You are on page 1of 25
PUZZLES ON GRAMMATICAL GENDER JASPER ROODENBURG* AAFKE HULK™ ABSTRACT: This article explores the link between the notion of “neuter” gender and default in Dutch, a language that does not make the familiar distinction between masculine and feminine nouns, but opposes non-neuter nouns to neuter nouns. In trying to extend to Dutch the analysis of “neuter” as a default pronoun in Spanish and Catalan (cf. Picallo, 2005, this volume) ~ which lacks both gender and number — we raise the following two puz- zles: (i) what distinguishes default her from non-default het, (ii) why isn’t het the default choice in Child Dutch? On the basis of an analysis of vari- ous constructions involving default het, we offer a new perspective on these puzzles that show that both in adult and child language the alleged default ‘determiners’ are of a different nature than previously assumed. KEYWORDS: grammatical gender, neuter, non countability, determiner ac- quisition, Dutch 1. INTRODUCTION The material presented in this article is the result of a collaboration based on some intriguing facts from Dutch child language, showing that children — at a certain stage — have recourse to a default determiner in what seemed to make part of their way towards acquiring grammatical gender. In our analysis of these data, the concept of “neuter” plays an important role — since Dutch is a language that does not make the familiar distinction between masculine and feminine nouns, but rather opposes non-neuter nouns to neuter nouns. More- over, we discovered related, equally puzzling, facts concerning the choice of the “default” option in adult vs. child language in Dutch. The relevance of the notion of a particular “default” follows from recent work by Picallo (2005, this volume), who shows not only that grammatical Gender, among the @-features Person/Number/Gender, plays a more substan- tial role within the syntactic computation than being only responsible for con- cord phenomena. Picallo also argues that the notion of “neuter” — in a [fem] * Internationales Graduiertenkolleg Stuttgart-Paris VIII, University of Stuttgart. ** Amsterdam Center for Language and Communication (ACLC), University of Amsterdam, Parts of the material in this paper have been presented at the workshop “The Syntax and Se- mantics of Grammatical Features”, organized by the University of Stuttgart on 23 February 2007, and at the XXXII Incontro di Grammatica Generativa, that took place at the University of Bologna from 1-3 March 2007. We would like to thank the audiences for useful comments. 67 LINGUE E LINGUAGGIO VIL.1 (2008) 67-91 JASPER ROODENBURG AND AAFKE HULK language like Spanish or Catalan — represents the absence of grammatical gen. der and in some sense fulfills the role of a default. Applying the implications of Picallo’s system to a language like Dutch, we offer a new perspective on our initial puzzles and show that in both adult and child language the alleged default ‘determiners’ are of a different nature than previously assumed. 1.1 Gender within the Minimalist Program: [+fem] languages In modern Romance languages like Spanish, Catalan, Italian and French, gen- der is a binary distinction, opposing masculine and feminine nouns. At the adnominal level, this binary distinction can be marked on the noun, the deter- miner, or both (in which case one of the two markings is supposed to be the result of concord), As for the way in which gender is represented within the functional nomi- nal structure, there is no widespread consensus in the literature. In this paper we adopt, as a working hypothesis, the implementation of grammatical gender into the minimalist program as proposed by Picallo (2005, this volume). Ac- cording to Picallo, Gender heads an independent functional projection ClassP, located directly above NP. This can be seen in (1) (cf. Picallo, 2005: 5): ag ft As illustrated in a simplified way in (2) and (3), the [+Gender] feature of Class* enters into a checking relation with the [++fem] feature present on the noun (which is morphosyntactically realized in Romance languages such as Spanish and Catalan). c ECLASS ] [ny Nittem II] (2) [ce [ CLASS ] [n corbataysfemy J] 3) (DD... fe [ CLASS[+femy ] [N Ni+femj I] ---] > — la/una corbata ‘the/a-FEM tie-FEM’ The gender distinction also exists at the pronominal level. As can be seen in example (4) — taken from Catalan — the feminine noun ploma is referred to by the feminine pronoun Ja, in (5) the masculine noun groc is referred to by the masculine pronoun el. (4) Laploma negra, encara no la puc fer servir. ‘The-FEM black-FEM pen-FEM, I still cannot use it-FEM.’ (5) Noel tanquis, el calaix groc. “Don’t close it-MASC, the-MASC yellow-MASC drawer-MASC.” (Picallo, 2006: 5) However, in both Spanish and Catalan, a third kind of expression exists at the pronominal level: the so-called “neuter” pronouns, corresponding respectively 68 PUZZLES ON GRAMMATICAL GENDER to lo (in Spanish) and ho (in Catalan). These “neuter” pronouns differ from masculine and feminine ones in the sense that they do not refer to “neuter” nouns since, in Spanish and Catalan, there does not exist a third gender spec- ification “neuter” that nouns can be specified for. As stated by Picallo, nomi- nal expressions cannot serve as linguistic antecedents of “neuter” pronominal forms. E.g. the masculine and feminine indefinite noun phrases in (6) and (7) respectively cannot be the linguistic antecedents of neuter /o in Catalan. (6) Quan un venedor té una calaixera; la; /el,;/; /ho,ijn, Ven. “When a seller-MASc has a drawer chest-FEM he sells it-FEM.SG / it-MASC.SG / it-NEUTER ” (7) Quan un venedora té un armari; lai; / el; /ho,iyn Ven. “When a seller-MASC has a closet-MASC he sells it-FEM.SG /it-MASC.SG /it-NEUTER ” (Picallo 2005) On the contrary, “neuter” pronouns are typically used to resume e.g. nominal predicates. This is illustrated in (8) and (9) for both languages. (8) En Joan és el cap de la seccid. Ho és des de fa temps. ‘John is the-MASC head-MASC of the department. He has been it-NEUTER for a long time.’ (9) Juan Carlos es rey de Espafia. Lo es desde hace tiempo. ‘Juan Carlos is king of Spain. He has been it-NEUTER for a long time.’ (Quer, p.c.) The “neuter” pronouns /o and ho do not resume any of the phi-features gen- der/number/person: postcopular adjectives and (pro)nominal predicates can be prototypical expressions of properties, which lack these p-features (cf. Lon- gobardi, to appear). In a similar vein, Picallo argues that “neuter” pronouns in Spanish and Catalan do not correspond to a ‘third’ gender category. On the contrary, the “neuter” pronoun is chosen when gender is arguably lacking. In this sense, one could assume that “neuter” pronouns are default forms, which are inserted in situations where gender is arguably absent, e.g. in the case of predicates. Thus, Spanish and Catalan “neuter” pronouns can be considered as default forms.! "As for the exact feature specification of these neuter (default) elements, we leave open the question of how they are specified exactly — in particular whether they lack features, or whether they are underspecified or negatively specified for them. At a structural level, Picallo proposes the derivation given in (i) and (ii), where it can be seen that neuter pronouns like fo and ho lack gender (and number). (i) LE... fe [ CLASS] LN Nifemj{num] J] (i) Fo. fe CLASSI IN Negemjinam| I] > fo/ho “itNEUTER’ 69 ‘ JASPER ROODENBURG AND AAFKE HULK, The hypothesis that “neuter” in Spanish and Catalan is the absence of gen. der, and (as a consequence) that “neuter” may correspond to the default (form), is an interesting idea since it raises several questions once that idea is embed. ded within a wider context. For instance, is the link between “neuter” ang default restricted to the two Romance languages cited, or does it extend to other languages where “neuter” is (traditionally considered as making) part of the gender system, like Dutch? Before examining this question, we will take closer look at the gender system in Dutch. 2. THE GENDER SYSTEM IN DUTCH: PRELIMINARY REMARKS Given the particular status of neuter within Picallo’s system, the question arises as to how the situation is in a language like Dutch: the traditional assumption about neuter in Dutch is that it is part of a grammatical gender system in which a binary distinction is made between non-neuter and neuter nouns. This raises the question whether “neuter” has a default status in Dutch as well. As we will see, this question is not trivial, in particular because the Dutch gender system turns out to be less homogeneous than is often assumed. In the next two sub-sections, we will briefly sketch how gender is expressed in Dutch by distinguishing the adnominal domain (2.1) from the pronominal domain (2.2) 2.1 Gender in the adnominal domain The Dutch gender system differs from that in Romance languages such as Spanish and Italian. Most strikingly, Dutch does not oppose masculine and feminine nouns. Instead, it distinguishes nouns introduced by the determiner de, called non-neuters, from nouns introduced by the determiner her, called neuters. Furthermore, the presence of gender is not very prominent within the nominal system. Firstly, gender distinctions are not morphologically ex- pressed on the noun. One (potential) exception is constituted by diminutives: nouns endowed with the diminutive suffix are always neuter, independently of the gender at the noun itself.? Secondly — and more importantly — the actual situation represents a stage of a gender system that gradually has lost an im- portant part of its visibility. While Dutch used to make a threeway distinction Anticipating the discussion below, the lack of gender and number often go together. For instance this is also the case for other arguments that have been argued to be genderless, like nominalized CPs in Spanish and Catalan as given in (iii). Nominalized CPs are e.g. resumed by neuter ho even when they appear inside a coordination. Gi) Sabem (que van bombardejar la ciutat] i [que van saquejar els magatzems] perqu® lo! els portava el diari d’ahir. “[I] know [that the city was bombarded] and [that the warehouses were looted] because yesterday's newspaper told it-NEUTER *them-MASC.PL.’ (Catalan, Picallo 2006) ? A detailed analysis of the diminutive is beyond the scope of the current paper. 70 PUZZLES ON GRAMMATICAL GENDER between masculine, feminine and neuter nouns, the distinction between mas- culine and feminine has totally disappeared in modern Dutch. What is left is a distinction between neuters and non-neuters. However, this distinction is only visible in a limited way. In fact, the neuter / non-neuter distinction is only vis- ible in the singular, while within the singular the distinction can only be seen with definite noun phrases. Let us illustrate the Dutch gender system with some examples. With sin- gular definites, the gender on the noun is reflected on the form that takes the definite article: non-neuter nouns (both mass and count nouns) are introduced by non-neuter de, while neuter nouns are introduced by neuter her. With singular indefinites, the gender distinction is not visible: all count nouns (both non-neuter and neuter) are introduced by the indefinite determiner een, while all mass nouns are bare. With plural nouns, gender is never visible either: all definite plural nouns: are introduced by de, while all indefinite plurals are ‘bare’. This situation just described is represented in Table 1. Singular Definites Non-neuter | de auto “the-NONNEUTER car” de wijn ‘the-NONNEUTER wine” Neuter het huis “the-NEUTER house” ‘het water ‘the-NEUTER water” ‘Singular Non Definites Non-neuter | een auto ‘acar’ @ wijn “@ wine’ Neuter een huis ‘a house” Owater —_‘@ water’ Plural Definites Non-neuter | de auto’s ‘the cars” Neuter de huizen ‘the houses” Plural Non Definites Non-neuter | @ auto's ‘@ cars’ Neuter @huizen —_“O houses’ Table 1: The Dutch gender system In the remainder of this paper, we will concentrate on singular nouns, es- pecially on the contrast between non-neuter de and neuter het. 3 Note that the gender distinction is partly reflected on adjectives (by means of the presence vs. absence of inflection), but since this involves not only gender but also other factors, this point will be left aside in the present paper. 1 JASPER ROODENBURG AND AAFKE HULK 2.2 Gender in the pronominal domain Ata pronominal level, gender plays a role as well in Dutch, although (as we will see) there are a number of mismatches. Nouns referring to non humans are referred to by means of a grammatical strategy. Non-neuter nouns like de auto are referred to with the pronominal form hij (literally ‘he’), while non- neuter nouns like het huis are resumed by pronominal her.’ Plural nouns are all referred back to by ze, independently of their gender (cf. Table 2). Singular Non-neuter | Waar is de auto? “Where is the-NONNEUTER car?” Hij staat buiten. ‘He-NONNEUTER is outside.” Neuter Waar is het boek? “Where is the-NEUTER book?’ Het ligt op tafel. “It-NEUTER is on the table.” (i Plural Non-neuter | Waar zijn de auto’s? ‘Where are the-NONNEUTER cars? Ze staan buiten. “They are outside.” Neuter Waar zijn de boeken? “Where are the-NEUTER books?” Ze liggen op tafel. “They are on the table.” Table 2: Gender in the Dutch pronominal system With nouns referring to humans, the grammatical gender can be “overruled”; that is, a grammatically neuter noun that is referring to a human, like het meisje (‘the girl’), cannot only be resumed by (the grammatically expected) neuter pronoun het, but also by the the feminine pronoun zij (‘she’). (10) Waar is het meisje gebleven? Zij/het is naar huis gegaan. “Where did the-NEUTER girl go to? She-FEM/it-NEUTER went home.’ In this case there is a “mismatch” between grammatical gender and natural gender. This “mismatch” with nouns referring to humans also shows up in other languages where pronominal reference is normally based on grammatical gender, Interestingly, in Dutch a comparable mismatch exists with nouns referring to non-humans. We return to this point in section 5.3, where we indicate to what extent this is relevant for the current discussion. When used as a subject referring to non humans, the strong form hij is sometimes replaced by the weak form ie (cf. Cardinaletti & Starke, 1994), However, ie cannot be used in preverbal subject position. With objects, non-neuters are referred to with hem /‘m, neuters with het. 72 = PUZZLES ON GRAMMATICAL GENDER 3. PUZZLE I: (ADULT) DUTCH AND THE DEFAULT STATUS OF NEUTER HET If we apply Picallo’s claim that in Spanish and Catalan “neuter” is the absence of gender to “neuter” nouns in Dutch, the question arises whether this hypoth- esis is feasible for a language like Dutch. That is, does “neuter” het in Dutch correspond to a default form as well? The answer to this question is not sim- ple, although it has received a positive answer in the literature — albeit stated in different terms (cf. Rooryck, 2003; Kester, 1996). As we will see, analysing het as a default form has consequences for various issues within the Dutch nominal domain. Firstly, it has direct consequences for the categorial status of “neuter” het (is it an adnominal element or a pronoun?), secondly it raises direct questions about the nature of its grammatical features and the order in which they are acquired. Both issues are dealt with by means of a puzzle. We will postpone the dis- cussion of the second puzzle until section 4. The first puzzle — which is raised in the current section — follows from the examination of the question as to what extent the notion of “neuter” is connected to the notion of default in Dutch. In the next two subsections, we will therefore closely examine the status of het in constructions in which gender (and eventually other @-features) can be shown to be absent on independent grounds. We will oppose instances of pronominal het (section 3.1) to instances of (apparent) adnominal het (section 3.2). 3.1 Pronominal het ‘As we have seen in section 2, “neuter” het can both function as an adnominal element (a determiner) and as a pronominal element. Let us start by con- fronting the notion of default to pronominal het. 3.1.1 “Dummy” het in impersonal constructions Neuter het is used in impersonal constructions of the type Het is leuk dat je dat zegt (‘It-NEUTER is nice that you are saying that’). Here, impersonal het clearly functions as a “dummy” pronoun. Although this dummy pronoun cor- responds to “neuter” het, it is not specified for gender. The examples in (11) and (12) show that her is blind to the gender specification of its nominal corre- late: het shows up both with a non-neuter noun like man (‘man’) and a neuter noun like huis (‘house’). (11) Heris een lange man. “It-NEUTER is a long man-NONNEUTER’ (12) Her is een groot huis. STLNEUTER is a big house-NEUTER.’ B JASPER ROODENBURG AND AAFKE HULK. 3.1.2 Predicative het Compared to the above, neuter her is the only available pronoun to resume nominal predicates, irrespective of the (lexical) gender specification involved (13) Juan-Carlos is (de) koning van Spanje. Hij is het / *hem / *de al sinds lange tijd. “Juan-Carlos is (the)-NONNEUTER king of Spain. He has been it-NEUTE, /*him / *DE since a long time.” (14) Jan is (het) hoofd van de afddeling. Hij is het /*hem al sinds lange tijd. “Jan is (the)-NEUTER head of the department. He has been it-NEUTER / *him for a long time.’ Since nominal predicates typically lack @-features (cf. the Spanish/Catalan examples given in (8) and (9) above), we have another instance of pronominal het inserted precisely where gender is lacking. In this sense, in examples such as (13) and (14) neuter het fulfils the same role as neuter /o and ho in Spanish and Catalan: namely, het corresponds to the default. 3.2 “Adnominal” het Interestingly, the default use of neuter het in Dutch is not restricted to the pronominal domain. Comparable evidence that neuter het can function as a default is found as well in cases that — at least traditionally — have been anal- ysed as “adnominal” instances of het (that is of “determiner” her). 3.2.1 Nominalized infinitives As is well known, Dutch allows for so-called “nominalized infinitives”. These nominalized infinitives have a sort of “in between” status between verbal and nominal elements. On the one hand they have verbal properties, which is e.g expressed by the fact that they can take syntactic arguments, On the other hand, they clearly have nominal properties, which e.g. is expressed ‘by the fact that they can be introduced by adjectives and other adnominal elements However, when nominalized infinitives are not used ‘bare’ but introduced by a determiner, the only legitimate form is neuter het. As can be seen in (16), the non-neuter determiner de cannot introduce an infinitive. (15) Door het urenlange knippen van de heg zaten mijn handen vol blaren. “Due to the-NEUTER for-hours-cut-INF of the hedge my hands were full of blisters.” (16) *De knippen van de heg. “The-NONNEUTER cut-INF of the hedge.’ 5 Anticipating the discussion below, “adnominal” must be understood as a descriptive label. 74 PUZZLES ON GRAMMATICAL GENDER Since it seems uncontroversial that nominalized infinitives lack gender (e.g. they are derived from the genderless category V), it is likely that het shows up precisely for this reason: because gender is lacking, het is inserted as a default form® 3.2.2 Predicative superlatives The behaviour of superlatives is a further interesting case. In Dutch, superla- tives are introduced by a definite determiner just as is the case e.g. in languages like French. When used attributively, the form of this determiner depends on the gender of the noun. When the attributive superlative is modifying a non- neuter noun like auto (‘car’), as in (17), the determiner used is non-neuter de. In the case of a neuter noun like huis (‘house’), the only possible determiner is neuter het, as shown in (18). (17) De mooiste auto. “The-NONNEUTER most-beautiful car.’ (18) Het mooiste huis. “The-NEUTER most-beautiful house.’ However, when the superlative is used predicatively — as in constructions of the type the red car is the most beautiful ~ all of the sudden the form of the de- terminer is blind to the lexical gender of the noun. As is illustrated in examples (19) and (20), predicative superlatives can only contain “neuter” het. (19) De rode auto is het mooist. ‘The-NONNEUTER red car is the-NEUTER most-beautiful’ (20) Het linker huis is het mooist. ‘The-NEUTER left house is the-NEUTER most-beautiful’ This situation is completely in line with the other cases discussed above. Since predicates in Dutch lack @-features — and so do not have a gender specification —it is natural to hypothesise that “neuter” het in predicative superlatives is the result of insertion by default. 3.2.3 “Abstract” de-adjectival noun constructions Finally, a special instance of “neuter” het shows up with a particular kind of “de-adjectival” nouns, namely so-called “abstract” de-adjectival nouns of the type het interessante (literally ‘the interesting’, which in English corresponds © As we will see in section 5.1 below, nominalized infinitives do not only lack gender, but are also non-specified with respect to another @-feature, namely number, 75 JASPER ROODENBURG AND AAFKE HULK to the interesting thing).’ In Dutch, where this construction is very productive, abstract de-adjectival nouns can only be combined with neuter het, as can be seen in (21). (21) het gekke / het interessante / etc. ‘the-NEUTER weird [thing] / the-NEUTER interesting [thing] / etc.” Without further explanation at this stage, we consider “neuter” her in this con- struction as another example of default her. As will become clear in the dis- cussion below, the reason why “neuter” her is taken here as the default differs slightly from the other cases discussed. However, it will turn out that it consti- tutes a crucial step towards a better understanding of the reasons why het can be chosen as a default form. 3.3 Recapitulation In the preceding sub-sections, we have argued that “neuter” her can function as a default element in (adult) Dutch: (22) “Neuter” het can function as a default element in (adult) Dutch. The situation described is thus compatible with the hypothesis of Picallo (2005, this volume) on the status of “neuter” for Spanish / Catalan. Just like the “neuter” pronouns in these languages — which are genderless and can be used by default — “neuter” her in Dutch has a default status as well. However, there seems to be an important difference between Dutch and Spanish / Catalan: in Dutch, the link between “neuter” and default does not seem to be limited to the pronominal domain. The properties described in section 3.2 show that “neuter” het can be used as a default as well in constructions where its use seems to be adnominal. However, the latter point raises a question in the light of the fact that, at the same time, her introduces neuter (definite) nouns as opposed to non-neuter (definite) nouns, that are introduced by de. Because traditionally her is consid- ered to be part of a binary gender system (cf. section 2), the fact that neuter het can be used as well as a default form leads to a paradox which is captured in the following puzzle, which we will call Puzzle I: (23) Puzzle I: ‘What is the common characteristic of the instances of default het in adult Dutch? To what extent is default her related to non-default het? ‘Abstract” de-adjectival noun constructions must be distinguished from “human” de- adjectival noun constructions like de rijke (‘the rich (person)'). The latter behave differently in several respects. We come back to this contrast in section 5.2.1 16 PUZZLES ON GRAMMATICAL GENDER This puzzle makes explicit what has already been suggested in the previous section: next to its well known use described in section 2, het can also be a default form which occurs precisely where gender (and perhaps also other fea- tures) is absent. In other words, “neuter” het seems less exclusively a member of a binary gender system than has traditionally been assumed. As we will see shortly, there is no simple answer to Puzzle I: it requires a closer look at the categorial status of het and its feature composition. Before turning to this, however, we take a closer look at the way in which the Dutch gender system is acquired by Dutch LI learners. As we will see in the next section, this immediately raises a second puzzle. 4. PUZZLE II: CHILD DUTCH AND THE DEFAULT STATUS OF NON-NEUTER As has been described above, Dutch differs from the discussed Romance lan- guages by not making a distinction between masculine and feminine nouns, but by opposing non-neuter and neuter nouns. However, the gender system of Dutch seems to differ more profoundly from these different binary gender sys- tems than might be thought at first sight: it is striking as to how the acquisition of the “gender” system of the definite determiner in Dutch differs with respect to other languages that make a binary distinction. This follows clearly from a comparison with a Romance language such as French. Firstly, the acquisition of neuter gender morphology on definite determin- ers in Dutch is a slow process, with a high “error rate” and is not completed before age 6, at the earliest. A striking point ~ made by Van der Velde (2003, 2004) and others (e.g., De Houwer & Gillis, 1998) — concerns the observation that the nature of the errors made is constrained. That is, it seems that the er- rors concerning the choice of the determiner made by Dutch children only go into one direction: they initially use the definite determiner de both with neuter and non-neuter singular nouns, but hardly ever the other way around. Van der Velde (2003, 2004), among others, studied the acquisition of gen- der morphology in definite articles by Dutch children in the same age group, between 3 and 6 years old, in a cross-sectional experiment. The children were expected to produce article + noun sequences. Van der Velde’s results show that these children tend to overgeneralize the non-neuter definite article de: they used de where neuter het is expected, until at least age 6. The children made errors of the type given in (24), where de is used instead of het (vliegtuig ‘airplane’ is a neuter noun in Dutch): (24) de vliegtuig Abel 2;10.0 ‘the-NONNEUTER plane-NEUTER’ 8 The properties described in this section also set apart Dutch from a Germanic language like German. This difference will not be discussed in the present paper. 7 JASPER ROODENBURG AND AAFKE HULK However, although one finds mistakes of the type given in (24), mistakes as given in (25) (where her would be used by error to introduce a non-neuter noun like appel ‘apple’) are hardly ever found at this stage of acquisition. (25) het appel (Not found) ‘the-NEUTER apple-NONNEUTER’ Dutch children appear then to use the non-neuter determiner de as a default. Indeed, Van der Velde (2003, 2004) argues that de should be analysed as the default or elsewhere form for Gender of the definite determiner in Dutch which may explain why it is acquired first, before the more specific form het, which comes in much later. In this respect, Dutch crucially differs from other languages, such as French, where the acquisition of gender morphology is fast and without errors: Van der Velde (2003, 2004) shows that French children not only do not start with a de- fault choice for the gender of the definite determiner: they do not make gender errors at all. This corresponds to what has been found/mentioned by others in the literature (Clark, 1985), namely that French monolingual children make hardly any gender errors. ‘The acquisition data just described are puzzling in the light of the present discussion on the status of neuter het. What makes the Dutch gender system so difficult to acquire (compared to the other languages mentioned)? Why is it the case that non-neuter de seems to be used as a default at a certain stage, contrary to neuter her which — as we have seen in sections 3.1 and 3.2 — is able to fulfill a default role in adult Dutch? This situation is captured in the following puzzle: (26) Puzzle II: Why is neuter het not chosen as a default in Child Dutch? Is there a link with the fact that het in Adult Dutch is used both as a default and as a non-default form? (Cf. Puzzle I in.(23)) In the remainder of this paper we try to shed light on both Puzzles I and Il, given in (23) and (26) respectively. In particular, we show that formulating an answer requires a closer look at the feature composition of the so-called neuter and non-neuter determiners in Dutch which goes beyond gender. The discussion will focus on neuter het. ° Het is more specific in the sense that it only shows up with singular definite nouns, and never with plurals. De, on the other hand, is also used with all definite plural nouns (irrespective of their gender). 78 PUZZLES ON GRAMMATICAL GENDER 5. REFLECTIONS ON PUZZLE I: DEFAULT HET AND NON-COUNTABILITY So far, we have tried to show that there is a connection in Dutch between the notions of “neuter” and default. For instance, we have seen that neuter het is used in constructions where @-features are lacking, like impersonal construc- tions (“dummy” her), nominal predicates, predicative superlatives and abstract de-adjectival noun constructions. Given the fact that her is traditionally an- alyzed as a “neuter” form, it is natural to examine the question whether het can be analyzed along the lines of Picallo’s analysis of “neuter” pronouns in Spanish and Catalan. This is in fact what we have suggested in section 3. How- ever, if the analysis of “neuter” pronouns in Spanish and Catalan can indeed be extended to Dutch, the parallelism is expected to go beyond gender: as we briefly mentioned in footnote 1, the lack of gender often goes together with the lack of number. We will explore this question in more detail in the following sub-sections, by focussing on Puzzle I. 5.1 Lack of Number An important observation in Picallo’s analysis of “neuter” expressions in Span- ish and Catalan, is that the lack of gender often goes together with the lack of number. This is e.g. the case with nominalized CPs in Spanish and Catalan, which are also analyzed as “neuter” expressions (cf. footnote | above, ex. (iii)) That nominalized CPs lack both gender and number clearly follows from their behaviour when used inside a coordination structure. As is shown in (27), co- ordinated nominalized CPs that are used in subject position trigger singular agreement on the verb.!° (27) [[Que detinguessin en Lluis] i [que escorllessin la casad’en Pere]] va /*van tenir lloc simultaniament / amb poques hores de diferencia. “[[That Llufs had been arrested] and [that Pere’s house had been searched] took-SG / *took-PL place simultaneously / within a few hours of dif- ference. The lack of gender and number is a property that characterizes “neuter” ex- pressions like nominalized CPs in Spanish and Catalan including the “neuter” pronouns /o and ho themselves. This has already been shown in example (iii) of footnote 1, where it is shown that the same CP coordinations can only be resumed by a “neuter” pronoun, and not by a pronoun bearing plural mor- phology. 10 At least, this corresponds to the standard (default) pattern. See Quer (this volume) and Zamparelli (this volume) for the claim that there are some instances of plural agreement that can receive an alternative explanation. 79 JASPER ROODENBURG AND AAFKE HULK, Indeed, the analysis of default her involves more factors as well. A com. parable situation is found in a related construction already discussed above, namely nominalized infinitives in Dutch. We have seen in section 3.2.1 that nominalized infinitives can only be introduced by default het. Interestingly, nominalized infinitives show the same agreement pattern as nominalized CPs in Spanish and Catalan: as is illustrated by (28), nominalized infinitives present singular agreement on the verb even when used as a coordinated subject. (28) [{Het lopen] en [het fietsen}] was / “waren vermoeiend. *[The-NEUTER walk-INF] and [the-NEUTER bike-INF]] was / *were exhausting, This fact shows then that default het in Dutch can be combined with an expres- sion that not only lacks gender, but also lacks number. This might be taken as a suggestion that default her cannot contain these features itself, or cannot be positively specified for them, since otherwise a feature clash is expected to obtain.!! In the light of what follows (in particular the acquisition data discussed in section 6), the observation that het cannot be combined with a plural expres- sion is important. Namely, it allows us to claim that default her introduces a non-countable element: (29) Default het introduces a non-countable element. At this point the behaviour of default her differs, however, from that of nor- default het. For instance coordinated neuter nouns introduced by non-default het trigger plural instead of singular agreement when contained inside a coor- dination: (30) [[Het huis] en [het stadion}] *wordt / worden gesloopt. *[The-NEUTER house] and [the-NEUTER stadium]] *will-sG / will-PL be demolished. 'l We leave open at this stage whether het lacks gender and number features altogether ot whether it is underspecified with respect to these features. 2 This situation is more complicated, because het can also introduce non-countable (ass) nouns, cases which are traditionally considered as involving non-default het (see e.g, below) However, the presence of het does not necessarily influence the agreement pattern of mass nouts that are used as coordinated subjects. For instance, (some) mass nouns which are used ‘bare’ & coordinated subjects do trigger (default) singular agreement on the verb: (i) Erlag/ Magen zand en stof op tafel. “There was / *were sand and dust on the table.’ {As can be seen in (i), the (default) singular agreement pattern remains unchanged in a defini context, with the mass nouns being introduced by her: (ii) Het zand en (het) stof was / "waren nog steeds niet opgezogen. “The-NEUTER and (the-NEUTER ) was / *were not picked up yet.’ ‘The behaviour of mass nouns requires further study. See also Zamparelli (this volume). 80 PUZZLES ON GRAMMATICAL GENDER This shows that non-default het, contrary to default her, can be combined as well with nouns that are countable. G1) Non-default het can introduce a countable element. The contrast between (29) and (31) — which crucially involves the notion of countability next to the notion of gender — raises the question as to how it can be accounted for (cf. Puzzle 1). We see two possibilitie + There are two instances of het: non-default he and default her. * There is only one instance of het: default het In the first option, both instances of her have the same form, but they proba- bly have a different feature specification, given that they combine with non- countable and countable nouns respectively. The second option — according to which non-default her is in fact the same element as default her — is more challenging. According to this second view, het would be a genderless, non- countable element across the board. Adopting this hypothesis implies that another solution must be found for the fact that he7 in its non-default use be- haves differently with respect to countability (and plurality) than het in its non-default use. A possible solution could be to assume that the behaviour of. non-default het is in fact due to other, independent, factors which “overrule” the default pattern. In the current paper, we will not choose between one of these hypotheses. However, in what follows we explore the second, more challenging option and try to argue why it might be on the right track. Our arguments at this point fall into two parts. In the next section, we continue our exploration of the idea that the core characteristic of (default) her is its non-countability, Subsequently, in section 4.3, we examine the generality of the claim that het is a non-countable element by having a closer look at pronominal het. 5.2 Het as a non-countable element In order to get grip on Puzzle I, we are exploring the idea that (default) het is a non-countable element, an idea that we tentatively assume to characterize all instances of het. In the next sub-section we will have a closer look at one of the default uses of het discussed in section 3.2, namely het showing up in “abstract” de-adjectival noun constructions. 1 The status of het inside “abstract” de-adjectival noun constructions As we have seen at the beginning of this article, Spanish has so-called “neuter” expressions (such as Jo), which - by hypothesis — are both genderless and non-countable. Interestingly, the neuter expression lo shows up precisely in 81 1D AAFKE HULK “abstract” de-adjectival noun constructions — which are productive in Spanish to the same extent as in Dutch (cf. section 3.2.3): (32) [Lo bueno} y [lo malo}] dela pelicula. “[[LO good [thing] and [LO bad [thing]] of the film.’ However, as mentioned by Kester (1996: 253), /o is not an ordinary determiner in this construction: Spanish does not have “neuter” nouns that constitute a third gender category next to masculine and feminine nouns. It would thus be an unnatural move to claim that /o in (32) is an (ad hoc) “neuter” deter- miner: it would be mysterious why this neuter “determiner” only shows up with “abstract” de-adjectival nouns. Alternatively, one could assume that lo is simply the “neuter” pronoun here as well. According to this view, “neuter” lo in (32) is the same /o that we have already seen in section 1.1, which is both genderless and non-countable. However, it does not automatically follow that “neuter” /o in (32) is indeed non-countable. Interestingly, the latter point can be further explored on the basis of Dutch. For this reason, we turn to “neuter” het in the equivalent construction in Dutch. The “abstract” de-adjectival nouns introduced by default her given in (21) above and repeated here in (33), are at first sight involving “adnominal” het. (33) het gekke / het interessante / etc. “the-NEUTER weird [thing] / the-NEUTER interesting [thing] / etc, However, the use of her in this construction is particular in two respects. Firstly, it is not so clear upon closer examination that het is really “adnominal” in this type of construction. This becomes clearer once we bear in mind that some de-adjectival nouns can be preceded by a pronoun. Longobardi (to appear) e.g. discusses “human” de-adjectival nouns introduced by a pronoun. ‘These cases are analyzed as constructions involving a pronoun that is inserted in the D® head. (34) [We rich] are neglecting certain issues. (35) [Noi ricchi] stiamo trascurando certi problemi. As illustrated in (36) and (37), the same construction exists in Dutch as well. (36) [Wij rijken] kunnen daarvan leren. “We rich can learn from that.’ (37) [Wij gekken] zaten er al vanaf 8 uur ‘s morgens. “We silly [persons] were already there from 8 oclock in the morning. ‘The de-adjectival noun construction is interesting, since it turns out to reveal something about the feature composition of (default) het (which ulti- mately can explain with which expressions het can combine). ‘This becomes 82 PUZZLES ON GRAMMATICAL GENDER clear when “abstract” de-adjectival noun constructions like those given in (33) — which do not exist in English (‘the weird *(thing)’) — are compared to “hu- man’” de-adjectival noun constructions like the rich (which exist in both Dutch and in English). In English, the latter have a striking restriction: contrary to ordinary plural and mass nouns in this language, “human” de-adjectival nouns cannot be used “bare”: (38) Students are neglecting certain issues. (39) Rich pro are neglecting certain issu (40) The rich pro are neglecting certain issues. This restriction is well known in the literature. The ungrammaticality of (39) has sometimes been accounted for in terms of a feature clash. More precisely, between the feature composition of the (empty) determiner and that of the empty nominal pro, that is present in this construction according to various linguists (see e.g. Kester, 1996). As is known from Rizzi (1986), pro in English can only have a default specification, in which case it corresponds to [+human, +generic, +plural]. As for the empty determiner in D° — which introduces “bare nouns” according to Longobardi 1994 — it has a default interpretation (expressed in terms of fea- tures) as well: according to Longobardi, the empty D° is specified by default as [+mass]. On the basis of these hypotheses, Kester (1996) argues that the ungram- maticality of (39) is the result of a feature clash: the default [+mass] feature of the empty D° is in conflict with the default [+human] and [+plural] features of (human) pro. On the other hand, (40) is grammatical, according to Kester, be- cause the determiner the in English is able to specify the D° head as [+count], which renders it compatible with the feature specification of pro. If we turn now to Dutch, an intriguing situation shows up, which teaches us more about the nature of (default) het. To a certain extent the “abstract” de- adjectival noun construction can be analyzed in the same way as the “human one”, although there is an important difference: “abstract” pro is specified as [-count] (see Kester, 1996). Strikingly, in Dutch — where ‘bare nouns’ exist to the same extent as in English — “abstract” de-adjectival nouns cannot be used ‘bare’ either, as shown by the contrast between (41) and (42): (41) *Nuitige pro kwam pas daama “Useful [thing] pro came only after-that. (42) Het nuttige pro kwam pas daarna. “The useful pro [thing] came only after that. This contrast, however, is a problem for the hypothesis put forward by Kester. Given the [-count] specification of pro the ungrammaticality of (41) is surpris- 83. haa JASPER ROODENBURG AND AAFKE HULK ing, because no feature clash is expected to exist between default D°, specified as [+mass], and default pro, specified as [-count]. However ~ and this is rel- evant in the present discussion — the fact that (42) is grammatical after the insertion of her is all the more surprising i the light of the idea that the fea- ture specification of the English determiner the corresponds to [+count]. This clearly shows that it might very well be the case that her that shows up in this construction corresponds to default non-countable het (and not to an instance of non-default countable her). Otherwise, it would have lead to a feature clash with [-count] pro. This idea, that het in the “abstract” de-adjectival construction is an instance of non-countable her, is perfectly in line with the (tentative) suggestion put forward in the previous section that het might be non-countable across the board. The question that remains to be answered now concerns the categorial sta- tus of het. It is very tentative to assume that her in this construction corresponds to pronominal het. This idea would also be compatible with Longobardi’s hy- pothesis of “human” de-adjectival nouns given above. A natural move would be then to hypothesize that the “abstract” de-adjectival noun construction in- volves a structure reminiscent of the ones given in (34) to (37) (cf. Longobardi, to appear): (43) [Het gekke] kwam pas daarna. ‘HET silly [thing] came only after-that.’ Within this structure, default het in (43) would be a pronoun residing in D°. On this hypothesis, no feature clash is supposed to obtain with the [-count] feature associated with pro. Summarizing, it can be said that the analysis put forward in the present sec- tion has shed some light on Puzzle I: default her is in fact pronominal het which is both genderless and non-countable.'3 This hypothesis raises the question, however, as to how general it is. In the light of the challenging idea pursued in the present article, namely that there is only one het — default het — it would be expected that the non-countable character of her is more general. We turn to this issue in the next section, where we have a closer look at an intriguing property of pronominal het referring to mass nouns. 5.3 Het as a non-countable pronoun In the previous section we have suggested that an important property of (de- fault) het concerns its non-countable character. More precisely, we have iso- 'S Given that default her in the “abstract” de-adjectival construction is analyzed as a pronoun, ‘one would expect this analysis to be extended to the other constructions involving default het, that is to nominalized infinitives and predicative superlatives. It will be left to future research how these constructions are to be analyzed. " 84 GRAMMATICAL GENDER. lated an interesting connection between non-countability and pronominal her occurring inside “abstract” de-adjectival noun constructions. In the current section, we will see that this connection is not an isolated fact of default pronominal her. In this respect, it strengthens the tentative hypothesis pursued here that her is a non-countable element across the board. This becomes par- ticularly clear once we take a closer look at pronominal het in spoken Dutch. At the end of section 2, we have mentioned that pronominal reference in Dutch is usually guaranteed by means of a grammatical strategy: that is, anaphoric reference is dependent on grammatical gender specification. The ex- ception we have mentioned in this respect — which is found in other languages as well — concerns nouns referring to humans. With nouns referring to humans, grammatical gender can be “overruled” by natural gender: in that case there is a “mismatch” between grammatical and natural gender (see (10) above). In- terestingly, there exists another kind of “mismatch” at the pronominal level in Dutch. This mismatch shows up with nouns referring to non-countable objects. Let us make this situation more concrete. As has been mentioned in section 2, mass nouns in Dutch can be either non-neuter or neuter: (44) de wijn, de puree, de sneeuw, . ‘the-NONNEUTER wine, the-NONNEUTER puree, the-NONNEUTER snow” (45) het water, het zand, het goud, . ‘the-NEUTER water, the-NEUTER sand, the-NEUTER gold’ In (44) some examples are given of non-neuter mass nouns introduced by the definite determiner de, while (45) gives examples of neuter mass nouns intro- duced by het. Given the fact that in Dutch pronominal reference is generally established via grammatical gender, non-neuter mass nouns are expected to be referred back to by hij (lit. ‘he’), while neuter mass nouns are expected to be referred to by neuter het (cf. Table 2 in section 2.2). Interestingly, a mismatch shows up in spoken Dutch which has often been overlooked, but that has been put to the foreground by Audring (2006). Audring shows that precisely this pattern = that would be expected if het was positively specified for gender, namely as a “neuter” element — does not correspond to (the most natural) pattern found in spoken Dutch. Namely, in spoken Dutch the following “mismatch” shows up in this case: all mass nouns, including those which are grammatically non- neuter, can be resumed by the “neuter” pronoun het. (46) Ik vind puree van echte aardappelen altijd lekkerder want het is wat steviger. ‘I always prefer mash-NONNEUTER made out of real potatoes, be- cause it-NEUTER is firmer.’ 85 JASPER ROODENBURG AND AAFKE HULK. (47) __ Ik draag geen merkkleding tenzij het erg goedkoop is ‘I don’t wear brand-name-clothing-NONNEUTER unless it-NEUTER is very cheap.’ (Audring, 2006: 96) The idea that non-countability is a property of het is further supported by the fact that the mismatch mentioned is never found with non-neuter count nouns, Summarizing the observations made in the current and previous sections, we can draw the following (tentative) conclusion: (48) Het is non-countable across the board. This hypothesis is supported by a variety of cases. Firstly, it is supported by the behaviour of het in pronominal contexts, where het can refer back to mass nouns independently of their gender specification. Secondly, it is supported by the hypotheses put forward in the previous section, where we suggested that the behaviour of het in “abstract” de-adjectival noun constructions fol- lows straightforwardly from the idea that het bears a [-count] feature. Thirdly, we have seen in section 3.2 that default het can be combined with several ex- pressions that lack number. This was shown most clearly with nominalized infinitives, that trigger singular agreement even when used inside a coordina- tion construction. Finally, the fact that her can never be combined with plural nouns, an observation that we mentioned in section 2, is also in line with the tentative conclusion put forward in (48), namely that het preferably combines with non-countable elements.'* '4 ‘This final hypothesis raises of course the question as to how to explain the cases where het does combine with a countable noun. This is clearly the case with “neuter” definite count nouns (het boek ‘the-NEUTER book’, het huis ‘the-NEUTER house’, etc.), but also with diminutives (which fell outside the scope of this article). It is well known that diminutives in Dutch always combine with “neuter” het, triggering a shift with nouns that are lexically non-neuter (de auto “the-NON-NEUTER Car’, het autootje ‘the-NEUTER car-DIM’, ‘the little car’). Although the answer to these problems cannot be given here, one could.explore the idea that both cases are the result of an “overruling” mechanism. Within this line of reasoning, the non- countable character of het would be “overruled” by specific features of the noun, like (semantic) count features. A comparable line of reasoning could explain the behaviour of diminutives, al- though additional factors must be taken into account here. Note, however, that our central idea about her — that (non) countability takes precedence over gender distinctions — is compatible with a more general change that the Dutch gender system seems to be subjected to: it seems to shift from a purely grammatical system to a system that becomes sensitive to semantic dit tions (see e.g. De Vogelaer, 2007). Finally, note that the idea that default expressions can be “overruled” by semantic factors is defended by others as well: see e.g. Quer (this volume) and Zamparelli (this volume) for “over- ruling” within agreement relations involving numberless clausal arguments. a 86 PUZZLES ON GRAMMATICAL GENDER, 6. BACK TO CHILD DUTCH AND PUZZLE II We have seen in the preceding sections that “neuter” her in Dutch can function as a default element. It turned out that the behaviour of default het is compati- ble with Picallo’s analysis of “neuter” pronouns in Spanish and Catalan, which are not only unspecified with respect to gender but also with respect to num- ber. More tentatively, we have made the hypothesis that her is a non-countable element across the board. As we will see in the present section, this hypothesis allows us to raise some new questions about the way in which the Dutch ‘gender’ system of the definite determiner is acquired. Contrary to children acquiring a Romance language such as French, Dutch children not only are slow in acquiring the gender-morphology of definite determiners, but also they initially use one of the two forms — the non-neuter de — as a default form with all nouns. This lead us to formulate Puzzle II in (26) above, repeated here: Why isn’t neuter her chosen as the default in child Dutch? Now that we have suggested that default het in adult Dutch could be analyzed not as a default gender marker, but as a default non-count element, we have to reconsider the features involved in the acquisition of de and het by Dutch children. 6.1 What does de mark in Child Dutch? Let us start by looking at the developmental path of the acquisition of deter- miners in Child Dutch. The first stage is characterized by the use of bare nouns only, just as in other languages: article omission is a well-documented phenomenon in early child language. Children differ, however, in terms of how extensively they use bare nouns depending on their age and the language. Different accounts have been proposed for this cross-linguistic variation, Here we will adopt the ideas first proposed by Chierchia et al. (2001) who oppose Germanic to Ro- mance languages:!5 children acquiring Romance languages quickly discover, on the basis of positive evidence, that nouns are consistently used with arti- cles. On the contrary, children acquiring Germanic languages, such as Dutch, are exposed to bare nouns and determiner+noun sequences seemingly in free variation: before they can correctly use determiners, they first have to figure out which nouns are count and which are mass. This does not say anything yet about the acquisition of the morpho- phonology of the determiner. As we have briefly mentioned above, Dutch children differ from others in that until a rather advanced age (6 years) they massively overgeneralize the definite determiner de, using it as the “default” 'S We abstract away here from the literature discussing the variations found within the Ger- ‘manic and the Romance languages, which challenges Chierchia’s original hypothesis. 87 JASPER ROODENBURG AND AAFKE HULK choice. In the literature (e.g. Van der Velde, 2004; Blom, PoliSenska & Weer- man, to appear) this has been analyzed as the default use of non-neuter gender, The question arises, however, whether Dutch children in this stage of develop- ment are aware of determiners having gender features. In fact there is no clear, salient evidence for a gender distinction in the determiner system in the input to Dutch children. As we have stressed in section 2, the gender distinction only shows up in a limited number of situations: it is never visible in the plural (all definites take de, all indefinites are “bare”), while in the singular it only shows up with definites (de vs. he) and not with indefinites (which all take een). Furthermore, with respect to definites, de appears not only with all plural nouns but also with the majority (75%) of singular nouns (Van Berkum, 1996). Therefore, it may very well be the case that children use de as a default defi- nite determiner, but this does not necessarily imply that de is the expression of default gender, as has been widely assumed until now. It could be the case that initially for Dutch children de is a default num- ber/+count marker, which is in line with the observation that the distinction [count] is acquired first. When they discover the existence of the (other) def- nite determiner het, they may then initially analyze it as an element having [-count] features, as opposed to [+count] de. In other words, this hypothesis, if proved to be on the right track, could be seen as supporting evidence for the suggestions made above about het in Dutch adult language being a [-count] element. Moreover, this hypothesis makes the interesting prediction that we would expect, once children have acquired the mass/count distinction, they would use het mainly with mass nouns and they would use and overgeneralize de specifically with count nouns, as in *de boek (‘the book’), rather than with non-count nouns such as water (‘water’). Some as yet unpublished results (Brugman, 2007) suggest that there is indeed some evidence for a difference between mass and count nouns in the acquisition of de and her: children appear to start using her with non-count nouns, rather than with count nouns. We hope to explore this hypothesis further in future research. Summarizing, the new hypothesis now is that when Dutch children ini- tially only use de as a definite determiner, they are not yet aware of its gender specifications, but only of its [+count] specification, using it with count nouns and plurals. In taking this hypothesis one step further, we assume that once the children become aware of het as (another) definite determiner, they will initially characterize it as [-count], using it with singular mass nouns, as in het water,!® a hypothesis to be verified in future research. The next question then is: what could trigger the last developmental step towards the target grammar? In particular, how do Dutch children become aware of the (semantic) factors that may “overrule” this (default) pattern, giv- '© Dutch children are never found to use het with plurals, 88 PUZZLES ON GRAMMATICAL GENDER ing rise to het with certain count nouns, as in het boek? The answer to ques- tions like these lies outside the scope of this article. 7. CONCLUSION In this article, we have investigated the link between the notions of “neuter” gender and default in Dutch. In trying to extend to Dutch Picallo’s analysis of “neuter” as a default pronoun in Spanish and Catalan — which lacks both gender and number — we came upon two puzzles: (i) what distinguishes default het from non-default het, (ii) why isn’t het the default choice in Child Dutch? On the basis of an analysis of various constructions involvi — including nominalized infinitives, predicative superlatives and adjectival nouns — we have argued that the crucial property behind the default form her is not absence of gender, but non-countability. Although we have left open to further research whether two instances of het must be distinguished (default het vs. non-default het), or whether all instances of her are in fact default her, here we have explored the second — more challenging — option. Firstly, we have seen that default het in Dutch turned out to be more closely connected to the “neuter” pronoun in Spanish than previously assumed. We have suggested in particular that default het can be analyzed as a pronoun, even in cases where it is apparently adnominal. Secondly, we have seen that non-countability is a more general characteristic of (pronominal) het, which is not limited to instances of default her. Secondly, the idea that her can basically be analysed as a [-count] element across the board, may find support in Child Dutch. In particular, we hypoth- esized that the overgeneralization of de in Child Dutch does not involve the gender specification and hence does not need to be analyzed as the default use of gender. We suggested that it appears more plausible to assume that the early use of de shows that it is a default marker of [+number/count]. We further hy- pothesized that when the children discover the existence of the (other) definite determiner het, they analyze it as an element having [-count] features, in line with our ideas about her in Dutch adult language. Summarizing our highly speculative answers to the two puzzles raised in the beginning of this article, we would like to suggest that both in adult and in child language the Dutch element her appears to be the “default” choice in [-count] contexts. 89 JASPER ROODENBURG AND AAFKE HULK REFERENCES, Audring, J. (2006). Pronominal gender in spoken Dutch. Journal of Germanic Lin- guistics, 18(2), 85-116. Berkum, J.J.A. van. (1996). The psycholinguistics of grammatical gender: Studies in Janguage comprehension and production. Unpublished doctoral dissertation, Max Planck Institute for Psycholinguistics, Nijmegen. Blom, E., Poligenska, D., & Weerman, F. (to appear). Articles, adjectives and age of onset: The acquisition of Dutch grammatical gender. Second Language Research. Brugman, M. (2007). Grammaticaal geslacht en telbaarheid. Unpublished master’s thesis, Utrecht University. Cardinaletti, A., & Starke, M. (1994). The typology of structural deficiency. On the three grammatical classes. University of Venice Working Papers in Linguistics, 4(2), 41-109. Chierchia, G., Guasti, M., & Gualmini, A. (2001). Nouns and articles in child gram- ‘mar and the syntat/semantics map. Unpublished manuscript, University of Milan / University of Siena / University of Maryland, College Park. Clark, E., & Slobin, D. (1985). The acquisition of Romance with special reference to French. Hillsdale: Lawrence Erlbaum. Comips, L., & Hulk, A. (2006). External and internal factors in bilingual and bidialee- tal language development: Grammatical gender of the Dutch definite determiner. In C. Lefebvre, L. White, & C. Jourdan (Eds.), L2 acquisition and creole genesis. di- alogues (p. 355-378). Amsterdam: Benjamins. Houwer, A. de, & Gillis, S. (1998). The acquisition of Dutch. Amsterdam: Ben- jamins. Kester, E.-P. (1996). The nature of adjectival inflection. Unpublished doctoral disser- tation, Utrecht University. Longobardi, G. (1994). Reference and proper names. Linguistic Inquiry, 25(4), 609. Longobardi, G. (to appear). Reference to individuals, person, and the variety of map- ping parameters. In A. Klinge & H. Milller (Eds.), Proceedings of the Copenhagen symposium on determination. Picallo, M.C. (2005). Some notes on grammatical gender and I-pronouns. In Fach- bereich Sprachwissenschaft, Arbeitspapier 119 (p. 107-121). Konstanz: Universitit Konstanz Picallo, M.C. (2006). On gender and number. Unpublished manuscript, Universitat Autonoma de Barcelona. Picallo, M.C, (2007). On gender and number. Unpublished manuscript, Universitat Autonoma de Barcelona. Picallo, M.C. (this volume, 2008). Gender and number in Romance. Lingue ¢ Lin- guaggio, VII(1), 47-66. Quer, J. (this volume, 2008). Argument clauses and nominal features. Lingue ¢ Linguaggio, VII(1), 93-110. Rizzi, L. (1986). Null objects in Italian and the theory of pro. Linguistic Inquiry, 17, 501-557. Rooryck, J. (2003). The morphosyntactic structure of articles and pronouns in Dutch. In J. Koster & H. van Riemsdijk (Eds.), Germania et alia. A linguistic webschrift for Hans den Besten. Groningen: University of Groningen, 90 PUZZLES ON GRAMMATICAL GENDER Velde, M. van der. (2003). Déterminants et pronoms en néerlandais et en francais: syntaxe et acquisition. Unpublished doctoral dissertation, Université Paris 8. Velde, M. van der. (2004). L’acquisition des déterminants en LI: une étude com- Parative entre le frangais et le néerlandais. Acquisition et Interaction en Langue Etrangére, 21, 9-46, Vogelaer, G. de. (2007). Transmission, diffusion, or both: gender change in Dutch. Paper presented at Transmission and Diffusion, January 17-19 2008, Max Plank Institute for Psycholinguistics Nijmegen. Zamparelli, R. (this volume, 2008). Features, facts and clausal agreement. Lingue € Linguaggio, VII(1), 111-134. Jasper Roodenburg Internationales Graduiertenkolleg Stuttgart-Paris VIL University of Stuttgart Postbus 15692 1001 ND Amsterdam The Netherlands e-mail: jasperro@xs4all.nl Aafke Hulk Amsterdam Center for Language and Communication (ACLC) University of Amsterdam Spuistraat 210 1012 VT Amsterdam The Netherlands e-mail: aafke.hulk¢ 91

You might also like