Gender (in)stability on Dutch relative pronouns
Ma Thesis European Linguistics
Rijksuniversiteit Groningen
Name Student: Louke Nadine Boelens Student Number: 1489100
E-mail: loukeboelens@hotmail.com
Supervisor: dr. A.M.H. van Hout
Faculty: Faculty of Arts
Module: Master’s Thesis Linguistics
Code: LTX998M20
Date: 29 September 2017
Abstract
Audring (2006, 2009) found in a corpus analysis of spoken Dutch that native speakers apply semantic agreement to anaphors (personal pronouns and demonstratives). She also found mismatches in gender between nouns and relative pronouns that were consistent with the system of semantic agreement of anaphors. This paper investigates how consistently Dutch native speakers apply syntactic agreement to relative pronouns and if they accept mismatches, whether this is consistent with semantic agreement. In an online questionnaire with a rating task, Dutch native speakers (n = 87) listened to four types of sentences containing a noun referring to an object: items with a syntactic match between noun and relative pronoun for common gender nouns (de tafel, die… “the.C table.C, which.C) and neuter gender
Table of contents
1 Introduction 3
2 Background
2.1 Agreement 4
2.2 Gender agreement in Dutch 7
2.3 Gender (in)stability in Dutch 19
1 Introduction
For most native speakers of northern Standard Dutch as it is spoken in the Netherlands, nouns have two genders: common and neuter. Historically, Dutch had three genders: masculine, feminine and neuter. Only personal and possessive pronouns still distinguish masculine and feminine. Dutch nouns have covert gender. That is, the gender of a noun cannot, in most cases, be deduced by morphological or phonological rules. Dutch speakers, therefore, gain their knowledge about a noun’s gender from agreeing elements, most notably the determiner. Since attributive modifiers have lost distinctive gender markings for masculine and feminine, the two genders have collapsed into the common gender. As a result, gender selection for personal pronouns has shifted from syntactic agreement to predominantly semantic agreement based on individuation (Audring, 2009).
Selecting the gender of agreement targets on the basis of semantic concepts rather than on the basis of the gender of nouns is far from unique in Germanic languages. English, having lost all gender distinctions on nouns, select personal pronouns on natural sex. Standard varieties of Danish and Swedish, languages in which the masculine and feminine genders have also collapsed into a common gender, use masculine and feminine personal pronouns exclusively in animate reference and the neuter and common demonstratives are used in inanimate reference according to the lexical gender of the noun (Audring, 2009, pp. 104, 186). In Dutch however, common demonstratives can be used in inanimate reference with neuter gender nouns and neuter personal pronouns and demonstratives can be used in inanimate reference with common gender nouns.
The frequency of semantic agreement compared to syntactic agreement seems to be related to the stability of lexical gender (Audring, 2009; Kraaikamp, 2012). The less stable lexical gender is, the more semantic agreement is used. Since Dutch has covert gender, the stability of lexical gender depends on the stability of gender on attributive agreement targets. Research in language acquisition among native speakers in minority communities (Blom, Polišenká, & Weerman, 2008; Cornips, Van der Hoek, & Verwer, 2006; Hulk & Cornips, 2006) suggests that gender in the attributive domain might become less stable. This might then lead to an increase in semantic gender agreement or to loss of gender distinctions. If the former is the case, semantic agreement is likely to be used on relative pronouns according to Corbett’s Agreement Hierarchy (1979, 1991). Indeed, Audring (2009) found mismatches in gender between nouns and relative pronouns that are consistent with the system of semantic agreement on personal pronouns based on individuation. However, relative pronouns are scarce in spoken language and therefore Audring’s data on relative pronouns was limited. The question is whether these mismatches are indeed the result of semantic agreement or if they are a result of gender instability.
This paper provides insight into the status of gender on relative pronouns. There are two main concerns: 1) how strictly do native speakers of Dutch apply syntactic gender agreement on relative pronouns and 2) are allowed mismatches consistent with semantic agreement based on individuation? In addition, this paper investigates whether sex, age, education, and regional background are factors in the (un)acceptability of mismatches. By means of an online questionnaire with a rating task, native speakers were asked to listen to and judge sentences containing relative pronouns that for 50% did not match the gender of the noun it belonged to.
2 Background
2.1 Agreement
Agreement is “systematic covariance between a semantic or formal property of one element and a formal property of another” (Steele, 1978, p. 610, cited in Audring, 2009, p. 15). That is, the formal properties of one or more elements, the agreement targets, adapt to the semantic or formal properties of the agreement controller. If the value of an agreement feature changes on the agreement controller, the agreement targets change accordingly. 1 illustrates this.
(1) a. il nuovo quadro
the.M.SG new.M.SG picture.M.SG
“the new picture”
b. i nuovi quadri
the. M.PL new. M.PL picture.M.PL
“the new pictures”
c. la nuova tela
the.F.SG new. F.SG painting. F.SG “the new painting”
d. le nuove tele
the. F.PL new. F.PL painting.F.PL
“the new paintings”
(Corbett, 2006, p. 9, ex. 14-17) In 1a-d, the nouns quadro (“picture”) and tela (“painting”) are agreement controllers of the features gender and number. The definite determiner and the attributive adjective are agreement targets of the features gender and number. Since the noun quadro in 1a has the values masculine and singular, so do the definite determiner and the attributive adjective. In 1b-d, the nouns have different values than the noun in 1a and the definite determiner and the attributive adjective adapt accordingly. However, the meaning of the definite determiner and the attributive adjective do not change. The meaning of the noun does change. Whereas in 1a and c there is one picture and one painting respectively, in 1b and d there is more than on picture or painting respectively. So the values of an agreement feature, in 1 the features gender and number, are determined by the agreement controller, in 1 the noun. The agreement targets duplicate those values.
Agreement exists on multiple syntactic levels: it occurs within the noun phrase, within the clause (outside the noun phrase), and across clause boundaries. Example 1 above is an example of agreement within the NP. An example of agreement within the clause is number agreement between subject and verb or gender agreement between subject and predicative adjective. 2 gives an example of subject-verb number agreement in Dutch.
(2) a. Het glas staat op tafel.
b. De glazen staan op tafel. the.N.PL tables.N.PL stand.PL on table “The glasses are on the table.”
In 2, the agreement controller is the subject. The verb changes according to the values of the subject. Agreement across clause boundaries occurs on anaphoric personal pronouns. In English, nouns do not specify gender, so personal pronouns take the natural gender of the referent. However, in languages in which nouns do distinguish lexical genders, personal pronouns may or must take the gender of the agreement controller, the antecedent noun. For example, French Majesté (“majesty”) is formally a feminine noun and often takes a feminine personal pronoun even though the referent is often male (3).
(3) Votre Majesté partira quand elle voudra.
your.PL majesty.F.SG leave.FUT.3.SG when she.F.SG wish.FUT.3.SG
“Your Majesty will leave when he (lit.: she) wishes.”
(Voltaire, quoted by Grevisse, 1964, p. 406, cited in Corbett, 1991, p. 227, ex. 3, and 2006, p. 229, ex. 45) The examples above are all examples of syntactic agreement. However, agreement can also occur on the basis of semantic concepts: semantic agreement. For example, although a feminine personal pronoun is preferred with Majesté, as in 3, it is also acceptable to use a masculine personal pronoun (4).
(4) Sa Majesté fut inquiète, et de nouveau
her.F.SG majesty .F.SG was worried.F.SG and of new
il envoya La Varenne à son ministre.
he.M.SG sent La Varenne to his.M.SG minister “His Majesty was worried, and again he sent La Varenne to his minister.”
(J. & J. Tharaud, quoted by Grevisse 1964, p. 405, cited in Corbett, 1991, p. 227, ex. 4) Another example of semantic agreement is given in 5. The police in 5 is formally a singular noun that would trigger singular agreement on the verb. However, conceptually, the police is a collective of police officers. Therefore, it can trigger plural agreement on the verb.
(5) Catalan police face their own test of independence
(Minder, 2017) Semantic agreement also exists on so-called hybrid nouns. These nouns have both a lexical gender and a semantic gender. A prime example is the German word das Mädchen (“the.N little.girl.N”). The lexical gender of Mädchen is neuter, as all diminutive nouns in German are. However, the semantic gender is the referent’s natural gender: feminine. Such a hybrid noun can trigger both genders on certain agreement targets. In this specific example, Mädchen can trigger both neuter and feminine gender on relative pronouns and personal pronouns, but not on determiners (6)1.
1
(6) Guck mal, das/*die Mädchen, das/die da läuft, kennst du Look once the.N/F girl.DIM.N REL.N/F there walks know you
die/das?
DEM.F/N
Look at that girl, who’s walking over there, do you know her?”
(Audring, 2013, p. 40, ex. 18) The likelihood of semantic agreement is not the same for every kind of agreement target. In a cross-linguistic study, Corbett (1979) investigated the use of semantic agreement on different kinds of agreement targets. He found that personal pronouns are more likely to take semantic agreement than relative pronouns. Predicative agreement targets are even less likely to take semantic agreement. Attributive agreement, such as agreement on determiners and on attributive adjectives, are the least likely to take semantic agreement. These differences are expressed in an agreement hierarchy (7). (7) attributive – predicative – relative pronoun – personal pronoun
(Corbett, 1979, 1991, 2006) If in a specific language any element on this agreement hierarchy can take semantic agreement for a specific agreement feature, then any element to the right can do so too. This hierarchy explains the example of Mädchen in 6, where the personal pronoun and relative pronoun can agree semantically, but the determiner, an attributive agreement target, agrees syntactically. Moreover, any element on the hierarchy that can take semantic agreement, will do so as frequently or more frequently than any element to its left. For example, the Russian vrač (“doctor”) is a masculine noun. However, in case of a female doctor, agreement targets can take the feminine gender (8 and 9, Corbett 1991).
(8) a. Ivanova – xoroš-ij vrač Ivanova good.M doctor.M
“Ivanova is a good doctor.”
b. Ivanova – xoroš-aja vrač Ivanova good.F doctor.M
“Ivanova is a good doctor.”
(Corbett, 1991, p. 231, ex. 14-15)
(i) a. das Mädchen, das ich gesehen habe…
the.N girl.N REL.N I seen have
“the girls I saw…”
b. Schau dir dieses Mädchen an, wie gut sie/es Tennis
look you DEM.N girl.DIM.N at how good she/it tennis
spielt. plays
“Do look at this girl, see how well she plays tennis.”
(9) a. vrač prišel-ø doctor.M came.M
“The doctor came.”
b. vrač prišel-a
doctor.M came.F
“The doctor came.”
(Corbett, 1991, p. 231-232, ex. 16-17) For the attributive agreement target, the attributive adjective in 8, Panov found that 16.9 percent of 3,835 respondents preferred 8b over 8a (1968, cited in Corbett, 1991). For the predicative agreement target, the verb in 9, the same study found that 51.7 percent of 3,806 respondents preferred 9b over 9a. Corbett found that these patterns hold cross-linguistically.
Grammatical distance or a distinction such as within and beyond the NP is insufficient to derive the agreement hierarchy (Corbett, 1991, p. 238). Although the attributive domain, the predicative domain, and the domain of personal pronouns do follow such distinctions, the domain of relative pronouns show that such generalizations are inadequate. Although relative pronouns cross clause boundaries, they exist within the NP. If such distinctions would be applied, the relative pronoun should either be included in the attributive domain or placed between the attributive and predicative domains. The agreement hierarchy, however, is built from cross-linguistic patterns that indicate that the relative pronoun is between the predicative domain and the domain of personal pronouns.
The agreement hierarchy can also be utilised to describe the process of changes in the gender system (Corbett, 1991, pp. 248-257). Changes in the gender system are instigated by semantic agreement on the personal pronoun. Semantic agreement can then move leftward over time along the agreement hierarchy. Ultimately, it can reach the attributive domain, after which the new system can be included in the (syntactic) gender system. However, it is not necessarily so that once semantic agreement is possible on personal pronouns it will eventually also be possible in the other domains. Neither is it necessarily so that semantic agreement on personal pronouns will inevitably lead to changes in the syntactic gender system.
Whether or not gender agreement occurs in a given domain depends on the language. In English, gender agreement is only found in the domain of personal pronouns. This paper will focus on gender agreement in Dutch.
2.2 Gender agreement in Dutch 2.2.1 Basic agreement in Dutch
In Dutch, gender agreement exists on attributive modifiers, relative pronouns, and personal pronouns, but not in the predicative domain. For most Dutch speakers, Dutch has only two genders: common and neuter. These genders, though not visible on the noun itself, are seen on agreement targets. Because the distinction between common and neuter nouns is visible on the definite determiner (10), common gender nouns are often called de-words (“the.C-words”) and neuter gender nouns het-words (“the.N -words”).
(10) a. de boot
b. het schip the.N ship.N
Gender distinctions on agreement targets in Dutch constitute a subtly complex system. For example, gender is marked on the determiner, but only on the definite determiner and not the indefinite one. In contrast, gender is marked on the indefinite but not on the definite attributive adjective. In addition, gender is only marked in the singular, not in the plural. Table 1 shows the paradigms for the determiner and attributive adjective in the singular.
Common Neuter
Determiner
Indefinite Een boot
A boat Een schip A ship Definite De boot The boat Het schip The ship Attributive Adjective
Indefinite Een groen-e boot
A green boat
Een groen-ø ship A green ship Definite De groen-e boot
The green boat
Het groen-e schip The green ship
Table 1. Gender agreement in Dutch on singular determiners and singular attributive adjectives.
Other agreement targets within the NP have a clearer two-way distinction. Table 2 gives an overview.
Common Neuter Quantifiers Elke boot Each boat Elk-ø ship Each ship Iedere boot Every boat Ieder-ø schip Every ship 1st person plural possessive pronoun Onze boot Our boat Ons-ø schip Our ship
Wh-element Welke boot
Which boat Welk-ø schip Which ship Proximal Demonstrative Pronoun Deze boot This boat Dit schip This ship Distal Demonstrative Pronoun Die boot That boat Dat schip That ship
Relative Pronoun …boot, die…
…boat that…
…schip, dat… …ship that…
Table 2. Gender agreement in Dutch within NP with clear two-way distinction.
the personal pronouns. Audring points out that personal pronouns are quite inconsistent in gender choice and make poor gauges. This paper will follow Audring: since attributive agreement targets in Dutch only distinguish two genders, nouns are also taken to distinguish only two genders.2
The original distinction between masculine and feminine is still visible on the third person singular possessive and personal pronouns (table 3). Personal pronouns have distinct forms for each of the three genders in the singular: masculine, feminine and neuter. Possessive pronouns, on the other hand, have a distinct form for the feminine singular, but the masculine and neuter singular possessive pronouns have the same form.
Masculine Feminine Neuter
Possessive Pronoun3 Full form Clitic zijn z’n haar d’r/’r zijn z’n Nominative Personal Pronoun Full form Clitic hij ie zij ze het ’t Oblique Personal Pronoun Full form Clitic hem ’m haar d’r/’r het ’t
Table 3. Gender agreement in Dutch on third person singular possessive and personal pronouns.
It is not uncommon for a language that suffers gender loss on the noun to retain the original gender system on personal pronouns (Corbett, 1991, p. 259).
2.2.2 Anaphoric semantic gender agreement in Dutch
Since Dutch nouns have only two genders and personal pronouns have three genders, the selection of personal pronouns cannot rely on syntactic agreement. In Dutch, sometimes demonstratives, which differentiate the same genders as nouns, are used instead of personal pronouns. It would seem an easy solution to the problem. Swedish and Danish, languages in which the former masculine and feminine gender have also collapsed into a common gender, in fact use demonstratives anaphorically according to the gender of the noun for inanimate nouns. Masculine and feminine personal pronouns in Swedish and Danish are exclusively used for animates (Corbett, 1991, p. 247; Audring, 2009, p. 104, 186). However, instead of relying solely on demonstratives, personal pronouns are still used in Dutch for anaphoric reference to inanimate nouns. In addition, demonstratives are also used for animate reference.
The problem of selecting the appropriate personal pronoun seems at first sight to be limited to common gender nouns. Both nouns and personal pronouns have a neuter gender, so it would seem logical to refer to neuter gender nouns with a neuter personal pronoun. However, Audring (2006, 2009) noticed that switches in gender between nouns and anaphora happen in a variety of ways and
2
Though this is true for Standard Northern Dutch, some southern Dutch dialects do still distinguish between masculine and feminine. For speakers of these dialects therefore, masculine and feminine nouns still exist. In addition, in formal language and among the higher educated there is a tendency to appoint feminine gender to nouns based on morphological suffixes or to nouns denoting collectives.
3
If the possessor is human, possessive pronouns in Dutch always agree with the natural gender of the referent, even if the antecedent is neuter (ii).
(ii) Het meisje deed haar/*zijn pop in haar/*zijn tas.
the.N girl.DIM.N did her/its doll in her/its bag
not just for common gender nouns. Considering that the gender agreement controller, the noun, has two values and that the gender agreement targets, personal pronouns and anaphoric demonstratives, combined have four values, eight gender combinations are possible (11).
(11) Gender of noun Gender of anaphor Example of anaphor
Common Masculine Hij (he)
Feminine Zij (she)
Common Deze, die (this, that)
Neuter Het, dit, dat (it, this, that)
Neuter Masculine Hij (he)
Feminine Zij (she)
Common Deze, die (this, that)
Neuter Het, dit, dat (it, this, that)
In six of the eight possible combinations, there is a mismatch or switch in gender between the noun and the anaphor. In her corpus analysis of the CGN (“Corpus Gesproken Nederlands”, Corpus Spoken Dutch) consisting of approximately 500,000 words, Audring found that all these switches occur. An example of each combination with a switch is given in 12.
(12) a. Common controller, neuter target
Ik vind puree van echte aardappelen altijd lekkerder I find puree.C of real potatoes always tastier
want het is wat steviger.
because it is what firmer
“I always think puree from real potatoes is better because it is more firm.”
(from CGN session 683, cited in Audring, 2009, p. 98, ex. 27c) b. Common controller, masculine target
omdat die vogel zo z’n best doet. Dat doet
because DEM.C bird.N so his best does that does
ie zeker, hij fluit van ‘s ochtends vroeg tot
he.CL sure he whistles from in.the morning early until
‘s avonds laat. in.the evening late
“because that bird is so much doing its best. He’s doing that for sure, because he’s singing from morning till night.”
(from CGN session 518, cited in Audring 2009, p. 85, ex. 6b) c. Common controller, feminine target
Dan kan ik je dame slaan met mijn toren.
Maar ik kan je ook dat we ze even wat but I can you also that we her PRT what beetje gaan martelen.
little go torture
“Then I can beat your queen with my rook. But then I can… that we’ll torture her a bit.”
(from CGN session 432, cited in Audring, 2009, p. 94, ex. 22) d. Neuter controller, common target
Dat meisje is dus eenentwintig en die gaat
DEM.N girl.DIM.N is thus twenty-one and DEM.C goes
trouwen. marry
“So that girl is twenty-one and she’s going to get married.”
(from CGN session 631, cited in Audring, 2009, p. 101, ex. 32b) e. Neuter controller, masculine target
Ken je ’t liedje niet? Hij is echt
know you the.N.CL song.DIM not he is really
leuk. ‘k heb ‘m één keer gehoord. Ik vond
nice I have him.CL one time listened I found
‘m echt leuk. him.CL really nice
“Don’t you know that song? I heard it once and I thought it’s really great.”
(from CGN session 686, cited in Audring, 2009, p. 87, ex. 12) f. Neuter controller, feminine target
Mijn zusje is dit jaar geslaagd dus ‘k ging
my sister.DIM is this year.N passed so I.CL went
effe met ‘r mee d’r cijferlijst halen.
PRT with her.CL with her.CL mark.list.C get
“My sister has passed her exams this year, so I went with her to get her marks list.” (from CGN session 619, cited in Audring, 2009, p. 93, ex. 20b) So when referring to a neuter gender noun, the neuter gender personal pronoun or anaphoric demonstrative is not always selected, even though there seems to be no reason why such switches would occur.
used in reference to specific masses if the antecedent noun is of common gender. Feminine personal pronouns are used for female humans (12f) and occasionally for other female animates. Audring (2006, 2009) summarizes this pattern according to an individuation hierarchy, a variation on Silverstein’s (1976, cited in Audring, 2009) Animacy Hierarchy. The more individuated the referent, the more likely a common or masculine (or even feminine) anaphor is. The less individuated the referent, the more likely a neuter anaphor is. Table 4 shows this hierarchy and the genders an agreement target can take if it does not take the lexical gender of the noun. That is, if the anaphor agrees semantically rather than syntactically with its antecedent noun.
Semantics of the referent
Humans Other animates Bounded abstract and objects Specific masses Unbounded abstracts and unspecific masses Gender of pronoun Common Masculine Feminine Common Masculine Feminine Common Masculine Neuter Common Masculine Neuter
Table 4. Individuation hierarchy and gender choices for anaphora based on semantic agreement. Audring also looked at relative and possessive pronouns. Audring found that switches between antecedent noun and relative pronouns showed a similar pattern to that found for personal pronouns (2009, pp. 108-110). Although relative pronouns are infrequent in spoken Dutch (Audring found only 167 instances), she found 22 instances of switches in gender between nouns and relative pronouns. The switches from a neuter gender noun to a common gender relative pronoun occurred when the referent denoted a human (13a), a non-human animate (13b), or a bounded object or abstract (13c).
(13) a. een koorlid die een aanval krijgt
a choir.member.N REL.C a fit.C gets “a choir member that has a fit”
(from CGN session 773, cited in Audring, 2009, p. 108, ex. 47a)
b. een schaap die mond- en klauwzeer kan krijgen
a sheep.N REL.C mouth- and foot.disease can get “that’s a sheep that can get foot and mouth disease”
(from CGN session 822, cited in Audring, 2009, p. 108, ex. 47b) c. een speciaal programma downloaden die dat ondersteunt
a special.N program.N download REL.C that supports “download a special program that supports this”
(14) dat er geen apparatuur onbeheerd is achtergebleven that there no equipment.C unattended is left.behind
dat aan staat
REL.N on stands
“that no equipment is left unattended that’s switched on”
(from CGN session 252, cited in Audring, 2009, p. 109, ex. 48b) One instance of a switch from common to neuter, denoting a bounded abstract, goes against the pattern. In this case, Audring speculated that the choice for a neuter relative pronoun may have been influenced by an interfering neuter gender noun. A summary of the switches between noun and relative pronoun that Audring found according to the category on the individuation hierarchy is given in table 5.
Semantics of the referent
Human Other animate Object/Bounded Abstract Specific Mass Unbounded abstract/Unspecific mass Common gender antecedent 1 3 Neuter gender antecedent 3 1 14
Table 5. Number of switches between noun and relative pronoun according to category on the individuation hierarchy and the gender of the noun. From Audring, 2009, p. 109.
The only switches in possessive pronouns Audring found were possessive pronouns following a neuter gender noun referencing a female human. These possessive pronouns agreed with the natural gender of the referent.
This system of semantic gender agreement based on individuation is common cross-linguistically. In a review of pronominal languages (languages that only distinguish gender on personal pronouns), Audring found that almost all these languages use basic semantic distinctions such as male versus female, human versus animal, animate versus inanimate, and count versus mass (2009, chap. 10; 2008). Audring’s individuation hierarchy can also be used to describe semantic agreement in other Germanic languages, but that these languages differ in where the cut-off points between the genders are on the hierarchy. In some languages, such as Standard English, Standard Danish, and Standard Swedish, the division is between animate and inanimate: masculine and feminine are only used in animate reference according to the natural sex of the referent. However, in some English dialects, masculine pronouns can be used in reference to objects (Audring, 2009, p. 182 and references there). She found that in all Germanic languages semantic agreement is possible, but it is used to a greater extent in the Germanic languages where the attributive domain and the domain of personal pronouns do not have the same gender values (Audring, 2009, chap. 9).
2.2.3 Factors influencing semantic gender agreement
1979, 1991). If in a given language, semantic agreement occurs on both personal and relative pronouns, personal pronouns take semantic agreement more often than relative pronouns do. Audring found this is true for Dutch as well: in her subcorpus of the CGN, personal pronouns and anaphoric demonstratives switched 65% of the time and relative pronouns only 25% of the time (2009, p. 160). Corbett also found that several forms of distance influence the availability of semantic agreement: the further an agreement target is removed from its noun, the more likely semantic agreement becomes (1991, p. 239-240). This is already somewhat reflected in the agreement hierarchy, but it also holds for agreeing elements of the same class in the hierarchy. This distance effect is also seen in Dutch (Audring, 2009, pp. 164-167).
In addition, semantic agreement in Dutch occurs more often towards the outer edges of the individuation hierarchy (Audring 2009, pp. 167-168). So pronouns referencing humans or unbounded abstracts and unspecific masses show semantic agreement more often than pronouns referencing bounded objects. This factor is also in effect for usages of the same word in different contexts. An unspecific reading of a given word moves it towards the less individuated end of the hierarchy. For example, partner (“partner.C”) in 15a can take a neuter anaphor, but in 15b, when the anaphor references a specific person, it cannot.
(15) a. Hij heeft een partner. Dat had ‘ie eerder niet
he has a partner .C DEM.N had he.CL earlier not
gehad. had
“He’s got a partner. He didn’t have one before.”
b. Hij heeft zijn partner meegenomen. *Dat heeft
he has his partner.C with.taken DEM.N has
‘ie daar ontmoet. he.CL there met
“He’s brought his partner. *He’s met that there.”
(Audring, 2009, p. 106-107, ex. 44) On the other hand, if the speaker has a personal relation with the referent, it moves towards the more individuated end of the hierarchy. Therefore, unspecific readings and personal relations increase the likelihood of semantic agreement.
In addition to linguistic factors that influence the use of semantic agreement, there is also interspeaker variability. For Dutch, younger speakers use semantic agreement more often than older speakers (Audring, 2009, pp. 168-169; De Vogelaer & De Sutter, 2011). In addition, the use of semantic agreement with the Russian word vrač (“doctor”) was found to depend on the speaker’s education, the region in Russia the speaker came from, on the speaker’s age, and whether or not the speaker is an industrial worker (Corbett, 1991, p. 241). There are even indications female speakers might be more inclined to use semantic agreement with words such as Russian vrač (“doctor”) and German Mädchen (“girl.DIM”). Audring, fearing that region could be a factor on the use of semantic agreement, excluded southern and northern dialects from her corpus study, focusing on Dutch as it is spoken in the Randstad (an area in the west of the Netherlands). Region may indeed be of influence on the use of semantic gender agreement in Dutch: De Vogelaer and De Sutter (2011) compared the use of grammatical and semantic agreement on anaphors by Dutch speakers from East- and West-Flanders, two provinces in Belgium. They found that speakers from West-Flanders used semantic agreement more than Dutch speakers from East Flanders. Although the different methods, a corpus analysis of colloquial Dutch versus and elicitation task in a questionnaire, do not allow for a direct comparison, it seems that semantic agreement is used more often by Dutch speakers in the Randstad (56%, Audring 2009, p. 186) than by Dutch speakers in Flanders (22.9%, De Vogelaer & De Sutter 2001, p. 198).4 De Vogelaer and De Sutter show that the effect of region in their study correlates with the complexity of the gender system of determiners: Participants with less gender distinctions on determiners in their dialect used more semantic agreement on pronouns.
2.2.4 Non-anaphoric semantic gender agreement in Dutch
Corbett found that a change in gender system is usually instigated by personal pronouns. These changes then go left-ward along the hierarchy over time. It is, however, not inevitable that it should bleed through to the other domains (Corbett, 1991, pp. 248-257). This section will investigate whether semantic agreement exists in Dutch outside the domain of personal pronouns.
Semplicini (2012) found that the system of semantic agreement based on individuation also has resonance with double gender nouns. Double gender nouns have an unstable lexical gender and can take both genders. The general assumption is that the choice of gender of agreement targets of these double gender nouns is speaker dependent, heavily influenced by geographical background (Semplicini, 2012, pp. 136-137). However, Semplicini found that the choice for the gender of the determiner could also be motivated by pragmatic and semantic context. In order to rule out interspeaker variability, Semplicini looked at switches in gender between determiners by the same speaker. She found that a switch in gender on the determiner could be pragmatically motivated by topicality and newness of information or semantically motivated by the degree of individuation of the referent (16).
(16) A heb jij die matras nodig? Do you need that.C mattress?
B nee
No
4
A die op gang staat. Mmm die kun je altijd gebruiken hoor kun echt gewoon uh
that.C is in the hall. Mmm you can use that.C as you want for sure you can just simply uh
B ja?
Really?
A ja, dat is het logeermatras van ’t huis een beetje … […] yeah, that.N is the.N guest bed of the house a bit … […]
B nou dan gebruiken we die toch gewoon die matras die in de gang staat now so we use that.C simply that.C mattress that.C is in the hall
(CGN session fn 00068, cited in Semplicini, 2012, p. 166-167, ex. 26) Semplicini argues that speaker A in 16 uses the common gender to refer to a concrete matrass, but shifts to the neuter gender when A is referencing an abstract concept, that of the guest mattress. Semplicini found that switches in gender like the one in 16 can be explained by the differences of the referent between specific and generic, between concrete and abstract, and between bounded and unbounded, all three distinctions being closely related to degree of individuation. This evidence from double gender nouns might indicate that semantic agreement based on individuation is possible in Dutch in the attributive domain.
Moreover, Kraaikamp (2012) believes that individuation is the root of Germanic gender assignment. This basis has been obscured over time due to gender becoming a lexically stored feature instead of a productive feature and due to meaning changes (Kraaikamp, 2012, pp. 222-226). However, even though the semantic basis of gender has been obscured over time on nouns, she believes that it has been retained in Dutch because semantic gender based on individuation is also present elsewhere in Dutch. In her view, the system of semantic agreement based on individuation is not restricted to personal pronouns and anaphoric demonstratives. She argues that constructions as in 17 are also an expression of semantic agreement.
(17) a. De tuin is complete vernield.
the.C garden.C is completely ruined
Wie heeft dat gedaan?
who has DEM.N done
“The garden is completely ruined. Who did that?
(Kraaikamp, 2012, p. 208-209, ex. 14a)
b. De broek is nu grijs,
the.C pants.C is now grey
maar dat verandert nog door de verf.
but DEM.N changes PART by the dye.C
The pants are grey now, but that will change with the dye.”
In 17, neuter personal or demonstrative pronouns are used to refer to what Kraaikamp calls nonentities: “non-nominal antecedents that denote activities, processes, states, properties, events, or propositions” (Kraaikamp, 2012, p. 208). The demonstratives in 17 refer to an activity (a) and a state (b). Corbett (2001, pp. 204-205) also shows that the use of neuter gender is common in Indo-European languages in case of genderless agreement controllers such as clauses and infinitival phrases. Therefore, Kraaikamp extends Audring’s individuation hierarchy to include nonentities on the rightmost side, giving that class even less individuation than unbounded abstracts and unspecific masses (2012, p. 212).
According to Kraaikamp, semantic gender is also still visible in the nominal domain on nominalised adjectives, double gender nouns, and nouns of certain semantic classes (2012, pp. 214-219). In Dutch, an adjective can be nominalised by preceding it with a determiner. The determiner in such nominalisations can take the gender that corresponds to the gender of the antecedent noun or the gender that corresponds to its place on the individuation hierarchy (18).5
(18) Er was een dik boek en een dun boek.
there was a thick.N book.N and a thin.N book.N
Ik heb het/de dunne gekocht.
I have the.N/C thin bought
“There was a thick book and a thin book. I bought the thin one.”
(Kraaikamp, 2012, p. 215-216, ex. 21b and 22) If there is no antecedent noun, the gender of the determiner always corresponds to its place on the individuation hierarchy (19).
(19) a. De lamp verandert van kleur. the.C lamp.C changes of colour.C
Dat is het leuke ervan.
that is the.N fun of.it
“The lamp changes colour. That is the fun of it.”
b. Eén van ons zal de gelukkige zijn.
one of us will the.C lucky be “One of us will be the lucky one.”
(Kraaikamp 2012, p. 215-216, ex. 23a and 23c)
5
These constructions are usually analysed as elliptical phrases. Kraaikamp points out that the fact they can take a determiner based on semantics suggests they are nominalisations and not elliptical phrases. If such determiner-adjective combinations are indeed nominalisations and not elliptical constructions, this could also explain why the construction in (iii) does not follow the normal agreement pattern expected for an indefinite neuter adjective.
(iii) Er waren dikke en dunne boeken.
There were thick and thin book.N.PL
Ik heb een dunne/*dun gekocht.
I have a thin.C/N bought.
Some double gender nouns change slightly in meaning depending on the gender of the determiner. This concerns nouns that can denote a substance with a neuter gender determiner or an object made of that substance with a common gender determiner (20).
(20) mass (neuter) count (common)
het steen de steen stone
het diamant de diamond diamond
het kurk de kurk cork
(Kraaikamp, 2012, p. 216, ex. 24) A reflection of gender based on individuation is also found among nouns of certain semantic classes. Following Zubin and Köpcke’s (1986, cited in Kraaikamp, 2012) observation on German that nouns referring to superordinate categories are usually neuter nouns and nouns referring to more specific terms are masculine or feminine, Kraaikamp shows that this division is also true for Dutch (21).
(21) superordinate (neuter) specific (common)
het instrument (the instrument) de gitaar (the guitar) de piano (the piano) de trompet (the trumpet) de fluit (the flute) het fruit (the fruit) de appel (the apple)
de peer (the pear) de kers (the cherry) de banaan (the banana) het meubel (the piece of furniture) de tafel (the table)
de stoel (the chair) de bank (the couch) de kast (the cabinet)
(Kraaikamp, 2012, p. 218, ex. 26) Kraaikamp argues that superordinate terms are less individuated than the more specific terms.
2.3 Gender (in)stability in Dutch
An increase in the use of semantic gender agreement among Germanic languages seems to correlate with a decrease in gender distinctions on agreement targets in the attributive domain (Audring, 2009, chap. 9; Kraaikamp, 2012). This implies that gender stability on personal pronouns, and perhaps on agreement targets in other domains, is dependent on gender stability in the attributive domain. Dutch has covert gender: the gender of a noun cannot be consistently be deduced by formal rules. An exception are diminutives, where the gender is determined by the morphological suffix. Because gender is covert, the main source of information about the gender of the noun comes from the attributive domain. If, however, the distinctions in the attributive domain become unstable, it seems that agreement targets in the other domains are insufficient to uphold the distinctions on the noun. In the past, the loss of the gender distinctions between masculine and feminine on attributive modifiers has indeed lead to the loss of the lexical genders masculine and feminine in Dutch, resulting in a common gender.
There is a growing gender instability on Dutch definite determiners. Especially among adult second language speakers, there is an overuse of the common definite determiner de (Blom, Polišenská, & Weerman, 2006, 2008; Orgassa & Weerman, 2008; Ziemann, Weerman, & Ruigendijk, 2011). This overuse of the common definite determiner de is common in Dutch gender acquisition. For most native speakers of Dutch this is a transitional period after which they acquire the neuter gender definite determiner het and have target-like grammatical gender at the age of five or six. However, child second language speakers and bilingual native speakers do not always reach this latter stage and seem to fossilise in the stage of overgeneralisation of de (Blom et al., 2008; Cornips, Van der Hoek, & Verwer, 2006; Hulk & Cornips, 2006). The quality and quantity of the input seems to be an important factor in the success or failure in acquiring a target-like grammatical gender. Cornips and Hulk (2006) found that bilingual children from ethnic minority communities are delayed in the acquisition of gender compared to monolingual children but bilingual children from communities where in addition to Dutch a dialect is spoken with a richer gender system on determiners than in standard Dutch are ahead compared to monolingual children. Brouwer, Cornips and Hulk (2008) found that bilingual children were more consistent in attributing gender with highly frequent words.
Attributive adjectives and relative pronouns seem to have an even more vulnerable gender. The child second language speakers and the child bilingual native speakers overuse the common forms with neuter nouns (Blom et al., 2008; Cornips et al., 2006; Hulk & Cornips, 2006; Orgassa & Weerman, 2008). The acquisition of the neuter forms of the determiner, attributive adjective and the relative pronoun is not simultaneous: the neuter form of the determiner is acquired before and/or used more frequently than neuter forms of the attributive adjective (Cornips et al., 2006) and both neuter forms of the determiner and attributive adjective are acquired before and/or used more often than the neuter form of the relative pronoun (Cornips et al., 2006; Hulk & Cornips, 2006). This is consistent with Corbett´s findings that attributive agreement targets are the last to lose gender and that attributive adjectives lose agreement before determiners (1991, p. 142-143). It seems possible then that in Dutch marking on the relative pronouns will disappear first, then on attributive adjectives and only later on the definite determiner.
province Groningen in the north of the Netherlands, definite determiners are frequently dropped, bilingual native speakers of Dutch and Gronings may also have relatively less input.
2.4 Gender loss and semantic gender agreement in Germanic languages
As mentioned before, Dutch is not the only Germanic language that has suffered gender loss or that utilises semantic gender agreement. Since gender loss and semantic agreement usually follow similar paths cross-linguistically, it is worthwhile to see how Dutch compares to other Germanic languages. Audring (2009, chap. 9) gives an overview of the attributive and pronominal gender systems in Germanic languages (table 6).
Attributive genders Pronominal genders
German
Masculine Feminine Neuter
Masculine Feminine Neuter Yiddish Icelandic Faroese Norwegian Norwegian Common Neuter Swedish Danish Dutch Frisian English - Afrikaans
Table 6. Attributive and pronominal genders in Germanic languages. From Audring, 2009, p. 179.
In English and Afrikaans all gender distinctions have disappeared over time except on personal pronouns. Standard Danish and Swedish have four genders: masculine, feminine, common, and neuter. The masculine and feminine genders only exist on personal pronouns and are only used in reference to humans and occasionally animals. German, Yiddish, Faroese, Icelandic, and some varieties of Norwegian have three genders in both domains.
In all Germanic languages semantic agreement exists, but it is used least frequently in languages that have the same genders in all domains. Kraaikamp (2012, 2016) argues that the rise of semantic agreement is not caused by the mismatch in number of features between nouns and personal pronouns, but rather that both phenomenon are a result of decreased gender marking on attributive agreement targets. Semantic agreement seems available in most if not all Germanic languages, but the degree to which it is available is dependent on the stability of lexical gender in the attributive domain (Kraaikamp, 2012, pp. 202-208). Audring (2009, chap. 9) also shows that the individuation hierarchy can be used to describe the different systems of semantic agreement in Germanic languages. For most, the cut-off point between masculine/feminine or common on the one hand and neuter on the other hand is between animate and inanimate. However, in Frisian and Afrikaans, masculine can be used in reference to objects.
Predicative adjectives modifying neuter nouns with human reference take the common gender (Holmes & Hinchcliffe, 2003, cited in Audring, 2009, p. 186). In Dutch, semantic agreement is used in the pronominal domain, similar to Afrikaans, Danish, and Swedish.
The question arises what the gender system in Dutch in the other domains is and will be. Four possible scenarios are for the gender systems in these domains 1) to remain strictly syntactic with two genders, 2) to lose all gender distinctions as it has in Afrikaans, 3) to increasingly utilise the semantic agreement system from the pronominal domain, or 4) to use semantic agreement based on the distinctions between in animate and inanimate similar to predicative agreement in the Scandinavian languages. Scenario 1 seems unlikely, since Audring found mismatches in gender between nouns and relative pronouns and the Dutch grammar E-ANS (the electronic version of the “Algemene Nederlands Spraakkunst”, General Dutch Speech Art) describes that common gender relative pronouns are possible with neuter gender nouns denoting humans (E-ANS, sec. 5.8.3.2). Scenario 2 is a possibility but there are no indications among native speakers of Standard Dutch as of yet that the distinction between common and neuter is disappearing. Scenarios 3 and 4 are perhaps most likely. The difference between scenarios 3 and 4 are the cut-off point between common and neuter. That is, the difference is whether or not common gender can be used on agreement targets with a neuter gender noun in reference to objects.
2.5 Gender agreement on Dutch relative pronouns
The loss or instability of gender distinctions on definite determiners is linked to both an instability of gender on attributive adjectives and relative pronouns and to an increase in semantic gender agreement. Research on gender acquisition seems to indicate that the distinction between common and neuter gender is also growing in instability. The question is then, whether instances of switches or mismatches on relative pronouns are a result of semantic agreement, as Audring (2009) concludes, or if they are a result of gender instability. It is possible that semantic agreement in Dutch is moving leftward on the agreement hierarchy and is to a lesser extent applicable on relative pronouns, but it is also possible that gender instability from the attributive domain is affecting the stability of gender on relative pronouns.
This study investigates how consistently Dutch native speakers apply syntactic gender agreement to relative pronouns. And if (some) Dutch native speakers allow a switch or mismatch in gender, whether these accepted switches are consistent with semantic agreement based on individuation. Since Audring found some indications that the latter is the case, the term switch will be used henceforth. In order to investigate this and to exclude any other linguistic factors that could contribute, the aim of this study was to have Dutch native speakers listen to pairs of sentences that were identical from the noun that controlled the relative pronoun onward. The sentences differed in gender of the relative pronoun: half of them matched the gender of the noun, half of them made a switch in gender. To limit the scope of this questionnaire study, only one class of nouns on the individuation hierarchy was selected: the category of objects and bounded abstracts. This is the class in which Audring (2009, p. 109) found most switches on relative pronouns (15 out of 22). This study focuses on objects, since Kraaikamp (2016) found in an elicitation task that Dutch native speakers did not make switches from neuter gender nouns denoting bounded abstracts to a common gender pronoun but did for neuter gender nouns denoting objects. If semantic agreement based on individuation is applicable on Dutch relative pronouns, switches from neuter gender nouns to common gender relative pronouns should be acceptable for objects but not switches from common gender nouns to neuter gender relative pronouns.
3 Method
3.1 Participants
87 people participated in this experiment. All participants were native speakers of Dutch from the Netherlands. 46 were male, 41 female (one unknown). The participants were between 19 and 73 years old (one unknown). The participants were from all over the Netherlands, though most were from the north: 64 participants were from the north, five were from the south, nine were from the west, and nine had an unknown or mixed background. Participants were placed in the north group if they had lived at least 12 of their first 18 years in the provinces of Groningen, Friesland, Drenthe, Overijssel, and the part of Gelderland that is north of the rivers and if at least one parent also grew up in this area. Participants were from the south if they had lived at least 12 of their first 18 years in the provinces of Limburg and Noord-Brabant and at least one of their parents did too. The west group consists of participants that grew up at least 12 of their first 18 years in any of the other provinces and at least one of their parents did too. Participants were put in the mixed/unknown group if they had not grown up at least 12 of their first 18 years in one these areas, or if both their parents were from a different area than the participants themselves, or if the area they and/or their parents had grown up in was unknown. They had a variety of educational backgrounds, though most attended or had attended university: 55 were being taught or had been taught at WO level (Wetenschappelijk Onderwijs, “Scientific Education”, meaning research universities) nineteen at HBO level (Hoger Beroeps Onderwijs, “Higher Vocational Education”, equivalent to universities of applied sciences), nine at MBO level (Middelbaar Beroeps Onderwijs, “Intermediate Vocational Education”), and four at VO level (Voortgezet Onderwijs, “Continued Education”, meaning secondary school). Table 7 gives an overview.
Count (N = 87) Sex Male 46 Female 40 Unknown 1 Age group -29 20 30-39 10 40-49 6 50-59 23 60+ 27 Unknown 1 Region North 64 West 9 South 5 Mixed/Unknown 9 Education WO 55 HBO 19 MBO 9 VO 4
3.2 Materials
Sixteen pairs of common and neuter nouns were selected using the CELEX database, which contains 42,380,000 words. The nouns were selected from the nouns with the highest frequencies in the database. The selected nouns had a frequency ranging from 1,545 to 26,719. Pairs of nouns were chosen that were as close as possible in frequency. All nouns were inanimate and concrete. Each pair consisted of a neuter and a common noun that were conceptually similar such as boot (“boat.C”) and schip (“ship.N”). However, in some cases a compromise had to be made between similar frequencies and similar in concept. For example, tafel (table) and bureau (desk) are very similar in concept, but had a difference in logarithm of 0.45. On the other hand, klok (clock) and schilderij (painting) only differed 0.01 in logarithm, but are conceptually farther apart than tafel and bureau are. In such cases, pairs were found that could be used in the exact same sentences. 22 lists all nouns according to gender. Table 1 in Appendix A lists all the nouns selected for the experimental sentences and their logarithmic frequencies in the CELEX database.
(22) a. Common gender nouns
bank (couch), boerderij (farm), boot (boat), brief (letter), fles (bottle), foto (photograph), jurk (dress), kerk (church), klok (clock), muur (wall), roman (novel), schaal (platter), school (school), tafel (table), trein (train), wijn (wine).
b. Neuter gender nouns
bed (bed), bier (beer), boek (book), bord (plate), bureau (desk), dak (roof), glas (glass), huis (house), kasteel (castle), museum (museum), papier (paper), rapport (report card), schilderij (painting), schip (ship), uniform (uniform), vliegtuig (airplane).
All 32 nouns appeared twice, once with a relative pronoun that agreed with the gender of the noun (henceforth called match items) and once with a relative pronoun that made a switch in gender (henceforth called switch items). The match and switch items had different subjects and relative pronouns but otherwise the sentences were exactly the same. In all experimental sentences the target noun was the direct object of the verb. In order to create some linear distance between the noun and the relative pronoun, the main verb, either a participle or infinitive, intervened between the noun and the relative pronoun, but no other material did (23).
(23) a. Je vader heeft de tafel geverfd, die bij het your father has the.C table.C painted REL.C at the
raam staat. window stands
“Your father has painted the table, which stands near the window.”
b. Je vader heeft het bureau geverfd, dat bij het your father has the.N desk.N painted REL.N at the
raam staat. window stands
As mentioned in section 2.2.3, Audring found that the sort of pronoun used, the form of the pronoun, case marking, linear distance, structural distance and the place of the referent on the individuation hierarchy influence the choice between syntactic and semantic agreement. In addition, unspecific readings and a personal relation to the referent influenced the choice for the gender of the pronoun. All experimental items used the same sort of pronoun, i.e. relative pronoun. Since relative pronouns only have one form for each gender and do not have any case markings, pronoun form and morphological markedness were a non-factor. The linear distance was the same in all experimental items, i.e. one word in between the antecedent noun and the relative pronoun. Though most of the relative pronouns functioned as subject in the relative clause, a few were objects. All antecedent nouns belonged to the category of objects on the individuation hierarchy.6 The use of the definite article het (N) or de (C) before the antecedent nouns excluded as much as possible an unspecific reading, which would be more likely if an indefinite article or no article was used, or a personal relation, which would be more likely if a possessive pronoun was used. The definite article also served as a cue to the gender of the noun, making a switch the more obvious. An example of all four conditions for the noun pair tafel/bureau (“table.C/desk.N)is given in table 8. Table 2 in Appendix A gives all experimental sentences.
Common antecedent noun Neuter antecedent noun
Match
Je vader heeft de tafel geverfd, Your father has the table painted die bij het raam staat.
which at the window stands
“Your father has painted the table, which stands near the window.”
Je vader heeft het bureau geverfd, Your father has the desk painted dat bij het raam staat.
which at the window stands
“Your father has painted the desk, which stands near the window.”
Switch
Je broer heeft de tafel geverfd, Your brother has the table painted dat bij het raam staat.
which at the window stands
“Your brother has painted the table, which stands near the window.”
Je broer heeft het bureau geverfd, Your brother has the desk painted die bij het raam staat.
which at the window stands
“Your brother has painted the desk, which stands near the window.”
Table 8. Example of experimental sentences. The relevant nouns and relative pronouns are underlined.
In addition, there were 48 filler items that varied in subject-verb agreement. The nouns for the experimental sentences were also used for the filler sentences. An additional eight nouns were selected that also had a high frequency, were inanimate and concrete. All nouns also appeared twice in the filler sentences, with half of them containing a mismatch in number between the subject and the verb. The format of these subject-verb agreement fillers was taken from Veenstra (2014). An example is given in 24. All sentences were recorded and edited with Adobe Audition.
(24) a. match
De tafel met de rode poten staat in de
the table.SG with the red legs stands.3SG in the kamer.
room
6
b. mismatch
De tafel met de rode poten staan in de
the table.SG with the red legs stand.PL in the kamer.
room
“The table with the red table legs is in the room.”
Two lists were made. Each list contained eight common gender nouns and eight neuter gender nouns: one of each pair. The relative pronoun match and switch experimental sentences and the number match and mismatch filler sentences of each noun occurred in the same list, so each participant would see any given noun twice in the test conditions and twice in the filler items. Both lists contained the filler sentences of the eight additional nouns. Both lists had a total of 80 items: 8 match sentences with a common noun and common relative pronoun, 8 switch sentences with a common noun and neuter relative pronoun, 8 match sentences with a neuter noun and neuter relative pronoun, 8 switch sentences with a neuter noun and common relative pronoun, and 48 filler sentences. Lest the participants lose concentration over 80 items, the lists were divided into 4 subsets of 20. The order of the items was semi-randomly distributed, with the provision that items with the same noun were divided from each other by at least three other items and that the first and last item of each subset was a filler sentence.
3.3 Procedure
4 Results
4.1 Group results
Table 9 gives the mean ratings and standard deviations for all match items with a common gender noun, match items with a neuter gender noun, switch items with a common gender noun, and switch items with a neuter gender noun. Table 1 in Appendix B shows the mean ratings and standard deviations for all filler items. The results of the filler items showed that the participants had understood the task: the match sentences had a mean rating of 4.6 and the mismatch sentences had a mean rating of 1.5. Mean (SD) Match Common 4.5 (1.00) Neuter 4.6 (0.98) Switch Common 2.0 (1.47) Neuter 2.3 (1.67)
Table 9. Mean ratings and standard deviations for experimental items for each of the four conditions.
Statistical analyses were done with IBM SPSS Statistics (22). A three-way mixed ANOVA with gender and match as within-subject variables and with list as between-subject variable showed main effects of gender, F (1, 85) = 15.192, p < 0.001 and match, F (1, 85) = 310.493, p < 0.001, but no main effect of list, F (1, 85) = 1.109, p = 0.295. On average, the items with a neuter gender noun were rated higher than the items with a common gender noun and match items were rated higher than switch items. There was also an interaction effect between gender and match, F (1,85) = 6.904, p = 0.010, but not between gender and list, F (1, 85) = 3.292, p = 0.073, nor between match and list, F (1, 85) = 0.052, p = 0.820. On average, switch items with a neuter gender noun were rated higher than switch items with a common gender noun. There was also a three-way interaction between gender, match and list, F (1, 85) = 6.754, p = 0.011. On average, match items with a neuter gender noun from list 1 were rated higher than the same items from list 2.
Figure 1. Percentages of each rating for items with common gender nouns and items with neuter gender nouns for match items. Darker bars represent common gender and lighter bars neuter gender.
Figure 2. Percentages of each rating for items with common gender nouns and items with neuter gender nouns for switch items. Darker bars represent common gender and lighter bars neuter gender.
Second most frequent was a difference of 0, more so for items with a neuter gender noun than for items with a common gender noun. This indicates that for most item pairs the match item was clearly preferred, but that there were also quite a lot of item pairs where the participants had no preference for either the match or the switch item. In only a very few cases was the switch item preferred over the match item for either gender. Since the number of negative difference was so low, a chi-square test was unreliable. Table 10 gives the differences with a new distribution collapsing them in 3 classes (figure 4). A difference of -4 to -2 meant the switch item was clearly preferred over the match item, a difference of -2 to 2 meant no clear preference and a difference of 2 up to and including 4 meant a clear preference for the match item over the switch item. A chi-square analysis revealed that the association between gender and the difference score was significant χ2 (2) = 11.729, p = 0.003. A difference score of -2 to 2 was found more frequently for items pairs with a neuter gender noun and a difference score of 2 to 4 was found more frequently for item pairs with a common gender noun.
Common Neuter
Match - Switch -4 to -2 5 8
-2 to 2 199 256
2 to 4 492 432
Table 10. Difference in rating between the match item and switch item with the same noun divided into three groups. N = 696 for each gender (number of participants times number of item pairs).
