Adverbial reinforcement of demonstratives in dialectal German

In the German dialects of Rhine and Moselle Franconian, demonstratives are reinforced by locative adverbs do/lo ‘here/there’ in order to emphasize their deictic strength. Interestingly, these adverbs can also appear in the intermediate position, i.e., between the demonstrative and the noun (e.g. das do Bier ‘that there beer’), which is not possible in most other varieties of European German. Our questionnaire study and several written and oral sources suggest that reinforcement has become mandatory in demonstrative contexts. We analyze this grammaticalization process as reanalysis of do/lo from a lexical head to the head of a functional Index Phrase. We also show that a functional DP-shell can better cope with this kind of syntactic change and with certain serialization facts concerning adjoined adjectives.


INTRODUCTION
In colloquial Standard German, the only difference between definite determiners (1a) and demonstratives is emphatic stress in demonstrative contexts (1b).
Kannst du die Frau SEHen? Can you det woman see 'Can you see the woman?' b.
Kannst du DIE Frau sehen? Can you dem woman see 'Can you see this woman?' Due to this functional overload, locative adverbs such as da/dort or hier 'there, here' help to reinforce the deictic strength of the demonstrative determiner. This kind of reinforcement is common in many Germanic and Romance languages (see Brugè 1996;Bernstein 1997;Putnam 2006;Roehrs 2010). The position of these adverbs in colloquial Standard German is restricted to peripheral positions, i.e. preceding the determiner (2a) or following the noun (2b). The intermediate position between determiner and noun yields an ungrammatical construction (2c).
give me adv dem beer 'Give me that beer (there).' b.
Gib mir das Bier da.
give me dem beer adv c. *Gib mir das da Bier.
give me dem adv beer Similar to Putnam's (2006) observations on Pennsylvania German, in the transition area between the German dialects of Rhine Franconian and Moselle Franconian, 1 the position of locative adverbs do/lo is not restricted to peripheral positions (3a, b). Additionally, they can appear between the demonstrative and the noun (3c). Rhine Franconian uses the phonetic variant do of Standard German da, while Moselle Franconian has established the lexical variant lo. Both variants, however, derive from the same Early New High German form allda 'there'.
Moselle Franconian (Fox 1955: 27) weil kommt lo die dreckig Kurwel dazeschen now comes adv dem dirty basket between 'Now this/that damn basket is getting in the way.' b.
Moselle Franconian (Braun & Peter 1999: 57) Awwer […] wat machschde da mét der Hónd lòò? but what make.you then with dem dog adv 'But what are you going to do with this/that dog then?' c.
Rhine Franconian (Lang 2011: 25) Fier all die wo unner der doo Hitz stöhne for all those who under dem adv heat moan 'For all of those who suffer from this/that heat.' In contrast to Girnth & Michel (2008: 205) who state that the intermediate position is on the decline, we posit, on the basis of abundant evidence (see below), that this position has become mandatory in the transition area between Rhine and Moselle Franconian and that peripheral positions cease to be used in demonstrative contexts. Dialectal texts, grammar books, audio recordings and informants show that speakers use dem+do/lo+N in the same contexts in which the Standard German demonstrative equivalent dies-'this/that' applies (see Rauth & Speyer 2018: 46-47). Thus, a new complex demonstrative has been grammaticalized in the area between Rhine and Moselle Franconian. This assumption is also supported by the fact In the present paper, we present data from a preliminary questionnaire study conducted in 2019 in order to identify the exact area where speakers use the new demonstrative (Section 2). We explain the emergence of dem+do/lo+N as a typical instance of grammaticalization involving the neutralization of semantic features of the locative adverb as well as an increase of bondedness between the demonstrative and do/lo (Section 3). As for the structural analysis, we act on the assumption that the lexical head do/lo has been reanalyzed as the head of a functional projection denoting deictic indexing (Section 4). Finally, we argue that the emergence of the new demonstrative can best be accounted for by the DP-hypothesis (Section 5).

QUESTIONNAIRE STUDY
In the course of a project in which we are going to investigate the syntactic peculiarities of the Rhine and Moselle Franconian varieties spoken in the border region of Germany, Lorraine and Luxembourg, we conducted a questionnaire study as a pretest. The questionnaire was sent to 121 townships, which are evenly distributed over the area of investigation. Ideally, six participants for each township were asked to complete the questionnaire, of which three should be over 60 years old, and three under 30 years old. This was done in order to trace possible changes in apparent time (see Labov 1994). The questionnaires were regionalized and there were three versions of the instructional text (German, French, Luxemburgish). Thus, the questionnaire came in seven versions (Southern Rhine Franconian: one version with French and one version with German instruction text; Moselle Franconian: one version with French and one version with German instruction text; Luxembourgish: one version with Luxembourgish instructional text). So far, we have received completed questionnaires from 36 townships, which are situated mostly in the German state of Saarland and the French department of Moselle, as well as one township in Luxembourg. Furthermore, only a subset of the expected six questionnaires per township has been received. The study is based on these questionnaires.
We decided to use multiple question types in the questionnaire. The tasks range from different types of judgment elicitation (selecting the best variant; rating variants on a 4-pointscale) to puzzles, cloze tests, descriptions of single pictures or of picture sequences. Our questionnaire also included questions regarding many other linguistic phenomena of Rhine and Moselle Franconian, which, at the same time, served as filler questions for the reinforced demonstratives. In that respect we follow the example of the project Syntax Hessischer Dialekte (SyHD, 'Syntax of Hessian Dialects'). The investigators involved in SyHD had good results with this broadness of tasks, so we are expecting good results for our project as well (see Fleischer, Kasper & Lenz 2012). One reason which guided us in making this decision was that the tasks address different aspects of language: While acceptability judgments pertain to competence, the actual performance is targeted by tasks such as picture descriptions. Thus, in order to gain a more complete picture, the inclusion of different tasks is necessary. This point was already behind the procedure in the SyHD project, to which we refer in this respect (Fleischer, Kasper & Lenz 2012). However, in analyzing the data from the pretest questionnaire, we noted that the picture description task is often prone to being misunderstood by the participants.
In this questionnaire, we tested the problem at hand with several different types of questions. One was acceptability judgment tasks (2 items) on a 4-point scale of clauses containing a noun phrase in the form dem+do/lo+N and one in the standard form dem+N+do/lo (example in Figure 1). Value 1 stands for 'completely natural', 2 for 'not completely natural', 3 for 'relatively unnatural, but still marginally acceptable', 4 for 'completely unnatural'. 2 4 Rauth and Speyer Glossa: a journal of general linguistics DOI: 10.5334/g jgl.1166

Rauth and Speyer
Glossa: a journal of general linguistics DOI: 10.5334/g jgl.1166 The acceptability study showed a clear area in which the use of the variant dem+do/lo+N was judged as grammatical (Figure 4), sometimes even better than the standard variant dem+N+do/ lo ( Figure 5). In the maps in Figures 4 and 5, the older and younger age groups were combined in order to increase the data set. The colors used in Figure 4 correspond to mean judgment values. The available scale 1.0 to 4.0 was divided into four parts; each part was assigned a color following a traffic light arrangement (1.0 to 1.749: dark green; 1.75 to 2.49: light green; 2.5 to 3.249: yellow; 3.25 to 4.0: red). The colors in Figure 5 reflect a heat map design. The numbers in Figure 5 are calculated by the formula s -d, where s = mean value of the standard dem+N+do/ lo variant, and d = mean value of the dialectal dem+do/lo+N variant. The larger the difference between the mean judgment values of the dem+do/lo+N variant to the dem+N+do/lo variant was, the darker the color. Red stands for preference of the dialectal dem+do/lo+N variant, blue for preference of the standard dem+N+do/lo variant.
Not only are judgments for the variant dem+do/lo+N relatively good (Figure 4) in this area, but in the same area, the variant dem+do/lo+N is judged even better than the standard dem+N+do/ lo ( Figure 5).

Rhineland-Palatinate Rauth and Speyer
Glossa: a journal of general linguistics DOI: 10.5334/g jgl.1166 spontaneously in the description task even in townships in which the acceptability judgments were poor, e.g. in the northern part of the area of investigation.
In general, we can say that the dem+do/lo+N variant is the preferred variant in the eastern parts of the transition area in Saarland and Lorraine, but that it is at least marginally possible in most other regions of the investigation area. This can be interpreted in such a way that this construction has been fully grammaticalized as the common way to express demonstrative force in the transition area.
An interesting point is that the area in which the dem+do/lo+N variant is preferred reaches across the border into Lorraine. This indicates either that the origin and spread of this variant predates the political separation of Lorraine and the Saarland, that is: World War I, or that the interchange between Lorraine and the Saarland was always so easily manageable that these two regions remained homogenous as a dialect area.

THE EMERGENCE OF A NEW DEMONSTRATIVE
Once the expressive force of a grammatical item is insufficient due to phonetic erosion or formal ambiguity, this item is often reinforced or replaced by a lexical item. For instance, the preverbal particle ne used to be the only means of coding sentence negation in Classical Latin. Ne has been phonetically weakened and therefore frequently accompanied by the numeral oenum 'one thing'. The prevalent juxtaposition with ne made it possible for both items to merge into the new strengthened negative marker non (< noenum). This process of reinforcement is commonly explained as an act of grammaticalization: The lexical item oenum is desemanticized by reducing its interpretable (semantic) features in order to acquire a second function, i.e. emphasizing the negative expressivity of the existing negation marker ne. As soon as the syntagma ne oenum is generally used to express sentence negation, an increase of bondedness between the two items is set in motion. As a result, ne and oenum are merged (univerbated) into the single word noenum and, again, successively weakened to the new negative marker non (see Jespersen 1917: 7;Lenz 1996: 183f.;Heine 2003: 579;Lehmann 2015: 24-26, 157-161).
The emergence of the new demonstrative dem+do/lo+N can be seen as an instance of grammaticalization very similar to the case of Latin non (see Girnth & Michel 2008: 211-213;Rauth & Speyer 2018: 53-56 As the adverbs do/lo carry a semantic feature of locative indexing, obligatory reinforcement is first expected to occur in situational demonstrative contexts where the referent of the DP is physically present and can be pointed at. This is the case in (4b), where the cellphone is ringing with a special ringtone. By das doo, the speaker refers to the audible ringtone which tells him that his dentist is calling. As a next step, do/lo lose their need for a physically present referent so that they are compatible with non-situational contexts of demonstratives. Such a case is illustrated in (4c): A 12-year-old boy disguised as St Nicholas has already left the room where the other children anaphorically referring to him by dat lòò are located. Additionally, (4c) illustrates that the dialectal reinforcers do/lo even appear in contexts where the reinforcer da would be banned or at least feel very odd in the (colloquial) Standard German equivalent (4d).
As a next step, the highly frequent use of the syntagma dem+do/lo gives rise to the formal part of the grammaticalization process, namely the initiation of the coalescence of dem and do/lo. Right after the loose juxtaposition where the adverb is prosodically more prominent than the demonstrative, grammaticalized do/lo is cliticized to the demonstrative. As a result, do/lo should be prosodically subordinated under the accent of the adjacent determiner and form a phonological word together with it (see Lehmann 2015: 157). In fact, it is not clear if this stage of coalescence has already been obtained. Steitz's (1981: 113-116) dialect grammar of Saarbrücken (Rhine Franconian) states that the demonstrative is regularly strengthened by do, and the syntagma bears initial-stress if the physically present referent is distal (ˈdɛːr.dɔː), while final-stress corresponds to a proximal referent (dɛːr.ˈdɔː). He clarifies the advanced coalescence status of dem+do/lo by consistently writing it in one word. Bonner's (1986: 134) grammar of Neunkirchen challenges Steitz's functional analysis by saying that he fails to take into account the contextual variability of deictic expressions. However, we can observe from the above-mentioned dialect grammarians that there actually is an initially stressed variant of dem+do/lo, whatever the function of the variable stress may be. The former locative adverb, thus, is currently changing its status from a juxtaposed to an encliticized demonstrative particle. Further evidence for an advanced status of coalescence can be seen in instances where both the demonstrative and the clitic are inflected: Since adverbs in German are non-inflectional, the examples in (5) suggest that do/lo have changed their categorial status. While inflected lo in (5a) and (5b) seem to behave like adjectives with weak inflectional morphemes (den Lo'en ↔ den gut-en 'the good one'), den do'ne in (5c) rather behaves like a determiner, since it bears a strong inflectional morpheme, by analogy with the form of the simple demonstrative den-ne ('this.acc.sg') in Rhine Franconian. Whatever the new status of do/lo, they most certainly do not act as typical adverbs anymore.
Once the coalescence of dem and do/lo is strong enough to be reanalyzed as a composite demonstrative, its use is extended from pronominal to adnominal contexts. The earliest (written) attestation of the adnominal use we found in our text corpus (see Table 1 for Rhine Franconian and Table 2 for Moselle Franconian) is from a Rhine Franconian text of the year 1900: (6) Rhine Franconian (Claus 1900: 15) In dem do Punkt bin ich ganz mit'm Franzel einverschtanne. in dem adv point am I fully with.the Franzel agreed 'In this/that respect I fully agree with Franzel.' The construction in dem do Punkt is used five times by the main protagonist of the text. Since there are no other instances of dem+do/lo+N, it is likely that it functions as a special stylistic element of the main protagonist. Dialect grammarians of the early 20 th century neglect to mention the reinforced demonstrative, with one crucial exception: Lehnert's (1926: 115) dialect grammar of Diefflen (Moselle Franconian) states that speakers putting lo in the intermediate position (der lo Monn 'this/that man') may appear odd. Thus, dem+do/lo+N must have been familiar to speakers of Rhine and Moselle Franconian at the turn of the century, but still seems to be stigmatized in a certain way. One generation after Claus (1900), the text of Jantzer (1933) reveals three lexically different instances of dem+do+N for Rhine Franconian, while in Moselle Franconian texts Fox (1955Fox ( [1924) and Lehnert & Augustin (1950[1939) lo is restricted to peripheral positions. As for adnominal demonstratives in recent Rhine Franconian (Lang 2011), do/lo is only attested in intermediate position. There actually are three instances of peripheral do/lo, but these function as temporal adverbs: Rhine Franconian (Lang 2011: 128) Auserdem fangd doo die Olympiaade […] aan. moreover begins adv det Olympics ptcl 'Moreover, the Olympic Games will begin at that time.' Modern Moselle Franconian (Braun & Peter 1999) finally also reveals two instances of dem+do/ lo+N, but in contrast to Rhine Franconian, peripheral do/lo is still a highly frequent alternative.
Apart from the adnominal use, reinforced demonstrative pronouns constitute a large portion of the attestations found in our text corpus, with the exception of Claus (1900  50 % in Moselle Franconian. Overall, the use of locative adverbs do/lo increased in both dialects, as the numbers of attestations in the text corpus show. In order to fill the data gap in the second half of the 20 th century, we analyzed audio recordings of the Mittelrheinischer Sprachatlas (MRhSA, 'Middle Rhine linguistic atlas', see Schmidt et al. 2008ff.), which have been collected in the 1980s. A younger (about 35 years old) and an older (about 65 years old) generation of informants translated Standard German words or sentences into their dialect. Girnth & Michel (2008) looked into the translations of the sentence Hast du dén Mann gekannt? 'Did you know that man?' and found that Rhine Franconian speakers solely used the disyllabic form denne Mann ('dem.acc.sg man') without reinforcing it by do, while Moselle Franconian informants only produced demonstratives with a peripheral reinforcer. They concluded that dem+do/lo+N is on the decline, which is contrary to our observations of modern spoken Rhine and Moselle Franconian. Apparently, disyllabic denne is an exception to the predominant similarity as it formally differs from the definite article de(n) in Rhine Franconian and thus does not need to be reinforced. 4 Instead, we analyzed the translations of the sentence Er soll mit dém Wagen fahren! 'He should drive that car!' produced by speakers living in the German federal state of Saarland. The results are given in Figures 8 and 9. A serious disadvantage of translation tasks is the possible influence of the originally presented sentence. Thus, the high number of demonstratives missing do/lo (white circles) in both figures could be traced back to the non-reinforced demonstrative in Standard German. As for the reinforced demonstratives, both figures reveal a division of the state of Saarland into two parts: The western part preferably uses peripheral positions of do/lo, while the eastern part seems to constitute the core region of the new demonstrative. 5 These findings are in line with the results of our recent questionnaire study (see Section 2).
The data from the older generation of informants also include two types of intermediate reinforcement by lexical items other than do/lo: (i) three tokens of reinforcement by hei 'here' (for instance, recorded in Falscheid, Moselle Franconian: dem hei Wòòn), and (ii) one token of reinforcement by sell 'the same' (recorded in Hoof, Rhine Franconian: dem sell'e Waan). As 4 It is the only case of formal difference of the demonstrative and the definite article. However, denne recently seems to be reinforced by do as well for the sake of a uniform inflectional paradigm of the new demonstrative (see example 8 for an illustration).

5
Comparing Figures 8 and 9, attestations of intermediate do/lo (i) decrease from 13 (older generation) to 8 (younger generation) and (ii) cease to appear in Moselle Franconian. However, the core region of dem+do/lo+N remains stable. the latter seems to be an isolated case, we discuss the former reinforcement by hei in more detail. Since most Germanic languages differentiate between a proximal and a distal locative adverb, it might be expected that reinforcement of the demonstrative be possible with each of the two adverbs. In colloquial Norwegian, for instance, the demonstrative den is compatible with both proximal her and distal der (see Julien 2005: 116). This also holds true for Moselle Franconian, which actually provides four locative adverbs: proximal hei and lei (both 'here', from Early New High German allhier), and distal do and lo (both 'there', from Early New High German allda) (see Lehnert 1926: 119, 203). The free choice between several locative adverbs could be the reason for the considerably later attestation of intermediate reinforcement in Moselle Franconian. As for Rhine Franconian locative adverbs, there is only a differentiation between proximal hie 'hier' and distal do 'there'. However, do can be used for both proximal and distal referents and is thus by far more frequent than hie. This could have been the promoting factor for the early emergence of a new formally strengthened demonstrative in Rhine Franconian. As a consequence, the recently rising occurrence of intermediate reinforcement by lo in Moselle Franconian might be influenced by the advanced state of grammaticalization in neighboring Rhine Franconian.
Further evidence for an early attestation of Rhine Franconian dem+do/lo+N is that the construction was taken across the ocean: Putnam (2006: 172) points out that contemporary Pennsylvania German provides two reinforcer constructions: Pennsylvania German (Putnam 2006: 172) a. der do mann the here man 'this man' b. seller datt mann that there man 'that man' The reinforcer construction in (8a) strikingly reminds of the Rhine Franconian equivalent discussed in this paper. The demonstrative seller in (8b) is also very common in many modern High German dialects. It is therefore worth mentioning that the German variety spoken by the religious community of the Amish as well as by many descendants of immigrants coming to the United States during the 19 th century is crucially influenced by the Rhine Franconian dialect of the Palatinate region ("Pfalz"), which served as a koiné dialect for the different immigrant groups of German origin (see Ammon 2015: 383-384). It seems to be quite likely that Rhine Franconian speaking immigrants brought the developing intermediate reinforcement of the demonstrative with them. While the coalescence of dem+do/lo has currently reached the status of cliticization by still inflecting dem, the youngest generation of Pennsylvania German speakers introduce a phonologically combined variant of (8b) (selldatt mann 'that man') without exhibiting any inflectional morphology (see Putnam 2006: 170-171). Consider that the development of the Standard German demonstrative dies-er/-e/-es also started out from an inflecting dem reinforced by sê and eventually developed into an invariable stem and an inflecting (former) reinforcer. 6 Thus, the next step on the grammaticalization path for Rhine and Moselle Franconian dem+do/lo might also be mono-inflection at the end of the new demonstrative as well as an invariable stem. In contrast to Pennsylvania German, there is still no evidence for that in Rhine and Moselle Franconian. In concordance with Pennsylvania German (see Putnam 2006: 168-170), however, the insertion of additional modifiers such as numerals (9a) or adjectives (9b,c) is restricted to the position between dem+do/lo and N in Rhine and Moselle Franconian.
Wenn mir denne do ganze Kram getrunk hann. once we dem adv whole stuff drunk have 'Once we have drunk all of this/that stuff.' c.
Das do verlängerde Wocheend war doch mo scheen! dem adv extended weekend was ptcl once nice 'This/that long weekend has been really nice!' To conclude this section, the new Rhine and Moselle Franconian demonstrative pronoun dem+do/ lo can occur in all semantico-pragmatic contexts where the occurrence of the Standard German equivalent dies-er/-e/-es is common as well. The former adverbs do/lo are currently changing their categorial status from juxtaposed to encliticized demonstrative particles. At the latest since the beginning of the 20 th century, the coalescence between dem and do/lo is sufficiently advanced to be extended to the adnominal use. As a formally reinforced demonstrative, it must precede noun-modifying numerals and adjectives.

REANALYSIS OF DESEMANTICIZED DO/LO AS A FUNCTIONAL HEAD
In the present section, we first briefly outline earlier approaches to the structural analysis of simple demonstratives and demonstrative reinforcement in Germanic and Romance languages and try to identify potential shortcomings of these approaches. We then propose our own structural analysis of the diachronic emergence of the new demonstrative dem+do/lo in Rhine and Moselle Franconian as a result of a common grammaticalization path.
In recent generative research on nominals, demonstratives and definite articles are commonly considered to differ with respect to their structural status: Demonstratives are regarded as phrasal elements that occupy specifier positions, whereas definite articles are classified as functional heads (see, for example, Giusti 1993;Brugè 1996;Bernstein 1997). Three reasons for structurally distinguishing demonstratives from definite articles are that (i) both can cooccur in many languages, (ii) demonstratives can be used pronominally, i.e. they function as an independent DP, and (iii) demonstratives share properties of adjectives such as similar inflection patterns or similar syntactical positions (see Bernstein 1997: 92-93). The base position of demonstratives is assumed to be in the specifier of a functional projection between 6 The demonstratives of many Germanic languages trace back to the univerbation of a stem and a reinforcer. As for German dies-er/-e/-es 'this-m/f/n', the Germanic demonstrative *þo lost its deictic force and was reanalyzed as the Old High German definite article dër/diu/daȝ 'det.m/f/n' (see Oubouzar 1992;Kraiss 2014). At the same time, the former demonstrative was strengthened by the deictic particle sê, which yielded the reinforced Old High German forms dëse/dësiu/diz 'this-m/f/n'. Although the particle sê originally was indeclinable, early attestations of the new demonstrative in Old High German already show inflexion on both dem and sê or even just on sê (see Wilmanns 1899: §429.2; Wilmanns 1909: §205; Lander 2015: 38-41). Today, the Standard German (proximal) demonstrative consists of the invariable stem dies-(with -s-being the remainder of sê) and its associated inflectional morphemes.

Rauth and Speyer
Glossa: a journal of general linguistics DOI: 10.5334/g jgl.1166 DP and the lexical NP. For instance, Brugè (1996) shows that Spanish demonstratives can occur in prenominal as well as in postnominal position. She therefore base generates the demonstrative in the specifier of a functional FP (Spec,FP) which is situated between several intermediate functional phrases hosting adjectives and the lexical NP (see Brugè 1996: 10). The postnominal position is obtained by movement of the noun into the head of a higher functional phrase, while the demonstrative stays in situ. Prenominal demonstratives, in turn, move from Spec,FP to Spec,DP and thus appear to the left of the noun. Regardless of whether the demonstrative is prenominal or postnominal, Brugè's (1996: 8) demonstratives in Spec,FP must always move to Spec,DP in order to check uninterpretable referentiality features of D. In case the demonstrative appears prenominally, this movement is carried out overtly at PF, while postnominal demonstratives covertly move to Spec,DP at LF.
As for demonstrative reinforcement in Spanish, Brugè (1996: 24) assumes that demonstratives like este 'this' and reinforcers like aquí 'here' are in a predicational-like relation established and dominated by the preposition de 'of'. The demonstrative and the reinforcer can stay in situ at PF, which yields the word order in (10a). Alternatively, este can rise to Spec,DP and de aquí stays in situ at PF, which gives us (10b).

(10)
Spanish (Brugè 1996: 20- Brugè (1996: 21) considers reinforcers as indicators for the base position of prenominal demonstratives. Example (11a) shows that de aquí is restricted to the same position as the postnominal demonstrative este would be (cf. 10a), i.e. between postnominal adjectives (viejo 'old') and postnominal PP-attributes (de sintaxis 'about syntax'). She then intends to expand this structural analysis of a low base position to languages that do not allow for postnominal demonstratives. As for Standard German, she claims that postnominal reinforcers such as hier 'here' are also restricted to a position higher than postnominal PP-modifiers (11b) (see Brugè 1996: 45, ex. 84c). This judgement is clearly wrong. German reinforcers can appear either before or after PP-modifiers of the noun. In our opinion, it is an economically more expensive operation to base generate German demonstratives in a low position where they can never occur, only to oblige them to always move up to Spec,DP at PF. Why not base generate German demonstratives in Spec,DP directly and adjoin possible postnominal reinforcers to the position they appear in?

Rauth and Speyer
Glossa: a journal of general linguistics DOI: 10.5334/g jgl.1166 This would also be a less costly option for Spanish demonstratives: (i) Prenominal demonstratives represent the unmarked case, while postnominal ones are preferably used in colloquial speech and restrict the nominal to a depreciatory meaning (see Brugè 1996: 1, fn. 3). We would expect a demonstrative in its base position to provide a neutral meaning as well. (ii) Brugè (1996: 8) explains the fact that postnominal demonstratives in Spanish must be accompanied by a prenominal definite article by proposing a Last Resort operation: The definite article must be inserted into D in order to validate its uninterpretable φ-features and make the relevant features visible at PF. Apparently, the covert movement of the demonstrative to Spec,DP at LF is not sufficient for this validation, while the overt movement can validate the φ-features. If we simply assume that cases like (10a) consist of the definite DP el libro and the adjoined DP este de aquí (see also Roehrs 2010: 255, fn. 32), neither the Last Resort insertion of the definite article nor the movement of the demonstrative to Spec,DP at PF or LF is necessary. The features of the first DP are validated by the definite article in D, the ones of the second DP by the demonstrative in Spec,DP.
The next approach we will focus on here is Roehrs (2010), who also proposes an analysis that copes with pre-and postnominal demonstratives in Germanic and Romance languages. He identifies four types of reinforced demonstratives: In type 1, a reinforcer is added after the demonstrative, but before adjectives and the head noun (dem reinf Adj N). This type, for instance, can be found in non-standard English (12a), Pennsylvania German (12b), Eastern Norwegian (12c) as well as in Swedish, Danish, Icelandic and Yiddish (see Roehrs 2010: 265). Rhine and Moselle Franconian also provide this kind of reinforced demonstratives (12d). Romance languages do not exhibit type 1 reinforcement.
Non-standard English (Bernstein 1997: 101)  In type 2 of demonstrative reinforcement, the reinforcer precedes the demonstrative and is thus located at the very beginning of the DP (reinf dem Adj N). This type is attested, among others, in Afrikaans (13a), Yiddish (13b), and colloquial Standard German (2a). Rhine and Moselle Franconian also provide a preceding reinforcer (13c). Again, type 2 reinforcement is not found in Romance languages.
Moselle Franconian (Fox 1955: 27) lo die dreckig Kurwel adv dem dirty basket 'this/that dirty basket' The third possibility is to insert the reinforcer at the very end of the DP after the head noun. Hence, demonstrative and reinforcer put the rest of the nominal in brackets (dem Adj N reinf).
French (Bernstein 1997: 98) ce livre jaune ci dem book yellow adv 'this yellow book' Type 4 reinforcement is restricted to Romance languages Spanish and Catalan. The crucial difference is that both the demonstrative and the reinforcer appear after the head noun and D must be filled by the definite article (10a), as has already become clear within the discussion of Brugè's approach.
Types 1, 2 and 3 of demonstrative reinforcement require three different positions for reinforcers inside the DP. Roehrs (2010: 240-244) therefore proposes that demonstratives are formed compositionally, consisting of (i) a lexical deictic part carrying an interpretable referentiality feature hosted in the head of a projection called Deictic Phrase (DeicP), (ii) a functional definite marker with an interpretable definiteness feature hosted in a functional Demonstrative Phrase (DemP), and, if available, (iii) an agreement suffix. In case of the Standard German demonstrative d+ies-er/-e/-es, the deictic part -ies-is hosted in DeicP. The deictic part can also be null (Ø) in case there is no such overtly audible part in a demonstrative like in the stressed variant of German d+Ø-er/-ie/-as. The definite marker d-is located in the head of DemP. Demonstratives are formed by raising the deictic part from Deic to the definite marker in Dem, where +ies-or +Ø-incorporates d+ yielding the complex formes d+ies-or d+Ø-. The specifiers of DeicP and DemP, then, provide free positions for reinforcements of types 1 and 2. Reinforcers of type 1 occupy Spec,DeicP and follow the demonstrative (12), while type 2 reinforcers are located in Spec,DemP and thus precede the demonstrative (13). As for type 3, Roehrs (2010: 254-255) right-adjoins a PP that contains the reinforcer to IndP in order for the reinforcer to follow the head noun (14). While in Romance languages the head of PP can be filled by an overt preposition, Germanic type 3 reinforcers obviously do not exhibit a preposition. The head of PP is therefore assumed be null in Germanic. The structure for reinforcement of types 1-3 is given in (15). Finally, the compositional demonstrative in DemP is moved to Spec,DP in order to check D's uninterpretable definiteness and referentiality features.

Rauth and Speyer
Glossa: a journal of general linguistics DOI: 10.5334/g jgl.1166 The structure of type 4 reinforcers in Spanish and Catalan differ from type 3 with regard to the position of the noun and the demonstrative as well as the obligatory co-occurrence of the definite article. In order to appear postnominally, the demonstrative stays in situ (i.e. in Spec,IndP), while the noun moves up to an additionally merged functional phrase (YP) dominated by DP. Roehrs (2010: 229, 236) follows Brugè in that the definite article is inserted as a Last Resort operation.
We would like to point out three problematic aspects of Roehrs' analysis: (i) Why is type 1 reinforcement not possible in Standard German (2c) and Dutch, if Spec,DeicP is a principally available landing site for reinforcers? Do languages that rule out type 1 differ structurally from languages allowing for type 1? (ii) Is it legitimate to split up a synchronically simplex morpheme like German dies-into a bipartite demonstrative whose parts cannot occur on their own? If we took the diachronic development consistently seriously, we would have to assume that, for instance, the German negative particle nicht 'not' is also still complex today (and consists of the negative marker ne and the noun Wicht 'dwarf, wight'). (iii) Roehrs (2010: 256) claims that Spanish demonstratives like este are not split up and can either stay in Deic or move to Dem, depending on whether they are marked for topic or focus. But, from a diachronic point of view, it would be consistent to also split up este, since it traces back to the Latin demonstrative is combined with the pronoun-like (or particle-like) element *-to-(see de Vaan 2008: 310). 7 The third approach to the structure of demonstratives discussed here is that of Leu (2015). He proposes that demonstratives and definite adjectival phrases are structurally alike (see Leu 2015: 12, 43). Therefore, he assumes demonstratives to consist of a strong inflecting definite marker and a silent or non-silent deictic adjective. The non-reinforced colloquial German dér in (16a) is an example for demonstratives that are composed of a definite marker d-, followed by a strong agreement marker -er, and an additional non-pronounced (silent) adjective represented by THERE. The adjective can also be pronounced (non-silent) as in Rhine Franconian der do Tisch (16b). What we have called an intermediate demonstrative reinforcer so far is thus seen as an overtly realized adjectival stem. According to Leu (2015: 22, 43-45, 71-72, 87), the Standard German demonstrative stem dies-is treated like an adjectival stem with a preceding unpronounced definite marker. Compare (16c) and (16d). In both cases, the definite marker is not merged and therefore the deictic adjective diesand the regular adjective schön move up to the left periphery and take the agreement marker with them. 8 As for peripheral reinforcers, Leu (2015: 19) looks into the case of Afrikaans hier-die (see 13a). The inverted order of the definite marker and the deictic adjective is achieved by movement of the adjective to the left of the definite marker. Rauth and Speyer Glossa: a journal of general linguistics DOI: 10.5334/g jgl.1166 Regarding demonstrative reinforcement, Leu's (2015) approach focuses on the representation of intermediate reinforcers that can be realized as non-silent deictic adjectives. Locative adverbs such as Rhine and Moselle Franconian do/lo are treated as lexical heads of an adjective phrase, the corresponding determiners (definite markers) occupy the specifier of the extended adjectival projection. Following Leu's analysis, we can only speak of demonstrative reinforcement if more than one reinforcer is present, since the first reinforcer is a formal part of the demonstrative.
However, (i) it remains unclear how, on the one hand, Romance demonstratives like Spanish este (de aquí) fit into the picture. Prenominal demonstratives could either be treated like German dies-, which means that they actually are deictic adjectives with a missing definite marker. Or they are analyzed as definite markers followed by a categorically silent deictic adjective, since PP-reinforcers like de aquí can never occur prenominally in Romance. On the other hand, Spanish postnominal demonstratives can be accompanied by reinforcers. Is it possible to analyze a PP like de aquí as a non-silent deictic adjective? (ii) While the status of German dies-as a deictic adjective without a definite marker is plausible from a synchronic perspective, the historic development forces us to question Leu's view. As has already been pointed out in Section 3, the form of the Standard German demonstrative stem dies-is the result of the historical reinforcement of the Old High German simple demonstrative dër/diu/da by the deictic element sê (see Lander 2015: 38-41). By means of Leu, sê can be treated as a non-silent deictic adjective preceded by the definite marker d-ër. In order for the definite marker to become an integral part of the deictic adjective -sê-(today -ies-), it must have changed its categorial status, while in the English cognate th-is the definite marker has not changed its status.
After having outlined several approaches to the analysis of demonstrative reinforcement, we propose a somewhat simplified structure which, leaving out the details for the moment, can cope with Germanic and Romance demonstrative reinforcement: As against most other recent analyses, we do not consider the movement of the demonstrative from a lower base position to Spec,DP necessary. In Germanic, the demonstrative can never occur postnominally. In Romance, the unmarked position of the demonstrative is also always prenominal. We therefore claim that the base position of prenominal demonstratives is in Spec,DP.
The demonstrative pronoun itself carries an interpretable identifiability feature as well as a feature of locative indexing and projects a Demonstrative Phrase (DemP), which in turn is situated in the specifier of a functional Index Phrase (IndP) (see example 15 and Roehrs 2010). On the one hand, the indexing feature must be of a rather neutral (or unspecified) nature in order to be suitable for physically present referents (e.g. table, car), abstract referents (e.g. rage, love) and anaphoric contexts. This feature can validate its uninterpretable counterpart situated in the head of IndP, which can be, depending on the pragmatic context, specified for the three different kinds of referents. Additionally, Ind can be overtly filled by a reinforcer, which we will elaborate down below. On the other hand, the identifiability feature is the main reason why the demonstrative is merged in Spec,DP. It satisfies the uninterpretable identifiability feature in D.
Regarding postnominal demonstratives in Romance, we claim that these can be analyzed as additional DP-adjuncts, since the primary DP is already validated by the definite article in D (see 10a). The different order of postnominal demonstratives, reinforcers, adjectives and nominal complements (see 11a and Brugè 1996) can also be achieved by extraposition to the right 18 Rauth and Speyer Glossa: a journal of general linguistics DOI: 10.5334/g jgl.1166 periphery, i.e. optional adjunction of a certain element to the outer DP-layer. A similar process is well known from German embedded sentences, where the finite verb usually comes last, but several constituents can be extraposed to the right of the verb in order to facilitate the interpretation of the sentence. First, we observed a stage in which Rhine and Moselle Franconian demonstratives can be reinforced by do/lo occupying the peripheral positions by either preceding the demonstrative or following the noun. In terms of Roehrs' (2010) classification and our structural proposal, do/lo has been adjoined to the left (type 2) or the right of the DP (type 3). However, the head of IndP (type 1) cannot have been a possible landing site for do/lo because the intermediate position was still ungrammatical in this early stage and lexical items cannot occupy functional heads without syntactic and/or semantic change. We therefore propose that, in order to get to type 1, reinforcers have to undergo a process of grammaticalization: In the pre-grammaticalization stage, the locative adverbs do/lo constitute the head of a lexical AdvP which is loosely adjoined to the outer DP-layer. This stage 1 is illustrated in (20) by the first and second line den Mann do/ lo 'dem man adv' and den do/lo 'dem adv'. While being in the state of an adjunct, do/lo still bear a specific indexing feature that requires a physically present referent.
Second, as we have already stated in Section 3, in the adnominal use dem+N+do/lo the reinforcers cannot be mistaken for being permanently adjacent to dem, i.e. to be mistaken for occupying the head of IndP. Therefore, the structural reanalysis must have resulted from a highly frequent use of dem+do/lo in pronominal contexts, where demonstratives and reinforcers are juxtaposed and no noun is intervening. 10 Stage 2 (third line) in (20) illustrates that den 'dem' and do/lo 'adv' 9 We thank an anonymous reviewer for bringing this important difference to our attention.
10 Reanalysis is defined as a structural reinterpretation of an ambiguous linear order of morphemes or lexemes (see Hopper & Traugott 2003: 50-52). Rauth and Speyer Glossa: a journal of general linguistics DOI: 10.5334/g jgl.1166 are always juxtaposed in pronominal use, both as head of an adjoined AdvP or head of the functional IndP.
Third, certain bridging contexts provide an ambiguous reading where the indexing feature is neutralized, i.e. the presence or absence of a physical referent which can be pointed at is negligible (see example 4b in Section 3). As a consequence, do/lo are reduced to the meaning of abstract indexing and their categorial status as a fully lexical item is weakened. This kind of semantic reduction represents a typical case of grammaticalization, which often results in reanalysis from a lexical to a functional head (see Roberts & Roussou 2003;van Gelderen 2007: 279, 286). Thus, the neutralized semantic feature of do/lo suggests that the former fully lexical adverb is now base generated as head of a functional phrase and constitutes an integral part of the demonstrative construction (see stage 3 in example 20). 11,12 Standard German reinforcers cannot occupy the head of the functional IndP because no reanalysis has taken place, which explains why the intermediate position of da in (2c) is ungrammatical.
Fourth, instances of double inflection on the demonstrative and the reinforcer in Rhine and Moselle Franconian (5b) can be explained by assuming that do/lo in the head of IndP start to agree with the noun. Since our proposal does not provide separate agreement phrases that the demonstrative has to cross in order to take up inflexion morphemes, morphological inflexion is not restricted to either the demonstrative stem in DemP or the reinforcer in Ind, but can also affect both elements. Moreover, we recognize that certain instances of do/lo could also have been reinterpreted as adjectives (5a) which have a different underlying construction (compare Standard German hie-s-ig 'here-s-adj' and dort-ig 'there-adj'). However, we do not consider double inflection to be an independent phase on the grammaticalization path.
A further development which can be observed in languages where there seem to be two intermediate reinforcers is not yet attested in Rhine and Moselle Franconian. For instance, in Eastern Norwegian (21a) and Pennsylvania German (21b) reinforcers can appear twice: (21) a.
Pennsylvania German (Roehrs 2010: 242) mit selldatt datt grosse mann with dem.adv adv tall man 'with that tall man' In Eastern Norwegian, the second reinforcer could either be analyzed as an indeclinable adjective or as an adverb left-adjoined somewhere inside NP. Leu (2015: 33) states that (21a) can only occur in locative and not in anaphorical contexts, which is why the functional head of IndP is not a possible landing site for Norwegian her 'here'. But since (21a) does not represent the standard or unmarked use of demonstratives in Eastern Norwegian, rarely found doubling phenomena like herre her should not lead to more functional structure than is generally needed. By contrast, if Pennsylvania German double reinforcement in (21b) actually was compatible with abstract referents or anaphoric contexts, this could be evidence for the next step on the grammaticalization path: The head of IndP has again been structurally reanalyzed, but this time as part of the adjacent head of DemP, which is the demonstrative stem itself. Thus, the demonstrative stem and the reinforcer have been fused (coalescence), and the head of IndP is available again for a newly reanalyzed item expressing abstract indexing. 13 11 The claim that reinforcers occupy the head of a functional phrase has already been made by Bernstein (1997: 100).
12 In Swedish, Norwegian, and English, the indexing features of the locative adverbs do not seem to be as fully neutralized as Rhine and Moselle Franconian do/lo, since there are two items specified for proximal (e.g. English the here) and distant (e.g. English the there) referents which have been reanalyzed as the head of IndP. At least in Swedish, however, there seems to be a difference between proximal and distant heads of IndP: While den här 'this' can be used for physically proximal referents and for anaphoric contexts, den där 'that' is restricted to physically distant referents (see Holmes & Hinchcliffe 2013: 175-176). Thus, här should be semantically more reduced than där.