This study examines 7-to-12-year-old Singaporean bilingual children's use of noun classifiers in Mandarin Chinese and their reasoning while making the associations between nouns and classifiers as a given task. The results show that the children made the association by applying their cognitive understanding of the properties and functions of the noun objects, comparing between them, as well as following self-generated or learned rules. To investigate other factors that may affect the children's learning, the children and their parents were given the task to make another 120 classifier phrases separately based on the list of the noun objects given. The results show that schooling, parents' language proficiency, father's age, mother's academic attainment, and family income had correlations with the children's result of this task-based language performance.
As a typological feature, noun classifiers in Chinese are unavoidable in everyday use of the language. For example, “a book” in English must be expressed in Chinese as “yì běn shū”, which has a classifier “běn” in between “a” and “book”. This compulsory structure requires Chinese speaking children acquire classifiers at an early age.
While relatively scarce in relation to noun studies, there is
still a sizeable volume of research done on classifier
acquisition and production
Research on other classifier languages, such as Thai
showed that Thai classifier usage seemed to be influenced
by the different aspects of a noun object that the speaker
wished to highlight (Juntanamalaga, 1989).
Mass count distinctions may also influence Chinese noun classifier usage. Contrary to the popular belief, the Chinese language does have a mass-count distinction for nouns and studies have shown that this distinction is made at the classifier and measure word level. For example, Yu et al’s (2003) comprehension-based study showed that Chinese speaking children could differentiate between count and mass nouns at an age comparable to English children (approximately 3 years-old) and improved steadily with time until adulthood. This explains the fact that generally count nouns require classifiers and mass nouns require measure words. Hence, whether a classifier is used or even which classifier is used may be dependent on whether the noun is a count or mass noun. The study done by Lee et al. (2008) further supports this claim. They found that children began acquiring classifiers by attending to shape, and became sensitive to solidity over a period of several years. Their study also shows that even at the age of six, Chinese speaking children still had not fully grasped the mass-count distinction. This may be partly due to the fact that in their study, they used objects that were unfamiliar to the children.
How bilingual Chinese speaking children acquire Chinese
classifiers was unknown until the study conducted by
As many studies show that bilingual children’s language acquisition is language specific. It involves both linguistic and non-linguistic factors. To investigate how Chinese classifiers are acquired by Singaporean English-Chinese bilingual children, this study aims to achieve the following three objectives. 1) to examine bilingual Chinese speaking children’s use of noun classifiers, 2) to understand the reasons behind their choices in making the association and 3) to identify the non-linguistic factors that may influence the level of the children’s classifier mastery.
This study includes two parts: children’s noun and classifier match task and a survey of children and their parents’ use of classifiers
Thirty English-Chinese bilingual Singaporean children were recruited (Mean=9.5 years; range=7 to 12 years; 16 girls, 12 boys). Quota sampling (approximately 5 children per age group) was employed to ensure that each age group was fairly represented. There were 6 age groups, corresponding to the ages 7, 8, 9, 10, 11, 12 and the school grades were Primary 1, 2, 3, 4, 5, 6 respectively.
The stimuli set contained 30 picture cards of everyday items. 27 of them required the use of a classifier in quantification and 3 required the use of measure words. Each card also had a helping phrase to elicit responses from the children (e.g., 一 （ ） 椅 子 one (classifier/measure word to be inserted) chair).
The children were tested individually in a separate room from their peers. They were asked to provide a classifier for each object on the picture card. The images were shown sequentially and the children were asked to answer. They were also given the option of saying ‘I don’t know.’
After all the images had been shown and their respective answers were collected, children were asked to provide reasons for their incorrect use of the classifiers with each noun object shown in the picture. However, they were not informed of their mistakes.
Thirty-five families that had children between the ages of 7 and 12 were recruited to fill in a questionnaire that includes a classifier-noun association task (for children and both of their parents to fill in) and enquires of family background. However, due to the incomplete answers, only 15 families’ responses were selected as usable data. Among the families, there were 15 children (Mean=8.73 years; range=7 to 12 years; 10 girls, 5 boys), 14 mothers (Mean=42.2 years; range=33 to 45 years), and 7 fathers (Mean=48.36 years; range=47 to 51). Same as the children who participated in the first task, they all spoke both English and Chinese at home and took Mandarin Chinese classes at school regularly.
The majority 64.29/%). of the mothers graduated from secondary schools. 14.2% of them completed university studies. 7.14% graduated from Junior college or Polytechnic institutes. The rest completed primary school.
The majority (45.45%) of the fathers completed primary school. The rest of the fathers graduated from universities (18.18%), junior colleges or polytechnic institutes (18.18%), or had no formal education (18.18%).
66.67% of the mothers’ dominant language was Mandarin Chinese and 53.33% of them spoke to their children mainly in Mandarin Chinese. The corresponding figures for fathers are 76.92% and 61.54% respectively.
In terms of housing, 40.00% of the families lived in 3room flats, 26.67% stayed in 4-room flats, 13.33% lived in 5-room flats, and the rest lived in other types of housing.
The majority (57.14%) of the families had incomes less than $3000 a month. That is a rough gauge for the lowest quintile of the monthly household income in Singapore.
A self-designed questionnaire with 120 noun objects and a series of enquiries about the participants’ language and SES backgrounds was ued. The children and both of their parents were required to fill in the questionnaire. The 120 objects are commonly seen or used objects which require different classifiers to count them. None of the objects required measure words for quantification.
The questionnaires were delivered to the children’s parents. For the 120 noun-classifier task, they were asked to fill in classifiers separately so that the child, the mother, and the father each gave answers of their own.
There are three main findings: The first is the lack of homogeneity in the children’s classifier usage. Out of the 30 items, only 6 had 80% of participants’ agreeing on their corresponding classifiers. The second is the significant deviation of the most often used classifiers from the correct answers. The third is the relatively high frequency of the incorrect use of the general classifier ‘ge’.
With regards to the first finding, given that certain items correspond to more than one classifier, it is not surprising that different people tend to use different classifiers. We examined the items that can be associated with only one classifier. There are 23 of such items, of which, 6 had homogeneity (at least 80%) in classifier usage. Since we’re discussing classifiers here, we do not include measure words (they will be discussed later). The final tally is then a total of 20 items with classifiers for 4 items being homogeneous. Respondents agreed on the appropriate classifier only one-fifth of the time.
.
An explanation for this lack of consensus would be participants making wild guesses when they did not know which classifier to use. If so, we should expect more homogenous answers from the older children (whom we assume to have a better command of the language). We divide the participants into two groups of ages 10-12 and 79 respectively to test this hypothesis. The results show that the older group had a higher percentage of same answers for 17 items, an equal percentage for 5 items, and a lower percentage for 8 items. In other words, the older group provided more uniformed (but not necessarily correct) answers than the younger group. This evidence proves that the lack of homogeneity was a result of the participants making wild guesses.
The second finding - the deviation of the most often used classifier from the correct answers. The most often used classifier for 5 of the 30 items was different from the correct answers. There are two possible reasons for this. The majority of the participants may have used the incorrect classifier simply due to low levels of Mandarin Chinese language competency. The deviation may also be due to the influence of a certain Chinese dialect spoken in Singapore. The two reasons are not mutually exclusive. It is also possible that the deviation stems from both reasons. For example, in Singapore, ‘li’ is used to classify anything that is round, regardless of size. However, in Mandarin Chinese, ‘li’ has a size restriction (small) but no shape restriction.
The third finding - the relatively high frequency of the incorrect use of ‘ge’. Twenty of the 30 items were incompatible with ‘ge’. Of this 20, ‘ge’ was one of the two most-used classifiers for 9 items. No other classifier was used with such regularity. A possible inference we may draw from this data is that ‘ge’, a generic classifier, was the default choice when the participants did not know the correct answer.
The reasons the participants provided as to why they used the classifiers the way they did can be broadly grouped into rule-based reasons, observation-based reasons, and others. The rules were either self-generated or learned. Observation-based reasons explain the usage based on observing others’ use of classifiers. The other reasons relate to perception, cognition, and the presence of default classifiers.
Noun and classifier association may be broadly grouped based on the rules by quantity, size, shape, and other aspects of the objects. In formal schooling, children were not taught that there are rules to all classifier usage. However, they were taught the specific rules for individual classifiers. They might have extended their understanding to all classifiers, resulting in both learnt and self-generated rules. Their selfgenerated rules were found to be more likely to be mistakeprone and inconsistent.
Quantity: The quantity of the objects in question influenced the classifier chosen. When questioned about the use of ‘ge’, many participants replied that ‘ge’ was used because there was only one item displayed. The use of ‘ge’ and the above reason were especially common when the item displayed usually came in multiples (e.g. shoe, hand, leg, and tree). Similar reasons were given to explain the use of ‘zhi’, ‘tiao’ and ‘zhang’.
Many participants used ‘shuang’ and ‘dui’ for items that are normally found in pairs (e.g. shoe, hand, and leg), even when there was clearly only one present. However, Chinese noun classifiers are not used based on different quantities. Quantities may be indicated by measure words. This result may be due to the confusion between measure words and classifiers.
Size: The size of the objects also determined the use of classifiers. A number of participants used ‘jia’ for a shelf, drawer, and computer because these items are big. This result is mainly due to the fact that there are size restrictions for certain classifiers.
Shape: Shape was also a factor in the choice of classifiers. Some children used ‘pian’ for paper and ‘zhang’ for door as both ‘pian’ and ‘zhang’ are supposed to be used for flat objects. Also, one child used ‘lun’ (the same word for a tire, something with a round shape) for keys as the keys in the picture were attached to a round key ring.
Other features of the noun objects: According to the children’s reasoning, ‘jia’ was for windows as they came in a set, ‘zhi’ for a bag as it could contain things, ‘ba’ for a pen as it could be held in one’s hand, and ‘tai’ for a telephone as it could be put on a table top. The rules or criteria that the children described for ‘jia’, ‘ba’, and ‘tai’ were correct although the application was incorrect. The rule for ‘zhi’ was incorrect. Some contradiction and inconsistency were observed in the rules that the children applied to classifier use. For example, one child first explained the use of ‘jia’ by referring to the large size of an object. However, the same child later explained the use of ‘jia’ for another object by referring to its small size. These cases were not very common though.
Perceived similarity with other objects. Participants tended to use the same classifier for objects that they perceived to be similar. Examples are provided in Table 1 below.
Participants’ reasoning show that they sometimes were confused between classifiers and temporary measure words. For example, a percentage of participants used ‘bei’ to classify a glass. The reason they gave for doing so was that ‘yi bei shui’ (a glass of water) is what they always said. However, ‘bei’ in ‘yi bei shui’ is a temporary measure word to quantify water. It is not a noun classifier. Furthermore, the object quantified in ‘yi bei shui’ is the water, not the glass. It is most likely that they used ‘bei’ because a glass is ‘bei’ in Chinese and they thought that ‘bei’ was a classifier.
Further examples of this type are the children’s use of ‘zuo’ (sit) to classify a chair, and ‘bao’ to classify a ‘shu bao’ (school bag)
Another finding is the prevalence of default classifiers. Many participants said that there were certain classifiers they used when they did not know the correct classifier. The most common default classifiers were ‘ge’ and ‘zhi’ respectively.
In this section, we attempt to find out whether gender, age, school level, and the nature of the objects and classifiers have any influence on the children’s correct rate of classifier usage. The mean score for all thirty children was 53.33%. This is comparable to the results of Gao’s (2010) study of Swedish-Chinese bilingual children.
Gender. The average score for boys (57.14%) was slightly higher than the average score for girls (54.17%). However, the difference was not statistically significant at the 95% confidence level.
Age/School level. On average, the older children (upperprimary level) scored better than the younger children (lower-primary level). The exact breakdown is shown in Table 2.
Age Primary Level Mean Score
Seven One 46.67%
Eight Two
Nine Three
Ten
Four 49.44%
A regression of scores on age found a weak but positive (beta=0.039, significant at 95% confidence level) relation between age and scores. Older children tended to do better than the younger children, albeit only slightly.
Properties of Objects. The three best and worst scored objects are presented Table 3.
Object Classifier Percentage Correct
If the frequency of usage of objects is a determiner of correct usage of corresponding classifiers, then all six of the objects presented in the above table ought to have high scores. In fact, it is likely that the participants uses chairs, keys, and lamps more often than they used dictionaries.
Among the objects that the participants scored best in, both ‘book’ and ‘dictionary’ require the same classifier ‘ben’. This classifier is used with objects that comprise of bound pages. It is inherently clear what objects ought to be paired with the classifier ‘ben’ and thus easy to acquire.
In comparing this to the objects that the participants
scored poorly, two of the low-scored objects, chair and key,
require the classifier ‘ba’. This classifier is meant to pick
out objects that can be gripped with one’s hand. Compared
to ‘ben’, ‘ba’ is a more difficult classifier to use as what
constitutes an object that can be gripped in one’s hand can
be rather opaque. According to
Properties of Classifiers. Table 4 lists the children’s classifier correct use rate. Some classifiers are used rarely. This may be an indication that they are not commonly known classifiers to children. They probably have to have a better command of the Chinese language. This is supported by the fact that they were mostly (75%) used by older children (9 years and above). If we compare classifiers that were used at least 20 times, the three classifiers that meet the criteria and were most used correctly are ‘ben’, ‘feng’, and ‘jian’. The classifiers used least correctly were ‘tai’, ‘zhi’, and ‘zhang’. Zhan 7 0 7 100.00% Dao 1 0 1 100.00% Ben 59 1 60 98.33% Feng 24 1 25 96.00% Tiao 14 1 15 93.33% chuan* 6 1 7 85.71% Jian 21 4 25 84.00% Ba 31 6 37 83.78% Ge 147 83 230 63.91% shuang* 58 36 94 61.70% Zhang 52 35 87 59.77% Zhi 40 43 83 48.19% Tai 13 14 27 48.15% *measure words
The three classifiers that were most often used correctly
are more specific in their requirements than those that were
most often used incorrectly. ‘Ben’ is used for bound
materials, ‘feng’ is used specifically for letters only, and
‘jian’ is most commonly used for articles of clothing worn
on the torso. On the other hand, ‘tai’ is used for mid-sized
electronics. ‘Zhi’ is used for a variety of objects from small
animals to certain body parts, and ‘zhang’ is used for things
that are flat or have a flat surface (usually paper products).
This shows that the higher the degree of specificity in its
requirement, the lower the probability of misusing the
classifier. This result contradicts with
Instead, the result shows that the higher the degree to which the object coincides with the requirements of the classifier, the higher the likelihood of correct classifier usage. Also, the more specific the requirements of the classifier, the more likely it will be used correctly.
Two types of incorrect usage stand out regarding the use of ‘shuang’. The first is its inappropriate association with items that are not a matching pair in the common sense of a Chinese language speaker. Many participants used ‘shuang’ with pants and scissors. Upon questioning, most participants said that such usage was the effect of spill-over from the English language. Both pants and scissors require the quantifying phrase ‘a pair of’ in English usage. This was then simply translated into ‘shuang’ in Chinese. This response was independent of age, suggesting that absolute levels of English exposure had little effect on the magnitude of transfer.
The other type of incorrect use is the matching of ‘shuang’ with paired objects even when the object is presented in singular. The percentage of participants who used ‘shuang’ in this manner is presented in Table 5.
The relatively high incidence of this improper usage
seems to suggest that the participants are not entirely aware
of the fact that ‘shuang’ is used for paired items when the
pair is present. If this is the case, then it is strong evidence
for the input-driven hypothesis whereby children, upon
hearing the phrase ‘yi shuang xie’ (a pair of shoes) simply
memorise ‘shuang’ coming before ‘xie’ (shoe). In a similar
study,
In this section, we use a quantitative approach to sift out which factors have an impact on classifier competency of bilingual Singaporean children. Acceptable but awkward uses of classifiers, such as “ge” in ‘yi ge xie zi’ (a [classifier] shoe) were considered incorrect to measure true competency. Hence, the scores presented in the following sections are not measures of proficiency in the usual sense but measures of mastery. However, the terms proficiency, mastery, and competency will be used interchangeably.
The factors we examined were gender, age, schooling, scores of parents, age of parents, education level of parents, the main language spoken to the child, the main language parents speak to others, housing, and income.
Schooling. A one-way ANOVA was conducted to check if schooling influenced classifier mastery. The participants were grouped into private lower primary, public lower primary, and public upper primary. The mean score for each group was 28.33%, 49.00%, and 40.00% respectively. There were no participants belonging to the other classes. The result was a significant difference at 90% (p=0.1) confidence level but not at 95% (p=0.05) confidence level.
Parents’ scores. The mean score for mother, father, and child was 51.73%, 52.14%, and 45.22% respectively. The scores of parents were moderately correlated to the children’s scores (Correlation coefficient for Father-Child=0.6273, Mother-Child=0.6768).
Interestingly, the children’s scores correlated positively to their mothers’ but negatively to their fathers’. To have a better picture, we regressed children’s scores on their parents’ scores, as shown in Table 6.
The negative coefficient of the father’s score and the positive coefficient of the mother’s score (both significant at 95% (p=0.05) confidence level) are further proof of the earlier point.
The positive correlation between the mothers’ scores and the children’s scores was not surprising. Children are likely to mimic what they hear and it is fair to assume that they hear their mothers’ speech most. Thus, children whose mothers are more proficient in classifier usage are likely to score better.
The reason for the negative relation between the fathers’ scores and the children’s scores is not quite clear. Perhaps fathers’ wrong usage prompted children to watch their own usage.
Fathers’ age. A regression of children’s score on fathers’ age was performed. The result shows that a weak but positive relation (beta=0.036) between them, significant at 90% (p=0.1) confidence level.
A one-way ANOVA was also conducted on the mean score of the children whose fathers belonged to different age groups (31-40years, 41-50years, and 51-60years) to confirm the above result. The differences between the means were significant at 95% (p=0.05) confidence level.
A possible explanation is that older fathers possibly used Mandarin Chinese more often, leading to the children being more exposed to Mandarin Chinese. As such, the children achieve a higher level of language (inclusive of classifier) competency. The probability of this being true will increase if older fathers really speak more Mandarin Chinese, and if children’s language environment has a significant impact on their language proficiency. We do not have any evidence of the former condition. The latter condition will be discussed in the later section.
Mothers’ Academic Achievement. Participants fell into four groups. Children whose mothers completed primary school, secondary school, tertiary studies (either junior college or polytechnic), or university. The mean score for each group was 49.99%, 46.67%, 62.5%, and 28.75% respectively. A one-way ANOVA at 95% (p=0.05) confidence level showed the differences between the means to be significant.
Income. The mean score of children from families of different income levels is shown in Table 7.
A one-way ANOVA shows that the differences between the means were significant at 99% (p=0.01) confidence level. This result probably stemmed from the correlation between education and income.
Gender, age, mothers’ age, father’s academic attainment,
housing, and the main language parents spoke to children
showed no statistically significant influence on children’s
scores. The main language parents spoke to children was
statistically insignificant (0.8<p<0.95). This is rather
surprising. Perhaps the language influence of parents on
children has been diluted.
Bilingual Singaporean children learn and use Chinese
classifiers by feature-based reasoning, observing, generating
or learning rules. Their feature-based reasoning was based
on their cognitive understanding of the properties of objects.
The rules based on which they applied to the use of
classifiers were either generated by themselves through
learning from the people around them, such as their parents
and teachers or the results of their own reasoning based on
their understanding of the functional use or perceptual
features of the noun objects. This is similar to what