As throughout the world, multilingualism is typical in South Africa owing to a culturally and linguistically diverse population, as well as the mobility of the population and the country’s historical and political background. Many people speak two or more languages in their day-to-day life (Williams & Stroud, 2013). Although the country’s constitution states that all 11 official languages should be treated with equal esteem, and that children have the right to receive education in any of the official languages, many parents want their children to be educated in English rather than in their home language owing to the perceived higher status of English (Webb, Lafon & Pare, 2010). This article focuses on difficulties in the bilingual speech sound acquisition of isiXhosa and English.

IsiXhosa is one of the 11 official languages of South Africa. It is the second most widely spoken language in South Africa, with 16% of the population speaking it as a home language (Statistics South Africa, 2012). IsiXhosa is a Bantu language from the Nguni group along with isiZulu, SiSwati and isiNdebele. It is one of the main languages spoken in the Western Cape region of the country, along with Afrikaans and South African English.

Previous research into Bantu languages has considered various aspects of speech and language acquisition and disorders. For example, Demuth (2003) summarises research considering the acquisition of various Bantu languages, looking at the acquisition of noun classes, verbal morphology and syntactic structures and phonology. Suzman and Tshabalala (2000) considered the nature of language impairment in two isiZulu-speaking children, analysing their phonology, morphology, syntax and pragmatics. Demuth and Suzman (1997) also describe the language impairments displayed by an isiZulu-speaking child. Their findings suggested that the child’s language was qualitatively different from a typically developing isiZulu-speaking child’s language, and not just delayed.

Considering isiXhosa, the phonological development in isiXhosa-speaking children has been investigated (Gxilishe, 2004; Lewis & Roux, 1996; Maphalala, Pascoe & Smouse, 2014; Mowrer & Burger, 1991; Pascoe et al., 2016; Tuomi, Gxilishe & Matomela, 2001). Maphalala et al. (2014) conducted a cross-sectional study, investigating the speech production of 24 isiXhosa-speaking children between the ages of 3 and 6 years. An isiXhosa speech assessment, Masincokoleni isiXhosa Speech Assessment (Maphalala, Pascoe & Smouse, 2012) was developed to more accurately describe the speech skills of the children in the study. Results suggested that most consonants are acquired by the age of 3 years, with aspirated plosives, affricates, fricatives and clicks being amongst the later developing phonemes. Affricates were also identified as later developing by Tuomi et al. (2001) and Mowrer and Burger (1991). Mowrer and Burger (1991) also suggested that isiXhosa-speaking children acquire most consonants earlier than their English-speaking peers. This was similarly noted in two case studies conducted by Pascoe et al. (2016), who found that the two children in their study had acquired all of their vowels and many of their consonants by the ages of 2 years and 5 months, and 2 years and 8 months. Clicks, one of the most well-studied aspects of isiXhosa phonology, are acquired between the ages of 1 and 3 years (Gxilishe, 2004; Tuomi et al., 2001). Research has also described the typical simplifications used by young isiXhosa-speaking children as they acquire clicks (Lewis & Roux, 1996). IsiXhosa-speaking children make use of phonological processes for other consonants, most commonly gliding of liquids, stopping, depalatalisation, deaspiration and denasalisation (Maphalala et al., 2014).

Pascoe et al. (2015) considered the acquisition of English phonology by 3-year olds in Cape Town. Their sample included a small number (n = 25) of isiXhosa–English bilingual speakers and provides some data regarding the English phonology of isiXhosa–English bilingual 3-year olds. Findings suggested that the isiXhosa bilingual children in their sample had the most complete consonant inventories in English when compared with monolingual English speakers, bilingual English–Afrikaans speakers and trilingual English–Afrikaans–isiXhosa speakers in the study. However, the isiXhosa–English children also had the lowest mean percentage of correct vowels. The researchers questioned the validity of this score, suggesting it may have been caused by typical vowel substitutions used by isiXhosa–English bilingual speakers, given the complexity of the English vowel system in comparison to the five vowels of isiXhosa. The study also determined that the phonological processes of cluster reduction and stopping were more prevalent in the English speech of the isiXhosa–English children than in the other groups, as well as the processes of backing and devoicing (Pascoe et al., 2015), often considered non-developmental in monolingual English speech (Dodd, Holm, Hua & Crosbie, 2003). Backing and devoicing may be typical in the development of English by isiXhosa–English bilingual children. None of the isiXhosa–English bilingual children were identified as having a speech difficulty, although it is important to note that the sample of isiXhosa–English children was small (Pascoe et al., 2015).

It has been hypothesised that if a bilingual child presents with a speech sound disorder, the disorder will fall in the same category in both languages, suggesting that although the child has two phonological systems, a single underlying deficit affects both (Holm & Dodd, 1999b; Holm, Dodd, Stow & Pert, 1999). This seems to align with interactional dual systems theory (Paradis, 2001), which suggests a child’s two phonological systems interact in various ways as the child’s phonology develops, resulting in cross-linguistic transfer. This has been documented across a range of language pairs by authors, including Holm, Dodd, Stow and Pert (1997), Holm and Dodd (1999a, 1999b, 2001) and Ray (2002). Where a child presents with speech errors in only one language, this may be indicative of cross-linguistic transfer rather than a speech sound disorder (McLeod, Verdon & International Expert Panel on Multilingual Children’s Speech, 2017). For example, a child who speaks both English and French may not be able to produce the fricative [θ] in English and substitutes it with the plosive [t], as it is not present in French. If this use of the phonological process of stopping is only present in English, it is not necessarily caused by a speech sound disorder, but rather by cross-linguistic transfer from French to English. This highlights the importance of a speech language therapist (SLT) being able to identify whether a bilingual child presents with a speech sound disorder or difference (McLeod et al., 2017). A child acquiring two or more languages may present with differences in their speech that are not true errors but rather are caused by the typical acquisition of two languages or dialects (McLeod, 2012). This, too, must be differentiated from a child who presents with a true phonological disorder or delay (McLeod et al., 2017). In order to do this, knowledge of the typical development of bilingual children acquiring various language pairs is essential. For many language pairs, this information is lacking.

Although there have been a large number of studies considering bilingual intervention for language disorders, there are fewer that describe intervention for a bilingual child presenting with a speech sound disorder. A literature search identified seven articles that present an intervention study of bilingual or multilingual children with speech sound disorders that clearly describes the participants’ speech sound disorder, the intervention provided and the changes in the child’s speech following intervention (see Table 1). The studies presented vary in approach to intervention, as well as language of intervention, and investigated a variety of language pairs. The majority of the studies provided intervention only in English (Holm & Dodd, 1999a, 2001; Holm et al., 1997; Ray, 2002), while one provided intervention only in Arabic (Mamdouh, 2008) – in all of these cases, intervention was provided in the language that was dominant within the community. In contrast, two of the case studies provided intervention in both of the children’s languages, Spanish and English (Gildersleeve-Neumann & Goldstein, 2015) and Portuguese and English (Ramos & Mead, 2014). No intervention studies were found that include a language pair that is common amongst bilingual children in South Africa or a Bantu language.

The procedures and approaches employed in the studies varied greatly, resulting in difficulty drawing conclusions. However, it is interesting to note the findings of Gildersleeve-Neumann and Goldstein (2015) and Ramos and Mead (2014) who investigated the effect of providing intervention in both languages. The use of both the child’s languages in intervention is more in line with the recommendations provided by various individuals and organisations (e.g. Gildersleeve-Neumann & Goldstein, 2012; Goldstein & Fabiano, 2007; International Expert Panel on Multilingual Children’s Speech, 2012). Results indicated an increase in overall accuracy in both languages (Gildersleeve-Neumann & Goldstein, 2015; Ramos & Mead, 2014). The results recorded by Ramos and Mead (2014) are worth mentioning, as they compared the results of a period of intervention in one language (English) to those following a period of intervention in two languages (English and Portuguese). They concluded that for their participant, even though some improvement of specific phonological processes was noted in Portuguese when the processes were targeted in English, there was a much greater improvement when she received bilingual intervention (1 h of English and 1 h of Portuguese per week). In addition, although providing English intervention improved her English speech production, bilingual intervention had a greater effect on improving her English production. However, they did note that some phonological processes (e.g. fronting of palatal fricatives) had to be specifically targeted in Portuguese, her weaker language, for the child to show indications of eliminating those specific processes from her Portuguese speech (Ramos & Mead, 2014).

This study aimed to describe changes occurring in an isiXhosa–English bilingual child’s speech following intervention targeted at her speech sound difficulties. We intended to add to the limited research considering intervention for speech sound disorders with bilingual children, as well as considering the challenges of assessing the speech of a bilingual child when little is known about the typical bilingual speech development of the language pair spoken by the child.

The objective of this study was to analyse and describe the isiXhosa and English speech of a bilingual isiXhosa–English-speaking child before and after providing intervention based on a tailor-made, theoretically motivated intervention plan. It followed an exploratory descriptive design in order to detail the changes that occurred in the child’s speech. A single subject pre-test and post-test design was used: this allowed the participant to be described as an individual case and act as her own control (Vance & Clegg, 2012), and included assessing the child’s speech before and after the intervention.

The study received ethical clearance from the University of Cape Town Faculty of Health Science Human Research Ethics Committee, HREC REF: 448/2015.

This study considered the changes in a bilingual child’s speech in both languages when provided with intervention targeting her speech sound disorder. The intervention approach of minimal pairs was chosen to target Gcobisa’s speech sound disorder (Baker, 2010), resulting in a change in Gcobisa’s phonological system, as she was able to accurately produce the phoneme [l] in both trained and untrained words and phrases in isiXhosa and in words in English.

Kohnert (2010) highlights three issues that clinicians need to take into account when considering bilingual children regardless of whether they have acquired their languages simultaneously or sequentially: (1) children may present with an uneven distribution of skills in their two languages, so their language skills may be distributed across both languages rather than duplicated from their stronger language to their weaker language; (2) some form of cross-linguistic interaction will take place; and (3) owing to the complex interaction between personal factors evident in all children (e.g. socioeconomic situation and general exposure to language at home) as well as factors specific to bilingual children (e.g. exposure to each language, age of acquisition of each language and opportunity to use each language), bilingual children present as a very heterogeneous group, making it difficult to compare a bilingual child’s development to norms, even when those norms are based on a similar group of bilingual children. This was evident in the case under discussion, and the lack of information regarding typical development in the isiXhosa–English population resulted in challenges in interpreting the data in the current study. Interpretation was based on monolingual norms, but this is problematic, as bilingual speech development is known to differ from the monolingual development of the languages (Holm & Dodd, 2006). This results in bilingual children often being under-referred or over-referred for intervention (Hambly, Wren, McLeod & Roulstone, 2013), as demonstrated in the current study. Although Gcobisa’s teachers were eager to refer her to the researcher and reported concerns regarding intelligibility, detailed analyses of phonological skills were difficult, as little is known about isiXhosa–English bilingual acquisition. In addition, her mother did not share the teacher’s concerns. Results were compared to monolingual norms, and the effect on Gcobisa’s participation in the classroom guided the decision to provide intervention. However, if one considers that Gcobisa is acquiring both English and isiXhosa, and this acquisition is different to monolingual development of each language, her speech development may be typical for a bilingual child, particularly considering the heterogeneity of the bilingual population.

Preliminary research into the typical development of South African English has included a small sample of isiXhosa–English bilingual children (n = 25) and provided some information regarding their English speech development (Pascoe et al., 2015). One of the findings in that study suggested that backing, a phonological process considered non-developmental in English, was prevalent in the English speech of isiXhosa–English bilingual participants. Gcobisa showed evidence of backing in isiXhosa. This adds to the data collected by Pascoe et al. (2015), adding to the evidence that backing may be a typical process in bilingual isiXhosa–English children. However, unlike Pascoe et al. (2015), this study also considered isiXhosa speech. Backing was evident in Gcobisa’s isiXhosa speech but was not evident in the monolingual isiXhosa sample under consideration by Maphalala et al. (2014). This process may be considered typical when evident in one or both languages of isiXhosa–English bilingual children, highlighting a difference between monolingual and bilingual speech acquisition of the two languages. However, a larger sample of isiXhosa–English children is required to determine whether this is typical of the larger population, as previous research has focused on fairly small samples.

Another process identified as being more prevalent in isiXhosa–English bilinguals than in monolinguals is cluster reduction (Pascoe et al., 2015). This may be the result of the effect of one language on the other: isiXhosa makes use of very few clusters (Maphalala et al., 2014). It is therefore not surprising that children who speak isiXhosa as a home language find the production of clusters difficult, often reducing them. This was noted in the English speech of Gcobisa. Cluster reduction is expected to be eliminated from the speech of monolingual English children by the age of 3 years and 11 months (Dodd et al., 2003) and yet this was the most prevalent process in the speech of Gcobisa (4 years and 6 months at the end of the study). This suggests that it may be typical for isiXhosa–English children to continue to use cluster reduction beyond the age of 4 years. However, again, this would need to be explored further with a larger sample of isiXhosa–English bilingual children.

South Africa’s multilingual environment creates challenges for SLTs. The current demographic of qualified SLTs is poorly representative of the South African population, as the majority of SLTs speak English and/or Afrikaans (Pascoe & Norman, 2011). Although this is gradually changing to reflect a more diverse demographic, there is often a language mismatch between SLT and client (Pascoe & Norman, 2011). Globally, this is also a common problem as discussed in a tutorial article by McLeod et al. (2017) entitled ‘Speech assessment for multilingual children who do not speak the same language(s) as the speech language pathologist’. Around the world, SLTs face further challenges in providing appropriate intervention for children who are multilingual, as there is limited research on the appropriate approach for intervention for this population (Goldstein & Fabiano, 2007).

In the recruitment process, a number of bilingual children were referred to the researcher and assessed. However, after analysing their speech using monolingual norms, very few of these children presented speech sound disorders. In South Africa, a preliminary study investigating the speech of 150 3-year-old children acquiring English identified that 6.66% of the children presented with speech sound disorders (Pascoe et al., 2015). Of the 6.66%, none were isiXhosa–English bilingual children. Further research into the speech development of bilingual children, particularly those acquiring isiXhosa and English, would be useful in order to aid SLTs in being able to differentiate between a speech disorder and speech differences.

Many previous studies considering intervention for bilingual children provided intervention in only one of the multilingual child’s languages (Holm & Dodd, 1999a, 2001; Holm et al., 1997; Mamdouh, 2008; Ray, 2002). Ramos and Mead (2014), however, provided intervention in both languages of a bilingual child. From their results, they concluded that intervention in both languages was more effective than intervention provided in only one of the child’s languages, even if generalisation takes place (Ramos & Mead, 2014). Considering the case of Gcobisa, although the targets chosen were in isiXhosa, the main language used within the sessions was English, with some simple isiXhosa instructions [e.g. hlala pantsi (sit down); mamela (listen); ewe (yes)]. This was a situation not well-documented in the literature, where the targets were in isiXhosa while the instructions and other interactions with the researcher took place in English. This may have had an impact on the generalisation from the treated isiXhosa to the untreated English.

Holm et al. (1997), Holm and Dodd (2001) and Mamdouh (2008) reported situations where generalisation did not occur from the treated to the untreated language. In these cases, the targeted error or phoneme appeared to be specific to the targeted language or uncommon in the untreated language. This would suggest that if a targeted error pattern or phoneme is common to both languages, intervention in one language may result in generalisation of results to the untreated language, as evidenced in Gcobisa’s results. Gcobisa made use of gliding of [l], producing the liquid as [j] in both English and isiXhosa. Intervention that used targets from only one language (isiXhosa) resulted in generalisation of results to the untreated language (English), as Gcobisa showed evidence of eliminating this phonological process from both her isiXhosa and her English speech.

This case study considered the changes in the speech of a bilingual child following intervention. Intervention sessions were conducted predominantly in English, while targets and stimuli chosen were in isiXhosa. These unique conditions appear to have resulted in change to the phonology of both languages. The language mismatch between the researcher and the child mirrors the situation often experienced by South African SLTs, who are not always able to speak all of the languages spoken by their clients. Although providing intervention in both languages is considered ideal (Gildersleeve-Neumann & Goldstein, 2012; Goldstein & Fabiano, 2007; International Expert Panel on Multilingual Children’s Speech, 2012), in situations where this is not possible and translators are not readily available, the use of targets in one language and intervention in the other language may result in change to both languages, although further research is necessary with a larger sample of children to generalise this to the wider population. Finally, this article also added to the small body of research considering isiXhosa–English typical speech acquisition. The child in the single case study presented with backing and cluster reduction that persisted beyond the typical age of elimination, both processes also noted by Pascoe et al. (2015) in the English speech of the isiXhosa–English bilingual children included in their study. This adds to the evidence that these processes may be typical in the acquisition of isiXhosa–English speech.