NPE , gender and the countable / mass distinction

The main purpose of this paper is to present arguments for the existence of a certain relation between formal gender features and the semantic notions ‘count noun’ and ‘mass noun’. In particular, facts relating to how NP ellipsis operates in indefinite DPs in Dutch and Afrikaans are taken to provide evidence for the proposed relation.


Introduction
This article explores a certain way of understanding the patterns emerging in Dutch and Afrikaans when NP ellipsis (NPE) applies in an indefinite singular noun phrase.NPE in definite and plural noun phrases will be left completely out of the picture here.
The motivation for concentrating on indefinite singular noun phrases is that the outcome of applying NPE in such phrases is determined by an interplay between grammatical gender and countability, suggesting a certain relation between gender and countability.

2.
The issues I begin by giving an outline of the issues I'll focus on.I will also identify some criteria that any successful analysis should meet.

Issue 1: NPE and gender change
In Dutch, adjectives modifying a singular common gender (C.SG) noun in an indefinite noun phrase have the agreement marker -e (a schwa): (1) Dutch C.SG N: een A-e N NPE may wipe out the noun leaving een A-e N (where N represents an N that NPE has applied to): (9) A-e licensed by phonology: (a) the A ends in d, g, f, s: droog "dry"  dro-ë (b) the A is polysyllabic: ewig "eternal"  ewig-e In addition, there is also an "emotiewe (emotional) -e" imposing a high degree reading on the adjective: 3 (10) A-e licensed by a high degree reading: 'n lang vergadering "a long meeting"  'n lang-e vergadering "a very long meeting" (see Combrink 1990: chapter 17;Donaldson 1993) But when NPE applies, the lowest adjective must have the suffix -e, even when it wouldn't have it when followed by an overt noun: (11) a .'n groot hond a big dog "a big dog" .b.
'n grot-e a big-E "a big one" But this -e doesn't spread to higher adjectives: (12) NPE in Afrikaans with As that don't have -e when followed by an overt N: a.
'n A-Ø N  'n A-e N . b.
'n A-Ø A-Ø N  'n A-Ø A-e N 3 In the Norwegian dialect spoken in Senja, a high degree reading of the adjective also licenses -e in indefinite singular noun phrases where -e otherwise never occurs on adjectives: (i) a. ei snill jenta (Senja Norwegian) a.F nice girl "a nice girl" b. ei snill-e jenta a.F nice-E girl "a very nice girl" When the modified noun has neuter gender, -t is always suffixed to the adjective in indefinite singular noun phrases, and this -t cannot co-occur with the "high degree -e" suggesting that both this -e and the neuter -t may spell out the X whose Spec hosts the AP: This -e also doesn't co-occur with a following een, although the -e of ( 9)-(10) does: (13) een-substitution in Afrikaans: 'n A-Ø een, except when the A requires -e by ( 9)-( 10) An account of the apparent gender change under NPE in Dutch must achieve the following: (14) Success criteria for an account of Dutch gender change: a. prevent gender change from occurring when an adjective precedes an overt N.SG noun or een "one" "substituting" for a N.SG noun b.
make gender change necessary when NPE applies to a N.SG noun in an indefinite noun phrase An account of the unexpected -e emerging under NPE in Afrikaans must meet similar criteria: (15) Success criteria for an account of the unexpected -e in Afrikaans : a. prevent this -e from occurring when an adjective precedes an overt noun or een "one" "substituting" for a noun b.
make the -e necessary when NPE applies to a noun following an adjective In addition, a successful analysis should arguably identify the force driving gender change with NPE in Dutch with the force imposing the unexpected -e on the adjective preceding the elided noun in in gender-less Afrikaans, since the unexpected -e in Afrikaans and the Dutch -e emerging under gender change have similar distributional properties: both must be directly adjacent to an adjective and only occur in conjunction with NPE.We note at the outset that the similarity between the unexpected -e in Dutch and its counterpart in Afrikaans indicates that a unified analysis is unlikely to involve actual gender change precisely because Afrikaans is gender-less. 4

Issue 2: NPE and the count/mass distinction
In Dutch, eliding an indefinite common gender mass noun gives a result that seems perfectly fine to most speakers, i.e.A-e __: (16) Jan heeft rode wijn gekocht en Marie heeft witt-e gekocht.Jan has red wine bought and Marie has white-E bought "Jan has bought red wine, and Marie bought white." (17) Dutch common gender mass noun under NPE: A-e N  A-e N 4 The following account of the unexpected -e in Afrikaans will also fail to provide a unified analysis: suppose adnominal adjectives are always followed by -e (in the head whose Spec hosts the AP), but this -e is eventually erased when followed by an overt AP or NP except when the conditions described in ( 9) and (10) are satisfied.
Then, -e would automatically surface on the final adjective under NPE.But this account cannot easily be extended to adjectives modifying neuter nouns in Dutch, since the unexpected -e in Dutch spreads to higher adjectives.Also, it doesn't capture the count/mass distinction described in the following section.

Gender and the count/mass distinction
The following outlines the analysis I will pursue.Several important questions will be left open until the following sections, including exactly how the -e on the A is involved in licensing NPE, an issue I essentially sidestep in this section.

The core of a proposal
Suppose that NPE (without post-adjectival een) in an indefinite singular noun phrase in both Dutch and Afrikaans requires the lowest adjective to be associated with Agr holding a certain feature F (spelled out by -e):6  This would be similar to Kester's (1996) proposal that English a big *(one) is due to English lacking adjectival inflection.
In From this point of view, the semantic feature [count] comes out as an interpretable counterpart of the grammatical gender feature [common gender] in a way similar to the relation between referential number and gender in Harbour's (2007) treatment of noun classes in Kiowa: 8 7 Taking NPE to be licensed by [count] on F is similar to Alexiadou and Gengel's (2012) proposal that NPE is licensed by classifiers, assuming classifiers to be [count].However, as argued in the text, this is compatible with NPE applying to common gender mass nouns only if we also assume a degree of equivalence between [ count] and the gender feature [common]. 8In Kiowa, the appearance of inverse marking on nouns and agreement inflection is conditioned by two factors: noun class and whether the noun is interpreted as a singular, a dual or a plural.Harbour accounts for this by associating the noun classes with different number features such as [+singular], [-augmented] etc.When associated with nouns (by appearing on Class, a nominal functional head), these features are uninterpretable in the sense that a noun in the class associated with [+singular] can perfectly well be interpreted as a dual or a plural.The interpretation is uniquely determined by interpretable number features on the node Num.But the distribution of inverse marking is captured by a feature matching algorithm that doesn't distinguish between interpretable and uninterpretable number features.In the next two subsections, I provide evidence from two other Germanic languages supporting the contention that common gender should be viewed as uninterpretable [count].9

Gender and count/mass in West Jutlandic
The West Jutlandic variety of Danish, like Afrikaans, doesn't have grammatical gender.But the inanimate demonstrative pronoun has two different singular forms: den "it", which matches the Standard Danish common gender form, is only used with a [count] denotation in West Jutlandic, while det "it", which matches the Standard Danish neuter, is only used with a mass denotation (Delsing 1993: 7): We can view the correspondences between Danish and West Jutlandic as the result of uninterpretable [count] ( = common gender) forms being reinterpreted as interpretable [count] when West Jutlandic lost grammatical gender. In

Gender and count/mass in Norwegian
In Norwegian, noen and noe, both roughly "some", only combine with plurals and mass nouns outside NPI contexts.Noen combines with plurals of any gender, and noe combines with mass nouns of any gender.But the inflection -en of noen matches the C.SG suffixed article on nouns, while the -e of noe corresponds to the suffixed N.SG definite article (written -et, but pronounced -e):

"Gender change" and NPE
The analysis sketched in the preceding section does not yet meet all the success criteria set up in section 2. In this section, I take a first step towards remedying this defect.

Unexpected -e and its connection to NPE
The unresolved issue I need to address is this: Why does the apparent gender change seen in ( 5) only happen in conjunction with NPE?: (5) Dutch neuter singular under NPE: That is, why is -e impossible in (40)?: (40) Dutch neuter singular with the N intact: een A(*-e) N with N = N.SG Similarly, why can a count noun not trigger the -e not licensed by phonology or a high degree reading in Afrikaans, except when it undergoes NPE?: (12) NPE in Afrikaans with As that don't have -e when followed by an overt N: a.
(13) een-substitution in Afrikaans: 'n A-Ø een except when the A requires -e by ( 9)-( 10) I'll start by looking more closely at Afrikaans.

Afrikaans ene
Afrikaans can have ene instead of een "one", both when there is a preceding A and when there is none: Notice that the ungrammaticality of (43c) tells us that the -e-in (43a-b) isn't just an epenthetic vowel inserted for morpho-phonological reasons, but must be the same unexpected -e that emerges under NPE and therefore cannot be followed by een.
Notice also that diminutives cannot be formed from een.Compare (44a) and (44b): (44) a. *'n mooi een-tjie a pretty one-DIM b. 'n been -tjie a leg/bone -DIM "a little leg/bone" Since the phonological properties of een relevant to diminutive formation are the same as those of been "leg" (Donaldson 1993: 87), (44a) (and not (43a)) would be the expected form, if een could combine with the diminutive suffix at all.Thus, I conclude that there is a special morphosyntactic element -e which licenses diminutive formation both in (43a) and (43b).That is, the unexpected -e that appears on adjectives in conjunction with NPE in Afrikaans is the same -e as the final vowel of ene. http://spil.journals.ac.za

Afrikaans ene formed by movement
To arrive at an analysis that identifies the final -e of Afrikaans ene with the unexpected -e on adjectives in Afrikaans, I suggest that both een and ene are composed of two distinct heads, as in ( 45), adapting a proposal by Corver and van Koppen ( 2011 One is to be understood as bimorphemic and in particular as 'wa+n', wa is the classifier and -n an indefinite article [footnote omitted].The necessary pronunciation of the n of one even before a consonant, as opposed to the necessary dropping of the n of an before a consonant, might just be phonology.Or it might also be related to syntax, especially if the order 'classifier -indefinite article' ('w + n) [footnote omitted] is produced by leftward movement from a structure in which the indefinite article precedes the classifier.
Endorsing the boldfaced suggestion at the end of this quote, I take en to be a classifier, but, like Corver and van Koppen (2011: 390), I diverge from Kayne by not analyzing -e as an indefinite article. 12Remaining agnostic as to its exact nature, I will label it '' to indicate a functional head in the nominal spine obligatorily merged on top of the classifier en.As indicated in (45b), ene is formed by raising en, the classifier, to the Specifier-position of -e.
I also assume that the two heads -e and en are on the main extended projection line of the noun (deviating in this respect from the treatment of the one of a black one in Kayne 2015): [

P [ClP NP [ClP en ]][P e ]]
I will also assume that NP in Spec-ClP or Spec-P cannot be pronounced, much as in a proposal by Kayne (2015: 10) for a blue one: 12 One reason for this is that the diminutive suffix -tjie does not attach to indefinite articles, but does attach to -e as shown above.Since een as analyzed in (48a) also contains e = n, we must then assume that -tjie needs to be adjacent to the morpheme spelling out , but cannot go inside the structure in (48a).In beentjie "small leg/bone" -tjie is adjacent to the morpheme spelling out , if the root been also spells out  as will be suggested at the end of section 5.1.
http://spil.journals.ac.zaThis is consistent with a more general hypothesis laid out in the following section: NPE is always the result of raising the NP to the Specifier position of a phase head. 13At this point, I would only like to point out that in any analysis, NPE must be prevented from eliding the entire extended N-projection in Dutch, Afrikaans and English, e.g.pancakes cannot be elided in I made pancakes.On the proposal adopted here, this is because NPE can only occur if the N is moved to some Spec-X within the extended N-projection in these languages.But then the head X itself will be an overt residue if syntactic constituents can only remain unpronounced in a Spec-position.Notice, though, that this does not exclude the possibility that other languages have the option of "eliding" the entire N-projection by moving it to the Spec of a phase head on the verbal spine (an option even available in Germanic in so-called "topic drop" constructions).

Movement across adjectives from Spec-e
On the assumptions just adopted, the fact that the presence of an adjective allows NPE to apply without een or ene, leads to the conclusion that adjectives make accessible a higher Spec-X and that movement to this position leads to NPE.I take this Spec-X to be higher than all adjectives.
To answer the question how the unexpected -e following the adjective is related to the e of ene, to reach the high Spec-X above the adjectives: Since a phrase in the relevant Spec-X is silent (just like a phrase in Spec-ClP), en will now be silent too, but the stranded -e will not be, and we can identify it with the unexpected -e that follows the last adjective under NPE in Afrikaans.
Taking the -e following the A in Afrikaans 'n A-e N to arise this way immediately accounts for (13): (13) een-substitution in Afrikaans: 'n A-Ø een except when the A requires -e by ( 9)-( 10) Since en moves only as part of a constituent containing N, ( 50) is accounted for too: (50) no -e when the N remains intact: *'n A-e N except when licensed by ( 9)-( 10) Thus, we both have a link between (43a) and (43b), and we can exclude (43c): http://spil.journals.ac.za (43) a.

*'n mooi-e-tjie een
In addition, we can account for the fact that adding the diminutive suffix imposes a [count] interpretation on mass nouns, by saying that e can only merge on top of the classifier en, noticing, however, that this makes it somewhat puzzling from the perspective of section 3 that the diminutive suffix also imposes neuter gender on the noun in Dutch and German.
But we still need to answer the following questions: (51) a.
Why can a noun not move to Spec-XP above the adjectives without first combining with en and -e, again yielding *'n A-Ø__?

5.
Movement and NPE at phase edges will now explore a way of providing a common answer to the questions in ( 51), based on a specific assumption about the high Spec-XP above the adjectives.The emerging analysis will then provide a link back to the discussion in section 3, which led to equating common gender with uninterpretable [count].
At a given phase level, only the head and material in the c-command domain of the head can (and must) be spelled out.b.
At a given phase level, no material within (or adjoined to) a lower phase can be spelled out.
(53) (52) is the only source of non-pronunciation (at least of elements that have a potential pronunciation).
In what follows, (53) will be as important as ( 52).This presupposes that -e and en are phase heads in ( 48): It also presupposes that the movement in (49) takes [ N [ en ]] to a phase edge, i.e. to a Spec-XP with X a phase head: http://spil.journals.ac.za The question now is what licenses this movement.Suppose this movement is triggered by a feature [count] associated with X attracting a phrase endowed with same feature.Then, question (51a) may be answered as follows: [count] is associated with en, but not with -e: (54 If nouns are not associated with [count], i.e. if [count] is only associated with classifiers, it also follows that ( 55) is illicit, and we also have an answer to question (51b): That is, *'n A-Ø __, provided the ClP headed by en needs to start out as the complement of the  e.
The fact that no mass noun undergoes NPE in Afrikaans now follows from the fact that mass nouns cannot combine with the classifier en -or rather, cease being interpreted as mass nouns, if they do.
The proposal that the feature [count] on the X above the adjectives may induce movement should not entail that it forces movement, unless either this X itself can fail to be present or X need not have the feature [count].Otherwise, Afrikaans could not have both (56a) and (56b), analyzed as in ( 57): (56) a.
'n A-e __ b.
'n A en-e Presumably, there also cannot be an X needing a phrase with the feature [count] as its Specifier in noun phrases with an overt noun following the adjective.
Instead of saying that [count] on X triggers movement, we might even say that movement to Spec-XP need not be triggered, but since movement to Spec-XP involves merging a phrase with another phrase, labeling the outcome requires that the head of the phrase merged to XP shares some feature with X, as in Chomsky (2013); and [count] might be the only option, e.g.there might not be a feature Either X selects for a ZP with an AP in its Spec, or anti-locality blocks the derivation in (59).That is, both the complement of X and the Spec of the complement of X might be too close to Spec-XP (as in Cinque 2005: 326). 15On the first option, X may be related to the head whose Spec the NP moves to acros, as in Romance and other languages according to Cinque (2005).
Returning to the claim that derivations like ( 55

Dutch
In Dutch, a neuter singular noun seems to undergo gender change when NPE applies to it in an indefinite DP (see section 1.1): ( heads between e and en and between en and N necessitating a weakening of the claim that silent elements must be in a Spec-position. 16As pointed out to me by Marjo van Koppen and a reviewer, Standard Dutch disallows een A(-e) een.The claim that (40) nevertheless is derived as in (49) then calls for an explanation why een cannot be left intact when there is a preceding A. This issue will be left unexplored here.Barbiers (2005) proposes an account for * een A(-e) een in Standard Dutch, but his proposal seems inconsistent with the fact that "een-substitution" is fine in Standard Dutch when there is no preceding A. http://spil.journals.ac.za But now we also need to make this analysis account for the fact that the unexpected -e spreads to higher As in Dutch: 17 For (60) to yield een A-e rather than *een A-e-e, one of two adjacent schwas must be elided by a phonological process; cf.Corver and van Koppen (2011). 18If overt count Ns are embedded under -e and en as proposed in section 5. From this point of view, the Ø on adjectives modifying an indefinite N.SG noun in Dutch does not reflect gender agreement, but rather absence of agreement.

Summary
Here is how the NPE patterns in indefinite singular noun phrases in Dutch and Afrikaans arise according to the analysis developed here: (III) no derivation leading to NPE: neuter mass nouns mass nouns

Conclusion
The main point that I have tried to make, is that the gender feature [common] in Dutch must be seen as the uninterpretable counterpart of the semantic feature [count].The advantage of this assumption is perhaps best illustrated by a comparison with Corver and van Koppen's (2011) proposal about the nature of the unexpected -e appearing on As modifying an elided N.SG N in Standard Dutch.Since an elided N must be licensed by overt gender/number inflection on the preceding A, and the preceding A has no overt gender/number inflection in indefinite singular noun phrases with a neuter N, Corver and van Koppen take this -e to be a pronoun, and to account for the fact that the -e spreads to all As, they add that it must be a C.SG pronoun.But this raises two questions: Why should -e necessarily be a C.SG pronoun, even when it replaces a N.SG noun and has a N.SG antecedent?In effect, this looks essentially like real gender change.By contrast, the assumption that [common] is uninterpretable [count] avoids assuming gender change as shown in section 5.2., allowing us instead to take advantage of the fact that neuter nouns too may be countable.
Nor is the unexpected -e a pronoun in the analysis proposed here.Rather, it is a functional head in the nominal spine left behind when [ClP NP [Cl P en ]] moves to Spec-XP above all As, as in section 4.5.The assumption that the X providing this Spec-position can only be merged on top of As, then accounts for the fact that -e must be preceded by an A, a fact that seems problematic for Corver and van Koppen's analysis even if -e is analyzed as a clitic pronoun, as pointed out in footnote 14.Admittedly, though, the claim that X can only merge on top of As also needs to be anchored to general principles, perhaps along the lines suggested at the end of section 5.1.
Finally, Corver and van Koppen's analysis of the unexpected -e as a C.SG pronoun doesn't explain why it cannot appear when the elided N is a neutral gender mass noun except by way of stipulating that the pronoun -e also must have the interpretable feature [count].But on the analysis proposed here, it follows automatically that -e will only emerge if merged above the classifier en, which has the interpretable feature [count] as an inherent property.
Needless to say, the rather sketchy analysis offered here comes with some loose ends which I can only hope can ultimately be tied up in a way consistent with the central ideas put forth in this article.

(
46) [P e [ClP en [NP N ]]]If head-movement inside an extended N-projection is excluded, as inCinque's (2005) account of Greenberg's Universal 20, the movement that brings the classifier en to Spec-e must therefore carry N along.I will take it that the first step in the derivation has P raising to Specen, and then either NP or [ClP NP [ClP en <NP>]] raises to Spec-P, yielding (47a-b) (with traces/copies not shown):(47) a.[P NP [P e [ClP en ]]] b.

(
48) a. een = [P NP [P e [ClP en ]]] b. ene = [P [ClP NP [ClP en ]][P e ]] 5) Dutch neuter singular under NPE: een A-Ø N  een A-e N with N = N.SG But gender change only occurs in conjunction with NPE: (40) Dutch neuter singular with the N intact: een A(*-e) N with N = N.SG This question can most easily be answered the same way as the question why Afrikaans disallows A-e preceding een or an undeleted N.That is, (5) has the same derivation as 'n A-e N in Afrikaans: 16 (49) [XP X [ A [P [ClP NP [ClP en ]] [P e ]]]]  [XP [ClP NP [ClP en ]] [XP X [ A [P e ]]]]] 15 If so, however, the movement of NP to Spec-ClP and of [ClP NP [ClP en ]] to Spec- under NPE: een A-Ø A-Ø N  een A-e A-e N At this point, we must again appeal to the assumption that common gender should be seen as uninterpretable[count]: when [ClP NP [ClP en ]] moves to the Specifier of -e (a phase head), the interpretable feature[count]  on en becomes visible to probes connected with adjectives and is treated on a par with the gender feature [common]: 17 (60) [ A [ Agr … [nP [ClP NP [ClP en ]] [nP e ]NP [ClP en ]] does not move to Spec-e, it remains invisible to higher agreement probes, since -e is a phase head: (61) [ AP [ Agr …[P e [ClP en [NP N ]]A(*-e) N with N = N.SG, since an overt count noun only appears when the NP doesn't move to Spec-P (by itself or as part of [ClP NP [ClP en ]]).Notice that again it is crucial that the N itself cannot have the interpretable feature [count]. 18With N = C.SG, the As agree with the N's gender feature (or -e, as in footnote 18): (62) [ AP [ Agr … [P NP [P e [ClP en ]A-*(e) een with N = C.SG just like een A-*(e) N with N = C.SG: 64), I assume that it is ungrammatical for the same reason as *ene without a preceding A in Dutch (leaving open what that reason is): NPE licensed by en = interpretable [count]: N.SG and C.SG count nouns: count nouns: een [XP [ClP NP [ClP en ]] [ A-e [P e ]]] 'n [XP [ClP NP [ClP en ]] [ A [P e ]]] In other words, the inflection of noen and noe is not determined by gender/number agreement with the accompanying noun, but rather by the countable/mass distinction:This would also account for the fact that -en and -e are the singular forms of the article, and that the article doesn't agree in gender with the overt noun.Saying instead that -en/-e reflects agreement with the grammatical gender of the silent noun, however, runs up against the objection that the overt counterpart of NUMBER is in fact neuter and the overt counterpart of Cinque 1988)journals.ac.za10Rather, noen by itself may denote both singularities and pluralities, as the "indefinite plural" in the Italian Bussano alla porta "They are/Someone is knocking on the door."(seeCinque1988),but can only range over human beings.http://spil.journals.ac.za Ths suggests that the silent nouns modified by noen/noen are not exactly silent counterparts of amount or number, but gender-less classifier-like nouns X and Y with meanings similar to number and amount such that X shares the interpretable feature [count] with number, but Y doesn't have this feature: (36) Norwegian: noen/noe outside NPI environments: a. no-en [ X [NP bøker]] "some books" [mass] or [-count].Correspondingly, when this account is extended to Dutch in the next subsection, there should be no feature [neuter] (seen as uninterpretable [mass] or [-count]).ClP NP [ClP en ]] [P e ]]  [XP [ClP NP [ClP en ]] X [P e ]] Starke (2009)d because the interpretable feature[count]can only be associated with the classifier en, never with N itself, we must also face the consequence that every noun interpreted as a countable noun must start out embedded under en, even when the noun does not undergo NPE.As already hinted in footnote 12, I will take it that when the NP in [P  [ClP Cl [NP N ]]] does not raise as in the derivations leading to NPE, a noun lexicalizes the whole structure in accordance withStarke (2009)and subsequent work in Nanosyntax.That is, every noun has a lexical entry of the form N  [P  [ClP Cl [NP N ]]], which allows it to replace any constituent of [P  [ClP Cl [NP N ]]], in accordance with the Superset Principle.When the NP raises, however, the resultant structure has no constituent containing  and Cl that matches a constituent of [P  [ClP Cl [NP N ]]].
But if a N is interpreted as a count N only if it combines with the classifier en (the only element that bears the interpretable feature[count]) as suggested at the end of section 5.1, (68) is not available even with common gender count nouns.
1, but does not raise to Spec-e, uninterpretable [count] corresponding to common gender must be associated with -e rather than (only) with N, since adjectives agree with overt common gender count nouns.http://spil.journals.ac.za