The acquisition of gender and declension class in a non-transparent system: Monolinguals and bilinguals

Abstract

This article presents a corpus study of the acquisition of grammatical gender in Norwegian in two monolingual and two bilingual Norwegian‐English children. Gender in Norwegian is expressed as agreement between the noun and other targets such as determiners and adjectives, while definiteness and plurality are expressed as suffixes on the noun itself, i.e. as part of the declension. Furthermore, the gender system is characterized by relatively opaque gender assignment, suggesting that there may be a delay in the acquisition process compared to languages with more transparent systems. Our results show that, while the acquisition of suffixed forms is unproblematic, the children experience considerable problems with gender agreement. Moreover, there is generally no qualitative difference between the monolingual and bilingual children. These findings are discussed in relation to a number of issues: gender vs. declension class, the role of frequency, knowledge of the concept of gender, and monolingual vs. bilingual acquisition.

Full text

THE ACQUISITION OF GENDER AND DECLENSION CLASS IN A NON-
TRANSPARENT SYSTEM: MONOLINGUALS AND BILINGUALS
Yulia Rodina & Marit Westergaard
Abstract. This article presents a corpus study of the acquisition of grammatical gender in
Norwegian in two monolingual and two bilingual Norwegian-English children. Gender in
Norwegian is expressed as agreement between the noun and other targets such as
determiners and adjectives, while definiteness and plurality are expressed as suffixes on
the noun itself, i.e. as part of the declension. Furthermore, the gender system is
characterized by relatively opaque gender assignment, suggesting that there may be a
delay in the acquisition process compared to languages with more transparent systems.
Our results show that, while the acquisition of suffixed forms is unproblematic, the
children experience considerable problems with gender agreement. Moreover, there is
generally no qualitative difference between the monolingual and bilingual children. These
findings are discussed in relation to a number of issues: gender vs. declension class, the
role of frequency, knowledge of the concept of gender, and monolingual vs. bilingual
acquisition.
1. Introduction
The goal of this paper is to study the acquisition of the gender system in Norwegian at an
early stage, i.e. before the age of three. We investigate the development of grammatical
gender in corpus data from two monolingual and two bilingual Norwegian-English
children, mainly focusing on two aspects of the gender system. First, gender in
Norwegian is expressed as agreement between a noun and other targets (e.g. determiners
and adjectives) and as declensional suffixes expressing definiteness and plurality in
addition to gender. For example, while the indefinite article is a free morpheme, e.g. ei
1

bok ‘a book’, the definite article is a suffix, e.g. boka ‘the book’. There is also a double
definite form (required when the noun is modified), displaying both a prenominal
determiner and a suffix, e.g. den nye boka ‘the new book’. Second, grammatical gender in
Norwegian is relatively opaque, in the sense that the gender of most nouns cannot be
inferred from the shape of the noun itself and has to be learned for individual lexical
items. This means that input frequency is expected to play an important role in the
acquisition process. Compared to what has been found in studies of the acquisition of
languages with more transparent gender systems such as Italian or Russian (Kupisch et al.
2002, Gvozdev 1961), where gender agreement is often found to be in place from early
on, the acquisition of target-consistent gender agreement may take somewhat longer in
Norwegian. This should especially be the case in bilingual situations, since bilingual
children typically have comparatively less input in Norwegian than monolingual children.
Children who are learning English simultaneously with Norwegian may also be slower in
acquiring the concept of grammatical gender, since this is not present on nouns in English
at all.
Our main findings show that the acquisition of gender expressed on suffixes (the
definite singular article and the definite and indefinite plural forms) is relatively
unproblematic for the children, whether they are monolingual or bilingual. In
comparison, the production of gender agreement, especially the indefinite article, shows a
considerable delay, as some of the children are found to massively overgeneralize the
masculine. In the double definite forms, we also find several cases of omission in the
child data. Due to great individual variation across the four children, we cannot
distinguish any clear qualitative differences between the monolinguals and the bilinguals.
2. The gender system of Norwegian (Tromsø dialect)
Norwegian generally has a three-gender system, distinguishing between masculine,
feminine and neuter.1 According to Trosterud (2001), masculine nouns constitute 52%,
feminine nouns 32%, and neuter nouns only 16% of all nouns in the Nynorsk Dictionary
1 There is some dialectal variation, and certain dialects only exhibit two genders, common and neuter (cf.
e.g. Lødrup 2011).
2

(based on a total of 31,500 nouns). While we have not studied this specifically, this
distribution should generally correspond to the situation in the dialect that the children in
this study are exposed to (Tromsø).
As shown in Table 1, the Tromsø dialect makes a three-way gender distinction.
Gender is expressed on indefinite and definite articles, possessive and demonstrative
determiners, as well as on some adjectives. Indefinite articles are free morphemes (e.g. en
hest ‘a horse’) while definite articles are bound (e.g. hesten ‘horse-the’). In addition there
are double definite forms, where a free prenominal determiner is combined with a bound
definite article, e.g. den hesten, ‘that horse-the’. Double definiteness is required when the
noun is demonstrative or modified by an adjective. Adjectives only express gender in the
indefinite singular, with a distinction between masculine and feminine on the one hand
(i.e. common gender) and neuter on the other (e.g. fin ‘nice.M/F vs. fint ‘nice.N’). Only
one exceptional adjective, liten/lita/lite ‘little/small’, distinguishes between all three
genders.
Table . The gender system of the Tromsø dialect2
Gender Masculine Feminine Neuter
SG Indefinite en hest a horse ei seng a bed et hus a house
Definite hesten horse-the senga bed-the huset house-the
Double
Definite
den hesten
that horse-the
den senga
that bed-the
det huset
that house-the
Adjective en fin hest
a nice horse
ei fin seng
a nice bed
et fint hus
a nice house
en liten hest
a small horse
ei lita seng
a small bed
et lite hus
a small house
Possessive min hest/hesten min
my horse
mi seng/senga mi
my bed
mitt hus/huset mitt
my house
PL Indefinite hesta
horses
senge
beds
hus
houses
Definite hestan
horses-the
sengen
beds-the
husan
houses-the
Adjective fine hestafine sengefine hus
2 The final consonant in the suffix –et in the neuter definite is silent, which means that for neuter nouns
ending in –e, there is no distinction between the indefinite and the definite form (e.g. et hode - hodet ‘a
head - the head’).
3

nice horses nice beds nice houses
Possessive mine hesta/
hestan mine
my horses
mine senge/ sengen
mine
my beds
mine hus/
husan mine
my houses
There is one group of feminine nouns in the Tromsø dialect that behave differently from
the noun presented in Table 1. These nouns end in -a and are often referred to as bare
definites (cf. e.g. Anderssen 2006), as they appear in the same form in the indefinite and
definite singular, e.g. ei jenta – jenta ‘a girl – the girl’. In many other dialects of
Norwegian, as well as in the written standards Bokmål and Nynorsk, these nouns end in –
e, and there is thus alternation between the indefinite and definite forms, ei jente – jenta.
As discussed in Anderssen (2006:207), some speakers of the Tromsø dialect alternate
between the indefinite forms ei jente/ei jenta. This may also be true for the children in
our study and hence cause some uncertainty in our analysis of the data. We return to this
in section 6.
Gender assignment in Norwegian has traditionally been viewed as arbitrary, since
gender is not manifested on the phonological form of the noun itself. It has also been
argued that a large number of idiosyncratic assignment rules cover small classes of nouns
(Trosterud 2001, Enger 2001, 2004, Conzett 2006, Nesset 2006). For example, Trosterud
(2001) proposes as many as 43 different gender assignment rules, including three general
rules, 28 semantic rules, nine morphological rules and three phonological rules. The
abundance of rules proposed by Trosterud may be problematic in terms of learnability,
since the rules cover relatively small groups of nouns, some of which are quite
infrequent, especially in the input to young children. These rules also have a considerable
number of exceptions. According to an experiment carried out by Gagliardi (2012), only
three cues show strong predictive power in Norwegian: Male human, female human, and
final –e, a morphophonological cue for feminine.
The category gender is traditionally defined as agreement between the noun and other
targets, e.g. determiners, adjectives or verbs (Hockett 1958, Corbett 1991). In Norwegian,
there is in general very little agreement, e.g. no subject-verb concord. This means that the
expression of gender is to a large extent restricted to agreement internal to the DP. As we
see in Table 1, this concerns agreement between the noun and the indefinite article and
4

possessive determiners (a three-way gender distinction), and between the noun on the one
hand and adjectives and the prenominal determiner in double definite forms on the other
(generally a two-way gender distinction). In much of the literature on gender in
Norwegian (e.g. Enger 2004, Lødrup 2011), the suffixed forms are not considered to
express gender, but declension classes. The following example may illustrate the complex
relationship between gender and declension class: Some neuter nouns end in -(e)r in the
indefinite plural, e.g. eple-r ‘apples’, which is different from the majority of neuter nouns
which have no ending in this form, e.g. hus-Ø ‘houses’ in Table 1. These two nouns thus
belong to different declension classes, but gender agreement is the same in both cases (et
stort eple ‘a big apple’, et stort hus ‘a big house’). Nevertheless, the suffixed forms are
often included in a discussion of gender in Norwegian, e.g. in Faarlund et al. (1997). In
the present article we consider the suffixed forms in the child data and compare them to
the forms that express gender as agreement.
3. Previous studies on gender acquisition
3.1 The acquisition of transparent vs. opaque gender systems
Grammatical gender has been investigated extensively in both bilingual and monolingual
acquisition of many languages, such as Italian, Spanish, German, French, Dutch, Hebrew,
Russian, Polish, Czech, English, Greek and Welsh (Gvozdev 1961, Popova 1973, Henzl
1975, Levy 1983, Smoszyńska 1985, Mills 1986, Müller 1994, 2000, Serratrice 2000,
Tsimpli 2004, Blom et al. 2008, Kupisch et al. 2002, Gathercole & Thomas 2005, Rodina
2008, Unsworth et al. 2011, Rodina & Westergaard 2012, forthcoming/2013. Generally,
gender is acquired relatively early, typically by the age of three. However, the time of
acquisition appears to be dependent on the morphophonological transparency of the target
system, transparent cues facilitating early acquisition. For example, despite the many
similarities between Slavic languages such as Polish, Russian and Czech, Polish has more
transparent cues for gender assignment and agreement (Corbett 1991). It is thus not
surprising that Polish children acquire a three-way gender distinction already at the age of
5

two (Smoszyńska 1985). Russian and Czech children, on the other hand, have been
observed to make agreement errors with certain non-transparent noun classes until the
age of three/four (Gvozdev 1961, Popova 1973, Henzl 1975, Rodina 2007, Rodina &
Westergaard 2012, forthcoming/2013). A similar observation has been made for Romance
languages such as French and Italian, based on monolingual and bilingual
(French/German and Italian/German) acquisition (Kupisch et al. 2002). French children
have been found to experience a delay in gender acquisition caused by the fact the French
gender system exhibits fewer transparent morphophonological cues than Italian.
The presence of certain transparent morphophonological regularities in languages like
German and Greek ensures that monolinguals acquire a three-way gender system earlier
than children acquiring less complex systems, such as e.g. Dutch, which only makes a
common vs. neuter distinction (Mills 1986, Tsimpli 2004, Blom et al. 2008). Tsimpli
(2004) shows that Greek monolinguals acquire gender between the ages of three and four,
while Dutch children make errors in the neuter until the age of six, since gender
regularities are limited and have many exceptions (Blom et al. 2008). The differences in
transparency of the gender systems of Greek and Dutch have also been shown to play a
role in the acquisition of these languages by bilingual Greek/English and Dutch/English
children (Unsworth et al. 2011). The acquisition of non-transparent forms has been found
to be particularly vulnerable in bilingual children, who generally have less exposure to
the target gender system. For example, Gathercole & Thomas (2005) show that complex
and opaque forms of the Welsh gender system are the most difficult to acquire for
Welsh/English children, and they suggest that children with little exposure to Welsh at
home and/or school may never converge on the target.
Other relevant observations have been made with regard to the acquisition of gender
marking on determiners. In German/French bilinguals, Müller (1994) found an overuse of
masculine indefinite articles with feminine nouns in both languages between the ages of
two and three. At the same time the children had no problems with gender marking on
definite articles in either German or French. To explain this result, Müller suggests that
the indefinite article initially has referential but no grammatical gender-marking function
and should be analyzed as the numeral ein/une ‘one’. In a study on the acquisition of
gender in Dutch by different groups of bilingual children, Cornips & Hulk (2008) observe
6

that the children initially make no distinction between common and neuter gender in the
definite form, and overuse the common gender article (i.e. de instead of het). Following
Hawkins & Franceschina (2004), who observe a similar effect in young monolingual and
bilingual children acquiring Dutch, they argue that the definite article de only has the
feature definite, while the gender feature is unspecified. Finally, Italian monolinguals
have been found to produce phonetically reduced forms of the indefinite article at an
early stage (approximately the age of 2). An analysis of this by Bottari et al. (1993/94)
suggests that the indefinite article is initially simply a placeholder with syntactic
(positional) properties, and that the morphophonological properties of this element (such
as gender) will develop later.
3.2 The acquisition of gender in Norwegian
Until recently there have been no studies focusing on the acquisition of grammatical
gender in Norwegian. Some facts about the acquisition of the noun phrase in
Scandinavian are reported in Plunkett & Strömquist (1992). Both Swedish and Danish
have a two-way gender distinction (common and neuter), and this has also been argued to
be the case in the Oslo dialect of Norwegian (e.g. Lødrup 2011). Comparing some
longitudinal data of one Swedish child, two Danish children, and one Norwegian child
growing up in Oslo (data from Vanvik 1971), Plunkett & Strömquist (1992:526-529) find
very few gender errors overall. However, the Norwegian child is making slightly more
mistakes than the Swedish and Danish children, and they speculate that this is because
the child is also exposed to Norwegian dialects with a three-gender system. Virtually all
the gender errors found involve overgeneralization of common gender to neuter nouns, as
shown by the following examples: In (1), both the definite suffix and the possessive are
marked for common gender‚ while in (2) the suffix is correctly marked for neuter while
the postnominal possessive again displays overgeneralization of common gender.
(1) eggen min
egg.DEF.COMM.SG my.COMM.SG
‘my egg’ Target: egget mitt (N)
7

(2) badekaret min
bathtub.DEF.N.SG my.COMM.SG
‘my bathtub’ Target: badekaret mitt (N)
Occasional examples of overgeneralization errors are also reported in Anderssen’s (2006)
longitudinal study of the acquisition of compositional definiteness. The monolingual
child in this study is found to occasionally overgeneralize the masculine to feminine and
neuter nouns (e.g. en dame ‘a.M woman.F’ (Ina 1;10.4), en spøkels ‘a.M ghost.N’ (Ina
2;10.2)), especially in the indefinite form. No quantitative studies are made of these error
types in previous literature.
More recently, two experimental studies (Gagliardi 2012, Rodina & Westergaard
forthcoming/2013) have focused on gender acquisition in definite and indefinite DPs.
Rodina & Westergaard find that masculine is frequently overgeneralized with feminine
and neuter nouns in the indefinite and that the feminine is most problematic for
Norwegian three-to-five-year-olds (mean age 4;4). They also report fewer gender errors
with suffixed definite articles in the double definite forms. Similar findings are also
reported in Gagliardi (2012), who examines elicited production of older pre-school and
school children (mean age 5;1 and 6;8). With regard to the children’s sensitivity to gender
cues, Gagliardi observes that children have a strong bias to classify novel nouns as
masculine.
4. Research questions and predictions
The previous studies reviewed in section 3.1 suggest that lack of transparency of a gender
system may cause a delay in the acquisition process. As discussed in section 2,
Norwegian has very few (if any) reliable gender regularities, and we therefore predict that
the acquisition of gender will be delayed in relation to other languages where gender is
more transparent. In light of previous acquisition findings from Norwegian as well as
other languages, such as German and French (Müller 1994, Kupisch et al. 2002), we also
expect that the children may have more problems with indefinite than with definite
8

articles. Importantly, in order to investigate the issue of gender vs. declension class in
Norwegian (cf. the discussion in section 2), we compare the children’s accuracy rates on
suffixed forms with forms that express gender as agreement.
The study investigates two monolingual and two bilingual children growing up in
Tromsø. According to De Houwer (2007), the acquisition of the majority language in
bilinguals is typically unproblematic, and the bilingual children are thus not expected to
experience more problems acquiring Norwegian than their monolingual peers. However,
the children’s daily exposure to Norwegian did not start until the age of one, and they
may therefore be considered early successive bilinguals, a category of learners where
cross-linguistic influence is typically found (cf. Unsworth et al. 2011). Furthermore, the
opaque nature of the gender system of Norwegian, which are problematic even for
monolingual children, may cause even greater delays in bilingual learners.
5. The child data
The data are extracted from spontaneous production corpora of four children, two
monolingual Norwegian children and two bilingual Norwegian-English children. The two
monolingual children, Ina and Ole, were recorded from the age of approximately 1;9 to
3;0 (see Anderssen 2006, Westergaard 2009).
The bilingual data come from relatively restricted corpora of two girls, Emma and
Sunniva. Emma’s data have been collected by Kristine Bentzen (Bentzen 2000), while
Sunniva’s data have been provided by Merete Anderssen (not previously published).
Emma’s data were collected between the ages of 2;7.10 and 2;10.9 and consist of seven
one-hour recordings in Norwegian. She grew up in an English-speaking home with an
American mother and a Norwegian father. She had no older siblings at the time of
recording and English was used as the home language. Her daily exposure to Norwegian
started in daycare at the age of one. Until then Emma was exposed to Norwegian outside
the home, through the media, and from Norwegian-speaking friends and family.
There are also seven Norwegian recordings of Sunniva, made between 1;8.8 and
2;7.24, i.e. at a considerably younger age than Emma. Sunniva has a similar background:
9

A Norwegian mother and a British father, with English as the home language. She had no
siblings during the recording period. She was in daycare from the age of one and until
then she was exposed to Norwegian in and outside the home in the same way as Emma.
The recordings of both bilingual children were made by native speakers of Norwegian
(speaking a northern dialect very similar to the Tromsø dialect). They were later
transcribed in CHAT and double-checked by another native speaker of Norwegian. The
investigators and the transcribers are all linguists.
An overview of the child data is provided in Table 2, which shows that we have
chosen 6-7 files of each of the monolingual children for comparison, at an age where
their MLU in words corresponds roughly to the MLU of Emma.
Table . Overview of the child data
Name/Language(s) Age No of files No of child
utterances
MLUw range
Emma/N-E 2;7.10-2;10.9 7 2.222 3.28-4.12
Sunniva/N-E 1;8.8-2;7.24 7 2.890 1.93-3.44
Ole/N 2;6.2-2;10.00 7 (13-19) 3.394 3.34-4.83
Ina/N 2;10.2-3;3.18 6 (22-27) 4.297 3.30-3.79
The results presented in the next section have been counted as tokens. We have
disregarded all cases where it is unclear whether the child has produced a definite or a
base form, i.e. all feminine bare definites such as jenta (‘girl’ or ‘the girl’) and all neuter
nouns ending in -e, e.g. hode(t) (‘head’ or ‘the head’), where the two forms have identical
pronunciations. Not all adjectives and possessives express gender distinctions, e.g. some
relatively frequent adjectives such as anna ‘other’ and farlig ‘dangerous’ as well as all
plural adjectives (fine bila ‘nice cars.M’, fine bøker ‘nice books.F’, fine hus ‘nice
houses.N’) and plural possessives (mine bila ‘cars.M’, mine bøker ‘my books.F’, mine hus
‘my houses.N’). Furthermore, there is no gender distinction in the 3rd person singular and
all plural possessive determiners (e.g. bilen hennes/våres ‘her/our car.M’, boka
hennes/våres ‘her/our book.F’, huset hennes/våres ‘her/our house.N’). All these forms
have therefore been disregarded.
10

6. Results
6.1 Gender expressed on suffixes
Table 3 shows the results of the children’s production of suffixed forms across the three
genders. Overall, the four children’s production is generally target-consistent with an
error rate of only 2% for simple definites, 4% for indefinite plural and 6% for definite
plural forms. The table also shows that simple definite forms are the most frequent and
least problematic. In the definite plural, the children’s production is virtually error-free
with masculine and neuter nouns, which is not surprising, given that masculine and
neuter have the same suffix in this context (-an). Given that feminine and neuter plural
forms are infrequent in the data, the percentages of the error rates must be treated with
caution.
Table 3. Suffixed forms, overall results (error/total, % error)
Form Masculine Feminine Neuter Total
Def. sg. 4/484 (1%) 3/138 (4%) 10/229 (4%) 17/851 (2%)
Indef. pl. 0/157 (0%) 3/12 (25%) 5/43 (12%) 8/212 (4%)
Def. pl. 1/106 (1%) 9/24 (37%) 0/31 (0%) 10/161 (6%)
The results in Table 3 show that masculine is virtually error free across the three suffixed
forms. An analysis of the children’s errors reveals overgeneralization of the masculine
suffixes -en and –a (definite singular and indefinite plural), as well as masculine/neuter –
an (definite plural). This is illustrated in (3)-(5).
(3) i vinduen (Emma 2;8.20)
in window.DEF.M.SG
‘in the window ’ Target: i vinduet(N)
(4) mange kua (Emma 2;9.25)
many cow.INDEF.M.PL
‘many cows’ Target: mange kue(F)
(5) alle kuan (Emma 2;9.25)
11

all cow.DEF.M.PL
‘all the cows’ Target: alle kuen(F)
Table 4 displays the suffixed forms produced by individual children. There is no major
difference between the monolinguals and the bilinguals, although bilingual Emma clearly
experiences the most problems with an overall error rate of 10%.
Table 4. Suffixed forms, bilinguals vs. monolinguals (error/total, % error)
Child Def. sg. Indef. pl. Def. pl. Total
Emma (N-E) 10/114 (8%) 7/57 (9%) 3/29 (10%) 20/210 (10%)
Sunniva (N-E) 3/144 (2%) 1/29 (3%) 1/22 (5%) 5/195 (3%)
Ina (N) 3/291 (1%) 0/74 (0%) 5/61 (8%) 8/427 (2%)
Ole (N) 1/292 (1%) 0/52 (0%) 1/49 (2%) 2/393 (1%)
6.2 Gender agreement
When we study the children’s production of the forms that mark gender through
agreement, a very different picture emerges. As illustrated in Table 5, the overall error
rates are rather high across all agreement forms, especially the indefinite article, where
non-target-consistent production is 63% and 71% for feminine and neuter respectively.
The majority of errors are overgeneralization of the masculine indefinite article en, as
illustrated in (6)-(7).
Table 5. Agreement forms, overall results (error/total, % error)
Form Masculine Feminine Neuter Total
Indef. sg. 2/272 (1%) 69/109 (63%) 89/126 (71%) 160/507 (32%)
Adjectives 13/58 (22%) 4/18 (22%) 19/35 (54%) 36/111 (32%)
Possessives 12/103 (12%) 16/48 (33%) 21/73 (29%) 49/224 (22%)
(6) en mus (Emma 2;7.21)
a.M mouse.F.SG Target: ei mus
(7) en hode (Ina 2;10.2)
12

a.M head.N.SG Target: et hode
Recall from Table 1 that most adjectives only distinguish between a common
(masculine/feminine) and a neuter form, which may explain why the highest percentage
of adjectival errors is in the neuter (54%), cf. example (8). An exception is the adjective
liten ‘little’, which distinguishes between all three genders, and perhaps not surprisingly,
the errors in the masculine and feminine mainly involve this adjective: In (9), the
feminine form lita has been substituted for the masculine liten, while in (10), the
masculine liten is replaced by the neuter form lite. The explanation for this is presumably
that the form lite also means ‘of small quantity’, e.g. lite mat ‘little food’, which is also
frequent in the children’s input.
(8) en ny ark (Ina 2;10.2)
a new.COMM.SG sheet.N.SG
‘a new sheet’ Target: et nytt ark
(9) ei liten jenta (Ina 2;11.26)
a.F little.M.SG girl.F.SG
‘a little girl’ Target: ei lita jenta
(10) en lite sjiraff (Emma 2;7.10)
a.M little.N.SG giraffe.M.SG
‘a little giraffe’ Target: en liten sjiraff
Possessive forms are also error-prone in all three genders, although the error rate in the
masculine is considerably lower than in the other two genders. Most of the errors are
overgeneralizations of masculine, as illustrated in (11).
(11) Mummimamma sin forkle (Sunniva 2;0.18)
Mummimamma her.M.SG apron.N.SG
‘Moominmamma’s apron’ Target: Mummimamma sitt forkle
13

The bilingual vs. monolingual children’s agreement production is illustrated in Table 6.
Again, we see no fundamental difference between the bilingual children on the one hand
and the monolinguals on the other: All the children make similar mistakes, and bilingual
Sunniva and monolingual Ole make considerably fewer mistakes than bilingual Emma
and monolingual Ina.
Table 6. Agreement forms, bilinguals vs. monolinguals (error/total, % error)
Child Indef. sg. Adjectives Possessives Total
Emma (N-E) 70/194 (36%) 13/24 (54%) 13/36 (36%) 96/254 (38%)
Sunniva (N-E) 6/46 (13%) 1/22 (5%) 9/45 (20%) 16/113 (14%)
Ina (N) 76/173 (44%) 20/43 (47%) 14/41 (34%) 110/257 (43%)
Ole (N) 8/94 (9%) 2/22 (9%) 13/102 (13%) 23/218 (11%)
Tables 3-6 show that the acquisition of gender is more problematic in the case of forms
that express gender through agreement rather than suffixation. This is clear across the
various noun forms as well as across the individual children. In order to investigate the
difference between the statistical likelihood of producing a declension or an agreement
error, we have used a logit mixed-effects model with form (suffix or agreement) as a
fixed effect, and intercepts and correlated effects of form which varied randomly by both
child and observation type (the six forms investigated). Once random variation have been
accounted for, the model estimates show that the children are on average 6.16 times as
likely to produce agreement errors compared to suffix errors (Z = 2.41, p = 0.016).3
6.3 Double definite forms
As we saw in Table 1, double definiteness is required when the DP is demonstrative or
the noun is modified by an adjective. In these cases the noun is preceded by a prenominal
determiner as well as a definite suffix. Double definite DPs are relatively infrequent in
early child data, mainly because adjectives are rarely used by young children. The
construction is also quite complex. Not surprisingly, the children produce few relevant
examples, with the exception of bilingual Emma, whose data on definiteness is discussed
3 We would like to thank Martin Corley for the statistical analysis of the data.
14

in more detail in Anderssen & Bentzen (this volume). Emma is different from the other
children also in the sense that very few of her double definites are target-consistent: Only
12% (16/134) have the correct gender form on both the prenominal determiner and the
suffix. Corresponding percentages for the other children are 46% (11/24) for Sunniva,
76% (40/53) for Ole and 48% (13/27) for Ina.
Table 7 shows the distribution of gender marking on the prenominal determiner and
the suffix across the three genders. There are at least two important points: First, the
totals for suffixes are lower than for determiners, which is mainly due to Emma’s
production where suffixes are spelled out less frequently than determiners (more on this
below). Second, most of the children’s mistakes occur in the neuter, where the
prenominal determiner det is replaced by the common gender form den. The error rate on
the suffix is considerably lower. An example is provided in (12).
Table 7. Gender marking on the prenominal determiner and the suffix in double definite
DPs in the child data (error/total)
Child Masculine Feminine Neuter
Det. Suffix Det. Suffix Det. Suffix
Emma (N-E) 0/78 0/9 0/48 0/7 5/6 0/0
Sunniva (N-E) 0/13 1/11 0/6 1/2 4/4 0/0
Ole (N) 2/35 0/31 1/5 0/4 4/9 0/11
Ina (N) 1/15 1/10 0/5 0/4 6/8 3/3
Total 3/141
(2%)
2/61
(3%)
1/64 (2%) 1/17 (6%) 19/27
(70%)
3/14
(21%)
(12) den store flyet (Ole 2;8.5)
the.COMM.SG big plane.DEF.N.SG
‘the big plane’ Target: det store flyet
Errors such as the one illustrated in (12) are typical of the monolingual children as well as
for bilingual Sunniva. Once again they reveal that gender marking through agreement on
the prenominal determiner is more problematic than gender marking on the suffix. There
are also many instances of the same lexical item occurring with correct marking on the
15

suffix, but not on the indefinite article or prenominal determiner. This is illustrated for
Ole in (13)-(14).
(13) en fly (Ole 2;8.5)
a.M.SG plane.N.SG
‘a plane’ Target: et fly
(14) inni flyet (Ole 2;8.5)
in plane.DEF.N.SG
‘in the plane’
In Table 7 there seems to be no fundamental difference between the bilingual and
monolingual children. However, on closer inspection bilingual Emma appears to be
different from the other three. First, she produces more forms requiring double definites,
mainly because she seems to favor demonstratives.4 She also omits suffixes to a
considerable extent.5
(15) den her skjei-Ø (Emma 2;8.7)
the.COMM.SG here spoon.DEF.F.SG
‘this spoon here’ Target: den her skjeia
In Table 8, the children’s errors of omission are considered across the three genders with
regard to the affected element, the determiner or the suffix. Both the prenominal
determiner and the suffix are occasionally omitted by Sunniva, Ole and Ina. In Emma’s
data, determiner omission is also infrequent, in contrast to suffix omission, which occurs
at a rate of 95% in the masculine and 100% in the feminine and neuter.6
4 Emma’s apparent preference for demonstratives may also be due to transfer from English, in the sense
that a regular definite is rendered as e.g. den bil instead of bilen (‘the car’). According to Anderssen &
Bentzen (this volume), 19 of the examples are of this type.
5 Strictly speaking, omissions are not gender errors and are therefore not considered in the other nominal
forms discussed in this paper.
6 With respect to the bare definite feminine nouns in the Tromsø dialect (e.g. ei jenta - jenta ‘a girl - the
girl’), Emma is found to vacillate between the dialect form ending in -a and the standard form ending in -e
in the indefinite. She therefore also produces examples such as den dukka as well as den dukke ‘that doll’.
While the latter examples are clearly non-target-consistent (due to suffix omission), all the former examples
16

Table 8. Omission of determiner and suffix in double definite DPs
Child Masculine Feminine Neuter Total
Det. Suffix Det. Suffix Det. Suffix Det/Suffix
Emma (N-E) 5 66 0 23 0 6 5/95
Sunniva (N-E) 1 2 0 2 0 1 1/5
Ole (N) 1 3 0 0 3 0 4/3
Ina (N) 0 5 1 0 0 2 1/7
7. Discussion
7.1 Gender vs. declension classes
Recall from section 2 that gender is typically defined as agreement between the noun and
other targets. In the case of Norwegian, these other elements are the indefinite articles,
certain adjectives and some possessive forms. This means that the definite singular and
definite and indefinite plural forms, which are suffixes on the noun itself, strictly
speaking do not express gender, although they do have distinct forms corresponding to
the three genders. These endings are more commonly referred to as expressions of
declension classes. Hence, in a form such as bilen ‘the car’, the suffix expresses that this
noun belongs to the -en declension class, which typically takes masculine gender
agreement on other targets. In the indefinite form en bil ‘a car’, on the other hand, the
choice of indefinite article agrees in gender with the masculine noun. One argument in
favor of such a distinction in Norwegian is that there are occasional mismatches between
the indefinite and definite forms in some dialects, e.g. en(M) bok – boka(F) ‘a book - the
book’ (Lødrup 2011).
Our findings from the acquisition data presented in this paper support this distinction
between gender and declension class. The declensional suffixes seem to be much easier
to acquire than gender agreement (Table 3 vs. Table 5). The definite form, for example, is
have been disregarded from the count in Table 8, as it is unclear whether the form dukka is the base form or
the definite. This is presumably the reason why our numbers for Emma are somewhat lower than those
found in Anderssen & Bentzen (this volume). It should be noted that most of the -e forms appear in early
files, while the examples with the -a ending mainly appear later, indicating that Emma is in fact making
even more mistakes with suffix omission than Table 8 shows.
17

produced with a target-consistent declensional suffix in 98% of the child data, while the
percentages for target-consistent plural suffixes is 96% and 94% for indefinites and
definites respectively. The difference between the definite singular and the two plural
forms is presumably that the former is much more frequent in the children’s input, also
evidenced by the different frequencies in the child data (cf. the raw figures). In
comparison, the forms that agree with the noun and thus express gender per se, are much
more error-prone: Adjectives are target-consistent 68%, possessives 78% and indefinite
articles 68%, and the statistical analysis shows that there is a highly significant effect of
the observation type (i.e. suffix vs. agreement) on the children’s accuracy.
One may then ask what triggers the acquisition of gender agreement in Norwegian.
As we stated in section 2, gender seems to be arbitrary, in the sense that the phonological
shape of the base form does not reveal the gender of a noun. Furthermore, there is not
much semantic agreement, at least not for the common everyday words that are frequent
in the vocabulary of a young child. One plausible explanation is thus that children learn
gender based on declensional endings. More specifically, we would argue that gender
acquisition is based on the suffixed definite form, i.e. bilen ‘the car’, boka ‘the book’ or
huset ‘the house’. The definite form is prosodically favored by young children in that it
typically constitutes a trochaic structure (Anderssen 2006). It is also the most frequent DP
form in the language, perhaps especially in the input to children, since child and child-
directed speech typically center around the “here and now”. This form is also the most
frequent one in the child data, which becomes evident if we compare the raw numbers in
Table 5 with any other table in this paper (recall also that a number of nouns have been
excluded from the count of the definite singular because they are ambiguous between a
definite and a bare form).
One example in our data that also points in this direction is Ole’s use of the noun
søppel ‘garbage’. This is one of the nouns where there is a mismatch between the definite
and the indefinite forms: The indefinite may only be neuter (et søppel), while the definite
form can be either søppelet(N) or søpla(F), the latter being the most commonly used form.
This is also the form that Ole uses in (16). Note that he is using the correct feminine form
of the possessive here, to correspond with the definite suffix on the noun. He has clearly
18

not learned the gender of this word from the indefinite form, for example, since ei søppel
‘a.F garbage’ simply does not exist.7
(16) søpla mi (Ole 2;10.0)
garbage my.F.SG
‘my garbage’
7.2 The role of frequency
In section 5 we predicted that the acquisition of an opaque gender system such as the
Norwegian one would be delayed in relation to languages with more transparent systems.
Input frequency should thus play a role, and we therefore speculated that bilingual
children, who may have less input in Norwegian, would need even more time to acquire
both the concept of gender and the corresponding morphology. The role of frequency in
gender acquisition has been discussed extensively in Szagun et al. (2007), who claim that
‘the role of frequency seems unclear at present’ (p. 450). Rodina (2007) on the other
hand, shows that frequency does play a role in the acquisition of exceptional noun classes
in Russian, i.e. where there is a mismatch between morphological and semantic cues (e.g.
the noun papa ‘daddy’, which is morphologically feminine but semantically masculine).
The results of the current study show that, when the children produce non-target-
consistent forms, they generally overgeneralize to the most frequent form in the input,
which in most cases is the masculine. For example, there is massive overgeneralization of
the masculine indefinite article en to feminine and neuter nouns. There is also
overgeneralization of the common gender (masculine/feminine) form to the neuter on
adjectives. Common gender is also overgeneralized to the neuter on the prenominal
determiner in the double definite DPs. Furthermore, there is also overgeneralization in the
declensional suffixes; the masculine suffix -a in the indefinite plural is overgeneralized to
the feminine and neuter, while in the indefinite plural it is mainly the masculine/neuter
7 Lødrup (2011) discusses a number of nouns that are not feminine but nevertheless take the suffix -a in
the definite. These may only take feminine agreement on a postnominal possessive‚ while in all other forms
there is normally agreement with the ‘original’ gender of the noun. This is also the case with søppel
‘garbage’: With a prenominal possessive the only correct form is neuter, i.e. mitt søppel ‘my.N garbage’,
while the use of feminine here is ungrammatical (*mi søppel).
19

suffix -an which is overgeneralized to the feminine. While all these cases differ to some
extent, they have one thing in common: It is always the most frequent gender form or
declensional suffix that is overgeneralized to other contexts. This provides support to the
claim that frequency plays a major role in the acquisition of a non-transparent gender
system.
However, frequency is a relative concept, and it is unclear how much is needed for
the acquisition of a particular grammatical feature, and whether all children need the
same amount of input. Frequency in the input is also extremely difficult to measure for
individual children. In our data, we find that the two bilingual children are very different
from each other; in fact, also the two monolingual children display considerable
differences with respect to the mastery of the gender forms. Bilingual Emma is more
similar to monolingual Ina, in that they produce more non-target-consistent forms than
the other two children. In this sense, therefore, our prediction that frequency should be
responsible for a distinction between bilingual children on the one hand and
monolinguals on the other clearly does not hold. This leads us to consider the claim made
by De Houwer & Bernstein (2011) that bilingual children do not always have only half
the input in each of their languages compared to monolingual children. They show that
the variation in the amount of child-directed speech that monolingual and bilingual
children are exposed to is quite extensive, and that it is impossible to conclude that
bilingual children always have much less input than monolinguals. Thus, given that we
do not have exact input frequencies of the different forms for the four children in our
study, we cannot make any conclusions with respect to the effect of frequency on mono-
vs. bilingual acquisition.
7.3 Knowledge of the concept of gender
We now turn to the question whether the children in our study actually display any
knowledge of the concept of gender in their production. For one monolingual and one
bilingual child, Ole and Sunniva, we would argue that this is clearly the case, in that they
make systematic choices in their production of forms that agree with the noun, e.g.
indefinite articles and possessives. We would also like to argue that Ina is in the process
20

of acquiring gender; however, she still has some problems producing the correct
morphological form in every case. Emma, on the other hand, does not seem to display
any knowledge of gender until the very last file of her data, when a few target-consistent
double definites appear. That is, although Emma produces target-consistent declensional
suffixes in many cases, she fails to distinguish the three genders on other targets:
Indefinite articles have a default form (the masculine en), while her production of
adjectives and possessive forms seems to be subject to chance.
The indefinite article en thus does not seem to have any gender marking function in
Emma’s grammar. This may also to some extent be the case for Ina, who seems to have
special problems with the neuter, while there may be remnants of such a system in the
production of Sunniva and Ole. We may therefore ask what the function of this
prenominal element is in early child grammar. As mentioned in section 3.1, several
studies have attested similar phenomena, e.g. an overuse of masculine indefinite articles
with feminine nouns in German/French bilinguals (Müller 1994). In this case, the
indefinite article was analyzed as a numeral. This would be a possible analysis also for
our data, as the masculine indefinite article is identical to the numeral also in Norwegian.
However, given that the children make similar mistakes also with adjectives and
possessives (albeit to a lesser extent), this does not seem to be a plausible analysis. At
best, it may be an additional factor, accounting for the somewhat greater extent to which
the children overgeneralize the indefinite article. An alternative analysis, proposed by
Bottari (1993/94) for Italian monolinguals, was that the indefinite article is simply a
placeholder with syntactic properties, but no morphophonological features. For our data,
we believe that the invariant indefinite article en does express the features indefinite and
singular in the children’s grammars; see also Anderssen (2006) for an analysis of the
acquisition of definiteness in Norwegian. Our finding thus seems to be more similar to
the attested overgeneralization to common gender in Dutch discussed by Cornips & Hulk
(2008): That is, the article en in the early Norwegian child data expresses definiteness and
number, but is unspecified for gender.
7.4 Bilinguals vs. monolinguals
21

Our results show that there is no fundamental difference between the bilingual children
on the one hand and the monolingual children on the other, as in most contexts,
monolingual Ina and bilingual Emma pattern together. Instead we find individual
differences, which are mostly quantitative. All the children experience problems with
gender agreement, but to different extents. Monolingual Ole, for example, makes the
fewest gender errors across the various DPs. Both bilingual Emma and monolingual Ina
display a relatively poor performance in the indefinite singular compared to bilingual
Sunniva and monolingual Ole. The same applies to gender marking on adjectives and
possessives.
Gender marking on double definite DPs is the only context where Emma seems to be
qualitatively different from the other children. Until the age of 2;8.20 she only uses the
demonstrative den and does not use a single suffix. There is thus no evidence that Emma
marks gender on double definites until the first and only occurrence of neuter det (which
marks gender unambiguously) is found at the age of 2;9.25. In the previous section we
discussed the children’s overgeneralization of the indefinite article en in the light of
similar findings from Dutch child language (Cornips & Hulk 2008) and argued that this
element expresses definiteness and number, but is unspecified for gender. Similarly, as
Emma overuses the prenominal determiner den in the double definite forms, this could
also be argued to only express definiteness and number in Emma’s grammar and not
gender. This could also explain why she omits the definite suffix: When definiteness is
marked on the prenominal determiner, the grammar assumes that there is no need to also
spell out definiteness on the suffix.
This account does not explain why the other children do not also pervasively omit the
suffix in double definite DPs. One reason could of course be that they are beyond this
stage and have already realized that the prenominal element expresses gender and
therefore include both the determiner and the suffix. That is, they may have already gone
through a (possibly brief) stage where their grammars were identical to Emma’s
grammar. However, suffix drop is hardly attested in the other children’s grammars. Thus,
a likely explanation is that this property of Emma’s grammar is due to transfer from
English, which lacks this double definite construction; see Anderssen & Bentzen (this
volume) for a more thorough discussion of these data.
22

Finally, however, we should note that the other bilingual child, Sunniva, does not
experience any delay with the double definites. Thus, if Emma’s problem with this
construction is due to language transfer, we have to conclude that not all bilingual
children are affected by it. That is, our data show that the individual variation among
bilingual children is considerable and at least as great as the variation among
monolinguals.
8. Concluding remarks
In this paper we have compared the acquisition of gender in Norwegian (Tromsø) in the
corpus data of two monolingual and two bilingual Norwegian-English children. We have
found that all four children pattern alike in that they have no problems with gender
marking on suffixed forms (e.g. the definite article), but experience great difficulties
when marking gender on other targets, such as indefinite articles or adjectives. The
difference between the children’s performance on declensional suffixes and gender
agreement is statistically significant, and our acquisition data may thus be taken as
support for the distinction between gender and declension class that is often made for
Norwegian. Furthermore, we have argued that gender acquisition is based on the suffixed
definite form.
Given the opaque nature of gender assignment in Norwegian we predicted that we
would find frequency effects, especially in the data of the bilingual children. Across the
three genders we found that the errors that the children make reveal overgeneralization of
the most frequent forms, mainly masculine or common gender. However, we do not find
that the monolinguals always outperform the bilinguals; instead there are substantial
individual differences among the four children.
References
23

ANDERSSEN, M. 2006. The Acquisition of Compositional Definiteness in Norwegian.
Doctoral dissertation, University of Tromsø.
ANDERSSEN, M. & K. BENTZEN. This volume. Cross-linguistic influence outside the
syntax-pragmatics interface: A case study of the acquisition of definiteness.
BENTZEN, K. 2000. I like not it like du like it! A Case Study of Language Transfer in
Bilingual First Language Acquisition. M. Phil. thesis, University of Tromsø.
BOTTARI, P., P. CIPRIANI & A. M. CHILOSI. 1993/94. Protosyntactic devices in the
acquisition of Italian morphology. Language Acquisition 3(4), 327-369.
CONZETT, P. 2006. Gender assignment and the structure of the lexicon. Sprachtypologie
und Universalienforschung 59.3, 223-240.
CORBETT, G. G. 1991. Gender. Cambridge: Cambridge University Press.
CORNIPS, L. & A. HULK. 2008. Factors of success and failure in the acquisition of
grammatical gender in Dutch. Second Language Research 24.3, 267-295.
DE HOUWER, A. 1990. The Acquisition of Two Languages from Birth: A Case Study.
Cambridge: Cambridge University Press.
DE HOUWER, A. 2007. Parental language input patterns and children’s bilingual use.
Applied Psycholinguistics 28, 411-424.
DE HOUWER, A. & M. H. BORNSTEIN. 2011. Maternal input to bilingual and monolingual
children. Talk given at IASCL 2011, Montreal.
ENGER, H.-O. 2001. Genus i norsk bør granskes grundigere. Norsk Lingvistisk Tidsskrift
19, 163-183.
ENGER, H.-O. 2004. On the relation between gender and declension: A diachronic
perspective from Norwegian. Studies in Language 28.1, 51–82.
FAARLUND, J. T., S. LIE & K.-I. VANNEBO. 1997. Norsk referansegrammatikk. Oslo:
Universitetsforlaget.
FANTINI, A. E. 1985. Language Acquisition of a Bilingual Child: A Sociolinguistic
Perspective (to age ten). Clevedon: Multilingual Matters.
GAGLIARDI, A. 2012. The Fundamentals of Language Acquisition. Doctoral dissertation,
University of Maryland.
24

GATHERCOLE, V. C. M. 2002. Grammatical gender in bilingual and monolingual children:
A Spanish morphosyntactic distinction. Language and Literacy in Bilingual
Children, eds. D. K. Oller & R. E. Eilers, 207-219. Clevedon: Multilingual Matters.
GATHERCOLE, V. C. M. & E. M. THOMAS. 2005. Minority language survival: Input
factors influencing the acquisition of Welsh. Proceedings of the 4th International
Symposium on Bilingualism, eds. J. Cohen, K. McAlister, K. Rolstad & J. MacSwan,
852-874. Somerville, MA: Cascadilla Press.
HAWKINS, R. & F. FRANCESCHINA. 2004. Explaining the acquisition and non-acquisition
of determiner-noun gender concord in French and Spanish. The Acquisition of
French in Different Contexts: Focus on Functional Categories, eds. P. Prevost & J.
Paradis, 175-205. Amsterdam: John Benjamins.
HOCKETT, C. F. 1958. A Course in Modern Linguistics. New York: MacMillan.
KUPISCH, T., N. MÜLLER & K. F. CANTONE. 2002. Gender in monolingual and bilingual
first language acquisition: Comparing Italian and French. Lingue e Linguaggio 1,
107-147.
LØDRUP, H. 2011. Hvor mange genus er det i Oslo-dialekten? Maal og Minne 2 2011,
120-36.
MÜLLER, N. 1994. Gender and Number agreement within DP. Two first languages. Early
Grammatical Development in Bilingual Children, ed. J. M. Meisel, 53-88.
Dordrecht: Foris.
NESSET, T. 2006. Gender meets the Usage-Based Model: Four Principles of Rule
Interaction in Gender Assignment, Lingua 116.9, 1369-1393.
PLUNKETT, K. & S. STRÖMQUIST. 1992. The acquisition of Scandinavian languages. The
Crosslinguistic Study of Language Acquisition, ed. D. I. Slobin, 457-556. Hillsdale,
NJ: Lawrence Erlbaum Associates.
RODINA, Y. 2007. What do asymmetries in children’s performance tell us about the nature
of their underlying knowledge? Nordlyd 34.3, 230-251.
RODINA, Y. 2008. Semantics and Morphology: The Acquisition of Grammatical Gender
in Russian. Ph.D. dissertation, University of Tromsø.
RODINA, Y. & M. WESTERGAARD. 2012. A cue-based approach to the acquisition of
grammatical gender in Russian. Journal of Child Language 39.5, 1077-1106.
25

RODINA, Y. & M. WESTERGAARD. FORTHCOMING/2013. Two gender systems in one
mind: The acquisition of grammatical gender in Norwegian-Russian bilinguals.
Multilingualism and Language Contact in Urban Areas: Acquisition-Development-
Teaching-Communication, Hamburg Studies on Linguistic Diversity. Amsterdam:
John Benjamins.
SERRATRICE, L. 2000. The Emergence of Functional Categories in Bilingual First
Language Acquisition. Ph.D. dissertation, University of Edinburgh.
SZAGUN, G., B. STUMPER, N. SONDAG & FRANIK, M. 2007. The acquisition of gender
marking by young German-speaking children: Evidence for learning guided by
phonological regularities. Journal of Child Language 34, 445-71.
TROSTERUD, T. 2001. Genustilordning i norsk er regelstyrt. Norsk Lingvistisk Tidsskrift
19, 29-57.
UNSWORTH, S., F. ARGYRI, L. CORNIPS, A. HULK, A. SORACE & I. TSIMPLI. 2011. On the
role of age of onset and input in early child bilingualism in Greek and Dutch.
Selected Proceedings of the 4th Conference on Generative Approaches to Language
Acquisition North America, eds. M. Pirvulescu, M. C. Cuervo, A. T. Pérez-Leroux, J.
Steele & N. Strik, 249-265. Somerville, MA: Cascadilla Press.
VANVIK, A. 1971. The phonetic-phonemic development of a Norwegian child. Norsk
Tidsskrift for Sprogvidenskap 24, 269-325.
WESTERGAARD, MARIT. 2009. The Acquisition of Word Order: Micro-cues, Information
Structure and Economy. [Linguistik Aktuell/Linguistics Today 145], Amsterdam:
John Benjamins.
Yulia Rodina & Marit Westergaard
University of Tromsø
CASTL/Department of Language and Linguistics
9037 Tromsø
Norway
yulia.rodina@uit.no, marit.westergaard@uit.no
26