1. Introduction
Any theory of grammar must seek to answer the question what mechanism licenses (only) the grammatical sentences or structures of a given language. One can roughly distinguish between derivational (lexicalist) and constructionist theories (Müller Reference Müller2006, Jacobs Reference Jacobs2008, Finkbeiner & Meibauer Reference d’Avis2016b). A derivational theory—for instance, the approaches usually conflated under the heading of Mainstream Generative Grammar (such as Chomsky Reference Chomsky1981, Reference Chomsky1995)—assumes that linguistic knowledge is rule-based: The properties of complex linguistic expressions can be derived by general rules of grammar. One basic strategy of derivational accounts of grammar is to model syntactic structures as projections of the properties of the words included. That is, the properties of complex expressions are derived via general mechanisms from the properties of its component parts.
By contrast, a constructionist theory—for instance, the approaches usually conflated under the heading of Construction Grammar (such as Fillmore et al. Reference Fillmore, Kay and O’Connor1988; Goldberg Reference Goldberg1995, Reference Goldberg2006; Kay & Fillmore Reference Kay and Fillmore1999; Croft Reference Croft2001)—assumes that linguistic knowledge consists of a large inventory of form-meaning pairings, that is, constructions. The basic strategy of constructionist accounts of grammar is to specify the constructions and classes of constructions of a given language and the relations between them. A constructionist account thus does not derive the properties of complex expressions from the properties of its component parts, but directly states them as properties of the construction.
A phenomenon at the heart of the controversy between derivational and constructionist approaches are schematic idioms (Fillmore et al. Reference Fillmore, Kay and O’Connor1988), that is, syntactic patterns that are lexically partly specified and partly open. Constructionists treat schematic idioms as prime candidates for complex constructions whose meanings cannot entirely be derived from their component parts (see Goldberg's Reference Goldberg1995:4 original definition of constructions: “C is a construction iffdef C is a form-meaning pair <Fi, Si> such that some aspect of Fi, or some aspect of Si is not strictly predictable from C's component parts or from other previously established constructions”).Footnote 1
By contrast, proponents of a derivational account may argue that schematic idioms, as productive patterns of a language, must share some properties with regular patterns of the language and try to derive these properties from more general principles of grammar. For example, while Jackendoff (Reference Jackendoff2008) claims that the schematic idiom illustrated by expressions such as student after student, book upon book (the NPN [Noun-Preposition-Noun] construction) is an idiosyncratic construction violating standard phrase structure rules, Müller (Reference Müller, Engelberg, Holler and Proost2011) argues that NPN may be derived as a special case of a prepositional phrase, assuming additional processes such as postsyntactic phonological copying.
In this paper, I treat the German schematic idiom N hin, N her ‘N here, N there’ as a new test case for the evaluation of different approaches to grammar.Footnote 2 N hin, N her is a productive pattern of German used frequently both in written and spoken discourse. However, no comprehensive linguistic analysis of this construction has been carried out so far. All one can find are sparse references in some early works on German phraseology (for example, Fleischer Reference Fleischer1982) and in some works on concessive conditionals (for example, Leuschner Reference Leuschner, Leuschner, Mortelmans and de Groodt2005, d’Avis Reference d’Avis2016). Standard grammar books, such as Duden Reference Duden2009, do not provide any information on the pattern. An example is given in 1 (all the examples in the paper are from DeReKo, German Reference Corpus, the electronic resource provided by Institut für deutsche Sprache, Mannheim).Footnote 3
-
(1)
N hin, N her is closely related, but not fully equivalent to, English X or no X, which—at first glance—seems to be a good approximation to an idiomatic translation (‘war or no war, …’). The two patterns are not interchangeable in every context, however.Footnote 4 In this paper, I therefore stick to the quasi-literal translation ‘N here, N there’ and refrain from providing an idiomatic translation.
Fleischer (Reference Fleischer1982:137) characterizes the pattern illustrated in 1 as a pattern ‘noun + hin (‘thither’), same noun + her (‘hither’)’ to which he assigns a concessive meaning.Footnote 5 Thus, one of the defining formal features of the pattern seems to be the identity of the two nouns. This raises the question of whether or not N hin, N her is an instance of syntactic reduplication. In this paper, I discuss this issue in greater detail. In particular, I discuss the question whether a rule-based copying approach to N hin, N her is fruitful.
Based on a close analysis of the specific linguistic properties of N hin, N her, I argue for an integrative model of grammar, as proposed more recently by Culicover & Jackendoff (Reference Culicover and Jackendoff2005), Jacobs (Reference Jacobs2008, Reference Jacobs2016), and Jackendoff (Reference Jackendoff2013).Footnote 6 An integrative model does not seek to account for all phenomena of a language in either a constructionist or a derivational manner; rather, it allows for both constructionist and derivational elements in the theory. For instance, the Simpler Syntax model of Culicover & Jackendoff (Reference Culicover and Jackendoff2005) includes not only an inventory of more or less specific constructions, but also abstract constructions and more general principles. Similarly, a “heterogeneous construction grammar” (Jackendoff Reference Jackendoff2013) does not take all linguistic units to be full Saussurean signs; instead, it contains both meaningful constructions and independent principles of syntactic form and semantic structure. In the same vein, Jacobs (Reference Jacobs2008) argues for a grammar that allows for complex phrasal constructions as well as for “projectionist” (derivational) elements, such as word-based valency restrictions and general linguistic (for example, phonological) laws.Footnote 7
A proponent of a purely derivational account may object that an integrative model leads to redundancy of analytical instruments.Footnote 8 I argue in this paper that redundancy in the analytical apparatus might be somehow inelegant, but is inescapable if one wants the grammar to account for phenomena that cannot be accounted for in a purely derivational way without additional ad hoc assumptions (see Jacobs Reference Jacobs2008). The proponent of a purely derivational account may further object that an integrative model is less explanative. Behind this is the idea that a theory is more explanatory if it can keep everything in the same general format, for example, in endocentric binary branching structures. However, such a homogenous theory does not automatically explain more examples, or is psycholinguistically more adequate than a heterogeneous theory. Moreover, homogeneity often is possible only at the cost of assuming empty elements in the grammar.Footnote 9 In general, an integrative account that incorporates both phrasal constructions and abstract rules seems to be equipped with an apparatus powerful enough to explain a broad range of phenomena.Footnote 10
The structure of this article is as follows: In section 2, I give a general description of the linguistic properties of N hin, N her. In section 3, I discuss some theoretical implications of the empirical observation that N hin, N her has a usage with distinct nominals. It is argued that this usage is to be seen as a case of creative modification. In section 4, I investigate the basic question of whether or not N hin, N her can be assigned a regular phrase structure. In sections 5 and 6, I then compare a rule-based copying analysis with a constructionist approach to N hin, N her. In section 7, I work out in detail an integrative account of N hin, N her, deriving the regular properties of the construction from properties of more abstract constructions, and from general syntactic and phonological principles. In section 8, I take the constructionist account one step further, outlining a more general motivation behind the idiosyncratic aspects of the construction from the perspective of a modular interaction of grammar and pragmatics. Section 9 is a conclusion.
2. General Description of N Hin, N Her
N hin, N her is a nonsentential sequence of two strings, each of which is composed of a noun followed by an adverb, the first lexically specified as hin ‘thither’, the second as her ‘hither’. The two strings form a prosodic unit. The main stress in each string is on the adverb: Krieg HIN, Krieg HER.
The prototypical filler categories are bare nouns in the nominative case, as in 1.Footnote 11 The nouns may be extended by adjectives, as illustrated in 2a,b, but are usually not specified by determiners, as in 2c.Footnote 12 I therefore make the generalization that the open slots in N hin, N her are to be filled by nominals, that is, maximal projections of N under a DP-analysis (see Huddleston et al. Reference Huddleston and Pullum2002, Zwarts Reference Zwarts2013).
-
(2)
As the examples show, in the standard case, the nominals are formally and semantically identical. This is also stated in the general description by Fleischer (Reference Fleischer1982), as cited in the introduction. In the next section, I discuss in more detail some cases that seem to violate the identity restriction.
N hin, N her is subject to a binary restriction: It combines exactly two components, no fewer and no more. Consider the contrast between 1 and 3.
-
(3)
Moreover, it is subject to an irreversibility restriction, as shown by the contrast between 1 and 4.Footnote 13
-
(4)
As to its syntactic distribution, N hin, N her cannot appear independently, but only in juxtaposition with a main clause. Juxtaposition may take place in the left periphery, as in 1, in the right periphery, as in 5a, or as a parenthetical, as in 5b. From 5c, it becomes clear that the construction does not have argument status: It cannot appear in the prefield in front of the finite verb.
-
(5)
As an approximation to its semantics, one can say that N hin, N her is the antecedent in a concessive conditional construction by which a speaker conveys that a certain state of affairs p is irrelevant for the truth of another state of affairs q. This may be paraphrased as in 6.
-
(6)
The construction cannot be used independently and therefore does not have illocutionary force on its own. The larger construction to which it is juxtaposed mostly is an assertion, but it can also instantiate other speech acts, such as questions, as in 7.
-
(7)
In actual usage, the construction allows for a number of different interpretations: For instance, it can be used as a mere concessive (p, anyway q), as in 8a. It can also be used as a concessive conditional, often with a fatalistic interpretation (“it's no use”, “it's in vain”), as in 8b. Moreover, there is a mock usage, as in 8c. In this usage, the speaker conveys a dismissive attitude toward something the discourse partner has uttered before (“what you say about X is pointless”).Footnote 15
-
(8)
Taken together, these data clearly show that N hin, N her exhibits a number of specific formal features, including identity of the nouns, binarity and irreversibility, and syntactic disintegration. The crucial question is how the syntactic features map onto the construction's specific meaning: Is there a rule-based relationship between form and meaning, or is one forced to stipulate the meaning of the construction as an arbitrary property? A comprehensive analysis should be able to account not only for the formal properties of the construction, but also for its restricted interpretational range.
3. The Problem of (Non)Identity
So far, I have taken for granted that the two nominals in instantiations of N hin, N her must be identical. This would call for a treatment of N hin, N her as a case of syntactic reduplication, either modeled via a rule-based copying process or, within a constructionist approach, via a specific identity requirement on the construction. However, a quick look into German usage data reveals that identity of the nouns is not mandatory. Consider 9.
-
(9)
Generally, it seems that there are two possibilities for dealing with the (non)identity of nominals in N hin, N her. First, one might assume one single construction (in the descriptive sense of the term), whose arguments may or may not be identical. Under this assumption, a reduplication analysis would be ruled out immediately. What is particularly unsatis-factory with this assumption is that it ascribes both variants the same status. This is against the intuition of native speakers that the two variants, in fact, have a different status, the variant with distinct nouns being in some sense secondary.
Second, one might assume two independent constructions. Under this assumption, one could maintain a reduplication analysis for the construction with identical nouns, while proposing a different analysis for the “nonidentical” construction. An obvious problem for this assumption is that the two variants are structurally parallel and near-synonymous—by both, speakers convey that a state of affairs described by p is irrelevant to the truth of q—which suggests that they are closely related.
I therefore opt for a third analysis. Under this analysis, there is only one construction, which requires identity of the nouns. The usages with distinct nouns are not taken to instantiate a different, independent construction, but to instantiate creative modifications of the established grammatical pattern, performed by speakers/writers to achieve a certain stylistic or discursive effect. This analysis has the advantage of allowing for a unified approach to N hin, N her.
In general, language users have an option of intentionally overriding grammatical restrictions in order to create a specific stylistic effect. This is well-known from text types such as poems or advertisement, where word play and allusion are major linguistic devices. In the domain of phraseology, creative modification is particularly ubiquitous (Nunberg et al. Reference Nunberg, Sag and Wasow1994, Gläser Reference Gläser and Cowie2001, Glucksberg Reference Glucksberg2001). In a way, it seems that frozen-ness actually invites modification.
For the case of N hin, N her, the newspaper data reveal that the main purpose of using the distinct variant seems to be to catch the reader's attention. For example, in 10, the construction is used to introduce an article about the growing popularity of exotic pet animals.
-
(10)
By initially enumerating two prototypical members of the category of usual Western pet animals, the speaker establishes a set that contrasts with the set of exotic pet animals mentioned later on in the text. The construction is used here primarily to catch the initial attention of the reader and to construe textual coherence, because it requires, for a full understanding, a continuation that provides a contrast. Thus, in 10, the main communicative intention of using X hin, X her is not to convey some kind of irrelevance semantics. Neither does 10 require the topic pet animals to be previously established; rather, the construction is used to introduce this topic. Interestingly, the variant with identical nouns does not produce this specific effect. If Dackel hin, Katze her is replaced by Dackel hin, Dackel her, the result is rather odd, because Dackel hin, Dackel her does not evoke a set to be contrasted with the set {Vögel, Reptilien, Spinnen}.
In the newspaper examples explored in this paper, the usage with distinct nouns is very widespread. Nevertheless, it is still recognizable as a noncanonical usage, because its effects largely build upon the feel that the “regular” pattern has been overridden. Furthermore, as is the case with creative idiom modification in general, the data show that the usage of distinct nouns in N hin, N her is not totally unrestricted. Instead, the referents of the two nouns must be conceivable as two instances of one and the same higher category. In being conceptually similar, the nouns, in fact, satisfy the identity requirement on the conceptual side (but not on the formal side). For instance, dachshunds and cats in 10 belong to the category of Western pet animals; energy revolution and nuclear power in 9a are topics that belong to the debate about the change of German energy politics; and ban and court decision in 9b represent legal measures that have been taken to prevent the leftists from demonstrating.Footnote 19 Taken together, these data suggest that the basic pattern to be investigated is the pattern with identical nouns, as described by Fleischer (Reference Fleischer1982). Therefore, the focus of this paper henceforth is on the variant with identical nominals.
4. Phrase-Structural Status of N Hin, N Her
In order to decide whether or not a derivational or a constructionist analysis is more apt to account for N hin, N her, it is crucial to find out if N hin, N her can be assigned a regular phrase structure. If it cannot be assigned regular phrase structure, this would be an argument against a derivational approach. I first address the question of whether or not the construction as a whole is an instance of syntactic coordination; then I discuss the internal structure of the components N hin and N her.
4.1. N Hin, N Her and Syntactic Coordination
If N hin, N her is an instance of a regular phrase, there must be some kind of relation between the two strings N hin and N her. It is obvious that N hin is not part of N her, and N her is not part of N hin. Thus, there is no subordination relation. Rather, these are two structurally parallel strings that appear on the same hierarchical level. The relation between the strings may be analyzed either as an instance of syntactic coordination or as an instance of mere juxtaposition. Only the syntactic coordination analysis states that there is a syntactic relation between the two components. By contrast, juxtaposition is not a syntactic relation, but rather a relation between elements in a text. If one strives for assigning a regular syntactic phrase structure to N hin, N her, one thus has to opt for the syntactic coordination analysis. However, this analysis is not without problems.
While in standard examples of syntactic coordination the two conjuncts are connected by a conjunction, in N hin, N her, there is no connector between the two components. However, this is not per se a problem for a rule-based approach. A standard way to solve this problem would be to assume a phonologically empty conjunction. Empty conjunctions have been argued for in the case of phrasal coordination of three or more conjuncts, as in Jennifer Ø Jason and Justin (Johannessen Reference Johannessen1998).
An empty conjunction analysis might be backed up by the observation that there are related instances in which the connector is overt. First, there is a nonreduplicative variant of the construction, which always exhibits the overt coordinative connector oder ‘or’, as in 11.
-
(11)
Second, one can find examples in which und ‘and’ or oder ‘or’ are inserted into the pattern N hin, N her, as in 12.
-
(12)
However, in order to claim that cases such as 11 are parallel to N hin, N her constructions, one first has to say more about the way in which N hin oder her ‘N here or there’ is related to N hin, N her. This is not trivial. Second, and more importantly, the fact that it is possible to insert a conjunction into the construction in 12 is a rather weak argument, because this process is hardly constrained. For example, one may connect the two sentences in 13a by inserting and, as in 13b. However, one certainly would not take this as sufficient evidence that 13a is a case of phrasal (CP) coordination with an empty conjunction.Footnote 20
-
(13)
Still, a proponent of a rule-based account might hold that N hin, N her is a coordinate phrase with an empty conjunction. The next step is then to determine the correct phrasal structure. There are at least three different options, as shown in 14.
-
(14)
The structure that can be inferred most directly from the input and that is closest to the speakers’ intuitions is 14c. However, the proponent of an X-bar theoretical account might hold that 14a is the correct structure, because it does not violate X-bar theoretical assumptions of binary branching, as 14c does; neither does it ascribe to the left conjunct the power of determining the features of the entire construction, as 14b does. In order to decide on the correct structure, one would need to test for the category of N hin, N her. However, N hin, N her is distributed to categorically unselective positions in the periphery of a main clause. Therefore, one cannot say whether or not it behaves more like YP (for example, an NP) or like &P.Footnote 21 Obviously, the category problem is further complicated by the fact that one does not know what the phrasal category of the basic string N hin (or N her) is. This brings me to the second issue, namely, the internal structure of the components.
4.2. Internal Structure of N Hin and N Her
With respect to the internal structure of the two components, N hin and N her, two questions arise: i) What is the category of hin and her, and ii) which of the two elements is the head, N or hin/her? Let me begin with the first question. In principle, hin and her could be analyzed as adverbs or as verbal particles. Most reference grammars of German opt for the adverb analysis (for example, Zifonun et al. Reference Zifonun, Hoffmann and Strecker1997:1162). A standard assumption is that adverbs may occupy the position in the prefield in front of the finite verb, while particles may not. Also, it is standardly assumed that adverbs can have the syntactic function of adverbials, while particles cannot. As the data in 15 show, hin and her can be fronted.Footnote 22
-
(15)
Furthermore, hin and her have a usage as (parts of) adverbials:
-
(16)
Thus, the possibility of fronting and adverbial usage support an adverb analysis of hin/her.
However, contra the view of prescriptive grammars, Müller (Reference Müller2002) argues that particles of all kinds can, in fact, be fronted.Footnote 25 Following his analysis, fronting is not a decisive criterion. At the same time, hin and her often combine with verbs of direction or motion, which supports their analysis as (verbal) particles. The status of verbal particles is subject to controversial debates (for example, Olsen Reference Olsen1997, Müller Reference Müller2002, Zeller Reference Zeller2002, Jacobs et al. Reference Jacobs, Heine and Külpmann2010). If one analyzes hin and her as verbal particles, the basic question is whether hin and her are heads or parts of the verbal head, and whether they are phrasal categories or not. In sum, the categorial status of hin/her is not a trivial matter.
Second, it is not immediately obvious which of the elements in the sequence N hin is the head. On the one hand, bare nouns normally do not select either adverbs or particles as their complements. While there are DPs with adverbial adjuncts, as in 17, combinations of bare noun + particle, such as 18, are syntactic fragments rather than regular phrases and should be treated differently. Thus, the phrasal status of 18 is unclear; hence the question, which of the elements is the head.
-
(17)
-
(18)
On the other hand, neither adverbs nor particles normally subcategorize for (bare) nouns. Apart from their usage with verbs of direction or motion, hin and her appear in adverbial prepositional phrases, as in 19a, or in word pairs, as in 20.Footnote 26 In 19a, hin is a nonhead, which is shown by the fact that it cannot fulfill the role of an adverbial on its own, as in 19b. This is also shown by the fact that it can be omitted, as in 19c.
-
(19)
By contrast, one might argue that hin/her in 20 (repeated from 16b–d) are heads, based on their distribution: Word pairs such as 20 have the syntactic distribution of adverbials.
-
(20)
Even though neither particles nor adverbs normally subcategorize for bare nouns, in structures such as 18, one is inclined to analyze hin as the element that selects the noun, not vice versa. This is due to the intuition that such structures are derived from sentences, in which hin/her are verbal particles: Gib das Geld her ‘Give the money prt (hither)’; Mach das Licht aus ‘Turn the light prt(off)’. Together with the verb, this particle would select the object NP. This is in line with the analysis of the bare nouns in 18 as accusative nouns.
Obviously, this argument presupposes some kind of ellipsis analysis, where parts of a full sentence are deleted. However, it is not entirely clear how one should apply this analysis to the case of N hin, N her. For instance, one might consider 21 as a potential full sentence.
-
(21)
However, the Deutsches Wörterbuch (DW), a large resource for older stages of German, does not provide structures such as 21 as examples of authentic usage of hin or her. What one does find are examples that confirm an analysis along the general lines of a VP with a direct object NP and a verb of reflecting or dithering, as shown in 22.
-
(22)
Clearly, a problem with an ellipsis analysis of N hin, N her is that the nouns in N hin, N her seem to carry nominative, not accusative case—at least, this is what examples such as …frei er Sturzraum hin, frei er Sturzraum her ‘… empty crash zones here, empty crash zones there’ suggest (see 5a above). Note, however, that the majority of usages of N hin, N her are bare noun usages, which would, in principle, allow for an accusative interpretation as well.Footnote 28
Moreover, from a theoretical point of view, the deletion in 23 seems to violate standard requirements of ellipsis (for example, Merchant Reference Merchant2004), in particular, the requirement that only constituents may be deleted.
-
(23)
Taken together, the data discussed in sections 4.1 and 4.2 suggest that an analysis of N hin, N her as (some kind of) coordination phrase may not be impossible within an X-bar theoretic framework, but it is far from convincing. First, there is not enough evidence supporting the analysis that N hin and N her are syntactically related. In other words, one might be dealing with mere juxtaposition rather than with coordination. Second, it is unclear what the correct phrase structure of a presumed coordination phrase should be. Finally, the internal composition of the two strings seems to defy regular phrase structure rules.
5. A Copying Approach to N Hin, N Her
I have argued above that N hin, N her is a potential instance of syntactic reduplication, as it requires identity of the two nominals. While I have shown that N hin, N her poses a number of problems for an X-bar theoretic phrase structure analysis, it might still be possible to account for the reduplication in N hin, N her via a rule-based approach to syntactic copying. In this section, I first introduce three recent copying analyses of different syntactic reduplication constructions and then make an attempt to apply them to the case of N hin, N her. It turns out that none of the versions of the copying approach provide a tailor-made solution for the N hin, N her problem. More generally, it is argued that a copying approach is not satisfactory as long as it does not systematically account for the specific meaning of the construction.
5.1. Three Versions of the Copying Approach
Copying approaches to syntactic reduplication have been developed within the Copy Theory of Movement (Chomsky Reference Chomsky1995). In the Copy Theory of Movement, traces are treated as unpronounced copies of a moved constituent. Sometimes, the traces (the copies) do not remain unpronounced, but are spelled out. Thus, syntactic copying would be the result of the spell-out of multiple positions in a movement chain. This can be illustrated by 24.
-
(24)
This approach raises the basic theoretical question of what mechanism licenses multiple spell-out. After all, in most cases, only one copy is overtly realized. It is often assumed that multiple spell-out is licensed by an abstract reduplication (RED) feature that needs to be checked (see Kimper Reference Kimper2008).
An influential approach to syntactic reduplication is the approach of Travis Reference Travis, Kim and Strauss2001, Reference Travis and Somesfalean2003. In her (2001) paper, Travis discusses different types of syntactic reduplication, which behave differently as to the role of the RED morpheme. For my purposes, only what she refers to as “phonological reduplication” and “syntactic reduplication” are of interest.Footnote 29
In phonological reduplication, RED is the head of a functional phrase and takes as its sister a head. The need for feature checking triggers head movement, which sets up the appropriate environment for the reduplication. In this case, the functional head is the final position of the copied constituent. An example of this kind of analysis is Ghomeshi et al.'s (Reference Ghomeshi, Jackendoff, Rosen and Russell2004) analysis of contrastive focus reduplication (salad-salad ‘lettuce, not tuna salad’).Footnote 30 Under this analysis, there is an abstract CR morpheme that heads a functional phrase CRP. The CR morpheme has the features [+prototypical, +contrast] and triggers head movement of the adjacent X0. The copy of X0 merges with CR, multiple spell-out results in the reduplication. This is illustrated in 25.
-
(25)
The second type discussed by Travis (Reference Travis, Kim and Strauss2001) is syntactic reduplication, as in 26. In this case, the structure also contains a functional phrase with a RED head, but there is no movement.
-
(26)
Instead, the head creates a copy of its phrasal sister which then appears in the Spec position of the functional phrase. In this case, RED is not the final position for the copy. As examples of this type of reduplication, Travis mentions cases such as English cup of coffee after cup of coffee or Malagasy na iza na iza ‘whoever’ (with iza meaning ‘who’).Footnote 32
A third alternative analysis related to Travis's syntactic reduplication is the analysis proposed by Grohmann & Nevins (Reference Grohmann and Nevins2004) for the case of shm-reduplication in English (for instance, money-shmoney), as in 27.
-
(27)
In this case, phrasal copying results from movement. However, Grohmann & Nevins (Reference Grohmann and Nevins2004) do not assume a specific reduplication projection (as in QP or CRP above). Instead, they assume two functional phrases: a topic phrase and a pejoration phrase. A topic expression (for example, money) is moved from the topic projection to the higher pejoration projection; multiple spell-out results in reduplication. Furthermore, they assume a restriction on the level of PF that requires minimal distinctness of the two copies. This restriction is responsible for the phonological distinctness of the second copy (money shmoney).
All analyses discussed so far concern cases of syntactic reduplication of lexical heads or of phrasal units with lexical heads. Thus, they should in principle be applicable to the case of N hin, N her, which involves reduplication of nouns or NPs. In the next subsection, I apply each of the copying approaches discussed above to the case of N hin, N her.
5.2. Is N Hin, N Her a Case of Syntactic Copying?
A first option is a head movement analysis, as illustrated in 28. Here, I assume a functional phrase with a RED head that triggers copying of the noun. To make the head movement analysis work, one must assume that the noun is the head of the sequence [N her].
-
(28)
Obviously, this analysis generates the wrong output, for example, phrases such as Krieg Krieg her (or, alternatively, Krieg Krieg hin). Thus, it can derive doubling of the noun, but it cannot derive the particle (or adverb) hin that intervenes between the two nominal copies. Furthermore, this analysis can only account for instances of N hin, N her that feature nouns, but not for the instances that feature NPs, as in 2 above.
A second option is spec-filling, as in 29.Footnote 33 The advantage of this analysis is that it can derive phrasal copying. Under this analysis, there is no movement, but a copy of XP is freely generated and inserted into the Spec position.Footnote 34 Here RED is not a landing position. Assuming a syntactic coordination analysis, one may suggest that RED attaches to the (empty) conjunctional head.
-
(29)
While this analysis correctly predicts that the whole string is copied, it results in the erroneous output Krieg her Krieg her (or, alternatively, Krieg hin Krieg hin). Thus, it cannot account for the distinctness of the particles (or adverbs), at least not without substantial additional assumptions.
As a third option, one may consider phrasal movement to a higher functional projection, as in 30. Like shm-reduplication, N hin, N her is also a case of topicalization and can be assumed to be base-generated in Spec, TopP.Footnote 35 Adapting Grohmann & Nevins's (Reference Grohmann and Nevins2004) analysis, one might postulate a higher functional projection IrrelP (Irrelevance Phrase), based on N hin, N her's irrelevance meaning. After movement to the Spec position of IrrelP, the copy that remains in Spec of TopP undergoes a process of phonological change (Krieg hin > Krieg her).
-
(30)
However, the analysis faces a number of problems. For example, it falsely predicts that Krieg hin should be able to appear alone in topic position, which is impossible, as shown by the contrast in 31.Footnote 36
-
(31)
Furthermore, it only accounts for N hin, N her in left-peripheral position, not for its right-peripheral or parenthetic usages. The explanation of the distinctness of hin versus her in terms of a postsyntactic PF process is also unwarranted. Finally, this analysis cannot explain how hin comes into the picture in the first place, as it is not part of the given discourse topic.
Taken together, these observations show that N hin, N her poses a number of technical and theoretical problems for a syntactic copying analysis, which mainly result from the fact that N hin, N her displays intervening material between the reduplicated elements.
These problems cannot be solved easily. In general, nonadjacent syntactic reduplication of lexical categories is rarely analyzed in the Copy Theory of Movement framework. There is a proposal of Kobele (Reference Kobele2008) for the case of English X or no X, which is not very detailed, though. Kobele (Reference Kobele2008:9) attempts to show, in contrast to Pullum & Rawlins (Reference Pullum and Rawlins2007), that X or no X is a case of copying, even if the strings are only partially identical (for example, war with Iraq or no fucking war). The assumption is that copying takes place at a certain point in the derivation. As the derivation continues, further operations may obscure the exactness of the copy relation. In particular, ellipsis and late adjunction are assumed to be at work here. Under Kobele's analysis, the derivation of war with Iraq or no fucking war would proceed as follows: First, the phrase war with Iraq is constructed, then it is copied, then with Iraq is deleted from the second disjunct, followed by late adjunction of the expressive adjective fucking.
In this analysis, it remains largely unclear how the copying process works exactly. A first problem is that it seems to violate the No-Tampering Condition (Chomsky Reference Chomsky, Freidin, Otero and Zubizarreta2008), which says that structure may not be changed after it is copied to another place.Footnote 37 Second, it is unclear whether this copying process involves movement or not. Probably not, because movement from one conjunct into the other is highly unwarranted, as it overrides a syntactic barrier. Furthermore, the Across the Board constraint states that if something is moved from the second conjunct, it should also be moved from the first conjunct, which would be incompatible with the analysis. Next, it is unclear what licenses the copying. Is there any RED feature, and if yes, where is it positioned? Also, Kobele seems to assume that the negation element is not part of the second string, that is, the structure is [X] or no [X]. This is rather unintuitive, however. The point is that if no is part of the second string, then the question is whether or not the two strings are identical at all at any point in derivation.
Returning to the case of N hin, N her, the crucial problem is maybe not so much the question of whether or not a copy analysis is technically possible. There certainly are technically sophisticated frameworks that might provide solutions for the problems mentioned above. The most fundamental flaw of a copy analysis is that there is nothing in a process of copying that predicts the resulting irrelevance meaning. In other words, the irrelevance aspect associated with the construction does not follow from the copying.
Thus, it appears that an analysis of N hin, N her within a rule-based framework of syntactic copying is not convincing. Not only is the copying analysis not straightforward, but also the basic assumption that N hin, N her can be assigned phrase-structural status is doubtful. Therefore, in the next section, I discuss an alternative, constructionist approach to N hin, N her.
6. A Constructionist Approach to N Hin, N Her
In this section, I sketch an alternative, constructionist approach to N hin, N her, taking Jackendoff's (Reference Jackendoff2008) analysis of the NPN construction as a model. It is argued that the constructionist approach is able to account for both regular and idiosyncratic features of N hin, N her in a systematic fashion. In general, a constructionist approach can cover idiosyncratic aspects of a construction by directly mapping meaning onto form. At the same time, the regular aspects can be accounted for in a “heterogeneous” constructionist approach by specifying the relations of the idiosyncratic construction to other, more general constructions and to overarching, general principles of grammar.
An example of a constructionist approach to syntactic reduplication is Jackendoff's (Reference Jackendoff2008) analysis of the NPN construction (for example, student after student, page by page, book (up)on book). This reduplicative construction may be used either as an argument (Student after student entered my office) or adverbially (He went through the book page by page), and comes with a variety of meanings, a central one being “succession”. Focusing on the empirical side of the phenomenon, Jackendoff argues that a rule-based, X-bar theoretical account is not feasible, as the construction “violates standard principles of phrase structure” (Reference Jackendoff2008:8).Footnote 38 In particular, he argues, it is unclear what syntactic category the NPN construction is—an NP or a PP—as the adjunct positions in which NPN can appear are categorically unselective. Also, the internal structure of NPN is puzzling, as prepositions normally do not take bare nouns as their complements, nor as their specifiers.
As an alternative to the X-bar approach, Jackendoff suggests a constructionist analysis. Under this analysis, there is a lexical entry for the NPN construction that maps meaning directly onto form, as in 32.
-
(32)
In the meaning representation, there is only one entity that is mapped onto two nouns in syntactic representation (index i). The preposition in syntax maps onto after in phonology but not in semantics; that is, the preposition does not contribute directly to the “succession” meaning. This is the noncompositional aspect of the construction. Thus, Jackendoff stipulates a constructional schema for NPN that imposes an identity restriction and is associated with the semantics of the entire construction.Footnote 39
Applying this analysis to the case at hand, one may conceive of N hin, N her as a syntactic schema with two open slots for whose fillers identity is required, and to which an irrelevance meaning is assigned. This schema may be represented as follows:
-
(33)
As in 32, there is only one entity in the meaning representation of 33 that is mapped onto two nominal constituents in syntactic representation. This provides the desired interpretation, which features only one referential entity (for example, Regardless of the war , the show must go on). The adverbs in syntax are not mapped onto semantics, only onto phonology. This is the idiomatic aspect of the construction.
Obviously, if one adopts 33 as the representation of N hin, N her, one does not have a problem with the fact that N hin, N her cannot be assigned a regular phrase structure. As there is no claim that the properties of the entire construction are projections of the head—instead, these properties are specified for the construction as a whole—one does not have to assign head status to any of the constituents. Here, I simply treat N hin, N her as an instance of the category U (utterance).
The constructionist approach fits nicely in with the intuition that N hin, N her is a frozen phraseological pattern. The constructional template states that this pattern is fixed in exactly this shape, with exactly the phonological, syntactic, and semantic features it has. However, it would be unsatisfactory to stop here. If one were to stipulate the features of N hin, N her in a template, one would reduce the construction merely to an arbitrary linguistic sign. However, not every feature of N hin, N her is arbitrary. Rather, there are several features in N hin, N her that correspond to well-established features of German grammar.
While constructionist approaches are often criticized for simply stipulating features, most major constructionist approaches, in fact, capture generalizations in the grammar.Footnote 41 Mostly, this is accomplished via default inheritance relations (for example, Goldberg Reference Goldberg1995, Reference Goldberg2006; Kay & Fillmore Reference Kay and Fillmore1999; Sag Reference Sag, Boas and Sag2012). Goldberg (Reference Goldberg2013:21) sketches a network of constructions into which the P N construction (to bed, in prison) is embedded, as shown in 34. The P N construction is peculiar in that it does not involve a determiner and is associated with a meaning “stereotypical activity”. However, it also has regular features: For instance, its word order follows the word order of the more general PP construction. More specifically, the PP construction reflects the fact that English has prepositions instead of postpositions.
-
(34)
While not detailed in Jackendoff's (Reference Jackendoff2008) paper, NPN could also be modeled as part of an inheritance hierarchy, thus covering the properties that can be related to more general constructions and to overarching principles of grammar. In Jackendoff Reference Jackendoff2008:17, one finds the following internal inheritance hierarchy for the NPN construction with its subconstructions (see also Jackendoff Reference Jackendoff2013:87):
-
(35)
In 35, the subconstructions to the left are idioms; the bold-faced subconstructions are fully productive; the N to N subconstruction is a case of what Jackendoff refers to as “semi-productive”. While Jackendoff does not specify the ways in which the general NPN construction is embedded into the larger network of other, more general constructions, the idea is that such a network should be possible in principle to model.
Overall, a constructionist approach that incorporates some kind of inheritance network provides an apparatus powerful enough to account for both, idiosyncratic and regular aspects of a given construction. However, as I argue below, the constructionist approach may be taken one step further if one also allows for its interaction with general pragmatic processes and principles. I show that this step also makes it possible to account for the idiosyncratic aspects of the construction.
7. An Integrative Account of N Hin, N Her
In 33, I stated a constructional template for N hin, N her which matches the specific formal configuration with a specific constructional meaning. In this section, I work out in more detail the hierarchy of constructions and more general principles into which the construction is embedded. In sketching this constructional network, one can account for its regular features.
My basic assumption here is that N hin, N her is not an instance of syntactic coordination, but an instance of mere juxtaposition of two strings. That is, I do not state a syntactic relation between the two strings, but rather a relation on the level of text, plus a high degree of conventionalization. From a surface-based perspective, relevant cases to which N hin, N her may be compared are expressions of the kind illustrated in 36.
-
(36)
The examples in 36 are parallelisms, which are stored as multiword expressions in memory. Each of them is a configuration of two structurally equivalent and semantically contrastive strings, with 36a being a pair of NPs; 36b a pair of past participles; 36c a pair of fragments, each of which is composed of an adverb and an interjection; and 36d a pair of sentences. The most plausible account of these structures treats them as frozen discourse units: They are not connected syntactically, but they are conventionalized as pairs, which can be seen in the fact that they are binary and irreversible, as shown in 37.
-
(37)
Now, if N hin, N her is an instance of a syntactic parallelism, how can one represent its internal composition, and how is it related to other constructions of German? As already stated in 33 above, I suggest the structural pattern in 38.
-
(38)
While an X-bar theoretic approach, as pointed out above, cannot deal with the structure of the substrings [N hin] and [N her] in a straightforward way, these substrings are not, from a constructional perspective, totally ill-formed or impossible. First, there is a well-established construction in German that may be referred to as the resultative directive. This is the construction illustrated in 39.Footnote 42
-
(39)
The examples in 39 are utterances conventionally associated with a directive illocutionary potential. They are composed of a noun functioning as THEME argument followed by a particle that denotes the endpoint, as in 39a,b, or the result, as in 39c,d, toward which the THEME argument is to be moved.
Second, there is a well-established directional adverb construction, as illustrated in 40.
-
(40)
The expressions in 40 are compounds composed of a noun denoting a type of landscape (Berg, Fluss) followed by a directional adverb (or particle) denoting a direction (auf, abwärts). As a whole, they have a directional meaning (for example, Sie folgten dem Tal flussabwärts ‘They followed the valley downstream’).
Both 39 and 40 can be analyzed as instantiations of a more general syntactic construction [N Part], which is independent of semantics.Footnote 43 To make this work, one must adopt a “heterogenous” constructionist approach, as outlined in Jackendoff Reference Jackendoff2013. Under this approach, meaningful constructions “are just one kind of abstract stored structure. The grammar of a language can also contain independent principles of syntactic form, as well as independent principles of semantic structure” (Jackendoff Reference Jackendoff2013:79). Thus, [N Part] can be analyzed as a piece of stored syntactic structure that does not specify any semantics and appears at the top of the inheritance hierarchy in 41.
-
(41)
According to this hierarchy, the syntactic configuration [N Part] may be instantiated either as an independent utterance, conveying the meaning “resultative directive”, or as an adverbial complement, conveying a directional meaning.
Now, the question is how N hin, N her fits into this hierarchy. First, note that the directional [N Part] construction may also appear as a pair of syntactically equivalent, semantically contrastive strings, as in 42. In this pair, the noun is duplicated and combined with particles denoting opposite directions, such as up/down.
-
(42)
While the construction in 42 is still directional, it differs from 40 in adding a meaning component of “exhaustive extension”, which is correlated with the doubled form. For example, landauf, landab means ‘all over the country, everywhere’. The same is possible with temporal expressions, such as tagein, tagaus meaning ‘every day’, as in 43.
-
(43)
These cases may be added to the hierarchy in 41, yielding 44.
-
(44)
Now, N hin, N her relates to landauf, landab not only in instantiating the same paired syntactic string of N followed by opposite adverbs/particles; it also has the same meaning component of exhaustive extension. By using N hin, N her, a speaker conveys the exhaustion of the cognitive space of alternative ways to look at something.Footnote 45 Thus, this pattern fits straightforwardly into the inheritance hierarchy, as in 45. As to the semantic relations, it is plausible that the meaning variants of spatial extension, temporal extension and cognitive extension are related to each other via metaphorical links.
-
(45)
Having established this partial inheritance hierarchy, it is possible to check which of the properties of N hin, N her can be derived from more general constructions or principles, and which should be treated as idiosyncratic. Recall the following general properties of N hin, N her (see section 2): i) identity; ii) binarity; iii) irreversibility; iv) main stress on the adverbs; v) no determiner; vi) word order N > Adv; vii) syntactic disintegration; viii) irrelevance meaning.
The constructional network in 45 already states the syntactic configuration of the construction discussed here: As the more general N Part construction, the subconstruction N hin, N her is a (doubled) configuration of a determinerless noun and a particle/adverb. Thus, property v) (no determiner) is an inherited property. Also, the word order restriction as per vi) (N > Adv) is inherited from the more general N Part construction. Again, it could be suggested that this word order is related to the more general word order NP > verbal particle in a main sentence, where the (verbal) particle remains in situ. This is straightforward for cases such as [ U N Part], for example, Geld her (<[Sie gibt] das Geld acc her ‘[She hands] the money over’), and [ Adv N-Part], such as bergauf (< [Sie geht] den Berg acc [hin]auf ‘[She goes] up the hill’, lit. ‘[She goes] the hill up’). However, for N hin, N her, this analysis is only straightforward under the assumption that N (originally) is an accusative, for example, under a reanalysis assumption (see section 4, note 28). In any case, as long as the word order N > Part is stated in the general [N Part] construction, N hin, N her inherits the correct word order.
The stress pattern of N hin, N her (property iv) is the same as the stress pattern of the more general [ Adv N-Part] construction. It follows a general phonological rule, according to which it is the contrasted element that carries stress in contrast pairs.Footnote 46 Note that the stress pattern does not follow from the abstract syntactic [N Part] construction; rather, [N Part] allows for different stress patterns, as shown in 46.
-
(46)
The irreversibility constraint (property iii), that is, hin > her, follows a general phonological restriction on syllable prominence (Müller Reference Müller1997:33, Lenz Reference Lenz, Hartmann and Wirrer2002; see also Ross Reference Ross1980). It states that a high nuclear vowel of a syllable dominates a low nuclear vowel of another syllable, with the dominating syllable sequentially preceding the dominated syllable. The validity of this principle can be observed, for example, in binomials and reduplicative words, as in 47 and 48, respectively.
-
(47)
-
(48)
Based on these considerations, the structure in 49 emerges.Footnote 47
-
(49)
What cannot be derived from the constructional network and from more general principles are the following properties of N hin, N her: i) identity, ii) binarity, vii) syntactic disintegration, and viii) irrelevance meaning.
First, it does not follow from a general rule of juxtaposition that the juxtaposed nouns must be identical (property i). Neither does it follow that one must have exactly two nouns (property ii). Second, it is not predictable from its structural composition that N hin, N her only can occur in the periphery of a sentence (property vii). After all, there are other instances of [N Part] that can occur in integrated position, namely, the adverbial cases (for example, bergauf or treppauf, treppab). Finally, it is obvious that the complex irrelevance meaning of the construction (property viii) cannot be derived simply from the meaning of its component parts. For example, there are two occurrences of the same noun in syntactic structure, but the noun contributes only once to the computational system. More generally, the problem is that syntactically, it is at most a syntactic fragment, but semantically, it is a proposition.
Obviously, the constructional approach can handle all of these nonderivational aspects by stating a constructional schema, which specifies the idiosyncratic properties of the construction. This is what was done in 33 above. However, I think that it is possible to go one step further and account for the idiosyncratic aspects of the construction as well. This can be accomplished if pragmatics is integrated into the picture.
8. A Pragmatic Perspective on N Hin, N Her
The constructional approach covers as many properties of the construction as possible via general regularities. The remaining properties are viewed as idiosyncratic aspects of this specific construction. What is still missing, however, is a more general motivation behind the construction. In short, why does the construction look the way it does? I argue that many of the idiosyncratic properties of N hin, N her can be explained by taking a closer look at different pragmatic domains that play a role here. A theoretical premise for such an enterprise is that one must adopt a modular theory of language in which grammar and pragmatics are thought of as independent but systematically interacting modules (for example, Bierwisch Reference Bierwisch, Searle, Ferenc and Manfred1980).
In Jackendoff's (Reference Jackendoff1997) tripartite parallel architecture, pragmatics is not a separate component. Thus, it is unclear how pragmatic principles or inferences become active in this model. The same is true for most constructionist approaches to grammar: They usually do not incorporate a systematic interface with pragmatics, but rather are conceptualized as nonmodular theories (Goldberg Reference Goldberg2013:16). This means, first, that a typical constructionist analysis represents both semantic (truth-functional) aspects and aspects traditionally thought of as pragmatic (for example, information-structural status, speech act force) as part of the meaning of a construction (Lakoff Reference Lakoff1987, Lambrecht Reference Lambrecht1994; see Nikiforidou Reference Nikiforidou, Brisard, Östman and Verschueren2009). Second, standard constructionist approaches do not incorporate a theoretical explication of the status of general pragmatic principles in the theory (as developed, for instance, by Grice Reference Grice, Cole and Morgan1975; for a recent discussion, see Ariel Reference Ariel2016).Footnote 48 In the following, it is shown that reference to pragmatic categories is indispensable if one wants to gain a deeper understanding of the idiosyncratic aspects of N hin, N her.
First, a pragmatic approach can account for property i), the identity of the nouns in N hin, N her. I rejected a formal-syntactic copying analysis of syntactic reduplication above. Instead, I would like to argue that doubling in N hin, N her is based on an operation of utterance repetition. The main difference between reduplication and repetition is that reduplication is a restricted, obligatory process located within the domain of grammar, while repetition is a pragmatic option used by speakers to create certain discursive effects (see Maas Reference Maas and Hurch2005, Stolz Reference Stolz, Ammann and Urdze2007, Stolz et al. Reference Stolz, Stroh and Urdze2011). Thus, while the effects of most well-known grammatical reduplication processes are on the level of grammatical functions or denotational semantics, repetition yields effects on the level of speaker attitude and illocution. For instance, an utterance such as “A fox, a fox is in the garden!” (Sperber & Wilson Reference Sperber and Wilson1995) does not convey that there are two foxes in the garden; instead, the speaker introduces an emotional component into the speech act performed. Similarly, it is on the level of speaker attitude and illocution where the irrelevance aspect of N hin, N her is to be described. This aspect is not a referential (propositional) type of meaning, which would be prototypical for reduplication.
From a repetition analysis, it also follows readily that the doubled noun contributes only once to the computational system. A linguistic expression that is repeated does not undergo changes in its structural composition or denotational semantics (see Knowles Reference Knowles1979). Thus, the utterance “A fox, a fox is in the garden!” is not an instance of uttering the syntactic structure a fox a fox; rather, the NP a fox is used twice. This is exactly what I described above as juxtaposition, as being different from coordination. From the viewpoint of classical rhetoric, of course, parallelism is nothing other than a figure of speech that belongs to the broader class of figures of addition and reordering (see Göttert Reference Göttert1991:44), to which figures of repetition also belong (for example, asyndeton, antithesis, chiasm).
Second, pragmatics provides a straightforward explanation for property viii), the general irrelevance meaning of N hin, N her. On the one hand, this meaning can be related to the repetition of the nouns. It is a well-known function of repetition to convey disaffiliation with the addressee (Tannen Reference Tannen1989, Bamford Reference Bamford2000, Bazzanella Reference Bazzanella2011). In particular, mock repeats can be used by speakers to convey a dismissive attitude toward a phenomenon, expressing that the phenomenon in question is of no relevance, as shown in 50.
-
(50)
On the other hand, the irrelevance meaning can be related to the pragmatic domain of deixis. Hin and her are indexical expressions, which incorporate reference to the speaker. Relative to the position of the speaker, hin and her point in two opposite directions. This brings about an interpretation under which the object referred to by the nominal is turned around and looked at from different, maximally contrasting perspectives. Hin and her thus can be conceptualized as contributing an aspect of exhausting the set of potential alternatives. At the same time, there is also an aspect of vainness, in that the reflection does not lead to the desired outcome.Footnote 49 Both aspects are in direct connection with the resulting irrelevance aspect (none of the potential preconditions p will influence the truth of q). From the conceptual requirement of contrasting two perspectives (property ii), the binarity restriction also follows readily: For a contrast to be established, what one minimally needs are exactly two contrasting elements.
Finally, by reference to the pragmatic domain of information structure one may account for property vii), the syntactic disintegration of N hin, N her. This positional restriction can be taken as an indicator of the information-structural status of N hin, N her as a discourse topic. A peripheral syntactic position is characteristic of topical constructions. Consider, for example, hanging topics, as in 51 and 52.
-
(51)
-
(52)
The construction can only be used felicitously if the discourse referent encoded by N is old information, either previously mentioned or provided by shared background knowledge of the discourse participants. While the N hin, N her part is correlated with the discourse topic, the juxtaposed main clause is correlated with a comment, or discourse-new information. The information-structural properties also have an impact on the distribution of communicative weight. According to Leuschner (Reference Leuschner, Leuschner, Mortelmans and de Groodt2005:296), the communicative weight in concessive conditionals is on the information presented in the second component (q), while the information in the first component (p) is “pragmatically downgraded” (see also Brandt Reference Brandt and Motsch1996).
Generally, it is reasonable to conclude that the construction is a piece of “frozen discourse”. As such, it is not a free discourse option any longer; rather, the syntactic and semantic properties of N hin, N her are conventionalized. This is in line with general assumptions of grammati-calization and pragmaticalization theory (Traugott & Trousdale Reference Traugott and Trousdale2014). Future work is needed to justify those assumptions empirically, by tracing N hin, N her back to earlier stages of German.
9. Conclusion
In this article, I have investigated in detail the formal and interpretational properties of the German schematic idiom N hin, N her ‘N here, N there’, treating it as a new test case for the evaluation of competing approaches to grammar. My conclusion is that a heterogeneous constructionist approach is most apt to capture both regular and idiosyncratic features of N hin, N her in a systematic fashion. In general, the constructionist approach can cover idiosyncratic aspects by directly mapping meaning onto form. Regular aspects can be accounted for, in a heterogeneous constructionist approach, by specifying those properties of the construction that can be related to properties of other, more general constructions, as well as to overarching, general principles of grammar. I have argued, however, that the analysis should not stop here, but should also seek a more general explanation for the idiosyncratic features of the construction that cannot be derived in grammar. This can be achieved within a theory that incorporates a systematic interface with pragmatics. As pragmatic domains and processes relevant to the case of N hin, N her, I have identified repetition as a conversational practice, deixis, and information structure. It is a task of future research to show whether or not a process of grammaticalization is involved in the development of N hin, N her.
Knowles (Reference Knowles1979:654) once remarked that processes of iteration are not linguistic in the narrow sense of being “reflexes of the formal system of rules we call ‘grammar’, but are in fact special stylistic phenomena, though based on linguistically defined data.” One may take the reverse to be possible as well: The special stylistic uses become grammaticalized via frequent usage, thus making their way back into the “formal system of rules we call ‘grammar’”.
