Comparing Reading Performance Between Xitsonga as The Language of Instruction and as the Second Language

Background: As shown by the Progress in International Reading Study (PIRLS) findings over four successive cycles (2006, 2011, 2016, and 2021), South African Grade 4 learners’ performance in reading literacy is low, irrespective of the language in which reading is assessed. Xitsonga home language (HL) learners are assessed in Grade 4, but very little research on early reading development in this language has been undertaken. Objectives: This article aims to examine whether Xitsonga HL reading develops differently when it is used as the language of instruction compared to being used as the second language (L2). Method: The quantitative case study was conducted in five schools in the Mopani district of Limpopo Province. The early grade reading assessment (EGRA) instrument was adapted to Xitsonga, and 75 Grade 1s were selected in this study simply because Grade 1 is an important reading milestone children have to go through to help them cope with more advanced reading skills in higher grades. Results: The findings showed significant differences between Xitsonga HL and L2 in favour of L2. It was unexpected for Xitsonga HL learners to be outscored by learners who were taught Xitsonga as an L2, given that HL learners often use the language at home. Conclusion: Based on the findings, it is evident that teachers need to consider the orthographic nature of African languages when teaching reading in the Foundation Phase (Grades 1-3) classroom. Moreover, teaching reading requires an expert reading teacher who is familiar with the cultural, linguistic, and educational backgrounds of their learners. Thus, this requires teachers to be constantly professionally developed through suitable, accurate, and updated materials to help them improve their teaching of reading.


INTRODUCTION
Early Grade Reading trajectories reflect learners' progress in learning to read according to mastery of various skills during different developmental stages (Torppa et al., 2019).However, being able to read with meaning is a challenge for South African learners.As shown by the Progress in International Reading Literacy Study (PIRLS), 81% of learners could not read for meaning (Department of Education, 2023) despite being assessed in their home language (HL).PIRLS assesses reading comprehension, not decoding skills, but what do such low levels of reading comprehension suggest about decoding?Children are expected to master early reading skills (phonological and phonemic awareness, letter-sound knowledge, word reading, and oral reading fluency) within their first three years in any of the official languages of the Republic of South Africa (these being Xitsonga, Tshivenda, Sepedi, Setswana, Sesotho, isiZulu, isiXhosa, isiSwati, isiNdebele, Afrikaans, English or Sign language) to help them cope with reading to learn in any language throughout schooling.However, despite being taught in their HL for three years, learning to read remains a challenge for many black African learners because most do not seem to have mastered early reading skills to help them read for meaning through Grade 3 (Foorman et al., 2016).There are several factors contributing to the learners' poor performance in reading, and this includes, inter alia, the "teachers' lack of understanding of how early reading develops and how each of the reading activities contributes to different aspects of this development" (Khosa, 2022a, p. 1).
Nonetheless, there is little research on early reading in Xitsonga.Results in Grade 3 from Spaull et al. (2020) show that Xitsonga learners performed at an average of 39.8 words correct per minute (wcpm) in reading fluency and a mean score of 3.4 in reading comprehension; no local data is available yet in Xitsonga Grade 1.This study aims to close the gap by examining whether Xitsonga HL reading develops differently when used as the language of instruction compared to the second language (L2).Because of the multilingual nature of our country, South Africa, although the focus was on Xitsonga HL learners, this study also included Northern Sotho, Shona, and Tshivenda HL speakers who used Xitsonga as their language of instruction (see Table 1).Before turning to the research focus, a highlight of teaching reading in the early grades is presented.This is followed by a discussion of findings from already available assessments of literacy attainments across South African linguistic groups.Another discussion on the presence of systematic confounds when attempting to disentangle second language instruction from home language instruction is presented.Next, a discussion on Xitsonga's psycholinguistic features and how these might influence what needs to be prioritised in Early Grade Reading Assessments (EGRA) is highlighted.Thereafter, the methodology used to collect and analyse data is described.This is followed by presenting and discussing the results.Next, the implications of the findings and limitations are presented.Finally, the concluding thoughts are given.

READING DEVELOPMENT ACROSS SOUTH AFRICAN LINGUISTIC GROUPS
Reading development in South Africa is still in crisis.Several researchers have shown that this may be linked to the quality of instruction rather than anything inherent in the language of instruction (Pretorius et al., 2016;Khosa, 2022b;Oyowe, 2022).A study that examined curriculum advisors' and teachers' perceptions of teaching early reading in the Foundation Phase classrooms in Limpopo Province revealed that teachers teach reading unsystematically, and neither do they abide by curriculum guidelines (Khosa, 2022b).This negatively impacts reading performance among learners.For example, the large-scale assessments of PIRLS 2006PIRLS , 2011PIRLS , 2016PIRLS , and 2021 in terms of assessing Grade 4 learners' reading development revealed that early reading across various linguistic groups is not satisfactory (Department of Basic Education, 2023).PIRLS is an international assessment of reading comprehension at the Grade 4 level that has been conducted every five years since 2001.PIRLS tests reading comprehension for literacy experience (narrative texts) and information (information texts) (Howie, et al., 2017).It also tests comprehension at four different levels of difficulty, such as the learners' ability to focus and retrieve explicitly stated information, make straightforward inferences, interpret and integrate ideas and information, and evaluate and critique content and textual elements.(Mullis & Martin, 2015, p. 16).Thus, if learners progress to higher grades without having mastered their basic skills, they are unlikely to cope with the levels of difficulties expected in higher grades.All four rounds of PIRLS results in South Africa show very poor performance, with most learners being unable to reach the lowest international benchmark of 400 points.In 2016, 78% of Grade 4s could not read for meaning in any language (Howie, et al., 2017).PIRLS 2021 results showed that the number of learners who could not read for meaning in any language increased to over 80% (Department of Basic Education, 2023).Although normal teaching and learning were disrupted by the outbreak of the coronavirus disease of 2019 (COVID-19), these outcomes cannot be blamed on the effects of the pandemic (which resulted in compelling the schooling system to transition to an unusual mode of teaching and learning), given that improvement in terms of performance in the previous rounds did not exceed 5%, as shown in Figure 1 below.% of Grade 4 learners who cannot read in four rounds (Zenex Foundation, 2023, p. 1) Xitsonga results from the four cycles of PIRLS 2006PIRLS , 2011PIRLS , 2016PIRLS , and 2021 showed that Grade 4s tested in Xitsonga failed to attain points higher than 300, and quite interestingly, PIRLS 2021 shows a decline in most African languages, as shown in Figure 2   In South African Foundation Phase classrooms (Grades 1-3), HL is used as the language of instruction (the language used in the classroom for teaching across the curriculum) alongside the L2s (the second to the learner's native language).HL and L2 instructions are determined by the level of proficiency.HL is generally allocated more time (e.g., 8 hours per week) for teaching four language skills (listening, speaking, reading, and writing) in the Grade 1 classroom.This is meant to provide children with opportunities to develop a strong literacy foundation in their HL (Department of Basic Education, 2011, p. 9).L2 is allocated 3 hours per week to provide learners with enough exposure to their oral language skills (e.g., listening and speaking skills).
Generally, children who speak the same language at home and at school have a better chance of experiencing better learning than those who speak a different language (Nag et al., 2019).Although there is relatively limited research examining whether children who speak the same language at home and school would perform better than those who speak a different language, findings from a previous study conducted eight years ago with regard to examining how HL (isiXhosa) as the language of instruction and English as an L2 affect learners' reading comprehension in the Grade 3 classroom revealed a significant difference between the learners' HL and the L2 favouring the HL (Cekiso, 2014).Another study elsewhere examined how students with Swedish as an L1 or L2 in Grades 1-3 perform in screening tests measuring vocabulary, decoding, and reading comprehension in Swedish.The findings revealed that L1 students had significantly performed better than L2 students (Fälth et al., 2023).Nonetheless, the 2011 prePIRLS results showed that South African Grade 4 learners performed significantly lower in their HLs than in English as the L2.Hence, van Staden et al. (2016) argue that performance of this nature stems from a failure to lay a robust foundation in the learners' HL.

XITSONGA LANGUAGE AND ITS PSYCHOLINGUISTIC FEATURES
Xitsonga is one of the 12 languages recognised officially in the South African constitution.Approximately 3 million speakers of Xitsonga are found in South Africa, Mozambique, Zimbabwe, and Swaziland.HL speakers of Xitsonga are in the minority, comprising less than 5% of the population of South Africa.Unlike the Nguni (e.g., isiZulu, isiXhosa, isiNdebele, and isiSwati) and Sesotho (e.g., Sepedi, Setswana, and Sesotho) languages, Xitsonga uses both conjunctive (longer word units) and disjunctive (shorter mean length of words) elements.Like other African indigenous languages, Xitsonga is rich in agglutinating, whereby prefixes, infixes, and suffixes are added to nouns and verb stems (Spaull et al., 2020).Because of its rich agglutinating structures, morphological processing poses challenges for children learning to read (Bosch et al., 2007).Xitsonga, like other agglutinating languages (e.g., Finnish, Turkish, African languages, etc.) also uses transparent (sound-symbol correspondence) orthographic systems in which letters always represent the same sound (Department of Basic Education, 2019).This is what makes agglutinative languages easier to read than opaque (phoneme-grapheme correspondence) languages such as English or French (Goswami et al., 1998;Landerl et al., 1997).
Based on what was discussed above, in terms of redressing the dearth of reading research in Xitsonga, this article reports on a study of African Xitsonga learners that tracked early reading development, focusing on foundational reading skills such as phonemic and phonological awareness (PA), letter-sound knowledge (LSK), word reading (WR), oral reading fluency (ORF), and oral reading comprehension (ORC).This article aims to determine whether achievement differences exist between Xitsonga HL as the language of instruction compared to when it is an L2 in five different schools.One school (School A) that used Xitsonga as an L2 was a middle-class socioeconomic status (SES) school situated in a semi-suburban area with better resources than the four under-resourced schools (Schools B, C, D, and E), which used Xitsonga as the language of instruction and were located in rural settings.For this article, Xitsonga as an L2 is used as a comparison group.Hence, the following research question is posed: • What is the Grade 1 learners' early reading performance of the Xitsonga home language compared to being taught as the second language?
METHODOLOGY A quantitative research methodology was used to compare the early reading performance of learners between Xitsonga as the language of instruction and L2.A quantitative case study research design was used to determine whether Xitsonga HL learners perform better in early reading compared to being used as L2.The assessment of early reading skills in Xitsonga happened over two phases (Term 1 and Term 3 of 2018) through the employment of the early grade reading assessment (EGRA) toolkit.The reliability and validity of the EGRA used for testing learners were ensured by using two untimed subtasks (PA and ORC) only for Term 3 data because the performance was still very low during the first phase.Cronbach's alpha was used to obtain internal consistency for the PA and ORC tasks.The results indicated that the test was valid and reliable (e.g., PA was 0.92 and ORC was 0.88).This study was conducted in five schools situated in the Mopani district of Limpopo Province in South Africa.Four of the five schools (referred to as Schools B, C, D, and E) selected in this study were quintile 2 schools, which used Xitsonga as the language of instruction and English as an L2, while one (referred to as School A) was a quintile 4 school, which used English as the language of instruction and Xitsonga as an L2.Convenience sampling was used to select five of these schools based on accessibility, given that they were situated in the same circuit where I used to work as a teacher.Systematic random sampling was used to select 15 learners per classroom in each of the five schools sampled.Thus, from a classroom of 62 learners, every fifth learner was selected from the class list.Besides Xitsonga learners, the survey also included Northern Sotho, Shona, and Tshivenda-speaking learners, as shown in Table 1.However, most learners at both language proficiency levels were Xitsonga-speaking learners.The Grade 1 learners tested were between 6 and 8 years old.Data in terms of testing learners was collected within a single academic year in two phases: Phase I (Term 1 March 2018) and again in Phase II (Term 3 September 2018).Learners were tested in a quiet classroom, one-on-one.It took approximately 12 minutes to assess each learner.Examples for each task were given beforehand to ensure that learners had initial practice time before attempting the actual task and followed the instructions accordingly.To reduce stress, if the learner could not read anything or got 6 items incorrect consecutively, they were asked to stop and move on to the next task.Below is a presentation of the reading assessment procedure for testing Grade 1 learners: • PA administering procedure: The PA chart comprised 13 items.As an oral task, the learner was asked to listen carefully to the words said aloud in order to delete, substitute, or identify the sounds of each word.One point was awarded for a correct response.
• LSK administering procedure: The letter-sound chart was presented to a learner, containing 110 letters, 10 per row.The learner was shown to read the letter sounds from left to right across each row.A timer was set for a minute so that the learner could begin sounding the letters.In the process of the learner reading, incorrect items were noted.After one minute, the learner was asked to stop, and a large circle was placed around the last letter that the learner had sounded.The total number of letters attempted was recorded, and the number of errors was subtracted from the total number of letters correct per minute.One point was assigned for each letter that sounded correct.
• WR administering procedures: A chart of 50 words was presented to the learner.Thereafter, the learner was given one minute to read each word, while errors were noted.After a minute, the same scoring procedure as above was used.
• ORF administering procedures: Here, the learner was asked to read aloud the first passage (containing 57 words) in one minute and answer questions linked to the passage before reading the second passage of 60 words.Errors were noted while the learner was reading.If the learner read very slowly and struggled, they were not asked to proceed with the second passage.The ORF score was the number of correct words read per minute.
• ORC administering procedures: After learners had read each passage for a minute, they were asked to respond to the five orally presented questions (4 literal and 1 inferential question) per passage.However, questions were asked to the point where they had read.The score awarded for reading comprehension was the number of correct answers given per item.Data from the learners' reading performance were analysed through the usual descriptive and inferential statistical procedures of the Statistical Package for the Social Sciences (SPSS) program.Before proceeding with the main analyses, preliminary analyses were done on the dataset by using Levene's test to check for homogeneity, and the results showed that the variance across the schools was not equal as the p-value in four subtasks (PA, LSK, WR, and ORF) was less than 0.05, except for the ORF subtask.The Shapiro-Wilk test used to check for normality revealed that four of the five schools showed a normal distribution, except for School E. However, because of the small sample size, a non-parametric Friedman test was used for the inferential statistics.

ETHICAL STATEMENTS
The institutional research ethics committee gave the study its approval.It was conducted in accordance with the Declaration of Helsinki, which involved seeking consent from the parents of the learners assessed, given that they were still underage.Verbal consent was also ensured with the learners before they agreed to participate in the study.A good rapport with learners was established to help them feel at ease during testing.Learners were also assured that their participation was voluntary and that they were free to withdraw at any time if they chose to discontinue.

RESULTS
Table 2 presents the descriptive results in Xitsonga as the HL and an L2 in terms of the average performance across the different reading measures in both assessment times (Term 1: baseline and Term 3: end line).It includes the percentage of the HL and an L2 performance based on the zero score.A composite score was also computed, comprising the mean derived from all five reading measures for baseline and end line, respectively.
Table 2 below shows that Xitsonga HL learners outscored L2 in the PA task at both assessment times.This is contrary to the LSK score, which looks higher for an L2 compared to the HL in Term 1 and Term 3. The same applies to the WR and ORF task end lines.In Term 3, most learners using Xitsonga as an L2 (80%) seemed unable to read for meaning in comparison to Xitsonga HL learners (68.3%).The results for Xitsonga as the HL and as L2 in terms of the composite score show that Xitsonga L2 learners performed slightly better than HL learners in Term 1, while the average gap in Term 3 in terms of the composite score widened in favour of Xitsonga L2.In both assessment terms, the proportion of floor effects in Xitsonga as an L2 seems higher compared to Xitsonga HL across the reading measures.The non-parametric Friedman test was used to test for significant differences across the reading measures between HL and L2.The results (Table 3) showed that the p-value is less than 0.05 in both language levels (HL and L2).In this case, we can reject the null hypothesis that Xitsonga HL reading develops differently when used as the language of instruction compared to being used as an L2.There is sufficient evidence to conclude that there are statistically significant differences between Xitsonga HL as the language of instruction and as an L2, where Xitsonga as an L2 seems to be performing better across the early reading measures compared to being used as the HL.The visual illustration clearly shows that, in terms of the mean score (shaded blue), learners in Xitsonga HL outscored L2 learners in the PA task in both assessment times.The standard deviation (shaded red) across the reading measures in HL and L2 is not widely spread out from the mean.

DISCUSSION
The descriptive results revealed that Xitsonga HL performed better than being taught as an L2 in the PA task in both assessment times.This was expected, given that learners enter Grade 1 with some background knowledge of PA skills in their HL.These findings corroborate the findings by Eslick et al. (2020), who demonstrated that first language (L1) learners performed better than L2 learners in the PA subtasks.Surprisingly, the overall results in terms of the composite score revealed that Xitsonga L2 learners performed better than HL learners in both terms.These results confirm Van Staden et al.'s (2016) findings, which established that testing in African languages predicts lower results as compared to English.The results also showed significant differences between HL and L2 across the EGRA measures, suggesting that English is most likely preferred for teaching and learning, given the belief that proficiency in English can guarantee better job opportunities (Massri, 2019).Although the proportion of learners who could not read for meaning in L2 was much higher than HL learners, the fact is that very few learners in both language levels could read for meaning by the end of Term 3.These results share similar findings with PIRLS 2006PIRLS , 2011PIRLS , 2016PIRLS , and 2021, which revealed that the majority of African language learners cannot read for meaning in any language.Although Xitsonga HL learners seemed to have performed better than Xitsonga L2 learners in the PA task, the difference did not show a strong trend between both levels of proficiency, suggesting that most of the Xitsonga HL learners enter Grade 1 classes with limited proficiency in the language they often speak at home.
Research consistently shows that children who enter school with lower levels of oral language are at risk of persistent literacy difficulties (Gillon, 2019).Moreover, research has also shown that learners can obtain low scores even if they are tested in the language often spoken at home and used as the language of instruction in the classroom (e.g., SACMEQ 2000SACMEQ , 2007SACMEQ , and 2013;;Department of Basic Education, 2016;PIRLS 2006PIRLS , 2011PIRLS , and 2016;;Howie et al. 2017).However, poor reading performance in the learners' HL does not have to be the norm if teachers could actually do their work in the classroom.

IMPLICATIONS OF THE STUDY
The findings showed significant differences between Xitsonga HL as the language of instruction and L2.It was not expected that Xitsonga learners at the level of using it as the HL (i.e., the language often spoken at home) could perform lower than those who used it as the L2, suggesting that most of these learners cannot use the language as their HL for academic purposes.This is often a disadvantage to learners because most are unlikely to perform well throughout schooling (Owen-Smith, 2010).

LIMITATIONS OF THE STUDY
Although data in this study was successfully gathered through testing the Grade 1 learners, it was not without limitations, as evidenced by the engagement of unequal sample sizes for Xitsonga as L2 (n = 15) and Xitsonga as HL (n = 75).The difference in sample sizes renders the employment of other statistical tests, such as ANOVAs, appropriate for addressing research questions involving comparisons.Another limitation is with regard to the sample size of learners who were tested; this requires caution in generalising the findings of all South African Xitsonga Grade 1 learners.

CONCLUSION
Although African languages are characterised by simple syllabic structures consisting of V or CV syllables (which make reading easier for the African languages because of their shallower orthography than English, which is opaque), the structures and features of these languages may have implications for reading development.Based on the findings in terms of the learners' reading performance in both language levels, it is important that Grade 1 teachers consider the orthographic nature of African languages when teaching reading in the FP classroom.Moreover, teaching reading requires an expert reading teacher.Thus, for FP teachers to know how they should support learners with diverse language and learning abilities, they need to be constantly professionally developed through suitable, accurate, and updated materials to help them improve their teaching of reading.

Figure 1
Figure 1 below shows performance trends in the PA task between HL and L2.

Figure 1
Figure 1Performance trends between HL and L2

Table 1
Home language statistics for parents of the learners

Table 2
The mean performance of Xitsonga HL and L2 across different reading measures