genitives and (proper) noun modifiers if an identifying meaning is (still) present, even though usually the classifying interpretation will prevail.

Lexicalising expressions form a special source of potential variation between genitives and (proper) noun modifiers and actually one that can be found already in Early Modern English: St Valentine's Day vs Valentine's Day vs Valentine day,Footnote ⁶³ or Barnardes Castle vs Barnard Castle, Gascoyne's-tower vs Gascoyne-tower. Footnote ⁶⁴ Breban & De Smet (Reference Breban and De Smet2019) suggest that the emergence of proper noun modifiers has been boosted by a general proper noun modifier construction subsuming various constructions with proper noun modifiers that surface in English at different points in time. The presence of (proper) noun modifiers resulting from lexicalisations (as in Valentine Day, Barnard Castle, Gascoyne-tower) may have helped to instantiate such a general (proper) noun modifier construction and may have paved the way for the emergence of human proper noun modifiers at a time, when, according to Breban & De Smet's study, proper noun modifiers in general were still very rare and basically restricted to time/religious feast (Midsomer term) and place (London warde).

3.6 Discussion: assessing equivalence between determiner genitives and (proper) noun modifiers

Breban (Reference Breban2018) makes the useful distinction between

1. (a) contexts which never allow alternation (e.g. the Abu Graibh prison),

2. (b) contexts which in principle allow alternation but with different interpretations

   of the semantic relation (cf. Kashmir's problem vs the Kashmir problem), and

3. (c) contexts which allow alternation and preserve the semantic relation of the

   possessor and the head (e.g. the Mairs’ dinner party vs the Mair dinner party).

Equivalence in the variationist sense holds only for (c), which corresponds to the choice contexts in figure 1 above; the challenge for a variation study is to identify (a) and (b) and then exclude these contexts from the analysis.

Breban (Reference Breban2018) put forward two main objections against the assumption that (proper) noun modifiers and determiner genitives are equivalent:

1. (i) Identifying proper noun modifiers are epithets and not referential anchors, so

   the way they identify differs from determiner genitives (see sections 3.1 and

   3.3).

2. (ii) There is ‘incomplete functional equivalence’ (2018: 385) between determiner

   genitives and proper noun modifiers as they don't exactly occur in the same

   environments and the same relations (see sections 3.2 and 3.4).

Schlücker (Reference Schlücker, Ackermann, Simon and Zimmer2018: 295) proceeds from the same notion of equivalence, concluding that ‘there is no true meaning equivalence in general between both constructions’, though she acknowledges that ‘particular kinds of proper name compounds and genitives … exhibit a substantial amount of semantic overlap which makes them interchangeable under certain conditions’ (2018: 295).

Breban's and Schlücker's interest is to give a comprehensive account of proper noun modifiers. They regard functional equivalence of all proper noun modifiers with determiner genitives as a prerequisite for the functional equivalence on the constructional level. Given this particular notion of equivalence, both Breban and Schlücker are of course right: there is indeed no complete functional overlap between constructions with proper noun modifiers and determiner genitives; on a constructional level they only partly overlap. Variation studies, however, naturally focus on those contexts where there is equivalence in the sense of semantic overlap, i.e. choice contexts. Categorical contexts that do not allow any choice naturally lie outside the scope of the variationist approach and their exclusion from the analysis is a heuristic necessity rather than an oversight.

I have argued that identifying (proper) noun modifiers are referential anchors (if unusual ones), in both definite and indefinite noun phrases. Note that from a variationist perspective it is sufficient that the variants are ‘two ways of saying the same thing’ rather than saying them in precisely the same way. Under this view it actually doesn't really matter whether we analyse these (proper) noun modifiers as referential anchors or as epithets as long as we acknowledge their function of restricting (or defining) the reference of the noun phrase and their ability to express a possessive relation, if by different routes (see section 3.1). The cognitive mechanisms or pragmatic differences that go with the way identification (reference restriction) works in these constructions are irrelevant on the level of descriptive synonymy.

While (proper) noun modifiers may have identifying meaning, they often also simultaneously express other meanings, in particular classifying meaning (e.g. the barn door). I have argued that in order to alternate, expressions don't need to be identical in all their potential descriptive meaning(s); there just needs to be sufficient semantic overlap, even if it's only partial. In other words, it is sufficient that an identifying meaning is present, no matter what other meaning(s) is/are co-present as well.

Note finally that what ultimately matters from a variationist perspective is that expressions are perceived as being sufficiently similar by speakers to alternate. This means that the equivalence between (proper) noun modifiers and determiner genitives and their ability to alternate ultimately rests on the level of usage, as repeatedly argued throughout this article. Of course language usage is not independent from the language system, but it allows for deviations and for not so perfect productions and interpretation by not so perfect speakers and hearers, especially in the ‘grey’ areas of grammar. We should also consider the possibility that we, as linguists, postulate fine distinctions that speakers may very well be oblivious to in actual use. We seem to proceed from an ideal speaker/hearer in our theoretical analyses, even within functional approaches focusing on language usage, but these fully competent speakers/hearers hardly exist and usage can be ‘messy’ (at least from a theoretical linguist's point of view).Footnote ⁶⁵ This is nicely demonstrated by Breban et al.’s (Reference Breban, Kolkmann and Payne2019) experimental study, where subjects repeatedly assigned interpretations to expressions that the researchers thought were ruled out for that particular construction. In addition, the ideal state formulated by linguistic theory often gets blurred by ambiguities or vagueness in meaning that indeed allow various interpretations in context, often with little if any practical difference in meaning for the language use as argued at various points in this article. Lexicalising expressions that vacillate between a determiner genitive, classifying or onomastic genitive and (proper) noun modifiers (see section 3.5 above) further contribute to the blurring of the distinction between the expression of determination (reference restriction) and classification (denotation/type restriction) or the designation of a name relation, as does the quasi-determiner genitive discussed in this article (a Merkel's third term). Footnote ⁶⁶

4.1 On equivalence

In this article I argued that determiner genitives and (proper) noun modifiers are equivalent in meaning (or rather, equivalent enough in meaning) to alternate in certain contexts. It has been shown that the term ‘equivalence’ is used in different ways within linguistics, actually reflecting the dividing line between theoretical linguistics giving comprehensive analyses of constructions on the system level and taking into account all levels of meaning, and the variationist framework studying alternative ways of ‘saying the same thing’ in usage and thus proceeding from a weaker version of equivalence in terms of ‘rough semantic equivalence’ or descriptive synonymy; for an overview and summary see table 1.

Table 1. Conceptions of equivalence

These two perspectives are of course not mutually exclusive. Although potential alternation and thus equivalence ultimately lies on the level of usage, where fine theoretical details may be ignored or overridden as argued throughout this article, theoretical analyses are indispensable for the variationist analysis to identify the classes of alternating vs non-alternating contexts and thus define and operationalise the scope of the alternation.

4.2 On variation and gradience

As initially argued, the variationist notion of equivalence is the one conceptually underlying my gradience approach (Rosenbach Reference Rosenbach2006, Reference Rosenbach2007a, Reference Rosenbach, Hundt, Nesselhauf and Biewer2007b, Reference Rosenbach, Traugott and Trousdale2010). There are, however, important differences in perspective and procedure between studying gradience and studying variation: for an argument of linguistic gradience we focus precisely on the ‘grey’ areas of grammar and the odd and ambiguous examples, where the distinctions between constructions get blurred, while a variationist study needs to ‘squeeze’ tokens into categories and quantify them. The latter has practical implications. For example, (proper) noun modifiers differ from determiner genitives and of-genitives in that they are not full noun phrases and their referentiality needs to be inferred from the context. It is therefore difficult to define the precise classes of (non-)alternating expressions in a straightforward way when analysing (and quantifying) corpus data as inferences can be difficult to operationalise. Note further that while the context of indefinite (possessive) noun phrases is an interesting one for a gradience approach and in principle constitutes a choice context, in a corpus study we would probably exclude it for heuristic reasons as the area of potential alternation is so small and thus not well quantifiable and may involve structurally deviant variants as the quasi-determiner genitive, apart from the fact that we are dealing here with a theoretical wasp's nest which makes it difficult to select potentially alternating variants in the first place.

Note, finally, that a quantitative variationist study has to follow the ‘principle of accountability’ (Labov Reference Labov1972), a core principle of variationist work, according to which we need to consider all possible occurrences of a linguistic variable. As such, the researcher needs to find out all contexts where a variant could have been used but wasn't. This includes the identification of all the variable ways of ‘saying the same thing’ and the contexts in which they can occur. As noted in figure 2 above, there are actually three genitive variants (the FBI's director vs the FBI director vs the director of the FBI). Rather than three congruent constructions, we are dealing with three partially overlapping constructions, which Szmrecsanyi et al. (Reference Szmrecsanyi, Biber, Egbert and Franco2016: 25) aptly call ‘a web of variation’. Figure 3 visualises how the three constructions are intertwined on a general level. A variationist analysis of English genitive variation should therefore, ideally, focus on the area of choice context where all three variants overlap semantically.Footnote ⁶⁷

Figure 3. The web of English genitive variation

As pointed out by Szmrecsanyi et al. (Reference Szmrecsanyi, Biber, Egbert and Franco2016: 25), a serious complication of widening genitive variation beyond the usual binary alternation of determiner genitives and of-genitives is the fact that there are potentially other variants that can intersect with s-genitives, of-genitives and noun modifiers (cf. also Feist Reference Feist2012: 292–3), such as other prepositional phrases For example, should the alternation of the FBI's director vs the FBI director vs the director of the FBI also include the director from the FBI? And what about adjectival variants, namely the alternation between Halliday's framework, the framework of Halliday, the Halliday framework and the Hallidayan framework? Opening up genitive variation to make it correspond more closely to the principle of accountability is certainly a desirable step, but it will increase the web of variation considerably, leaving us with further partly overlapping variants, which creates a level of complexity for the quantitative analysis that I'm not sure we can handle (yet) methodologically.

So, while in principle (proper) noun modifiers can alternate with determiner genitives and form a variant in genitive variation, it is methodologically challenging to include them in an actual variation study.Footnote ⁶⁸

4.3 Notes on method

To further ascertain the precise classes of (non-)alternating expressions, both experimental studies and translations seem to me very promising methodological tools. In experimental studies we can test hypotheses and see in which contexts the two constructions can be used interchangeably or not, where they express distinct meanings, and where they are ambiguous in interpretation, especially for contexts only rarely attested in corpora, such as indefinite possessive noun phrases (see sections 3.2 and 3.4 above). Experiments tap the intuition of more or less verbally skilled individuals and may bring to light ambiguous or vague contexts (see also Breban et al. Reference Breban, Kolkmann and Payne2019).

In the case of translations, we are actually dealing with linguistically highly skilled speakers who have a high competence in assessing equivalence, because this is part of their job as translators. So, if a determiner genitive in one language gets translated as a (proper) noun modifier in another language (or vice versa), this constitutes evidence for their equivalence. In this article I've also used translations as evidence for assessing the referentiality of (proper) noun modifiers. Parallel corpora are now available that contain both original texts and their translations. They have become an excellent tool for contrastive studies (see, e.g., Aijmer & Altenberg Reference Aijmer and Altenberg2013). Ström Herold & Levin (Reference Ström Herold and Levin2019) is a first demonstration of a quantitative contrastive study of proper noun modifiers, genitives and prepositional possessives in a translation corpus covering English, German and Swedish.

Eventually, the synchronic nature of noun modifiers and their overlap with genitives can only be fully understood when taking a diachronic perspective. Future research ought to look in more detail at how the evolution of a general (proper) noun modifier construction described by Breban & De Smet (Reference Breban and De Smet2019) intersects with the development of genitive constructions, and how it fits into the wider context of the English noun phrase, in particular how it relates to the observed increasing complexity within the prenominal string of the noun phrase (e.g. Biber et al. Reference Biber, Grieve, Iberri-Shea, Rohdenburg and Schlüter2009; Biber & Gray Reference Biber, Clark, Fanego, López-Couso and Pérez-Guerra2002, Reference Biber and Gray2011, Reference Biber and Gray2016; Günther Reference Günther2019).