Demonstrative Position in Michif

Abstract

This article consists of a generative analysis of the variable surface ordering of demonstratives in Michif, a mixed language historically derived from French and Cree, and spoken by some Métis. It is claimed that all demonstratives in Michif originate in [Spec, DemP] and raise to [Spec, DP]. Prenominal demonstratives occur when the head of the movement chain is pronounced, while postnominal demonstratives are the result of two factors: first, the pronunciation of the tail rather than the head of the demonstrative's movement chain, and second, the noun undergoing a last resort p-movement, adjoining to DemP. The different patterning is motivated via meaning differences in the corresponding patterns, appealing to the differences in the featural makeup of demonstratives. Pragmatic information, said here to be a contrastive focus feature, is posited on some demonstratives while not on others, yielding the different ordering and also a different interpretation. The variable nature of demonstratives cross-linguistically is also discussed.

Résumé

Cet article présente une analyse générative de l'ordre variable des démonstratifs en Michif, une langue mixte historiquement dérivée du francüais et du cri, parlée par certains Métis encore aujourd'hui. Il est proposé que tous les démonstratifs en Michif proviennent de [Spec, DemP] et sont déplacés [Spec, DP]. La position prénominale est le résultat de la prononciation de la tête de la chaîne de mouvement, alors que la position postnominale peut s'expliquer par deux facteurs : premièrement, la prononciation de la queue plutôt que de la tête de la chaîne de mouvement du démonstratif, et deuxièmement, le nom subissant un mouvement de dernier recours pour des raisons phonologiques, s'adjoignant DemP. L'ordre différent est motivé par des traits différents associés avec les démonstratifs, ce qui mène un sens différent entre les deux ordres possibles. De l'information pragmatique, ici un trait de focus contrastif, est présente sur certains démonstratifs, expliquant les différences dans l'ordre et dans l'interprétation. La nature variable des démonstratifs dans d'autres langues est également discutée.

The goal of this article is to begin the study of the DP structure in Michif within a generative framework. Michif is a mixed language historically derived from French and Cree, and spoken today by some Métis in parts of Manitoba, Saskatchewan, Montana, and North Dakota. Recent work claims Michif to be stratified on the basis of historical source, with two distinct grammars: that of French for the French vocabulary, and that of Cree for the Cree vocabulary (Bakker 1997; Bakker and Papen 1997). Though this claim will not be overtly discussed here, a synchronic account for the Michif DP facts is presented, focussing on the ordering of the demonstratives, implying that two separate grammars may not be necessary to account for Michif data. In section 1, the Michif DP data is presented, using a comparison with French as a starting point, given that the Michif DP vocabulary is primarily of French extraction. In section 2, a syntactic structure for demonstratives in Michif is proposed, based on structures hypothesized for some other Romance languages by authors such as Giusti (1991), and Bernstein (1993, 1997, 2001), and the different ordering possibilities of the Michif demonstrative are derived from one basic structure. Details of the account are expanded upon in section 3, and section 4 presents further evidence in support of such an analysis.

1. DP Internal Word Ordering in Michif

This section outlines the Michif DP, including the forms of the DP-internal elements, their ordering, and the agreement patterns. It has been claimed that the Michif DP follows the same patterning as the French DP¹ (Bakker 1997, Bakker and Papen 1997). However, in comparing the two,two obvious differences become apparent. First, Michif, unlike French, allows the cooccurrence of a determiner and a demonstrative, and in fact, a determiner is always required, as shown in the following data:²

(1) Michif

awa  la    fij

DEM DET girl

'that girl'

*awa  fij

DEM girl

'that girl'

(2) French

*cette la      fille

DEM DET girl

'that girl'

cette fille

DEM girl

'that girl'

The second difference with regards to French is the fact that the position of the demonstrative is not completely fixed. Note the data in (3a–e).

(3)

1. a. trwa  anIkIk li    zvo

   three DEM    det horse

   'those three horses'

2. b. awa  la    fij

   DEM DET girl

   'that girl'

3. c. *la    awa  fij

   DET DEM girl

4. d. la    tab    anIma

   DET table DEM

   'that table (there)'

5. e. li      fij    smart okIk

   DET girl smart DEM

   'those smart girls (there)'

Given the possibilities, we must be able to account for the following orders in Michif:

(4)

1. a. DEM DET NP

2. b. *DET DEM NP

3. c. DET NP DEM

The bulk of this article will concentrate on the position of the demonstrative, and on deriving its ordering within the DP. However, in order for the data in the following sections to be clear, the remainder of this section will lay out the forms of the determiners and demonstratives, as well as the agreement patterns in the Michif DP data.

It may be of interest to note that the Michif determiners, adjectives, numerals, and nouns can be derived historically from French, while the Michif demonstratives are derived from the Cree forms. Both Dem and Det in Michif mark number, while the Dem marks animacy and the Det marks masculine or feminine gender, as is illustrated in the following examples:

(5)

1. a. anIma          la          ∫ez  ver

   DEM.INT.SG.INAN DET.FEM.SG chair green

   'that green chair'

2. b. anIkIk          li      ʒvo

   DEM.INT.PL.ANIM DET.PL horse

   'those horses'

Bakker and Papen (1997) and Papen (2002) claim that, in most cases, Michif borrows the grammatical gender and animacy along with the lexical item from the source language, though the details of how this would be remain unclear. For example, if a French noun is masculine, the corresponding noun in Michif would be predicted to be masculine. The forms of the Michif demonstratives and determiners are given in Tables 1 and 2, respectively.³

Table 1.

Michif demonstratives

Table 2.

Michif determiners

Let us now turn to Michif adjectives. While the majority of adjectives in Michif are postnominal, a small group appear only prenominally. These adjectives agree in gender (though sometimes the feminine morpheme may be Ø), and can never appear postnominally.⁴ The group includes, among a few others; gro/gros 'big', pt ∫i/pt ∫it 'small', ʒœn 'young', vjœ/vjej 'old', grã/grãd 'big', bo/bεl 'beautiful'. However, the vast majority of adjectives, including any innovations such as English borrowings or Cree-source deverbal adjectives, are postnominal, and can never be prenominal, leading to the conclusion that the postnominal position is the adjective's syntactic position, while the prenominal adjectives are lexically determined. The example in (6) shows the ungrammaticality of a lexically determined prenominal adjective such as gros 'big' appearing postnominally, while the example in (7) shows the ungrammaticality of a "regular" adjective such as vεr 'green' appearing prenominally.

(6) *εn          fij  gro(s)

INDEF.FEM.SG girl big-FEM

'a big girl'

(7) *la         vεr    fij

DEF.FEM.SG green girl

'the green girl'

In Michif, the lexically determined adjectives show gender alternations, as in the contrast below:

(8)

1. a. εn          gros      fij

   INDEF.FEM.SG big-FEM girl

   'a big girl'

2. b.           gro     garsõ

   INDEF.MASC.SG big-MASC boy

   'a big boy'

3. c.           gro     t∫a∫t∫ãgõ

   INDEF.MASC.SG. big-MASC teeter-totter

   'a big teeter-totter'

However, postnominal adjectives do not display any agreement, as illustrated in (9).

(9)

1. a. la         fij  ver

   DEF.FEM.SG girl green-Ø

   'the green girl'

2. b. lI          garsõ ver

   DEF.MASC.SG boy   green-Ø

   'the green boy'

If the postnominal position is the syntactic position for adjectives, and there is no postnominal adjectival agreement, then this amounts to saying that there is no adjectival agreement in Michif. There are, in fact, a few cases of apparent postnominal agreement. However, these cases are better thought of as lexicalized expressions rather than true adjectival agreement. This is illustrated in the minimal pair below:

(10)

1. a. εn        fij   blã-∫

   IND.FEM.SG girl white-FEM

   'an ethnically white/Caucasian girl'

2. b. εn        fij   blã

   IND.FEM.SG girl white

   'a white-coloured girl' (as in a painting)

Given this data, we may assume that postnominal agreement is not part of the regular Michif grammar, and fossilized occurrences of French-type agreement patterns occur solely in a few lexicalized expressions. A summary of the agreement marking is shown in the table in Table 3.

Table 3.

Nominal agreement marking in Michif

The main generalization is that postnominal adjectives always remain invariable in Michif, while all other DP elements agree in at least one of number, gender, and animacy. Now that the details of the data have been laid out, let us turn to the question of ordering of elements within the Michif DP.

2. Position of Dem and Det

In this section, the cooccurrence of Dem and Det in Michif is discussed in comparison with some other Romance languages. It is then shown that the Dem should be considered to be in the specifier of a functional position between DP and NP, following other proposals made in the literature. Finally, an overview of the analysis for the different word order possibilities of the Dem in Michif is presented. The details of such a proposal are developed in sections 3 and 4.

2.1. Cooccurrence of Dem and Det

The first question to be addressed is the ordering of the demonstrative and the definite determiner shown in (3). Given that it has been assumed that the Michif DP simply follows the rules of the French DP (Bakker 1997; Bakker and Papen 1997), it should be noted that the very fact that demonstratives and determiners cooccur in Michif is in itself unusual, since in French the two elements are in complementary distribution. The following examples illustrate the ungrammaticality for the French equivalents of acceptable Michif DP word orders in (3).

(11)

1. a. *trois  ces   les   chevaux

   three DEM DET horses

   'those 3 horses'

2. b. cette fille / *cette la    fille

   DEM  N        DEM  DET N

   'that girl'

3. c. la   table *cette / *-là

   DET table   DEM

   'that table'

4. d. *les  filles intelligentes ces

   DET girls smart          DEM

   'those smart girls'

The Michif facts are not inherently unusual, but are unexpected given an assumption that the Michif DP patterns in the same way as the French DP. Demonstratives and determiners can certainly cooccur in other languages such as Greek, Javanese, and Welsh, among others:

(12) Greek (from Giusti 1992, citing Lundeby 1965 and Heinrichs 1954):

• autòs ó   anēr

• this   the man

• 'this man'

(13) Javanese (from Giusti 1992, citing Lundeby 1965 and Heinrichs 1954):

• ika ń   anak

  this the boy

  'this boy'

(14) Welsh (from Dryer 1992, citing Jones and Thomas 1977:167):

• y   ty     ma

  the house this

  'this house'

Although French and Italian do not permit the cooccurrence of both Dem and Det (see Giusti 1991, Brugè 1996, and Bernstein 1993, 1997, 2001 for the details regarding Romance demonstratives), Romanian allows constructions with both the definite determiner and Dem marked. The definite determiner is always realized as an enclitic in Romanian (15a). Dem occur prenominally when there is no definite determiner (15b) but must occur postnominally when a determiner is present (15c–d). (Examples from Gabriela Alboiu, p.c.)

(15)

1. a. băiat-ul

   boy-DET

   'the boy'

2. b. acest băiat

   DEM  boy

   'this boy'

3. c. băiat-ul  acesta

   boy-DET DEM-a

   'this boy'

4. d. *acest băiat-ul

   DEM  boy-DET

Giusti (1992) follows Grosu (1988) in deriving the order in (15c) from the order in (15b), by N-movement of băiat 'boy' over the demonstrative acest. The details of the analysis of the Romanian data are not relevant for the present article (see Grosu 1988, Cornilescu 1992 for details), but note that Romanian is a language which marks both DET and DEM, though it differs from Michif in that Michif obligatorily fully realizes both the demonstrative and the determiner.

The fact that demonstratives and articles are frequently in complementary distribution across languages has led to the preliminary assumption that they simply occupy the same slot in syntax. Of course, Det and Dem normally represent different semantic features. A [+definite] feature is normally attributed to definite determiners to distinguish them from indefinite determiners, while an additional feature relating to deixis or specificity is normally attributed to demonstratives such as [±Demonstrative], [±Proximate], or [±Specific] (Lyons 1999). Such a differentiation between definite determiners and demonstratives will be assumed here, though the specifics of the feature used to distinguish the two elements is outside the scope of this discussion, and will not be touched upon.⁵

In keeping with the different semantics, further work on DPs since the early 1990s has shown that the demonstrative and the definite article do not correspond to the same structural position (Giusti 1991, 1992; Cornilescu 1992; Brugè 1996; Roca 1996; Bernstein 1993, 1997). Most notably, as we have seen in the examples in (12)–(15), as well as in the Michif examples, several languages allow cooccurrence of the two elements, pointing to two positions structurally.

There are also typological reasons for splitting the categories determiner and demonstrative. For instance, Dryer (1992) reports that VO languages show a great tendency to have Article-Noun order, and OV languages show a tendency to have Noun-Article order. He generalizes that articles are what he calls "verb patterners", meaning that the Article-Noun word order is predictable based on whether it is a VO or OV language. Demonstratives, on the other hand, follow no such pattern, which may be further evidence they should not be thought of as the same element. These tendencies found by Dryer (1992) offer cross-linguistic support to an analysis which would see a split between Dem and Det positions.

2.2. Dem position

Assuming, now, that Dem is in a different structural position than Det, then the next step is to determine in what position it is generated. One possibility is that Dem is a type of adjective, as Cinque (1996) suggests. He tentatively places Dem in a high position in his universal serialization of adjectives, above numeral and descriptive adjectives. However, we will reject this possibility for Michif due to the fact that Dem patterns differently than adjectives. In Michif, all demonstratives may appear either pre- or postnominally, while we saw in section 1 that adjectives are generally fixed in a postnominal position. It is the case that there are some lexically determined pre-nominal adjectives as well, but these cannot be moved into a postnominal position:

(16)

1. a. ẽ            gro garsõ

   INDEF.MASC.SG big boy

   'a big boy'

2. b. *ẽ            garsõ gro

   INDEF.MASC.SG boy   big

   'a big boy'

3. c. la         fij   ver

   DEF.FEM.SG girl green

   'the green girl'

4. d. *la         ver    fij

   DEF.FEM.SG green girl

   'the green girl'

Unlike the strict positioning of the adjectives, we have seen in (3d) and (5a) that the same Dem can appear on either side of the NP.⁶ Note as well that even Cinque (1996:454) allows the possibility that the Dem may indeed be generated in a lower position and moved higher in some languages, while staying in situ in others. Given the patterning differences between adjectives and demonstratives, we argue that Dem is not a type of adjective in Michif, and is rather a different element.

A second analysis, the option argued for by other authors studying DPs in Romance languages such as Spanish, Italian, Catalan, and French (Giusti 1992; Bernstein 1993, 1997, 2001), is one where the Dem is base-generated in the specifier of a functional projection between DP and NP. The following section addresses the question of how to derive the different orderings of Michif under such an analysis.

2.3. Prenominal Dem ordering

The working assumption for the remainder of this article will be that the merge position of Dem is in the specifier of a functional projection between DP and NP, following the Romance DP authors cited above. This functional projection will be called DemP.⁷ The determiner, in contrast, is base-generated as the head of D. The structure in (17) is for Michif, but is identical to the structures proposed by Bernstein for the Romance languages she discusses.

Both Giusti and Bernstein propose analyses involving the raising of Dem out of its base-generated position; the analyses differ with regards to the landing sites of the raised Dem. Giusti proposes that demonstratives undergo regular specifier movement into [Spec, DP] for feature checking, while Bernstein claims that they undergo raising and substitution into the head of D.⁸ This substitution is motivated by the fact that definite articles and demonstratives do not cooccur in most Romance languages. She argues that by adopting an analysis which substitutes the demonstrative into the head of D, we are able to predict this impossible demonstrative-determiner cooccurrence without resort to filters, as needed in Giusti's proposal. While this proposal may account for the facts in most Romance languages, it clearly does not account for the Michif facts, given that demonstratives and determiners cooccur. Bernstein admits that in languages which allow cooccurrence of determiners and demonstratives, an analysis where demonstratives raise to a specifier position may be preferable. Given that Michif is such a language, we suggest that demonstratives in Michif are base generated in [Spec, DemP], following these assumptions, and that the demonstrative undergoes regular specifier-to-specifier movement to yield the word order in (3a–b), where the demonstrative is prenominal. This is shown in the structure in (18).

2.4. Postnominal Dem ordering

In addition to the prenominal demonstratives, we also get cases as in (3d–e), where the demonstrative is postnominal. We propose that this ordering may be derived from the base order given in (17) through a prosodically motivated scrambling process such as proposed by Zubizarreta (1998). Zubizarreta proposes this movement for clausal focus-driven movement, where it appears that the object gets scrambled leftward to yield a subject-final sentence. Zubizarreta's proposal will be discussed in some detail in the next section before moving on to its relevance to the Michif facts.

2.4.1. Zubizarreta's theory of p-movement

Zubizarreta (1998) examines a well-known property of many Romance languages, specifically that although they are characterized as underlyingly SVO, VOS ordering is possible just in the cases where the clause-final subject receives main sentential stress (see also Ordóñez 1997, 1998). In the derived VOS sentences below (taken from Bernstein 2001, citing Ordóñez 1997), main sentential stress is required on the clause-final subject in order to yield a grammatical sentence:⁹

(19) Spanish

1. a. A quién le presto el diccionario JUAN?

   to whom CL-lent the dictionary Juan

   'Who did JUAN lend the dictionary to?

2. b. Espero que te devuelva el libro JUAN.

   I-hope that CL-you-return the book Juan

   'I hope that JUAN returns the book to you.'

(20) Catalan:

• Demà comprar un libre LA MAGDA.

• tomorrow will buy a book Magda

• 'Tomorrow MAGDA will buy a book.'

(21) Italian:

• Ha recensito il libro UN PROFESSORE.

• has summarized the book a professor

• 'A PROFESSOR has summarized the book.'

According to Ordóñez and Zubizarreta, in the sentences where stress is on the subject, the object is scrambled leftward to the specifier of a functional projection. Although Ordóñez does not specify what triggers the scrambling, Zubizarreta claims that it is for prosodic reasons. Let us examine this claim further.

The Nuclear Stress Rule (NSR; see Chomsky 1971, Jackendoff 1972) in Romance languages such as French, Spanish, Catalan, and Italian, assigns prosodic prominence to the clause-final, or right-most constituent. Normally, then, in an SVO sentence in these languages, the object will receive main sentential (nuclear) stress. In Spanish, according to Zubizarreta (1998:125), both VSO and SVO structures are compatible with no particular focus interpretation, in which case sentential stress falls on the last constituent.¹⁰

(22)

1. a. Maria me regalo la botella de vino.

   Maria to-me gave the bottle of wine

   'Maria gave me the bottle of wine.'

2. b. Me regalo Maria botella de vino.

   to-me gave Maria bottle of wine

   'Maria gave me a bottle of wine.'

However, if the subject is sentence-final, then an emphatic/focus interpretation is obtained, and the neutral focus interpretation is no longer available.

(23) Me regalo la botella de vino MARIA.

to-me gave the bottle of wine Maria

'Maria gave me the bottle of wine.'

Zubizarreta (1998) concludes that due to the difference in interpretation, the prominence must not be assigned by the regular NSR, as in the examples in (22). She proposes, rather, that the subject in (23) receives stress due to her Focus Prominence Principle (FPR, 1998:21). This principle states simply that focussed constituents receive prosodic prominence. Zubizarreta therefore derives the sentence in (23) with VOS order from the VSO sentence in (22b). This VSO sentence was in turn derived from base-generated SVO (22a) by raising the verb out of the VP projection. Since the derivation from SVO to VSO is not important for our purposes, we will simply concentrate on Zubizarreta's explanation of VOS from VSO.

In the sentences in (19)–(21), as well as in (23), the subject is being focussed. Given the FPR, the focussed subject must receive prosodic prominence in order to be appropriately interpreted. However, if the subject remains in a non-final position, it is unable to receive prominence due to the NSR, which assigns prominence to the right-most element. In order to resolve this contradictory prosodic prominence assignment, the right-most element, in this case the object, is scrambled leftward to allow the stress assignment onto the subject, thus deriving the VOS order. Zubizarreta calls this type of scrambling prosodic, or p-movement, a type of non-feature-driven movement. It is a Last Resort operation carried out to ensure the resolution of a prosodically contradictory situation between the two prosodic assigners, and ensuring that the focalised subject is in a position to receive prominence via the NSR, thus ensuring that the output is compatible with the FPR (1998:124).

Zubizarreta's proposal accounts for the different word orderings within the clause. Recall that similar differences in word orderings are available within the Michif DP (and in other languages, as will be shown in section 4). An analogous proposal is possible which would account for the ordering of elements within the DP, in the spirit of Abney's (1987) proposal for a functional category above NP, seen as analogous to IP over VP. Bernstein (2001) argues just this, the details of which are outlined in the next section.

2.4.2. p-movement within the DP

Bernstein (2001) extends Zubizarreta's theory of p-movement within the clausal domain to account for DP element ordering patterns. Specifically, across Romance languages, various contrastively focussed elements are found as the right-most elements of the DP along with prosodic prominence. These elements appear prenominally in the unmarked case, but postnominally when focussed, as in the following examples taken from Bernstein (2001), with the exception of (26) added here for comparison purposes.

(24) Spanish:

1. a. este libro interesante

   this book interesting

   'this interesting book'

2. b. el libro interesante ESTE

   the book interesting THIS

   'THIS interesting book'

(25) Italian:

1. a. il mio libro importante

   the my book important

   'my important book'

2. b. il libro importante MIO

   the book important my

   'MY important book'

(26) French:¹¹

1. a. cette femme intelligente

   this woman intelligent

   'this intelligent woman'

2. b. cette femme intelligente CI

   this woman intelligent here

   'THIS intelligent woman'

Bernstein shows that the elements occupying the DP-final position in Romance generally receive a focus interpretation, generalizing that Romance languages express focus on the right periphery of DP. Crucial to her argument is the observation that there is a meaning difference between the (a) and (b) orderings in the phrases in (24)–(26), just as there is a meaning difference in the subject-final sentences in Zubizarreta's work. Bernstein shows the similarity of the ordering facts within the Romance DP with the ordering facts within the clause, where the right-most element receives stress in both. She claims that the DP-final elements seen in (24b), (25b), and (26b) are in fact base generated prenominally, and that their postnominal position is derived from leftward XP movement of another element. Bernstein's proposal for the Spanish data in (24b) is illustrated in the tree in (27). Bernstein argues that this movement is an instance of p-movement, analogous to the clausal movement for focus interpretation in Zubizarreta (1998).

In the next section, we will apply Bernstein's (2001) analysis to the Michif DP in order to account for the demonstrative ordering differences within the DP.

2.5. Back to the Michif postnominal Dem

Given the similarities between French and Michif DPs, applying Bernstein's (2001) analysis to Michif is fairly straightforward. First, we have assumed that the demonstrative is base generated prenominally in the specifier of the DemP, a functional projection above NP and below DP. The demonstrative appears post-nominally on the surface in examples like those in (3d-e) because of leftward scrambling of the NP over DemP. This scrambling is motivated by the need to assign prosodic prominence for emphasis of the demonstrative on the right periphery of the DP. In order to conclude that Michif behaves akin to Bernstein's analysis of other Romance DPs, however, certain facts need to be made clear. Specifically, it is crucial that there be meaning differences between the prenominal and postnominal instantiations of Dem. This does appear to be the case. In the Michif data, like the other Romance data shown in (24b) and (25b), a postnominal Dem yields a different interpretation.

(28)

1. a. li  fij    smart OKIK

   the girls smart these

   'THESE smart girls (not those)'

2. b. dã la    gro∫∫eʒ   ANIMA

   on DET big   chair DEM

   'on that big chair (no other one)'

3. c. LI      nɔm AWA stε ẽ    

   DET young man DEM  is    DET drunk

   'This young man here is a drunk.'

4. d. Awa lI       nɔ Inεhwε∫tew.

   DEM DET young man drinks.a.lot

   'This young man drinks a lot.'

Though it is clear that there is a difference in contextual usage of the two orderings, this difference is not always easy for our consultant to define. However, we can be sure about certain facts: when a contrastive reading is desired, and whenever the demonstrative is stressed, the demonstrative must appear to the right of the NP, as in (28a–c). As well, the difference in the translation of the sentences in (28c) and (28d) shows that the locative or deictic features of the demonstrative are being emphasized in (28c), while in (28d), the demonstrative is neutral. The examples with postnominal demonstratives are consistently translated by adding the reinforcing English adverb 'here' or 'there' to 'this' and 'that', which is not the case for the prenominal demonstratives. This follows the same pattern as other Romance DPs, if, as Bernstein suggests, the right-most element within the DP is the locus of focus or emphasis. Given the similarities, we will follow Bernstein (2001) and analyse these cases of postnominal demonstratives as cases of leftward scrambling of the NP to a specifier position above DemP, as in the tree in (29), for the DP in (28a).

This proposal is in line with Bernstein's (2001) analysis of Spanish and French. However, a few important questions remain. Note that in the structure in (29), the demonstrative does not raise to [Spec, DP]. Recall from section 2.3 that when the demonstrative is prenominal, it does not remain in its base-generated position in [Spec, DemP], but rather raises to [Spec, DP]. One may wonder why there is Dem movement when there is no NP movement, but no Dem movement when there is NP movement. This issue is addressed in the next section.

3. Pronunciation and Agreement

An analysis such as has been proposed in this article, where the demonstrative appears to move in one instance and remain in its base-generated position in another, requires some explanation. It is proposed that these differences in word orders may be attributed to different pronunciation positions within the framework of the copy theory of movement, following Chomsky (1995, 2000) and Richards (1999). Section 3.1 will outline this proposal while section 3.2 will motivate intra-categorial differences between demonstratives, which allow them to pattern differently. Section 3.3 will deal briefly with issues of agreement within the Michif DP.

3.1. Demonstrative pronunciation

A basic tenet of Chomsky (1995) is that feature checking is the driving force behind movement. That is, there are uninterpretable features which attract like features, and these pairs must be checked off during syntax in order to complete the derivation, and if not, then the derivation crashes, resulting in an ungrammatical sentence. These features may be strong, in which case they trigger overt movement of a syntactic element up the structure. Alternatively, these features may be weak, in which case they attract only the features themselves, leaving the syntactic element in its base position, resulting in what is called covert movement.

This overt/covert movement distinction has been recast by Richards (1999) and Chomsky (2000) within the copy theory of movement. Essentially, when an element is attracted upwards in the structure by a strong feature, it leaves behind a copy of itself, rather than a trace, forming a chain between the two (or more) positions in the structure. At the head of the chain is the element in its higher position, and at the tail is the copy in its lower, base-generated position. If the feature is weak, there is no movement, and there is no chain created. When there is a chain, instructions are sent to pronounce the higher copy, or the head, of the chain. Under this theory of movement, we are able to derive the surface position of the Michif demonstratives through different pronunciation instructions at PF.

The base-generated positions of the Dem and Det elements in Michif, as given above, are shown again here in (30).

In order to get the surface order of the Michif data in (3a–b), a straightforward analysis would be to posit D-features on the head of DP. Given that Dem agrees with animacy and number, these are the features we propose here. Therefore, uninterpretable agreement features for animacy and number on D attract Dem, triggering specifier-to-specifier movement of Dem from its Merge position in [Spec, DemP] up to [Spec, DP], while leaving behind a copy of the Dem as seen in the schema in (31).¹²

For the cases where the demonstratives appear prenominally,the presence of strong features accounts for the instructions at PF to pronounce the higher copy, resulting in the word order of the Michif prenominal demonstratives above the determiner. However, in the postnominal cases as in (3d–e), where the NP scrambles above the demonstrative, it appears that the demonstrative must be pronounced in its base-generated position, yielding the pronounced order as represented in the tree in (32).

This Michif data is somewhat reminiscent of data in Alboiu (2000, 2001), where the pronunciation site of contrastively focussed elements in Romanian appears to be either at the head or the tail of a syntactic chain. The claim for Michif here is that in some cases we must pronounce the tail of the chain, raising the question of why it would be that in some cases, the head is pronounced, and in others, the tail.

3.2. Why different pronunciation instructions?

In this section, we discuss the reasons behind the different pronunciation instructions of the elements of the chain, proposing that these differences are related to the features associated with the Dem in question. Essentially, the proposal is that demonstratives may have two different properties associated with them. Demonstratives always have regular uninterpretable functional features, but in some cases, they also have parasitic deictic or pragmatic properties associated with them, which cause them to pattern differently. It should be no surprise that Michif demonstratives carry pragmatic information, as one of the principal differences between definite determiners and demonstratives is the fact that demonstratives pick out a specific, contrastive reference (Hawkins 1978). Given this contrasting quality of demonstratives, we propose a treatment of these features as contrastive focus features. This is consistent with Reinholtz (2002), who also considers demonstratives in Cree to be contrastively focussing elements, though her treatment of these demonstratives is quite different from the one proposed here.¹³ These contrastive focus features are merged along with the demonstrative into [Spec, DemP], the eventual tail of the Dem chain. Though they contain no phonology of their own, phonetic realization of the pragmatic properties is required. Given that these demonstratives are stressed when they appear in their deictically charged, or focussed use, it seems reasonable to posit that stress is a requirement in the pronunciation of focussed material in Michif.¹⁴ Therefore, even though the default pronunciation of Dem in Michif is at the head of the chain, if we posit that stressed constituents cannot delete, and that the element that needs stress is the contrastive focus feature merged in at the tail, this explains the pronunciation of the tail in these cases. The trees in (33) and (34) show the two structures for the two different Dem orderings.

In (33), there are no contrastively focussing properties present, and so the regular default pronunciation of the head of the chain occurs. This default pronunciation of the head (and not the tail) is in keeping with assumptions in Richards (1999) that it is always the head of a chain which gets pronounced when there is a strong feature. To pronounce the tail would imply a weak feature, as only the features would be attracted, and thus there would be no chain formation at all. The morpho-syntactic feature checking would not be able to take place in Michif, and the derivation would crash.¹⁵

(33) Pronunciation: head of chain, awa la fij

In (34), as well as having grammatical features, the contrastive focus feature present in the structure needs to be interpreted. These features affect the structure in two ways: they provide an emphatic reading of the demonstrative, causing the p-movement of the NP to the left of the Dem, as discussed in section 2.4.2, in order to yield the correct prosodic prominence, and they cause the pronunciation of the tail of the Dem chain. The proposal here is that the contrastive features are present on the tail, and that they require phonetic realization.¹⁷ The pronunciation of the head and not the tail in these cases would yield a wrong interpretation, as no focus would be present.

(34) Pronunciation: base-generated position la fij awa

Returning to the Romanian data mentioned above, we find that this proposal is also along the lines of Alboiu (2000), who argues that the pronunciation site of a contrastively focussed element in Romanian may be at either the head or the tail of the chain due to featural differences between the elements. According to Alboiu, the [+focus] feature on I0 is non-selectional, unlike the [wh-] feature, and therefore PF will not be instructed as to which of the copies to pronounce (Alboiu 2000:317). In this case, the pronunciation site is optionally either the head or the tail of the chain, as long as there is prosodic marking on the element. The claim presented here is similar in that pronunciation is phonologically driven, and signalled prosodically, rather than a result of morpho-syntactic feature checking. It is also in line with the treatment of contrastive focus as a feature on the element.

Such a proposal makes certain predictions,raising questions regarding demonstratives. First, it assumes a somewhat heterogeneous Dem category, which would carry extra pragmatic features in certain contexts and not in others. Further, we are left to wonder in the cases where these focus features are not present, how this demonstrative would differ from a definite determiner. These questions are discussed in section 4, with support for the present proposal coming from a cross-linguistic examination of the nature of demonstratives. Before turning to this section, however, a proposal for an account of the Michif DP agreement facts is given.

3.3. Michif DP agreement

Although agreement within the DP elements does not seem to have been a source of concern within the Romance demonstrative literature, one possibility of its mechanics is given here. Recall from section 1 that all prenominal elements and all demonstratives agree with the noun in at least one of number (#), gender (G), or animacy (A), while regular adjectives in postnominal position do not mark agreement. As we can see from the structure in (35), all agreeing elements in the Michif DP are in specifiers of prenominal heads.

The prenominal elements agree with the features of the head of NP through the movement of these features up through each head on its way to D. Note that in more recent minimalist theory (Chomsky 2000), we may simply posit agreement relations between each head. If Michif agreement is done via agreement through the prenominal heads, it is no surprise then that the regular postnominal adjectives do not agree with the head noun.¹⁸

Now that it has been established that there should not be any problems deriving the agreement facts, let us return to the problem of why demonstratives are not patterning in a homogenous way.

4. Nature of Demonstratives

This section discusses the nature of the demonstrative category in grammar, to support the idea that elements in this category may carry both discourse-based and functional features (see Giorgi and Pianesi 1996 for the idea that both types of features may be associated with one constituent). Section 4.1 examines the general semantics and function of demonstratives, showing that demonstratives are dual-functioning elements, as both functional and pragmatic, or discourse-based items. Section 4.2 investigates differences between demonstratives and definite determiners, while section 4.3 shows some patterning differences among demonstratives in different positions with relation to the noun, and in different languages.

4.1. Semantics of Demonstratives

Demonstratives are place or spatial deictics, which normally encode two types of features: deictic features and qualitative features (Diessel 1999). For Diessel, deictic features are features which refer to the location of the referent with respect to the speaker, and qualitative features are those which encode the classificatory information about the referent. Translating this characterization into more familiar generative grammar terminology, these "qualitative" features can be considered uninterpretable agreement features, while the deictic features¹⁹ associated with demonstratives may be thought of as the pragmatic part of the information encoded by demonstratives.

We often think of demonstratives as distance markers, marking at least a two-way contrast often labelled proximate and distal. However, Himmelmann (1997:53–62) argues that demonstratives do not crucially encode this deictic contrast. Based on colloquial German, he shows that German dies 'this' and das 'the' must be accompanied by a demonstrative adverb²⁰ in order to contrast location, for example, das Haus da 'the house there'. While Anderson and Keenan (1985:280) argue that a deictic expression unmarked for distance would be little different from a definite article, Himmelman shows that it serves the same pragmatic function as demonstratives, which is to focus the hearer's attention on entities in the speech situation. In order to indicate any distance features, however, some sort of pointing gesture is required. This is a crucial departure from what we sometimes think of as the function of demonstratives; that is, that rather than necessarily encoding a distal contrast, we may think of some demonstratives as a "value-added" definite article, which focusses the attention of the hearer on the speech situation. As mentioned in section 2.1, the value [± specific] is one that has been attributed to demonstratives, but not to definite determiners, which do not necessarily refer to distance.

Diessel (1999:38) agrees that there are distance-neutral demonstratives, and that they are often reinforced with distance-marked demonstratives. Note that these reinforcing distance-marking demonstratives may be omitted if we have some sort of gestural information indicating location. This is the case in English as well, where it can be argued that this car does not really indicate any contrastive distance features on its own without the addition of a pointing gesture, or a different intonation, focussing or stressing the demonstrative. However, these extra-linguistic elements may be replaced with here or there, as in this car here (or that car there) presumably due to the deictic or pragmatic features encoded in here and there.

We may think of demonstratives, therefore, as encoding two types of features: both grammatical, and pragmatic or distal features. Feature syncretism (Giorgi and Pianesi 1996) allows for both grammatical features and discourse-based features to be associated with one constituent.

4.2. Typological differences within demonstratives

In section 2.1, typological differences between articles and demonstratives found in Dryer (1989, 1992) were offered as evidence that the two elements should not occupy the same syntactic slot. In this discussion of demonstratives, Dryer elaborates that there appear to be two distinct types of demonstratives, those that pattern along with adjectives, and those that pattern along with articles. Demonstratives in some languages are ordered like adjectives, as in Welsh (see example (14)), where demonstratives appear postnominally, like adjectives, whereas determiners appear prenominally.²¹ However, in English, determiners and demonstratives always appear in complementary distribution, but demonstratives and adjectives are not subject to any such restriction. Dryer predicts that if we were to divide the languages he studied into two categories; those in which the demonstratives were adjective-like and those in which they were article-like, we would be able to see the word-order pattern for articles and nouns mentioned in section 2.1 with the article-type demonstratives as well.

Mithun (1987) also argues that demonstratives in languages such as Tuscarora are functionally very different from demonstratives in languages such as English. She argues that in some languages, demonstratives and the nouns they modify do not form syntactic units like those in English. It is not the goal of this article to defend or argue against this point of view, but rather to show that there are clearly cross-linguistic differences between items said to be within the category labelled demonstrative. These differences clearly imply that the category of demonstrative is not necessarily homogeneous cross-linguistically. Bernstein, in her work on demonstratives in Romance, is reluctant to admit that this is the case (1997:95, 2001), which seems overly conservative given the different patterning of demonstratives in different languages.

4.2.1. Interpretation differences based on demonstrative position

An interesting fact which seems to be prevalent across several languages is that there are differences in interpretation of demonstratives within a given language, depending on the position of the demonstrative with relation to the noun. An example of patterning differences is Romanian, where prenominal demonstratives are inflected differently from the postnominal demonstratives. Prenominal demonstratives are inflected according to the adjectival paradigm, while postnominal demonstratives are inflected according to their pronominal counterparts (Alboiu, p.c.). Of special interest to us here are the different interpretations available for the demonstrative based on its position relative to the noun it modifies.

Dryer (1989) reports that in Swahili, the form of the demonstrative is the same as that of the determiner, but the interpretation of the element is dependent on its positioning with respect to the noun. When the element is prenominal, it functions as a determiner, and when it is postnominal, it functions as a demonstrative, as illustrated in (36).

(36)

1. a. yule      mtu

   DEM/DET man

   'the man'

2. b. mtu yule

   man DEM/DET

   'this man'

This also appears to be the case in Montagnais, an Algonquian language. Cyr (1993:68) reports that:

In Montagnais, it is the case that preposed demonstratives correspond closely to definite articles in other languages while postposed demonstratives behave more like demonstratives ... on the second mention of a definite referent, and/or at paragraph boundaries ... we can infer that the preposed demonstratives of Montagnais are in fact definite articles, while the postposed and circumfixed ones seem to be genuine demonstratives.

Cyr argues that the preposed demonstratives function identically to a definite article, contrary to the general perception that Algonquian languages do not have articles. It is interesting to note that Montagnais is related to Michif closely enough that they share identical forms for the demonstrative vocabulary. An example of a prenominal demonstrative in Montagnais taken from Cyr (1993:72) is given in (37).

(37) ketahtawe  esa       awa  ostesimaw kamahmamitoneyihtahk

at one time apparently DEM oldest      was.always.thinking

enohteitohtet     otenahk

he.wanted.to.go to.town

'Then there came a time when the oldest was thinking he wanted to go to town.'

The patterning described in Cyr reinforces the proposal we have outlined in this article, where a difference in demonstrative position entails a difference in interpretation.

We have attempted to show in this section that demonstratives display different patterns, both cross-linguistically and within a single language. In many of the languages that we have seen in this article (Tuscarora, Swahili, Spanish, French), there is a pattern where prenominal demonstratives get a more determiner-type interpretation, and postnominal demonstratives get a more pragmatically relevant interpretation. These tendencies support a view of demonstrative movement such as proposed here, where functional demonstratives will generally appear prenominally, while those with pragmatic features will appear postnominally.

4.3. Further support

The goal of the previous section has been to motivate the positing of different features, both pragmatic and functional, on demonstratives, based on evidence that demonstratives do not appear to pattern as a homogeneous group. Let us now turn to some further data from French, Cree, and Boulogne Picard which also seem to support this analysis.

4.3.1. French

First, let us examine some French data, similar to the examples already given in (26), which show demonstrative elements on both sides of the NP:

(38)

1. a. ce         garçon

   DEM-SG.MASC boy

   'this/that boy'

2. b. ce         garçon- ci

   DEM-SG.MASC boy      DEM-PROX

   'THIS boy', 'this boy here'

3. c. ce         garçon- là

   DEM-SG.MASC boy    DEM-DIST

   'THAT boy', 'that boy there'

4. d. ce         garçon intelligent- là

   DEM-SG.MASC boy    smart      DEM-DIST

   'THAT smart boy', 'that smart boy there'

Bernstein (1997) analyses the postnominal demonstrative element in Germanic and Romance as a reinforcer, which is obligatorily dependent on the prenominal demonstrative. She generates reinforcers in the head of an FP between DP and NP (recall that her FP is called DemP here, see footnote 7), with the prenominal demonstrative in [Spec, FP]. Under her analysis, reinforcers should never occur without demonstratives. This analysis does not hold for English, however, where the presence of a prenominal demonstrative is not necessary for the presence of the reinforcer (Foreman 2002); see the examples in (39), which have no overt determiner.

(39)

1. a. Bill here disagrees with you.

2. b. My friend there bet me five dollars that I couldn't get a date with you.

(Foreman 2002:2)

Foreman also argues that the elements Bernstein labels "reinforcers" do not merely reinforce the demonstratives, but rather introduce new information.²²

It is interesting to note that in French, the demonstrative elements appear on both sides of the NP. Recalling the analysis where the demonstrative is moved to check uninterpretable features, forming a chain from the tail in [Spec, DemP] to the head in [Spec, DP], the above data may be analysed as a case where there needs to be some phonological realization in both the head and the tail. French requires the Det feature to be overtly checked in [Spec, DP], but the pragmatic features in the merge position of [Spec, DemP] must also be satisfied. Therefore, a deictic element is merged into [Spec, DemP] to satisfy both phonological requirements. The tree in (40) shows the structure of the French DP in (38b).

(40) Pronunciation: both head and tail of chain, ce garçon-ci

It is particularly interesting that the head of the chain is the form which agrees with the NP, while the element in [Spec, DemP] is invariable with regards to agreement, and holds the distal or pragmatic features (see glosses in (38)).

Under the present analysis, French would have a language-particular rule which would force phonological realization of both positions of the Dem, perhaps due to some sort of necessity for overt agreement material in DP, whether it be in [Spec, DP] or D. Note that although there is a language-particular requirement for pronunciation of both copies when there are pragmatic features on the demonstrative, the overall mechanics is the same for French as for the other Romance languages we have seen here. Let us now turn to a language where, unlike French, both copies of the demonstrative are identical.

4.3.2. Plains Cree

Plains Cree frequently permits two demonstratives to be pronounced, on either side of the noun. Note the example in (41), from Wolfart (2002).

(41) kaa-sipweehteeyaahk awa  nisiimis              awa

when took off         DEM younger sister DEM

'when this little sister here of mine and I took off'

Although an analysis of the Cree DP is outside the scope of the present article, this example is mentioned for two reasons. First, Cree being one of the vocabulary sources for Michif, it is of interest to find similar data in Cree. Second, Ghomeshi et al. (to appear), based on Richards (1999), discuss the logical possibility that in addition to the options of pronouncing either the head or the tail of a chain, a further option to pronounce both the head and the tail of a chain should be possible. The example in (41) may provide support for this view, whereby Cree would be a language which permits the pronunciation of both the head and the tail of the Dem chain under certain contexts, leading to possible support for the copying analysis argued for here. Matthewson and Reinholtz (1996) and Reinholtz (2002) also claim that postnominal demonstratives in (Swampy) Cree are different from prenominal demonstratives, though their analysis is different that that proposed here.²³ As we have seen, Cyr has analysed prenominal demonstratives as functionally playing the role of definite determiners in Montagnais, a closely related language to Cree. It seems, in any case, as though the application of an analysis such as proposed in this article could be in lines with the patterns seen in the Plains Cree DP.

4.3.3. Boulogne Picard

If a pragmatically locative interpretation is only available when we have a demonstrative with the relevant pragmatic features, and if these pragmatic features trigger leftward scrambling of the NP in order for it to receive the correct prosodic stress, one may question what the role of the demonstratives without the pragmatic features is, and why they would exist alongside of definite determiners. Interestingly, Boulogne Picard is a dialect where, as Bernstein (1997) reports, the prenominal demonstrative receives a definite determiner interpretation. Recall that Bernstein (1997) argues for a substitution analysis, shown in (43) for the phrase in (42).²⁴

(42) chele école

DEM  school

'the school'

Bernstein claims that properties of the head contribute the definite interpretation (shown by the English translation). It is possible to consider this data in the light of the proposal of the present article.

If we claim that prenominal demonstratives carry primarily grammatical information rather than pragmatic information, their role can be seen as primarily definite determiners with added reference to deictic information pertaining to the speaker. In fact, the two categories would be quite similar, which, functionally speaking, may explain why they are in complementary distribution in so many languages. However, if the two elements are so close in meaning, it would be reasonable to find the contrast lost in some languages, giving rise to systems without the definite/deictic contrast. This may be the case for Boulogne Picard, which has a definite determiner interpretation for what appear to be demonstrative lexical items. It is possible that in the absence of a pragmatic contrast between prenominal demonstratives without pragmatic features and definite determiners, a language begins to not differentiate between the two, and therefore the expected interpretation would be one akin to a definite determiner.

As support for this notion, it is common to have demonstratives which develop historically into definite determiners and noun class markers (Greenberg 1978, cited by Diessel 1999:128–129). Examples of such languages include, among others, English, French, Swedish, Turkana, and Xhosa. An increase in use of demonstratives is said to correlate with the rise of phonologically distinct definite determiners (Diessel 1999, Ng 2000). An example of this is the forms of the Latin demonstrative ille which gave rise to definite articles in Romance languages (Faingold 1996, cited by Ng 2000). Ng (2000) claims that Passamaquoddy, an Algonquian language, may also be presently in such a stage of change, given the increase in frequency of demonstrative words, and the same could be said of Montagnais given Cyr's analysis discussed above. This development of definite determiners could be seen as a natural evolution from demonstratives without pragmatic features.

5. Conclusions

The goal of this article has been to investigate the patterning of demonstratives in Michif and the nature of demonstratives cross-linguistically. First, a structure for the Michif DP was proposed along the lines of the French structure argued for most recently by Bernstein (1997, 2001), though one where the Dem is base generated in [Spec, DemP] and raises into [Spec, DP] to check features. Following Bernstein (2001), who extends the use of Zubizarreta's (1998) p-movement analysis to account for different word ordering in the French DP, we explain the different positions possible for Dem in Michif. The application of p-movement is motivated based on the association of different features, functional and focus related, with prenominal and postnominal demonstratives. While all demonstratives obligatorily encode uninterpretable grammatical features, not all demonstratives also encode interpretable focus features which contribute to their interpretation as well as their pronunciation site in Michif.

While principles of language contact have not been the focus of this article,it is interesting to note that, other than agreement, the Michif DP patterns much like that of French. However, the one element of the Michif DP which is not from French vocabulary—the demonstratives—pattern differently from the demonstratives in French. This should be of interest for studies in language contact, in that perhaps syntactic features are borrowed (or intertwined) from the source language along with the lexical items. This would follow Bakker's (1997) theory of language intertwining for Michif. Though we do not defend Bakker's view that we need to posit two distinct grammars at work in Michif today, it seems likely that both Cree and French grammars have influenced synchronic Michif. This article has proposed an analysis of word order within the Michif DP based on synchronic principles found elsewhere in the DP literature, rather than historical explanations: according to this analysis,no separation of historically Cree and historically French items is necessary.