Children learn their first words from social interaction with people in their lives. This interaction is grounded in the concrete, here-and-now of daily events and actions of the young child’s life, and provides a potentially rich oral communicative context within which words are acquired.

Although many words are explicitly ‘taught’ to young children during social interaction, it is generally agreed that there is a large amount of vocabulary that, for all practical purposes, could not have been explicitly taught and must therefore have been incidentally acquired. In other words, it is generally held that children infer the meanings of a large number of unfamiliar words from oral contexts before they start school (Nagy & Herman 1987; Sternberg 1987). By the time 6-year-olds start school, they have a ‘basic’ vocabulary that consists mostly of high frequency words that occur in oral contexts. But how large is a ‘basic’ vocabulary? There are no easy answers to the thorny issue of estimating vocabulary size. With regard to English, Chall et al. (1990) put it at about 4000 words at age 6, while Sénéchal and Cornell (1993) posit 8000 words.

Estimates about vocabulary size and rate of development vary, depending on how the concept of ‘word’ is defined, how word counts in a particular language are undertaken, how word knowledge is defined and assessed, and whether one is referring to HL or FAL vocabulary development. Some scholars define a word as a graphically distinct type (e.g. White, Graves & Slater 1990). However, several scholars such as Laufer and Nation, and Anderson, Nagy and colleagues work with the notion of word families, where a word is defined ‘as a base form with its inflected and some derived word forms’ (Laufer & Nation 1995:312). Thus, for example, happy, unhappy, happily and happiness belong to the same word family. Knowing the meaning of one of these words is assumed to increase the chances of being able to infer the meaning of the other members of the word family. There are usually three to four words per word family.

Furthermore, vocabulary development does not only involve the quantitative accumulation of new words to the lexicon; qualitative development is also an integral part of vocabulary acquisition. Many words in a language’s lexicon are polysemous, conveying different meanings in different contexts (e.g. light comes from the sun; a light wind; light blue). Vocabulary development should therefore also take into account the degree of knowledge, which progresses from receptive knowledge (recognising words when heard or seen) to active or productive knowledge (using words in speech or writing). We always know more words than we use. There are a large number of words that never progress to free active use but remain part of a learner’s receptive vocabulary repertoire.

As the discussion so far indicates, vocabulary development is quantitatively and qualitatively cumulative, and for HL English learners, fairly rapid too, with about 1000 word families (equivalent to about 3000 words) added each year. Increasing age typically reflects increasing word knowledge, unless children have language or learning difficulties, in which case their vocabulary is often below that of their age peers. Proposing a developmental sequence for HL vocabulary, Nation (1990) suggests that by age 5 at least 1528 word families are known (more than 4000 words), by age 10 a child knows about 7020 word families, by age 15 approximately 12 000 word families are known and by age 18 there are at least 17 600 known word families. Researchers (Nagy & Herman 1987; White et al. 1990) estimate that the average school child learns more than 3000 new words each year during the school years, of which at least half are estimated to be acquired from written contexts and the other half from oral contexts. However, there is considerable variation within this average, with some children learning 5000 new words a year and others only about 1000.

Research indicates that variables such as home and school background as well as individual ability affect vocabulary development considerably. As already indicated, socioeconomic status has been found to have an early and persistent effect on children’s vocabulary. In their groundbreaking research, Hart and Risley (1995, 2003) found that children from middle-class homes had vocabularies five times larger than children from very disadvantaged homes. In their comparative study of children’s vocabulary from Grade 1 to Grade 5, White et al. (1990) found that by the end of Grade 4, middle-class children knew an estimated 15 000 words, while the same grade SES children knew an estimated 10 000 words (these estimates were based on graphically distinct words, not word families). In other words, middle-class children’s vocabulary skill grew at a rate of 5200 words per annum, while low-SE status children had a growth rate of around 3500 words.

Further evidence of vocabulary differences between socioeconomic groups comes from Corson’s British research (1983, 1997). Corson found not only quantitative but also qualitative vocabulary differences between working-class and middle-class children. Corson suggested that speakers of some social dialects in English, especially working-class dialects, are hindered from accessing knowledge categories in their school curriculum because of differences in the types of words they acquire. Research evidence in the USA points to SE status differences in vocabulary growing larger with age. In other words, unless there is active intervention to narrow the vocabulary gap, differences between SES groups increase rather than level off (Hart & Risley 1995, 2003).

What about FAL vocabulary development, and what are the estimates with regard to the size of English FAL learners’ vocabularies? The notion of incidental word learning in context is obviously less problematic when applied to the HL because, by the time children start school, they have already acquired basic competence in the language, which provides a scaffold for inferring word meaning from context. Furthermore, they are extensively exposed to the language in print or oral form on a daily basis. The underlying linguistic competence of FAL students is not commensurate with that of HL students, and they typically have less exposure to the language. Thus, not only do they have fewer bootstrapping mechanisms to infer word meaning but they also have fewer opportunities in which to apply word inferencing processes. Laufer (1992) estimates that for English FAL students a threshold of 3000 word families (about 5000 lexical items) is needed for ‘minimal comprehension’ in English. She argues that for learners below this threshold, no amount of general academic ability or reading skills in the HL will enable them to read in a satisfactory manner (Laufer 1992:100).

Turning to South Africa, in the Threshold report, Macdonald (1990) pointed to the immense vocabulary gap that exists between the words that Grade 4 FAL learners know at the end of their school year and the words they were then expected to know the following year in order to understand their textbooks in Grade 5. The learners in her study had a vocabulary of about 800 words, and they did not seem to have encountered more than half of the words used in their science textbooks. Twenty-four years later, Sibanda’s study (2014) had similar findings, where Grade 4 learners tested on 60 high-frequency words that occur in South African textbooks performed poorly. These findings point to the critical need for special emphasis to be placed on vocabulary interventions in schools.

Although there is a strong association between vocabulary and reading, the nature of the relationship changes at different stages of schooling, with vocabulary becoming an increasingly stronger determinant of reading ability as children move up the schooling ladder. The following trends have been observed:

Given the multilingual nature of South African schooling, what guidelines does the curriculum provide regarding vocabulary development? The Foundation Phase² Curriculum and Assessment Policy Statement document (referred to as CAPS) for English FAL states that by the end of Grade 3 learners should understand 1500–2500 English words in context (Department of Basic Education 2011:22). No comparable recommendations about vocabulary size are found in the English or African Home Language CAPS document. The somewhat inconsistent guidelines for vocabulary development across HL and FAL documents suggest that further research is called for in this domain.

Mindful of two factors – the expected differences in vocabulary knowledge in HL and FAL, and the barriers that low SE status can impose on learning and vocabulary knowledge across languages – the main aim of this study was to investigate the vocabulary levels of Grade 3 learners in schools that serve disadvantaged communities. This is a small-scale exploratory study of Grade 3 learners in four schools in the Western and Eastern Cape provinces. We captured the aim in the following research questions:

Two different low-SE urban primary schools in the Eastern Cape and two in the Western Cape participated in the study. The two schools in the Western Cape were English HL primary schools, situated in the Cape Flats and serving mainly low-income mixed race communities. The schools had Afrikaans as their First Additional Language in Foundation Phase. These will be referred to as schools WC1 and WC2.

The two schools in the Eastern Cape had Xhosa Home Language as their LoLT and English as their FAL in Foundation Phase. These are Quintile 3 schools situated in township areas of Port Elizabeth. These will be referred to as schools EC1 and EC2. From each school, two classes participated in the vocabulary research, amounting to a total of eight classes, four from each province.

All the schools were part of a larger three year project, the Zenex Literacy Project (2015-2017), that worked with Foundation Phase teachers in Quintile 1–3 schools³ across three provinces, the aim of which was to improve learner literacy levels through improved classroom literacy practices. The central focus of this intervention was on improving teacher reading instruction and increasing reading opportunities in the Foundation Phase, primarily in the HL but also in the FAL. The vocabulary study formed a small part of this larger intervention. The larger study is not discussed further in this article and serves merely to contextualise the vocabulary study.

In total, 118 Grade 3 learners from eight different classes from the four selected schools participated in an active vocabulary test. Their active vocabulary was assessed twice: at the start of the year (January 2016) and at the end of the year (November 2016). Their average age at the start of the year was 8.4 years and towards the end of the year it was 9.1 years. The genders were fairly evenly distributed across the classes, except for school EC1 where there were more girls than boys in the class (for details, see data collection procedures). The dominant HL spoken by the learners’ parents was Xhosa in the Eastern Cape and English in the Western Cape (see Table 1). Another 284 learners from the same eight Grade 3 classes participated in a receptive vocabulary test at the end of the year.

The tests were administered by two external researchers during school hours. Because of time and budgetary constraints, for the one-on-one active vocabulary assessment, a purposive stratified sample of 15 learners from each of the eight Grade 3 classes was used. The classroom teachers were asked to select participants based on their class performance: five top learners, five average learners and five weak learners, respectively. Demographic characteristics of the test participants are presented in Table 3.

Sibanda’s receptive word knowledge test of high-frequency words was administered in November to all 284 learners from the same eight Grade 3 classes. The pen-and-paper test was administered to the whole class, in the classroom, by two external researchers during regular class hours. Demographic characteristics of the test participants are presented in Table 4.

For both samples, the demographic items on gender, age and HL were averaged per school. The average score on the vocabulary test was calculated for all participants. The Kolmogorov–Smirnov test was used to determine normal distributions within the two separate groups (Western Cape [WC)]and Eastern Cape [EC]). Scores were normally distributed within each sample: the EC, D(60) = 0.09, p > 0.05; and the WC, D(58) = 0.09, p > 0.05. For these normally distributed data, independent t-tests (independent samples) and one-way analysis of variance (ANOVA) were used to compare subgroups, that is, schools within each province, followed by post hoc tests in case of significance. Dependent t-tests were used to compare pre- and post-test results within groups, and an effect size analysis on the pre-post results for each group was calculated, using Cohen’s d (e.g. Becker 2000). Pearson’s correlation coefficient was used to investigate the relationship between active and receptive word knowledge. Finally, a simple regression analysis was used to determine the role of initial active vocabulary knowledge on vocabulary development during the year.

All the schools were part of a larger three year project that worked with Foundation Phase teachers, the aim of which was to improve learner literacy levels through improved classroom literacy practices. Permission to work in these schools was obtained at national and provincial levels. The focus of this larger intervention was on improving teacher reading instruction and increasing reading opportunities, primarily in the HL and also in the FAL.

All the schools were informed about the vocabulary study at the beginning of the year and two schools from each province were asked to volunteer to participate. Permission to use the Sibanda’s high-frequency word tests was obtained in March 2016.

Results at province level are presented in Table 5. The scores have been converted to percentages and the standard deviation is given in brackets. On average, participants in the Western Cape scored significantly higher on the post-test (mean [M] = 68.8, standard error [SE] = 1.7) than on the pre-test (M = 60.3, SE = 1.8, t[57] = -12.01, p = 0.001). Cohen’s effect size value suggested a medium to large vocabulary learning effect (d = 0.65). Participants in the Eastern Cape also scored significantly higher on the post-test (M = 41.2, SE = 1.1) than on the pre-test (M = 31.4, SE = 1.2, t[59] = -13.23, p = 0.001), and the effect size shows a large vocabulary learning effect (d = 1.09). As to be expected, Grade 3 English HL learners knew almost double the number of words the English FAL learners did. What is of interest, however, is the fact that on the whole the learners from both English HL and English FAL schools increased by about 9% from January (pre-test) to November (post-test). At the end of the year, the active word knowledge gains of learners from English FAL schools in the Eastern Cape (M = 9.7, SE = 0.7) had increased slightly more than the active word knowledge score of learners from the Western Cape schools (M = 8.5, SE = 0.7), but this difference is not statistically significant (t [22] = -1.21, p > 0.05).

By the end of the year, this gap remained more or less the same, with the mean score of English FAL learners in the Eastern Cape at 41.2%, whereas the mean score of the English HL learners in the Western Cape was 68.8%. What is also interesting to note is that across all the schools, the vocabulary growth of children at the 25th percentile (i.e. those with lower scores) was lower, at around 5%, than the better performing learners at the 75th percentile, who showed a vocabulary growth of 12% – 13%.

The second research question focuses on the receptive vocabulary knowledge of high-frequency words of Grade 3 learners at the end of the Foundation Phase, using two of Sibanda’s receptive knowledge tests.

Results of the high-frequency word test are presented in Table 6, which includes the separate test results (word meaning and cloze test), as well as the combined results (average receptive vocabulary knowledge of high-frequency words). Here too, as expected, the data show that the English FAL schools in the Eastern Cape obtained lower scores on both receptive tests than the English HL schools in the Western Cape. The average scores of the learners in the Western Cape (M = 60.6, SE = 2.5) are twice as high as the scores of learners in the Eastern Cape (M = 27.4, SE = 1.8). This difference is statistically significant (t [282] = 10.94, p = 0.001). Not unexpectedly, the English HL learners knew twice as many words as their English FAL peers. Furthermore, the data show that all learners found the word meaning test challenging and obtained a lower score on the word meaning test than on the gap-filling cloze test.

The results illustrate two important things. Firstly, there are large differences in terms of distribution of scores between the HL and FAL learners. Secondly, the score differences within the language groups are equally large, suggesting that the vocabulary levels of the learners are highly unequal.

We turn now to age and gender differences in active and receptive word knowledge of Grade 3 learners. Age distribution at the province level is presented in Figure 1. With an average age of 9.4 years, the Grade 3 learners in the Eastern Cape were older than the learners in the Western Cape (8.9 years).

Results of the receptive vocabulary knowledge average (word meaning and cloze tests) across the age groups are shown in Table 7.

There were significant differences between the age groups in the Western Cape (F [3,138] = 3.50, p < 0.05, ω = 0.22). Post hoc tests indicated that both 8- and 9-year-olds had a significantly higher score on receptive knowledge of high-frequency words than 10-year-olds. No other significant differences were found. In the Eastern Cape, there were also significant differences between the age groups (F [4,136] = 2.54, p < 0.05, ω = 0.20). Post hoc tests indicated that the 9-year-olds scored significantly higher on receptive vocabulary knowledge than their 10-year-old peers.

There is typically an increase in vocabulary knowledge of high-frequency words with age during the school years; exceptions do of course occur, and children with learning problems often present with lower vocabulary knowledge. The majority of Grade 3 children in this cohort were 9 years old. The older learners are likely to be learners who repeated a grade in Foundation Phase. The results seem to confirm this: the 10-, 11- and 12-year olds in particular seemed to display much lower vocabulary knowledge, suggesting that they were weak learners.

Table 8 reflects gender differences in active and receptive vocabulary knowledge across English HL and English FAL Grade 3 learners.

Although female FAL learners in the Eastern Cape scored slightly higher on active vocabulary (M = 41.8, SE = 7.1) than their male peers (M = 40.4, SE = 10.3), this difference was not significant (t[58] = 0.62, p > 0.05). The difference in active vocabulary knowledge between HL boys (M = 68.9, SE = 12.8) and girls (M = 68.7, SE = 12.5) in the Western Cape was also not significant (t[56] = -0.04, p > 0.05).

Girls in the Eastern Cape scored slightly higher in receptive word knowledge (M = 28.6, SE = 21.2) than their male peers (M = 25.8, SE = 21.1), but this was not significant (t[139] = 0.77, p > 0.05). The difference in receptive vocabulary knowledge between boys (M = 57.4, SE = 30.5) and girls (M = 64.0, SE = 28.0) in the Western Cape was also not significant (t[141] = 1.35, p > 0.05).

The fourth research question investigated the relationship between the learners’ active and receptive word knowledge. Pearson’s correlation coefficient was calculated for active and receptive vocabulary knowledge, and the results are presented in Table 9. No significant correlations were found between both active vocabulary test results (January and November) and the receptive vocabulary results. The correlation between the active vocabulary tests of January and November was highly significant.

The fact that no correlation was found between learners’ active and receptive vocabulary knowledge might be explained by the different levels of vocabulary that are measured by both tests. In the Sibanda tests, 97.1% of the words are from the 1000 word family list (i.e. the most common words in English). In contrast, the Woodcock-Muñoz test only has 39.5% words at the 1000 word family level and includes far more difficult vocabulary (up to the 14 000 word family list). Details of the word family levels per test are presented in Table 10.

The final research question looked at the role of initial vocabulary knowledge on vocabulary development during the year. To what extent did the learners’ active vocabulary knowledge at the beginning of the year predict their vocabulary growth during the year, as reflected in their final active vocabulary knowledge scores and their scores on the high-frequency words?

In both English HL and English FAL, learners who knew more words at the beginning of the year seemed to acquire more words by the end of the year. In order to investigate the role of initial active vocabulary in vocabulary development during the year, a simple linear regression was calculated to predict post-test active vocabulary knowledge based on pre-test active vocabulary knowledge. A significant regression equation was found in the Western Cape: F [1, 56] = 288.647, p < 0.001, with an R² of 0.838. Similarly, in the Eastern Cape, a significant regression equation was found: F (1, 58) = 100.697, p < 0.001, with an R² of 0.797.

Another simple linear regression was calculated to predict receptive vocabulary knowledge at the end of the year based on active vocabulary knowledge at the beginning of the year. No significant regression equation was found either in the Western Cape (F [1, 56] = 1.157, p > 0.05) or in the Eastern Cape (F [1, 58] = 0.476, p > 0.05).