Meaning-focused vs Form-focused L 2 Instruction : Implications for Writing Educational Materials for South African Learners of English

CITATION: Ollerhead, S. & Oosthuizen, J. 2005. Meaning-focused vs form-focused L2 instruction : implications for writing educational materials for South African learners of English. Stellenbosch Papers in Linguistics, 36:59-84, doi:0.5774/36-0-13.


Introduction
In a report on the Third International Mathematics and Science Repeat Study (Human Sciences Research Council 2000), conducted in 1998/1999 to measure school learners' proficiency in mathematics and science, it was stated that South African learners achieved the lowest results of the 38 countries that took part.One reason cited for this was that the majority of South African learners were not fluent in English, the language of the test, and thus struggled to understand the questions and to communicate their ideas clearly.Significantly, however, learners from other participating countries, such as Malaysia and Singapore, who also face the challenge of having to learn in a second language (L2), performed comparatively well.Yet, unlike these countries, which have one common language in which all learners receive their instruction, the South African language policy of "additive multilingualism" (National Department of Education 2002) dictates that learners learn their home language and at least one additional official language.This could be English, or one of the other ten official South African languages.Therefore, in those schools where English is not the official medium of instruction, and where learners are likely to converse with their peers, teachers and community members in their first language (L1), exposure to English is significantly limited.As stated by the Human Sciences Research Council (2002): Clearly the language issue contributes to the poor subject knowledge of both teacher and pupil in South Africa, and if there is to be a commitment to improving the levels of pupils' performance … then solving the language issue is a critical part of the solution.
As Ellis (1994:243) states, all researchers recognise the need for input in L2 acquisition, although they may differ on the role that it plays.It is reasonable to assume, therefore, that the input that South African learners are receiving in the form of their L2 instruction, is in some significant way deficient.The purpose of this paper is to identify the source of the deficiency of the input, as it appears in educational texts specifically, and to put forward suggestions as to how this deficiency can be addressed.We report on a study undertaken by us to determine how educational texts written for the South African English-as-an-additional language (EAL) syllabus (which adopt a focus-on-meaning approach to English language instruction) compare with educational texts written for the Kenyan EAL syllabus (which favour a focus-on-form instructional approach), in terms of their effectiveness in enabling South African primary school learners to acquire interrogative constructions.
As a point of departure, we will first describe three different viewpoints concerning the role of instruction in L2 acquisition.We will then describe the three different types of instruction used in L2 classrooms, namely focus-on-meaning, focus-on-form and focus-on-formS instruction.A brief discussion of cumulative findings based on recent L2 type-of-instruction research will serve to contextualise our study which we discuss in this paper.The second half of the paper will proceed in the form of a report on the study itself, including the rationale behind it, an exposition of the participants and procedure, and a presentation and discussion of the results.In the conclusion we will suggest ways in which the instructional approach adopted by South African educational texts can be enhanced, so as to enable learners to acquire English more effectively.

Theoretical approaches to different types of instruction
In L2 acquisition research that examines the effects of form-focused instruction (FFI), three different positions may be identified, as summarised in Norris and Ortega (2001:159-160).These are the noninterface, strong interface and weak interface positions.

The noninterface position
The noninterface position, adopted by researchers such as Krashen (1985) and Schwartz (1993), holds that the only kind of information that is usable for L2 language acquisition is naturally occurring instances of the language, otherwise referred to as "positive evidence", or, as Krashen (1985) terms it, "comprehensible input".This position holds that teaching grammar or correcting learner errors has no effect on the learner's acquired linguistic knowledge or interlanguage.Krashen (1985:1-3) maintains that there is no interface between learned knowledge, which results from conscious learning, and acquired knowledge, which results from learners' exposure to comprehensible input.

The strong interface position
In contrast to the noninterface position, the strong interface position holds that, through repeated practice, learned knowledge can be converted to acquired knowledge, which will result in natural L2 use.Proponents of this view are concerned with the question of how this conversion may take place.One such proponent is De Keyser (1998), whose research indicates that explicit FFI leads to significantly larger gains in L2 learning than does implicit learning.

The weak interface position
Researchers who hold the weak-interface position maintain that if L2 material is placed within a meaningful context in an inconspicuous way, but is made sufficiently salient for further processing, it may draw learners' attention to "notice" the form of the target language, and thus eventually to acquire it (Norris and Ortega 2001:159).This position finds support amongst researchers such as White (1989), who suggests that although much of an L2 can be learned on the basis of exposure to positive evidence (viz.information about what is permissible in the target language grammar), learners may need negative evidence (viz.information about what is not Sue Ollerhead and Johan Oosthuizen 62 permissible), when their interlanguage contains rules that are more general than the rules of the target language (White 1989:50).White claims that the parameters linked to L2 learners' principles of universal grammar (UG) have been fixed according to their L1 grammar and that for the L2 learners to change these parameter settings, they need negative evidence, i.e. evidence that a certain form does not occur in the target language.Positive evidence will not suffice in this case, as it will not contain the non-occurring utterances (Le Roux 1994:23).
Similarly, in his discussion of consciousness raising (later referred to as "input enhancement"), Sharwood Smith (1991), claims that for acquisition to take place, learners must consciously notice forms and the meanings they represent in the input.He provides evidence indicating that highlighting forms in the input increases the likelihood of them being noticed and subsequently used.This evidence contradicts Krashen's claim that L2 learners acquire language unconsciously.
Further evidence in this regard was provided by Spada and Lightbown's (1993) study of the effects of FFI and corrective feedback on the development of question formation in the speech of ESL learners.Their findings support the hypothesis that FFI and corrective feedback can effect the aligning of the L2 learner's interlanguage more closely with the target language.
Long's interaction hypothesis posits that instruction helps learning, but only if it coincides with the natural process of acquisition (cf.Ellis 2001:10).This premise is informed by Pienemann's teachability hypothesis (cf.Ellis 2001:7), which predicts that some structures are best learned if the specific instruction coincides with the learner's next stage of development.Long views breakdowns in communication as an opportunity for learners to negotiate for meaning, which in turn highlights problematic forms.This helps learners to notice the gap between the input and their own interlanguage, and stimulates them to restructure the incorrect forms that led to the initial breakdown in communication (Ellis 2001:10).

Three different types of L2 instruction
The interface positions outlined above underlie three significantly different types of L2 instruction, namely focus-on-meaning, focus-on-form and focus-on-formS instruction, which are described below.

Focus-on-meaning instruction
The focus-on-meaning (FonM) approach to L2 instruction corresponds with the noninterface view, by providing exposure to rich input and meaningful use of the L2 in context, which is intended to lead to incidental acquisition of the L2 (Norris and Ortega 2001:160).A FonM instructional approach can be widely found in contemporary English Language classrooms, in techniques such as Krashen and Terrell's Natural Approach, some content-based ESL instruction and immersion programmes (Ellis 1994:571).

Form-focused instruction
The term "form-focused instruction" (FFI) is defined by Ellis (2001:2) as "any planned or incidental instructional activity that is intended to induce language learners to pay attention to linguistic form".It serves as a generic term for "analytic teaching", "focus on form", "focus on forms", "corrective feedback/error correction" and "negotiation of form".The term "form-focused instruction" is used to describe both approaches to teaching forms based on artificial syllabi, as well as more communicative approaches, where attention to form arises out of activities that are primarily meaning-focused (cf.Long and Robinson 1998).FFI comprises two subcategories, namely: focus-on-formS and focus-on-form instruction.

Focus-on-formS instruction
Focus-on-formS (FonFS) instruction is informed by a strong interface view, and occurs when parts of a grammar are taught as discrete units, in order of their linguistic complexity.This is the traditional approach to grammar teaching, and is based on an artificially reproduced, as opposed to an "organic", syllabus.In this approach, language is treated as an object to be studied and language teaching is viewed to be an actvity to be practised systematically.Furthermore, learners are seen as students, rather than users of the language (Ellis 2001:14).

Focus-on-form instruction
Focus-on-form (FonF) instruction, which draws on a weak interface view, involves strategies that draw learners' attention to the form or properties of target structures within a meaningful context.This is done primarily with structures that are potentially difficult, that are learnable according to the stages put forward in Pienemann's teachability hypothesis (Spada and Lightbown 1993:207) and that are likely to be used or needed in future communication.
According to Norris and Ortega (2001:167), instruction may be regarded as FonF if it meets the following criteria: (i) that learners engage with the meaning of a structure before paying attention to its form, through tasks that ensure that target forms are crucial to the successful completion of the tasks; (ii) that instruction in a particular form occurs as a result of analysing learner needs; and (iii) that learners' attention be drawn to a form briefly yet noticeably, "thus achieving a difficult balance between unobtrusiveness and salience".Ellis (2001: 20-23) delineates a useful distinction between planned and incidental FonF.In planned FonF, learners attend to a specific form many times.This may be done in the form of (i) input flooding (proposed by, amongst others, Sharwood Smith (1991)), where input is enriched with numerous examples of the target language without overtly drawing attention to it, and (ii) input enhancement, which highlights target features and draws learners' attention to them.In planned FonF, learners still focus primarily on meaning, and the activities based on the enriched input have a communicative focus, where the language to be learned is useful and natural.
Incidental FonF occurs either as a result of a communicative problem, when users have not understood each other, or when learners want to focus on a form that has been used in communication.Incidental FonF can be pre-emptive, when the teacher or the learner draws attention to a form that seems to be problematic, even though no error has yet been made, or it can be reactive (Ellis 2001:22).Reactive feedback arises when a teacher provides negative feedback in response to a learner's actual or perceived error.A teacher may provide this negative feedback in an implicit way, where instead of negatively judging an incorrect utterance, she repeats the incorrect form with exaggerated intonation, for example: You goed shopping yesterday?or uses recasting (cf.Nicholas, Lightbown and Spada 2001:720), where she correctly reformulates all or part of the incorrect utterance, for example: Oh really?So you went shopping yesterday?
In contrast, planned FoF can take one of three forms (Ellis 2001), namely (i) explicit correction, where a teacher clearly indicates that the learner has said something incorrectly, and provides the correct form; (ii) metalinguistic feedback, where the teacher provides information about the correctness of a learner's utterance; or (iii) elicitation, where the teacher attempts to draw out the correct forms from learners (Ellis 2001).
An important distinction between incidental and planned FonF is that with incidental FonF, various linguistic forms, namely grammatical, lexical, phonological and pragmatic, compete for the learner's attention, whereas with planned FonF, the teacher can select the linguistic form to which the learner should pay attention.According to Ellis (2001:16), this distinction in strategies has important implications for the effectiveness of the way L2 English is taught in the classroom: … it raises the question as to whether language learning benefits most from focusing on a few problematic linguistic forms intensively or from a "scatter-gun approach", where multitudinous problematic forms are treated randomly and cursorily and where the treatment may/may not be repeated.

Findings of type-of-instruction research
Norris and Ortega ( 2001) provide a valuable synthesis of recent research findings regarding the relative effectiveness of different types of L2 instruction.Overall, they found that form-focused L2 instruction results in significant improvements in the acquisition of target language structures.
Delayed post-test results furthermore indicate that the positive effects of L2 instruction are durable.
In general, the findings suggest that explicit FonF instruction leads to more effective acquisition than implicit FonF instruction.Significantly too, it was found that FonF instruction that is integrated into a meaningful context is as effective as instruction that involved a focus on forms.
Enlightening as these cumulative research findings may be, it is important to recognise that they emanate mostly from studies by the likes of Spada and Lightbown (1993) and Trahey and White (1993), which have been conducted in L2 learning situations where the L2 was easily accessible, for example, in immersion programmes in Canada.The average South African child's L2 learning situation is vastly different, in that access to English input is limited, due to factors already mentioned.
Thus, it is encouraging to note that there is of late a growing trend within L2 type-of-instruction research, that has become concerned with the relative effectiveness of types of instruction when matched with learner characteristics, including the learner's age, language aptitude and L1 background as well as the state of a learner's interlanguage.This can only bode well for the relevance of L2 type-of-instruction research results to a society with as complex and challenging a language learning scenario as South Africa.

Examining educational texts as instructional tools
Our reasons for examining instructional effects in the form of educational texts specifically, were twofold.Firstly, in South Africa, where many schools are dogged by challenges in the form of inadequate resources, violence, teacher and pupil absenteeism, and, perhaps most significantly, inadequate teacher training, teachers tend to rely heavily on learner support materials in the form of textbooks, worksheets and so on, to the extent that the methodological approach of the teachers is often informed by the content of these materials.
Secondly, and perhaps of even more relevance, the approval of textbooks is governed by a national submissions process, which stipulates that their content must reflect the rationale of the South African EAL syllabus, through incorporating its critical outcomes, learning outcomes and assessment standards.Educational texts therefore tend to represent the pedagogical views of the National Department of Education's curriculum developers, with regard to what learners must know, how they must learn it and how this knowledge will be assessed.Thus, in the South African context, there is clearly a close link between L2 educational texts and L2 instruction.
For the purpose of this paper, we have chosen to compare the South African EAL curriculum's pedagogical stance with that of the Kenyan ESL curriculum, which informed the other type of educational text used in this study.The first author is an educational materials developer, who has been closely involved in the writing and publishing of English language textbooks for various markets in sub-Saharan Africa over the past five years, including the Kenyan Primary schools market.We therefore had access to both South African and Kenyan English language textbooks

An overview of the South African EAL curriculum
The South African EAL curriculum is informed by an outcomes-based methodology, which encourages a learner-centred and activity-based approach to language instruction, and sets the learning outcomes (LOs) to be achieved at the end of each unit of work.Learners' attainment of these outcomes is assessed according to specific assessment standards, which prescribe the level and depth of knowledge that is to be achieved in each grade.
Due to its complex design features, the currently used Curriculum 2005 (C2005) has undergone revision in the form of the Revised National Curriculum Statement (RNCS) (Department of Education 2002), which is being implemented incrementally between 2004 and 2006, beginning with grades R to 3 in 2004.The RNCS derives its LOs from critical and developmental outcomes, which are "inspired by the Constitution and developed in a democratic process".Thus, it is clear that, in combining knowledge and skills with a curriculum of values that need to be acquired, the RNCS does not only have an educational agenda, but is also concerned with what may be called a "social justice" curriculum.This is because it specifies distinctive issues that need to be addressed in the language classroom, such as poverty, inequality, race, gender, age, disability and HIV/Aids.Some might even say that the RNCS has a political agenda, when reading its assertion that "the promotion of values is important … to ensure that a national South African identity is built on values very different from those that underpinned apartheid education" (National Department of Education 2002:3).
The South African EAL curriculum has six LOs.The first four LOs cover the fundamental language skills, namely listening, speaking, reading and writing; the fifth LO concerns the use of language for thinking and reasoning and the sixth LO, referred to as "language structure and use", specifies what it terms "sounds, words and grammar" that need to be learned through the creation and interpretation of texts, with the caveat that these texts should have embedded in them the necessary values enshrined by the South African Constitution.Thus we find assessment standards such as those in (1) below (National Department of Education 2002), where AS stands for assessment standard: (1) LO2 Speaking, AS: discusses how language constructs knowledge and identity and positions people; critically discusses naming practices (e.g.how slaves and workers were named by their owners / naming people as "makwerekwere") LO3 Reading: notices how characters and plots are constructed to represent a particular view of the world (e.g.How are boys/girls stereotyped?How are different races represented?) We would argue that such assessment standards demand a sophisticated grasp of English from the EAL learner.Significantly, they also emphasise a focus on meaning over form, and favour implicit over explicit instruction.
For purposes of this study, the assessment standard relating to the acquisition of interrogative structures in Grade 5 is noteworthy (Department of Education 2002).This assessment standard is given in (2) below.
(2) LO6 Language structure and use: Understands and uses some question forms, such as 'Why couldn't …?' and 'What/How do you think …?' From our experience, this assessment standard is not particularly helpful to EAL learners, or teachers, who are looking for specific language patterns on which to base their understanding of the fairly challenging task of forming interrogative structures in English.

An overview of the Kenyan ESL syllabus
In contrast to the South African curriculum, the Kenyan Education syllabus (Kenya Institute of Education 2002:2) describes literary texts and language structure as "two autonomous but related entities" that need to be integrated so that learners can learn to use language in a variety of ways.
While it concedes that language skills should not be taught in isolation, but in the context of meaningful communication, it does prescribe the teaching of grammar in the form of "idioms".
These are described as "structures and expressions in the English language that are fixed and unchanging, whose meaning differ from that of their individual words".
The Kenyan Primary English Grade 5 syllabus is broken up into various themes, which comprise learning objectives that focus on fundamental language skills, as well as specific language patterns for certain grammatical structures, such as that for interrogatives: question word + auxiliary + subject + main verb (Kenya Institute of Education 2002).
At secondary level, the Kenyan ESL syllabus ceases to prescribe themes, in favour of a carefully planned curriculum of grammatical elements that need to be acquired at each stage in the school year.One can thus see that the Kenyan syllabus, while incorporating focus-on-form in a planned and explicit manner, becomes more and more focused-on-formS as it progresses from lower to higher school grades.

A description of the participants
The participants in this study were Grade 5 Xhosa-speaking learners of English, who attend a primary school in Masiphumelele informal settlement in the Western Cape.They were educated according to the Grade 5 South African Schools Curriculum 2005.The participants received most of their subject instruction in Xhosa, and thus for the majority of them, English was learned as an additional language, and in some cases even as a second or third language, in addition to Zulu and Afrikaans.
Although the standard age of Grade 5 learners in South Africa is 11 or 12 years, the ages of the participants in this study varied between 10 and 15 years.This is significant in view of research done on the effects of age on L2 acquisition by the likes of Scovel (1998:2), who claims that language is best learned during the early years of childhood, and that after approximately the first 12 years of life, everybody faces certain constraints in the ability to acquire a new language.
Research done by various non-governmental organisations working in the area revealed that the majority of the learners in this area came from a low socio-economic background with few books or educational aids in the home.The educational level of the learners' parents was generally very low, with the highest level usually not exceeding Grade 7.Although the learners lived with several people at home, 20% of them were orphans, and of those with parents, only 40% had both parents living at home.Unemployment ran at 65-70%, and alcohol and child abuse was high.
Learners' generally low educational achievement could be related to all of these conditions.
The school timetable provided learners with 120 minutes of English tuition per week (three Through the first author's participation in a teacher support programme early in 2003, run by the Masiphumelele Corporation (a non-governmental organisation committed to uplifting the community), we have observed that teachers at this particular school prepared lessons chiefly from a selection of textbooks, and that language exercises were selected and photocopied on an ad hoc basis.The teachers then attempted to relate these exercises to the outcomes-based syllabus, a difficult and sometimes impossible task.Quite often, teachers would use concurrent translation, from English into Xhosa, when teaching parts of speech or vocabulary, as well as encouraging rote learning of metalinguistic information, for example: "a noun is the name given to a thing or a person".

A description of the procedure
During a three-week period, the first author gave nine hours of intensive tuition in question formation to two Grade 5 classes, from hereon referred to as "the FonM class" and "the FonF class", using two different sets of learner support materials (one set per class).Having co-written and developed both sets of materials, the first author was well-acquainted with their respective pedagogical approaches.The FonM class was instructed using materials based on a meaningfocused instructional approach, whereas the FonF class was instructed using materials based on a form-focused instructional approach.The L2 type-of-instruction approach that underlie the two different textbooks used to instruct the two groups, was examined in relation to the participants' written performance.A third class, which we shall refer to as the "control class", received no specific instruction in interrogative construction.In this class, the school's Grade 5 English teacher continued to teach according to the C2005 syllabus.
Five days prior to the three-week period of instruction, all three classes completed a baseline assessment test so that we could gain an independent measure of their proficiency in the use of interrogative structures.A post-test was administered three days after the period of instruction.
Because of the limited scope of this study, a delayed post-test to measure the durability of the results, was not administered, yet may well provide an opportunity for follow-up research.

The instructional materials and approach
The participants in the FonM group were instructed using English Matters (Montgomery and Ollerhead 2003), a South African EAL textbook for Grade 5 learners based on a meaningfocused, implicit instructional approach.All activities in this textbook are based on a variety of texts from a reader, which is a separate component comprising an anthology of stories.It is assumed that the acquisition of grammatical structures will occur naturally and fairly unconsciously through the learner's interaction with relevant, engaging texts.
Throughout the three-week instruction period, the FonM class worked through one theme of the book, which is based on a story called Some of us are leopards, some of us are lions.Activities comprised the following: (i) comprehension questions relating to the meaning of the text in the form of yes/no or specific information questions; (ii) an oral exercise in which participants had to work in pairs and discuss questions relating to how the main character of the story felt when faced with different challenges, for example: How do you think Mpumelelo felt when his classmates laughed at him?; and (iii) a writing exercise in which participants had to write a letter to the main character, giving him advice on how to solve his problems.
Potential structures for L2 investigation in these activities were interrogative, passive, pronominal and past tense structures.However, each of these structures was treated fairly superficially, in that they were simply necessary to answer comprehension questions, and were competing for attention with various other aspects of language, including vocabulary, phonology and pragmatic elements.Whereas the theme provided practice in the four main language skills (reading, writing, listening and speaking), the acquisition of forms, in our opinion, was left very much to chance, seeing that nowhere in the learner's book was there any explanation of how to form a question in English.
In instructing the FonM class, errors made by participants in written and oral production were frequently observed, such as How Mpumelelo did feel?or How he is feel?attempting to maintain a strongly meaning-focused instructional approach, however, corrective feedback was not provided to learners.
The participants in the FonF group were instructed using Explore English (Ollerhead and Njoroge-Gachuhi, in  To supplement the textbook's form-focused instructional approach, incidental corrective feedback was provided in response to participants' errors, both written and oral.This was done both preemptively and reactively.Pre-emptive feedback was provided in the form of metalinguistic information.This focused on the role of auxiliaries in question formation, demonstration of subject-verb inversion on the whiteboard using arrows that showed the direction of inversion, and highlighting the different components of the interrogative structures (auxiliary, verb, object and so on) by writing them on the whiteboard in different colours.Reactive feedback was given mostly in the form of recasts; for example, in response to a participant's incorrect question, such as Why the woman is smiling?, the feedback would be Oh, good question.Why is the woman smiling?

The formation of interrogative structures in Xhosa
In order to identify what influence, if any, the participants' first language may have had on their formation of interrogative constructions in English, we examined the salient differences between interrogative structures in Xhosa and English, as described in (Kirsch, Skorge and Magona 1999).
As in English, there are two main types of question in Xhosa.The first requires a yes/no answer.
In Xhosa, at times, only intonation distinguishes statements and questions.To make the distinction clearer, na is also usually added at the end of a question, as illustrated in (3) below.
You-speak Xhosa well You-speak Xhosa well QUESTION MARKER "You speak Xhosa well" "Do you speak Xhosa well?" The second type of question requires answers supplying specific information.Question words usually follow the verb or predicate, and do not occur at the beginning of a question, as is the case in English.An example is provided in (4) below.
You-are-alive how "How are you?" It is common for Xhosa-speaking EAL learners to experience L1 interference in their frequent failure to employ subject-verb inversion, resulting in constructions such as: You are going where? and When you will come back?

The results of the study
The participants had to complete three different types of question-formation tasks at both the pretest and the post-test stage.The post-test was identical in design to the pre-test.This in itself may well have resulted in slight gains in accuracy over the three weeks, according to Norris and Ortega (2001:181), who claim with reference to their study that "up to 18% of the change observed over the course of the investigation may be due to … test practice effects." Each table in the subsections below provides horizontal progress data for each class, expressed as a deviation percentage from pre-test to post-test stage, as well as a vertical comparison of the three classes' overall achievement on each specific task.

Task 1: The ordering task
The ordering task consisted of 12 questions in which a series of words had to be placed in the correct order to form a correct question, for example: smiling why woman the is ("Why is the woman smiling?").Each series of words was accompanied by a picture, which provided contextual clues to assist participants in completing the task.Accuracy was determined in terms of correct word order.Participants obtained one mark for correctly placing the question word and the auxiliary verb in relation to the subject, and a further mark if the remaining word order was also correct.The results of Task 1 are presented in Table 1.In Task 1, all three classes produced a greater proportion of correct questions on the post-test than on the pre-test; however, the control groups' progress was significantly lower (by 5%).This suggests that both FonM and FonF instruction were effective in helping learners to acquire the correct syntax (subject-verb inversion as well as the correct use of the auxiliary in relation to the subject) needed for accurate question formation, as opposed to the control class, which received no specific instruction.
The FonF class started out at a higher level of accuracy on the pre-test than the control and FonM classes; however, the FonM class made marginally better progress on Task 1 from pre-test to post-test (10% versus 9%).The FonF class attained the highest level of accuracy in the post-test stage.Significantly though, despite the 10% and 9% improvement in Task 1 observed in the FonM and FonF classes, respectively, the overall attainment in this task remained poor, with only 39% of learners achieving accuracy in the FonM group and 43% in the FonF group.Common errors encountered in this task included echo constructions such as There are how many pigs?
This suggests that there was little or no understanding either of the correct placement of the question word and the auxiliary in relation to the subject or of subject-verb inversion.When the question word was correctly placed, it was often not followed by subject-verb inversion, for example: Why he is shouting?
If one considers the structure of questions in Xhosa (cf.5.8), one could perhaps attribute the participants' overall lack of accuracy to the fact that their parameters associated with question formation may still have been set to the L1 value, which dictates that no subject-verb inversion takes place, and that the word na is placed at the end of a sentence to indicate a question, rather than placing a question word at the beginning of a sentence.

Task 2: The preference task
This task comprised ten questions ordered in pairs of which the two members differed only in terms of the relative positioning of the subject and verb.An example of such a question pair is given in (5) below.Learners had to indicate which question was correctly formed by underlining it.The results of task one is presented in Table 1.
(5) (i) What game you are playing?
(ii) What game are you playing?In Task 2, participants in all three classes chose a greater number of correctly formed questions on the post-test than on the pre-test.The FonF and control classes both attained a higher level of accuracy on the pre-test than the FonM class (61% versus 51%), yet, once again, the FonM group made greater gains in progress from pre-test to post-test than both the control and FonF groups (22% versus 13% and 19%, respectively).The FonF class attained the highest level of accuracy in the post-test stage.
If one examines the cumulative post-test results, however, there is very little difference in ultimate attainment levels.Although the FonF class attained the highest level of accuracy at posttest stage (78%), this result was not markedly higher than the 74% achieved by the FonM and control classes.One could conclude from this result that neither FonM nor FonF instruction had a significant effect on correct subject-verb inversion.It could of course be argued that this result was not due to factors pertaining to type of instruction but to an inherent weakness in the design of the test: participants may have realised that they had a 50% chance of choosing the correct member of each question pair, and could therefore perhaps have been tempted to guess the correct answer, rather than to reflect on the correctness of each question.Task 2 could have been more effective had an additional variable been added to options (i) and (ii), such as Say whether Sue Ollerhead and Johan Oosthuizen 78 both sentences are correct, both are incorrect, or only (i) or (ii) is correct.This may well have resulted in a more accurate reflection of participants' proficiency.
In examining the results of this task, an interesting pattern emerged with regard to the relative accuracy of the usage of the auxiliaries do and will.Whereas the incidence of correct answers was high on the construction Why do you like going to the library?, a common error was perceived in the construction When you will visit the Transkei?which was frequently selected over When will you visit the Transkei?This could suggest that the auxiliary do is more salient to L2 learners because it is used more frequently than the auxiliary will.It must be stated, however, that this is no more than a tentative suggestion.

Task 3: The yes/no question formation task
In Task 3, participants were required to form yes/no questions, deduced from a series of eight statements.This was a challenging exercise in that it demanded that participants use subject-verb inversion and employ the auxiliaries do, does, must, can, in the absence of the infinitival auxiliary form to be, in both the present and the past tense.In both the pre-and post-test, exemplars of correct yes/no questions derived from statements were provided for participants' reference.Table 3 contains the results of the participants' performance on this task.Significant deviations become apparent when one examines the difference in performance from pre-test to post-test stage of all three classes.The FonM class's performance increased by 6%, while the control group decreased by 2%.This is a puzzling result in that it contradicts Norris and Ortega's (2001:181) observation that test practice effects should result in as much as an 18% improvement observed over the course of an investigation.One possible reason for this is that four new class members wrote the post-test.As they had not taken part in the pre-test, they would thus not have been familiar with its design.
Most striking of all, however, was the percentage gain noticed in the FonF class, in which the initial 26% obtained on the pre-test rose to 51% on post-testing, resulting in a gain of 25% overall.This would seem to suggest that the intensive FonF instruction that the class received was significantly beneficial to participants when they had to integrate all aspects of question formation (including auxiliary placement, subject-verb inversion and word order) in a task where the components of the forms were not provided.
Again, however, it must be noted that post-test results were not satisfactory, with only the FonM class obtaining over 50% accuracy.Generally, there seemed to be a weakness in the correct use of auxiliaries

Conclusions and suggestions
The analysis of the data clearly shows that both FonM and FonF groups outperformed the control group in all three tasks, thus suggesting that both types of instruction were better than no instruction at all.As questions are a naturally occurring feature of classroom interaction, it is certain that the control group would have been exposed to interrogative constructions in some form -however, the results suggest that this input was not sufficient to allow learners to acquire interrogative constructions naturalistically.Significantly, many errors were observed in the natural English input provided by the teacher with regard to interrogative constructions during the five weeks the first author spent working at the school.
While there was considerable variation in individual performance in each class, it was clear that overall proficiency in English was very low.Despite the significant gains that were made in the development from pre-test to post-test stage on all three tasks, it was only on Task 2 that all three classes obtained over 50% accuracy.
In order to obtain more definitive results, a more protracted instruction period and the administration of a delayed post-test (to determine whether gains increased incrementally and were maintained over an extended period), would have been valuable.This provides an opportunity for further research.
While there was very little difference in attainment between the FonM group and the FonF group in Tasks 1 and 2, it became apparent that the FonF group outperformed the FonM and control groups on Task 3, which was the most challenging in that it demanded that participants focus on all aspects of interrogative construction in a variety of tenses, i.e. subject-verb inversion, verb agreement, word choice, word order and so on.
We maintain that this result can be attributed to the fact that, in the FonF class, the textbook activities were organised in such a way as to draw participants' attention to the forms of the target language, within a relevant, communicative context.This was complemented by drawing the participants' attention to errors, in the form of negative evidence.This would seem to support White's (1989:50) claim that positive evidence alone does not result in parameter resetting.In other words, the Xhosa-speaking learners needed to know (in the form of negative evidence) that an utterance such as You are going where? does not reflect a permissible parameter setting for a non-echo wh-question in English.
Although the study has several limitations, these findings have significant implications for the writing of educational materials for South African learners.The poor results achieved by all classes at pre-test stage suggest that the current input provided in the form of South Africa's EAL textbooks is not effective in helping learners to acquire the relevant forms.A frequent observation in educational publishing circles in this regard, is that teachers "aren't being trained to use the textbooks".We would put forward an alternative observation, namely that the textbooks are not addressing the realities and needs of the teachers and learners who are using them.Certainly, in the case of the school visited during this study, which is typical of the majority of South African schools in which English is not the medium of instruction, the input provided in the textbooks that learners were using did not have the desired effect.
The South African EAL curriculum (Department of Education 2002), which draws heavily on a FonM approach to instruction, or at best, a very implict FonF approach, intends language forms to be acquired naturalistically through learners' interaction with "meaningful texts".
Simultaneously, the teaching of "values and attitudes" such as social justice and human rights is given equal weight to the teaching of language skills.This has resulted in what Ellis (2001:16) refers to as the "scatter-gun approach", where forms are dealt with superficially and arbitrarily.
The results of this small-scale study lend credence to Norris and Ortega's (2001:178) observation that explicit focus-on-form instruction, followed by explicit focus-on-formS instruction, is the most effective for L2 learners.We would suggest that incorporating more form-focused instruction into South African educational texts, and providing clearer guidelines in teacher support materials as to how negative feedback can assist learners in identifying errors in their interlanguage, will not only help to address the significant vacuum of English language input that occurs in the average South African classroom, but would also address the deficiencies, inconsistencies and errors that may occur in teachers' speech (cf.Spada and Lightbown 1993).
In our opinion, this shift in perspective on the part of textbook writers away from a seemingly ineffectual implicit and meaning-focused approach and towards a more explicit, form-focused approach, will go a long way towards enabling learners of English to know what forms they are expected to learn and how they should use them.

*
We wish to thank Frenette Southwood for her insightful comments and careful editing, and two anonymous reviewers for their helpful comments on a previous version of this paper.
of 40 minutes each), and there was a period of 40 minutes set aside for English reading each day.Many learners had obtained their foundational schooling at rural schools in the Eastern Cape before migrating to Cape Town.In general, learners displayed low levels of EAL development.The participants belonged to three classes, comprising 56, 58 and 57 learners respectively.Prior to May 2003, the school had only two Grade 5 classes, comprising more than 80 learners each; however, the appointment of a new teacher (specifically for English) in June of 2003 and the building of an extra classroom has greatly alleviated crowding and improved learning conditions significantly.
press), a Kenyan ESL textbook for Grade 5 learners, which is based on a form-focused, fairly explicit instructional approach.Although this textbook provides substantial FonF practice of interrogative structures, it engages participants in meaningful, communicative reading passages.The Explore English material included exercises and activities that emphasised both the form and function of yes/no questions such as Do you live in Cape Town?, prepositional object questions such as In which town do you live?, adverb questions such as How often do you watch television?, and direct object questions such as Which sport do you like best?Activities included the following: (i) ordering activities, where participants had to construct questions by correctly ordering scrambled words in a "word pool" based on a picture which provided them with contextual clues; (ii) problem-solving oral activities, where participants had to assume the identities of celebrities and then guess the identities of other celebrities by asking them questions such as Are you a man or a woman?or Do you appear on television?; and (iii) a class survey, where participants had to find out what the class's favourite sport was -this elicited questions such as What is your favourite sport?and Do you like soccer better than rugby?

Table 1 :
Accuracy of question formation ordering task

Table 2 :
Accuracy of question formation in preference task

Table 3 .
Accuracy of question formation in yes/no questions task Participants' performance in this task was poor overall.At pre-test stage, the lowest overall result came from the FonF group (26%), whereas the best pre-test performance came from the FonM group (32%).
. For example, in forming a question from a statement such as Martin shouted very loudly, learners often simply added a question mark, i.e.Martin shouted very loudly?One possible reason for this could be L1 transfer, as Xhosa makes use of neither subject-verb inversion nor fronted question words or auxiliaries in forming questions.This suggests that the learners had not yet reset the Xhosa (L1) parameters associated with question formation to the English (L2) value.Those learners who did use auxiliaries either failed to use the infinitive form of the main verb, resulting in constructions such as Did Thandi borrowed the textbook?and Did Martin shouted very loudly, or overused the auxiliary, for example Did Lona and Sipho were late for school this morning?Very few learners identified must as an auxiliary, and in sentences where it appeared opted for adding the verb do, resulting in constructions such as Does Aphiwe must win the running race?