Perceptions of isiZulu-speaking pre-service teachers’ classroom English proficiency

Despite less than 10% of South Africans claiming English as their home language, it has become the de facto language of instruction. Yet we cannot assume that teachers have sufficient command of this language when using it for instructional purposes. As a sub-study, in this article we report on the oral proficiency of isiZulu-speaking pre-service teachers who use English when expounding content in rural schools. The conceptual framework draws primarily on research relating to instructional communication and Classroom English. For this mixed methods case study, using questionnaire data from 52 pre-service teachers and 18 tutors, we sought to establish the perceptions that respondents had of students’ oral proficiency while teaching in situ. Responses were statistically analysed using computing software. Unedited audio recordings of lessons presented in rural KwaZulu-Natal schools during pre-service teachers’ work-integrated learning stint provided oral data from which to gauge proficiency using a self-designed rubric. Findings correspond with those of previous studies, pointing to pre-service teachers’ oral proficiency being less than ideal for effectively facilitating learning. However, what is considered adequate proficiency and what is ideal is yet to be agreed upon. We recommend that interventions which address the development of oral proficiency required for classroom use be considered. Our pilot rubric may serve as a useful data-gathering tool in future research.


Introduction
Internationally the number of non-native English speakers who use English as a medium of instruction, has surpassed that of native English-speaking teachers (Richards, JC 2017;Tsang, 2017). These statistics hold true for South Africa as well. The racial and linguistic composition of especially urban schools is so diverse that it has commonly been accepted that English will serve as a medium of instruction right from the first school day (Alexander, 1997; Department of Basic Education [DBE], Republic of South Africa [RSA], 2010;Evans, 2005). While learners in rural communities i are largely taught in a monolingual context for the first 3 years of formal schooling (as permitted and encouraged by policy), there is an abrupt switch to English as the language of learning and teaching (LoLT) in Grade 4.
While only 9.6% of South Africans claim English as their first language (Statistics South Africa, 2011), 79% of learners are taught through the medium of English (DBE, RSA, 2010). This may not seem problematic, yet those teaching them often lack the proficiency or confidence to do so.
Recent studies (Evans & Cleghorn, 2012;Evans & Nthulana, 2018, Hugo & Nieman, 2010Reyneke, 2014;Taylor & Mayet, 2015;Van der Walt & Ruiters, 2011) point towards the majority of prospective teachers entering the education system annually not being mother-tongue speakers of English. It is, nevertheless, assumed that since they speak English socially with reasonable fluency, they can teach content effectively using English. Low levels of English proficiency among teachers and learners have been identified as one of several academic challenges that manifest in South African schools (National Education Evaluation & Development Unit, 2013), as in many other countries grappling with (im)migrant/refugee learners, or post-colonial policy debates (Freeman, Katz, Gomez & Burns, 2015;Gan, 2012;Low, Chong & Ellis, 2014;Moon, 2014;Pasternak & Bailey, 2004, Tsang, 2017. Complex cognitive processes related to learning take place as teachers and learners interact with each other and grapple with mastering content material during each lesson. In such an educational context, learning could ultimately be affected negatively by instructional dissonance (Evans, 2005) and ineffective classroom communication.
In this article we report on the outcome of a sub-study which sought to establish the perceptions that 52 isiZulu-speaking final year Bachelor of Education (BEd) students specialising in the Intermediate Phase had of their English proficiency when expounding content. These perceptions were augmented by the views of 18 tutors while researcher perception was captured by assessing the authentic speech recorded while teaching.

Literature Review
Complex cognitive processes take place as teachers and learners interact with each other and grapple with mastering content material during each lesson. Being an effective teacher thus requires a high level of language proficiency to "provide meaningful explanations, rich language input for learners and respond spontaneously and knowledgeably to their learners' questions" (Richards, H, Conway, Roskvist & Harvey, 2013:244). Willis (1985:5) defines Classroom English as "the specialised and idiomatic forms of the English used when teaching that enables teachers to use English effectively and imaginatively as a means of instruction or as a means of organising a class or even a means of communicating with their learners as individuals about their life outside the classroom." Freeman et al. (2015:129) also focus on instructional language as "a specialized subset of language skills required to prepare and communicate lesson content and assess learners", which includes managing the classroom, and giving feedback. Being proficient in the language of instruction includes the ability to determine learners' prior knowledge, give clear, executable instructions (Chadwick, 2012), provide meaningful explanations, while responding spontaneously and knowledgeably to learners' questions (Richards, H et al., 2013). Without such communicative skills and linguistic agility in an educational context, learning could ultimately be affected negatively by instructional dissonance (Evans, 2015), hence the importance of prospective teachers being proficient in the LoLT. Tsang (2017) points out that there is still deliberation about what constitutes proficiency and how/when to declare a teacher orally proficient.
Policy requirements documented in the Government Gazette of 12 February 2015 (Department of Higher Education and Training [DHET], RSA) require of South African beginner teachers "to know how to communicate effectively in general, as well as in relation to their subject(s) in order to mediate learning" (p. 62). The policy also states that "all new I[initial] T[eacher] E[ducation] qualifications must be endorsed to indicate the holder's level of competence in specific languages by using appropriate labels…" (p. 13). Yet no guidance is offered on how a pre-service teacher's language proficiency ought to be determined or assessed.
While language proficiency testing is not new (Burt & Dulay, 1978;Elder, 1993;Peyper, 2014;Richards, JC 2010;Shulman, 1986) and many researchers offer criteria to assess, not all measures apply to how chalk-face teachers communicate orally while teaching. Butler (2004) includes listening, oral fluency, vocabulary, pronunciation, grammatical accuracy, reading and writing in his study of Korean, Taiwanese and Japanese teachers' perceptions of their English proficiency levels. Gan (2012) focussed on the oral skills of English second language pre-service teachers (PSTs), identifying their English vocabulary, grammar, pronunciation and intonation as problematic areas. Some studies have a broader focus than mere spoken proficiency, such as Low et al. (2014) who studied teachers' English communication skills using the International English Language Testing System that assesses reading, writing, listening and speaking skills. Moon (2014) tested the written literacy skills of teacher undergraduates by assessing their spelling, vocabulary and word building, their punctuation, sentence construction and grammar.
Fewer researchers focus on oral skills specifically required for teaching. Pioneers like Flanders, Richmond, Mottet, Beebe, Hurt, Scott, Wells, Willis and the McCroskeys, have greatly enriched the field of instructional communication since the middle of the previous century. More recently, names like Butler (2007), Mercer and Dawes (2014) and Walsh and Li (2016), along with local studies like De Jager (2012), Erasmus (2018), Evans and Cleghorn (2012), Peyper (2014) and AHC Uys (2006) have contributed to the field of teacher talk and classroom discourse.
H Richards et al. (2013) studied the language proficiency and teaching practices of foreign language teachers but focussed more on their subject knowledge, use of language resources, modelling of correct language use, provision of corrective feedback, management of learner behaviour in the target language, giving meaningful explanations, providing rich language input and their ability to improvise. Tsang (2017) judges teachers' English proficiency by their ability to engage learners in classroom activities. Theron and Nel's South African study (2005) asked Grade 4 teachers of English who taught speakers of other languages to rate their basic interpersonal language skills and their cognitive academic language skills on a scale from poor to excellent. Their findings suggest that these teachers needed language support. Hugo and Nieman (2010) focussed on concerns and needs of primary school teachers in South Africa using English as a second language and identified teachers' pronunciation, vocabulary and confidence as challenges. Cummins (1979Cummins ( , 2000 differentiates between two levels of proficiencylanguage used for everyday communication (basic interpersonal communication skills -BICS) and higher order language required for teaching and learning (Cognitive Academic Language Proficiency -CALP). A person's ability to converse in a particular language (BICS) is associated with fluent, grammatically accurate speech, clear pronunciation, and a broad vocabulary (Elder, 1993;International English Language Testing System, 2015;Padilla & Sung, 1999). This general language proficiency does not routinely translate into high levels of CALP but does influence the level and nature of classroom discourse (Canh & Renandya, 2017). It can thus not be assumed that speaking English socially with reasonable fluency, enables one to teach content and manage instructional contexts effectively through the medium of English. We believe that teachers need high levels of both BICS and CALP as teacher talk used during instructional contexts shuttles along a BICS-CALP continuum.

Conceptual Framework
Research that informed the content and structure of our conceptual framework included that on instructional communication (Evans, 2015;McCroskey & Richmond, 1992;Meyers, 2010); teachers' use of language in the classroom (Chadwick, 2012;Erasmus, 2018;Freeman, 2017;Freeman et al., 2015;Loughran, 2010) and language skills required by teachers (Elder, 2001;Loughran, 2010;Pasternak & Bailey, 2004). We synthesised and categorised language skills, structuring them into a hierarchy of three proficiency tiers (Chadwick, 2012;Cummins, 2000;Lucas, Villegas & Freedson-Gonzales, 2008) suggesting that the skills and competencies required for using English as a medium of instruction are: Oral English proficiency which is foundational and equivalent to BICS, Classroom English proficiency which includes the ability to give instructions, praise, reprimand, and pose questions skilfully; as well as the sophisticated linguistic skills required to have mastered Instructional Communication competence the ability to engage learners actively while expounding the new knowledge/skills competently. Figure 1 visually illustrates our thinking. Arrows indicate which skills directly support others, with the bi-directional arrow linking vocabulary and subject content knowledge indicating their interdependence. Our study focused on oral and Classroom English proficiencies.

Methodology Design
We chose a mixed-methods triangular design for this case study as different sources of data were required to elicit two sets of views. Mixed methods improve research quality by providing better, more reliable insight into complex research problems than a single method could (Johnson & Onwuegbuzie, 2004).
A pragmatic paradigm framed the study, incorporating interpretivist and constructivist views implying that we acknowledge that the respondents' perceptions of the PSTs' English proficiency were informed by a language schema shaped in a community of non-native English speakers.

Participants
The oral English proficiency of isiZulu-speaking final year PSTs aiming to graduate with a BEd (Intermediate Phase) formed the unit of analysis. Respondents were drawn from this cohort specifically since policy dictates a switch from mother tongue to English at the start of this phase (Grade 4-7). The switch is sudden and has several ramifications, especially for rural teachers (Evans & Nthulana, 2018).
All internationally accepted ethical considerations were adhered to and after receiving institutional clearance, tutor participants were contacted telephonically to explain the purpose and intended data collection procedures, received a letter of invitation to participate in the study via email and distributed these letters to PST participants. Convenience sampling based on accessibility to the principal researcher resulted in 52 respondents drawn from a cohort of 325 isiZulu-speaking PSTs who had enrolled at a private higher education institution that offered distance education in Kwa-Zulu Natal, South Africa. PST respondents, male and female, were between 22 and 39 years old with a mean age of 29 years. Furthermore, 18 of their 21 tutors voluntarily consented to participate. Tutors knew the PSTs well as they had provided regular academic support at learning centres and had observed them in situ over a three-and-a-half-year period. Tutor respondents drew on their perceptions of the whole student cohort rather than only those who participated in the study.

Data Gathering
Data were gathered from respondents through questionnaires and audio recorded lessons presented during the mandatory period of work integrated learning (WIL). These methods were chosen with the researchers' knowledge of best practices, at the time, and taking the scope, logistics and resource constraints of the study into account.
Questionnaire data were sourced from 52 PSTs and 18 tutor respondents. A pre-service teacher and a tutor version of the questionnaire was developed based on variables of oral and Classroom English proficiency identified from literature (Butler, 2004;Elder, 1993;Gan, 2012;Hugo & Nieman, 2010;Low et al., 2014;Moon, 2014;Peyper, 2014;Richards, H et al., 2013;Theron & Nel, 2005;Uys, AHC 2006). Both versions garnered biographical details and a language profile (Section A), while Section B gauged the perceptions of the PSTs' spoken English while teaching. Each questionnaire contained 14 open-ended questions (used for qualitative analysis) and 38 closed-ended/Likertscale questions (used for quantitative analysis). The questionnaire items are summarised in Table 1. In addition, all students were invited to submit audio recordings of full lessons. Voice, rather than video, recordings were used to simplify adherence to ethical requirements. Only eight partial lesson recordings materialised (see Table 2). Requests for more full lesson recordings were made but could not be obtained within the time and resource constraints of the study. Data from lesson recordings, mainly qualitative, some quantified, provided further insight into oral proficiency levels and helped triangulate perceptions to improve trustworthiness.

Data Analysis
Quantitative and qualitative data were analysed independently before being triangulated. Quantitative questionnaire data were statistically analysed using the Statistical Package for the Social Sciences (SPSS), specifically tested the reliability of the questionnaire using Cronbach's alpha coefficients ( ). After reverse-scoring was done for some items to ensure that all items of a construct were in the same direction, constructs were created by averaging over items belonging to the same construct. Since the p-values of the Shapiro-Wilk test were less than 0.05, the constructs were not normally distributed and, accordingly, nonparametric statistics were used. The Spearman correlation (inferential) was run where p-values less than 0.05 indicate significant correlations and the Mann-Whitney (MW) test (inferential) was run where p-values less than 0.05 indicate significant differences between the tutor and student responses.
The open-ended questions and audio recordings (also transcribed) were initially analysed by manual coding to identify patterns, and later supported by software (ATLAS.ti) and oral English proficiency rubrics. Comparison of qualitative data with relevant quantitative data improved the trustworthiness within the context of this study. Due to the small sample size, findings have limited transferability to a wider group.
A rubric mitigates the influence of preconceptions to some extent by endeavouring to provide a uniform frame of reference against which proficiency can be rated. Our self-designed rubric (Appendix A) was informed by our conceptual framework and descriptors from the following sources: the International English Language Testing System (IELTS) Speaking: band descriptors (public version) of the British Council, IDP: IELTS Australia and Cambridge Assessment English (2015), the Oral English Proficiency Test (OEPT) of Purdue University (2012), the Stanford Foreign Language Oral Skills Evaluation Matrix (FLOSEM) designed by Padilla and Sung (1999) and adapted by Butler (2004), and descriptors included in research done by Elder (1993Elder ( , 2001, Peyper (2014), and U Uys, Van der Walt, Van den Berg and Botha (2007).

Reliability, Validity and Trustworthiness
In Table 1, for each construct, the items and 's are provided. For conciseness, the item-level descriptions for only the tutor questionnaire are given; for the student questionnaire, the items are similar, but phrased so that the questions ask students about themselves. When testing reliability of a questionnaire, 's of 0.60 or greater are generally accepted by researchers in the social sciences (Ghazali, 2008).
From Table 1 it can be seen that some 's were below 0.6. In such cases, we conducted an item analysis guiding us in deciding on which item(s) to drop. For the construct "use of code switching", the item analysis suggested the removal of one item (How often do students code switch to isiZulu when learners struggle to understand?) which increases the Cronbach's alpha for the tutor questionnaire from 0.511 to 0.786 and for the student questionnaire from 0.620 to 0.758, which are acceptable. The lowest Cronbach's alpha coefficient equals 0.361 for the construct "pronunciation" for the tutor questionnaire. In this case, items could not be removed to increase the Cronbach's alpha values, as this construct consisted of only two items. The unreliability of the two student teacher pronunciation items on the tutor questionnairerelating respectively to clarity of pronunciation and likeness of pronunciation to that of English first language speakersmay be due to PSTs' varying degrees of accurate pronunciation leading to uncertainty as tutors were expected to provide one answer based on their group of students. Thus, after the removal of one item, all 's were acceptable, except for the construct "pronunciation" (for the tutor questionnaire only). A recommendation for future research is that more items be added to this construct as it is well known that as the number of items of a construct increases, the will also increase (Field, 2018). p < 0.001 *How good are students' English vocabularies required for managing the classroom? Do students generally easily find the right words to explain a concept to learners? How often do students find it easy to introduce a new topic in a lesson? How often do students explain concepts in a way that learners easily understand? How often do students rephrase their explanations if one or more learners did not understand? Do students help learners understand new vocabulary when introducing a new topic? Do students sometimes "get stuck" using English in the classroom? When learners answer questions during a lesson, do students sometimes rephrase learners' answers to make it clearer to the rest of the class what they mean? U = 270.000 p = 0.010 When learners answer questions during a lesson, do students sometimes rephrase learners' answers to replace basic words they have used with more academic words? When learners answer questions during a lesson, do students sometimes rephrase learners' answers to correct errors in their language use? = 2.907; Mdn = 3.000; SD = 0.694 = 3.367; Mdn = 3.333; SD = 0.487 Note. *Items flagged by a star had options "1 = poor", "2 = not so good", "3 = good", "4 = excellent." All the other items (not flagged by a star) had options "1 = usually not", "2 = sometimes", "3 = most of the time", "4 = almost always".
Saunders, Lewis and Thornhill (2016) define face validity as "agreement that a question, scale, or measure appears logically to reflect accurately what it was intended to measure" (p. 716). All members of the research team agreed that the questionnaire met these criteria.
For the qualitative data, credibility of openended questions was supported by triangulation with quantitative questionnaire data and transcriptions of voice recordings. Credibility of the voice recordings was accepted at face value. To ensure accuracy and support trustworthiness, the transcriptions were analysed alongside the actual recordings to accurately capture more nuanced elements, such as intonation and speech patterns. Findings based on voice recordings alone were interpreted with caution as the data set was limited to eight extracts from lesson presentations and were thus not considered representative of the full range of PSTs' oral proficiencies.

Findings
In this section we present the respondents' perceptions of the PSTs' oral English proficiency based on what we gleaned from questionnaires and audio recordings. We report in the same order as the constructs appear in Table 1. Before considering them separately, it should be noted that for English proficiency and the eight proficiency skills, the tutor and student responses differed significantly. This is evident from Table 1 where it can be seen that all the p-values of the MW test are less than 0.05. The mean ( ), median (Mdn) and standard deviation (SD) are provided per construct for the tutors and students respectively, and by investigating these statistics it can be seen that the students' responses were statistically significantly higher than those of the tutors for each construct. This finding can be attributed to the Dunning-Kruger effect (Kruger & Dunning, 2009) which is a psychological concept based on the tendency of individuals to have overtly favourable (i.e. overconfidence) or optimistic views of their abilities in social, cognitive and intellectual domains.

English Proficiency
The PSTs' English proficiency was rated as excellent by 23% of students though only 6% of tutors. A good rating was given by 72% of the students, while 61% of the tutors agreed. Tutor (33%) and student respondents (4%) rated PSTs' proficiency as not so good. None gave the PSTs a poor rating.
These perceptions were likely influenced by peripheral normativity (Blommaert, Muyllaert, Huysmans & Dyers, 2005), meaning that the PSTs' English was compared against other members of the community and deemed better by most respondents. Tutor perceptions appear to be less affected by this consideration which explains the difference in their responses across proficiency skills. Comparison of expressed English proficiency to specific language proficiency indicators (such as those on authors' oral proficiency rubric), suggests this perception is not fully accurate when a more widely accepted view of English proficiency is applied.

Vocabulary
Hugo and Nieman (2010:66) claim that " [v]ocabulary is one of the most important components in language acquisition. A lack of vocabulary can lead to a breakdown in communication, which forms a vital part of a teacher's instruction in a classroom." We attempted to gauge whether the PSTs' vocabulary range required for teaching was sufficient. The student respondents believed so, with 71% considering their range good, and 23% as excellent. Approximately two-thirds of the tutor respondents agreed, with 67% rating the PSTs' range as good while 11% considered it excellent. However, when asked how often PSTs found the appropriate words to explain a concept, there were 10% fewer positive responses.
Analysis of lesson recordings supported these findings. While the respondents' BICS vocabulary was adequate, they struggled with common expression, at times resorting to code switching. For example: "Where do we put margarine? Other than, other than, arh. On the [isiZulu word]. We don't say on the [isiZulu word]. We say on the bread."; incorrect word choice: "You are going to write number A"numbers instead of letters); and use of incorrect word forms: disability instead of disabled.
Applying our self-designed rubric, the PSTs' vocabulary range fit the third level descriptor: Able to express content knowledge and engage learners at a basic level. Occasionally "gets stuck" explaining complex concepts.

Grammatical Accuracy
Questionnaire data correlated well with PST respondents' perceptions that they made few grammatical errors, spoke English well during classroom interaction (significant at the 1% level; rs = 0.379, p-value < 0.001) and possessed high levels of confidence in teaching in English (significant at the 1% level; rs = 0.321, p-value < 0.001).
When asked how often PSTs used correct grammar, one-third of tutors (33%) answered sometimes while the remainder (67%) answered most of the time. Asked how often PSTs used the correct tense, tutors answered sometimes (39%), most of the time (56%) and almost always (6%), respectively. Student respondents, on the other hand, had a better perception of their grammar usage as indicated by these responses: sometimes (13%), most of the time (66%) and almost always (19%). They also believed that they used tenses correctly as shown by these responses: sometimes (9%), most of the time (45%) and almost always (42%). Yet we found that grammatical errors abounded in all eight lessons recorded.
We focussed, not on the nature of these grammatical errors, but on the frequency of occurrence and the effect on comprehension. The number of clearly identified grammatical errors per lesson are presented in Table 2.

Pronunciation
The respondents were asked how the PSTs' pronunciation compared to that of first language speakers. Our statistical analysis showed results that were neither reliable nor valid (see Table 1). Our analysis of the recorded lessons showed that PSTs' sporadic mispronunciation, together with hesitancy and garbled sentence construction, created a less than ideal impression of their English proficiency. In five of the eight lessons recorded, PSTs pronounced several words so oddly that they were unrecognisable and affected our ability to follow the lesson. The rubric descriptors adhere to a languagefor-specific purposes stance, which Freeman et al. (2015) explain as a position in which the language standard is determined by others who work in comparable contexts of use. As applied here, this implies pronunciation commonly heard and understood in the community. We thus rated the PSTs' expression on level three: Occasional unclear pronunciation, comprehensible with limited effort. Meaning largely uninfluenced.

Language Sophistication Expressed through Engagement of Learners
We acknowledge that it is difficult to separate the degree of oral proficiency and teaching prowess as they generally operate in unison to effect various levels of learner engagement. Since previous studies also identified the combination of these skills as indicators of language proficiency (Elder, 1993(Elder, , 2001Peyper, 2014;Uys, U et al., 2007) we did not attempt to isolate them.
Quantitative items relating to learner engagement asked how learners were engaged, whether they were encouraged to ask questions and whether they were asked to share their knowledge on a topic. Over 75% of respondents indicated that engagement occurred most of the time or almost always. This engagement seemingly manifested in opportunities to ask questions, use learning support materials, do tasks or participate in group work. However, apart from limited questioning, this apparent high level of learner engagement was not evident in the recordings, possibly due to the limited number of audio minutes available. We thus rated this aspect at the lowest level: Any encouragement of engagement is at a basic level and does not require cognitive demand of learners.

Language Use Expressed through Teaching Techniques
Six items (see Table 1) on both versions of the questionnaire gauged this skill as used most of the time or almost always by between 46 and 91% of the PSTs. Student respondents reported more frequent use than tutors for four of the six teaching techniques. This difference may be an example of the Dunning-Kruger effect mentioned previously.
As audio data were lean, and we had no access to written lesson plans, we anticipated not being able to gauge this aspect satisfactorily. Code switching (not included as a teaching technique item on the questionnaires) was the only attempt to help learners understand unfamiliar vocabulary. Inaccurate triangulation of data sources resulted in our not confidently assigning a level to language expressed through teaching techniques.

Expression of Subject Content Knowledge
When asked how often PSTs found it easy to introduce a new topic in a lesson, the tutor responses were scattered across all possible response options: usually not (6%), sometimes (17%), most of the time (44%) and almost always (28%). Student responses to this item were noted as sometimes (11%), most of the time (49%) and almost always (38%). Asked how often students explained concepts so that learners easily understand, tutors answered usually not (6%), sometimes (33%) and most of the time (56%). Student respondents on the other hand, perceived their ability better by responding with most of the time (62%) and almost always (34%).
Five instances of facts incorrectly presented were identified across three lessons. One example, in particular, highlights how limited language proficiency could hinder learning. In the Science lesson on circuits, while connecting different objects to the circuit to establish whether they were insulators or conductors, the respondent stated that "[c]eramic tiles doesn't turn on." What we believe was meant was that when the ceramic tile is connected, no current flowed to illuminate the lightbulb.
Based on the quantitative results and frequency of errors identified in recordings, the PSTs were tentatively placed at level three on our rubric: Subject content knowledge expression is occasionally unclear, though correct meaning can be derived.

Use of Code Switching
In the South African context, many learners are still becoming proficient in English while simultaneously trying to understand the learning material presented in this language. Responsible code switching is thus encouraged by policy and could be a productive teaching choice for both classroom management and content collaboration (King & Chetty, 2014;Van der Walt & Ruiters, 2011). Yet at times, it may be used due to deficiencies in the teacher's language proficiency.
Questionnaire items were designed to identify the frequency of as well as the reasons for code switching. It was evident that PSTs primarily used English to settle learners before a lesson and to give instructions prior to completing an activity. However, 46% of the PSTs code switched when learners struggled to understand.
From lesson recordings we differentiated between code switching for convenience, to support learner understanding or due to respondent inadequacy. We identified 33 instances where code switching was used to assist comprehension since the respondent followed up the isiZulu with an English equivalent. In 38 instances we were uncertain why the respondent had code switched but ascribed it to likely being convenience or habit. Only in one instance, (Life Skills baking lesson) did the PST struggle to remember a simple English word; possibly as more than one appellation for bread exists in isiZulu.
Although the reason why PSTs code switched was often unclear, all instances were brief. They best matched the level four descriptor on the rubric -Code switches occasionally only in short phrases, mostly to translate what has been said in English -as code switching was seldom used when the preservice teacher did not know a word (required to meet level 3) nor was it used only to support learner understanding (required to meet level 5).

Support of Learners' English Development
Respondents believed that the following strategies were used to support learners' English development: explanations, use of learning and teaching support materials, group work, rephrasing, and code switching. Although we did not ask how frequently these strategies were used, their implementation was limited to the explanations of words and to aid comprehension rather than purposefully develop learners' English. Examples from recordings are: So, if you can distinguish them, I mean the difference between the two … (Natural Sciences and Technology circuit lesson). Anonymous, yes. The person who wrote the poem does not want to be known (English poetry lesson).
With the misalignment between questionnaire responses and limited findings from the recordings, we did not score this construct.

Discussion
Triangulated data indicate a mismatch between the perceptions that the PSTs had of their English proficiency and those held by the tutors. Researcher assessment using a non-standardised protocol suggests that the PSTs had several linguistic lacunae which needed addressing to reach a level sufficient for effectively facilitating learning through the medium of English. These mismatched perceptions that graduates entering the teaching profession have sufficient linguistic prowess to mediate learning and the actual inadequacy of many beginner teachers' English proficiency in meeting the facilitation demands of the classroom, create false expectations in parent communities and among other role players in education.
These PSTs possessed a fair BICS proficiency judging by their general ability to express themselves. Despite their frequent grammatical inaccuracies and unclear pronunciation, overall comprehension was not compromised. While their code switching was infrequent and supportive of learning, purposeful attention to developing learners' English proficiency was not evident. The PSTs possessed a limited academic vocabulary and lacked the strategic and discourse competence required for quality teaching. Being able to develop sophisticated linguistic abilities of their own, PSTs should be able to encourage a high cognitive level by learners.
Anecdotal evidence from tutor respondents indicates that specific PSTs were requested to make recordings, typically "better" and reliable students who were also likely to have higher levels of oral English proficiency than their peers. It is thus reasonable to surmise that the remaining PST cohort may have been even less proficient than those sampled for this study.
Our findings should be extrapolated to a different context with a similar population with much caution, yet it is reasonable to argue that many PSTs expected to use a medium of instruction that they have not fully mastered would display oral inadequacies which may impact learning opportunities. It is indicative that not only the respondents (who may be more proficient than many of their peers), but possibly many non-native speakers teaching through English, require more support to develop their oral proficiency as concluded by Hugo and Nieman (2010:68) in their study of South African teachers: It is clear that a basic knowledge of English grammar and vocabulary is not sufficient to properly teach in English as a second language. The basic knowledge of English that many South African teachers have often masks their deficits in using the language effectively in the classroom. A consequence is that teachers' poor English ability is unfortunately passed on to many of their learners, with far-reaching consequences.
This study contributes to the literature base of instructional communication by synthesising several applicable variables to frame oral proficiency on the BICS-CALP continuum as foundational to enabling teachers to facilitate learning effectively. Our findings support earlier research indicating how high levels of oral proficiency enable facilitation of teaching and learning. The significance of our study lies in foregrounding the continued need to empower PSTs not only to mediate learning appropriately but also managing learner engagement and behaviour using English. This should not be limited to support during their internship experiences but should be sufficient to permit linguistic confidence and efficacy in their careers in classrooms that are rich in linguistic and cultural diversity. In such classes especially, the teacher is often the sole model of English as the target language and thus a high level of proficiency would assist vicariously in developing the learners' English skills as well. Some of our findings may already be useful considerations to the appropriate redesigning of BEd programmes nationally and may even be useful for the professional development of in-service teachers as well, aligning with the national priority of improving teacher education and development as outlined in the Integrated Strategic Planning Framework for Teacher Education and Development in South Africa 2011-2025(DBE & DHET, 2011. Focussing on instructional communication as a discipline-specific skill also requires attention. It is thus imperative that teacher education programmes include interventions that support PSTs in improving their linguistic proficiency and communication skills necessary for effectively facilitating learningregardless of the LoLT used. Curriculum developers should identify how, within existing programme structures, they could better support oral proficiency development across the curriculum, by for example, providing frequently used instructional phrases required for elicitation, explanation, questioning, praise or reprimand. Other recommendations relate to mentors formally assessing and providing feedback on the communicative skills observed during lessons. In addition, watching and then dissecting lessons taught by proficient speakers of English may serve as linguistic exemplars. Guided peer and self-assessment of (video-recorded) lessons in situ are also recommended.
Drawing on existing programmes designed to equip persons who intend teaching English to speakers of other languages, materials developers could incorporate the pedagogical and communicative skills essential to teaching non-native learners. As national policy (DHET, RSA, 2015) dictates that Intermediate Phase teachers specialise in teaching a first and an additional language, it would be meaningful to alert prospective teachers to how language teaching could be integrated into content subjects.
Content and Language Integrated Learning (CLIL) is widely implementedin at least 30 countries in Europe alone (Coyle, 2007). While not actively, thoughtfully and purposefully applied in South African schools, CLIL appears to be taking place out of necessity, with learners learning through English before they have mastered English. Further research on the implementation of the CLIL model in South African classrooms is required.
The massification of higher education has resulted in large classes which militate against effective development of oral proficiency but we recommend that some creative (technological) intervention be designed to purposefully teach the underlying oral skills identified in this study using a practice-based approach. Finally, our self-designed rubric upon refinement may serve as a standardised protocol to gauge the level of PSTs' oral proficiency in any LoLT and could serve to determine the scope of the intervention required. A more comprehensive rubric may also mitigate the influence of preconceptions.

Conclusion
Unsurprisingly, our study confirmed that yet another cohort of PSTslikely to be appointed to monolingual, rural schoolslacked well-developed oral English proficiency. They also perceived their English to be better than it was. This is especially disconcerting considering their target audience has just entered a new scholastic phase with all its concomitant unfamiliaritykey being an abrupt switch to a new LoLT. We acknowledge the limited scope of this sub-study as well as the need to formulate more nuanced rubric descriptors pertaining exclusively to linguistic proficiency.
Further larger-scale studies may provide richer oral data from which to refine an assessment tool. However, what is considered adequate proficiency to meaningfully facilitate quality teaching and learning and what is ideal is yet to be agreed upon.
This article emanates from Jessica Kellerman's master's degree in Curriculum and Instructional Design in 2018, under the guidance of Prof. Rinelle Evans. Prof. Marien Graham provided support in the form of statistical analysis and guidance on the reporting of this data. All authors reviewed the final manuscript.

Notes
i.
For several political-historical reasons, Afrikaansspeaking learners are the only language group (aside from English mother tongue speakers) able to undergo their entire scholastic and -until recently -their postschooling education in their mother tongue regardless of their geographical location. When referring to rural learners in this article we exclude this language group. ii. This is an approximate value, as pauses of varying lengths and minimal learner responses are included in the total length of the recording. iii.
Published Often struggles to find the right word and/or uses words incorrectly, influencing meaning.
Able to express content knowledge and engage learners at a basic level.
Some evidence of academic vocabulary. Speaks fluently and with ease.
Extensive vocabulary is evident, as required for presenting subject content and managing teaching and learning. Grammatical accuracy Grammar errors are frequent and significantly influence meaning.
Moving toward accurate grammar use. Meaning is sometimes unclear or ambiguous.
Grammar is accurate more than half the time. The correct meaning can be deduced with little effort.
Grammar is mostly accurate. The few errors that occur do not influence meaning.
Grammar is consistently accurate.

Pronunciation
Difficulty in pronouncing many words clearly, requiring significant effort to understand what is said.
Some effort is required to understand what is said. Pronunciation of a number of words is unclear, influencing meaning at times.
Occasional unclear pronunciation, comprehensible with limited effort. Meaning largely uninfluenced.
What is being said can be understood with very little effort; pronunciation is mostly clear and comprehensible.
Pronunciation is consistently clear and comprehensible.
Language sophistication expressed through engagement of learners Any encouragement of engagement is at a basic level and does not require cognitive demand of learners.
Brief interactions are occasionally identified, limited to basic question and answer.
Well-phrased questions and extending/meaningful rephrasing of learner answers is occasionally observed. Group work is used.
Meaningful engagement occurs (asking follow-up questions, extending learners' answers, encouraging and answering learner questions). Group work is facilitated well.
A range of learner engagement techniques is used and showcase sophisticated language use. Attempts to explain a word that learners do not understand when asked for an explanation.
Spontaneously offers basic explanations for words used in a lesson that learners may not understand.
Purposefully provides thorough explanation of new or complex words before they are used in the lesson.
Opportunities to support vocabulary are purposefully built into the lesson and used.