Children who are not faring well in acquiring their language(s) or those who present with language disorders are at risk for poor literacy acquisition, which can negatively affect educational progress (Lundberg, 2006; see Duff & Tomblin, 2018). Language assessment enables the identification of these children but requires reliable, valid language assessment tools if underdiagnosis or overdiagnosis of language problems is to be avoided (Buac & Jarzynski, 2022). Furthermore, accurate language assessment with linguistically and culturally appropriate assessment tools forms the cornerstone of effective language intervention (Pascoe & Norman, 2011).

This study considers child language assessment tool development for the official languages of South Africa (SA). Tool development during apartheid reflected the general unequal treatment of languages, with more work performed for the then only two official languages (Afrikaans and English) than for African languages (isiNdebele, isiXhosa, isiZulu, Sesotho, Sesotho sa Leboa, Setswana, Siswati, Tshivenda, Xitsonga) (see e.g. Penn, 1998). Since 1994, the government has started to redress past inequities, among others, introducing policies and initiatives to elevate the status of African languages in education, the media and public life (Sokani, 2024). However, as will be shown in this paper, research on language acquisition and the development of tools with which to measure child language skills and diagnose language problems mostly remains neglected for children of all ages, particularly in African languages. Given the linguistic demographics of the country’s population – 80.7% use a language other than Afrikaans or English as home language (Statistics South Africa, 2023) – this imbalance between the country’s languages is problematic.

There have been five reviews of language assessment tools for use with SA children: Penn’s (1998) overview and Mphahlele’s (2006) inventory were written from a speech-language therapy (SLT) perspective; Pascoe and Norman (2011) gave an overview of contextually relevant resources for SA SLTs and audiologists; Pascoe et al. (2020) did a scoping review of health resources available, and Pascoe and colleagues (2011, 2020) and Pascoe and Singh (2023) included resources for use with children and adults. Except in the Pascoe and Singh (2023) publication, the scientific readiness-for-use of the available tools was not critically evaluated.

In 1998, Penn concluded that ‘resources are lacking’ (1998, pp. 265–266). Pascoe et al. (2020), 22 years later, reported 57 resources for SA languages, a seemingly substantial improvement since 1998. However, these 57 resources are not limited to child language nor to assessment, and Pascoe and Singh (2023) listed only 16 published tools for assessing SA children’s speech or language (21 if counting all language versions). For these available tools to be of diagnostic and research value, they need to have been adapted culturally and linguistically, standardised locally and have norms that are still valid (American Psychological Association, 2017), otherwise they do not allow for benchmarking of a child’s language performance and are therefore not necessarily ready for wider use.

This study critically reviews progress in tool development for measuring the language of children up to 12 years, for research and diagnostic purposes, in SA’s official languages. Considered for review were published tools and those in development, and for each, it was indicated whether (1) it was purpose-developed, translated and/or adapted; (2) there were any pre-pilot activities to increase linguistic and cultural appropriateness; (3) tools that were a pilot study was conducted; (4) the tool had been standardised; (5) reliability and validity were tested and reported and (6) the tool had been normed. The following were excluded: tools (1) assessing speech sound and phonological development and (pre)literacy skills; (2) lacking evidence of principled linguistic and cultural adaptation – for instance, tools that were merely translated or were adapted informally and/or live (e.g., Feris, 2017); (3) that were cited in other publications but were not traceable for verification purposes; (4) devised as intervention material that could incidentally be used for informal assessment and/or (5) that do not assess language exclusively but include a language subtest when assessing child development more broadly.

The development of child language assessment tools commenced during apartheid, focusing on Afrikaans and English, the then well-resourced, official languages. Original tools developed for Afrikaans by 1994 include Afrikaanse Reseptiewe Woordeskattoets (ARW, ‘Afrikaans Receptive Vocabulary Test’; Buitendag, 1994) and Afrikaanse Semantiese Taalevalueringsmedium (AST; ‘Afrikaans Semantic Language Evaluation Medium’; Pretorius, 1989). An Afrikaans version of two tools that were originally developed elsewhere for child speakers of English were created during apartheid: Illinois Test of Psycholinguistic Abilities (ITPA; first edition: Kirk et al., 1967) was translated into and adapted for Afrikaans (Lotter, 1971); and Peabody Picture Vocabulary Test (PPVT; first edition: Dunn, 1959) was translated into Afrikaans (Alant & Beukes, 1986). Pre-democracy, SA English-speaking children were mostly assessed with tools developed elsewhere in minority-world contexts–and this practice is ongoing (cf. Van Dulm & Southwood, 2013)–or with informally adapted versions thereof (such adaptation typically entailing replacing words with their SA English equivalents), without validating or renorming the tools for SA children. Such tools are excluded from this review based mainly on exclusion criterion (2) above. When considering exclusion criteria (1) to (5) above, no tools developed during apartheid for African languages could be included in this review.

Below, an overview of tools developed for SA languages during and after apartheid is presented. A summary of these tools and their selected psychometric properties appears in Table 1 (see also Appendix 1), showing the methodological steps completed during tool development.

For the SA language versions of the following 10 tools that were included in the list in Table 1, either (1) no reliability, validity or norming study was conducted; (2) reliability and/or validity was studied but the obtained statistical values were not reported, and therefore reliability and validity could not be verified or (3) the norming study is dated:

For vocabulary or semantics: The translation of the PPVT (first edition: Dunn, 1959) for Afrikaans overseen by Alant and Beukes (1986); translation and adaptation of the PPVT for isiZulu (Naidoo, 1994); and the translation and adaptation of PPVT-Revised for Sesotho sa Leboa (Pakendorf & Alant, 1997); AST (Pretorius, 1989) and ARW (Buitendag, 1994). The last two were developed more than 30 years ago. Given that the lexico-semantics of languages change over time, reconsideration of the linguistic and visual content of these tools is indicated; therefore, they are not necessarily currently suitable for diagnostic purposes. Language Impairment Testing in Multilingual Settings (LITMUS)-Crosslinguistic Lexical Tasks (CLT; cf. Haman et al., 2015) was developed for Afrikaans, isiXhosa, and SA English (cf. Potgieter & Southwood, 2016) using the same protocol for each language version, to increase comparability across languages in terms of item difficulty. For each language version, three mother-tongue speakers provided general information on approximately 1000 words and on the appropriateness of the picture accompanying each word. From these, 165 clearly depictable words were selected, and three mother-tongue speakers provided information on the phonetic, morphological, semantic and etymological characteristics of each. These were used to calculate a complexity score. The latter was considered together with the reported age of acquisition (of 20 adults per language) to calculate each word’s complexity index. Items were selected based on their complexity index. CLTs were piloted with 26 Afrikaans-, 34 English- and 10 isiXhosa-speaking children. Amendments were made to the tool, and larger-scale data collection for validation purposes is underway. Each 120-item tool assesses the comprehension and production of nouns and verbs based on pictures and is meant for use with 3- to 5-year-olds.

For morphosyntax: Palmer’s (2022) sentence repetition test, the only tool in SA Sign Language considered for this review, consists of 20 sentences (simple, moderate or complex, based on their estimated grammatical complexity). It was piloted with 40 Deaf children aged 7 to 9 years. Scores increased with an increase in child age and an increase in length of exposure to SA Sign Language.

For narration: LITMUS-Multilingual Assessment Instrument for Narratives (LITMUS-MAIN; cf. Gagarina et al., 2012) has an Afrikaans version by Klop and Visser (cf. Gagarina et al., 2019a); isiXhosa version (Klop & colleagues, 2019; cf. Gagarina et al., 2019b), Tshivenda version (Mabaso & Tshikonelo, 2019) and isiZulu version (Ndlovu et al. cited in Gagarina et al., 2019c; Ndlovu & Klop, 2023 – piloted with 28 children aged 9 to 11 years). LITMUS-MAIN allows for assessment of multilingual children’s understanding and production of narratives in two or more languages using picture-based stories presented through different elicitation modes (model story, telling, retelling). Stimulus material is controlled to be (1) equivalent in terms of cultural appropriateness and cognitive and linguistic complexity (see Klop & Visser, 2020; Ndlovu & Klop, 2023), and (2) parallel in macro- and microstructure within and across language versions (see Gagarina et al., 2012). For all language versions of LITMUS-MAIN, aspects of content validity were addressed by adapting the picture stimuli for cross-cultural appropriateness.

For tools assessing more than one language domain: ITPA, originally designed to measure spoken and written language abilities of English-speaking children (Hammil et al., 2001), was preliminary translated into and adapted for Afrikaans (Lotter, 1971); South African Language Assessments (SALA; Bortz, 1997) for Setswana, Sesotho, Tshivenda and Xitsonga, viz. the translated versions of Zulu Expressive and Receptive Language Assessment test (ZERLA; discussed below) and the Afrikaans and SA English Diagnostic Evaluation of Language Variation (DELV; Seymour et al., 2015) of which the development is ongoing (cf. Van Dulm & Southwood, 2008). DELV is not yet ready for clinical use but has stimulated some research (cf., e.g., Southwood 2013; Southwood & White, 2021a, 2021b). It is a picture-based tool assessing the comprehension and production of various aspects of syntax, pragmatics, semantics and phonology in children aged 4 to 9 years. The Afrikaans and SA versions were linguistically and culturally adapted in conjunction with SLTs who spoke non-mainstream varieties of these languages. The Afrikaans version was piloted among 48 children aged 5 to 9 years, all speakers of the so-called Kaaps variety of Afrikaans, and their DELV scores correlated well with their ARW scores (Southwood & Van Dulm, 2009). Norming is underway and nearing completion. However, given that norming commenced in 2008, some data points might need replacement before norm calculation (F. Southwood, personal communication).

Table 2 shows the sample sizes and statistical values of the tools listed in Table 1 for which the statistical analyses of the reliability, validity and/or norming studies were reported, barring the ARW and AST for reasons mentioned above. The content of Table 2 is discussed further.

Since the onset of democracy, 12 child language tools have been added to the small number of pre-1994 tools in SA’s now-official languages (see Table 1), covering the age range from infancy to 12 years. Of these, six are translations and/or adaptations of tools developed elsewhere (CDI, DELV, all but the Afrikaans version of the LITMUS-MAIN, PPVT, Receptive and Expressive One-Word Picture Vocabulary Test, Test of Ability to Explain) and six were originally developed by SA researchers (IRVT, LITMUS-CLT, Productive Vocabulary Test, SALA, Sentence repetition test, ZERLA). Eleven of these post-1994 tools are available in at least one of SA’s non-West-Germanic official languages, with DELV being available only in Afrikaans and English. When each language version is counted separately and when LDS (a validated, pre-democracy tool) is added, 39 language assessment tools, in varying states of psychometric readiness, are available for use with SA children (at least one per official language).

However, this calculation presents an overly positive picture. In total, when excluding tools that need norming or renorming, there are four validated tools for use with SA children: IRVT, LDS, Productive Vocabulary Test and Test of Ability to Explain. These collectively have six language versions, with LDS being available in English only. The language for which most validated tools are available is isiZulu (3), followed by English (2) and Siswati (1). Based on the above, some progress has been made in language assessment tool development post-1994; however, there are no currently normed tools available. (Norming of two tools–DELV, and the Afrikaans and isiXhosa versions of CDI–is about to commence; H. Brookes & F. Southwood, personal communication). Scientifically sound child language assessment tools are unevenly spread across languages. This, coupled with insufficient support for research in indigenous languages, is disconcerting: tool development needs to be theory-driven and evidence-based if tools are to be valid (APA, 2017), which means that they need to be grounded in the findings of robust research into the acquisition of SA’s languages by child speakers of a wide range of ages. Such research is emerging slowly. Furthermore, children growing up multilingually, which includes many children in SA, need to be assessed in each of their languages (see, e.g., Buac & Jarzynski, 2022), which requires appropriate assessment tools in those languages. This adds gravity to the general unavailability of SA child language assessment tools: not only is there a need for tools to be developed to completion, but multiple language versions of these tools are also needed. In addition, the fact that SA languages have different varieties requires tool developers to specify the geographic area(s) from which their participants were sampled in order to allow the end-users of a tool to evaluate the applicability of the tool for the child(ren) to whom they wish to administer it.

To address the lack of valid and standardised measurement tools, the research methodologies used during tool development may require critical evaluation, as more community co-design may be needed, which could be well served by adopting participatory action research. Furthermore, larger teams of experts in child language development from different disciplines, as well as well-trained multilingual research teams, are needed to conduct large-sample studies in SA languages to validate and standardise tools. Data from large studies will also contribute to improving existing theories in the field of child language acquisition that have been built on findings predominantly generated in minority-world contexts. Scientifically sound child language research embedded in local child language socialisation practices will allow for accurate child language assessment. This is more than a diagnosis- or research-related ideal; it is a moral obligation. Assessment tools that do not consider linguistic and cultural variation and do not attempt to avoid linguistic and cultural bias cannot deliver accurate assessment results (cf. Pascoe & Norman, 2011), and without accurate assessment, child language intervention flounders. Conversely, scientifically sound, contextually appropriate early childhood interventions can lead to SA children gaining a strong language foundation on which to build literacy skills.