2.1. Basic Discussions of Determiners

Authors vary in the range and depth of their remarks on determiners. Some authors seem to restrict their attention to the structure of definite determiners. For instance, Vater (Reference Vater1984:21, note 3) states that it is possible to analyze definite articles in German as involving a root morpheme and an inflectional morpheme (d-er). Similarly, Katzir (Reference Katzir2011) proposes for Danish den that the d- is presumably a dummy and -en is a licensor for a definiteness feature. Chomsky (Reference Chomsky1995:338) suggests that English demonstratives are presumably complex elements, with th representing D (as in the and there) and the remaining part possibly involving some kind of adjective. Similarly, den Dikken (2006:198) observes that demonstratives in Germanic are usually amalgams of the definite determiner in D and some deictic marker. Finally, Bernstein et al. (Reference Bernstein, Cowart and McDaniel1999:500) propose that demonstratives in English in their definite interpretation consist of components involving definiteness (the), number, and deixis.

Interestingly, other proposals are wider in scope of their discussion of the relevant internal structures. Specifically, Bernstein (Reference Bernstein, Høeg Müller and Klinge2008) discusses lexical elements such as the, this, they, and existential there arguing that the th-morpheme represents a 3rd person marker that is unspecified for number and gender and that does not encode definiteness and deixis. Déchaine & Wiltschko (2002:422) treat th in words such as the, this, and them as a bound D-morpheme. Finally, Wiltschko (Reference Wiltschko1998: 149) proposes that determiners in general can be analyzed as a bound morpheme that involves d- for definite determiners. This morpheme occupies D and takes AgrDP as its complement hosting the inflectional ending (see also section 6.2).

To be clear, all these authors suggest that definite determiners are complex elements with inner structure. However, they differ on a number of points, namely (i) what lexical items are involved or not (for example, personal pronouns, existential there) and in what languages, (ii) what exact form the initial element has (for example, th versus the), and (iii) what status the initial element has with regard to definiteness.

It is also interesting to point out that some of the above proposals are couched in rather cautious terms and provide no explicit discussion about the syntactic structures involved; that is, it remains unclear where the relevant elements are in the syntactic structure and how they combine to yield the relevant surface forms. To be fair, though, I hasten to add that not all proposals above set out to explain what I am interested in in this paper.

2.2. More Detailed Discussions of Determiners

I turn to three recent proposals that discuss the decomposition of determiners in quite some detail: Klinge Reference Klinge, Høeg Müller and Klinge2008, Leu Reference Leu2008, and Roehrs Reference Roehrs2010. As might be expected, these proposals also vary in (intended) empirical coverage. I dedicate most of the space to Leu's work, which is somewhat programmatic but has been quite influential.

Klinge (Reference Klinge, Høeg Müller and Klinge2008) discusses þ-related and hw-related elements in the history of the Germanic languages with special focus on German, English, and Danish. Illustrating his proposal with Modern German here, Klinge delineates four components for the demonstrative dieses ‘this' in 16a. Specifically, the stem of the demonstrative consists of a d-root with a deictic feature (DEIC), an epenthetic vowel (V), and a proximal element (PROX). This complex stem combines with an inflectional ending (INFL), as in 16b.

1. (16)

   

2. 

The distal demonstrative das ‘that' is as follows:

1. (17)

   

Klinge (2008:234–235) ties adverbial elements such as German da ‘there' and dann ‘then' to the demonstratives above by showing that the former can be given paraphrases involving the latter. All these indexical elements are contrasted with w-elements such as German welcher ‘which', wo ‘where', and wann ‘where'. Klinge argues that d-related and w-related elements are determiners that involve the same makeup. The basic difference is as follows: D-related determiners indicate that the speaker has a specific entity in mind that the hearer is instructed to identify, whereas w-related elements contain a variable that the hearer is instructed to fill. Connecting all these elements in one account is a nice achievement.

Focusing on the demonstratives in 16 and 17, Klinge posits a maximum of four elements in the makeup of demonstratives. However, as far as I can tell, Klinge winds up with two different decompositions: In the parsing in 16b and 17b, there are differences in the epenthetic vowels, in the presence versus absence of a second deictic element, and in the shape of inflectional suffixes. As a result, the inventory of inflectional suffixes is increased to seven, where the two additional ones, -as and -ie, are restricted to simple demonstratives, as in 17a.

In my view, it would be more desirable to postulate the same number and types of components for both demonstratives and to keep the inflectional inventory the same. Also, it remains unclear how non-d-stems such as German jen- ‘that', Yiddish yen- ‘that', or Pennsylvania German sell- ‘that' fit in. Finally, while Klinge (2008:245) provides the decompositions in 16b and 17b, he discusses neither the details of the internal structures nor how the individual segments combine. For Klinge, all determiners—including demonstratives—are under D.

Leu (Reference Leu2008) is more explicit about the morphosyntactic structure of determiners. Drawing a parallel to other complex projections, he provides a syntactic account of demonstratives. He proposes that these types of determiners involve extended adjectival projections (Dryer Reference Dryer1992:120–122; see Grimshaw Reference Grimshaw2005). Leu assumes that these elements consist of a—what he calls—definite marker d-, an adjectival agreement head (AgrA), and a demonstrative adjective at the bottom of the tree.Footnote ⁶ The example in 18a is analyzed as in 18b.

1. (18)

   

The structure in 18b sits in Spec, DP. Cases where the demonstrative adjective does not surface are explained by the assumption that the moved demonstrative adjective remains unpronounced (see Kayne Reference Kayne, Broekhuis, Corver, Huybregts, Kleinhenz and Koster2005). Non-pronunciation is marked by strikethrough. The derivation of 19a is shown in 19b.

1. (19)

   

Treating both demonstratives and adjectives in the same way affords Leu an account of the strong/weak alternation of adjective endings. Pointing out the well-known fact that the inflections on definite determiners and those on strong adjectives are the same, Leu proposes that AgrA hosts the strong inflection. Now, if d- is merged inside xAP, as in 18b, it precedes AgrA, and the determiner surfaces with the strong ending while the adjective has a weak ending. In contrast, if d- is not merged, then the adjective moves in front of AgrA “picking up” the strong inflection. This is an elegant account of the identical inflections on determiners and adjectives. Later in his discussion, Leu extends this analysis to the interaction between complementizers such as dass ‘that' and the position of finite verbs. Furthermore, Leu applies these adjectival structures to non-d-determiners such as welch ‘which', solch ‘such', mein ‘my' and others. To sum up, abstracting away from movement, Leu postulates three syntactic components in the makeup of demonstratives.

Leu (Reference Leu2008) takes evidence for the higher position of the unpronounced demonstrative adjective in Norwegian in 19b from Afrikaans, which shows this distribution overtly. For instance, with the complex demonstrative hierdie ‘this', the demonstrative adjective hier precedes the determiner die (I return to this example below). If I interpret Leu Reference Leu2008:28 correctly, he claims that Norwegian differs from Afrikaans either with regard to crosslinguistic differences in the (non)pronunciation of demonstrative adjectives or with regard to the exact landing site of the demonstrative adjectives, which in turn regulates the actual pronunciation more generally. It should be mentioned that these two options are not discussed in much detail in Leu's work. It is instructive though to explore these two suggestions a little further here.

Starting with the pronunciation of the demonstrative adjective in Afrikaans versus its nonpronunciation in Norwegian, note that this dichotomy should ideally be tied to another property. A potentially relevant distinction can be identified in the different stress patterns of these complex demonstratives. To see this, notice first that noun phrases with an article have their nuclear stress on the noun, but noun phrases with a demonstrative are stressed on the determiner itself (for Afrikaans, see Ponelis Reference Ponelis1993:168 and Donaldson Reference Donaldson1994:491; for Yiddish, see Margolis Reference Margolis2011:122). Compare 20a,c to 20b,d.

1. (20)

   
   

Now, as explicitly stated in Ponelis Reference Ponelis1993:168, the Afrikaans demonstrative has stress on the adverbial element, as in 21a. In contrast, Weinreich (Reference Weinreich1999:191) and Schaechter (Reference Schaechter2003:58) point out that stress falls on the determiner in Yiddish, as shown in 21b.

1. (21)

   

In other words, while Afrikaans híerdie seems to have compound stress, the Yiddish sequence ot dér does not. I believe that this makes it unlikely that ot and dér form a compound-like element in the latter language; that is, that ot is part of dér.Footnote ⁷ To drive this point home, note also that another element (-o) can optionally intervene between ot and dér yielding oto dér ‘this'.Footnote ⁸ This then presents a clear difference between Afrikaans and Yiddish. However, note that the preceding demonstrative adjective is pronounced in both languages, and that the different stress patterns cannot help explain the unpronounced demonstrative adjective in Norwegian.

As a second option to explain the difference in pronunciation of the demonstrative adjective, Leu suggests different landing sites for these elements. Note first that when regular adjectives move to pick up the strong inflection, d- is not merged; that is, d- is absent (for example, guter ‘good'). In contrast, movement of the demonstrative adjective is not affected by the presence of d-. In fact, d- must be present: *hierie and *oter (note that the obligatory presence of d- has nothing to do with the semantics; see the discussion of preposition-article contractions in Leu Reference Leu2008:28). One could interpret this difference in the presence versus absence of d- by claiming that demonstrative adjectives move higher than regular adjectives. In other words, there are at least two positions in front of AgrA.

Now, in order to explain the difference in pronunciation of the demonstrative adjective in Afrikaans versus Norwegian, one could suggest that Norwegian moves its demonstrative adjective to a third, even higher, landing site where nonpronunciation is licensed. As Leu (2007:147) tentatively suggests, this is perhaps so because the demonstrative adjective is at the edge of a phase in the sense of Chomsky Reference Chomsky, Martin, Micheals and Uriagereka2000 and subsequent work. Taking stock of the discussion of this option, one winds up with three different landing sites for adjectives preceding AgrA. Consequently, the adjectival structure must involve a fairly complex left periphery for which no independent evidence is provided.

To sum up, both accounts of the unpronounced demonstrative adjective in Norwegian are not without issue. At best, this state of affairs calls for further detailed investigation and ideally some empirical evidence to argue in favor of any one option. At worst, this casts doubt on Leu's proposal of demonstrative structures (for more critical discussion, see Roehrs Reference Roehrs2010, section 2.1).Footnote ⁹ Finally, Leu does not discuss complex demonstratives, determiners involving the deictic element -s-, in much structural detail (for some brief remarks, see, for instance, Leu Reference Leu2008:18, 32, 35–36). The internal structure of these demonstratives is of primary interest to me here.

Turning to the third proposal, Roehrs (Reference Roehrs2010) assumes a different internal structure of demonstratives. Illustrating the different components with German dér ‘that' and dieser ‘this', the deictic part of the demonstrative, null +Ø- and +ies-, projects a Deictic Phrase (DeicP) and the definite marker d+ is housed in a Demonstrative Phrase (DemP). FP is an optional functional phrase projected when certain adverbial elements are present. The structure in 22 is located in Spec, DP.

1. (22)

   

Based on Brugè Reference Brugè1996 and Bernstein Reference Bernstein1997, Roehrs Reference Roehrs2010 provides a fairly comprehensive discussion of the different combinations of determiners and adverbial elements—the latter called REINFORCERS—in the Germanic and Romance languages (for discussion, see section 4.1). Identifying four different types of demonstrative-reinforcer constructions, this paper connects the varying distributions of these types to the different determiner systems in the two language families. As a consequence, Roehrs Reference Roehrs2010 provides good evidence that demonstratives are complex elements that can surface in different specifier positions in the larger noun phrase.

One nontrivial aspect of the morphosyntax of demonstratives that Roehrs' proposal does not address is the status of the inflectional morphology on the determiners in 22. It is important to discuss this type of element, since attributive adjectives such as guter ‘good' have the same inflection. Not doing so misses an important generalization. Furthermore, demonstratives such as jener ‘that' and other types of determiners were not discussed in much detail in that paper.

In the next section, I argue that the structure for the demonstratives in 22 is not quite accurate in two ways: First, it is not fine-grained enough to take the inflection on the demonstratives into consideration. Second, it postulates two phrases (DemP and FP) that are not well motivated. To be more specific, I argue below that the inflectional morphology on determiners involves a separate head in the syntax (Infl), and that the demonstrative stem involves just one head (Deic). I consider this in more detail below.

3.1. Strengthening Hypothesis 1

In the introduction, I briefly discussed the definite determiners in German, English, and Icelandic. In 23, I contrast these determiners with the adjective lucky in the relevant languages.

1. (23)

   

As can be seen in 23a, German has three related determiners sharing the definiteness marker, and all the determiners have the same inflections as the adjective. In other words, the two elements under discussion occur in tandem allowing a transparent parse into separate components. This is summarized in 24a. However, I have also shown that definiteness markers and regular inflection can occur separately. English in 23b shows related determiners but irregular endings, and German jener ‘that' in 23a shows that an unrelated determiner can have a regular inflection. This is summed up in 24bi. Finally, Icelandic þessi ‘this' and sá ‘that' in 23c illustrate that both the definiteness marker and regular inflection can be absent at the same time. This is stated in 24bii.

1. (24) Hypothesis 1: Synchronic Relatedness of Definite Determiners

   1. a. Definite determiners are transparent with regard to the definiteness marker and the inflection.

   2. b. Definite determiners may also be partially or completely opaque:

      1. (i) Partial opacity: determiners do not share either the definiteness marker or the inflection.

      2. (ii) Complete opacity: determiners share neither the definiteness marker nor the inflection.

Now I cast the empirical net wider. The objective here is to determine if the initial juxtaposition of German, English, and Icelandic above is representative of all the Germanic languages. If so, one can strengthen this hypothesis. For that purpose, I have investigated the definite determiners of twelve Germanic languages. First, I focus on the determiner stem forms.

The twelve Germanic languages investigated here are, in alphabetical order: Afrikaans (Af), Danish (Da), Dutch (Du), English (En), Frisian (Fr), German (Ge), Icelandic (Ic), Norwegian (No), Pennsylvania German (Penn. Ge), Swedish (Sw), West Jutlandic (W. Jutl), and Yiddish (Yi). Table 1 below provides an empirical overview of the determiner systems of these languages. The data are sorted by the type and number of determiners: The columns of the table are organized by the number and type of article, and the rows—by the number and type of demonstrative. Specifically, d-art and h-art stand for articles starting in d- and h-; d-dem and h-dem stand for demonstratives starting in these elements. These are the related stem forms of articles and demonstratives. Determiners without a d- or h- are unrelated and underlined. In the cells containing the concrete data, a dash separates the article from the demonstrative(s); commas separate proximal from distal demonstratives; slash signs separate different distal demonstratives.

To illustrate the presentation of the data in Table 1 for one language, Icelandic is in column one and row four of the data cells. This language has four determiners: one article and three demonstratives, where one of the demonstratives is related to the article (both share h-). The other two demonstratives are unrelated and underlined. The article can be free-standing in preadjectival position in literary Icelandic (hinn) or suffixed to the head noun in common Icelandic (-inn). These two options are represented by putting the h- in parenthesis (the Mainland Scandinavian languages are different in this respect, see section 3.3).Footnote ¹⁰ Furthermore, Icelandic has one proximal demonstrative (þessi) and two distal demonstratives (sá, hinn).Footnote ¹¹

Studying Table 1, one can observe that all contemporary Germanic languages have at least one article and at least one demonstrative. In fact, it is striking that every member of this language family has at least one article and at least one demonstrative that share the definiteness marker. This marker is either a dental sound or h-, the latter only occurring as a definiteness marker in the Insular Scandinavian languages (note that the Dutch article het is an isolated form).Footnote ¹² In other words, all Germanic languages have related determiner stems. This means that the above-mentioned fact is not accidental and should be taken seriously.

Table 1.

Definite determiners in Germanic.

There is one minor qualification to this general statement. Languages such as German have a second demonstrative element that is not related to the article (jener ‘that' versus der ‘the'), and languages such as Dutch have a second article that is not related to the demonstrative elements (het ‘the' versus dit ‘this'). In fact, there is a third type of language, Pennsylvania German, that combines the general patterns of German and Dutch. Specifically, this language has related articles and demonstratives but also an unrelated article (es) and unrelated demonstratives (sell-). This means that there is no connection between the presence of an un-related article and the presence of an unrelated demonstrative. To be clear, though, with the exception of a few determiner forms, all Germanic languages have related definite determiners.Footnote ¹³

1. (25) Hypothesis 1a more generally

   All contemporary Germanic languages have at least one article and at least one demonstrative that share the definiteness marker. In addition, these languages may have an unrelated article, or an un-related demonstrative/demonstratives, or both at the same time.

Note that while this hypothesis may not be very revealing if one only considers the Germanic languages, it becomes more important once other languages are taken into account.

Consider some of the Romance languages. While there are many interesting points to observe, it is particularly striking that the Romance articles are not related to their demonstratives in any obvious way (see also Kayne & Pollock 2010; Roehrs Reference Roehrs2010:238, a response to a reviewer's comment):Footnote ¹⁴

1. (26)

   

I take this difference between the Germanic and Romance languages to be significant. In other words, I claim that transparency in definite determiner stems is another typological difference that sets Germanic apart from Romance. I interpret this common trait in Germanic such that the article and the demonstrative are morphosyntactically related. While I provide a fairly comprehensive treatment of the contemporary Germanic languages, I only briefly comment on the Romance languages restricting most comments to Spanish and Sardinian. Next, I turn to the inflections on determiners.

As just discussed, determiners can be grouped with regard to their initial element, the definiteness marker. However, they can also be classified in terms of the regularity of their endings. I define regularity of inflection as follows: If a determiner has the same inflection as an adjective, then the inflection on the determiner is regular. Before classifying the determiner systems in terms of their inflections, I need to make some general remarks about inflections on adjectives.

As is well known, the Germanic languages have two types of adjectival inflections, traditionally called WEAK and STRONG. Weak endings are fewer in number making fewer overt distinctions than their strong counterparts. The stereotypical distribution is that weak inflections occur on adjectives following definite articles, as in 27a, and strong inflections surface on adjectives occurring with indefinite articles, as in 27b.

1. (27)

   

All Germanic languages exhibit this pattern, except for English, which does not have adjective endings anymore, and Afrikaans, which is special (see section 3.4). With these general remarks in place, it is important to decide which of the two inflections, weak or strong, is relevant for classifying the determiners. To make an informed decision, I consult some recent work on this topic.

Leu Reference Leu2008 and Schoorlemmer Reference Schoorlemmer2009 provide uniform accounts for the strong/weak alternation of adjective inflections across the Germanic languages. As discussed in section 2.2, Leu (Reference Leu2008) proposes that under certain conditions, adjectives move to “pick up” the strong inflection. Schoorlemmer (Reference Schoorlemmer2009) provides an Agree-based account. To be clear, each author argues for one basic mechanism held accountable for the alternation.

Based on Harbert Reference Harbert2007:135, Roehrs & Julien (2012) have recently challenged the view that the Germanic languages can be given a uniform analysis. Discussing nine definite contexts, Roehrs & Julien (2012) demonstrate that German consistently shows a strong ending, and Norwegian—a weak ending. They conclude that German is lexicostructural in nature, but Norwegian involves the semantics.

To briefly illustrate, German allows a weak ending in an indefinite context, as in 28a, and a strong ending in vocatives, which are definite in interpretation, as in 28b.

1. (28)

   

Without going into too much detail here, weak endings in German are triggered by certain determiners in a regular DP structure. Specifically, the example in 28a involves a regular structure and is explained if determiners are lexically marked to trigger a weak inflection on the adjective; 28b is accounted for if vocatives have a structure different from regular DPs, where a weak ending is not licensed. In contrast, Norwegian is semantic in nature (see Lohrmann Reference Lohrmann2011). If the noun phrase containing the adjective is definite in interpretation, the adjectival inflection is weak, as in 29a. However, lack of definiteness in the predicative use of the adjective brings about a strong ending, as in 29b.

1. (29)

   

To be clear, Roehrs & Julien (2012) argue for different language-specific mechanisms that bring about the weak adjective inflection; that is, weak inflections are subject to different language-specific conditions. These authors go on to claim that once these differing conditions for the weak inflections in German and Norwegian are identified, the distribution of the strong endings can be interpreted as the elsewhere case in both languages.

I assume that inflections spell out abstract features on terminal nodes. Recalling that the strong inflections have more diverse forms, Roehrs & Julien (2012) state that they make more underlying features visible. While they undergo regular concord agreement in nominal features, strong endings are not subject to specific conditions but wear their underlying features on their sleeves. As such, I take the strong inflections on adjectives as the reference point when I determine the regularity of the inflection on determiners. With this in mind, one can distinguish three types of languages.

German and Yiddish belong to the first type. In these languages, definite determiners have the same inflections as adjectives (the Pennsylvania German determiner system is similar to Yiddish):

1. (30)

   

Returning to the Romance data set in 26 for a moment, note that Sardinian is the only language with regular inflection (see section 5.2).

The second type involves eight languages all showing—what I have called—irregular endings. These languages can be put in four subgroups, where the languages in parentheses behave in a similar way to the language preceding them: Danish (Norwegian, Swedish), Dutch (Frisian), English (Afrikaans), and West Jutlandic have the following patterns (31d is adopted from Julien Reference Julien2002:265, where I added (') to the adjective, which represents a glottal stop; see Delsing Reference Delsing1993:230).

1. (31)

   

To use one language for the purposes of illustration, the neuter article det and the distal demonstrative dét in Danish in 31a are pronounced without the final -t thus differing from the adjectival inflection.Footnote ¹⁵ I turn to the third language type and twelfth language in my data set.

At first glance, Icelandic determiners seem to pattern with the determiners in English-type languages in that they have different inflections from the adjectives, as shown in 32a. However, when one considers adjectives ending in -in (and most participles formed from strong verbs), then the homophonous article and demonstrative do pattern with the adjectives, as shown in 32b.Footnote ¹⁶

1. (32)

   

To be clear, related determiners in this language have regular inflection, but unrelated ones do not.

In recent work, Katzir (2011:66) makes the convincing case that the different types of endings on the adjectives in 32a,b can be given a phonological explanation. Exemplifying with the masculine nominative singular forms, there are two (independently motivated) rules at play, where 33a applies before 33b.

1. (33)

   

In more detail, the application of 33a to /heppinr/ will yield heppinn (and turns /hinr/ into hinn), and the application of 33b to /ríkr/ will yield ríkur.

More generally then, Icelandic is interesting in that the article (h-)inn and the demonstrative hinn share the same stem, and that they have the same inflections as adjectives.Footnote ¹⁷ At the same time, there are two types of demonstratives (þessi, sá) that have different stems and irregular endings. This makes Icelandic a “mixed” type of language: With one set of determiners, inflections are regular as in German, and with the remaining determiners, the endings are irregular as in English. Cross-linguistically, the West and North Germanic languages have determiners either with regular inflections (German, Icelandic) or with irregular endings (English, Icelandic). This is summed up as follows:

1. (34) Hypothesis 1b more generally

   The contemporary Germanic languages differ in the way their definite determiners exhibit inflectional regularity. There are three cases: All the determiners in a language exhibit the same inflection as adjectives; none do; some determiners do, but others do not.

   To sum up this subsection, I have provided a survey of definite determiners with regard to relatedness of stems and regularity of inflections strengthening the initial hypothesis to cover all the Germanic languages.

3.2. Internal Structure of Definite Determiners and DPs

The main goal of this paper is to illuminate the nature of the definiteness marker and inflection with regard to syntax, and to identify some consequences for the DP as a whole. In the last subsection, I identifed individual components of the determiners and formulated hypotheses about their occurrence. With this in place, I can now proceed to the second step in the segmentation approach, namely, to matching these overt components with heads in the syntactic structure and determining the sequence of these heads.

More to the point, I interpret the components identified above as overt exponents of abstract features. I assume that these features project heads in the syntax. To be clear, matching the overt components with abstract features—and thus structure—yields the inner makeup of determiners. This allows me to develop an explicit and detailed analysis of the structure of determiners, in particular the definiteness marker and the inflection, compatible with general (minimal) structural assumptions in the literature.

Recall from the introduction that I assume that the definiteness marker enjoys a special status. Now, I provide evidence that the presence of this element cannot be used as a diagnostic for definiteness (Lyons Reference Lyons1999); that is, the presence of a definiteness marker does not entail a definite interpretation, and vice versa. I provide four pieces of evidence for each direction of a potential entailment relation. For ease of reference, I keep the traditional term for definiteness marker.

I start by pointing out that a definite interpretation of the nominal does not entail the presence of a definiteness marker. Recalling Table 1, seven languages have unrelated definite determiner forms where the article and/or the demonstrative do not start with the definiteness marker, as in 35a. Furthermore, personal pronouns lack this element, as in 35b. Third, proper names in many languages do not involve a definiteness marker at all, as shown in 35c. To give the last piece of evidence, Lyons (Reference Lyons1999:15) points out that possessives and universal quantifiers are also definite elements, as in 35d.

1. (35)

   

In each of the cases in 35, the interpretation is definite, but there is no definiteness marker. In section 3.3, I return to this part of the discussion in the context of the suffixal article in the Scandinavian languages.

Conversely, and more importantly, the presence of the definiteness marker does not entail definite interpretation. First, Longobardi (Reference Longobardi1994) argues that definite articles preceding proper names are expletive elements, as in 36a; that is, they have no semantics of their own and are merged in definite contexts. Second, certain subordinating conjunctions, such as in 36b, that developed from the neuter form of the demonstrative (see van Gelderen Reference Gelderen2011:259; also Diessel Reference Diessel1999:123), are often held to be devoid of definiteness semantics (for example, Lasnik & Uriagereka Reference Lasnik and Uriagereka1988:97, Paul et al. Reference Paul, Wiehl and Grosse1989:395, 438).Footnote ¹⁸ Third, there is evidence from English that even—what looks like—demonstratives can be used in indefinite contexts, as in 36c (Ionin Reference Ionin2006:179).Footnote ¹⁹ Finally, Bernstein (Reference Bernstein, Høeg Müller and Klinge2008) argues for English that the definiteness marker does not indicate definiteness. This is particularly clear for existential there in 36c. In a similar vein and following the basic parsing in Klinge Reference Klinge, Høeg Müller and Klinge2008:234, existential-do and a second adverbial do can appear in existential contexts in Yiddish (Kahn Reference Kahn2012:41, Zucker Reference Zucker1994:20), as shown in 36d.

1. (36)

   

In fact, the English translation in 36d clearly indicates that the definiteness marker on there cannot be tied to (distal) deixis either as this would lead to a semantic clash with the proximal adverbial here (see Hazout Reference Hazout2004:396).

To sum up, it is clear that the definiteness marker on the determiners does not bring about definiteness. In fact, with the possible exception of Afrikaans, there are at least eleven (out of the twelve) languages in Table 1 that militate against the idea that definiteness markers are closely tied to definiteness semantics. It is clear, then, that d by itself cannot be responsible for definiteness (see also Leu Reference Leu2010). Rather, I claim that the definiteness marker is a semantically vacuous element that performs a different task.

I propose that the definiteness marker does not point to syntactic structure (for now, I put the definiteness marker in parentheses, but I elaborate on its status in section 3.3). As such, articles only involve an inflection, and demonstratives only consist of an inflection and a deictic part. I assume that the inflection and the deictic part involve abstract features. I propose that the inflectional features involve a head in the syntax (I refer to it as Infl), and so does the deictic feature (I refer to it as Deic). I assume that the latter attracts stress. Locating the semantic and phonotactic properties on Deic captures the frequently expressed intuition noted in section 2 that the demonstrative consists of a definite article (here: Infl) and another element (here: Deic). Abstracting away from the definiteness marker, notice that there is a one-to-one match between the overt components and the abstract syntactic structure. To be clear, articles involve one head in the syntax, and demonstratives involve two. I can now update the main hypothesis from the introduction:

1. (37) Hypothesis 3: Inner Makeup of Definite Determiners

   Articles only involve an inflection, which means they have one syntactic head: Infl.

   Demonstratives only involve a deictic element and an inflection, which means they have two syntactic heads: Deic and Infl.

Having discussed the number of heads of determiners, I turn to the order of these heads in the syntax. I argue for extended projections where a head builds a complex projection, similar to 22 above. Unlike in 22, however, this complex projection is closed off by an inflectional phrase (InflP) at the top. For articles, the head of the projection is the head noun of the DP; for demonstrative, the head is the deictic part. Schematically, the two structural proposals are given in 38a and 38b.

1. (38)

   

2. 

Note that this proposal borrows ideas from Klinge Reference Klinge, Høeg Müller and Klinge2008, Leu Reference Leu2008, and Roehrs Reference Roehrs2010. It is similar to Klinge in that I follow some of his assumptions about the morphological components of the demonstrative; it is related to Leu in that I assume an adjective-like extended projection of the demonstrative. Unlike Klinge, however, I propose that all demonstratives have a deictic element, and I depart from Leu (following Roehrs Reference Roehrs2010) in assuming that the deictic part of the demonstrative (and not the adjectival element) is the head of the extended projection. An important consequence is that the structures in 38a,b are morphosyntactically the same. This makes the DP as a whole and its subparts uniform in structure. In the next part of this subsection, I provide more details and motivation for these structures. I begin by discussing the internal structure of demonstratives.

In section 3.1, I defined regularity in inflection with regard to adjectives. Determiners, adjectives, and quantifiers may take the same inflection (for reasons of space, I do not discuss quantifiers in much detail). In order to avoid issues with regard to the strong/weak alternation of adjective endings, I provide the relevant examples in the dative plural, where the strong and weak endings are identical:

1. (39)

   

Crucially, adjectives and demonstratives also share other properties. To begin with, note that modifiers have been used to explore the structure of their containing phrases (for example, for adjectives in the noun phrase, see Cinque Reference Cinque, Cinque, Koster, Pollock, Rizzi and Zanuttini1994). Now, adjectives and demonstratives both allow modifiers: Adjectives take degree words (Corver Reference Corver1997) and demonstratives take adverbial elements (Bernstein Reference Bernstein1997). In more detail, one can observe that the degree word for enough may either follow or precede the adjective. Contrast English with Yiddish:

1. (40)

   

Similarly, adverbial elements for here may either follow or precede the demonstrative:

1. (41)

   

To be clear, with both the adjective and the demonstrative, the same modifying elements are in different positions in different languages. The same basic point can be made in another way, namely, lexically different modifiers can also occur in different positions in one and the same language. Compare 40a to 42a, where English modifiers can appear on either side of the adjective, and 41b to 42b, where Yiddish modifiers can surface on either side of the demonstrative.

1. (42)

   

To sum up, modifying elements can be in different positions both cross-linguistically and language-internally.

In formal linguistics, differences in the syntactic distribution of elements are often captured by movement of these elements. If the same approach is applied to the combinations involving modifiers, then one can conclude that modifiers are part of a complex structure accommodating this displacement (I give the details of the syntactic interaction of demonstratives and adverbial elements in section 4.1). Now, given that the inflectional and distributional patterns are so similar, one can hypothesize that adjectives and demonstratives involve parallel structures (see Dryer Reference Dryer1992, Leu Reference Leu2008, Roehrs Reference Roehrs2010).

I propose that all elements in the noun phrase are split up into an inflectional ending and a stem (see Chomsky Reference Chomsky1957 on verbal morphology; see Leu Reference Leu2008 on the DP). In other words, the phrase projected by the stem has an inflectional phrase on top of it:

1. (43)

   

To provide a principled account for the sequence of these heads, I interpret the structure in 43 as an extended projection (Grimshaw Reference Grimshaw2005).

To make this idea more concrete, I follow Corver Reference Corver1997 in that adjectives build up extended projections. To capture the commonalities between adjectives and demonstratives, I propose the same for demonstratives. Making minimal configurational assumptions, the adjective projects an Adjective Phrase (AP), and the demonstrative—a Deictic Phrase (DeicP). Both involve extended projections that are closed off by an Inflectional Phrase (InflP):Footnote ²⁰

1. (44)

   

I assume here but argue in section 4.1 that Deic in 44b undergoes head movement to Infl:

1. (45) Demonstrative Structure after Head Movement

   

Note that there are some immediate advantages to this parallelism. With a syntactically active component (Infl) and a component with semantic content (Deic) present, 44b accommodates both elements on minimal structural assumptions. Note that this avoids the proliferation of abstract functional structure not warranted by overt components. Further-more, both structures have only one root element with semantic content (either A or Deic) that builds up the extended projection. In other words, if demonstratives involve extended projections, they have only one semantic component, and that component is at the bottom of the projection. Given this theoretical claim, one may expect demonstratives to behave similarly crosslinguistically.Footnote ²¹

Finally, if the structures of adjectives and demonstratives are indeed similar, then the inflectional and distributional similarities discussed above are no longer coincidental but receive a principled account. This explains the regular inflections of demonstratives in hypothesis 1b in 34.Footnote ²² Compared to demonstratives, definite articles lack a deictic element. I propose that they involve only an inflectional head in the syntax. Recalling the hypothesis that the definiteness marker is semantically vacuous, definite determiners per se cannot be responsible for definiteness.

I propose that there is a feature in the syntax, [DEF], that determiners make visible. I follow Roehrs Reference Roehrs2009 and Schoorlemmer Reference Schoorlemmer2012 in that this feature originates lower in the nominal structure and undergoes head movement to D. With certain assumptions, this explains the double definiteness in some of the Scandinavian languages, a pattern that shows the presence of two definite elements at the same time (among many others, Delsing Reference Delsing1993, Anderssen Reference Anderssen2005, and Julien Reference Julien2005). To make the proposed structure fully parallel to the structure in 43, I replace D, which traditionally stands for determiner (Abney Reference Abney1987), with Infl_N. To avoid confusion, the diacritic _N on Infl indicates that this inflectional head closes off the extended projection of the head noun. In other words, demonstratives and articles both involve extended projections but differ in the head of the projection (Deic versus N). On par with demonstratives, then, this explains the regular inflections on articles in hypothesis 1b. With this in place, I locate the different definite determiners inside the larger nominal structure.

Recalling the general structure in 43, I propose that adjectives and quantifiers are in specifier positions. Specifically, the quantifier structure is located in the specifier of a Cardinal Phrase (CardP), and the adjective structure is housed in the specifier of an Agreement Phrase (AgrP). Consider 46. The definiteness feature [DEF] originates in the head position of a Definiteness Phrase (DefP) and moves to adjoin to Infl_N.Footnote ²³ Finally, the head noun is under XP at the bottom, which is fleshed out below (for more detailed discussion of the structure of the noun phrase, see Julien Reference Julien2005, Roehrs Reference Roehrs2009, and references cited therein). Recalling that stems move to pick up their inflection, the nominal in 39a is analyzed as follows:

1. (46)

   

Now I flesh out the XP at the bottom of the tree in 46. Since Abney Reference Abney1987, most scholars take determiners to be part of the extended projection of the noun. What is interesting from the current perspective is that unlike the extended projection of demonstratives and adjectives, the extended projection of the noun has dual exponents where both the determiner and the head noun show inflections. Consider again the example in the dative plural shortened to d-en Kind-er-n ‘to the children'. Importantly, though, these inflections are not identical. The head noun provides the gender feature, Num gives number, and K supplies case:

1. (47)

   

The head noun undergoes movement to Num and then to K resulting in Kindern ‘children'. Returning to 46, XP stands for KP. Now, the first crucial difference between the inflection on Infl_N and the inflections inside KP is that there is no overt marking for gender inside KP. Second, number and case involve two separate heads inside KP. Besides these structural differences, there is also a functional one. To use Corbett's (2006) terminology, note that the features inside KP are agreement controllers, while the features on Infl_N are agreement targets. Now, since the features inside KP and the ones on Infl_N are different in nature, this is not a case of true dual exponence after all, and two sets of inflections as part of the same extended projection are not unexpected.

Before I turn to the discussion of demonstratives inside the DP, I would like to be more specific about the interaction between the definite article and [DEF]. After head movement of [DEF] from Def to Infl_N (formerly D), one obtains the following simplified adjunction structure:

1. (48)

   

There are two elements in 48 that need licensing: [DEF] and the inflectional suffix.

With Longobardi Reference Longobardi1994, I assume that definite DPs must be made overt, either in Infl_N by the article or in Spec, Infl_NP by the demonstrative (see also Julien Reference Julien2005, Roehrs Reference Roehrs2009, among many others). In other words, an overt element inside Infl_NP licenses the presence of [DEF]. As to the inflectional suffix, these bound morphemes require a host (see Lasnik's Reference Lasnik, Hornstein and Lightfoot1981 Stranded Affix Filter). In section 3.3, I propose that similar to do-support in English, the definiteness marker is inserted to provide such a host for the inflection. Note now that the inflection supported by the definiteness marker makes Infl_NP overt and thus licenses [DEF] at the same time. I show below that providing a host for the inflection and making [DEF] visible coincide in most of the languages (one exception is English, see section 3.4). I return to the main line of argumentation.

Unlike articles, demonstratives involve two heads. As phrases, I locate them in Spec, Infl_NP. The example in 39b is analyzed as 49 (see below). With the demonstrative present, the definiteness feature is made visible. I propose that the presence of an element in Spec, Infl_NP licenses the non-pronunciation of Infl_N (nonpronunciation is marked by strike-through).Footnote ²⁴

Some remarks are in order here. The reason why the demonstrative (but not the article) is pronounced presumably has to do with the fact that the demonstrative comprises all the features of the article. As such, the deletion of the article can be recovered when the demonstrative is present but not vice versa. Furthermore, putting the demonstrative structure InflP in a specifier position of the matrix noun phrase is fully compatible with Cinque Reference Cinque, Cinque, Koster, Pollock, Rizzi and Zanuttini1994, Reference Cinque2010, where it is argued that adjectives are specifiers as well. Moreover, there is evidence that the demonstrative and the article are indeed in different positions.

1. (49)

   

Conditions on the licensing of the nonpronunciation of overt material vary crosslinguistically. For instance, the Doubly Filled COMP Filter originally formulated by Chomsky & Lasnik (Reference Chomsky and Lasnik1977) has been shown to hold in some languages only. Specifically, Haegeman (Reference Haegeman1994:382–384) provides a number of examples where the fronted question word and the complementizer can both surface at the same time. In view of the often-claimed parallelism between the sentence and the noun phrase (for instance, Abney Reference Abney1987), one might expect that both the demonstrative and the definite article can be spelled out too (Giusti Reference Giusti and Haegeman1997:109). While not possible in West Germanic, a doubly filled Infl_NP can be found in all the major North Germanic languages. This is exemplified here with Nor-wegian (see Roehrs Reference Roehrs2010:236; 50a is adopted with a slight modification from Julien Reference Julien2005:109). Importantly, none of the Germanic languages allows a definite article to precede a demonstrative, as in 50b.

1. (50)

   

The general absence of the pattern in 50b follows from the structure in 49, where the demonstrative is in the specifier position and the article is in the head position of the same phrase (that is, Infl_NP).

To give a second piece of evidence in favor of the different positions of definite determiners, Giusti (Reference Giusti1996:107) points out for Italian that extraction from a DP involving a definite article, as in 51b, is fine, but not when a demonstrative is present, as in 51c. The example in 51a shows the basic structure.

1. (51)

   

Note that both DPs are definite in interpretation, the crucial difference being the presence of the article versus the demonstrative. I follow Giusti Reference Giusti1996 in that Spec, DP (here: Spec, Infl_NP) is an escape hatch for wh-movement. Locating the demonstrative in this specifier position blocks movement in 51c. In contrast, the article is in a head position and does not block this movement in 51b. I take this as evidence that the structure in 49 is on the right track. In the next subsections, I employ these structural assumptions to account for the bulk of the data.

3.3. Hypothesis 1a: Related and Unrelated Stem Forms

The account of related versus unrelated determiner forms consists of two parts that work in tandem: the special status of the definiteness marker and the dichotomy of bound and free morphemes. To make a stronger case for the special status of the definiteness marker, I first lay the relevant groundwork by discussing bound and free morphemes. After detailing my assumptions about the definiteness marker, I discuss some isolated exceptions.

As stated in hypothesis 1a in 25, all Germanic languages have related articles and demonstratives. In addition, some languages have a few unrelated determiner forms. I illustrate this again with examples from Pennsylvania German and Icelandic, where the former are separated by semicolon from the latter. Related determiners are provided in 52a, and unrelated forms in 52b.

1. (52)

   

There is a well-known dichotomy between free and bound morphemes (for example, Matthews Reference Matthews1991:210). For current purposes, this can be exemplified with spatial expressions in German, where adjectives derived from nouns can be free, as in 53a, or bound, as in 53b.

1. (53)

   

In what follows, I propose to extend this well-established dichotomy to the determiner system in Germanic. Returning to Pennsylvania German and Icelandic, I propose that determiners can involve free morphemes, as in 54a, and bound morphemes, as in 54b (for English, see Kayne & Pollock 2010).

1. (54)

   

Unlike free morphemes, their bound counterparts are in need of another element on their left side, the definiteness marker d in Pennsylvania German or the definiteness marker h in Icelandic. Note also that another element can, at least with some forms, be added to the right of free morphemes, as in 55a. This is usually not possible with bound morphemes such as inflections, as shown in 55b.

1. (55)

   

With this in place, I discuss my assumptions about vocabulary insertion.

To be inserted in the syntactic tree, a vocabulary item—a word or an affix—must contain a subset of the features of the terminal node (for example, Halle & Marantz Reference Halle, Marantz, Hale and Keyser1993:122). Continuing with masculine forms from Pennsylvania German, I suggest the following vocabulary items for seller ‘that' and dér ‘this', where 56a,b yield the relevant stem forms, and 56c yields the inflection. The simplified feature specifications on the left define the featural context of the insertion of these vocabulary items under the relevant terminal nodes.

1. (56)

   

Note that the feature [PROX] in 56b has no overt vocabulary item. For convenience, this is indicated by Ø in this paper. Note though that demonstratives are typically stressed. I assume that the abstract deictic feature attracts stress. If so, then one can state that this segmentally unrealized deictic feature shows its existence in the phonology, where this particular demonstrative can be stressed once a vowel has been inserted. Note that this does not exclude the possibility that articles can also be stressed; for instance, English the can be pronounced as [Ði:] under certain conditions. As to 56c, I assume that this specification includes features for lexical contexts such that it is applicable to both determiner and adjective structures, but it cannot surface on the head noun itself.

Returning to the demonstrative structures from above, suppose that the Deic-head is specified as [PROX] in 57a and as [DIST] in 57b. The Infl-head is marked as [MASC] in both cases. Recall that Deic undergoes head movement to Infl. Given the feature specifications of the Deic-head, 56a can only be inserted in 57b, and 56b can only be inserted in 57a. As for the Infl-head, 56c is inserted in both.

1. (57)

   

The derivations for 57a,b are finalized below. Notice also that 56b does not “block” the insertion of other (overt) elements, as there is no other lexical item specified for [PROX] that is less specified than 56b itself. The vocabulary items in the other languages with homophonous demonstratives (Afrikaans, German, the Mainland Scandinavian languages, and Yiddish) can be represented in a similar fashion. To save space, I do not do this for all languages. Returning to 57, the same applies to Yiddish modulo some minor differences.

German is a bit more complicated. This language has three demonstratives, one proximal (dieser) and two distal (dér, jener). Since the distal demonstratives are very similar in their specifications, they are competing with one another for insertion. In order to account for the proper vocabulary insertion, I assume with Halle & Marantz Reference Halle, Marantz, Hale and Keyser1993:122 that vocabulary items with more features matching those of the terminal node get inserted blocking the insertion of items with fewer matching features. Now, it seems that jener ‘that' is typically used in discourse (but not situation) deixis.Footnote ²⁵ I propose the following two vocabulary items for the distal demonstratives in 58a,b and provide the entry for the proximal demonstrative dieser in 58c.

1. (58)

   

The presence of the feature [DISCOURSE] under Deic allows the insertion of 58a blocking that of 58b.

I turn to the articles. Unlike demonstratives, articles involve only one head in syntax. Like demonstratives, I propose that they also involve bound and free morphemes. One can suggest then for Pennsylvania German that the masculine article der involves a bound morpheme, and the neuter article es is a free form. After vocabulary insertion in Infl_N, one obtains the following:

1. (59)

   

As a bound morpheme, the masculine article in 59a still requires a host, but its free neuter counterpart in 59b does not. More generally, I propose that all unrelated articles in Table 1 (that is, Dutch het, Frisian it, West Jutlandic æ) are free morphemes with the same representation as in 59b.

To summarize, in order to capture the synchronic relatedness of determiner stem forms (hypothesis 1a), I proposed that determiners involve bound or free morphemes. As I show more clearly below, bound morphemes lead to related stem forms; free morphemes appear as unrelated determiners. To finalize the above derivations, two more steps are required: Bound morphemes must be supported by the definiteness marker and a vowel must be inserted in front of a consonantal inflection. This is what I turn to in the next part of this subsection.

Above, I argued that the definiteness marker has a different status. This element is semantically vacuous. I propose in more detail now that it is not an integrated part of definite determiners. Rather, the definiteness marker is inserted by a different mechanism. Similar to Last Resort operations such as do-support in English, this mechanism inserts the definiteness marker. I refer to this operation as D-SUPPORT. With d-support occurring postsyntactically, one can maintain that the demonstrative involves two and the article one syntactic head. Besides requiring minimal structural assumptions, this proposal also allows one to account for the related determiner forms in Germanic synchronically and diachronically.Footnote ²⁶

In more detail, I provided four pieces of evidence in section 3.2 that definiteness does not entail the presence of the definiteness marker. The same basic point can be made in a different way for the Scandinavian languages. Here the definiteness marker is not present under certain morphosyntactic conditions. Exemplifying with Danish, the unmodified noun has a suffixal article, but the modified one has a free-standing article. Compare 60a to 60b. Crucially, while both DPs are definite, only the noun phrase in 60b exhibits the definiteness marker:Footnote ²⁷

1. (60)

   

This observation adds to the data that suggest a special status of the definiteness marker.

To account for Swedish data similar to 60, Santelmann (Reference Santelmann1993) proposes that under certain conditions, den in 60b is inserted in D to support the strong definiteness feature.Footnote ²⁸ Pointing out some parallels to do-support in English, she interprets this as a Last Resort operation. She labels the operation den-support (see also Heck et al. 2008:229). Discussing parallel data, Embick & Noyer (2001:581–583), Roehrs (Reference Roehrs2006), and Schoorlemmer (2009:19, 28) propose that it is not the entire den that is inserted but only the dummy part d, which supports a stranded affix. To reflect this difference, this operation is called d-support.

I follow the latter authors in claiming that d is inserted to support a bound morpheme. However, I show below that the inserted material can vary in size: It can consist of d only or of a combination of d and a vowel. I assume that both of these elements have no semantics. For simplicity, I continue using the name d-support. Now, if d is inserted late, then this implies that it does not project its own syntactic head. In other words, the definiteness marker is not structurally integrated with determiners but only linearly attached. While the definiteness marker has no semantics of its own, it is inserted in a featural context that includes a specification for definiteness (or perhaps 3rd person, as Bernstein Reference Bernstein, Høeg Müller and Klinge2008 claims for English). In what follows, I propose to extend d-support to the other Germanic languages. For this idea to be acceptable, I need to address some forms that do not seem straightforward. This is particularly relevant for German, Yiddish, and Pennsylvania German, all of which have regular inflections on their determiners.

Starting with German, I juxtapose the definite article, the three demonstratives, and an inflected adjective. Consider 61. Starting with the latter elements, the vocalic part of the inflection on the proximal demonstrative dieser, the distal demonstrative jener and the adjective guter ‘good' is schwa.Footnote ²⁹ In comparison, the definite article der and its demonstrative counterpart dér exhibit a number of irregular forms. These differences are particularly clear with the stressed and unreduced demonstratives. Specifically, the forms in 61a have an /e/-type sound. Depending on the following consonantal element, this vowel receives a slightly different pronunciation (for details, see Wiese Reference Wiese1988:34). More clearly and thus importantly, the feminine/plural nominative/accusative forms are special in that they have no independent inflection, as shown in 61b, and the neuter nominative/accusative instances show a different vowel following the dental element, as shown in 61c. Finally, notice that the vowel immediately following the definiteness marker of the proximal demonstrative dies- is uniformly pronounced as [i:].

1. (61)

   

In order to explain the different vocalic components in the definite article and simple demonstrative versus the remaining elements, I need to be more precise about the actual process of d-support. I follow the basic system laid out in Roehrs Reference Roehrs2009:132. In particular, I suggest that German does not simply have support by d; rather, it inserts the dental element along with a certain vowel depending on the featural context. The following different forms of the definiteness marker are ordered according to the specificity of their insertion context:Footnote ³⁰

1. (62)

   

To be clear, the vowel following d is not an epenthetic vowel but part of the support element itself. Furthermore, the distribution of the forms in 62 is not the result of a morphosyntactic agreement relation between elements in the syntax. Rather, the insertion of these forms is sensitive to syntactic and semantic features.Footnote ³¹ This is consistent with the above claim that the definiteness marker does not involve definiteness itself. To sum up this far, the definiteness marker is postsyntactically inserted in certain featural contexts. D-support is a Last Resort operation that provides an overt host for inflectional suffixes making the [DEF] feature visible.

The forms of the article are straightforward. After vocabulary insertion, the inflection under Infl_N is supported by the relevant d-vowel combination, as shown in 63a–c, with the minor adjustment that the schwa-inflection in the feminine/plural receives no independent realization, as shown in 63b. The related simple demonstratives are exemplified by the masculine form in 63d.

1. (63)

   

Now I turn to the complex demonstrative. After vocabulary insertion and d-support, there are two more steps. First, I assume that the vowel before the consonantal inflection in 64a,c is due to schwa-epenthesis, similar to ordinary adjectives such as guter ‘good'. Second, recall that the stem vowel of the proximal demonstrative stem -ies- is uniformly pronounced as [i:]. This contrasts with some of the vowels in the support elements. In order to avoid a hiatus, one vowel is deleted.Footnote ³² Assuming that semantic information cannot be lost in the course of the derivation, the proper deletion of the vowel follows from the assumption that the vowel on the (dummy) definiteness marker in 62 is not associated with semantics, but the one in the deictic component -ies- is. In other words, only the vowel of the support element can be deleted. This yields the correct final forms:

1. (64)

   

Finally, the distal demonstrative jener involves a free stem morpheme and does not undergo d-support. The discussion can be extended to Yiddish and Pennsylvania German. I show that these languages have slightly different types of d-support.Footnote ³³

As in German, the masculine determiner forms in Yiddish and Pennsylvania German are regular. Putting aside differences in spelling, the feminine/plural definite articles and simple demonstratives in these languages are also the same as in German. Consider 65a,b and 66a,b. However, there are also some differences. Recall that the regularity of inflection is determined by comparing the endings on the determiners to the strong endings on the adjectives. Taking the inflections on the adjectives into account, Yiddish patterns with German, as shown in 65b, but Pennsylvania German has the ending -i on the distal demonstrative sell- and on the feminine adjective as well, as shown in 66b.

1. (65)

   

2. (66)

   

Interestingly, there is a second feature according to which the three languages fall into two groups. Whereas German and Yiddish have a third form of the definite article and simple demonstrative in the neuter, as in 65c, their counterparts in Pennsylvania German exhibit the same type of vowel as in the masculine nominative, as in 66c.Footnote ³⁴

I propose that like German, Yiddish has three support forms. However, there is one minor difference between these two languages: While the distribution of the endings in the masculine and neuter in Yiddish is essentially like the one in German, Yiddish di is inserted not only in the nominative/accusative feminine/plural but also in the dative plural. This means that the conditions for d-support in Yiddish are different from those in German. In this regard, note also that the strong ending in the dative plural in Yiddish is -e (and not -en as in German). This makes the ending -e the same as in the nominative/accusative feminine/plural. In other words, di and strong -e have the same distribution in Yiddish.

I tentatively suggest that the insertion conditions in Yiddish can be stated in terms of the following inflection (and not in terms of features as in German). Consider 67. Note that this is consistent with the assumption that insertion of vocabulary—including inflections—precedes d-support. Also, as in German, when di is inserted before -e, the determiner is pronounced as [di(:)]. In contrast, Pennsylvania German has a simpler support inventory. Putting the ending -e on the plural adjective aside for a moment, I propose that -i is a regular inflection in the feminine and plural. It can be stated then that bound morphemes are uniformly supported by a sole d in the latter language:

1. (67)

   

It can be seen now that the three languages with regular inflections on their determiners have slightly different rules for d-support.

This concludes the discussion of the derivations begun earlier in this subsection. Specifically, bound morphemes supported by the definiteness marker account for the relatedness of the determiner stem forms. This accounts for the related forms in hypothesis 1a. The unrelated forms were accounted for by the insertion of free morphemes. In the next part of this subsection, I take up the discussion of the few isolated exceptions left open above.

Languages with regular inflections provide clear evidence for the presence of features for gender and number and, in some instances, also case. As pointed out above, there is a small number of unsystematic exceptions. What I mean by that is that given a regular inflectional system, these exceptions involve unexpected forms that surface in isolated, specific contexts. In other words, they appear only in certain combinations of gender, number, and case. As such, they are few in number and do not appear throughout the entire inflectional paradigm. Nonetheless, I believe they deserve attention because they reveal some interesting properties, especially in the neuter gender.

I assume that structures are assembled in the syntax, with abstract features on their terminal nodes. Adopting the general framework of Distributed Morphology (Halle & Marantz Reference Halle, Marantz, Hale and Keyser1993, Reference Halle and Marantz1994), these terminal nodes can be manipulated after syntax by certain operations (for example, Impoverishment and Fusion). Vocabulary items are inserted under the altered terminal nodes. I suggest that the isolated exceptions are due to the mechanism Impoverishment.

In more detail, recall that vocabulary items have feature specifications for their insertion under the terminal nodes. Now, Impoverishment rules remove certain features from these terminal nodes in specific featural contexts. With these features deleted, the expected vocabulary item cannot be inserted allowing a less specified vocabulary item to appear (Halle Reference Halle1997). Depending on how many features are deleted, a vocabulary item with fewer feature specifications is inserted, or no vocabulary item at all. Both scenarios are important for the discussion in this subsection. Relevant for current purposes, I assume that all inflectional suffixes have at least one feature that determines in what lexical context they are inserted. Among other consequences, this prevents adjectival inflection from surfacing on nouns (for simplicity's sake, though, these features are not provided). Before I turn to the exceptional cases, I illustrate the basic workings of Impoverishment.

In section 3.1, I briefly discussed the strong/weak alternation of adjective inflections in German. The basic principle is that adjectives not preceded by a definite determiner show a strong ending, as in 68a, but adjectives that are preceded by a definite determiner show a weak ending, as in 68b.

1. (68)

   

Fleshing out the ideas in Roehrs Reference Roehrs2009 originally inspired by Sauerland Reference Sauerland1996 and Schlenker Reference Schlenker1999, Roehrs & Julien (2012) propose that the weak endings are the result of an Impoverishment rule that deletes a certain feature. Without going into too much detail, this feature is part of the case system and is labeled [STRUCTURAL]. Now, strong inflections are more diverse and their vocabulary items are more specified. Importantly, they all involve the feature [STRUCTURAL]. In contrast, weak endings are less diverse and less specified. None of them have the [STRUCTURAL] feature. Roehrs & Julien (2012) formulate the relevant Impoverishment rule in such a way that it removes the feature [STRUCTURAL] from the Infl-head within the AdjP when the adjective is in the context of a determiner in a regular DP structure. As a result, the strong ending cannot be inserted, but the less-specified weak ending is allowed to surface. This accounts for the strong/weak alternation in 68. With this brief illustration of Impoverishment in place, I turn to some exceptional instances in German, Pennsylvania German, and Yiddish.

As stated above, the inflections on definite determiners and strong adjectives are identical in German. There is only one well-known exception to this. In the masculine/neuter genitive, an adjective without a definite determiner exhibits a weak (rather than an expected strong) ending. Compare 69a to 69b.

1. (69)

   

In line with the discussion of the strong/weak alternation above, I assume that an Impoverishment rule removes the feature [STRUCTURAL]. However, this particular Impoverishment rule is different from the one discussed above in that it applies regardless of the presence of a definite determiner and only in masculine/neuter genitive contexts. As a result, the strong ending cannot be inserted and only the weak ending can surface.

The second exception involves adjectival endings in Pennsylvania German. Recall that the ending on the plural adjective in Pennsylvania German is -e (rather than the expected -i). Compare 70a to 70b.

1. (70)

   

To account for this isolated exception, observe that in this language, all adjectives in the plural, be they strong, weak, or mixed, have -e (van Ness Reference Ness1994:429). Again, I suggest that due to an Impoverishment rule, this inflection is not spelled out as -i but as -e. Specifically, this rule removes a certain feature from adjective structures in the plural. As a consequence, the insertion of the more specified strong ending is blocked and only the less specified weak ending is allowed to surface.Footnote ³⁵

The third exception involves the neuter gender that exhibits some interesting language-specific peculiarities. In Pennsylvania German, the article es is different in that it does not have d, and the distal demonstrative sell is different in that it does not have the inflection -s, as shown in 71a.Footnote ³⁶ Turning to adjectives, note that in Pennsylvania German, the inflection on the adjective is optional, as shown in 71a. In contrast, in Yiddish, the adjective does not have an inflection when the noun follows, as in 71b, but the inflection reemerges when no noun occurs after the adjective, as in 71c.Footnote ³⁷

1. (71)

   

While these forms are exceptional from a certain point of view, note that they only occur in very restricted, isolated contexts. In fact, the “missing” elements always surface on related forms. Specifically, the proximal demonstrative des in Pennsylvania German has both d- and -s, and the adjectival inflection may also appear optionally in Pennsylvania German and in Yiddish when the adjective follows the noun. I hypothesize that the relevant abstract features are always present in the syntax and that these context-specific particularities only hold on the surface. In other words, I suggest that these exceptions are not part of the underlying linguistic system but rather a reflex of a subsequent operation.

Here I distinguish between two types of Impoverishment. As discussed in the first couple of cases, one type deletes a certain feature on the terminal node allowing a less specified lexical item to surface. This explains the insertion of a weak ending as opposed to an expected strong ending. The second type, which I refer to as RADICAL IMPOVERISHMENT, deletes all features, including those specifying the lexical context of insertion (perhaps it deletes even the entire terminal node Infl). As a result, no inflection is inserted at all, which accounts for the isolated instances of inflectionless forms.

I propose that the neuter cases in 71 involve radical Impoverishment removing all the relevant features in certain contexts. In particular, in Pennsylvania German, an Impoverishment rule optionally removes these abstract features of the inflection from the adjective and obligatorily—from the distal demonstrative.Footnote ³⁸ In Yiddish, these features are removed from the adjective when a noun follows. With all the features deleted, neither the strong nor the weak inflections are inserted bringing about the uninflected surface forms above. As for the article in Pennsylvania German, I proposed above that similarly to Dutch het, es is a free morpheme inserted under Infl_N. If this is accepted, then one can continue to claim that Pennsylvania German and Yiddish are similar to German in the relevant respects.Footnote ³⁹

The question arises why neuter is special in these languages allowing radical Impoverishment to occur. Steinmetz (Reference Steinmetz, Rauch and Carr2001) argues that the West Germanic languages have changed from gender systems where the default was neuter to systems where the default is masculine. This is particularly clear in German and Yiddish, where the number of neuter nouns has decreased over time (I assume the same is true for Pennsyl-vania German). I interpret this change in default gender as an indication that neuter has become the marked gender in these languages. In this regard, Nevins (Reference Nevins2011) proposes that marked features are the target of Impoverishment, whose purpose is to reduce markedness. If his discussion can be extended to the neuter gender in West Germanic, one has begun to answer the question why neuter exhibits so many uninflected forms.

More generally, notice that Impoverishment manipulates abstract features on Infl. This leads to the spellout of unexpected inflectional suffixes or none at all. Importantly, the Infl-head belongs to both closed-class items such as demonstratives and open-class elements such as adjectives. It seems unlikely that the inflectional patterns of the adjectives, including the isolated exceptions, are memorized on a case-by-case basis. Having argued that demonstratives and adjectives have the same basic inflectional and distributional patterns, I submit again that demonstrative forms are not memorized as unanalyzed chunks either. Next, I turn to unrelated demonstrative stems with irregular inflections.

3.4. Hypothesis 1b: Irregular Inflections on Demonstratives

Recall from above that irregular forms of the definite article are accounted for based on the assumption that free morphemes are inserted under Infl_N. With articles involving one head in the syntax, the insertion of a sole free morpheme accounts for both unrelated stem forms and irregular endings. Unlike articles, demonstratives consist of two heads, Infl and Deic. Unrelated demonstrative stem forms are accounted for by the insertion of free morphemes under Deic. As for the irregular endings on demonstratives, more needs to be said.

English has related determiner stem forms, as shown in 72. As discussed above, I assume that this language has bound morphemes for definite determiners and the dental element is inserted late. However, English is interesting in two other ways: On the one hand, it is well known that adjectives in this language have no inflections at all; on the other hand, the endings on the demonstratives cannot easily be parsed into an inflectional (for example, number) component and a deictic component.Footnote ⁴⁰

1. (72)

   

One can observe that these endings are irregular in that they combine inflectional and deictic features into one element. Factoring out d-support, note now that the proposed abstract two-headed demonstrative structure seems to have one head too many. The same basic point can be made for Icelandic.

Recall that Icelandic is a mixed language. It has one set of determiners that are related and have regular inflections: (h)-inn and hinn. In addition, Icelandic also has demonstratives with unrelated stem forms and irregular endings: þessi and sá. These forms are given below again, in the nominative singular:

1. (73)

   

To be clear, similarly to English, Icelandic also has some opaque forms that combine inflection and deixis into one element. One has arrived then at two issues. On the one hand, one has irregular forms; on the other hand, one seems to have one head too many in the demonstrative structure, and one may wonder where exactly these demonstratives are located in the structure. Specifically, one may wonder if these elements are inserted under Deic, Infl, or somewhere else. To account for these types of demonstratives, I employ a second operation from Distributed Morphology.

In order to motivate the next part of the proposal, consider an intriguing interaction between certain prepositions and definite articles in French. As in many other languages, prepositions select noun phrases with articles, as shown in 74a. Interestingly, when the prepositions à ‘to' and de ‘from' combine with the definite articles in the masculine (le) and plural (les), they yield special forms. For instance, the combination of à and le surfaces as au. Consider all the cases in 74b,c.

1. (74)

   

Comparing 74a to 74b,c, one can observe that there is a mismatch between the syntax and the morphology in the latter set of examples. Specifically, there are two syntactic heads in the structure (that is, P, Infl_N), but there is only one overt vocabulary item (for example, au ‘to the'). Elements with this type of multiple exponence are often called portmanteau morphemes (Spencer Reference Spencer1991:50).

To account for these forms in the current framework, one can employ Fusion (Halle & Marantz Reference Halle, Marantz, Hale and Keyser1993, Reference Halle and Marantz1994). Fusion is a postsyntactic operation that targets two terminal nodes standing in a sisterhood relation and “welds” these two heads together into one head. As a result, the number of nodes is reduced. This operation is followed by vocabulary insertion. To illustrate, I use au in 74b and make the standard assumption that prepositions take a DP (here: Infl_NP) as a complement. First, to bring about the sisterhood relation, I assume here for concreteness that the masculine Infl_N moves to P, as in 75a. Second, Fusion applies to P and Infl_N, as in 75b. This is followed by vocabulary insertion.

1. (75)

   

To be clear, the combination of Fusion and late vocabulary insertion allows one to explain syntax-morphology mismatches where more terminal nodes in the syntactic output are realized by fewer overt vocabulary items in PF. In other words, Fusion “masks” isomorphic correspondences between abstract syntactic structures and overt phonetic realizations. With this in place, I turn to the demonstratives, elements that also involve Infl.

Recall that these irregular demonstrative forms combine features for inflection and deixis into one unparsable element. I propose that after head movement of Deic, Infl and Deic undergo Fusion. Consider 76. Once these two heads are fused into one terminal head, a sole vocabulary item is inserted. Now, in order to account for the nontransparent demonstratives, I propose that partially opaque forms such as English that involve bound morphemes (also Kayne & Pollock 2010) and completely opaque forms such as Icelandic sá involve free morphemes.

1. (76)

   

Continuing with English, I suggest that the feature specifications for insertion of some of the demonstratives are a blend of inflectional features (that is, [PL]) and deictic features (for example, [PROX]). Given my assumptions about vocabulary insertion, the other items can remain underspecified for number:

1. (77)

   

To bring about the final forms, the bound morphemes in English have to be supported (for the discussion of the article in English, see below). As for Icelandic, I assume, for the sake of simplicity, that each feature combination of the þessi-type and the sá-type demonstratives is associated with a vocabulary item. Inserting free-standing morphemes, Icelandic does not require support for these forms. To be clear, Fusion in conjunction with late vocabulary insertion of bound and free morphemes explains the irregular endings in English and the suppletive forms in Icelandic.

Putting Afrikaans aside for a moment, the remaining languages of this type (Dutch, Frisian, the Mainland Scandinavian languages including West Jutlandic) work like English with a minor qualification. Unlike English, these languages have inflections on their adjectives. In section 3.3, I pointed out that the featural specifications of vocabulary items must contain information about the lexical context of their insertion. This prevents adjectival inflections from occurring on nouns. This is an important point since not all vocabulary items in the current proposal have complete feature specifications (that is, they may be under-specified). Now, the specification of a lexical context for insertion also explains this minor difference where the adjectival inflections in Dutch, Frisian, and the Mainland Scandinavian languages only occur on adjectives but not on the demonstratives and definite articles (the latter involving an unmanipulated Infl-head). This situates these languages between the German-type languages and English.

More generally, note that Fusion explains both the irregular forms of the demonstratives, which is part of hypothesis 1b, and the apparent superfluity of a second head in the proposed demonstrative structure. In other words, one can maintain that all Germanic languages start out with a two-headed demonstrative structure. Languages vary as to whether or not they have Fusion in their grammar that “masks” the underlying structure (in section 5.1, I formulate a diagnostic for Fusion). For clarity, I briefly delineate some differences between Impoverishment and Fusion.

Both Impoverishment and Fusion are postsyntactic operations that manipulate Infl. While Impoverishment deletes certain features, Fusion welds together Infl and Deic. Furthermore, Impoverishment works in a feature-specific manner, but Fusion applies across the entire paradigm of certain demonstratives. Recalling that as a mixed language, Icelandic also has regular determiners, I assume that Fusion does not apply in a language-general fashion but only throughout certain paradigms.Footnote ⁴¹ In the last part of this subsection, I discuss Afrikaans, which has nonsyntactic inflections on adjectives, and I complete the discussion of the determiner system in English, which has no inflections on adjectives at all.

Compared to the other Germanic languages, Afrikaans has the simplest determiner system. It has just one basic pair:

1. (78)

   

Ponelis (Reference Ponelis1993:168) and Donaldson (Reference Donaldson1994:491) state that dié can be both proximal and distal. In other words, the deixis of the demonstrative has no inherent specification. It seems that the value for deixis depends on other factors; for instance, adverbial elements can be added that specify its deictic value. Below, I illustrate the deictic component of the demonstrative as unspecified [DEIC]. Furthermore, according to Donaldson (Reference Donaldson1994:486–488), adjectives in Afrikaans may exhibit the inflection -e. This inflection depends on certain phonological properties of the adjective stem and, partially, on the semantics of the noun phrase.

Given these properties of the inflection, I assume that -e is not syntactically determined; that is, the inflection does not take part in concord of phi-features within the noun phrase (also Corver & van Koppen 2011). However, given the very existence of inflections on adjectives (albeit of a different nature), I assume that the determiners have an inflection too. This determiner ending is uniformly -ie. I assume that, similarly to Dutch, Frisian, and the Mainland Scandinavian languages, the inflections on the determiners and adjectives in Afrikaans have separate featural specifications for their insertion. Unlike these languages, I assume that Afrikaans does not allow Fusion of Infl and Deic. With this in place, I suggest the following featural specifications for the determiners, where 79a applies to the Deic-head and 79b is inserted under Infl of Deic or N.

1. (79)

   

The inflection -ie is supported by the definiteness marker. Recall that the abstract deictic feature attracts stress.

Returning to the discussion of English, this language has five definite determiner forms, each with a different ending. Recall 72 above. Further-more, English is the only language in the data where adjectives have no inflections at all. While one might claim that English has no features for grammatical gender, it has features for number and definiteness. I assume that at least number is present in fused Infl+Deic and Infl_N; [DEF] has moved to adjoin to Infl_N. Since English has no inflectional endings, I assume that vocabulary insertion does not merge any element under Infl_N, the position of the definite article. In other words, Infl_N does not contain a bound morpheme.Footnote ⁴²

To explain the presence of the overt article, I propose that the entire form the is inserted under Infl_N to make [DEF] visible (see section 3.2). As to the demonstratives, I suggest that similarly to d-support in German (see 64), -e on the definiteness marker the is later deleted; that is, cases such as the-is surface as this without the loss of any deictic information. This implies that d-support in English has a slightly different status: The definiteness marker only supports the bound morpheme under Infl+Deic, but it provides no support for the suffixless Infl_N.Footnote ⁴³

3.5. Interim Summary

This paper studies the definiteness marker and the inflection on definite determiners. The aim is to figure out the inner makeup of these determiners and to identify some consequences for the underlying syntax of the noun phrase as a whole. I began this section by formulating the hypothesis that all Germanic languages have related definite articles and demonstratives. In addition, some languages may have a few unrelated determiner forms. With regard to inflection, I have sorted the Germanic languages into three groups. Taking adjectives with strong inflections as the point of reference, some languages have determiners with regular inflections, some languages have determiners with irregular endings, and some languages have both. I have established that the two dimensions of variation—that is, stem form and inflection—may occur separately or together. This interaction gives rise to four possibilities represented in Table 2 for articles and in Table 3 for demonstratives.

Table 2.

Cross-classification of articles.

Table 3.

Cross-classification of demonstratives.

As all combinations of stem forms and inflections are possible, the different operations of the account must be able to occur independently but they must also be compatible with each other at the same time.

I have provided evidence that the definiteness marker is special. It is semantically vacuous and cannot be directly tied to definiteness. I have proposed that the definiteness marker is not part of the syntactic structure per se but rather it supports inflectional suffixes making the feature [DEF] visible at the same time. Languages vary as to the details of the support mechanism. I have proposed that definite articles involve one head in the syntax (Infl) but demonstratives—two (Infl, Deic). The latter type of determiner has been shown to have distributional properties similar to those of adjectives.

I have proposed that all these elements, including articles, involve extended projections closed off by the inflection. The difference is that demonstratives make up the entire extended projection whereas articles form only one part of it, with the head noun forming the other part (Giusti Reference Giusti and Haegeman1997:100). As a general and, I believe, attractive consequence, all these elements receive a uniform structure explaining their commonalities in a principled way. Furthermore, all the determiners in Table 2 and 3 follow from the well-established distinction between bound and free morphemes, where the former have to be supported by the definiteness marker, and the operation Fusion.

Specifically, the Pennsylvania German article der is analyzed as a bound morpheme that is supported by the definiteness marker. English the functions as the definiteness marker. Pennsylvania German es and West Jutlandic æ involve free morphemes. As to the demonstratives, they involve two heads. Like the articles, they also involve bound and free morphemes. Unlike the articles, they may undergo Fusion, a post-syntactic operation that welds together Infl and Deic. In more detail, German dieser involves a bound morpheme supported by the definiteness marker. English that is basically like German dieser but, in addition, involves Fusion. Pennsylvania German seller involves a free morpheme. Icelandic sá is similar to Pennsylvania German seller but also involves Fusion. In other words, the distinction between bound and free morphemes is compatible with the presence or absence of Fusion in the grammar. Finally, a number of isolated, specific exceptions were accounted for by Impoverishment rules that manipulate regular underlying features thereby allowing exceptional forms to appear.

The flow chart in 80 summarizes the basic model of grammar adopted here. I illustrate the derivational progression with two examples: English that, which has a related stem and an irregular ending, and Pennsylvania German sell, which has an unrelated stem and no inflection at all. Given that two heads project in the syntax, I have proposed that English that involves Fusion of Infl and Deic. This results in one node where the bound morpheme -at is inserted. Bound morphemes in English are supported by the definiteness marker the. The correct final form is obtained by deleting the vowel -e of the vacuous definiteness marker. As to Pennsylvania German sell, this demonstrative form only occurs in the neuter. Rather than appealing to Fusion, I have proposed that Impoverishment deletes all features in this featural context (this context-specific application is indicated in 80 by square brackets).Footnote ⁴⁴ Consequently, no inflectional morpheme is inserted and Infl remains empty. The unrelated stem form is accounted for by the insertion of a free morpheme.

1. (80)

   

Finally, I briefly comment on the Romance languages. All these languages have unrelated determiner forms and, except Sardinian, they all have irregular endings. I propose that all demonstrative structures in the Romance languages undergo Fusion, following which free morphemes are inserted. This explains the typological difference between Germanic and Romance.Footnote ⁴⁵

4.1. Adverbial Elements inside Demonstrative Structures

Demonstratives can occur with adverbial elements (for example, Brugè Reference Brugè1996, Bernstein Reference Bernstein1997, Roehrs Reference Roehrs2010). These adverbial elements are often referred to as reinforcers. A simple example is this here house in non-standard English. Relevant for current purposes, Brugè (Reference Brugè1996) and Roehrs (Reference Roehrs2010) argue in detail that these adverbials are part of the demonstrative structure. The question arises as to where exactly the adverbial element is located in this structure.

The demonstrative structure proposed above involves two heads and three phrasal positions: two specifiers and one complement position. If the lower head of the demonstrative structure moves up, then there is one specifier position above the complex head structure and two phrasal positions below it. While I know of no Germanic language that fills all three positions at the same time, Yiddish and Eastern Norwegian show, when examined together, that all three phrasal positions are necessary parts of the demonstrative structure.

Although this is not a very frequent pattern, in some languages adverbial elements can precede the demonstrative. As discussed in section 2.2, Afrikaans and Yiddish are such languages (for other languages, see Roehrs Reference Roehrs2010). Focusing on Yiddish, ot ‘here' appears in the left periphery of the noun phrase, as shown in 82a. In line with Roehrs Reference Roehrs2010, I propose that this element surfaces in the highest specifier position of the demonstrative structure. In current terms, this position is labeled Spec, InflP in 82b.

1. (82)

   

The inflectional suffix in 82a is supported by the definiteness marker. Also, assuming the structure of the noun phrase in section 3.2, 82b itself is located in the specifier position of Infl_N.

Certain dialects in the Scandinavian languages may exhibit two adverbial elements immediately to the right of the demonstrative, as shown in 83a. Pace Leu Reference Leu2007, Reference Leu2008, I propose that one is in Spec,DeicP and the other is in the complement position of Deic, as in 83b.

1. (83)

   

D-support brings about the final form. Recall also that Infl and Deic in English-type languages undergo Fusion. Comparing 82b and 83b, note that adverbial elements can be accommodated in the specifier positions independent of whether or not the complex head structure undergoes postsyntactic Fusion. In other words, Fusion does not affect the syntax of the demonstrative structures, and adverbial elements are possible in English-type languages.Footnote ⁴⁶

To sum up, Yiddish and Eastern Norwegian show that demonstrative structures have three phrasal positions. The two-headed structure is fully compatible with the distribution of these adverbial elements. In fact, as far as I am aware, only one adverbial element can precede the demonstrative, but two may follow. I take this as evidence that the Deic-head underwent syntactic movement, which is upwards.

The discussion of these adverbial elements allows one to rule out alternative derivations that might combine the two heads in 82b by different means. For instance, Embick & Noyer (2001) suggest that two heads can undergo Morphological Merger. This postsyntactic operation has two subtypes: Lowering and Local Dislocation. For current purposes, I focus on Lowering, which allows displacement across adverbial elements. On this analysis, the Infl-head in 82b would attach onto the Deic-head. In other words, this operation involves downward displacement. Now, if this option were correct, then two adverbial elements should be able to occur in front of the demonstrative but only one after it, contrary to fact (recall that the -o on Yiddish ot-o has a special status, see note 8).Footnote ⁴⁷

As I show below, the analysis developed so far also provides a straightforward account for the development of complex demonstratives that contain s, such as German dieser ‘this'. To the best of my knowledge, this change has never been given a structural account before.

4.2. Development of Complex Demonstratives

I turn to an interesting diachronic change in Germanic that involves the old deictic particle si. In the last subsection, I pointed out that adverbial elements are in specifier positions. If one assumes that these elements can undergo Spec-Head reanalysis, then one can explain the diachronic development of complex demonstratives in Germanic. Specifically, employing a two-headed demonstrative structure allows one to tie the reanalysis of si to the change of the definiteness marker from a Deic-head to a support element. Besides that, I claim that this change had consequences for the entire system of definite determiners in Germanic.

Factoring in late d-insertion, I have analyzed simple demonstratives such as 84a as phrases with a segmentally unrealized Deic-head, as in 84b.

1. (84)

   

By hypothesis, OHG also involved two heads in its demonstrative structure. Unlike the contemporary language, OHG made two overt heads available; that is, I propose that d was inserted in Deic by vocabulary insertion (and not by d-support):

1. (85)

   

I propose below that over time, the deictic head thë was replaced by si forming the complex demonstrative dieser ‘this'.Footnote ⁴⁸ As a result, the simple demonstrative thër also changed such that Deic no longer received an overt realization. In other words, the definite determiner system was reorganized: A new demonstrative element came into being, as in 86a, the old demonstrative changed its segmental composition, as in 86b, and the definite article emerged, as in 86c.

1. (86)

   

This development coincided with the status of the definiteness marker changing from a deictic head in OHG to a dummy in Modern German. As discussed above, now this element by itself has no semantics and is inserted late in the derivation. I consider this in more detail below.

Simple demonstratives in the older Germanic languages could be reinforced by si.Footnote ⁴⁹ Haspelmath (Reference Haspelmath1993:282) glosses this element as ‘behold' and points out that there are three stages in the development of the complex demonstrative. First, si followed the inflection on the simple demonstrative, as in 87a. Second, there were hybrid forms with the deictic reinforcer sandwiched between two inflections, as in 87b. Third, the inflection could only follow si, as in 87c. Haspelmath refers to this process as the “externalization of inflection.” I gloss the definiteness marker as D and the deictic element as SI (data are taken from Haspelmath).Footnote ⁵⁰

1. (87)

   

According to Braune & Eggers Reference Braune and Eggers1987:246, most of the complex demonstrative forms in OHG have reached the third stage shown in 87c (masculine/neuter genitive forms are still at the second stage; see Paul et al. Reference Paul, Wiehl and Grosse1989:228).

With these data in mind, I propose that the deictic particle underwent some changes that eventually resulted in it becoming an integral part of the demonstrative. Two fundamental changes can be found in the new shape of the deictic element (that is, si > s > ies) and its new, internal position. The proposal of a two-headed demonstrative structure relates the change in the shape of this element to the change in its position. I sketch out a three-stage development to show how the shift took place.

I assume that at the beginning, the particle si behaved like other adverbial elements. I propose that it was merged into Spec,DeicP. The tree in 88a shows the underlying demonstrative structure, and the tree in 88b shows the structure after head movement of the definiteness marker.

1. (88)

   

With Deic undergoing head movement, si winds up in the right periphery of the demonstrative structure.

In the second stage, si underwent a phonological reduction that resulted in a clitic-like element. With clitic-like s following, I assume that the inflection on the demonstrative was no longer fully “active” morphosyntactically, and a second, external inflection could (had to?) now be realized in Infl_N. I return to this below. First, though, consider the internal structure of the demonstrative in detail:

1. (89)

   

As discussed above, the demonstrative structure as a whole surfaces in Spec, InflP_N. As to the second, external inflection, I propose that it is inserted under Infl_N:Footnote ⁵¹

1. (90) Intermediate Stage: Realization of the Inflection in Infl_N

   

Before I turn to the third stage, I consider two points in more detail. First, while pursuing a different account, Haspelmath (Reference Haspelmath1993:298) states that a second, pleonastic inflectional affix occurs when the original form is “irregular and unproductive;” that is, when it is no longer morpho-semantically transparent. If so, one expects to find hybrid forms where the inflections differ in shape. This is indeed the case with these demonstratives. For instance, older upper German dialects formed the simple demonstrative in the nominative/accusative plural neuter as dei. importantly, the reinforced demonstrative shows two different inflections (see Braune and Eggers Reference Braune and Eggers1987:247):

1. (91)

   

To explain the co-occurrence of two different inflections more technically, I tentatively propose that the (inner) irregular inflection undergoes only “partial” concord. There are at least two ways to think of this: All concord features are present in Infl of the demonstrative structure, but only some of them are licensed/valued; alternatively, only some of the concord features are present in Infl and all of them are licensed/valued. While I am not able to decide between these scenarios, either one involves inflections that are in some sense incomplete and defective. In contrast, I suggest that the second (outer) inflectional morpheme is regular in that it undergoes “full” concord where all features are present and licensed/valued. Now, if the inner inflection was indeed somehow incomplete and defective, it is not surprising that dei was later replaced and the reinforced demonstrative developed into (the expected) disiu.

Second, one must also schematize the development of si itself. To illustrate this developmental sequence, I show how si moved from the status of phrase via clitic to head. In 92a,b, the reinforcer has the status of a phrase and is merged in Spec,DeicP. Unlike the complex phrase in 92a, s in 92b is a clitic-like element. Finally, 92c shows a head that is merged in Deic.

1. (92)

   

Next, consider what happens once the clitic has changed into a head. In the third and final stage, the reinforcer undergoes Spec-Head reanalysis à la van Gelderen Reference Gelderen2004. In other words, the s in 92c is merged in the Deic-head of the two-headed structure:

1. (93) Spec-Head Reanalysis of the Reinforcer

   

As for the stem vowel of the demonstrative, the change from /ë/ to /i/ begins with forms such as disiu and spreads by analogy to the other demonstrative forms in the paradigm (for details, see Braune & Eggers Reference Braune and Eggers1987:247, Paul et al. Reference Paul, Wiehl and Grosse1989:228). Once this process was completed, the vowel, which formed part of the simple demonstrative, came to be associated with the deictic element resulting in ies. A direct consequence of this syntactic and morphophonological reanalysis is that the former deictic particle replaced the definiteness marker as the deictic head. In other words, si changed from an adverbial particle to -ies, which now forms an integral part of the proximal demonstrative:

1. (94) Third Stage: After Spec-Head Reanalysis of the Reinforcer

   

I suggest that this replacement of the definiteness marker by the former reinforcer changed the status of the definiteness marker from a deictic head to a support element:Footnote ⁵²

1. (95)

   

Comparing 95a to 95b, this reanalysis led to a change in the morpheme boundaries.Footnote ⁵³ More generally, note that the two-headed demonstrative structure ties the change of si to the change of the definiteness marker in a direct structural manner.

To recapitulate, OHG and Modern German demonstratives both involve two-headed structures. Semantically, they are definite; that is, there is a [DEF] feature present in the syntax. They differ, however, in that OHG makes this feature visible by vocabulary insertion, while Modern German makes it visible by d-support of the inflectional suffix. Given the assumption that these are late operations, one does not expect differences in the syntax. To the best of my knowledge, this is indeed the case. Finally, one may wonder why the demonstrative system was reorganized in the first place. Considering the three-stage developmental sequence sketched above, two important questions arise: (i) Why was si inserted at all? and (ii) Why was d replaced by s? I briefly discuss a traditional type of account and a more formal solution.

In the traditional literature, diachronic changes are often explained on the assumption that old and new elements are related through a chain reaction. Specifically, a “push” or a “pull” operation is invoked in these types of accounts. In the current context, one could claim that the reanalysis of -s as a head pushes d out of the Deic-head position, or that the reanalysis of d as a dummy pulls -s into Deic. Turning to the two questions above, this traditional account has no clear answer to the first question: The initial appearance of si seems to be accidental.

As to the second question, there is also no straightforward way to determine whether this is a push or a pull operation. More importantly, it is not clear why reanalysis has taken place at all. Furthermore, note that a pull operation could, in principle, remain incomplete. In other words, by staying a clitic, s does not necessarily have to move into Deic. This would result in the ungrammatical 96a, where in Modern German, a segmentally unrealized deitic head and a regular inflection would be followed by the reinforcing clitic -s, as in 96b.

1. (96)

   

While 96a may be grammatical in other types of languages, note that an incomplete pull operation does not tie the change that the definiteness marker underwent to the presence of the deictic particle in a dependence relation. To rule this option out for Modern German, it must be stipulated that s is pulled all the way into a head position.Footnote ⁵⁴ To be clear, neither question receives a satisfying answer under the traditional type of “push or pull” approach.

A more promising account was suggested to me by Elly van Gelderen (personal communication), and the proposals in van Gelderen Reference Gelderen2011 provide better answers to these questions. As part of formal economy considerations, an answer to the first question is found in Feature Economy, which claims that semantic and interpretable features are to be minimized in the derivation. More precisely, semantic features change to formal interpretable features [iF]. Interpretable features turn into uninter-pretable features [uF] and may eventually get lost. Crucially, when a feature has become uninterpretable, it becomes a probe in the sense of Chomsky Reference Chomsky, Martin, Micheals and Uriagereka2000 and requires a goal to be valued. Now, assuming that the deictic feature on þ in þaim ‘that' has become uninterpretable, si is inserted to value it, which results in þaimsi. This accounts for the presence of the deictic particle in the first stage above.

The second question is answered by van Gelderen's Head Preference Principle. According to this principle, it is more economical to insert a (structurally) smaller element—a head, than a (structurally) larger element—a phrase. This principle provides a plausible rationale for the frequently employed Spec-Head reanalysis. Now, if the uninterpretable feature on d is lost, then clitic s can “renew” the deictic system by merging as a head in Deic complying with the Head Preference Principle (eventually, this renewal of the featural makeup may result in an uninterpretable deictic feature again). This accounts for s replacing d in the third stage above (note that the second stage above has nothing to do with Deic but with concord on Infl in the demonstrative structure).Footnote ⁵⁵

To sum up, van Gelderen's proposals based on formal economy considerations provide better answers to the two questions above than the traditional type of account.Footnote ⁵⁶ With the definiteness marker losing its deictic specification, the old simple demonstrative changed its segmental composition and the definite article could emerge.

4.3. Simple Demonstratives and Definite Articles

As is well known, articles often derive from demonstratives (Diessel Reference Diessel1999:128, Lyons Reference Lyons1999:331, Harbert Reference Harbert2007:142; for Icelandic, see next subsection). Recall that the complex demonstrative was basically established in OHG; that is, most attested demonstrative forms are in stage 3 discussed above. Now, as Paul et al. (Reference Paul, Wiehl and Grosse1989:226, 383) point out, the definite article could still be missing in some contexts in Middle High German (MHG) where it is used today. It seems clear, then, that the emergence of the definite article took a long time and came into being after the complex demonstrative had been established.

In her discussion of the Definiteness Cycle, van Gelderen (2007, 2011) proposes that demonstratives in phrasal positions are reanalyzed as articles in head positions. She argues that this structural reanalysis coincides with a change in the featural makeup of these elements. As pointed out above, semantic features change to formal features, and the latter may eventually disappear (there is one proviso here, namely, the resultant definite article retains one uninterpretable feature requiring the presence of a noun). Van Gelderen does not discuss the inner structure of demonstratives in much detail. One interesting difference is that I have aligned her [LOC] (that is, deictic) feature with a deictic head. I believe though that her proposal and the current account are largely compatible.

As discussed above, at the end of the development of the proximal demonstrative, the definiteness marker was no longer inserted under Deic and became a dummy that supported bound morphemes. In other words, the deictic feature was no longer associated with the dental stem. I propose that as a consequence, the Deic-head could be realized by -ies or by no overt segment at all (represented by Ø). The former option derives the proximal demonstrative (see previous subsection); the latter option captures the reanalysis of the old simple demonstrative as its later counterpart. For the latter change, compare 97a to 97b. Furthermore, with a support element now available in the grammar, inflectional suffixes could successfully be inserted under Infl_N. This derives the emergence of the definite article in 97c.

1. (97)

   

In other words, the definiteness marker changed from the deictic head thë to the dummy d that supports bound morphemes. Taken together, these developments led to the reorganization of the definite determiner system in German (and other languages). Finally, I discuss the diachronic and synchronic relatedness of articles and demonstratives in North Germanic exemplified here by Icelandic.

4.4. Definite Articles and Simple Demonstratives in Icelandic

Modern Icelandic has an almost homophonous word pair consisting of the demonstrative hinn and the article (h)-inn. The only exception to this pattern is in the neuter nominative/accusative singular, where the demonstrative is hitt, but the article is (h)-iÐ. Given these (almost) identical forms, one might claim that similarly to the article in German, the article in Icelandic derived from the demonstrative. However, there are reasons to be doubtful about such a conjecture.Footnote ⁵⁷ Rather, it seems more likely that the demonstrative and the definite article are two different words with two different histories.Footnote ⁵⁸

Specifically, hinn (fem. hin, neut. hitt) developed from the PIE root ⁺k^jo + eno, but inn (fem. in, neut. it) developed from PIE ⁺eno (see de Vries Reference Vries1962). In early skaldic poetry, the two are strictly separated: Hinn always carries stress, but inn never does. In later Icelandic and Nor-wegian, the hinn forms often replace the inn forms. In such cases, the former were unstressed and the strongly stressed hitt was replaced by hit, either because of analogy with it, or because the -tt lost quantity in unstressed position (see Noreen Reference Noreen1970:§§471, 285). For instance, with an adjective present, one can find examples with or without h: hit fyrra sumar = it fyrra sumar ‘the last summer'. At the same time, one can find unreduced forms: hitt var fyrr ‘that was earlier'. In 14th-century Ice-landic, final -t changed to -Ð in weakly stressed syllables. Hence, it and hit became iÐ and hiÐ (see Noreen Reference Noreen1970:§248).

In more general terms, Icelandic started out with two separate forms. Later, these forms fell together with stress being the main distinguishing factor. It is not clear to me if stress always coincided with deixis. Be that as it may, differences in stress led to the separation of the determiners into today's articles and demonstratives, which is only visible in the nominative/accusative neuter. For the purposes of this paper, I assume that contemporary speakers have no access to the varying diachronic developments of these different determiners. As such, I treat these elements as morphosyntactically related in the current grammar of this language. I assume that the difference in inflection between Icelandic hitt and (h)-iÐ has a phonological explanation.Footnote ⁵⁹

To sum up this section, I have shown that adverbial elements are straightforwardly accommodated by the two-headed demonstrative structure. In fact, the distribution of these adverbial elements provides an argument that Deic undergoes head movement to Infl. Furthermore, I have proposed that the proximal demonstrative and the definite article both derived from the distal demonstrative, whereby the latter underwent a change as well. This development resulted in a new determiner system. The proposal that the deictic particle changed from a free to a bound morpheme (si > s > -ies) implies that the definiteness marker changed from a free deictic morpheme to a support element. These two changes are tied together in the two-headed demonstrative structure where the Deic-head is filled by different vocabulary items over time. Moreover, the change to bound and unrealized morphemes (-ies, -Ø) supported by the definiteness marker allows one to relate the definite determiners both diachronically and synchronically. Finally, I have briefly discussed the North Germanic languages. I have suggested that although their determiners seem to have followed separate paths of development at the beginning, they can be integrated into the current proposal.

5.1. Logically Conceivable Forms

Hypothesis 1 states that every Germanic language has at least one article and at least one demonstrative that share a definiteness marker (a dental sound or h-). To capture the semantic and phonological differences between articles and demonstratives, I have proposed that the latter contained a deictic element. This deictic element has been argued to involve an abstract head in the syntax that is filled by subsequent vocabulary insertion. Exemplifying with German, I have made the distinction between bound morphemes (-ies-) and free morphemes (jen-). In conjunction with d-support, the former accounted for the essence of hypothesis 1.

To identify gaps in the data, I propose to make the structural patterns of the definite determiners more abstract. I use the following notation: D- for a shared support element (for example, d-), (Ir-)RegInfl for (ir-) regular inflection, Ø for a segmentally present but overtly unrealized element, X for a putative third syntactic head in the demonstrative structure, Y for a bound morpheme (for instance, -ies-), and Z for a free morpheme (for example, jen-). Unattested but potentially possible combinations of an article and demonstrative are designated by ??; unattested and, by hypothesis, impossible combinations are marked by *.

Consider 99 and 100: The former shows the combinations of an article with regular inflection, and the latter contains the combinations of an article with irregular endings. For clarity, one concrete example is provided for each of the attested types of determiners. Abstracting away from unrelated article forms, one obtains the following logically conceivable options:Footnote ⁶⁰

1. (99)

   

2. (100)

   

Before discussing these combinations in some detail, I step back for a moment and recast the above proposal in more general terms.

If articles involve one head in the syntax and demonstratives involve two, one can formulate two diagnostics that help parse these (and other) determiners with regard to their inner makeup. Starting with definite articles, if these elements exhibit regular inflection, then one can infer that the other part, if present, must be a dummy. Since inflections are bound morphemes, this dummy must be a support element. In other words, to the extent that the one-headed structure of articles is correct, one has a means to parse the inner structure of articles. Note that articles with irregular endings can only be parsed when compared to the other definite determiners (for example, by ascertaining that other determiners share the definiteness marker).

Turning to demonstratives, these elements involve two heads in the syntax, a deictic one and an inflectional one. I argue in the next sub-section that the c-examples and g-examples in 99 and 100 above are systematic gaps. As for the remaining combinations in 99 and 100, the fully transparent forms are derived by inserting bound morphemes in the a- and b-examples, and free morphemes in the d-examples. The partially transparent forms in the e- and f-examples follow from the assumption that Fusion welds together the two syntactic heads. Fusion is marked by a plus sign in 99 and 100. The resultant head is filled by a bound morpheme supported later by the relevant definiteness marker. Finally, the opaque, nonparsable forms in the h-examples are derived by Fusion and the insertion of free morphemes.

What the e-, f- and h-examples have in common is the presence of irregular inflection that was derived by Fusion of the deictic and inflectional heads. If so, one can infer from irregular endings on demonstratives that Fusion is at play, which results in one head. If another component is present, that component must be a support element. To avoid an ad hoc claim, one should find evidence that this dummy is present elsewhere in the language (for example, on articles).

To sum up, given the proposed determiner structures, one has arrived at two diagnostics for parsing the inner makeup of definite determiners:Footnote ⁶⁵

1. (101) Diagnostics for Parsing Determiners

   1. (a) Regular inflection on articles -> Identification of support elements

   2. (b) Irregular endings on demonstratives -> Identification of Fusion -> Identification of support elements

Formulating these diagnostics in very general terms, I mean for them to be applicable beyond the languages discussed here (for some cross-linguistic information, see Diessel Reference Diessel1999, Lyons Reference Lyons1999:107, Rijkhoff Reference Rijkhoff2002: 178).

There are many interesting points to discuss with regard to the determiner systems delineated in 99 and 100. This is especially true for the potentially existing yet unattested systems marked by ??. For considerations of space, I do not investigate all these combinations here in detail. To discuss just one case briefly: Icelandic is interesting in that it has related stem forms with regular inflections, as shown in 102a,b. The unrelated stem forms have irregular endings, as shown in 102c.

1. (102)

   

Logically speaking, it is possible to conceive of a system that has related stems with irregular endings, as in 103a,b, and unrelated stems with regular inflections, as in 103c. I label this imaginary language Ice-landic2, and the forms are exemplified by a combination of English and German determiners.

1. (103)

   

Although I currently know of no such language, I believe Icelandic2 is possible. Note first that the determiners of Icelandic2 are structurally compatible with the current analysis. As for the concrete inner makeup, this language would be similar to English with two qualifications: Adjectives have inflections and the distal demonstrative that is replaced by a German-type demonstrative. To be clear, while this system does not exist in Germanic, nothing in the current proposal militates against the existence of this type of determiner system in principle. As such, I take this and the other potential determiner systems as accidental gaps in the data. More relevant for current purposes, I discuss systematic gaps in the next subsection.

5.2 Demonstratives with Three Syntactic Components

If demonstratives involve only two heads in the syntax and D- is the only element inserted late, then all combinations of fully transparent X-Y-RegInfl or partially transparent X-Y+IrregInfl are expected to be non-existent, provided there is indeed evidence that X stands for a syntactic head. The relevant cases are repeated here for convenience:

As far as I can tell, these are indeed systematic gaps. This is predicted by the current analysis. In simplified terms, the present proposal claims that X is too “many” in that this contentful element cannot be accommodated by the two-headed demonstrative structure. Note that the hypothesized parallelism between demonstratives and adjectives provides a natural motivation for this gap as both elements involve extended projections built from one contentful root element: the deictic part and the adjective stem. Below, I argue against a three-headed structure more directly.

Earlier work on this topic proposed a variety of analyses in terms of three-headed structures (for instance, Leu Reference Leu2008, Roehrs Reference Roehrs2010). In that vein, one could claim that German dieser ‘this' has three heads where the ending involves the head of an Inflectional Phrase (InflP), the deictic element instantiates the head of a Deictic Phrase (DeicP), and the definiteness marker is interpreted as the head of—what one might call—a Referential Phrase (RefP) (see Longobardi Reference Longobardi1994):

1. (104)

   

Combining these elements by head movement, this alternative is similar to the current proposal with the difference being that the definiteness marker is taken to indicate a syntactic head. Turning to some distal demonstratives, the case of German jener, Yiddish yener, and Pennsyl-vania German seller seem less straightforward at first glance. Parsing jener and the other two elements as in 105 does not seem revealing; that is, I am not aware of any evidence that j/y/s play a role as referential parts somewhere else in the determiner system, or that n/n/ll play a role as deictic parts somewhere else in the deictic system:Footnote ⁶⁶

1. (105)

   

Rather, jen-, yen-, and sell- seem to be unanalyzed, opaque stems. Taking the inflection into account, there seem to be two components. Proponents of an analysis in terms of a three-headed structure could propose that the Deic-head and the Ref-head in 105 undergo Fusion. As to the definite article, these scholars would then claim that German der ‘the' involves two heads (Infl, Ref) and West Jutlandic æ involves Fusion of these two heads. To be clear, on this alternative analysis, the definiteness marker instantiates a syntactic head (see Leu Reference Leu2008, Roehrs Reference Roehrs2010). As a consequence, definite articles involve two heads in the syntax, demonstratives involve three, and Fusion is extended to opaque article forms and to more demonstrative forms.

As far as I can see, most of the discussion in sections 3 and 4 can also be captured by this more elaborate structural analysis. However, there are also some differences between the two proposals: theory-internal, acquisitional, and empirical. Focusing the discussion on demonstratives, all of the following points speak unanimously in favor of the analysis of demonstratives as two-headed structures.

Starting with a theory-internal argument, recall that I have explained irregular endings on demonstratives by Fusion, which applies to two heads in a sisterhood relation. Proposing a three-headed structure, one faces the following problem: After head movement, only two heads (out of the three) are in a sisterhood relation under standard binary branching. In order to explain the irregular forms in Icelandic, one would have to make additional assumptions: Either one would have to give up the sisterhood relation, or one would have to postulate that Fusion can recur. The current proposal fares better in this regard.

In section 3, I showed that the definiteness marker is special and proposed that it is not part of the syntax. Proponents of a three-headed structure could accept this argumentation. However, this comes at a price. An analysis in terms of a three-headed structure would have to posit that simple demonstratives consist of two segmentally unrealized elements, dér ‘that': Ø-Ø-r. Recall that one of these segmentally unrealized heads was motivated by the observation that demonstratives—complex or simple—are stressed, with the abstract deictic feature, segmentally realized (-ies-) or not (-Ø-), attracting stress. To make a strong case for a second segmentally unrealized head, one should find some good motivating evidence. Until such evidence is found, the assumption of one segmentally unrealized head is, from the perspective of language acquisition, to be preferred over two.

Finally and more importantly, I turn to some empirical considerations. A three-headed structure involves four phrasal positions. Assuming that head movement proceeds all the way to the top, one might expect to find languages where demonstratives can occur with three adverbial elements to their immediate right. As far as I know, this is not the case (although this might be an accidental gap). A second, stronger empirical argument against the three-headed structure can be derived from certain nonexistent demonstrative forms. First, I briefly discuss Germanic and then focus on Romance in more detail.

If one were to suggest that German had a three-headed demonstrative structure after all, then one could expect forms such as X-Deic-Infl or d-Deic-X-Infl in German, where X stands for an element other than a dummy. Combinations not entirely implausible are given in 106a. Note that a quantifier can occur with other deictic elements, as shown in 106b.

1. (106)

   

However, the forms in 106a do not exist—either in German or, as far as I know, in any other Germanic language. An analysis in terms of a two-headed structure explains the absence of a third (contentful) component straightforwardly, as a third head is simply not present in the syntax. On the current analysis and abstracting away from reinforcers, only late insertion of a dummy can bring about a third element.

Another empirical argument can be derived from certain cross-linguistic considerations. Specifically, there is potential evidence from the Romance languages that demonstratives may involve three heads, and articles—two. To give this alternative a fair shot, I approach this issue in three different ways. First, I discuss Spanish, then I turn to Sardinian, and finally, I return to Spanish.

To begin, the claim about additional syntactic heads could be made on the basis of the feminine determiner forms in Spanish. At first glance, the demonstratives could be parsed as e-st-a ‘this', e-s-a ‘that', a-quell-a (or aque-ll-a, etc.) ‘that' and the article as l-a. As pointed out in section 3.1, the demonstratives and articles do not share the same initial elements; that is, they are not morphosyntactically related. These cases then present a first potential counterexample to the claim that demonstratives involve two heads, and articles—just one.

However, the masculine forms cannot be as straightforwardly accounted for in terms of these larger structures. The demonstratives in question are este, ese, aquel, and the article is el. Note first that not all these determiners share the same inflection. In fact, the typical gender marker for masculine does not occur on these determiners at all: Neither the demonstratives nor the article end in -o.Footnote ⁶⁷ In other words, gender marking on the determiners is, in the case of masculine, exceptional. Just as in the case discussed above, proponents of a three-headed structure could claim that the inflectional head has undergone Fusion with another head, thereby maintaining the larger structures. Put simply, the resulting structures could look as follows (Fusion of the two relevant heads is marked by curly brackets): e-{ste}, e-{se}, a-{quel}, and {el}. In conjunction with Fusion, the account in terms of three syntactic heads yields demonstratives with two stem forms: e- and a-.

Notice though that this analysis has to find a way to rule out the ungrammatical combinations where the fused heads co-occur with another stem form: *a-{ste}, *a-{se}, and *e-{quel} (similar problems arise under the alternative parse aque-ll-a, etc.). Note that one cannot claim that e- and a- are only compatible with certain deictic values: e- is part of the proximal and one of the distal demonstratives. These issues disappear if one abandons the larger structures and maintains that the underlying structure of demonstratives and articles in the Romance languages is the same as the structure of their Germanic counterparts, with the demonstratives having two heads, and the article—just one.Footnote ⁶⁸ Before I tackle the Spanish data from a different perspective, I briefly turn to Sardinian.

As far as I know, Sardinian is the only Romance language that has completely regular inflections on the definite determiners when compared to adjectives. The latter is exemplified in 107 by the word beautiful. Just as in German, in Sardinian these are fully transparent elements. Unlike their Germanic counterparts, the article and the demonstratives have different initial elements, s- versus ku-.Footnote ⁶⁹

1. (107)

   

To make this language fit into the current proposal one could claim that Sardinian is the only Romance language that has bound morphemes, which makes it similar to the Germanic languages. Unlike Germanic support elements, the support elements in Sardinian would be sensitive to their contexts of insertion, such that s- is inserted before the inflection in Infl_N and ku- is inserted before the deictic element. This may turn out to be an ad hoc solution. Note, though, that there are other ways to explain away this potential counterexample. Pending an indepth analysis, I leave the discussion of Sardinian and return to Spanish.

A reviewer asks what a language would have to look like so that it cannot be captured by the current system at all. In other words, the reviewer wonders if the main hypothesis about the syntactic structure is falsifiable. I believe the answer to this question is affirmative. Returning to the issue of an analysis in terms of three syntactic heads, Romance demonstratives are traditionally presented by reference to three grammatical persons essentially making demonstratives parallel to personal pronouns. Exemplifying with Spanish again, the tripartite system is organized as follows:

1. (108)

   

I propose to capture closeness to the speaker by [+SPEAKER], closeness to the addressee by [+ADDRESSEE], and remoteness by a negative value of these features. Employing these two binary features, one could suggest that each binary feature is located under one head. If so, one essentially splits the above Deic-head in two, which results in the following three-headed structure: Deic1-Deic2-Infl. There are two options: Either Deic1 could host the feature [±SPEAKER] and Deic2 could host [±ADDRESSEE], or Deic1 could involve [±ADDRESSEE] and Deic2—[±SPEAKER]. The analysis of the demonstratives can now be updated as follows:Footnote ⁷⁰

1. (109)

   

Note though that some exponents (for example, e-) are not consistently employed to make certain features visible, and conversely, some features (for example, [–ADDRESSEE]) are not consistently made visible by specific exponents. Recall also that the inflection -o on demonstratives has a special status and is not the result of an agreement relation with another element in the masculine gender. To be clear, then, Spanish cannot straightforwardly be analyzed in this way. This is consistent with the discussion above. However, such languages are easily conceivable.

To make a strong case for an approach that involves three heads, I discuss two constructed cases. These imaginary languages are labeled Spanish2 and Spanish3, and their demonstratives appear in capital letters:

1. (110)

   

Beginning with Spanish2, the first element (that is, E versus A) realizes the feature [SPEAKER] and the second element (that is, ST versus S) involves [ADDRESSEE]. As for Spanish3, the first element realizes the feature [ADDRESSEE] and the second element—[SPEAKER]. Now, both Spanish2 and Spanish3 would make a strong case for an analysis in terms of three heads provided that the inflection -O is a regular exponent of some morphosyntactic feature, and that none of the two Deic-heads turns out to be an expletive/dummy element (inserted after syntax).

While I know of no such language in Germanic, Romance, or in general, it remains to be seen if such a language exists. I believe not (for some potentially challenging data, see Diessel Reference Diessel1999). However, if such a language can indeed be discovered, the current proposal that involves two heads would have to be expanded to accommodate a third head. Now, while a number of motivating assumptions would have to be rethought, I believe that many points of this paper would still carry over assuming that languages simply differ in the way they encode deixis in the syntax; that is, either a deictic head is unsplit, as in Germanic and Romance, or it is split in two, as in the other (as yet undiscovered) type of language. A technical way to capture such a difference is proposed by Bobaljik & Thráinsson (Reference Bobaljik and Thráinsson1998), who discuss different inflectional systems in the sentential domain in Germanic.

To sum up this section, I have approached the empirical domain under discussion in more abstract terms. On the one hand, this allowed me to identify gaps in the data; on the other, I could formulate two diagnostics for the parsing of the makeup of determiners across languages. In the second part of this section, I argued directly against a structural analysis of demonstratives that posits three heads.

6.1. Other Types of Determiners

In this subsection, I extend the discussion of definite articles to their indefinite counterparts, and the analysis of demonstratives—to question words, often referred to as WH-ELEMENTS. Like the definite determiners, indefinite articles and wh-elements both involve an inflection. To side-step the issue of certain irregular forms of ein ‘a', I illustrate this with a masculine accusative example:

1. (111)

   

Putting special cases such as manch ein ‘some' or irgendein ‘any' aside, the definite and indefinite articles both consist of two components at most: an inflection, and either the definiteness marker or ein. Interestingly, there is evidence that like the definiteness marker, ein is also semantically vacuous. This becomes evident with predicates involving kind nouns where ein makes no independent semantic contribution, as shown in 112a. Turning to wh-elements, one can point out that like demonstratives, they can be modified, as shown in 112b.Footnote ⁷¹

1. (112)

   

A second similarity between demonstratives and wh-elements is semantic. I have shown above that different demonstrative stems are tied to different semantics (for example, dieser ‘this' versus dér ‘that'). The same basic point can be made for wh-elements: welcher ‘which' versus wer ‘who'. Given this initial list of commonalities, I hypothesize that indefinite articles are structurally similar to definite articles and wh-elements are structurally parallel to demonstratives.

In section 3.2, I followed Roehrs Reference Roehrs2009 and Schoorlemmer Reference Schoorlemmer2012 in assuming that the [DEF] feature is base-generated low in the structure. Consider 113a. After movement to Infl_NP, [DEF] is made visible by the inflectional suffix supported by the definiteness marker. I propose something similar for the indefinite article. Following de Swart et al. 2007, kind nouns require the presence of a realization operator (REL). Specifically, REL maps kind nouns to sets of entities, which allows them to function as predicates. This operator is assumed to be in Num, as shown in 113b. I assume that as an operator REL moves to Infl_NP. De Swart et al. propose that this operator is made visible by the indefinite article. Slightly diverging from them, I propose that this is achieved by the inflectional suffix supported by ein.

1. (113)

   

To be clear, one arrives at the simplified representations in 113 that form the basis for parallel derivations of the definite and the indefinite articles.

Turning to demonstratives and wh-elements, these items also involve operators, and I assume they also surface in Infl_NP. As to their internal structure, I propose that wh-elements also involve an extended projection closed off by the inflection. Compare 114a to 114b.

1. (114)

   

Again, just as the proposal respecting definite articles extends to indefinite articles, the current proposal respecting demonstratives extends straightforwardly to wh-elements. In other words, the current analysis can be generalized to cover a larger empirical territory. I take this to be a good indication that the current proposal is on the right track. Finally, I briefly address how personal pronouns and intransitive determiners relate to the transitive determiners discussed so far.

6.2. Personal Pronouns and Intransitive Demonstratives

I start with the often-made observation that personal pronouns are very similar to determiners (Postal Reference Postal and Dinneen1966). This resemblance can easily be seen in Spanish, where some personal pronouns are basically identical to the definite articles: él ‘he/it' versus el libro ‘the book'. Turning to German, it is well known that 3rd person pronouns share the same inflections as determiners. Consider the juxtaposed forms in Table 4, where the personal pronoun precedes the determiner (I give more details below).

Table 4.

3rd person pronouns and determiners in German.

However, personal pronouns seem to behave more like demonstratives than articles with regard to at least two coinciding properties: transitivity and form (see also Harbert Reference Harbert2007:177). First, personal pronouns are intransitive in that they cannot occur with an overt noun, as shown in 115a. Interestingly, demonstratives have two forms in the dative plural (and all the genitive instances), where the intransitive version has the additional ending -en, as shown in 115b. In contrast, the transitive version or the definite article does not, as shown in 115c.

1. (115)

   

Intransitivity coincides with form. Comparing the personal pronoun ihnen in 115a to the accusative masculine forms ihn and den in Table 4, one observes that just like the intransitive demonstrative in 15b, this pronoun also has an additional -en. In fact, there is diachronic evidence that -en has been added. This can be gleaned from the fact that MHG had in and dën as personal pronoun and demonstrative form, respectively, in the dative plural. This additional -en in 115a,b and the other cases is separated by hyphen in Table 4.Footnote ⁷² To conclude, personal pronouns in German pattern with intransitive demonstratives both in terms of their intransitivity (no co-occurring noun) and their form (presence of an additional -en in some instances).

However, there are also some important differences between personal pronouns and demonstratives. To name just two, in contrast to demonstratives, 3rd person pronouns have three different initial elements (e-, s-, ih-) in one and the same paradigm. In this regard, observe also the difference in the vowel between the neuter forms of the pronoun es and the demonstrative dás. Second, unlike the genitive of demonstratives, the genitive of the personal pronouns is stylistically marked in that it only occurs in a very elevated and poetic style. These forms are put in parentheses in Table 4.

While an in-depth investigation of these differences goes beyond the scope of this article, I show how personal pronouns and intranstitive demonstratives fit into the current system. The similarities with regard to inflection can be accounted for by extending the basic demonstrative structure from above to these cases. To capture the semantic differences, I propose, for simplicity's sake, to replace the Deic-head by another head that hosts its own feature. I label the head Ana(phoric). Consider 116a. The differences in terms of stem form can be accounted for by assuming that the three initial elements of the personal pronouns (e-, s-, ih-) are inserted by vocabulary insertion. After head movement and vocabulary insertion, one obtains the forms in 116. To be clear, the derivation of es ‘it' is similar to that of demonstratives such as jenes ‘that'.

1. (116)

   

To finalize the derivations, the additional mechanism of d-support applies to the form in 116b, and 116c involves schwa-epenthesis.

The difference with regard to transitivity is derived by embedding InflP in 116a inside a larger noun phrase structure. Now, similarly to the structure of transitive determiners above, the personal pronoun structure surfaces in Spec, InflP_N. Unlike transitive determiners, personal pronouns have the noun position occupied by another element, the additional -en. This blocks the appearance of an overt noun:Footnote ⁷³

1. (117)

   

For intransitive elements without another overt element (for example, es ‘it'), one has to assume a null element in N.Footnote ⁷⁴ Note also that the structural analysis in 117 fits well with the fact that personal pronouns and demonstratives function as arguments. At least in their base-generated position, these arguments appear in phrasal positions where they receive a theta role. More generally, intransitive elements such as personal pronouns and certain cases of demonstratives find a straightforward account in the current proposal.

To sum up this section, I have briefly sketched how the current proposal can be generalized to cover more empirical ground. Specifically, I have discussed indefinite articles and wh-elements. Finally, I have discussed intransitive elements such as personal pronouns and certain demonstratives and how they fit into the current analysis.