This study seeks to investigate the development of noun and predicate comprehension and production in isiZulu-speaking children between the ages of 25 and 36 months. It compares lexical comprehension and production in isiZulu, using an Italian developed and validated vocabulary assessment tool: The Picture Naming Game (PiNG) developed by Bello, Giannantoni, Pettenati, Stefanini and Caselli (
Research on the emergence of early lexical comprehension and production is very important for both enhancing our understanding of language acquisition and diagnostic purposes. In the South African context, improving numeracy and literacy skills remains a huge challenge in the public education sector (Department of Basic Education,
The other challenge is finding locally appropriate standardised tools that are relatively easy to use in any given language. It is heartening, however, to note that several researchers in speech and language therapy have adapted several international standardised tests to the local context and are starting to include crucial information in the measurement of children’s typical and atypical development. Having normative data allows for interventions in delayed or impaired language as well as effective language teaching and learning (Kathard
Despite an increase, there are still too few academic studies on the comprehension and production of South African Bantu languages. Those available include studies focusing on phonological development (Gxilishe,
Although these studies have greatly advanced our knowledge of less-studied languages, no one has managed to present comprehensive linguistic research on the developmental aspects of the lexicon inventory of the Bantu-speaking child, similar to that regarding children who speak western languages, such as English, French, and Italian. Researchers tend to focus on particular aspects within linguistic theory (Berko Gleason & Bernstein Ratner,
This study seeks to investigate the development of noun and predicate comprehension (PC) and production in isiZulu-speaking children between the ages of 25 and 36 months. The study is part of an international research collaboration that aims to investigate speech and co-speech gesture production and comprehension development in children. It compares lexical comprehension and production of two romance languages, Italian and French, and two South African Bantu languages, isiZulu and Sesotho, using an Italian developed and validated vocabulary assessment tool. This paper presents the preliminary findings of the lexical development of isiZulu speakers.
We use words to communicate about everything related to our physical environment, including events, activities, people, objects, places, relations, properties, and states of being (Clark,
It is well documented that typically developing children are accurately able, by 3.5 years of age, to produce most of the basic morphosyntactic structures of their languages such as relative clauses, the passive construction and other complex forms (Bates & Goodman,
In the cross-linguistic study by Caselli
Bantu languages are typologically similar and share several typical grammatical features. IsiZulu is a South Eastern Bantu language of the Nguni cluster spoken primarily in South Africa (especially the southeastern areas of Kwa-Zulu Natal), but it also has speakers in neighbouring countries. IsiZulu is highly mutually intelligible with other Nguni languages, such as isiNdebele, isiXhosa, and siSwati. In 2011, South Africans citing isiZulu as their home language numbered 11.5 million, or 22.7% of the population, the language that has the highest number of speakers (Census,
IsiZulu is a Subject-Verb-Object (SVO) language with a high number (about 15) of noun classes, triggering the agreement of verbs, adjectives, and other elements. In other words, ‘nominal and verbal modifiers follow the noun and verb respectively, and grammatical morphology is prefixed to both nouns and verbs’ (Demuth & Suzman,
Demuth (
Despite the numerous studies on the isiZulu verb (or related languages), we have not come across literature that documents the acquisition of verbs, adjectives, and adverbs or the noun and its morphology.
Studies on South African Bantu languages are definitely increasing, but as yet there has been no study that has looked at simultaneous comprehension and production during lexical development.
We focus on the lexical development of comprehension and the production of nouns and predicates from a speech perspective. Specifically, this article seeks to explore the lexical development of isiZulu using the adapted assessment tool with three main objectives. These are to:
Test the effectiveness of the adaptation of a vocabulary tool to measure isiZulu comprehension and production development.
Test the universal finding that both comprehension and production performance increase with age for a less-studied language, isiZulu.
Present our findings on the comprehension and production of the linguistic categories of nouns and predicates.
Early childhood development research shows a strong interdependence between vocabulary, phonology, and grammar in both typical and atypical populations (Marchman & Thal,
Constructed and validated in Italy, the Picture Naming Game (PiNG) was specifically developed to assess lexicon production and comprehension in children between the ages of 19 and 37 months, involving the consideration of both nouns and predicates, and based on the Italian MB-CDI. Previous studies have shown that it is extremely relevant to investigate the relationship between vocabulary comprehension and production as well as between nouns and predicates, as these skills and their relationships are indicators of both the level of language development and conceptual organisation. In general, studies using the PiNG tool with Italian children proved that the comprehension subtests were easier than the production subtests, thus allowing for their administration in younger children and resulting in fewer errors. Similarly, children found the noun subtests easier than the predicates subtests. Therefore, lower variability was found in vocabulary comprehension compared with production, and in nouns compared with predicates for hearing children (Bello
The PiNG tool consists of two sets of colour pictures and contains two tasks; comprehension and production tasks which in turn contain four subtests. The first set has 22 images (20 test pictures and two pre-test pictures) of objects and tools, animals, food and clothing (e.g. a fork, a lion, bananas, gloves) and is used in evaluating the comprehension and production of nouns in the noun comprehension subtest (NC) and the noun production subtest (NP), respectively. The second set contains 22 images (20 test pictures and two pre-test pictures) showing actions, location adverbs, and/or adjectives (for example, to push, close by or far away, big or small) and is used to evaluate the comprehension and production of predicates in the PC subtest and the predicate production subtest (PP), respectively. The original PiNG test for Italian children was adapted from the Italian MB- CDI and the items had different levels of difficulty. It included items that were ‘easy’, ‘moderately easy’, and ‘difficult’, based on the Italian normative sample (Bello
In this paper, we report on the adaptation of PiNG to isiZulu. The PiNG tool has already been successfully adapted to other languages and cultures. For example, a study by Pettenati and colleagues provided the first occasion for a cross-cultural comparison of gestures and vocabulary production and comprehension of 22 Italian and 22 Japanese children between 25 and 37 months of age (Pettenati
Translation of the set of nouns (20 target nouns in the comprehension task +20 target nouns in the production task +2 × 2 = 4 lexical items for the pre-tests) and the set of predicates (20 target verbs and/or adverbs and/or adjectives in the comprehension task +20 target verbs and/or adverbs and/or adjectives in the production task +2 × 2 = 4 lexical items for the pre-tests) was carried out in isiZulu by the researcher, who is a native speaker of isiZulu and a linguist, together with two isiZulu-speaking research assistants, who are also linguists. The translation was further tested in a pilot study of native Zulu adults for validation (see Stage 2).
Of particular interest in the international collaborative comparative study is the different language typology of romance languages that are analytic and Bantu languages that are agglutinative. In the initial adaptation, a conscious decision was made to adapt the protocol questions as closely as possible to the original Italian version, that is, questions were to be ‘neutral’ so as not to give a clue to the answer. For example, in the Italian version, the question would be translated to ‘show me running’, which does not provide any clue to the participant for the comprehension task and, therefore, the speaker cannot get a clue on the referent. A participant could choose any item he or she deemed fit. However, because of the Noun Class system, agreement concords, and morphosyntax structure of Bantu languages, the question must have the relevant noun class and subject concord, which may give a clue to the item. For instance, in isiZulu, a semantic translation for the above example would read
Twenty two adults (11 males and 11 females) participated in the Zulu adult pilot study. Participants were drawn from the pre-dominant isiZulu speakers of Kwa-Zulu Natal, the south-east region of South Africa. Participants were university students from various communities in the Kwa-Zulu Natal area, 60% of whom were from the Empangeni area. The other 40% were from surrounding areas: Pongola, Harrismith, Durban, Eshowe, and Pietermaritzburg. Applying a neutral questioning style did not work. Participants would reformulate the question or stop the interviewer to ask for more clarity. If the questions were amended, participants answered with no difficulty. The inclusion of the class prefixes did not affect the results but rather assisted the participant in understanding what was requested of him or her. This was indicated by the fact that once the correct class prefixes were used, the participants would indicate that they did not recognise an item, or they would give an answer if they did. With the neutral questions, the participant would simply halt the interview, seek clarification, and personally supply the class prefixes. When the interviewer asked why the participant reformulated the utterances, all participants stated that the ‘neutral’ utterances were grammatically correct, but ambiguous and confusing. It is interesting to note that all 22 participants corrected the utterances.
Agreement between participants was 95% for the comprehension subset and 86% for the production subset. Four items under the NP subset produced either no responses or ‘I do not know’ answers, referring specifically to bidet, radiator, penguin, and seal. Under the PP, two predicate task words produced a low frequency of correct target words (spinning, heavy, far apart).
From the adult pilot results, it became clear that the isiZulu version of PiNG needed further adaptation before the pilot study with children was initiated. The adults seldom produced words in isiZulu for ‘seal’ and ‘penguin’ and so these items were changed to ‘snail’ and ‘crocodile’, respectively. As both ‘radiator’ and ‘bidet’ are foreign cultural objects, these two items were replaced by ‘heater’ and ‘toilet’. Some pictures were specifically cultural, such as the picture of the ‘roof’, which was a European type of roof, but in order to allow a systematic comparison with other languages in the four-language project, some items were retained for future adaptation.
After the changes to the above-mentioned picture items, a pilot study was conducted with 15 Zulu children. The group included five children aged 25 months (±1 month), five children aged 30 months (±1 month), and five children aged 36 months (±1 month). This was done in order to test the corresponding adaptation of PiNG on a small sample in case further adaptation was needed before going onto the main study. Participants were drawn from Kwa-Zulu Natal, the same area where the adult pilot study was conducted. Data was collected from the urban townships of Empangeni and Ngwelezane on the northern coast of Kwa-Zulu Natal. The principals and caregivers were very helpful in providing the researchers with the children’s clinic cards. These vaccination cards aided the researchers in selecting participants for the appropriate age cohorts. The files provided by the teachers also gave the researchers additional information, including, for example, whether a child had been born prematurely or had a learning disability and needed to be excluded from the selected participants.
We worked with nine crèches to ensure that our three age groups were exact. Many crèches could not form part of our participant sample because those children were bilingual and would alternate between naming items in English and isiZulu. We finally chose two schools from the city centre of Empangeni and four schools in the Ngwelezane township. Administration of the PiNG tool began with a familiarisation phase that involved playing various card games, counting games, and naming games with the children. Researchers also played with the children in the school playgrounds on the swings and slides. Once the researchers felt that the children were comfortable enough, they asked the children to play a game with them in front of the camera.
The main study for the isiZulu data was collected from Soweto in the Gauteng province. The move from KZN was purely logistical as the researchers were all Gauteng based. Monolingual isiZulu-speaking children were chosen with the help of their caregivers. Children’s vaccination cards were examined to exclude premature babies or those with any recorded pathologies. All crèches require clinic or vaccination cards in order to enrol the child. The caregivers also assisted in selecting children, who they said showed no language delays in comparison to their peers. All selected children had parental consent (see the Ethics section).
Forty-nine children from four neighbouring crèches participated in the study. Nine children were excluded from the data sample for various reasons: some children did not complete the two tasks, some children were bilingual and code-switched regularly, some children spoke too softly for the camcorder to record sound, and one child was sleepy and had to go for a nap. For this study, 36 participants were chosen with 12 per age cohort in consideration of gender balance. There were a total of 19 females and 17 males across the different age cohorts (
Zulu participants age groups.
Group in months | Age range | Average age m; d | SD m; d | Number of children | ||
---|---|---|---|---|---|---|
Min | Max | Female | Male | |||
25 | 23.08 | 26.14 | 24; 29 | 0m; 29d | 7 | 5 |
30 | 28.01 | 31.26 | 30; 4 | 1m; 13d | 7 | 5 |
36 | 34.16 | 38.29 | 36; 13 | 1m; 7d | 5 | 7 |
The procedure of this study followed those of previous studies (Bello
After the familiarisation period, during which we played different card games with the children, all children were tested individually at their schools. Three sets of pictures per set were presented to each child on a small table. The first part of the task was comprehension, in which a child was asked ‘Where is the cat? Show me the cat’ for a noun comprehension item or ‘What is this child doing? What is this one doing?’ for the PC item. The second part of the each subset was production, in which the child was asked, ‘What is this?’ for the NP subset or ‘What is he doing?’ for the PP subset. The third card was a distractor to eliminate the choice by luck. A total of 22 cards were presented for the noun subtest, and another 22 cards were presented for the predicate subtest. The first two sets were pre-test cards to ensure that the child understood what was expected. The data was based on the remaining 20-card set per subtest. For the comprehension task, only one prompt was considered. For the production task, if the child struggled with producing the correct item, a second prompt was used. All elicitations were filmed for later data coding and analysis.
Two research assistants and the researcher of this study collected the data. All researchers are first-language, native speakers of isiZulu, linguists with fieldwork experience in the collection of data from children.
The coding system was adapted from previous studies (Bello
For the comprehension subtests (NC and PC), if the child indicated (either by pointing, showing, or verbalising) the photograph corresponding to the item indicated by the adult researcher, the answer was considered correct. If the child selected the no target photograph or did not respond at all, the response was coded as incorrect or no response, respectively. Similarly, in the production subtests, if the child produced the target lexical item, their response was coded as correct. If the child produced a non-target item or did not respond at all, their response was coded as incorrect or no response, respectively. For some photographs, more than one answer was accepted as correct; for instance, for the ‘diaper’ item, some children called it
Synonymous items were considered to be correct synonyms, for instance
Three trained native isiZulu-speaking research assistants and the researcher independently coded the verbal transcription, that is, orthographical transcription directly from the film footage. Two different research assistants, who are also trained linguists, coded the classification of the speech responses as well as those of gesture. Disagreements were resolved through discussion.
After the annotation phase was completed, all data was exported to Excel for an ultimate verification (internal consistency on the coding) and statistical analysis.
Ethical considerations guided the pilot study as well as the main study. All children who participated were recruited on a voluntary basis after their caregivers signed an informed consent form, which was provided in their language, and after they themselves agreed to participate at the start of the task (‘Nouns’ or ‘Predicates’). Parents or members of the crèche were welcomed in the room during the administration of the tool. The tasks were interrupted or ended if a child verbalised a desire to stop, or expressed discomfort by crying and/or withdrawing. Children’s identities were kept confidential, and data obtained from this project were not disclosed to any third party. The Wits HREC Non-Medical Ethics Committee granted ethical clearance for the study (protocol number H13/08/43).
Children’s responses were analysed according to the coding criteria listed in the method section. For our first objective, we analysed the comprehension and production tasks across the three age groups to test whether the PiNG assessment tool was effective in detecting the development of comprehension and production in isiZulu. Our second objective overlapped our first, and so our first finding addresses both of our objectives.
For the comprehension and production tasks, an analysis of variance, ANOVA, was run with the age group as the independent variable.
The correct answers for the comprehension and production task items are illustrated in
Comprehension and production task scores per age group.
Variables | Comprehension task | Production task | ||||
---|---|---|---|---|---|---|
Age group | G25 | G30 | G36 | G25 | G30 | G36 |
Mean | 12.8 | 13.6 | 14.7 | 8.1 | 8.8 | 10.7 |
SD | 2.9 | 2.4 | 2.8 | 2.7 | 2.5 | 2.5 |
A significant difference across the age groups emerged at (
In order to test the comprehension and production tasks per lexical categories, we looked at the performance of the nouns and predicates across the age groups. We performed an ANOVA between groups per lexical subset.
The comprehension of the noun and predicate subtests was better performed than the production of noun and predicate subtests across the age groups. In
Lexical comprehension noun and predicate subtest means (SD) per age group.
Age group | Noun comprehension mean | Predicate comprehension mean | ||
---|---|---|---|---|
SD | SD | |||
Group 25-months | 13.75 | 2.2 | 11.75 | 3.2 |
Group 30-months | 14.67 | 1.6 | 12.58 | 2.8 |
Group 36-months | 16.08 | 2.5 | 13.33 | 2.4 |
Lexical production noun and predicate subtest means (SD) per age group.
Age group | Noun production mean | Predicate production mean | ||
---|---|---|---|---|
SD | SD | |||
Group 25-months | 7.75 | 2.3 | 8.50 | 3.0 |
Group 30-months | 8.50 | 1.9 | 9.08 | 2.9 |
Group 36-months | 11.42 | 2.9 | 10.00 | 1.8 |
For the production of nouns and predicates, we noted a similar pattern in that children performed better at labelling the correct items in the NP subtest (
For our third objective, we present our findings on the performance of the lexical items. In order to have a better understanding of the performance on the comprehension and production task, as well as noun and predicate subtests, we ranked the items in terms of correctness according to the total sample shown in
Percentage of correct answers for comprehension and production provided by (
Variables | Percentage of correct answers by children |
---|---|
Doll | 100 |
Hat | 100 |
Boots | 100 |
Bidet/toilet | 97 |
Motorcycle | 94 |
Spoon | 94 |
Apples | 92 |
Sofa | 92 |
Iron | 86 |
Cow | 86 |
Watch | 83 |
Glasses | 81 |
Box | 72 |
Clouds | 53 |
Terrace | 50 |
Mountain | 44 |
Penguin/snail | 44 |
Elephant | 44 |
Bib | 42 |
Hammer | 39 |
Comb | 92 |
Socks | 92 |
Bananas | 89 |
Bag | 81 |
Umbrella | 78 |
Hen | 75 |
Fork | 78 |
Glass | 78 |
Table | 72 |
Radiator | 53 |
Gloves | 33 |
Diaper | 31 |
Picture | 28 |
Truck | 19 |
Book | 19 |
Beach | 14 |
Lion | 8 |
Seal/crocodile | 3 |
Roof | 0 |
Flags | 0 |
To pull | 94 |
To sweep | 92 |
To comb | 92 |
To drink | 89 |
To bite | 86 |
Dirty | 83 |
To swing | 81 |
To run | 81 |
To scramble up | 75 |
Big | 69 |
To build | 69 |
High | 56 |
To greet | 56 |
To walk | 47 |
Behind | 44 |
Close | 39 |
Full | 31 |
Outside | 31 |
To embrace | 31 |
Short | 25 |
To push | 94 |
To drive | 89 |
To eat | 89 |
To phone | 83 |
To wash | 72 |
To play | 69 |
To kiss | 67 |
To open | 64 |
To fall | 64 |
To laugh | 61 |
To swim | 53 |
Small | 50 |
Clean | 28 |
To turn | 17 |
Empty | 6 |
In front of | 6 |
Heavy | 3 |
Far | 3 |
Inside | 0 |
Long | 0 |
Under the noun comprehension subtest, three items were perceived correctly by all 36 children; ‘doll’, ‘hat’, and ‘boots’ had a 100% response rate across all ages. Five items had less than 50% success; ‘mountain’, ‘snail’, ‘elephant’, ‘bib’, and ‘hammer’, meaning that fewer than 18 children across the three age groups found these items ‘difficult’.
The photograph of the mountain showed a rising mountain of the European Alps, a type of geographical feature that is not commonly seen in South Africa. It was interesting to note that the children had difficulty identifying ‘snail’ and ‘elephant’, but managed easily to identify the domestic animals such as ‘cow’. The ‘bib’ and ‘hammer’ were also not easily identified: the children gave answers that focused more on the baby who was wearing the ‘bib’ and did not respond at all to the ‘hammer’ or said,
Under the PC subset, there were seven items that had less than 50% success; ‘walk’, ‘behind’, ‘close’, ‘full’, ‘outside’, ‘to embrace’, and ‘short’.
Some of the responses for ‘walk’ were
Overall, production items scored lower than comprehension items, as seen in
The 36-month-old group was more familiar with the leather gloves depicted in the picture with most children saying
The ‘picture’ item was also more familiar to the 36-month-old group, even though most children across the age groups identified it as a TV, because the picture looked like a flat screen TV. The picture depicted a lone beach with a blue sky, the ocean, and a strip of sand. The picture also garnered ambiguous responses from the adults in the pilot study.
The item ‘truck’ produced the semantically related ‘a car or a bus or a big car’. This item did not display any developmental trajectory, as 36-month-old children were equally likely to produce ‘car’ rather than the target word
The ‘book’ item mostly produced
The wild animals ‘lion’ and ‘crocodile’ produced very interesting responses such as
The ‘roof’ item depicted a European type roof and a portion of a house with trees surrounding it. Most responses were either
In the PP subtest, the following seven items were difficult for the children: ‘to turn’, ‘empty’, ‘in front of’, ‘heavy’, ‘far’, ‘inside’, and ‘long’.
The item for ‘to turn’ depicted a group of children on a merry-go-round. Children’s responses included
For the adjectives and adverbs ‘empty’, ‘in front of’, ‘far’, ‘inside’, and ‘long’, the responses did not show correct labelling across all ages. These items showed a developmental pattern when considered in the light of that the responses showed that the children acquired this concept with age.
Interestingly enough, five of these items were in the same set at the PC subtest, which shows consistency in the production of these categories. All the children had difficulty perceiving the item ‘heavy’. This item depicted a young child carrying a brown torn box while grimacing. Children’s responses focused on the child or the box being torn.
Our study begins with an Italian picture naming assessment tool being adapted to a Bantu language, isiZulu. The tool is designed to directly observe the lexical composition of vocabulary in two related tasks; comprehension and production. To our knowledge, most studies on isiZulu or related Bantu languages have either directly observed either comprehension or production, but never both at the same time. Several pilot studies in both adult and children populations enabled us to alter obvious elements of cultural bias. Although some other problematic items remained, we decided to preserve our initial goal by keeping as many items as necessary for our systematic comparison with two romance languages, Italian and French, as well as another South African Bantu language, Sesotho.
Certain items such as the picture of a ‘roof’ did not depict a ‘roof’ as seen by many South African children. Some items such as ‘to turn’, which depicted a merry-go-round produced unexpected results in that one would expect most children to have been exposed to a merry-go-round as these are commonly found in parks, but instead the children focused more on the people in the picture than on what they were doing. Despite the cultural differences that may have stemmed from the images of our stimulus, we note that our findings confirm what has been long documented in literature: that comprehension comes before production. In a related cross-linguistic comparison, Japanese children showed a lower lexical production compared with the Italian children (Pettenati
In terms of development, our results show how age affects both comprehension and production: 36-month-old children performed better than 30-month-old children who in turn performed better than 25-month-old children. Our statistical analysis did not discern a significant difference between the 25-month-old group and the 30-month-old group for comprehension, but it detected a significant difference between the 36-month to the 25-month and 30-month groups in terms of production. This was not surprising as, in the first study by Bello
In terms of production, our findings showed an overall effect of age, which meant that production does indeed lag behind comprehension. Moreover, the larger number of culturally foreign pictures in the noun comprehension subtest suggests that it may be necessary to further adapt the assessment tool to more effectively evaluate the isiZulu lexicon.
Children used a semantic description strategy to try and explain unfamiliar items, which shows that though they may have the concept, the lexical item is not ‘concrete’ enough for them to relate to their physical environment. If perception is difficult, it is harder to retrieve the semantic representation from the lexicon and, as such, retrieval will be impaired (Harley,
The production of the noun category showed an age effect but the predicate category did not show a similar difference. The higher number of adverbs and adjectives in the PP subtest was difficult for all the age groups, which could explain the lack of a significant difference. Alternatively, if verbs are acquired earlier, it shows that the children have reached a plateau phase between the ages of 24 and 36 months. Further research on the vocabulary spurt in isiZulu would shed more light on this issue.
The lack of a child inventory like the MacArthur Bates CDI for isiZulu and related languages is a disadvantage. We have no idea which items are acquired first, nor do we have the exact age of the vocabulary spurt in isiZulu. Future analysis should look at the gender effect. Although gender-related data is available in this study, it has not yet been analysed. It would be interesting to see whether girls have an advantage over boys, as has frequently been reported in western languages.
Developing normative lexical data on Zulu-speaking children or those speaking other related Bantu languages is important for research for both acquisition and clinical purposes. Finding a standardised assessment tool that can be used for South African Bantu languages is an ongoing challenge. The PiNG assessment tool has proved to be robust and effective in directly assessing a child’s lexical vocabulary. For further research into isiZulu, the tool would need further adaptation by replacing some images with items of local content.
The literature states that children begin with language comprehension and, when the motoric and cognitive apparatus develops, language production follows suit. This is a universal linguistic parameter. Children start with objects and events around their immediate environment and quickly learn people’s names, concrete objects around them, and familiar routines coming from their home environment. The first words will therefore largely depend on this input, and cultural as well as linguistic constraints may affect this development.
This study shows that as children get older, comprehension and production improve. It would therefore be very important for researchers or speech therapists to factor input into their intervention therapies. Children will understand and talk more about what they know and what surrounds them. Some linguistic phenomena like adjectives and adverbs are complex and not yet acquired by the Zulu child at 36 months.
The authors declare that they have no financial or personal relationships that may have inappropriately influenced them in writing this article.
R.K.N. was the project leader and was responsible for the experiment, data collection and data management. S.A. performed most of the coding, prepared the samples, and contributed to the conceptualising and analysing of the data samples. A qualitative analysis was done by both authors.