The effects of translanguaging on the bi-literate inferencing strategies of fourth grade learners

Previous research suggests that enhanced cognitive and meta­ cognitive skills are achieved when translanguaging techniques are applied in a multilingual classroom. This paper presents findings on the effects of translanguaging techniques on teaching grade 4 learners how to apply relevant background knowledge when drawing inferences during reading. It examines the efficacy of simultaneously using the learners’ home language and second language in reading development among bilingual Xhosa­English readers in a rural school in the Eastern Cape, South Africa. The study adopted a quasi-experimental design where the participants attempted pre-tests in the targeted languages and then after an intervention were provided interventions using translanguaging techniques, thereafter they attempted post-tests. The findings indicate improved performance in terms of learners’ use of background knowledge when drawing inferences, instead of heavily relying on the reading text. The researchers argue for a literacy model that integrates skills and practices drawn from all accessible linguistic repertoires of learners when dealing with reading development at elementary grades since this helps learners develop a sense of self, which in return allows them to be active participants in their learning.


Introduction
The ability to draw inferences is probably one of the essential requirements for reading comprehension.It may help a reader understand complex and subtle implicit messages conveyed through the choice of particular vocabulary by the writer.Drawing inferences possibly requires a reader to draw from their background knowledge in order to understand the text better.However, researchers have observed that learners in lower grades cannot draw inferences from the texts they read (Madiba, 2013;Manyike, 2012).The learners often decode print information without any clear evidence of comprehension whatsoever, whereas reading involves a comprehensive amount of cognitive processing, which includes inferring, integrating and evaluating a text for any inconsistencies and loose ends (Pretorius, 2000).Even though a number of factors may contribute to learners' inability to draw relevant inferences (Modisaotsile, 2012), these may be exacerbated by using monolingual models, practices and teaching approaches that tend to perpetuate the belief that education should only be conducted in one language.
On the one hand, current research questions the validity of language boundaries in literacy and applied language studies (Baker, Jones & Lewis, 2012) while the orthodox practices on the other, are still virtually linear and monolingual in orientation.Reading development is done sequentially -in one language at a time -and languages are treated as separate and unrelated entities in multilingual classrooms regardless of the common underlying proficiencies that aid reading ability across languages.It is against this backdrop that this study sought to investigate an alternative teaching approach in which reading development can be enhanced by using two languages in the same lesson in grade 4 classes.This approach is referred to as translanguaging (Garcia, 2009;Hornberger & Link, 2012).
Translanguaging has been found to be one of the models that enhance learners' higher cognitive skills in reading development (Garcia, 2009;Hornberger & Link, 2012).This approach deliberately alternates the language of input and output during a lesson, thus allowing a learner to think and express their ideas in whichever language they are comfortable.So far, however, there is paucity of research that has investigated the effects of translanguaging when teaching learners how to apply appropriate background knowledge in order to draw inferences.Research on teaching inferencing skills tends to embed this skill into teaching comprehension skills in general, which clouds the significance of teaching inferencing as a unique skill that aids comprehension.This paper, therefore, presents the effects of translanguaging as a pedagogy when teaching grade 4 learners how to apply relevant background knowledge when drawing inferences during reading.First, the paper presents a literature review, in which we highlight the need for learners to be taught inferencing skills at lower grades when reading since research shows that they are unable to draw relevant inferences.Then, we present the theoretical framework around which the study was based.Finally, we show the effects of translanguaging as a strategy that can be utilised to enhance multilingual learners' ability to apply background knowledge when making inferences during reading.The study focused on grade 4 learners who learn English as an additional language (L2) and isiXhosa as the home language (L1) in the Eastern Cape in South Africa.

Related literature
Inferencing, in simple terms, can be defined as reading between the lines.In other words, it occurs when a reader blends pieces of information from a text with the reader's background knowledge in order to arrive at a conclusion (Kispal, 2008).The ability to draw inferences helps a reader understand complex and subtle implicit messages conveyed through the choice of particular vocabulary by the writer.Currently, there is no consensus regarding a suitable pedagogy that seeks to address how to teach inferencing skills.The teaching of inferencing skills tends to be embedded into the teaching of reading comprehension.This clouds the importance of teaching inferencing as a distinct skill that aids comprehension.Nonetheless, three main components ought to be addressed when teaching inferencing skills.These are lexical training, which focuses on the function and meaning of certain individual words in a text; question generation, in which learners are asked questions that would require them to make inferences and obtain supporting evidence from the text and prediction, in which learners use contextual clues to provide pieces of information that are missing from a text.
Texts that are conducive for teaching inferencing skills at elementary levels should elicit more interest, prompt more explanations and enable one to make predictions.
Research in South Africa shows that learners in lower grades cannot draw inferences from the texts they read (DBE, 2014).While reading involves a comprehensive amount of cognitive processing, learners in these grades often recite print information without drawing inferences from the text or noticing any inconsistencies and loose ends.Manyike (2012) investigated reading and writing performance among Xitsonga grade 7 learners in township schools.The learners attempted a reading and writing performance test in their home language, Xitsonga.The reading component of the test tested, inter alia, text reorganisation skills and making inferences.Findings indicated that learners performed poorly in the reading and writing skills, which concurs with the observation made by the national Department of Basic Education (DBE, 2014) regarding the learners' poor reading abilities in lower grades.
Little research locally and abroad has attempted to find out more about teaching inferencing skills in elementary grades.The common trend many researchers follow is to focus more on how to teach comprehension in general rather than inferencing specifically.It is assumed that methods that have been developed for general comprehension work in the classroom have the hallmarks of strategies known to be useful for imparting inferencing skills (Kispal, 2008), whereas it is one's ability to make correct inferences that aid comprehension and not the other-way-round.In essence, the ability to draw inferences predetermines one's reading skills.That is, poor inferencing skills cause poor comprehension.
The above studies clearly indicate that there is a need for learners to be explicitly taught inferencing skills in lower grades since they are unable to draw correct inferences when reading.While the national curriculum and assessment policy statement (CAPS) document mentions drawing inferences as one of the strategies learners can use to understand texts better (DBE, 2011), numerous annual national assessment reports on the other hand, acknowledge that teaching inferencing skills is not an easy task for teachers to perform (DBE, 2013;2014).The CAPS document, nonetheless, is silent on how inferencing skills should be taught.

Theoretical framework
This study is premised on the translanguaging paradigm.Translanguaging is a concept that came about in the 1920s because of, inter alia, the struggle against English hegemony in Wales, in which Welsh was threatened in addition to countering the fallacy that stated that bilingualism caused mental confusion (Baker, et al., 2012).The concept is attributed to Cen Williams (García, Bartlett, & Kleifgen, 2007) and refers to a pedagogical practice in a multilingual classroom in which a learner receives input in one language and gives output in a different one.The practice deliberately switches the language mode of input and output in a well-planned and organised manner to mediate information processing (Lewis, Jones & Baker, 2012).Garcia has broadened the concept to include multiple discursive language practices a multilingual person engages in at school, the street and beyond to formulate and express thoughts to make sense of the world (Garcia, 2009).In other words, translanguaging aids meaning making using two or more languages utilised in a dynamic and functionally integrated manner.
Translanguaging differs from other theories of language acquisition such as the linguistic and the threshold hypotheses in that it recognises the various linguistic varieties a learner has acquired as a single repertoire which a learner utilises to communicate (Celic & Seltzer, 2011).
Unlike the linguistic interdependency and the threshold hypotheses which claim that the successful development of a second language at school depends on L1-L2 transfer (Cummins, 2005), translanguaging advocates for the simultaneous use of the learners' tapestry of languages to acquire, construct and communicate knowledge.The linguistic and the threshold hypotheses give the impression that they assume a linear approach to language development -a skill has to be obtained in one language first and then transferred to another at a later stage.Furthermore, these hypotheses seem to engage a monoglossic approach to language acquisition in that they treat the languages in a learner's repertoire as separate and unrelated entities.In essence, these hypotheses do not appear to cater for common underlying proficiencies that aid abilities across languages, which might allow for simultaneous development in two or more languages.
One may wonder how translanguaging differs from code switching.Translanguaging posits that bilinguals have a single integrated linguistic repertoire from which they strategically draw appropriate features to communicate effectively.Code switching differs from translanguaging in that the former assumes that the linguistic repertoire of a multilingual person comprises multiple separate language systems between which the speaker switches to express a thought.As a pedagogical approach, the teacher often code switches by switching from the prescribed official language of teaching to another language or the learners' home language and back again (Probyn, 2015).Unlike code switching, translanguaging as a pedagogical approach is a result of deliberate alternating of languages of input and output between the teacher and the learners in order to optimise understanding.In essence, translanguaging may be regarded as a process through which a teacher affords learners the opportunity to formulate and share ideas using a tapestry of vocabulary in their entire linguistic repertoire.It allows a space for multilingual language learners to apply different dimensions of their experiences and linguistic knowledge into one coordinated and meaningful performance.
Previous research demonstrates that enhanced cognitive and metacognitive skills are achieved when translanguaging techniques are applied in a multilingual classroom (Garcia et al., 2007).Canagarajah (2011) observes that translanguaging empowers one to be confident in one's identity and to draw from one's background to communicate effectively with a clear voice.It may also afford one the opportunity to interact and negotiate meaning on an equal footing.Creese and Blackledge (2010) investigated bilingual interaction in two complementary schools to establish whether translingual strategies work in the classroom.They found that their participants simultaneously used two languages to convey clearer messages to each other.In order to increase the inclusion, participation and enhanced understanding during the learning process, the teachers employed flexible bilingualism as an instructional strategy to engage students.In some instance, a teacher introduced a concept in one language and explained it in another or asked a question in one language and expected the learners to respond in another.They conclude that one language needed another in order to make complete sense.
A translanguaging model was adopted for this study to facilitate inference-making strategies using the learners' home language and English in the same lesson in a grade four classroom.Simultaneously using two languages would possibly reflect the learners' identities in a positive light in the learning process, which could encourage them to consciously engage and take ownership, resulting in deeper understanding and desirable outcomes of the process.

The study
The study used quantitative and qualitative means of data collection.For the quantitative methodology, a battery of tests that included pre-tests and post-tests, comprising mostly open-ended questions was used.Observations and field notes were used to collect data for the qualitative section of the data collection.

Research design
This study adopted a pre-and post-test quasi-experimental design.According to Dornyei (2007), quasi-experimental research seeks to find out whether the participants' behaviour might be changed after they have been exposed to some intervention or planned learning experience.This design suited this study well because the researchers would be able to establish the cause-effect connection between the intervention and the learner performance.Furthermore, this design would allow for the data obtained to be analysed within and between the groups of participants.The purpose of the pre-and post-tests was to establish the participants' reading comprehension and their ability to draw inferences at the baseline and at the end.After the pre-test, a translanguaging intervention was introduced, after which the participants completed a post-test.
One of the authors was actively involved in the translanguaging intervention; he co-taught the class with the subject teachers.The intervention was done during the scheduled class time, as provided for in the school timetable.It was also done in the regular classrooms used for teaching reading in the languages targeted for the research, which are English and isiXhosa as subjects.The intervention of the study was conducted over two weeks, which comprised five class periods of 5 hours in total.
Furthermore, qualitative methods of data collection in the form of observation and field notes were also used.The researcher observed whether the participants seemed confused in any way by the exercise.Field notes were recorded to capture occasions that might suggest possible confusion due to the process.

Population and sampling
After all ethical issues had been cleared; one rural school in quintile 2 in the Matatiele District in the Eastern Cape was selected to participate.Schools classified as quintiles 1 to 3 are mostly rural and have the poorest learners, measured by the socio-economic conditions of the surrounding communities in which the schools are situated.Some learners in these schools are exempt from paying school fees and the schools are eligible to benefit from the government's national school nutrition programme.A rural school was considered for the study because schools located in rural areas perform badly in the annual national assessments and need more support to improve the quality of education (DBE, 2014).The Matatiele District was chosen because it is one of the districts in the Eastern Cape that performs poorly in the annual national assessments.The research also used a purposive sampling method in that the school was chosen based on the languages used for teaching and learning from grade R to grade 3. The L1 for most of the learners at the school is supposed to be IsiXhosa and learners were transitioning into English as the language of teaching and learning in grade 4. English, as the First Additional Language and IsiXhosa, as Mother Tongue, were offered at the school in grade 4.
Initially, there were two groups of participants from two different rural schools in Matatiele, Eastern Cape.One group, which was the control group, had to be excluded from the study due to an irregularity that occurred during data collection.This irregularity would negatively influence the results of the study.As a result, 61 grade 4 learners aged between 9 and 12 years from one rural school in Matatiele participated in the study.They attempted a pre-test at baseline and a post-test at the end.

Data collection
The research adopted quantitative means of data collection in which two tests, a pre-and a post-test, were used.The tests were designed to assess the learners' ability to draw inferences from the texts.The texts on which the test questions were based were written in IsiXhosa, which was the learners' L1 and English, their L2.Each test was written during the class period according to the school class schedule.Each test had a separate reading passage for each language followed by two sets of comprehension check and text-based open-ended questions.The questions sought to assess the learners' higher cognitive and critical thinking skills.To this effect, the learners were required to respond in writing to questions in which they had to use relevant background knowledge to provide the correct inference that could be drawn from the reading text.Each test had 11 questions, five of which were considered for this paper and deemed sufficient to enable the researcher to make deductions from the answers to determine whether learners could draw inferences.The reading passages were fictional narrative texts about animals and contained approximately 300 words.To ensure the quality and the appropriate level of the language, the reading texts were adopted from the learners' language textbooks.
During the tests, each reading passage was read aloud by the teacher while learners listened to and read along silently.The questions for the pre-and post-tests and for each language had been typed and printed out onto A4 size sheets.Each participant had their copy to read from and answered the questions on the answer spaces provided for each question/ item.Even though the content was different, similar text types (i.e.narratives of approximately the same length) for both languages were used for the pre-tests and post-tests.Soon thereafter, the learners answered questions based on the passage using the language the text was written in.Even though the purpose of the study was to investigate the effects of translanguaging as a teaching approach, in which two or more languages are juxtaposed in the same text or lesson, for test taking it was deemed appropriate to require learners to provide answers in the same language as the one the passage and assessment questions had been asked.This was done in accordance with the standard assessment practice in South Africa.
The learners' responses were scored from 1 to 4 per item, with 4 being the correct answer in which the learner had applied appropriate background knowledge, 3 was an acceptable answer which the learner obtained from the reading text, 2 was used to indicate an incorrect answer and 1 indicated an incorrect answer obtained as is from the reading text.

Translanguaging treatment conditions
During the translanguaging treatment, learners were explicitly taught how to draw inferences when reading.The reading passages that were used for the translanguaging treatment were obtained from the learners' textbooks and an online source, Nal'ibali (Nal'ibali, 2016).Each reading passage was written in two languages.English and isiXhosa had been used in different paragraphs of the same text and in various sentences in the same paragraph.One of the authors actively taught and sometimes, co-taught the class with the subject teachers.
Short passages of approximately 50 words were used to introduce the concept of reading a text in one language and answering questions in a different one.After two encounters, the length of the reading passages was increased to approximately 300 words.Learners were provided with copies of printed reading passages in which isiXhosa and English had been used in respective paragraphs.The following procedure, details of which were outlined in the lesson plan, was followed.
• Before reading -learners were encouraged to make predictions about the text that was going to be read, including what they thought the text would be about based on the keywords that were usually provided at the initial stages of the process.
• During reading -the teacher/researcher often checked with the learners to verify or redirect their predictions.Think-aloud strategies, as outlined in Pannell (2014) were adopted, in which the teacher/researcher read the text aloud, paused at certain points and asked questions whose answers could only be provided by making logical inferences and then the teacher/researcher verbalised his thinking in order to help learners make the correct inferences.Learners were guided on how to apply prior knowledge and the facts from the text to draw correct inferences.For questions whose answers could be located in the passage, even though not explicitly stated in one sentence, the learners were required to provide evidence on which they based their reasoning.In essence, each answer a learner provided was often followed by the question "how do you know?Please, support your answer" which was asked in a different language from the one the text was written in.
Having learners explain their answers helped them to think deeper about what they were doing.The teacher/researcher modelled the inferencing procedure until the students could begin to take over the necessary steps, finally successfully reasoning on their own.
• After reading -when the teacher/researcher had finished reading, learners read the passage and asked each other questions about what they had just read.They were asked to summarise in one or two sentences what the passage was about (the main idea/s).They were also encouraged to ask questions about vocabulary and/or a statement they did not understand.
In essence, translanguaging methods were used to explicitly teach the participants how to draw inferences from a text by applying relevant background knowledge.The participants were asked questions in a different language from the one in which the answers (clues) were provided in the passage.They were allowed to use either language to provide their answers.During the pre-reading and the reading stages of each lesson, the participants were required to provide answers orally but for the post-reading activities, they were required to work in pairs and write their answers down.For questions whose answers could be located in the passage, even though not explicitly stated in one sentence, learners were required to provide evidence by referring to a sentence on which they based their reasoning.

Data analysis
To establish whether translanguaging techniques enhanced the learners' ability to apply background knowledge when drawing inferences, the results of the pre-test and the post-test were compared using the descriptive Statistical Package for Social Sciences (SPSS) to obtain measures of central tendencies (mean) and dispersion (standard deviations).A paired t-test was used to examine the results of the two tests in order to compare means and to establish whether the results were statistically significant.The calculations for the t-test were pitched at an alpha value of 0.05 to measure statistical significance.

Results
Sixty-six grade 4 learners in one school participated in the study.The results showed that at the baseline, the participants obtained a mean of 6.90 in isiXhosa and 6.08 in English.The standard deviation was 1.89 in isiXhosa and 1.88 in English.With regard to the post-tests, 61 participants attempted the test and obtained a mean score of 10.51, with a standard deviation of 1.94 in isiXhosa, while the mean score was 9.28 with a standard deviation of 2.78 in English.The results showed that the participants performed better in the post-tests than they did in the pre-tests.For instance, a comparison of the pre-and post-test results of isiXhosa showed that the participants gained a mean score of 3.61, while they gained a mean score of 3.20 in English.These remarkable gains indicate that the translanguaging intervention had a positive contribution to the participants' performance in the two languages.The results of a paired t-test showed a significant difference between the pre-test and the post-test for isiXhosa (t=-9.831;df=60; p<0.005) and for English (t=-8.047;df=60; p<0.05).As it may be observed in table 1, five learners were absent from school on the day the post-tests were written and as such, their responses were excluded when a paired t-test was run.
The group was found to be homogenous in terms of applying higher cognitive and critical thinking skills when reading in either language at baseline.In other words, the performance of most of the participants could be located around a central point.This was reflected by the standard deviation of 1.89 in isiXhosa and 1.88 in English, each of which shows a stronger level of homogeneity since it is far-off from the mean.At the end, the level of homogeneity seemed to have marginally increased.The dispersion levels of the group increased to 1.94 in isiXhosa and 2.78 in English.This means that participants who could apply background knowledge at the baseline became even better due to the intervention.
In essence, the results show that translanguaging enhances learners' ability to apply background knowledge when reading, thus improving reading comprehension.

Discussion
Previous studies have shown that translanguaging has the potential to enhance cognitive abilities of multilingual learners (Baker, et al., 2012;García, et al., 2007;Madiba, 2012).However, there has been paucity of research on the effects of translingual techniques in reading trajectories among isiXhosa-English bilinguals in elementary grades.This study has demonstrated that translanguaging techniques improve learners' ability to draw relevant inferences when reading in either isiXhosa or English.Presented below are the effects of translanguaging on teaching reading and inferencing skills on each language.

The effects of translanguaging techniques on isiXhosa
The results of the research reflected significant improvement on the participants' ability to draw appropriate inferences when reading (t=-9.831; df=60; p<0.005).This rejects the belief that education should only be carried out in one language and corroborates previous research in this area (see for instance Baker et al., 2012).The results reflected that the participants gained more in isiXhosa than they did in English.They gained a mean of score of 3.61 in isiXhosa, while the gain in English was 3.20.
The reason for the gap between isiXhosa and English gains may be due to the fact that learners had been taught in isiXhosa, their mother tongue, in the previous grades and their English proficiency was still limited. .Madiba (2012) observes that more benefits can be realised when learners' different language repertoires are utilised in a manner that complements each other in the classroom instead of competes against each other.As learners in this grade are transitioning into English as the medium of instruction, translanguaging in this regard neutralises the potential cognitive deficit that might be caused by the early exit from mother tongue tuition to English and facilitates a smooth transition and crosspollination of cognitive competences across the languages.

The effects of translanguaging techniques on English
Translanguaging substantially enhanced learners' ability to apply background knowledge when answering questions based on the English reading text (t=-8.047;pf=60; p<0.05).These findings are consistent with previous research on the impact of translanguaging on higher cognitive abilities (see Lewis, et al., 2012).It was observed that the gains in isiXhosa were higher than English.The seemingly less gains in performance in additional language in the current study might be because it takes learners considerably longer to master a language learning skill in L2 than L1.This corroborates Makalela's (2012) findings on the effects of bi-literacy development on Sepedi, which is an African language and English.He observed that Sepedi showed significant gains while gains in English were statistically insignificant.Furthermore, the findings in this study contradict assertions from bilingual research that claim that more exposure to L2 reading improves L1 reading abilities (see Pretorius & Mampuru, 2007).
What can be deduced from these results is that the learners' L1 and L2 need not be seen as competing against each other but as complementary to one another (Madiba, 2013) in order to facilitate cognitive skills transferrable across languages.This contradicts the observation made by the linguistic interdependency hypotheses, which posit that language transfer is sequential and one-directional.In other words, successful L2 acquisition always comes after and depends on L1.

Incidental observations
What has also been established in the study is that the participants have remained immune from the negative effects assumed could result from translanguaging -i.e. that translanguaging causes mental confusion.During the intervention, no indication of mental confusion or any behaviour that could be associated with confusion due to the process was observed.Instead, the participants were interested and enthusiastic about the undertaking.When they noticed that the languages were being interchanged during the facilitation and had been switched on the reading text as well, they became more interested in the process and seemed freer to use either language in class.This defends translanguaging as a multilingual pedagogy even further from being viewed from the negative light and as Baker (2012: 643) puts it, "from causing mental confusion to the benefits of dual language capability, [and] from solitudes to synergies".

Limitations of the study
One may have noted that during the test taking, the teachers read out the passages and the questions to the learners, while learners listened and read along silently.This deviates from the standard test/exam-taking procedure in which a learner reads on their own, silently and then answers the questions.Moreover, the participants were assessed on the ability to draw inferences only whereas the annual national assessments cover a comprehensive amount of aspects, which, due to the limited scope of the paper, could not be considered.During a normal examination, for instance, learners are often required to produce a piece of writing or answer questions using correct language, grammar and punctuation.It can be argued that if the teachers had not read out to the learners and the scope of the research had not been limited to the participants' ability to draw inferences only, it is not unlikely that the results of the study could have been different.Furthermore, the study would have been strengthened by the data from the control group, the results of which had to be discarded since one of the teachers interfered with the data collection process and suggested correct answers to the participants.
The results of this study cannot be generalised since the participants and the research context cannot be representative of all grade 4 learners in South Africa.However, within the scope of this study it is possible to claim validity since the researchers had a recognisable number of participants from whom they collected data.Furthermore, the data collected from the study was read several times to check for any possibility of factors that might require further analysis until a saturation point was reached.

Conclusion
This study sought to investigate the effects of translanguaging in teaching grade 4 learners how to apply relevant background knowledge when reading texts.The post-test results of the experimental group showed that learners' application of background knowledge improved since the day they took the pre-tests and the extent of improvement is higher in isiXhosa than in English.It can be presupposed that their seemingly less gains in performance in the additional language might be due to limited proficiency in the target language and the fact that they had not yet gained full control of flexible reading strategies due to limited reading instruction in both languages.This paper contributes to the current research on reading development at elementary levels and affirms translanguaging as an appropriate approach that should be considered when teaching.Therefore, it can be concluded that translanguaging had a substantial impact on enhancing learners' ability to apply background knowledge when answering questions based on a reading text.These findings are consistent with research on the impact of translanguaging on higher cognitive skills such as drawing inferences.The paper challenges the prevailing monolithic ideologies about language in education and argues for the engagement of an alternative multilingual approach, such as translanguaging, to enhance reading trajectories in multilingual elementary classrooms.There is a need for more studies to augment the view of translanguaging as an effective pedagogy to enhance higher order thinking skills.

Table 1 :
A summary of the results showing the end-point gains between the two languages