2.1 Basic data

Besides the typical presence of head nouns, phrasal PN may also involve articles – often called proprial articles – (4a), possessive determiners (4b), and prepositions (4c):

Considering the elements in (4), one may observe that phrasal PN involve both lexical and functional words. Indeed, lexical and functional elements can be combined in a wide range of ways yielding fairly complex strings. A representative sample is given below, where besides prepositions, determiners, and nouns, Saxon Genitives, adjectives, numerals, or particles can be present:

Like common DPs, phrasal PN can involve several adjectives, and these can occur in different orders:

Phrasal PN can be even more complex. They can involve coordinations and embeddings:

In section 3.2.1, I discuss the presence of the articles and other determiner-like elements in (4–5) and their absence in (6–7).³ Finally, it is well known that PN are semantically definite. Intriguingly, they can be syntactically indefinite. Examples can be found in the singular and plural:⁴

Note that the above examples all consist of multiple individual words. As such, they seem to be phrasal in structure – hence the name for these constructions. They are different from cases like Bahnhofstraße ‘Station Street’, which appear to be compound-like and are briefly discussed in the next subsection.

Before we proceed, I would like to make a comment about spelling (also Nübling et al. 2015: 86ff). In general, only nouns and adjectives based on geographical names are capitalized in German. Note that deutsch ‘German’ and historisch ‘historical’ are not such adjectives. Their capitalization in das Deutsche Historische Museum in (2c) indicates that these elements are part of the PN. In some cases, the entire noun is made up of capital letters (e.g., Der SPIEGEL ‘The Mirror’). In this paper, I follow the capitalization of words as they appear in their original context. Thus, the capitalization of determiner(-like) elements, adjectives, quantifiers and numerals indicate that these elements belong to the proper name. This means that das in (2c) is not part of the PN. Unfortunately, capitalization is not always consistent (see the examples containing kleine in (5b) or voller in (8a)). Particles like als are usually not capitalized, not even in proper names. Thus, spelling can only be used as an indication but not evidence per se of the proper namehood of these elements. Capitalization in the gloss indicates what belongs to the PN. As to the translation, it is not always clear how to render these proper names in English. For the most part, I provide a literal translation. With some minor adjustments, I retain the capitalization of the gloss in the translation line.

2.2 Previous proposals

Inherent PN have been discussed in much detail (e.g., Longobardi 1994; 2005; Anderson 2004), but phrasal PN much less so. If they are addressed, they are usually only briefly discussed, often just in passing. Detailed discussions are the exception. It is probably fair to state that phrasal PN have not received much attention in the literature. First, I summarize the insights of the works that only briefly address phrasal PN, and then I turn to one analysis that provides more details.

Anderson (2003: 372, 386) states that names tend to be desemanticized. However, although they are listed lexically, complex names like the University of Queensland retain some descriptive content. Anderson (2007: 106) categorizes nouns like University as, in his terms, “classifiers”. Karnowski & Pafel (2005: 52) state that phrasal PN like Deutsche Gesellschaft für Sprachwissenschaft ‘German Society for Linguistics’ have the same semantic properties as inherent PN (for these authors, PN are predicates). Following work by Kripke, Sturm (2005: 72) classifies cases like the Holy Roman Empire as a borderline case in that syntactically they belong to the class of definite expressions, but semantically they belong to the class of rigid designators. Finally, Weber (2004: 286–287) observes that new (i.e., phrasal) PN conform to regular morpho-syntactic patterns of DPs.

Formal, structural proposals are very rare. If they are provided, they usually do not involve many details. As one of the early works, Allerton (1987) provides a number of morpho-syntactic arguments showing that phrasal PN are special (for detailed discussion, see section 3.3). He proposes that (phrasal) PN are “much more like … lexical unit[s], and the semantic status of [their] parts [are] more akin to that of the morphemes in a compound word.” (page 64) Importantly, the morphological derivation of compounding is not further elaborated on. Note that the author seems to realize though that this type of compounding is not of the regular kind using hedges like “more akin to” in his discussion.

A notable exception to the shortcomings pointed out above is Köhnlein (2015). This author provides a detailed proposal of place names in Dutch. He argues that names like Amsterdam are morphologically and semantically complex: Amster-dam. While Köhnlein takes the different accent patterns of these PN as his point of departure (for German, see Nübling et al. 2015: 67), he also provides many syntactic and semantic details in his discussion. Focusing on the syntax and semantics here, Köhnlein observes that the individual parts of these PN do not occur as independent words and that the meaning of the parts differs from their etymological origin. He proposes that the first element (Amster-) has no meaning at all but provides a referential pointer to a unique place; the second element (-dam) indicates that the name is a settlement (distinguishing this name for people’s last names). This yields the following general bipartide representation of the cases discussed in Köhnlein:

The different status of these two parts from their etymological origin is also reflected by their differing morphological gender (place names are all neuter in Dutch). Crucially, following Bermúdez-Otero (2012), elements of the type in (9) are claimed not to be computed online but to be stored in the lexicon as complex units, that is, as analytical listings. The proposal is that the grammar still has access to the individual parts of this complex unit.

There are three components to the lexical entries in (9): a semantic, a syntactic, and a phonological one, all corresponding to the different modules of grammar: semantics, syntax, and phonology. Leaving the latter aside, let us illustrate this with the place name Amsterdam in more detail. Köhnlein parses this name as in (10a) providing the semantics in (10b) and the syntax in (10b’). Semantically, Amster- provides the referential pointer and the classifier -dam indicates that this name refers to a settlement (10b). As to the (morpho-)syntax (10b’), name morphemes are taken to have the syntactic feature [+proper], which distinguishes them from common nouns. Specifically, Amsterdam is a complex word that consists of two proper name stems combining into a proper name compound noun. Coindexation associates the different components of the representation across the various modules (syntax, semantics). The double arrow indicates that the different modules interact with one another (i.e., the syntax interacts with the semantics):⁵

Note that both parts of the word Amsterdam are heavily underspecified in their semantics, and this is taken to explain why these elements cannot occur as independent words. In sum, we can observe that unlike previous work, Köhnlein (2015) provides a very detailed structural analysis of PN, specifically of place names in Dutch.

Let me point out already here that the data examined in Köhnlein (2015) are different from the cases under discussion. In the current paper, the individual parts of phrasal PN typically have transparent meanings, and they can occur independently as regular words. Thus, while the above proposal explains complex place names in Dutch, it does not transfer straightforwardly to phrasal PN in German. As should be clear, Köhnlein’s proposal does not involve a plain-vanilla type of derivation for compounds. Before I proceed, let me illustrate this with data from German. The differences between the above compound-type PN, which have exceptional features, and the regular compounds and phrasal PN discussed below, which have many regular features, call for a novel (but closely related) proposal.

Nübling et al. (2015) provide observations that compound-type PN are not regular compounds. As in Dutch, regular compounds show the expected gender, but their place name counterparts are of neuter gender (see Nübling et al. 2015: 75):

The right-hand head rule explains the gender of regular compounds. As to phrasal PN, there are different possible solutions: either -stadt is a toponymic classifier of neuter gender (cf. Köhnlein’s 2015: 197 discussion of dam vs. -dam), or the right-head head rule is not at work here anymore. Rather, as Fahlbusch & Nübling (2014: 246) argue, certain classes of objects develop their own referential gender.

Second, compounds often involve a linking element. Nübling et al. (2015: 53) observe that the common noun Bahnhof ‘(train) station’ is always followed by the linking element -s- if it is the first part of a compound (12a). This is different if this element is part of a PN (12b) (see also Hochzeit ‘wedding’ in this regard):

Third, a reviewer points out that ordinary compounds and compound PN in the masculine (and neuter) show some interesting differences as regards case inflection. In a corpus study, Fritzinger (2018) discovered that there is a categorial difference in that the former must take a genitive inflection (13a) but the latter may not (13b) (-Ø marks the lack of an inflection). Second, there is also a difference in the preference of these endings such that regular compounds prefer the long genitive ending -es but that the compound PN have a slight preference for the short ending -s. The relative preferences of these inflections are provided in the example line in descending order going from left to right:

Finally, regular compounds allow ellipsis in coordinations where the second element of the first compound is elided (14a). As discussed in Nübling (2018: 16), this is not possible with PN (14b):

Taken together, this shows that compound-type PN are different from regular compounds. As such, compound-type PN have a special status.

To sum up, while Köhnlein (2015) makes a plausible, detailed proposal for complex proper names in Dutch, it seems clear that the proposed type of compounding is not of the regular kind when compared to ordinary compounds. As will become clear in the next section, the phrasal PN under discussion here also show many regular morpho-syntactic features. Most importantly, the individual parts can occur as independent words and (typically) have regular lexical meanings. In addition, unlike Köhnlein’s compound cases, they seem to be phrasal in nature. Nevertheless, although Köhnlein’s (2015) account does not extend to the current cases, I will make use of many insights of that analysis providing a new but closely related proposal in section 4.1.

3.1 Phrasal proper names are like regular DPs

While this paper is mostly about the morpho-syntax of phrasal PN, let us start with the stress pattern. Comparing the common DP in (15a) to the phrasal PN in (15b), we can point out that there is no difference in the stress pattern:

Similarly, phrasal PN exhibit strings like common DPs. In fact, if we take the interplay of articles, adjectives, and nouns as an example, we can point out that just like common DPs, we only find instances patterning like (16a), where a determiner precedes an adjective and an adjective precedes a noun. There are no cases of (16b–f), where these elements are in an order different from (16a):

The same holds for other elements: only strings similar to common DPs are possible with phrasal PN.

Like common DPs, phrasal PN are integrated in their syntactic context with regard to case. In other words, PN can change their morphological forms. This goes for the change of the article as well as for case suffixes added to the noun. Compare the common DPs in (17) to the phrasal PN in (18). As a rewiewer notes, the genitive inflection can also be left out in (18b) (see also Nübling 2012):⁶

Furthermore, phrasal PN behave like common DPs as regards adjective endings as well. To begin, consider the following common DPs (19). As is well known for German, if a definite article is present, a following adjective has a weak ending; if it is absent, the adjective has a strong ending (for background, see, among many others, Harbert 2007: 134–35 and references cited therein):

Again, like common DPs, adjectives in phrasal PN have a weak ending if the definite determiner precedes them (20a). We briefly mentioned above that phrasal PN can occur without an article. If the article is absent, adjectives have a strong ending (20b):

Thus, phrasal PN behave morphologically like common DPs as well. In fact, it is clear that these aspects of PN are not lexical but are computed during the derivation depending on the syntactic context (i.e., the presence of case assigners and determiners). Similar to the syntactic restrictions on word order above, there are certain cases that are never found, neither with common DPs nor phrasal PN. Schematically illustrating, there are no instances in German where an adjective with a strong ending is preceded by a definite article (21a), or where an adjective with a weak ending is not preceded by a definite article (21b).⁷ In addition, German does not exhibit instances (in the nominative masculine/neuter and the accusative neuter) where an adjective with a weak ending is preceded by an indefinite article (21c):

To take stock thus far, examining (15) though (21), we notice that there are no unexpected patterns in phrasal PN in German, syntactically or morphologically (or phonologically). We can side then with Payne & Huddelston (2002: 517), Weber (2004), and many others that all of these PN involve regular DP patterns. This is a strong indication that the phrasal PN discussed here are different from the compound-type PN in section 2.2. If proper names have surface forms similar to common DPs, we are in need of some morpho-syntactic diagnostics to establish their special status as PN. These diagnostics are meant to stand independently of the semantic property of PN being rigid designators. Note that both of these properties, special morpho-syntax and unique reference, along with the way these PN are used in context help language users recognize these strings as PN. First, I discuss the interaction between phrasal PN and determiners, and then I turn to the frozen aspects of phrasal PN.

3.2 Phrasal proper names and determiners

Above, we have seen examples of phrasal PN where articles and determiner-like elements are absent and some where they are present. In the next subsections, I discuss when the article can or must be left out and when it must be present. Finally, I discuss possessives showing that they cannot be left out or substituted and that they cannot enter into Binding relations.

3.3 Phrasal proper names are frozen

In this subsection, we will see that phrasal PN are lexically and syntactically frozen; that is, certain aspects of their forms cannot be manipulated. Allerton (1987: 64–69) observes for English that PN are fixed with regard to morphological number, they tolerate the addition of non-restrictive (non-stressed) modifiers, but they disallow restrictive modifiers, degree words, determiners, or quantifiers. Furthermore, he states that none of the elements of a PN can be substituted by a different element. I illustrate these properties for German and add some more to the discussion.

4.1 Proprialization

Ordinary name giving, sometimes called nomination (Anderson 2003: 354), provides an individual with a unique label, often restricted to a certain social context (cf. given names vs. nicknames). With inherent PN, the name is typically drawn from a common, existing stock (e.g., Anderson 2004: 442). Phrasal PN differ in that they usually involve new names. Thus, while the basic function of naming is the same, the mechanism for phrasal PN must be different: a name is created. Note that all PN must have or receive designations that mark them as PN. This can be seen with common nouns, which are not referential, inherent PN, which are, and some nouns that can function as either. For now, this marking is indicated below as [PN]:¹⁹

The new proposal can be briefly summarized as follows. The individual elements of phrasal PN are taken from the common stock of (typically) regular lexical and functional words, they receive each a referential marker, and they are stored as a set in the lexicon. During the derivation, the individual parts are combined into strings similar to common DPs. Recall that some elements of the PN do not undergo certain syntactic operations. Rather than exempting those elements, I propose that given the presence of markers, all elements of the PN participate in these syntactic operations equally. The net effect is that no individual element can be singled out and that the entire nominal is frozen to certain operations.

In more detail, I propose that the type of naming of phrasal PN involves one linguistic operation. Following Nübling et al. (2015: 16f), I will call this proprialization. Although this is usually conceived of as a diachronic process, here I will focus on cases where a new phrasal name is created (almost) instantaneously, perhaps during an official name-giving act (for some tentative remarks about cases involving a longer-lasting process, see section 5). Specifically, this operation is basically a memorization procedure that marks a set of lexical and functional items as being (part of) a proper name. More formally, there are three parts to this procedure.

First, this operation picks out regular lexical and functional vocabulary items from the lexicon. Second, elements receive – what I have called thus far – a referential marker. Recall though that Köhnlein (2015) argues for a clear bipartide structure of place names in Dutch. Semantically, there is a division of labor between the first part (the referential morpheme with the referential pointer) and the second part (the classifier with the category label [+settlement]). Syntactically, all elements have the feature [+proper]. The structure of the cases under discussion here is different.

There can be one to three or more individual parts. As far as I can tell, all elements contribute to the semantics equally. However, what all phrasal PN discussed here have in common is a head noun. So, in order to avoid redundancy and employing Köhnlein’s general system, I assume that the head noun of the phrasal PN has a referential pointer (i.e., an index) and a category label. This means that the pointer and category label coincide on the same element.²⁰ Furthermore, I propose that with the exception of the definite proprial article, all vocabulary items are assigned the feature [+proper]. As the third part, proprialization collects these marked elements in a set as part of the lexicon. I turn to the lexical entries.

As discussed above, there are three basic cases of phrasal PN: instances involving possessives, cases where the definite proprial article is present in the original context, and instances without such an article. Recall that there are no robust differences between the last two cases in German. I argued above that definite articles are not part of the PN. I propose then that possessive elements are part of the lexical entry of phrasal PN and receive the relevant markings (66a). As to the other two cases, if the article is in the original context, it is part of the stored set, but it does not receive a marking (66b). Phrasal PN without an article in the original context do not have an article in the stored set (66c). The markings are indicated as follows. Leaving out the category label for now, the referential pointer is marked by ↑ and the feature [+proper] by [+p]. Stored sets are indicated by curly brackets, and embeddings are marked by round brackets:

The different assumptions about the proprial determiner are meant to capture the following facts. Since the possessive determiner has the feature [+proper] (66a), it is obligatorily present and cannot be substituted. As to (66b), where the article is present without a feature, it does not belong to the referential part of the PN and can be left out or substituted under certain circumstances. Finally, (66c) does not have an article in the stored set. However, if an article is added for a syntactic reason (i.e., the PN is in argument position), then the presence of the feature [+proper] on the other elements will guarantee that that article must be a definite one.

Above, we also illustrated the possible presence of demonstratives and non-restrictive modifiers. Now, let us assume that the elements with the feature [+proper] are part of the core of the phrasal PN. Other elements belong to the syntactic periphery, and they may be part of the lexically stored set (der in (66b)) or not (demonstratives, non-restrictive modifiers). I discuss the latter elements in more detail in section 4.3.

Leaving the phonology aside again, the lexical entries above can be fleshed out employing Köhnlein’s (2015) system with a few modifications. Like Köhnlein (2015), the lexical entries involve different components. Unlike his proposal, the entries are complex in that they may involve multiple, individual vocabulary items. As an example, consider (66a) from above, where (67a) involves the noun Telefonladen, (67b) consists of the possessive determiner dein, and (67c) is the number feature [-PL]. Recalling that head nouns may be monomorphemic, I assume that the referential pointer and the category label relate to the same element. This is indicated by subscripts in (67a) below. To be clear, the vocabulary entry stored in the lexicon consists of a list of unordered, multi-component vocabulary items involving semantics and syntax (and phonology):

As discussed above, inflectional endings vary with syntactic context. In other words, these elements are not part of the lexical entries but are determined during the regular morpho-syntactic derivation.

The lexical entry for Der SPIEGEL ‘The Mirror’ looks similar to (67), with the exception that the feature [+proper] is missing on the determiner. Furthermore, as this is an expletive element, the determiner has no semantics. The entry for Institut für Deutsche Sprache ‘Institute for the German Language’ is more complex in that it contains an embedding, that is, two nominals, one of them preceded by a preposition. The embedding could be indicated by round brackets, as in the short-hand formulation in (66c) above, or by a different superscript on the elements of the embedding. This will guarantee that the elements belonging to the same nominal are merged together.

Making some adjustments, it seems possible to apply Köhnlein’s (2015) system to a new empirical domain. In my view, this provides confirmation that this system is on the right track. Interestingly, there is an intriguing difference. Köhnlein’s cases involve compound PN with special properties that require some aspects of the structure to be part of the lexical entry. Note again that the structure is specific to the individual lexical items. However, the stored structure is at the word level, that is, minimal in size. The cases discussed here involve sets of several individual lexical items yielding a complex entry. The structure is not part of the lexical entry but derives from the general syntactic derivation as discussed momentarily. While this may seem like a trade-off, it is possible that both empirical domains (and analyses) may share some points of contact (see section 5 for some avenues for future research in this regard).

4.2 Derivation of simple phrasal proper names

During the derivation, the indexed items are taken out of the set in the lexicon and are merged in a regular fashion (in fact, one may speculate that the stored set is taken out of the lexicon as a whole and functions as part of the Numeration in the sense of Chomsky 1995 and much subsequent work). As in common DPs, nouns project NPs, number specification projects a NumP, adjectives are in Spec,AgrP, and determiners reside in the DP (for detailed background discussion, see Julien 2005, Alexiadou et al. 2007, and many others). Furthermore, I assume that the syntactic feature [+proper] projects from the head to its corresponding phrase (68a) and that phrasal elements like adjectives transfer their feature by Spec-head agreement and subsequent projection of the feature to the hosting phrase (68b):

The result is that all parts of the syntactic tree receive the feature [+proper]. An exception to this is the DP-level when the determiner does not have the feature as in (66b).

With that in mind, I derive the example in (66a) in the familiar way yielding the following simplified structure:

The example in (66b) is similar to (69) but without the feature [+proper] in the DP-level. Turning to the more complex case in (66c) and proceeding bottom-up in (70), the innermost subset indicated by round brackets in (66c) is built first until all these elements are merged. This is followed by the remaining elements thereby embedding the first nominal under the head noun Institut ‘institute’. This yields the following structure:

In certain aspects, phrasal PN are regular definite expressions. This was illustrated above in that they require an overt article when in argument position. In that case, a DP-level is projected in (70). Again, the feature [+proper] ensures that only a definite article will surface. Furthermore, I showed that the definite article can and sometimes must be left out. Nevertheless, the phrasal PN remains referential. If we assume that the DP-level is also absent in those cases, then we can suggest that the DP-level is not related to the referentiality of phrasal PN. The referential pointer brings about referentiality.

Consider how the above-mentioned properties of phrasal PN are accounted for. Proprialization creates a fixed set of vocabulary items yielding complex lexical entries. This accounts for the facts that phrasal PN are lexically frozen (no element can be added, removed, or changed inside the stored set),²¹ it explains the hybrid character as regards phrasal PN being rigid designators with descriptive content (regular vocabulary items get the feature [+proper] and nouns also a referential pointer), and it allows formally indefinite nominals to receive a definite interpretation (phrasal PN are referential and thus definite). Furthermore, note that once the elements are part of the stored set, (almost) all elements are marked by the feature [+proper]. This makes them diacritically different from their ordinary counterparts. Given this marking, they can now undergo long-term changes, semantically and in other ways (Nübling et al. 2015: 50ff).

It is evident that linguistic operations involving phrasal PN are sensitive to certain conditions. As seen above, ellipsis of the head noun is possible when two phrasal PN are coordinated. A reviewer points out that such coordinations involving a phrasal PN and a common DP yield marked results (the first case seems to be a bit worse than the second; the data are due to the reviewer):

Recalling that the head nouns of phrasal PN have the feature [+proper], the above effect follows if we assume that ellipsis must involve identical elements. This provides evidence that linguistic operations are sensitive to the feature [+proper].

As for the syntactic properties of PN, it has been well known since Longobardi (1994) that (inherent) PN are syntactically special. Unlike common nouns, PN may undergo movement to a higher position in Italian. Compare (72b) to (73b). Longobardi argues that PN move to D in cases like (73b):²²

Having argued above that proper names, inherent and phrasal, have the feature [+proper], it is clear that syntactic operations involving movement are also sensitive to this feature.

Now, note again that all elements in (70) have the feature [+proper]. Metaphorically speaking, one might say that all elements are concatenated by this feature. I propose that the presence of these features makes all these elements essentially identical as regards certain operations; that is, operations involving movement apply to all elements equally such that no subpart can be singled out. This explains why elements inside these DPs cannot be reordered and that these PN are islands for subextraction. The presence of this feature also restricts the behavior of possessive determiners.²³ This includes the fact that these possessives cannot establish Binding relations. In other words, certain syntactic operations are severly constrained. However, considering (70), we can observe again that phrasal PN wind up with a regular DP structure yielding ordinary word order patterns. Furthermore, this type of regular derivation immediately accounts for the transparency of phrasal PN to syntactic operations like agreement in gender, number, and case. It is also consistent with the fact that phrasal PN have the same inflectional alternation on adjectives as common DPs.

4.3 Derivation of more complex phrasal proper names

In section 3.3.1, we have seen that phrasal PN have a core that is lexically frozen with regard to the presence of its elements – no restrictive modifier can be added, and no element can be substituted or removed. The core excludes the (expletive) definite article, which can be left out or substituted by a demonstrative. This holds whether or not the article is part of the original context (i.e., in the lexically stored set). I also illustrated there that non-restricted adjectives and relative clauses can be added to the left and right peripheries of PN.

It is worth pointing out that other elements can be added as well, namely inherent and phrasal PN tolerate the occurence of unstressed, non-restrictive possessives in the left periphery:

Often used in exclamations, these possessives establish a close emotional relation to the name and its bearer.

For expository purposes, we can parse the aforementioned complex nominal strings into the core and the pheriphery yielding two domains. Focusing on the left pheriphery, (76a) contains a definite article and a non-restrictive adjective and (76b) a demonstrative or possessive. This yields the delineation in (76c):

While not part of the referential domain of the PN, it is clear that the elements in the left periphery belong to the same nominal structure as the core. This can be deduced from the fact that all adjectives, non-restrictive and those part of the core, must have a weak ending if a relevant determiner precedes them:

Overall, this indicates that these phrasal PN have a regular syntactic structure in German, even in the transition from one domain to another (see Roehrs 2015). The example in (77a) can be illustrated in (78). The definite article is an expletive determiner in D, and the non-restrictive adjective is projected above the core (the arch separates the left periphery from the core):

As for non-restrictive relative clauses, I assume that they are right-adjoined above the core. Finally, assuming that demonstratives and possessives are merged in Spec,DP, their presence leads to the absence of the expletive article. This follows from the general restriction that German only tolerates one element in the DP-layer, either in Spec,DP or in D.

To summarize, complex phrasal PN (without embedding) involve a nominal structure with the head noun of the PN at the bottom and the remaining stored elements in higher positions. Items without the feature [+proper] are projected on top of the core of the PN. Consider two possible alternatives that might also explain the data in this paper.

Above, I proposed that phrasal PN involve regular syntactic derivations constrained by vocabulary items with the feature [+proper]. Let us call this type of analysis option A. To account for the above data, one might be tempted to suggest two other options. First, option B would, metaphorically speaking, take a snapshot of the string of words making up the phrasal PN. Here the PN would be lexico-morpho-syntactically fixed, i.e., completely stored in the lexicon (cf. Anderson’s 2003: 386 discussion of “freezing” of place names). This option would work for some surface patterns. However, it would leave unexplained the regular strong/weak alternation of adjectives in German shown in (20a–b) and the optional additions or substitutions in the peripheries.

As a second alternative, option C, one could suggest that vocabulary items and abstract structures, call them templates, are stored in the lexicon (cf. Culicover & Jackendoff 2005: 29). During the derivation, the vocabulary items and templates would be taken out of the lexicon, and the vocabulary items would be inserted into these templates. While not implausible, this option is not very attractive as many different templates would be needed. In addition, many ordinary structural patterns would be stored as templates and thus duplicated in the lexicon. Furthermore, regular syntactic processes (e.g., agreement in general and the strong/weak alternation of adjectives in particular) would have to apply inside frozen templates. Finally, it would not be clear how optional elements could be added or substituted inside a fixed template.

To conclude this section, I provided an account of the basic facts in German proposing regular derivations in conjunction with the operation proprialization extended to the current cases. Employing Köhnlein’s (2015) system, this accounted for the hybrid properties of phrasal PN in German. Additions or substitutions in the peripheries can be explained by more complex structures that involve one nominal consisting of the core and phrasal layers projected on top.