Native language grammar knowledge awareness in children and the
role of explicit teaching
Michael Flint 10619119
MA thesis in the program of General Linguistics Faculty of Humanities, University of Amsterdam
June 2014
Contents
1. Introduction: The nature of child L1 knowledge 2. Background
2.1 Child L1 knowledge studies
2.2 The hierarchy of child language knowledge 2.3 The impact of formal schooling
3. Research questions and Hypotheses 4. Method
4.1 Subjects
4.2 Experimental Design 4.2.1 Stimuli 4.2.2 Procedure
4.3 Task justification and considerations 4.4 Scoring
5. Results
5.1 Overview
5.2 Error detection x construction type + proficiency 5.3 Error correction x construction type + proficiency 5.4 Explanation x construction type + proficiency 5.5 The effect of teaching
6. Discussion
6.1 Research Question 1 6.3 Research Question 1b 6.4 Research Question 1c 6.5 Research Question 2
6.6 Limitations and suggestions for further research 7. Conclusion
8. References Appendix
1. Introduction: The nature of child L1 knowledge
The nature of the process which allows us to acquire our first language (L1) has divided researchers with, arguably, more vigour and resulting separatism than any other area of linguistic intrigue. In spite of the divisive quality of this field, the fact that a child acquires their L1 with relatively little conscious awareness of its occurrence is a widely accepted facet of the acquisition process. In societies which can be broadly grouped according to having a formal education system which starts around the age of 5, children’s pre-schooling linguistic acquisition is predominantly characterised as largely unconscious, resulting in implicit knowledge of language. The gaining of this implicit knowledge, is said by many, to be the result of uninhibitable processes which detect and deduce probabilistic based regularities from frequency features in the input (Saffran, Aslin & Newport, 1996; Gomez & Gerken 2000; Rebuschat & Williams, 20012). The result of continued exposure to the L1, these frequency induced regularities form the basis of our language knowledge and subsequent proficiency at all levels of processing from perception to production. At this “pre-formal schooling” stage in language development, a child’s knowledge of their language is taken for granted to be inaccessible via conscious introspection and is thus considered implicit.
In the years between birth and the onset of formal schooling, a child is habitually assumed to have acquired the majority of the fundamental rudiments of their L1. The exact level of L1 proficiency is determined by the specifics of the child’s linguistic environment, including factors such as guardian education levels and exposure to input. Although the precise character of children’s environment undoubtedly leads to some degree of individual variation, a large amount of uniformity is supposed for all children approaching school age (Meisel, 2011), an assumption questioned by Hulstijn (forthcoming). As previously mentioned, up until this point, language
knowledge is assumed to be predominantly implicit in that the child is unaware of and unable to consciously access the rules and regularities upon which their emerging language capabilities are based. This is not to say, however, that children’s knowledge of a language has to be, or indeed is, implicitly represented for the remainder of their lives. As children’s cognition matures, their increasing ability to “think about
language (Myhill, Jones, Lines & Watson, 2012;143). This examination may result in the development of explicit or declarative knowledge as they consciously reflect on regularities that govern their language (Bialystok, 2001; Karmiloff Smith, 1995). Such introspection can be exemplified by an English speaking child who comes to consciously realise, via their own investigative thoughts, that in the case of most nouns an /s/, /z/ or /iz/ sound added to the noun means that more than one instance of the noun is being referred to. Proving that this knowledge is conscious is one of the biggest challenges that faces researchers.
Another way in which explicit language knowledge regarding the L1 can be gained is through explicit instruction. Previously implicit L1 language knowledge can be described in such a way by an educator that a child comes to acquire an explicit structural understanding of this knowledge (Bialystok, 2001). The acquisition of explicit knowledge via teaching can be exemplified as follows; A child who may already be proficiently applying a plural rule for nouns in their speech production and perception but who is unable to demonstrate any conscious knowledge of the rule, has the rule explained to them by their educator. This situation, if successful, results in the child acquiring explicit knowledge of the rule. The quantity and quality of explicit language teaching is contingent on government policy and teachers’ confidence in their own grammatical knowledge amongst other factors. With some form of literacy introduced in the early years of school, children may well be expected to move from a presumed purely implicit representation of language to one which combines both implicit and some explicit understanding (Hulstijn, forthcoming).
Bialystok (2001;125) sees explicit metalinguistic knowledge as an “optional embellishment”, distinct from, and additional to, basic linguistic competence. In principle, therefore, it is possible for explicit language awareness to follow a development trajectory distinct from general language ability, but in reality, the degree to which these two skills can be teased apart is, as yet, unclear. With general intelligence likely to impact aptitude in both abilities, the impression that the two are indeed linked may be created. In contrast to variation in general L1 ability in young children, the variation in metalinguistic ability is more often recognized. Bialystok (2001) especially emphasises the link between general academic capacity and metalinguistic ability suggesting that children who more readily acquire explicit
knowledge about their L1 might well also represent a broad grouping of more academically gifted pupils.
Investigations into the nature of the knowledge that children posses about their own language is a linguistic road which is not well trodden. Few studies have addressed the issue directly, with the majority of research into whether learners have a more implicit or explicit representation of language knowledge falling into the realm of second language (L2) research. This dearth in L1 type research is likely to be a result of the fact that L1 acquisition is seen, particularly by generative linguists, as a
universally successful process in children not affected by mental disabilities/disorders. This study aims to contribute to the slowly increasing knowledge concerning the extent to which children are explicitly aware of the grammar of their L1. Such a contribution hopes to be achieved as a result of testing children of varying ages and general academic proficiencies on the level of explicit grammatical knowledge they have regarding a number of grammatical constructions (on which they have received varying degrees of explicit instruction) in their native tounge.
It is worth clearing up definitional issues regarding the use of the terms implicit and explicit in order to ensure clarity henceforth. Commonly used definitions of the terms have been adopted as provided by (Hulstijn, forthcoming). Implicit knowledge here is defined as unconscious knowledge which is inaccessible via conscious introspection. It is exemplified in language via our ability to follow the rules and regularities of our L1 with great consistency and accuracy without knowledge of the defining properties of said rules and regularities or even the awareness that they exist. There must be some level of representation of these rules in order for us to adhere to them but the fact that we often, although not always, cannot consciously access the rules, makes the knowledge implicit. From here on implicit and unconscious will be used interchangeably. Explicit knowledge or metalinguistic awareness is defined as conscious knowledge that has either been taught or deduced by the learner through other explicit strategies and examination of linguistic evidence. Explicit language knowledge is more easily exemplified by a stereotypical L2 learner who is presented with a rule to follow as part of an instructional process and therefore has an explicit conscious awareness of the formula for correct language use in the given tongue.
Explicit knowledge will be interchangeably referred to as conscious knowledge and metalinguistic knowledge.
2. Background
2.1 Child L1 knowledge studies
A handful of studies have attempted to investigate the age of onset of explicit grammatical knowledge in children. These studies have varied greatly in their methodologies and subjects which has unsurprisingly resulted in varying results. Evidence that children have the ability to reason grammatical errors in their first language via explicit reference to linguistic cues has been claimed to have been found in subjects as young as 5-6 (Gleitman, Gleitman & Shipley, 1972; Hakes, 1980). It is worth mentioning though that Gleitman et al (1972) make it very clear that their results cannot be generalized to the population at large. Instead, they explicitly state that they made an effort to find particularly articulate children who were more likely to be unrepresentative of the average ability of their age group than vice versa.
Karmiloff-Smith (1986) claimed that by the age of 4, children are beginning to exhibit sensitivity to linguistic markers with respect to their own spontaneous speech repairs. This sensitivity to markers is only expanded on to encompass explicit metalinguistic judgments referencing linguistic cues by the age of 6 (Karmiloff-Smith, 1986). Again, a number of factors are worth noting in relation to the Karmiloff-Smith study. One is that the only construction students were tested on was determiners. Experimenters elicited spontaneous production and comprehension of different types of determiners in French native speaking subjects. Subsequently, participants were asked to verbally reflect on the procedures they had used via a “think out loud” protocol. While testing different types of determiners added a little to the variety of the explanations which were required by participants, the study does not really enlighten us as to how far reaching children’s applicability of this ability is across different grammatical constructions. It is conceivable, for example, that children of a certain age are
particularly good at explaining determiners but are unable to replicate such an ability for past tense or progressive verbs. In addition to the lack of variety in constructions, it is important to recognize that the study made use of a 5-stage scale of awareness against which participants’ responses were pitched. Only the two most advanced stages of this scale can be reached by making explicit reference to linguistic cues in
the examples, with the first 3 stages representing different levels of semantic
reasoning. The statement regarding the level of awareness beginning to be shown at the age of 6 refers to answers given which were ranked in the highest two levels of awareness. Importantly the results show that only some 6-year olds are beginning to exhibit the ability to reason errors with reference to linguistic features of the
ungrammatical phrase. The most common responses for this age group still rely on semantic reasoning without explicit reference to cues. An example of such a response is, in a situation where a specific referent rather than other possible candidates had to be specified, a child replied, “I chose the red book because it’s red”. In fact, the only age at which the majority of reasoning moves into the two stages which require explicit mention of linguistic cues is at the age of 10.
Clark’s (1978) developmental-literature review outlines three stages in the
development of children’s linguistic awareness. Around the age of 2/3 Clark claims that children develop some form of linguistic awareness in recognising the fact that they have progressed from early babbling to real words and grammatical
constructions. After this, Clark’s three stages commence with children’s spontaneous repairing of their own speech. This first stage is claimed to first be represented by phonetic corrections with lexical and grammatical corrections following not long after. The self-repairing stage is followed by the development of the confidence to correct others, with the final stage representing the ability to add explanation and justification to the grammaticality of utterances and their possible interpretations. Galambos & Goldin-Meadow (1990) used this three-stage model of development as the basis for their experimental design in examining the effect of bilingualism on children’s metalinguistic awareness. As this three-stage model supposedly represents increasing levels of awareness, its use by Galambos & Goldin-Meadow allowed for the incremental examination of the development that children go through with regards to their ability to consciously reflect on language and the type of knowledge they possess. Importantly, the three stages were adapted slightly by Galambos & Goldin-Meadow so that stages 1 and 2 were assimilated into a new stage 2 and a separate stage 1 was created in which the only skill necessary was detecting an error. Galambos & Goldin-Meadow (1990) did not find evidence in favour of their three stage division reflecting children’s development. The monolingual subjects tested (aged 4:5-8) showed similar ability in detecting and correcting errors. Offering correct
explanations as to why something was ungrammatical, on the other hand, was found to be a qualitatively different skill. The researchers also found that the ability to perform at different levels is linked to age, with the ability to correct increasing between the ages of 4/5-5/6 and the ability to give a grammar-oriented explanation increasing between the ages of 5/6-7/8.
2.2 The hierarchy of child language knowledge
Karmiloff-Smith (1986) and Marshall & Morton (1978) proposed that the initial two stages of Clarks observations, whereby children repair their own speech and
subsequently the speech of others, are not reliant on any conscious awareness but rather an unconscious system for detecting and correcting errors. The ability to correct makes use of the capability to hold in short term memory long enough, process and generate a correct variant of an ungrammatically formed utterance (Fowler, 1998) but is not necessarily reliant on conscious retrieval of rules. On account of the fact that stages 1 and 2 appear to involve essentially the same two skills but exhibited in scenarios which require alternative levels of confidence, Galambos &
Goldin-Meadow created a new stage 1 which only involves noticing the ungrammaticality of an utterance with a unifying noticing+correcting stage as the second level of
development. The new initial stage of Galambos & Goldin-Meadow represents a level at which, as Bialystok (2001) points out, the only prerequisite in children is that some representation of a language “rule” must be present.
The final stage of Clark’s model is left unaltered by Galambos & Goldin-Meadow. Successful demonstration of this stage of development requires the presence of explicit knowledge and or conscious introspection of language. After the ability to detect and correct an error, in order to explain it, a child must be able to access some concept of a rule or regularity of their language which they can use to articulate a justification for the ungrammaticality of the given utterance. Galambos and Goldin-Meadow (1990), like Bialystok (2001) and Karmiloff-Smith (1986) argue that this stage of explicit metalinguistic ability is not directly linked to the development of language capability but rather reflects a more general capacity to reason and explain. This skill is said to be nurtured independently of linguistic ability but shows up in a linguistic environment when it is employed by a child to explain language. An interesting point made by Galambos and Goldin-Meadow is that, according to their
results, younger children tend to give more content-based explanations for errors whereas older children are more likely to give grammar-oriented explanations. This distinction suggests that being able to give some form of explanation for
ungrammaticality is not necessarily as difficult a skill to acquire as the ability to give an accurate grammar-oriented explanation.
As indicated previously, progression to the point of accurate grammar-oriented explanations is not compulsory for all learners, reinforcing the view that we are dealing with a skill developing independently to language itself. While we would expect most learners to pass through stages 1 and 2, stage 3 clearly represents a different kind of challenge and one which is perhaps not so intuitively useful, especially with regards to L1 acquisition. As Bialystok (2001) recognizes, explicit metalinguistic ability is difficult to acquire for L1 learners for a number of reasons including the lack of transparency of language rules, the lack of conscious attention that needs to be paid to grammar in language use and the generally negative opinion associated with learning “grammar rules”.
Karmiloff-Smith (1986) and Galambos & Goldin-Meadow (1990) offer a somewhat vague account of the development into the third level of awareness. These researchers attribute the development of conscious access to previously implicitly represented language rules and regularities to the persistent redescription of a child’s internal language representation. According to this account, the isolated procedures which allow a child to correct speech must be restructured so that these isolated procedures form one system. Once the isolated procedures have been unified and redescribed they become accessible to the child via conscious introspection. Bialystok (2001) offers a similar line of reasoning with the assertion that the analysis of internal representations of structures allows children to develop exponentially complex and accessible forms of these structures. This analysis, turning formally implicit
representations into explicit ones, is said to be instigated by the explicit description of implicit knowledge.
Hulstijn (forthcoming) poses a number of questions to be addressed by researchers concerned with the implicit vs explicit knowledge distinction. He cites research by Dienes and Perner (1998) and Dienes (2012) which suggest that the object of learning
affects the learning process to the extent that the amount of exposure and or
distribution of learning episodes may warrant individual values per object. Galambos and Goldin-Meadow (1990) appear to have some initial evidence in favour of this view. To address the question of whether different grammatical constructions are represented equally by different knowledge types, they chose to test their participants (age 4;5 to 8;0 years old) on 15 different grammatical constructions. The 15
grammatical constructions were subsequently divided into three groups according to the proportion of errors noted, with an easy (Mass noun, word order, lack of article, number agreement (sing.n.), case), medium (adverb/verb agreement, reflexive pronoun, possessive marker, selection restriction, preposition) and hard (irregular verb, number agreement (coor.n.), pronoun agreement, adverb/adjective, comparative) group of constructions. An interesting observation that can be taken from the
Galambos and Goldin-Meadow (1990) study is that constructions in which errors were easier to detect were no more likely to be explained by children tested than constructions in which errors were difficult to note. More work should be done with another native English population of children in order to verify the construction ranking found by Galambos and Goldin-Meadow (1990) and also to test a wider variety of constructions.
Inspired, in part, by the work of Galambos and Goldin-Meadow (1990), the current study aims to offer more insight into the nature of child language knowledge, both for different age groups and different general academic abilities within the age groups in relation to different grammatical constructions.
2.3 The impact of formal schooling
Towards the end of their article, Galambos & Goldin-Meadow recognise the fact that their research does not allow for an examination of the possibility that schooling contributes to children’s metalinguistic awareness. With a rise in their subjects’ grammar-oriented explanations of ungrammaticalities around the age of formal schooling, it would seem a plausible explanation that explicit metalinguistic ability is nurtured in the classroom.
Van Lier (1998) suggested that metalinguistic knowledge does not support writing or any other form of language development. In strong opposition of this claim, the 2013
UK National English Primary Curriculum prescribes the explicit teaching of English grammar in a detailed grammar appendix. With the 2014 curriculum supposedly giving an even greater role to grammar teaching and the introduction of a mandatory ‘spelling, punctuation and grammar test’ for pupils leaving primary education (12 year olds), the opportunity to evaluate the impact of explicit grammar teaching on children is evident. The current curriculum features a series of tables which state, by academic-year group, which constructions should be taught to pupils along with the linguistic terminology to be introduced. The justification for the focus on grammar is outlined in the following statement:
Explicit knowledge of grammar is very important, as it gives us more conscious control and choice in our language. (Department for Education, English programmes of study: key stages 1 and 2, 2013;64)
This statement does not express a new view, in fact, arguably a relatively well-seasoned one which is seeing a revival in teaching practises, especially under the current UK government. Advocates of the value of grammar teaching in schools include Clark (2010) who sees its merit in introducing pupils to terminology which allows them a greater understanding of the connection between grammatical
constructions and their literal and symbolic meaning. The view that metacognition is a valuable skill for language was previously argued by Bereiter and Scardamalia (1982) who emphasised the value of having labels that make implicit knowledge accessible and by Englert, Raphael, and Anderson (1992) who also placed enormous emphasis on the value of metacognition in language development. Hudson (1990) justifies explicit grammar teaching with the claim that it is valuable in the acquisition of Standard English, increases students’ self-respect, improves analytical thinking and adds to pupils’ overall language knowledge. Finally, in line with the Department for Education’s justification, Carter (1990) refers to the decrease in grammatical teaching as detrimental to pupils’ ability to partake in conscious control and choice of their language; a deficiency which stunts their ability to use language with the same systematicity and precision.
Much of the research concerning the effect of explicit grammar teaching on language development has specifically looked at the connection between teaching and writing ability. This approach was typified by the review of international evidence on the effects of teaching grammar by the Evidence for Policy and Practice Information and
Co-ordinating Centre (EPPI-Centre) Review Group for English (2004). Interestingly, the EPPI-Centre review concluded against the value of explicit grammar teaching, stating that it found no good evidence in favour of the view that grammar teaching improved the writing ability of 5-16 year olds; for similar skepticism regarding the role of grammar teaching see Lockwood (2006) and Cajkler & Dymoke (2005). The EPPI review subsequently received a great deal of criticism. Myhill et al (2012;141) suggested that the EPPI review typified the flaws in the field with a research question which took an “over-simplified view of causal relationships between grammar
teaching and written composition”. Additionally Myhill et al (2012) point out that the review’s “clear conclusion” is based only on three studies, of which two are over 25 years old (Bateman and Zidonis 1966; Elley, Barham, & Wylie 1975) and none of which were carried out in the UK.
Myhill et al (2012) conducted an intervention study involving writing task oriented explicit grammar intervention lessons. Explicit grammar teaching was found to have a significantly positive effect on post-test writing development of 14 year olds.
Interestingly, baring in mind the popular view that metalinguistic ability is more readily acquired by children with a higher academic aptitude, the study indeed found that intervention had a more significant effect on pupils who were already able writers. With some caution, it could be suggested that Myhill et al’s results show that more able students have learnt from the explicit grammar teaching and subsequently implemented their new knowledge in their writing. Aside from the obviously tentative nature with which such a claim should be made, it is also worth noting that the
intervention used in the Myhill study consisted of embedded grammatical teaching which was specifically designed to be highly relevant to particular writing tasks. In addition, teachers were given free reign to adapt and alter the material. With regards to pedagogy, the nature of the intervention should alert us to the crucial role of the context in which grammar is taught. It would seem that if we can extract any pedagogical message from Myhill et al’s research it would be something along the very narrow lines of that the introduction of explicit grammatical terminology and structures in a context in which they are directly applicable as part of a lesson structure with clear goals for it’s use, might well have a positive impact on pupils of 14 years of age who are already above average in terms of their writing capability.
What is very difficult to know in such a study, however, is the extent to which students’ explicit knowledge of grammar is improved by the teaching and to what extent it was the specific grammar orientation of the lesson that caused the positive effect in subsequent writing ability. Some qualitative evidence was gathered by Myhill et al (2012) in the form of post-intervention interviews in which subjects were encouraged to talk about linguistic choices made in their own writing and by the author of a written text. Despite the apparent gains in writing ability by some pupils, with respect to metasyntactic knowledge, Myhill et al conclude that many of the students did not yet have a sufficient level of explicit syntactic knowledge to explain their syntactic understanding. It would therefore seem less likely that the gain in writing ability is a knock-on effect of increased conscious syntactic awareness.
Without controlling for all of the other factors which could differ between a control group lesson and a grammar oriented lesson, it is very difficult to prove that it is really the grammar orientation of the intervention which causes subsequent success. If, however, as has been claimed, and is explicitly stated in the current UK English curriculum, acquiring explicit knowledge about grammar has a knock-on effect for linguistic competency, we first need to know more about the success of explicit grammar teaching. It is only once we have really established that what is being taught to children is actually being taken in that we can then try to establish the truth in Van Lier’s (1998) concern that metalinguistic knowledge does not support writing or any other form of language development.
3. Research questions and Hypotheses
The current study has been shaped by previous research and assertions in the literature in a number of ways. The study was designed so that children of different general academic abilities aged 6-7 and 8-9 were tested on their ability to detect, correct and explain errors in stimuli representing a number of commonly occurring grammatical constructions of English. Constructions had variously been taught or not taught to the children according to their age group. It has therefore been designed in order to shed more light on a number of questions.
1. To what extent are children (aged 6-9) explicitly aware of their grammatical language knowledge?
a. Does the extent to which children are explicitly aware of the grammatical regularities of their language vary with age? b. Does the extent to which children are explicitly aware of the
grammatical regularities of their language vary with construction type? c. Is development of conscious access to language knowledge determined
by general academic ability?
2. Does explicit grammatical language teaching impact the extent to which children are explicitly aware of the grammatical regularities of their language?
A number of testable hypotheses result from these questions. Each hypothesis will be presented in relation to the research question it is associated with.
1. Children become more consciously aware of their grammatical knowledge with age. (Research question 1a)
For this hypothesis to be supported, the older children tested will have to show a greater ability to exhibit conscious awareness of grammatical knowledge.
2. Different grammatical constructions will vary with respect to the level of conscious access children have to the rules and regularities underlying them. (Research question 1b)
For this hypothesis to be supported, the data will have to show that there is a
difference in the level of conscious access children have to the rules and regularities of varying constructions.
3. Students’ general academic proficiency will show a positive correlation with their conscious access to grammatical knowledge. (Research question 1c) In order for this hypothesis to be supported we would expect to see children with higher academic proficiency performing more competently with regards to explaining their grammatical knowledge followed by average achievers and finally lower
achievers.
4. Explicit grammatical language teaching will result in more conscious awareness of the rules and regularities underlying language. (Research question 2)
Without a control group and a more tightly controlled intervention strategy, the design of the current study dictates that any evidence could only be considered indicative of support for the hypothesis. Subjects having a significantly greater ability to exhibit conscious access to the rules and regularities of constructions that have been explicitly taught over those which haven’t would be considered suggestive evidence in favour of the hypothesis.
4. Method 4.1 Subjects
Subjects recruited for the study were 24 primary-school aged monolingual British English speakers. Participants were recruited through a contact at the school which they attend and the study was officially deemed appropriate by the head teacher of the school. Consent to interview the children was given on the grounds that, as the experimenter, the author would interview children in the immediate vicinity of a member of staff and that no names of students would be taken. Participants were chosen from two age ranges within the primary-school year-group system. Twelve participants represented year 4 (ages 8-9) and the remaining twelve represented year 2 (ages 6-7). Both clusters of participants represented the male/female gender divide equally. All participants attended a Church of England primary school in Berkshire, England.
The motivation behind testing two separate age groups was to facilitate the
examination of any significant increase in explicit grammatical awareness over time (Research question 1a). As was discussed in the introduction, it is often said that children’s propensity for conscious introspection of language develops as a result of other developmental cognitive maturation. In testing two age groups we hope to be able to further examine such claims. An additional motivation in comparing two age groups was that, according to the English curriculum, there are a number of
grammatical constructs which should have been introduced explicitly to one age group but not to the other. The age group divide therefore lets us analyse what, if any, effect the prescribed grammar teaching has on pupils’ explicit grammatical awareness (Research question 2).
Different proficiencies within the age groups are also represented. Of each of the 12 participants in each age group, 3 were identified as higher achievers, 6 as average and a further 3 as lower achievers. Ranking was decided based on pupils’ preliminary grading on government set Standard Assessment Tests (SATs) which primary school children in England have to take in English, Maths and Science at the ages of 7 and 11. Grouping of children into proficiency groups via preliminary SAT results was carried out in order to help assess the claim that explicit metalinguistic knowledge is more readily acquired by children who already have a higher than average academic ability (Hulstijn, forthcomig; Bialystok, 2001)(Research question 1c). The decision to have half the group for each age band representing average ability with the lower and higher achievers representing 25% each was in order to more accurately represent the overall ability of the year groups. If, for example, an equal number of higher, average and lower ability pupils were selected for each age group, this would not accurately show the real distribution of academic ability which realistically comprises of a far greater number of averagely performing students.
4.2 Design of the study
Participants were each subjected to a 3 stage task. The task aimed not only to obtain a greater understanding of the extent of the subjects’ grammatical knowledge but also to probe the type of knowledge they have regarding the grammar of their L1.
4.2.1 Stimuli
Subjects were tested on their knowledge of 8 different regularly occurring
grammatical constructions in English (Plural nouns, past tense verbs, progressive verbs in the present tense, present perfect, possessive-s, mass nouns, prepositions and comparatives). The inclusion of 8 constructions was, in part, motivated by the desire to better understand whether the object of learning affects the type of knowledge acquired. Do, for example, pupils find it difficult to consciously conceptualise and explain past tense constructions as compared to plurals? The constructions were selected based on a combination of those explicitly mentioned in the Key Stage 1 (KS1 - ages 5-7) and Key Stage 2 (KS2 - 7-11) sections of the English curriculum and constructions not explicitly stated in the curriculum but which had previously been successfully tested on children of the same age by Galambos and Goldin-Meadow (1990). The reason for the selection of stimuli from the curriculums of the two
key-stage groups was to see whether constructions which have been explicitly taught, have been subsequently internalized as explicit knowledge (Research question 2). In testing constructions which are supposed to have been introduced at different ages (KS1= 3 constructions, KS2= 3 constructions) we can also examine the extent to which explicit instruction is needed. We might, for example, find that students already have an explicit understanding of a construction before they are taught it. On top of the 6 constructions which are mentioned in the curriculum, 2 constructions not prescribed to be taught by teachers were also included. The justification for their inclusion is in order to examine whether students from both year groups have a natural capacity for consciously accessing the rules and regularities of their language and subsequently articulating them without explicit instruction.
A slight risk was taken in the construction choice in that, although 6 of the constructions were chosen based on their inclusion in the curriculum, this did not totally ensure that they would be taught. This lack of control over what had and hadn’t been taught, in principle, further weakens the opportunity to provide support for hypothesis 4. The reason teachers were not asked which constructions they had taught was to avoid the scenario where they knew so much about the study that they were able to train students artificially to perform well in it. The majority of details surrounding the tasks were kept from the teachers for this reason. In order to
determine which construction had been taught by teachers, they were asked after the testing to indicate the extent to which the constructions tested had been introduced and discussed in class. Their feedback closely matched the curriculum but wasn’t an identical match. Out of the 8 constructions, 2 had not been taught to either year group, 3 had been taught to year 2 and year 4 and 3 had only been taught to year 4. This 3x3x2 combination was exactly what we had hoped for but ended up being distributed differently across the constructions to our expectation. For the exact construction to curriculum prescription to teacher feedback breakdown, see table 1 below.
For each of the 8 constructions, 4 stimuli were created. These stimuli were in the form of sentences representing the constructions. In each group of 4 stimuli there were 3 ungrammatical tokens and 1 grammatical one. Grammatical tokens were included in the test to ensure that participants did not hypothesise that all of the stimuli were ungrammatical and start detecting errors without properly paying attention to the task. Ungrammatical stimuli were only incorrect as a result of one feature change that
affected the grammaticality of the construction with regards to the target structure (e.g. in the target structure “past tense”, one past tense suffix in the sentence might be missing). All 36 stimulus sentences were recorded by a female native speaker of English and were presented to the participants via laptop speakers. The decision to present stimuli audibly was taken in order to diminish any effect of reading
proficiency. A list of all stimuli is given in Appendix 1.
Table 1. Construction type teaching information. Curriculum stated introduction year + teacher stated evidence of instruction for each year group
Construction type Year of
curriculum stated introduction Taught in year 2 (as evidenced by teacher feedback) Taught in year 4 (as evidenced by teacher feedback) Plural Noun Suffixes
Year 1 YES YES
Past Tense Suffix Year1/2 YES YES
Progressive in Present Tense
Year 2 NO YES
Present Perfect Year 3 NO NO
Possessive –‘s Year 4 YES YES
Mass Nouns No recognition NO NO
Prepositions Year 3 NO YES
Comparatives No recognition NO YES
4.2.2 Procedure
Participants were tested individually in one-on-one interviews with an experimenter (the author of this study). The interviews were conducted on a table outside the participants’ school classroom. The procedure followed three stages. For each stimulus presented, participants were asked a possible succession of three questions. On hearing a stimulus (which they were informed they were allowed to listen to multiple times), the subject was asked whether the sentence sounded ok to them or whether there was anything wrong with the English used. In this error detection stage the pupil was expected to respond with a “yes” or “no” answer. If an error was correctly detected, the experimenter would then proceed to the second question in asking the participant to correct the sentence (correction stage) and verbally reproduce it in its correct form. The final stage of the test required subjects, when prompted by the experimenter, to give an explanation as to why they had corrected the initial
sentence therefore explaining the ungrammaticality. All interviews were audio recorded and analysed post interview.
4.3 Task justification and considerations
It was important that the task ensured that grammaticality was really the factor being tested. Many previous studies have presented children with sentences which
semantically contrast their world knowledge (De Villiers & De Villiers, 1972; Hakes, 1980). These tests do not measure the grammatical knowledge of participants. By contrast, the stimuli used in the current study complied semantically with children’s world knowledge but instead featured grammatical irregularities based on the 8 target constructions. In judging a sentence ungrammatical we can infer that the child has recognized that a grammatical rule they must posses knowledge of at a certain level has been flouted. Such an approach is exemplified by Gleitman et al (1972) and Galambos & Goldin-Meadow (1990) whose stimuli violated various grammatical regularities such as number agreement (e.g. “*Many door are completely broken”) or possessive markers (e.g. “*The teacher coat is very dirty”).
The degree to which correcting is a different skill to error detection is expected to be shown in differentiating the skills in the experimental set up. If a participant corrects an error in a target-like manner, the presumption is that in addition to noticing a violation of a rule, the participant is also able to temporarily store the sentence, locate the object of correction and generate a grammatical alternative. The previous research has suggested that in spite of the presumed additional processing involved in
correcting an error, children do not appear to have any more difficulty in carrying out this ability. This study aims to further assess this claim.
The final question of the task is employed to probe a more metalinguistic level of awareness, namely the ability to explain why something is ungrammatical (“wrong”). This level of awareness is not so commonly tested in young children (Karmiloff-Smith, 1986; Gleitman et al., 1972; Galambos & Goldin-Meadow, 1990). The previous research which has been done, however, clearly shows that this skill is markedly different from the skills that are required to answer questions 1 and 2. It has also been shown to be the most difficult skill to acquire and perhaps the reserve of only the most academically gifted students (Hulstijn, forthcoming; Bialystok, 2001). The reason behind this stage’s inclusion in the test is that it is designed to test
participants’ explicit knowledge of the rules that guide their language use. So-called retrospective verbal reports have often been used to gauge the knowledge type of language learners with respect to the implicit vs explicit distinction. The basic premise of their use is that if a subject is able to verbalise any sort of appropriate language rule, they have explicit knowledge. If, on the other hand, they are unable to articulate a rule but perform above chance in a grammaticality or acceptability judgment task, their performance is presumed to be based on implicit knowledge. Many have put forward criticism of retrospective verbal reports including Berry and Dienes (1993) who claimed that a subjects’ failure to articulate any form of rule-based knowledge does not necessarily mean they do not possess some form of explicit knowledge. Dienes (2012) hypothesized that subjects not verbalizing their knowledge could either be a reflection of the lack of confidence they have in this knowledge type or of the fact that the type of knowledge is difficult for them to articulate, especially without the correct grammatical vocabulary. In support of this view comes evidence from Erdelyi and Becker (1974) who claimed that if you offer subjects multiple attempts to verbalize knowledge they may be more successful in doing so on subsequent attempts. This evidence highlights the problem with concluding that a subject has no explicit knowledge if they fail to exhibit it at the first and only opportunity to do so. Based on his review of tools to measure implicit and explicit knowledge, Rebuschat (2013) concludes that perhaps verbal reports might not be “sensitive enough” to really show us what is accessible via conscious access by subjects.
In spite of the concession that any tool, which relies on active subject-participation, is unlikely to be 100% successful in ascertaining what is accessible via a subject’s conscious retrieval, it can also be argued that verbal reports actually represent a very effective and appropriate gauge of explicit language knowledge. In giving participants the ability to try to explain their answers and consequently their knowledge, they are not restricted to having to recognize that they are using a rule as some subjective measures tools ensure. Additionally, as a researcher you do not expose yourself to the problem of participants’ self-criterion setting in confidence reports. Instead, with the correct balance of encouragement and creation of a comfortable setting for the subject, you allow them to explore their own reasoning verbally. This freedom to hypothesise may very well be a good indicator as to exactly the degree of explicit
knowledge subjects do have, whether it be rule based, moving towards the expression of some sort of regularity or based on a very definite rule with the metalinguistic terminology to express it. In summary, despite its detractors, retrospective verbal reports arguably offer a good deal of sensitivity to the individual subjects. In asking subjects to explain their knowledge it is much easier to gain a comprehensive
understanding of the extent of what they know. For the purposes of the current study, where the idea of rules in relation to language may only be preliminary or not
developed at all, asking students to try to verbalise their thoughts maybe be the simplest and most subject-comprehensible way of attaining the data desired. 4.4 Scoring
Scoring can be divided into 3 sections relating to each of the three questions in the task. For error detection, participants simply received a positive or negative mark relating to whether they did or did not detect an error in a given stimulus. Stimuli which were not ungrammatical also received a positive mark but were analysed in isolation from the ungrammatical stimuli.
In cases of detected errors, students were required to correct the stimuli. The scoring system for corrections consisted of 4 options, “no response” for instances when the subject failed to give a correction, “incorrect correction” when a subject attempted a correction but failed to give an alternative which was grammatical, “grammatical correction” for corrections which were grammatical paraphrases retaining the essence of the semantic content of the stimulus and “target correction” for stimuli which were corrected accurately according to the target structure. Originally there was a fifth option for semantic corrections but it was found to be unwarranted.
The final stage of the task, in which participants had to explain the ungrammaticality prompting their correction, was scored according to a 5-point scale. A “0” mark was given when no response at all was offered, or a response which was in no way
relevant. A “1” was awarded for responses which surmounted to any of the following three options, “because it doesn’t make sense”, “because it doesn’t sound right” or “because it’s not proper English”. A “2” was given in cases where the subject had made a good attempt at accurately describing a relevant rule or regularity but only in a semantic sense. An example of such a response is, when presented with the sentence “*Yesterday I cycle my bike to school”, the participant’s explanation for its
ungrammaticality was “because they’re not cycling anymore”. A “3” was assigned to answers which were correct and were rule-based/grammar oriented. Such responses are exemplified by “because there’s more than 1” as an explanation for a missing plural –s or “because it happened yesterday” for a missing past tense –ed suffix. The final possible grade was “4” and it was only awarded for a correct rule-based response which also made use of appropriate grammatical terminology. Examples of such explanations were typified by responses such as “You have to use possessive –s because it’s his” for a missing possessive –s or “ you have to add -ed because it’s in the past tense” for a missing past tense –ed.
5. Results 5.1 Overview
Table 2 provides the percentage of each year group’s errors detected, corrected and given a GO/RBE according to each construction type.
Table 2. Percentage of each year group’s errors detected, corrected and given a GO/RBE by construction type.
Year 2 Year 4 Constructio n Error Detectio n Correctio n GO/RB Explanatio n Error Detectio n Correctio n GO/RB Explanatio n Plural 94% 100% 59% 97% 100% 51% Past Tense 75% 85% 34% 86% 93% 37% Progressive Present Tense 97% 57% 12% 100% 61% 20% Present Perfect 56% 68% 0% 86% 76% 9% Possessive 86% 89% 28% 86% 82% 35% Mass Noun 97% 100% 22% 97% 100% 14% Preposition 61% 90% 39% 92% 100% 47% Comparativ e 97% 97% 16% 97% 88% 24%
There was a clear difference between year 2 and year 4 students’ ability to detect errors in the ungrammatical stimuli. Year 4 students detected 11% more errors than year 2 students. Of the 288 possible errors to be detected by each year group, year 2 students detected 236 (82%) and year 4 students detected 267 (93%). Year 4 pupils individual results ranged from 83% of errors detected to 100% therefore performing relatively proficiently, as would be expected of native speakers. Arguably it is slightly surprising that year 2 pupils, whose individual results ranged from 63%-96%, missed on average around 1 in 5 errors. This lack of ability, however, is not in contrast to previous research which showed an even poorer error detection ability for this age group (Galambos & Goldin-Meadow, 1990).
Of those errors accurately detected, the ability of each year group to correct them was not nearly as contrasting. Year 4 students gave grammatical corrections for 257 of the 267 errors they detected, representing a 96% conversion percentage. Year 2 students managed a similar percentage, 94%, correcting 223 of the 236 errors they detected. Of the grammatical corrections, 87% of year 4 corrections were target corrections with a range of 72%-100%, as were 88% of year 2’s with a range of 60%-100%. As was found in the Galambos and Goldin-Meadow study, it would appear then that both year groups are better at correcting an error they have detected than detecting an error in the first place. In combining the results for each year group, after the first two questions we can say that year 2 pupils were able to detect and give a grammatical correction for 77% of stimuli whereas year 4 were able to do so for 89% of stimuli. Representing a difference of 12%, the real difference between the two age groups is in their ability to detect errors, not correct them.
Of the stimuli which had been correctly detected, year 2 and 4 children were able to give a grammar oriented/rule-based explanation (GO/RBE) as to their
ungrammaticality for 24% and 26% of stimuli respectively. Students from both year groups showed a slightly greater ability to give a GO/RBE for errors which they had not only detected but also grammatically corrected. Year 2 students gave such an explanation for detected and corrected stimuli 27% of the time and year 4 students were able to do so 30% of the time. Similarly to their ability to correct errors, year 4 students only showed a small advantage in their ability to give a GO/RBE over year 2
students. What is clear though is that the ability to give such an explanation represents a markedly more difficult ability than detecting or correcting an error.
5.2 Error detection x construction type + proficiency
Table 3. Number of errors detected over possible errors to detect + percentage of possible errors detected by year group. Year 2 Year 4 Construction Rank No. Detected Errors Percentage Construction Rank No. Detected Errors Percentage Progressive Present Tense 35/36 97% Progressive Present Tense 36/36 100%
Mass Nouns 35/36 97% Mass Nouns 35/36 97%
Comparatives 34/36 97% Comparative
s
35/36 97%
Plural –s 34/36 94% Plural -s 35/36 97%
Possessive -s 31/36 86% Prepositions 33/36 92%
Past Tense 27/36 75% Possessive -s 31/36 86%
Prepositions 22/36 61% Past Tense 31/36 86%
Present Perfect
20/36 56% Present
Perfect
31/36 86%
In order to analyse why year 2 pupils detected 11% less errors than year 4 pupils, each of the 8 constructions tested was examined to see whether there were any particular structures which posed more problems for the younger age group (overview in table 3). This allows us to see the relative ease of error detection for each construction. For each age group there was a total of 36 ungrammatical stimuli representing each construction. Year 4 error detection was very high for plurals, progressives, mass nouns and comparatives with percentages ranging from 97-100%. Detection of errors regarding past tense, present perfect, possessives and prepositions was slightly lower ranging from 86-92%. Similarly to year 4 pupils, year 2s also performed most proficiently on plurals, progressives, mass nouns and comparatives with percentage rates ranging from 94-97%. Also similarly then was their less proficient error
though is that, while year 4 pupils were less able to detect errors in these
constructions, their percentage of detections was not drastically different from their top 4. Year two students, on the other hand, scored very poorly on present perfect (56%) and prepositions (61%), with past tense (75%) quite a bit better and possessives (86%) just about representing the average. It would appear then that year 2 pupils’ significant struggle detecting errors in present perfect constructions and constructions with an incorrect preposition might be held accountable for their deficiency in error detection as compared to year 4s.
Another possible source of the variance between the two year group’s error detection capability could be varying relationships between the proficiency bands, as
determined by the subjects’ SAT results, within the age groups. For year 4 students, higher achievers detected 99% of errors with their year 2 counterparts detecting 88%. The difference between these two proficiency groups is therefore roughly in line with the overall contrast between the age groups. The difference between average year 4 and 2 achievers also didn’t divergence much from the overall average at 94% and 81% respectively. In terms of lower achievers, the year 2 pupils’ error detection percentage was actually a little closer to their year 4 counterparts’ than in the overall contrast, 79% compared to 85%. There is not nearly as much variation when
comparing the proficiency groups across age bands in relation to the overall contrast than in comparing construction type. Each age group’s varying ability to detect errors in different construction types then appears to be the biggest influencing factor between the two groups, especially in the three construction types that each year group jointly struggled with the most. It is also worth mentioning that there is a bigger range between the error detection percentages for higher and lower achievers in year 4 (14%) than in year 2 (9%). This distribution of the range across the two proficiency gaps, higher - average and average - lower, is far more pronounced for the difference between average and lower ability students in year 4 with a difference of 9% as compared to 2% in year 2. This difference shows that the year 4 average is brought down by their lower achievers to a greater extent than the year 2 average is.
On the surface, the ability to correct an error does not appear to be more difficult than the ability to detect an error. As stated in the “Results Overview”, on average, both year groups were actually able to correct a higher percentage of errors than they were able to detect (year 4=96%, year 2=94%). However, when the averages are broken down by proficiency groups for each year group, an interesting similarity emerges. For both year groups, higher and average pupils performed similarly proficiently but lower achievers in each year group were a good deal less proficient. In year 4, higher and average pupils corrected 99% of errors but lower achievers only corrected 89%. In year 2, higher and average pupils corrected 98% and 97% of errors with lower achievers managing 84%. Correction then seems only to be a more difficult skill for pupils with of lower general academic ability.
In contrast to just looking at the data for grammatical corrections, examining the percentage of target corrections allows us a better understanding of how confident pupils are in their ability to make use of the 8 constructions tested (overview in table 4). Interestingly, the constructions group differently for target correction percentages
Table 4. Number of target correction/detected errors (No. TC/DE) + Percentage of detected errors which received a target correction (Percentage of TC/DE) by year group
Year 2 Year 4 Constructio n Rank No. TC/DE Percentage of TC/DE Constructio n rank No. TC/DE Percentage of TC/DE Plural 34/34 100% Plural 35/35 100%
Mass Noun 35/35 100% Preposition 32/32 100%
Comparativ e
32/33 97% Mass Noun 35/35 100%
Preposition 18/20 90% Past Tense 27/29 93%
Possessive 25/28 89% Comparativ
e
29/33 88%
Past Tense 23/27 85% Possessive 23/28 82%
Pres Perfect 13/19 68% Pres Perfect 23/30 76%
as compared to error detection. For year 4 pupils, plurals, past tense, mass noun, prepositions and comparatives range between 88-100% for target corrections. By contrast, progressives, present perfect and possessives range between 61% and 82%. Correcting constructions which should have a progressive verb seems to be
particularly unnatural for year 4 students with most student’s correction technique involving opting to change the construction into the past tense. Year 2 pupils were also less confident correcting progressives and present perfect constructions (57%, 68%) but not so with possessives (89%).
It is important not to become too focused on the 3 proficiency groups but rather to take individual variation into account. It is worth noting, for example, that in both year 2 and 4 average achiever groups there were a number of pupils who performed in a similar manner to the lower achievers with regards to target corrections. The overall averages of the average achiever groups are brought up by them having more
participants than the higher and lower groups. The fact that this small number of “average achievers” performed more like “lower achievers” in terms of their ability to give target corrections highlights the pitfalls of grouping children according to “general academic ability”. Individual assessment is always more enlightening and shows that small sub-skills of grammatical proficiency vary between students at all levels.
5.4 Explanation x construction type + proficiency
As previously mentioned, both year 2 and 4 pupils showed themselves to be notably worse at explaining errors than detecting or correcting them. Year 4 pupils could only give an accurate GO/RBE 29% of the time and year 2 pupils could only do so 26% of the time. This is strong evidence in favour of the view that explaining an error is a distinctly separate skill from error detection and correction. But, are some
constructions easier to explain errors in than others and are some general academic proficiency groups better than others at doing so than others?
With regards to the 8 constructions, for both year 2 and year 4 pupils, of all the GO/RBEs of target corrected errors, plurals represented the highest proportion, 35% and 26% of the total. Target corrected plurals also represent the construction which received the highest percentage of grammar oriented/rule based explanations. Year 4 students gave such an explanation to 51% of target corrected plural errors and year 2 pupils did so for 59%. Additionally, both year groups also found prepositions, past tense and possessive constructions the next 3 easiest constructions to give a GO/RBE for, in that order. Both year groups also shared present perfect constructions as the most difficult to give a GO/RBE for. Year 2 pupils didn’t offer any GO/RBEs for present perfect constructions and year 4 pupils gave such an explanation for only 9% of target corrected errors.
Out of all the constructions, of particular interest is the mass noun category which received a higher percentage of GO/RBEs and subsequent ranking from year 2 pupils than year 4s. Year 2 pupils gave GO/RBEs for 22% of target corrected errors
compared to year 4’s 14%. It is tempting to overstate the significance of this finding. When the numbers are examined, year 2 and 4 pupils both gave 35 target corrections for mass noun constructions. Of these 35 target corrections, year 2 pupils gave 8
Table 5. Percentage of target corrected errors with a grammar oriented/rule-based explanation (Percentage of TC with a GO/RBE) + Distributional percentage of total grammar oriented/rule-based explanations (Precentage of total GO/RBEs) by year group
Year 2 Year 4 Constructio n Rank Percentage of TC with a GO/RBE Percentage of total GO/RBEs Construction rank Percentage of TC with a GO/RBE Percentage of total GO/RBEs Plural 59% 35% Plural 51% 26% Preposition 39% 12% Preposition 47% 22%
Past Tense 34% 14% Past Tense 37% 14%
Possessive 28% 14% Possessive 35% 12%
Mass Noun 22% 12% Comparative 24% 10%
Comparativ e
16% 9% Progressive 20% 6%
Progressive 12% 4% Mass Noun 14% 7%
GO/RBEs and year 4 pupils gave 5. A difference of 3 explanations should not be taken as evidence that year 2 pupils are, on average, much better at explaining mass nouns than year 4 pupils. The most important fact remains, at 22% and 14%, neither year group are particularly good.
In comparing the GO/RBE results to the target correction results we can also observe that there is not a strong correlation between constructions which are easy to correct and those which are easy to explain as was also found by Galambos and Goldin Meadow (1990). This lack of correlation can be exemplified by constructions featuring mass nouns which received the most or joint most amount of target correction for both age groups but ranked 4th from bottom for year 2 pupils and 2nd from bottom for year 4s in terms of explanations. A similar story is true for comparative and possessive constructions. The same cannot be said, however, for plurals which were corrected with high accuracy and were also the easiest to explain. In addition, present perfect constructions had a very low count for corrections in both age groups and were also the most difficult to explain.
When the GO/RBE results are broken down according to each proficiency group for the two age groups, the average ability group unsurprisingly comes out representing the highest number of GO/RBEs because the group is twice a big as the other two. In order to really understand which proficiency group is the most able to explain errors it is necessary to examine the percentage of target corrected stimuli which received a GO/RBE. For both year 2 and 4, the higher achievers are better at giving a GO/RBE than any other proficiency group, explaining 42% and 45% of errors they had
corrected in a target like manner. The average achievers were the second most able to do so for both age groups, giving a GO/RBE 27% of the time in the case of the year 2s and 31% in the case of the year 4s. Subsequently the lower achievers performed the most poorly with year 2 lower achievers offering GO/RBEs for 9% of target corrected stimuli and year 4 lower achievers doing so for 17%. Once again, these results
exemplify how far behind the rest of their peers the lower achievers are. Interestingly though is that there is more of a difference between the year 2 and year 4 lower achievers than between any of the other aged matched proficiency groups. 5.5 The effect of teaching
Below, table 6 shows which constructions were and weren’t taught to pupils in each age group and their relative ability to explain errors in these constructions.
Table 6. Teacher feedback for constructions taught to each year group + percentage of target corrections with a grammar oriented/rule-based explanation.
Construction type Taught in year 2 Percentage of TC with a GO/RBE Taught in year 4 Percentage of TC with a GO/RBE Plural Noun Suffixes YES 59 YES 51 Past Tense Suffix YES 34 YES 37 Progressive in Present Tense NO 12 YES 20 Present Perfect NO 0 NO 9
Possessive –‘s YES 28 YES 35
Mass Nouns NO 22 NO 14
Prepositions NO 39 YES 47
Comparatives NO 16 YES 24
In terms of the 3 constructions which had not been taught to year 2 but had to year 4, in each case year 4 children were able to give 8% more GO/RBEs of target corrected errors. In terms of real numbers, once totaled combining all three constructions, year 4 pupils gave 12 more (26 compared to 14) GO/RBEs than year 2 pupils for
constructions they had exclusively been taught. When examining the constructions which had been taught to both year groups the gap between their ability to give a GO/RBE almost diminishes to 0.7 percent. The two constructions which were taught to neither group resulted in differing outcomes. Present perfect constructions were taught to neither group which is perhaps reflected in both groups’ lack of ability to explain the constructions (Year 2 GO/RBE= 0%, Year 4 GO/RBE= 9%). Mass nouns not being taught to year 4 pupils appears to correlate with the fact that they were explained the second lowest percentage of the time out of all constructions (14%). Year 2 pupils, on the other hand, gave a GO/RBE for 22% of mass noun errors they had target corrected, which only represented their 4th worst conversion rate.
Combined in GO/RBEs are explanations which did and did not make explicit use of grammatical terminology. Based on the presumption that a GO/RBE without
terminology is possible without the influence of explicit teaching but that one with terminology is not, table 7, below, shows the number of GO/RBE which did make use of grammatical terminology.
Table 7. Teacher feedback for constructions taught to each year group + number of target corrections (TC) with a grammar oriented/rule-based explanation which used explicit terminology.
Construction type Taught in year 2 Number “4” categorized explanations over TC. Taught in year 4 Number “4” categorized explanations over TC. Plural Noun Suffixes YES 0/34 YES 6/35 Past Tense Suffix YES 3/23 YES 10/27 Progressive in Present Tense NO 0/17 YES 0/20 Present Perfect NO 0/13 NO 0/23
Possessive –‘s YES 0/25 YES 0/23
Mass Nouns NO 0/35 NO 0/35
Prepositions NO 0/18 YES 0/32
Comparatives NO 0/32 YES 3/29
The table shows that year 2 pupils only gave 3 explanations using terminology, both for past tense constructions. Closer inspection of the data shows that these 3
explanations were from only 2 pupils. Year 4 pupils gave a good deal more explanations using terminology especially for past tense constructions and to a slightly lesser extent for plurals. Neither year group gave any GO/RBE with
terminology for constructions they hadn’t been taught but both also failed to give any for some constructions they had been taught.
6.1 Research Question 1 - To what extent are children (aged 6-9) aware of their grammatical language knowledge?
It was proposed to investigate the extent to which children (aged 6-9) are aware of their grammatical native language knowledge. The results from both year groups suggest that at this age, the knowledge that children have does not extend a great degree further than the presumed implicit knowledge necessary in order to detect and give and accurate correction to a grammatical error. Their awareness of the rule-based characteristics of this knowledge is far smaller. As has been suggested by Bialystok (2001), the fact that children have little conscious awareness of the rules that underlie their language proficiency is likely to be because this sort of access is largely
unnecessary for communication. Without a motivating factor for such access, children’s ability to achieve it appears to be greatly diminished. This is not to say, however, that conscious access to grammatical knowledge is unachievable at a young age. In both year 2 and year 4 pupils we found plenty of examples of grammar-based understanding of various constructions. This finding importantly shows us that some children within this age range are perfectly capable of understanding and retaining rule-based knowledge or independently hypothesising via conscious introspection regarding the constructions tested.
6.2 Research Question 1a - Does the extent to which children are explicitly aware of the grammatical regularities of their language vary with age?
Evidence was not found in favour of a considerable change in subjects’ ability to express explicit knowledge between the ages of 6/7 and 8/9. The main difference between the two age groups was, in fact, the ability to detect an error, which increased by 11% from year 2 to year 4. With regards to conscious grammar knowledge, both year groups were able to give a GO/RBE for a very similar amount of corrected errors. This result is perhaps a little surprising because, after two more years of schooling, year 4 pupils’ appeared not to have improved in their ability to explain a grammatical error using a grammar oriented explanation. This is also in spite of the fact that the evidence from their teachers suggests the year 4 pupils had been exposed to grammar oriented teaching for a number of constructions which the year 2 pupils had not and had also received additional class-time devoted to constructions they had
