The distribution of personal pronouns in the pronominal
reference system of Dutch children
Maartje Hoekstra
4th of August, 2017
University of Amsterdam
Supervisor:
2 Abbreviations
Glossing follows the Leipzig Glossing Rules
(http://www.eva.mpg.de/lingua/resources/glossing-rules.php) 3 third person ART article C common gender DEF definite DEM demonstrative DIM diminutive F feminine gender INDF indefinite M masculine gender N neuter gender NUM numeral OBJ object PL plural SBJ subject SG singular
3
Table of content
Introduction ... 5
1. Gender and gender agreement in Dutch ... 7
1.1 The role of lexical gender in pronominal reference ... 11
1.2 The role of individuation in pronominal reference ... 14
1.3 Competition between lexical gender and semantic gender in modern spoken Dutch ... 17
2. Lexical gender in Dutch ... 22
2.1 Lexical gender assignment ... 22
2.2 Lexical gender agreement ... 24
2.3. Previous research on the acquisition of lexical gender ... 27
2.4 Implications for this study ... 28
3. Semantic features in Dutch ... 30
3.1 Count – mass distinction ... 30
3.1.1 Previous research on the acquisition of the count – mass distinction ... 33
3.2 Implications for this study ... 34
4. Hypotheses and predictions ... 36
5. Method ... 38
5.1 Participants ... 38
5.2 Test-items ... 38
5.3 Test-design and procedure ... 39
6. Results ... 42
7. Discussion ... 45
7.1 comparison between children and adults ... 45
7.2 concrete versus abstract nouns ... 46
4
8. Conclusion ... 51
References ... 52
Appendix I: Test-items and questions version 1 ... 56
Appendix II: Test-items and questions version 2 ... 66
Appendix III: Data per item per child – participant group 1 ... 76
5 Introduction
Contemporary Standard Dutch has a nominal two-gender system where nouns can be common gendered (a conflation of the former masculine and feminine gender) or neuter gendered. Personal pronouns, on the other hand, still have a distinct form for masculine, feminine and neuter gender.
In the pronominal reference system of Standard Dutch, personal pronouns can show two types of agreement with their antecedent noun: lexical gender agreement, agreement with the lexical gender of the noun, and semantic gender agreement, agreement with semantic properties of the noun’s referent. It is widely acknowledged that in reference to nouns denoting humans, the gender of the personal pronoun agrees with the natural
gender/biological sex of the noun’s referent. For inanimate referents, traditionally, the gender of the personal pronouns agree with the lexical gender of the noun, in reference to nouns denoting inanimate entities. The masculine pronouns hij/hem (‘he/him’) refer to common inanimate nouns, and the neuter pronouns het/het (‘it/it’) refer to neuter inanimate nouns. However, research by Audring (2009) has shown that in reference to inanimate nouns in spoken standard Dutch, rather than distributing pronouns based on the lexical gender of the noun, adult speakers increasingly distribute personal pronouns based on a semantic property of their antecedent noun: it’s degree of individuation, or, countability, Adult Dutch speakers tend to use masculine personal pronouns for common and neuter nouns denoting referents that are higher in individuation (countable) and neuter personal pronouns for common and neuter nouns denoting referents that are lower in individuation (uncountable/masses). Hence, in the pronominal reference system of adults, both the semantic feature [±animate] and the semantic feature [±count] appear to play a role in pronoun distribution.
Little is known about the pronominal reference system of children and if children distribute masculine personal pronouns and neuter personal pronouns based on the degree of individuation of the noun’s referent, i.e. based on the features [±animate] and [±count]. A pilot study by Bijmans et al. (2017) indicates that the pronominal reference system of Dutch children is undergoing a process of neutralization, where the boundary of the use of the neuter personal pronoun seems to be shifting towards the domain of countable referents and other animates. Hence the influence of the feature [±count] seems to be decreasing in the pronoun choice of children. If this is indeed the case, the pronominal reference system of Dutch could be moving towards a system found in English, where the feature [±animate] plays an
important role in pronoun distribution and boundary of the use of the masculine pronouns and neuter pronouns is situated somewhere between animate and inanimate referents. To see if
6
this is indeed the case, and if so, if this is merely a temporary shift in pronoun distribution or a more permanent shift, this research looks at the distribution of masculine personal pronouns in reference to common nouns by children aged 5;1 – 6;5 and by children aged 7;7 – 8;8.
7 1. Gender and gender agreement in Dutch
Originally, Dutch had three nominal genders: masculine, feminine, and neuter gender. Due to a process of deflection in the noun phrase, masculine and feminine gender became
indistinguishable from each other and modern Standard Dutch is left with two nominal genders: common gender (the conflation of the former masculine and feminine gender) and neuter gender. The gender specification of a noun is largely arbitrary, often not visible on the noun itself, and is traditionally deduced from the gender of its agreeing definite articles. Nouns that take the common definite article de (‘ the’) are considered to be common nouns, while nouns that take the neuter definite article het (‘the’) are considered to be neuter nouns.
Different from the adnominal elements (and most pronominal elements), which distinguish between two genders, the personal pronoun has maintained separate forms for each of the former three genders. These different genders are visible in the third person singular personal pronouns which have the masculine/common forms hij/hem (‘he/him’), the feminine forms zij/haar (‘she/her’), and the neuter forms het/het (‘it/it’). Whereas gender agreement inside the noun phrase is determined by the lexical gender of the noun, as can be seen on the aforementioned definite articles, pronominal gender can either agree with the lexical gender of the antecedent noun, or with semantic properties of the noun’s referent (De Vos 2013: 1). These two types of agreement are known as lexical gender agreement and
semantic gender agreement respectively (Audring 2009; Kraaikamp 2012, 2017).
It is not uncommon that personal pronouns can agree with semantic properties of the noun’s referent, while other elements agree with the lexical gender of the noun. Corbett (1979: 204) noticed that cross-linguistically, the likelihood of agreement targets showing formal (lexical) or semantic gender agreement can be represented by a hierarchy he calls the Agreement Hierarchy:
(1) The Agreement Hierarchy
Attributive < predicative < relative pronoun < personal pronoun formal semantic
agreement agreement
The elements on this hierarchy differ in their syntactic distance to the noun. Attributive elements, such as definite articles, which are syntactically closest to the noun are most likely to show formal agreement (lexical agreement). The more the syntactic distance increases, the greater the likelihood of semantic agreement. Hence, the personal pronoun, which is
8
syntactically furthest from the noun, is the element that is most likely to show semantic gender agreement.
In Dutch, semantic gender agreement on personal pronouns is often found in reference to persons, and is based the natural gender or biological sex of the noun’s referent (example taken from Kraaikamp 2017: 15):
(2) Het meisje ging snel naar school,
The girl(N) went quickly to school
want ze was te laat. Because 3SG.F was to late
‘The girl left for school quickly, because she was late.’
In example (2) the neuter noun meisje ‘little girl’ clearly refers to a female referent and hence the feminine personal pronoun zij ‘she’ is used (here presented in its abbreviated version ze). Similarly, a neuter noun such as jongetje (‘little boy) is often referred to with the masculine pronouns hij/hem (‘he/him’) (Booij 2002: 36). This type of semantic agreement (based on the biological sex of the noun’s referent) in reference to persons is widely acknowledged by the official spelling dictionary (the Woordenlijst Nederlandse Taal), the standard reference grammar Algemene Nederlandse Spraakkunst (ANS) and the main dictionary, the Van Dale (Audring 2009: 117). Semantic gender agreement based on the biological sex of the noun’s referent is also found in reference to other animates (animals). However, nouns denoting animals are mostly pronominalized with the masculine personal pronoun (Audring 2009: 43)
In reference to inanimate nouns, the claim is that lexical gender agreement prevails (de Vos 2013:1; Hulk & Cornips 2009: 4; Audring 2009: 119). Neuter nouns take neuter
pronouns and common gender nouns combine with masculine pronouns (a process also known as masculinization) (Audring 2009: 118). (3) is an example of such lexical gender agreement (example taken from Kraaikamp 2017: 8):
(3) Moet je nog wat informatie over dat boek hebben? need you more some information about that book(N) have
Dan moet ‘k ‘t nog niet gaan inleveren. then should I 3SG.N yet not go return
9
‘Do you need some information about that book? Then I shouldn’t return it yet.’
The noun boek (‘book’) in (3) is a neuter noun and is accordingly referred to with the neuter personal pronoun het (‘it’) (here presented in its abbreviated version ‘t).
Interestingly, research by Audring (2006, 2009) shows that in the pronominal reference system of spoken modern Dutch, semantic gender agreement is not only found in reference to animate entities, but is also increasingly found in reference to inanimate entities. This semantically motivated agreement system is not based on the biological sex of the referent, but on the referent’s boundedness, or, countability. In her corpus study, Audring (2006, 2009) found that in spoken Dutch, the masculine personal pronouns hij/hem (‘he/him’) are often used for referents that are countable and have clear boundaries, such as objects. The neuter noun boek (‘book’) in spoken Standard Dutch is alternatively referred to with the masculine pronoun (example taken from Audring 2006: 95):
(4) Moet je nog wat informatie over dat boek hebben? need you more some information about that book (N) have
Dan moet ‘k ‘m nog niet gaan inleveren. then should I 3SG.M yet not go return
‘Do you need some information about that book? Then I shouldn’t return it yet.’
The neuter personal pronouns het/het (‘it/it’), on the other hand, are often used for referents that are uncountable and have unclear boundaries, such as masses (example taken from Audring 2009: 122):
(5) Groente moet je niet te lang koken omdat het anders vegetables(C) must you not too long cook because 3SG.N otherwise
zijn voedingsstoffen verliest. his nutrients loses
10
Audring (2009: 127) proposes that the semantic distribution of Dutch pronouns can be aligned to a conceptual hierarchy called the Individuation Hierarchy. This Individuation Hierarchy is an offspring of Silverstein’s (1976) Animacy Hierarchy, but captures more subtle semantic differences regarding inanimate nouns. On the Individuation Hierarchy, referents are ordered in semantic classes, ranging from a high degree of individuation (left) to a low degree of individuation (right):
(6) The Individuation Hierarchy
Human – Other – Bounded – Specific – Unbounded/Unspecific animate object/ mass abstract mass
Bounded abstract
Nouns denoting animate entities are high in individuation, with human as the highest
individuated referents, followed by nouns referring to other animates. Inanimate referents are less individuated, with bounded and countable objects/abstracts (boek/idee – ‘book/idea’) as highest in individuation, unbounded and uncountable abstracts/unspecific masses
(pijn/sneeuw – ‘pain/snow’) as least individuated, and specific masses, which combine properties of objects and substances (zijn thee – ‘his tea’) taking the middle ground. On the Individuation Hierarchy, each personal pronoun has its own domain:
Table (1). The different semantic categories including examples and their corresponding pronouns in Dutch.
On the left of the hierarchy, with nouns denoting human referents, the pronominal gender carries between masculine and feminine, based on the biological sex of the noun’s referent.
Degree of individuation Male human/ female human Man/Woman Other animate Chicken Bounded object/ bounded abstract Book/Idea Specific mass His tea Unbounded abstract/ Unspecific mass Pain/Snow Pronoun used in Dutch
Masculine/ Masculine Neuter
11
For nouns denoting other animates, the feminine pronouns are occasionally used, but often other animates are referred to with the masculine pronouns (Audring 2009; Kraaikamp 2012). For inanimate referents the pronominal gender carries between masculine and neuter. The boundary between the domain of the masculine pronouns and the domain of the neuter pronouns is situated somewhere between bounded referents (count nouns) and unbounded referents (mass nouns) (Kraaikamp 2012: 20).
Thus, in the pronominal reference system of spoken Standard Dutch, the distribution of pronouns can be based on two types of agreement with the antecedent noun: lexical gender agreement, based on the lexical gender of the antecedent noun, or semantic gender agreement, based on the degree of individuation of the noun’s referent. Feminine and masculine personal pronouns are used for animate referents, which are high in individuation, based on their biological sex (although animals are usually referred to with masculine pronouns). For inanimate referents, the masculine pronouns are used for referents that are higher in individuation and bounded/countable, while the neuter personal pronouns are used in for referents that are lower in individuation and unbounded/uncountable. Hence, both the feature [±animate] and the feature [±count] appear to play a role in the pronoun distribution of spoken Dutch. In the next two paragraphs, the role of lexical gender and the role of the individuation of the noun’s referent in (Dutch) pronominal reference will be further discussed. The last paragraph will further explore the ongoing competition between lexical and semantic gender agreement.
1.1 The role of lexical gender in pronominal reference
The pronominal gender system of spoken modern Dutch seems to be undergoing a process called resemanticization (Wurzel 1986) where the old grammatical gender system in which pronouns agree syntactically is gradually replaced by a new system that uses semantically motivated pronouns, based on the degree of individuation of the noun’s referent (De Vos 2013: 2; Audring 2009).
An explanation for the rise of semantic gender agreement is given by Audring (2009: 172) who states that this rise “is connected to the paradigmatic mismatches that have uprooted the traditional Dutch gender system”. As previously mentioned, the Dutch language
historically had a grammatical three-gender system (masculine, feminine, and neuter gender), similar to other Germanic languages. This system underwent a process of deflection, which lead to the conflation of masculine and feminine gender. Hence, modern-day Standard Dutch knows two nominal and adnominal genders (common and neuter gender) but retained its three
12
pronominal genders, resulting in a mismatch between the nominal and pronominal genders. This paradigmatic mismatch between nominal and pronominal gender is shown in table (2) (table taken from Kraaikamp 2017: 24):
Common gender noun Masculine personal pronoun Feminine personal pronoun Neuter gender noun Neuter personal pronoun
Table (2). Paradigmatic mismatch in gender values between nouns and pronouns.
Due to this paradigmatic mismatch, it became unclear which pronoun should go with a common noun, since there was no pronoun directly corresponding to this noun. For animate nouns, speakers could fall back on the natural gender/biological sex of the referent. For inanimate common nouns, this was is not the case. To compensate for the reduced paradigms, a functional reinterpretation of pronominal genders emerged, based on conceptual semantic criteria rather than the less identifiable lexical criteria (Audring 2006, 2009).
However, Kraaikamp (2017: 24) observed that the idea of paradigmatic mismatches cannot be the (full) explanation for the rise of semantic gender agreement in the whole pronominal domain. She points out that this paradigmatic mismatch only affects common nouns, not neuter nouns. Hence, the mismatches between nominal and pronominal gender should only have to lead to a new division of labor between the masculine and feminine pronouns in reference to common nouns. Studies by Kraaikamp (2017: 25) and Geerts (1966: 192) show that this new division of labor between the masculine and feminine pronouns was already happening in Early Modern Dutch (around the 17th century), where the feminine pronoun “became reserved for agreement with female referents”. With the feminine pronoun obtaining this semantic function, the masculine pronoun increasingly became toreplace this feminine pronoun in reference to all other common nouns, thus resolving the mismatch in nominal and pronominal gender (Kraaikamp 2017: 25). The results of the corpus study carried out by Audring (2009: 114) show that the feminine pronoun has maintained this semantic function in present-day spoken Dutch, as it only occurred in reference to female
persons/animals. If the feminine personal pronouns are strictly used for female referents and no longer plays a role in lexical gender agreement, there is no necessary mismatch between noun and pronoun gender with respect to the lexical gender agreement system (Kraaikamp 2012: 204). Consequently, there would be no need for a semantic reinterpretation of masculine and neuter gender pronouns.
13
The observation that the rise of semantic gender agreement is not necessarily related to a paradigmatic mismatch is supported by research on the system pronominal reference in southern dialects of Dutch,such as Flemish varieties (De Vogelaer & De Stutter 2011). Since these southern varieties still have a nominal, adnominal, and pronominal three-gender system, there is no paradigmatic mismatch between noun and pronoun gender. Yet, De Vogelaer & De Stutter (2011: 203) showed that in the Dutch varieties spoken in East and West Flanders by adults, a tendency towards the resemanticization of pronouns is emerging. However, De Vogelaer & De Stutter (2011) found that the tendency towards resemanticization was higher in participants speaking the West-Flemish variety than in participants speaking the East-Flemish variety. The research showed that an important factor in these different tendencies towards resemanticization had to do with the richness of adnominal morphology of the two varieties. Although both West-Flemish and East-Flemish make a distinction between masculine, feminine, and neuter gender in the adnominal domain, this distinction is more overtly marked in the East-Flemish variety. Hence, although the paradigmatic mismatch resulting from the loss of a nominal gender might not be the direct reason for the rise of semantic gender agreement, the process of deflection and the decreasing visibility of lexical gender (that also underlies the collapse of masculine and feminine gender) seems to be a motivating factor in the occurrence of semantic gender agreement
Since adnominal elements are known to agree with the lexical gender of the noun and in many languages lexical gender is mostly, if not only, visible on adnominal elements, covariance between the richness of the adnominal morphology of a language and the amount of lexical gender agreement shown in the pronominal domain is not peculiar. Klom & De Vogelaer (2017: 4) observed that, when taking a comparative perspective regarding Germanic languages, “a correlation is observed between pronominal gender agreement and adnominal (gender) morphology, with High German and English as the most ‘extreme varieties”. Historically, English and German, like Dutch, had a grammatical three-gender system. English lost its gender morphology in the adnominal domain but maintained these genders in the pronominal domain. With the loss of grammatical gender in the adnominal domain, the lexical gender agreement system in the pronominal domain was abandoned in favor of a semantic gender agreement system. German, on the other hand, still clearly marks three grammatical genders in its adnominal and pronominal domain. Correspondingly, pronominal
14
reference in German is (predominantly) based on lexical gender agreement1. Standard Dutch appears to occupy the middle ground between German and English. Although modern-day Standard Dutch still knows two adnominal genders, its adnominal gender morphology is far from being overt. Firstly, gender-marking on agreeing adnominal elements contrasting the distinction between common and neuter nouns is only visible in their singular forms, all plural nouns take the common agreement patterns. Secondly, indefinite determiners do not mark gender at all. Thirdly, attributive adjectives only mark gender when encountered in an
indefinite context. Hence, the visibility of lexical gender in its entirety is relatively low in the Dutch language. It has been argued that, due to the low visibility of lexical gender per se in the adnominal domain, the role of lexical gender in Dutch pronominal reference is becoming compromised, resulting in the emergence of a semantic system of gender agreement
(Kraaikamp 2012, 2017; De Klom & De Vogelaer 2017). “This process can be considered an instance of morphological regularization, in other words, the rise of an innovative rule system when the traditional system becomes too opaque to be successfully acquired” (De Vos & De Vogelaer 2011: 276).
The next paragraph will go further into the role of semantic gender agreement in pronominal reference, and the association between gender and individuation.
1.2 The role of individuation in pronominal reference
The semantics-based distribution of personal pronouns in the pronominal reference system of spoken Standard Dutch has been related to the degree of individuation of the noun’s referent. For animate referents, specifically humans, which are high in individuation, the masculine and feminine pronouns are used based on the biological sex of the referent. However, for animals, the masculine pronoun is most often used. For inanimate referents, the masculine pronouns are used for referent that are higher in individuation and bounded/countable, and the neuter pronoun is used for referents that are low in individuation and unbounded/uncountable.
Dutch is not the only language that distributes gendered and neuter personal pronouns based on the degree of individuation of the noun’s referent. Siemund (2002) observed that the distribution of pronouns in the system of pronominal reference in both the standard and the regional varieties of (several) Germanic languages can also be related to the degree of
individuation of the noun’s referent. The cut-off point between the use of gendered and neuter
1
Although research by Audring (2009) and Kraaikamp (2016) suggests that even in German, semantic gender agreement can be found with referents on the outer side of the ends of the Individuation Hierarchy (human – unbounded abstracts/unspecific masses).
15
pronouns is merely situated at different points of the hierarchy. This can be illustrated by comparing the distribution of pronouns in languages such as English to that of Dutch. For English, which has lost its lexical gender altogether, the cut-off point between the use of gendered and neuter pronouns is on the left of the hierarchy between nouns denoting animate and nouns denoting inanimate referents. Hence, the feature [±animate] plays the important role in pronominal reference in Standard English. In spoken Standard Dutch, the cut-off point between the use of gendered and neuter pronouns is situated further to the right, between nouns denoting bounded/countable referents and nouns denoting unbounded/non-countable referents. Apart from the feature [±animate], additionally the feature [±count] influences pronoun distribution in spoken Dutch. Table (3) illustrates the different cut-off points of Standard English and (spoken) Standard Dutch:
Degree of individuation Male human/ female human Man/Woman Other animate Chicken Bounded object/ bounded abstract Book/Idea Specific mass His tea Unbounded abstract/ Unspecific mass Pain/Snow Animate Inanimate Count Mass Standard English Masculine/ Neuter Feminine Standard Dutch Masculine/ Feminine Masculine Neuter Table (3). The different cut-off points in pronominal gender for Standard English and (spoken) Standard Dutch
The distribution of the masculine and neuter pronoun based on the boundedness/countability of the noun’s referent can also be found in other Germanic language varieties. Siemund (2002) showed that in regional varieties of English, Danish, Frisian, and German, the distribution of gendered and neuter pronouns also reflects a distinction between bounded an unbounded referents. In several of these language varieties, a similar distribution of masculine
16
and neuter pronouns is found as for spoken modern Dutch. As Siemund (2002: 213-214) states:
“the distribution of animate (mainly masculine) and inanimate pronouns does not depend on the animacy of the referents picked out, as in the standard varieties, but on the categorization of referents into those that are countable and concrete on the on hand and those that are non-countable (masses) or abstract on the other. In such varieties, the masculine pronouns are used for reference to humans, animals and objects [...] whereas the neuter pronoun is used for materials, liquids, substances and the like”
Similar to Dutch, these language varieties underwent a process of deflection in their gender systems and are left with two-gender or zero-gender systems. Interestingly, these language varieties, in which a distinction between gendered and neuter pronouns based on a count – mass distinction is visible, also show gender agreement within the noun phrase. Hence, Kraaikamp (2017: 151-153) suggests that “the preservation of gender distinctions in the noun phrase makes semantic agreement move in a different direction than when all gender
agreement in the noun phrase is lost”. According to Kraaikamp (2017: 151), in languages such as English which have completely lost their lexical gender, the use of the neuter pronoun spreads to the left of the Individuation Hierarchy. In languages that have maintained lexical gender, but in which this gender system is weakened, might still have an association between masculine gender and inanimate entities and hence the use of the masculine pronoun spreads to the right of the Individuation Hierarchy.
Furthermore, Siemund (2002: 229) observed that “systems of pronominal gender – or gender systems in general for that matter – based on the mass/count distinction of nominals do not represent a feature unique to Germanic languages”. In areas in the Romance world, a count/mass distinction (in nominal gender) can also be found. In Central Italian Dialects, for example, neuter nouns roughly correspond to the class of mass nouns and count nouns belong to the other gender classes (Siemund 2002:228). Although, the count-mass distinction in (pro)nominal gender might not be unique to Germanic languages, it appears to be relatively unique to Indo-European languages. Although it has been observed that there are languages from other language families that seem to incorporate “a contrast in terms of individuation”, the count-mass distinction in (pro)nominal gender appears to be restricted to Indo-European languages (Siemund 2002: 229). A possible explanation for the finding that an association between count/mass nouns, or individuation, and (pro)nominal gender seems to be restricted
17
to Indo-European languages can be found in research by Kraaikamp (2012, 2016, 2017). Kraaikamp (2012, 2016, 2017) hypothesised that the original Proto-Indo-European genders were already associated with different degrees of individuation as shown in (7) (taken from Kraaikamp 2017: 42):
(7) The semantic basis of Indo-European gender
count/ individual --- male individual ‘masculine gender’ collective/abstract --- female individual ‘feminine gender’ mass --- sexless ‘neuter gender’
“Therefore, the observed semantic agreement [in personal pronouns] in present-day Germanic varieties may be an increased surfacing of a tendency that was already present before the nominal gender systems of these varieties changed” (Kraaikamp 2016: 5). Kraaikamp (2012, 2016, 2017) suggests that the association between biological sex and pronominal gender developed in a later stadium in addition to these original classifications.
Hence, the distribution of gendered and neuter pronouns based on the degree of individuation of the noun’s referent is not a feature unique to Dutch, but has been found to occur in other Germanic languages and to a lesser degree in Romance languages. This could be due to the hypothesised original classifications of genders in Proto-Indo European. In Dutch, both the individuation distinction as the biological sex distinction are visible in the pronominal reference system alongside the traditional lexical system of agreement
(Kraaikamp 2017). The next paragraph will go further into the factors that can influence whether a Dutch speaker makes his choice in pronouns based on the lexical gender of the noun or on the degree of individuation of the noun’s referent.
1.3 Competition between lexical gender and semantic gender in modern spoken Dutch
There are several factors related to whether a Dutch speaker makes his choice in pronoun based on the lexical gender of the noun or the semantic gender of the noun’s referent. Two of
18
the factors that are important to consider for this study are the position of the noun’s referent
on the Individuation Hierarchy, and the age of the speaker.
The “pull from semantic gender”, in the words of Kraaikamp (2012: 227), is not equally strong for al referents on the Individuation Hierarchy. Audring (2009: 167) found that semantic gender agreement in pronominal reference is more likely to occur on the outer sides of the Individuation Hierarchy, i.e. with nouns referring to animate referents (especially nouns referring to persons) and nouns referring to unspecific masses. Table (4) gives an overview of this data (note that the nouns in the object category were lexically neuter, and the nouns in the categories specific masses and unspecific masses were lexically common):
Person Animal Object Specific mass Unspecific mass Switch 129 (94%) 13 (81%) 241 (52%) 22 (25%) 75 (88%)
Match 8 (6%) 3 (19%) 222 (48%) 65 (75%) 10 (12%)
Table (4). Pronoun switch and match per degree of individuation as provided by Audring (2009:167). Swith = semantic gender agreement; match = lexical gender agreement.
Audring (2009: 167) gives the following explanation for the finding that semantic gender agreement is more likely to occur at the outer ends of the hierarchy:
“If the degree of individuation is viewed as a gradual phenomenon that can be expressed in a hierarchy, then the extreme ends of the hierarchy are most conflictive. The more individuated a referent is, the more this classes with neuter gender, while extremely low individuation conflicts maximally with common gender”
The reason for the low amount of gender switches in the category of specific mass could have to do with the idea that this semantic category combines properties of both objects and
substances/masses. As objects are referred to with the masculine pronoun and
substances/masses with the neuter pronoun, there is less necessity to switch from a pronoun that agrees with the lexical gender of the noun to a pronoun that semantically feels more “correct”. That is, due to the double semantic status of nouns denoting specific masses, the lexically agreeing pronoun does not often feel semantically “wrong” (Audring 2009: 168).
Moreover, Audring (2009: 169) found that the older the speaker, the more likely that the pronouns used in pronominal reference agreed with the lexical gender of their antecedent. Correspondingly, the younger the speaker the more likely that the pronouns used in
19
pronominal reference agreed with the semantic properties of the noun’s referent. Table (5) gives an overview of this data:
≤ 20 21-30 31-40 41-50 51-60 > 60
Switch 106 (72%) 293 (65%) 30 (42%) 22 (38%) 24 (52%) 21 (35%) Match 41 (28%) 151 (35%) 22 (46) 38 (62%) 23 (48%) 28 (65%)
Table (5). Pronoun switch and match per age group as provided by Audring (2009: 169). Switch = semantic gender agreement; match = lexical gender agreement.
This data shows that the pronominal reference system is indeed undergoing a process of resemanticization, where in spoken Dutch, the semantically motivated system is spreading and becoming more influential than the lexically motivated gender system.
Contemporary research on pronominal reference in spoken Standard Dutch has mainly focused on the distribution of gendered and neuter pronouns in the pronominal reference system of adult speaker and few studies focus on the distribution of pronouns in the pronominal reference system of Dutch children. A pilot study by Hulk & Cornips (2010) found that for children aged 4;4 – 6;6, the feature [±count] plays a significant role in the choice of masculine and neuter personal pronouns. The masculine pronoun hem (‘him’) was often used to refer to nouns denoting countable objects and use the neuter pronoun het (‘it’) was often used to refer to nouns denoting uncountable masses, regardless of the lexical gender of the noun. These results indicate that children base their choice in pronoun on the degree of individuation or countability of the noun’s referent for nous denoting inanimate entities. However, the test-items used in this study either belonged to the semantic class of bounded objects or to the semantic class of unspecific masses and no further semantic distinctions were made relating to the degree of individuation of the noun’s referent.
A pilot study by Bijman et al. (2017) aimed to see if similar patterns in pronoun distribution could be found between the pronominal reference system of adults and that of children around the age of 5, taking the different degrees of individuation of the noun’s referent into account. Different from the corpus study performed by Audring (2009), in this study the semantic categories of bounded objects and bounded abstract were investigated separately. The results showed that that for adults, the boundary between the use of the masculine pronoun and the neuter pronoun, in reference to common nouns, was situated somewhere between the bounded abstracts and the specific masses:
20 Pronouns used Other animate Bounded object Bounded abstract Specific mass Unspecific mass Hem (‘him’) 93,8% 89% 60% 7,5% 2,4% Het (‘it’) 6,3% 11% 40% 92,5% 97,6%
Table (6). Percentages of the use of the masculine pronoun hem (‘him’) and the neuter pronoun het (‘it’) in pronominal reference by adults for each semantic category as provided by Bijman et al. (2017: 14)
Although this boundary is situated somewhat more to the left than in the result found by Audring (2009), these results still indicate that in the pronominal reference system of adults, a distinction is made between referents with clear boundaries (countable objects/abstracts) and referents with unclear boundaries (specific masses/unspecific masses).
Interestingly, Bijman et al. (2017: 14) found that the distribution of the masculine pronoun hem (‘him’) and the neuter pronoun het (‘it’) for children around the age of 5 differed from that of adults. The children were found to mostly make use of the masculine personal pronoun hem (‘him’) in reference to other animates and (countable) bounded objects, but the neuter pronoun het (‘it’) was mostly used in reference to (countable) bounded
abstracts, specific masses and unspecific masses:
Pronouns used Other animate Bounded object Bounded abstract Specific mass Unspecific mass Hem (‘him’) 71,4% 64,5% 17,4% 3,7% 6% Het (‘it’) 28,6% 35,5% 82,6% 96,3% 94%
Table (7). Percentages of the use the masculine pronoun hem (‘him’) and the neuter pronoun het (‘it’) in pronominal reference by children around the age of 5 for each semantic category as provided by Bijman et al. (2017:14).
Bijman et al. (2017: 16) found that there was as significant difference between the use of the neuter personal pronoun het (‘it’) in reference to bounded abstract for the children and for the adults. Hence, for children around the age of 5, the cut-off point between the use of the masculine and neuter pronouns seems to be situated more towards the left on the Individuation Hierarchy, between the semantic categories bounded object and bounded abstract. In addition, Bijman et al. (2017) found that children used the neuter pronoun het (‘it’) significantly more often in the categories bounded object and other animate than adults. This data suggests a process of neutralization in the pronominal reference system of Dutch children, where the neuter pronoun is gaining ground in all semantic categories on the
21
left side of the hierarchy (bounded referents and other animates). This could indicate that the feature [±count] is becoming to play less of a role in pronoun distribution in the Dutch pronominal reference system. If so, the Dutch pronominal reference system could be moving towards a pronominal reference system as found in English, were the boundary between the use of the masculine and neuter personal pronoun is based on the feature [±animate] and lies somewhere between animate and inanimate referents.
To better understand the possible distribution patterns of masculine and neuter personal pronouns in child pronominal reference and to be able to compare them to the distribution of pronoun gender found in adults, it is important to consider the underlying knowledge and mechanisms that bring about the patterns found in adult pronoun distribution. Hence, the next chapters will further discuss how lexical gender and the semantic feature [±count] manifest themselves in Standard Dutch, and when these features are acquired by children.
22 2. Lexical gender in Dutch
Generally, the Dutch language is considered to have “a complex gender assignment system with opaque rules [and] the gender system of this language is particularly hard to acquire” (Audring 2014:16). If lexical gender is a difficult feature to acquire, it is important to take the acquisition process into consideration when looking at the child system of pronominal
reference. After all, if children are unaware of or have difficulties with lexical gender, that is, gender assignment and the agreement patterns that this lexical gender invokes, they cannot be expected to uniformly show lexical gender agreement when pronominally referring to a noun.
The following paragraphs will focus on gender assignment, the agreement patterns the gender of nouns trigger and the saliency of lexical gender and lexical gender agreement. In addition, the last paragraph will focus on previous research on the acquisition of gender. In doing so, a more complete picture of lexical gender agreement can be presented, which needs to be considered when looking at the types of agreement children show in pronominal
reference.
2.1 Lexical gender assignment
As mentioned in the first chapter, nouns in contemporary Dutch have a lexical gender and can either be common (masculine/feminine) or neuter. Whether a noun is assigned a common or a neuter gender is largely arbitrary and there are no unambiguous semantic principles
underlying this assignment. Moreover, the gender of a root noun is not visible from its morphophonological form (Booij 2002; Blom, Polisenska & Weerman 2008; Kraaikamp 2012). Contrary to, for example, Romance languages, where nouns often have typical
feminine or masculine endings, the gender of root nouns in Dutch has no direct morphological expression. As a result, the gender of Dutch root nouns has to be learned and lexically stored (Booij 2002).
Although the gender of root nouns cannot be systematically organized along semantic or morphophonological lines, there are a few semantic regularities in gender assignment. As Kraaikamp (2017: 36-39) points out, examples of semantic patterns in lexical gender can be found in double gender nouns and nouns that belong to the same semantic domain. Double gender nouns can either take common or neuter gender but their meaning differs according to the gender they take. Instances of such double gender nouns are nouns relating to materials such as diamant (‘diamond’) and steen (‘stone’). When these nouns are common they denote an object made of the material, while their neuter forms denote the material itself (Kraaikamp 2012: 216). In nouns that belong to the same semantic domain the observed pattern is that of
23
superordinate categories which often take neuter gender and their subordinate, specific sub-categories which often take common gender. This can be illustrated by nouns denoting instruments. The word instrument ‘instrument’ itself denotes a category and is neuter, while specific instruments such as gitaar ‘guitar’ are often common (Kraaikamp 2012:18).
As mentioned in the previous chapter, Kraaikamp (2012; 2017) hypothesized that nominal gender based on individuation dates back to Indo-European, where nouns referring to objects/individuals would be masculine, nouns referring to collectives/abstracts would be feminine and nouns referring to masses would be neuter. Kraaikamp (2012) suggests that the semantic regularities found in Dutch nominal gender can be explained by the idea that Dutch, being of the Indo-European language family, used to have a nominal gender system based on the degree of individuation of the noun’s referent. However, this semantic basis became disrupted when gender stopped being a feature that modified the noun’s meaning and became a largely arbitrary feature that had to be lexically stored2. This semantic basis would explain why the nouns in the aforementioned examples that denote more specific objects and are higher in individuation are lexically common, while the nouns that have less bounded referents or denote a category and are lower in individuation are lexically neuter.
Apart from these semantic regularities, there are a few morphophonological patterns in derived nouns that can give some indication as to the gender of these nouns. For example, nouns ending with the suffix –isme are mostly neuter and nouns ending with the suffix –heid or –ing are mostly common (Blom & Polisenska 2008; de Vos & Vogelaer 2011). In addition, there is one unambiguous morphophonological rule that assigns gender, which applies to diminutives. All diminutivized nouns are neuter, regardless of the gender of their root nouns (Blom & Polisenska 2008).
Although there are semantic regularities as well as morphophonological patterns found in derived nouns that can provide some clarity on the gender of a given noun, these factors cannot by far predict the gender of all nouns and there are still many exceptions to the rules. Whether a noun is assigned a common or neuter gender in Dutch remains largely arbitrary and has to be learned on a case-by-case basis. As a result, from the (root) nouns themselves, children cannot gain sufficient information that there is such a thing as gender, and consequently, which gender a noun is assigned. Instead, information on gender has to be deduced from the different agreement patterns that common and neuter nouns take. This will be discussed in the next paragraph.
2Why and how this disruption happened exactly remains unclear, see Kraaikamp 2017
24 2.2 Lexical gender agreement
While the gender of a given noun is often not clearly visible from its semantic or morpho(phono)logical properties, nominal gender is “spelled out [...] on functional morphemes via the process of agreement” (Blom & Polisenska 2008: 260). In Dutch, the gender of a noun generally manifests itself indirectly through formal patterns of agreement on determiners, attributive adjectives and pronouns. Table (8) gives an overview of the gender marked forms:
Table (8). Overview of the Dutch gender marked forms.
Despite marking gender in elements that are in an agreement relation with nouns as shown in table (8), Dutch does not have a particularly transparent gender agreement system (Blom, Polisenska & Weerman 2008; Kraaikamp 2012, 2016, 2017; Audring 2016, 2009). To begin with, neuter and common nouns only trigger different agreement patterns when they appear in their singular form. All plural nouns, regardless of their lexical gender, take the common agreement patterns. In other words, all nouns except for singular neuter nouns (and in the case of the attributive adjective, only singular neuter nouns in an indefinite context) trigger the same agreement patterns.
Determiners Definite determiner (article) Demonstrative determiner Interrogative determiner Possessive Determiner (1PL) Collective determiner common de ‘the’ deze ‘this’ die ‘that’ Welke ‘which’ onze ‘our’ iedere/elke ‘every’ neuter het ‘the’ dit ‘this’ dat ‘that welk ‘which’ ons ‘our’ ieder/elk ‘every’
Attributive adjective (in indefinite contexts)
common -e suffix neuter no suffix Pronouns Relative pronoun Demonstrative pronoun Possessive pronoun Personal Pronoun (3SG) common die ‘that’ deze ‘this’ die ‘that’ Masc. Zijn ‘his’ Masc. Hij/hem ‘he/him’ Fem. Haar ‘her’ Fem. Zij/haar ‘she/her’ neuter Dat ‘that’ dit ‘this’ dat ‘that’ Zijn ‘his’ Het ‘it’
25
Moreover, as Unsworth (2008: 370) mentions, in Dutch “as for many other languages, gender attribution is based on the form of the definite determiner” (370). If gender attribution is based on the form of the definite determiner, gender assignment in Dutch becomes a
slightly unreliable undertaking. Table (8) shows that common nouns take the common definite article de (‘the’) and neuter nouns take the neuter definite article het (‘the’). However, as mentioned, all plural nouns, regardless of whether they are common or neuter, take the common agreement patterns and thus the definite article de (‘the’):
(8) a. het huis - de huiz -en
ART.DEF.N.SG house(N) - ART.DEF.PL house-PL
‘the house’ - ‘the houses’
b. de auto - de auto-s
ART.DEF.C car(C) - ART.DEF.PL car -PL
‘the car’ - ‘the cars’
Thus, while common nouns always take the common definite determiner, neuter nouns can be encountered with both the neuter definite article and the common definite article (plural nouns). In addition, an estimate of 75% of all Dutch nouns are common and only around 25% of all nouns are neuter nouns (Hulk & Cornips 2006; Blom et al. 2008). Van Berkum (1996), in his analysis of the CELEX database, found that in running texts, common nouns are generally overrepresented in the Dutch vocabulary by a ratio of 2:1 considering type
frequencies. This number is upped to 3:1 when token frequencies are counted. When taking into account that neuter nouns can appear with both the neuter and the common definite article, and that common nouns are already overrepresented in the language input, the definite article de (‘the’) predominates in the language input. This distribution of common and neuter nouns in the Dutch language together with the distribution of common and neuter definite articles for neuter nouns makes the distinction between common and neuter gender quite opaque and reduces the visibility of lexical gender as a whole.
Furthermore, nominal gender is not (clearly) marked on other adnominal elements (as presented by Audring 2014: 16). Firstly, the indefinite article een (‘a’) does not mark gender at all and can be used for both common and neuter singular nouns without any morphological adjustments (9)
26
(9) a. een huis - een auto
ART.INDF. house(N) - ART.INDF. car(C)
‘a house’ - ‘a car’
Secondly, as can be seen in table n, attributive adjectives mark neuter gender but only with neuter singular nouns in an indefinite context. In all other contexts, the attributive adjective has an –e suffix (table taken from Blom et al. 2008: 304):
Context suffix Example
Definite, neuter, singular -e Het mooie huis ‘The nice house’ Indefinite, neuter, singular -ø Een mooi huis ‘A nice house’
Definite, common, singular -e De mooie auto ‘The nice car’
Indefinite, common, singular -e Een mooie auto ‘A nice car’
Definite, neuter, plural -e De mooie huizen ‘The nice houses’
Indefinite, neuter, plural -e Mooie huizen ‘Nice houses’
Definite, common, plural -e De mooie auto’s ‘The nice cars’ Indefinite, common, plural -e Mooie auto’s ‘Nice cars’
Table (9). Attributive adjectival flexion in Dutch.
Furthermore, relative, personal and demonstrative pronouns show variation in the amount of genders they distinguish. The relative pronouns and demonstrative pronouns distinguish between neuter and common gender, whereas personal pronouns can be masculine, feminine, or neuter. Moreover these different types of pronouns show different types of agreement with their antecedent. Following Corbett’s (1991) Agreement Hierarchy, the relative pronoun is syntactically closer to its antecedent and less likely to show semantic gender agreement than the personal (and demonstrative) pronouns, which indeed appears to be the case in Dutch (Audring 2009). When referring to the same noun, the types of agreement in these types of pronouns can differ, resulting in a distribution of various gendered pronouns. (9) illustrates this variation (example taken from Audring 2016: 26):
(9) Mijn broertje die was toen was ‘ie nog klein My brother-DIM(N) DEM.C was then was 3SG.M still small
27
The diminutive broertje (‘little brother’) is neuter but triggers common gender agreement on the demonstrative pronounand masculine gender agreement on the personal pronoun.
All in all, there are several problematic sides to lexical gender and gender agreement in Dutch. As has been discussed in the previous paragraph, the gender of nouns is (often) not visible from the noun’s semantic or morphophonological properties and has to be lexically stored. Gender is, however, visible through the process of agreement it triggers in other linguistic elements. Yet, the gender of nouns is only visible on singular nouns and even for singular nouns gender is not always clearly marked in their agreeing elements. Definite articles only clearly mark gender in singular nouns and indefinite articles do not mark gender at all. Attributive adjectives only mark a distinction between common and neuter in indefinite contexts, and pronouns show variation in gender. Hence, the saliency of lexical gender per se in Dutch is rather low. Unsurprisingly, the acquisition of lexical gender has been found to be a lengthy undertaking for children. The next paragraph will discuss previous research on how and when children acquire lexical gender.
2.3 Previous research on gender acquisition
The previous paragraphs have shown that the lexical gender of a given noun in Dutch (mostly) becomes visible through its agreeing elements. Moreover, Corbett’s Agreement Hierarchy states that attributive elements, such as definite articles and attributive adjectives, are most likely to agree with the lexical gender of a noun. Consequently, when looking at the acquisition of nominal gender, it has to be considered when the feature [gender] on these attributive elements is acquired.
Previous research has shown that grammatical gender is a difficult feature to acquire in the Dutch language, that is, L1 Dutch children do not fully master grammatical gender until a relatively late age. Van der Velde (2004) compared the acquisition of definite articles in L1 Dutch children and L1 French children of the ages 3, 4 and 6. The results showed that the L1 Dutch children overgeneralize the common definite article de (‘the’) with neuter nouns, which require the neuter definite article het (‘the’), until at least the age of 6. The L1 French
children, on the other hand, did not make gender errors at all.
This overgeneralization of the common definite article de (‘the’) with neuter nouns in L1 Dutch children was also found by Blom et al. (2008a). Blom et al. (2008a) looked at the acquisition of grammatical gender in definite articles an attributive adjectives in (amongst other groups) Dutch monolingual children from the ages 3-7. By means of a sentence
28
they found that Dutch monolingual children overgeneralize the common definite article de ‘the’ with neuter nouns until (at least) the age of 7. Moreover, this overgeneralization was (mostly) unidirectional. Although the neuter definite article was occasionally used to modify common nouns (mostly by the older children), these occurrences were infrequent. This unidirectional overgeneralization was also visible for the attributive adjectives.The
participants overgeneralized the inflected attributive adjective until (at least) the age of 7. The bare adjective (which modifies only singular neuter nouns in an indefinite context), on the other hand, was never incorrectly used. These results indicate that the [common] gender value is initially the default gender, resulting in the overgeneralization of common agreement patterns. Blom et al. (2008a) did observe a parallel development path of the correct
distribution of the definite article het (‘it’) and the bare adjective between the ages 3 and 7. Hence, it appears that in this stage during the acquisition process, children become aware of the more specified [neuter] gender value. Once the feature [±neuter] (Blom et al. 2008a call this the [gender] feature) is activated, it seems to influence the linguistic performance in both the use of the definite article and the use of the attributive adjective, hence in multiple gender domains. However, at the age of 7, the participants still performed “below Brown’s (1973) criterion, which states that a grammatical morpheme is acquired once it appears in at least 90% of the obligatory context” indicating that the grammatical gender feature was not sufficiently acquired (Blom et al. 2008a: 23).
Although it is not clear at which precise age monolingual children acquire the feature [gender] a study by Hulk & Cornips (2006) indicates that the feature [gender] has at least been acquired at the of 9;3.
Thus, studies have shown that the gender feature is a difficult feature to acquire. Until the age of (at least) 6/7, monolingual children have not fully mastered grammatical gender on definite articles and attributive adjectives. This feature appears to become acquired
somewhere between the ages 7-9.
2.4 Implications for this study
The previous paragraphs have shown that whether nouns are assignment neuter or common gender is rather arbitrary and the gender of a given noun becomes clear through the form of its agreeing adnominal elements, mainly through the common or neuter definite article that a noun takes. However, definite articles only mark gender for singular nouns and the common definite article de (‘the’) is overrepresented in the input. Accordingly, previous research on gender acquisition found that children until (at least) the age of 6/7 have not fully acquired
29
grammatical gender on definite articles. If the gender feature on definite determiners, which most clearly indicate the lexical gender of a given noun, is not fully acquired until (at least) the ages of 6/7, monolingual Dutch children of and before the aforementioned ages have little to no evidence indicating that there is such a thing as grammatical gender, and hence a
difference in lexical gender between common and neuter nouns.
Hence, in the present study, it is unlikely that the younger participants (aged 5;1 – 6;5) will distribute masculine and neuter pronouns based on the lexical gender of the antecedent noun. If the lexical gender of the noun plays a role in the pronominal reference system of children, lexical gender agreement might be found with the older age group (aged 7;8 – 8;8), as they could have acquired the gender feature.
30 3. Semantic features in Dutch
The first chapter has shown that when pronominally referring in Dutch, the degree of individuation of the noun’s referent can influence the choice in pronoun.For animate
referents, the boundary between the use of both the masculine and feminine pronouns, and the use of the masculine pronoun seems to be situated between nouns denoting animates and nouns denoting other animates. Audring 2009; Kraaikamp 2012). For inanimate referents the pronominal gender carries between masculine and neuter gender, where the boundary
between the domain of masculine and neuter gendered pronouns is situated somewhere between bounded (countable) referents, which take masculine pronouns, and unbounded (non-countable/mass) referents, which take neuter pronouns.
To be able to base the distribution of neuter pronouns on the degree of individuation of the noun’s referent, or its countability, a child must have knowledge of a distinction between count nouns and mass nouns. Hence, the semantic features [±count] has to have been
acquired. If this feature is not acquired, a child cannot be expected to show semantic agreement in pronominal reference based on the countability of the noun’s referent. Consequently, when looking at pronominal reference by Dutch children of different age groups, it is important to know how the distinction between count nouns and mass nouns manifest itself in Dutch and in which stage during acquisition, that is, at what age a child acquires the feature [±count].
The following paragraphs the difference between Dutch count and mass nouns and the acquisition of the mechanisms underlying this count – mass distinction.
3.1 The count – mass distinction
Dutch is a number marking language where the mass – count distinction is reflected in the singular and plural morphology of the noun (Witteloostuijn 2013; ). There are several
syntactic cues that indicate this count – mass distinction. Count nouns are assumed to have the following properties (here illustrated with the count noun beker ‘cup’):
(i) they can be pluralized; beker - beker-s cup - cup -PL
‘cup - cups’
(ii) they can be preceded by numerals; twee bekers
31
NUM. cups
‘two cups’
(iii) they can be preceded by an indefinite article. een beker
ART.INDF. cup
‘a cup’
Mass nouns, on the other hand, are assumed to have the following syntactic properties, her illustrated with the mass noun sneeuw (‘snow’) (Chiercia 1998; Krifka 2007; Witteloostuijn 2013):
(i) They cannot be pluralized;
*Heb je de sneeuw-en buiten gezien? Have you the snow-PL outside seen?
*‘Have you seen the snows outside’? (ii) They cannot be preceded by numerals; * Er ligt twee sneeuw buiten There lies NUM. snow outside
* ‘There is two snow outside’.
(iii) They cannot be preceded by an indefinite article;
* Er ligt een sneeuw buiten
There lies ART.INDF snow outside
* ‘There is a snow outside’.
(iv) They need classifier or measure phrases to be quantized. Er ligt veel sneeuw buiten There lies much snow outside ‘There is a lot of snow outside’
At first glance, there appears to be clear evidence available for a classification of Dutch nouns in count nouns an mass nouns in the input to children, facilitating the acquisition of a count – mass distinction. However, as Wittenloostuijn (2013) points out, using the aforementioned properties as syntactic cues distinguishing count nouns from mass nouns is not completely
32
unproblematic. She mentions several aspects of these properties that can be confusing when trying to differentiate between mass and count nouns, three of which apply to Dutch nouns. Firstly, some mass nouns seem to be able to occur in a plural form. For example, the noun water ‘water’ in Dutch is a mass noun. Yet, a plural form of this noun, wateren
(‘water-PL’) also exists. However, this alters its meaning such that it is interpreted as quantifying over
individuals” (Witteloostuijn 2013: 20). The plural form wateren (‘water-PL’) can then be
interpreted as a collective form for different bodies or kinds of water, in which the meaning of the mass noun shifts to a “count” noun. That is, in this case, the noun behaves syntactically and semantically as similar to a count noun.
Secondly, in general mass nouns cannot be modified by numerals, unless “an
expression is added that indicates a unit of counting or a measure” (also known as a ‘measure term’), as in the utterance twee glazen appelsap (‘two glasses of apple juice’) (Witteloostuijn 2013: 21). However, there are exceptions to this rule. In Dutch, mass nouns that have some sort of standard serving of the relevant mass noun can be used in combination with a numeral without using a measure term. It is not uncommon for Dutch speakers to order twee appelsap (‘two glasses of apple juice’; lit. ‘two apple juice’) instead of twee glazen appelsap (‘two glasses of apple juice’). Although the noun appelsap (‘apple juice’) in this context
syntactically behaves as a mass noun (it does not take pluralisation as count nouns do when preceded by a numeral), it gets a count interpretation.
Thirdly, although there are quantifiers that can only be used to modify either mass or plural count nouns, some quantifiers can be used to modify both mass and plural count nouns. Examples of these quantifiers are veel (‘much/many’), weinig (‘little/few’), minder
(‘less/fewer’), een (hele)boel (a lot). Hence, from the use of these quantifiers alone it is not possible to distil whether the noun it modifies is a plural count noun or a mass noun.
These examples show that there is not always a clear syntactic binary split between count nouns and mass nouns. Witteloostuijn (2013: 25) provides the following overview of availability of syntactic cues for the acquisition of the count – mass distinction in Dutch:
Pluralization Numerals Indefinite article Quantifiers
+ ± + ±
Table (10). The available syntactic cues for the acquisition of the count – mass distinction in Dutch
Following this overview, the morphosyntactic mechanisms that most clearly underlie the count – mass distinction is whether or not a noun can be pluralized and preceded by an
33
indefinite article. As has been previously discussed, generally, count nouns can do both, while mass nouns can do neither.
When considering the acquisition of the count – mass distinction it is useful to look at previous research focussing on the acquisition of these two morphosyntactic mechanisms to see at what age the use of definite articles with count nouns and pluralisation starts to play a role for children in distinguishing count nouns from mass nouns. The next paragraph will go further into previous research on these morphosyntactic mechanisms in particular and on the acquisition of the count – mass distinction in general.
3.1.1 The acquisition of the count – mass distinction
According to Roodenburg & Hulk (2009: 157), Dutch children go through several stadia when acquiring indefinite articles. In the first stadium children only utter bare nouns, without articles all together. Subsequently children start to produce a sjwa (/ǝ/), which can be
analyzed as either the indefinite article een (‘a’) or the common definite article de (‘the’). In the next stadium children start to the indefinite article een (‘a’) and are aware of which nouns can be preceded by this definite article and which nouns cannot. Initially the production of the indefinite article with count nouns might function as a numeral rather than a definite article. Nonetheless, at this stage of the acquisition children become aware that the feature [count] can be used to divide nouns into count nouns and mass nouns.
Roodenburg & Hulk (2009) do not specify at which age children generally go through this different stadia. However, a study by Cornips & Hulk (2013), which initially aimed to elicit the distribution of common and neuter definite determiners, found that children aged 4;2 – 5;11 also produce the indefinite determiner een (‘a’) in combination with both common nouns and neuter nouns. Furthermore, all participants only used the indefinite determiner in combination with [+count] nouns and (almost) never with [-count] nouns (only 1 out of 210 tokens) nor with plurals (Hulk & Cornips 2013: 62). These results suggest that children from (at least) the age of 4 can distinguish [+count] nouns from [-count] nouns.
The feature [+count] most likely also plays a role in the acquisition of pluralization, since only count nouns can be pluralized (Roodenburg & Hulk 2009). In Dutch the plural can be marked by either the suffix –en or the suffix –s. Research by Schaerlaekens (1977) shows that children start producing plurals from the age of 2. Although children tend to use the wrong suffix for irregular plural forms until the age of 6, regarding the pluralization of regular forms, children appear to perform adult-like at the age of 4 (this was found for Flemish
34
produce plurals at a young age, although children tend to overgeneralize the plural marker –en until the age of 5, they have seem to acquired plural forms at the age of 3.
Moreover, Witteloostuijn (2013) looked at the interpretation of different mass nouns and plural (count) nouns in Dutch children aged 4;1 – 12;6 to see in which case and at which age children would base their quantity judgment for mass nouns on overall volume and for plural count nouns on number. By presenting these children with a quantity judgment task (who has more X), Witteloostuijn (2013) found that children at the age of 4;1 become sensitive to the count – mass distinction between plural nouns specifically denoting objects and substance mass nouns. At the age of 4;1 children were aware that more X quantified over individuals for plural count nouns and quantified over overall mass for substance mass nouns.
Aditionally, Witteloostuijn (2013) found that children at the age of 6;2 become sensitive to the count – mass distinction when this distinction only relies on morphosyntax. That is, at the of 6;2, children become aware that nouns which can either be interpreted as count nouns or as mass nouns (depending on the presence or absence of the plural marker) quantify over individuals when they are pluralized but quantify over overall volume when the plural marker is absent. These results indicate that children are more sensitive to a count – mass distinction when this distinction is aided by a conceptual object – mass distinction.
The abovementioned studies show that Dutch children around the age of 4 are aware of the [count] feature in the nominal domain. They have acquired the plural suffix and can interpret its connection to nouns quantifying over individuals (for nouns denoting concrete countable objects). Moreover, around the age of 4 children only use the indefinite article een (‘a’) for singular [+count] nouns, but never for plural nouns or [+mass] nouns.
3.2 Implications for this study
The knowledge of [+count] and [-count] in the nominal domain is a prerequisite for the distribution of pronouns based on the [count] feature in the pronominal domain. The previous studies on the count – mass distinction suggest that Dutch children have acquired the
phenomena involving the [count] feature (pluralization and the use of the indefinite
determiner with count nouns) around the age of 4. Hence, around this age, Dutch children are aware of the feature [count] in the nominal domain and the morphosyntactic mechanism that this feature brings about.
Whereas it is less likely that children of aged 5;1 – 6;5 and the children aged 7;8 – 8;8 have acquired lexical gender, they can be assumed to have acquired the feature [±count]. As the research by Audring (2009) shows, for adults, the [±count] feature increasingly plays a
35
role in the distribution of the masculine personal pronouns and the neuter personal pronouns. Assuming that the role of the [±count] feature in the pronominal domain is present in the input of children increases the likelihood that children will base their choice in pronoun on the [±count] feature rather than on the lexical gender of the noun.
