This page presents text recovered from a no-longer supported file format.
Chapter 4 of Understanding Chemical Bonding: The development of A level students' understanding of the concept of chemical bonding
Justifying Methodology
§4.0: The purpose and structure of the chapter.
In educational research, unlike mature branches of the natural sciences, there is not a consensus on the research approaches and techniques that lead to acceptable work. There is no single 'disciplinary matrix' of the type described by Kuhn (1977, p. 297) that would make 'educational researchers' a unified 'scientific community' in his sense (cf. Gilbert and Swift's 1985 description of the constructivist program: "terminology has not been agreed upon, a common methodology not shared, ultimate aims not stated", p.682). Therefore it is appropriate that I should set out my own position as a researcher to demonstrate that:
1) I am using accepted research methodology;
2) my specific data collection and analysis techniques are congruent, and are consistent with my stated methodological position;
3) my methodology is consistent with my conceptualisation of the research field presented in the preceding chapters.
The present chapter acts therefore as a bridge between the introductory chapters, which outline my research focus and locate this in a field of enquiry, and chapter 5 which provides technical details regarding the techniques used in the research.
First the notion of disparate research paradigms, or traditions, in educational research is considered (§4.1). The significance of locating the present research within an accepted methodological position is considered in terms of problematic aspects of the research process (the origin of the researcher's conjectures, and relationship between research findings and conclusions, §4.2). The study was undertaken with a specific ethical stance, and this is explained (§4.3). The present study is then described as developing grounded theory (§4.4). The characteristics of grounded theory are presented and related to the present research. The selection of data collection techniques used in the study is discussed (§4.5 – §4.9). Finally some specific issues of authenticity and generalisability are considered (§4.10).
§4.1: Research paradigms in educational research.
"The documentation of students' scientific conceptions and the way these progress is a field of work that has its roots in the ethnographic tradition with its recognition of the centrality of personal meaning and of individual and cultural differences. Yet despite this orientation, there appears to be strong messages about apparent commonalities in students' conceptions that may have implications for future directions of work in this field."
Driver, 1989, p.488, emphasis added
Mortimore points out that educational researchers usually have training and backgrounds in other fields (1991, p.210). Research in education takes a number of forms, which are described variously using such terms as new paradigm (Reason and Rowan, 1981), post-modernist (Jennings, 1994), critical (Carr and Kemmis, 1986), feminist (Griffiths, 1995) etc. However, probably the most well recognised division is between those studies which seek general statistically valid conclusions about some average epistemic subject – and which seek to negate the effects of individual differences and idiosyncrasies – and those studies which deliberately focus on understanding the individual. This distinction is between two clusters of approaches rather than a sharp dichotomy. So Cohen and Manion form an analysis focussed on the subjective–objective dimension (1989, p.9), where work at the objectivist pole follows the positivist model borrowed from the physical sciences. This positivist approach assumes that the aims, concept, methods and model of explanation employed in the natural sciences may be applied unproblematically (Carr and Kemmis, 1986, p.62; Walford, 1991b, p.2). Carr and Kemmis refer to the natural scientific view compared to the interpretive view. Hitchcock and Hughes (1989) refer to positivistic and interpretative researchers, and refer to Spradley's analogy with petroleum engineers and explorers (pp.15-16). Gilbert and Watts (1983) refer to the former cluster as paradigm 1, or the erklären tradition ("in which explanation is the goal", p.64) and associate it with such descriptors as 'scientific', 'experimental', and 'traditional'. By contrast the verstehen tradition ("in which understanding is the goal", p.64) forms the basis of their paradigm 2 which is described in such terms as 'holistic' and 'naturalistic'. P aradigm 1 research is nomothetic – concerned with general laws – and is commonly associated with quantitative research methodology, where paradigm 2 research is idiographic – concerned with the individual case – and is usually associated with qualitative methods. Reynolds has described this distinction as "an intellectual either/or situation where the two oppositional groups used only one method each, a method which in both cases was supported and buttressed by a supporting ideology about the nature of social science knowledge" (1991, p.194). Hammersley claims that this can lead to "clashes among researchers with different purposes who tend to see the others as engaged in the same enterprise as themselves, but simply doing it badly" (1993b, p.xix).
§4.1.1: The nature of the present research
The purpose of my study was to investigate the development of understanding, and therefore it was necessary to work with the same individuals at different stages in their courses so that I could observe any changes in their thinking (§1.8). In order to understand how students relate their knowledge about chemical bonding it was necessary to use sequences of questions that went beyond finding-out which diagrams were considered to include bonds, and which categories of chemical bond were used, but to ask 'why' each response was given until a detailed picture of the colearners' thinking – what has been described in chapter 2 as intuitive theories or alternative frameworks – emerged. As Driver points out "in order to investigate such alternative frameworks, pupils' thinking has to be probed in some detail; it is the reasons pupils give for their answers, not the answers themselves, which are important" (1983, p.26). This requirement for detailed attention to the nuances of data from individual learners locates the research in the idiographic and interpretative tradition. Indeed, Pope and Denicolo have suggested "that the very choice of intuitive theories as a focus of investigation represents an epistemological stance consistent with the qualitative-interpretative approach" (1986, p.154). The work could also be described as 'clinical', and following the interview approach used by Piaget (see §2.2.1).
Chemical bonding is an abstract topic that can be understood to varying extents, and which relies on a range of prerequisite knowledge (see the earlier discussions in chapter 1, §1.7.1, and in chapter 3). It is therefore reasonable to conjecture that a learner's ideas about chemical bonding could be confused and multifaceted. In chapter 2 a case was made for accepting that learners could have multiple frameworks for what scientists may consider a unified topic area (§2.9). In order to discuss changes in student thinking it was therefore necessary to work with students intensely so that as much as possible of the complexity and nuances of their ideas could be revealed. In effect, a case study approach was required. By definition, a case study is "the examination of an instance in action" (Walker, 1993, p.165) and is said to involve "some commitment to the study and portrayal of the idiosyncratic and the particular as legitimate in themselves" (p.166). However, it was my intention to identify any "apparent commonalities in students' conceptions" (to borrow Driver's phrase from the motto above), to attempt to devise a model of developing student understanding of chemical bonding that might have some more general applicability. (Thus the grounded theory approach, discussed below, §4.4).
Carr and Kemmis suggest that participants in what they label 'critical educational science' (§4.1.2) should be the teachers and learners themselves (1986, p.158). The present study may be seen to fit this criterion. I have undertaken research in my own institution, working with some of my own students as colearners (§4.3.2).
My research focus derives from a concern that as a teacher I was not helping students to develop their understanding of a curriculum topic as effectively as I would wish. My research programme then was
- to find about more about how students' learning in this topic takes place (and therefore to understand aspects of student thinking at the start of the course, and how this may change over time);
- to be able to diagnose impediments to desired learning;
- and to inform the development of strategies to improve my teaching and students' learning.
This programme may be seen as having an action research flavour. This present thesis is primarily concerned with the first part of this agenda, with constructing a model of developing student understanding of chemical bonding. In the final chapter, chapter 12, I consider the extent to which this model contributes to the programme I have set out. Action research is not tied to a specific methodology: rather it is characterised by "integrating various methods in a methodologically consistent strategy" (Altrichter, 1993, p.40). In order to develop my model, I have applied principles of 'grounded theory' (see §4.4). This has enabled me to utilise research techniques that are established and accepted within the field of constructivism in science education, and which have been considered to relate to the idiographic research tradition, yet to work towards a model of wide applicability. §4.1.2: The case for critical action research. Carr and Kemmis have questioned the dichotomy between the two traditions (erklären and verstehen) discussed above: what they describe as the apparent assumption held by "those on both sides of this intellectual divide [who] adhere to a conception of science which ensures that scientific explanation and interpretative understanding are mutually exclusive categories" (1986, p.105; see also Atkinson and Delamont, 1993, p.214; Delamont and Hamilton, 1993, p.26, p.36; Hammersley, 1993c, p.47; May, 1993, p.26; Walford, 1991b, p.2). They argue that as education is a practical activity (rather than the seeking of knowledge for its own sake as in 'pure' science), then "educational research cannot be defined by reference to the aims appropriate to research activities concerned to resolve theoretical problems, but, instead, must operate within the framework of practical ends in terms of which educational activities are conducted" (Carr and Kemmis, 1986, p.108). According to this view, educational research is undertaken to solve an educational (i.e. practical) problem – that is the mismatch between the practitioner's theory and practice (p. 112). So the purpose of educational research is seen to be to produce theory that is grounded in the educational practice (p.122).
Carr and Kemmis suggest a model of educational research based on 'criticaltheory' (pp.133), founded on Habermas's rejection of the notion that knowledge is the product of disinterested intellectual activity (p.134). Rather, this view considers that knowledge arises from the interests and needs of individuals in a particular sociohistorical context (p.134). Giddens (1985) has suggested that critical theory offers an alternative paradigm to those traditionally used in the empirical-analytical sciences and the historical-hermeneutic disciplines (i.e. paradigm 1 and paradigm 2 respectively).
Car and Kemmis argue that the interpretive perspective is useful but does not fully recognise the inherent limitations of any specific research context (1986, p.135). The critical approach attempts to overcome this disadvantage by making these contingent conditions explicit (p.137), allowing what Habermas called 'ideal-speech situations' to develop (May, 1993, p.28). Carr and Kemmis apply this approach to the educational context by identifying a new role for educational researchers, such that the research activity is recognised – and justified – as a social and political act (p.152). From this view the participants in 'critical educational science' should be the teachers and learners themselves (p.158). Carr and Kemmis conclude their analysis by advocating action research as a suitable basis for a critical educational science. Stenhouse goes further than this, and proposes the ideal of "an educational science in which each classroom is a laboratory, each teacher a member of the scientific community" (1993, p.222).
Action research arises from the practitioner's professional concerns, rather than from existing established theory (Hustler et al., 1986, p.3), and – like critical theory (May, 1993, p.28) – aims to improve practice and the understanding of practice (Carr and Kemmis, 1986, p.165; Elliott, 1991, p.49), rather than to develop abstract theory. Thus Carr and Kemmis' reference to grounded theory (discussed further below, §4.4) where the "relevant concepts, hypotheses and problems must be inductively developed from the 'raw data' provided by a study of the substantive area" (p.125, my italics).
§4.2: The process of induction
As this present research has been carried out from such a perspective, and as the notion of induction is seen as problematic, it is important to examine how hypotheses are inductively developed from the raw data. My discussion will take the following form:
1. that the emergence of hypotheses is not well understood, but rather occurs at a sub-conscious level that is not open to rational evaluation (also see chapter 2, figure 2.3);
2. that in natural science this is not seen as problematic as there are algorithms for testing hypotheses, so that theory should be solely judged in terms of its match with the results of reproducible controlled empirical investigations;
3. that as in educational research of the type reported here the methods most associated with natural science are neither available nor appropriate (see above, §4.1.2), then the nature of the researcher's intuitions are of more concern;
4. so it is important to be aware of the possible sources of bias that may channel the researcher's thinking,
5. and a methodology is required that ensures that hypotheses may be authenticated in terms of the data collected.
§4.2.1: The problems of induction
In the positivist model of science derived from Bacon, an observer notices a pattern inherent in the data, and then sets about testing the hypothesised pattern by making a systematic and controlled set of observations, that is, to "interrogate nature to tabulate both the circumstances under which a phenomenon is present and also those under which it is absent" (Bynum et al., 1981, p.203). There are two aspects of this method which may be considered problematic: the process by which patterns are spotted, and the process by which they are validated against nature. The latter aspect is sometimes termed 'the problem of induction', and is considered below (§4.2.2). In case study work a researcher is dealing with the singular case and the problem of induction (which concerns general laws) does not apply (§4.2.3). However, it has been suggested that even in case study work the researcher may be considered to be looking for patterns (Freakely, 1996, pp.230-231), so the former process remains important.
Pattern spotting concerns a process that may also be labelled induction: how specific conjectures are induced in the researcher's mind. That is, the process by which patterns are recognised and categories or concepts initially emerge during analysis. This is a creative act of intuition, or imagination, which can not be justified logically, and probably depends on subconscious processes: Koestler suggests that "the particular type of mental activity which takes place in the so- called 'period of incubation' [prior to the awareness of an original insight] does not meet the criteria of articulateness and logical decency required for admission into the focal awareness of the wide-awake state" (1982 {1967} , p.361). Kuhn (1970 {1962} ) has pointed out how the development of science often depends on the recognition of the significance of some anomaly, i.e., the interpretation of some datum as not being adequately explained within the existing theory. This is also an act of imagination, that takes the individual beyond the communal view "that blinds him towards truths which, once perceived by a seer, become so heartbreakingly obvious" as Koestler expressed it (1959, p.10).
There are accounts of this moment of inspiration in science, perhaps the most famous being Kekulé's description of the benzene ring structure, although – ironically – it has been suggested that Kekulé invented his stories as part of a strategy to claim priority for the structure (Noe and Bader, 1993; Bader, 1996, c.f. §2.8.2). Glaser and Strauss suggest that "everyone knows" how such insights can occur at any time, during any activity, and may 'dawn' suddenly or slowly (1967, p. 251). Barbara McClintock, the Nobel prize winning geneticist, has described to her biographer how much of her scientific work depended on a kind of subconscious thinking that she labelled 'integration' (Keller, 1983, pp.102-3, p.115). This type of thinking process is not only below the level of conscious awareness, but outside of conscious control, thus Lloyd Morgan's recommendation to "saturate yourself through and through with your subject, and wait" (as quoted in Koestler, 1982 {1967} , p.363). However, in general this aspect of the scientific process has tended to be underplayed, and once a hypothesis has been subjected to rigorous scientific testing, the mysterious nature of its initial induction in the mind is ignored. Indeed, Medawar (1963) claimed that – post-hoc – the hypothesis tends to be presented as if logically emerging from the data that was collected whilst it was being tested; and in this sense the scientific research paper is fraudulent, "because it misrepresents the process of thought that accompanied or gave rise to the work that is described in the paper", so that "the scientific paper in its orthodox form does embody a totally mistaken conception, even a travesty, of the nature of scientific thought" (Medawar, 1963, p.228).
In my own research there were moments during the analysis of data that I became aware of hypotheses about relevant categories that seemed to describe aspects of the data (such as for Annie's meaning for 'charge', §7.2.2, and for Tajinder's notion of 'conservation of force', §8.2.5). Such a hypothesis may be judged to be authentic if it resonates with the data: that is if the hypothesis is found to match other parts of the data set, and is not significantly challenged by incommensurate data. In my research I referred to this process of matching, of checking hypothesised categories against data, as post-inductive resonance. It is my belief, based on my own experience of the data analysis, that to a large extent the process of post-inductive resonance occurs at a sub-conscious level. Over a period of time, immersion in a data set leads to the sudden realisation that one has interpretations that seem to fit ('resonate with') the data, but which one has not up to that point consciously thought through. One may be able to offer a post-hoc reconstruction of the match between data and interpretation, but one is not able to describe the inductive process. (I originally intended to use the term 'inductive resonance', suggested by the title of a piece by the musician Robert Fripp. However, I understand that this term derives from systems theory, and is a measure of a system's capacity to hold together under stress (Fripp, personal communication, 14.3.96). As my own intended use was different to this existing technical meaning, I decided to add the prefix 'post-', as the process being considered is the resonance between an induced category and the data – which occurs after the creative process of initial category induction itself.)
§4.2.2: Induction and the methods of natural science
Medawar's argument was that induction has at its origin nothing more than guesswork, but that the initial origin of a hypothesis did not invalidate the research which followed. Scientific theory is judged by the match of theory to observation, which is independent of the creative act of forming the initial hypothesis. Whilst Medawar thought this 'fraud' gave an unfortunate distortion to accounts of scientific work, this conventionalisation of accounts (often required by journals), did not affect the validity of the conclusions, as these depended on the controlled experimental method.
A problematic aspect of positivist science is the logical impossibility of demonstrating the truth of general statements from any finite set of particular instances: there will always be alternative (albeit perhaps less parsimonious) interpretations that are consistent with a limited data set, and there is always the possibility that the instances not studied would refute the hypothesis. This is formally known as 'the problem of induction'. Consequently, Popper (1959 {1934} ; 1989 {1963} ) has discussed how in principle scientists should proceed by conjecture and refutation, and seek falsification rather than confirmation of their theories. Most natural scientists may be considered to work within a disciplinary matrix (Kuhn, 1970 {1962} ) or research programme (Lakatos, 1965) where there is a theoretical core (e.g. Lakatos' hard core) which is generally considered secure, and which – in practice, rather than in terms of pure logic – effectively limits the range of acceptable interpretations of a data set. For example, within physics or chemistry, carrying out an experiment on a Tuesday rather than a Wednesday would not be considered a variable worth controlling or exploring, and explanations that violate certain conservation laws would not usually be entertained. All such assumptions can only be formally justified in terms of the existing theoretical framework of science, which in Popper's terms should be considered provisional.
Some commentators have accordingly taken a 'relativist' position (c.f. §2.0 and §2.3.9), often inspired by Kuhn's work on scientific revolutions (1970 {1962} ), and suggested that the development of science is itself irrational, and depends more on the power of rhetoric than on logical argument. So Feyerabend has suggested that "the events, procedures and results that constitute the sciences have no common structure" (1988, p.1). Regardless of the rationality of science, there certainly are standards of evidence that are generally accepted in scientific fields which are tied closely to accepted methodology: in terms of experimental design, data collection
techniques, acceptable instrumentation, and approved procedures for data analysis. Further, each of these aspects has to be – in principal – described in sufficient detail for work to be independently reproduced. Where work cannot be reliably replicated, such as Fleischmann and Pon's work on 'cold fusion' (Close, 1990), it is considered to be pathological science, and is not generally accepted.
§4.2.3: The criterion of authenticity in educational research.
In qualitative studies in education, such as this present thesis, neither controlled experiments, nor statistical testing are possible, and the notion of proving theories is not appropriate. There is no agreed canon of core theoretical ideas that must be taken as axiomatic in all educational research (or even within science education – see chapter 2, §2.5); there are too many variables to control to follow Bacon's methodology; and research often involves unique, sentient, feeling others, rather than reproducible inanimate samples. Consequently the present work presents a model or theory which is supported by evidence from the data base, but there is no suggestion that my findings have been, or are capable of being 'proved'. This model is intended to relate to a concern from my own professional practice, and to inform my – and, I would hope, other teachers' – future practice (c.f. §4.1.2).
An important distinction here is between general and singular problems (Wenham, 1987). Science is generally concerned with general problems, whereas in fields such as teaching or the practice of medicine (as opposed to medical research) the practitioners are concerned with 'diagnosing' and 'treating' individual cases. As Wenham has pointed out, the approach to singular problems (where traditional notions of induction are irrelevant) should be different to that used with general problems. In particular, in the singular case it is quite acceptable to seek confirmatory evidence for a hypothesis, providing such tests could conceivably falsify that hypothesis (p.50).
In the research reported here, I have attempted to describe student thinking about a topic at particular stages in their developing understanding. During an interview I would form hypotheses, and then set out to test them by asking particular questions. However once the interview was finished, that opportunity was lost. A new hypothesis that occurred to me as I read a transcript – maybe months later in the light of some subsequent interview, perhaps with another learner, or just because I had time to reflect on the data – can not be fully tested. I could still interrogate the data, but I could no longer interrogate the student at the same stage in their studies. Nor can my study ever be truly replicated, in the the sense that failure to find similar results with other learners at some time in the future would not invalidate my own findings as an account of my colearners' cases (although it might refute the suggestion that my model has wider application). It is often pointed out that in work of the kind presented in this thesis it is not sensible to discuss the validity of the findings as one would when hypotheses may be tested statistically, but rather to refer to the authenticity of the findings (see below, §4.10.1). In view of these very significant limitations, it is important to be explicit as possible about the process through which my findings have been obtained.
§4.2.4: Sources of bias in research
One aspect of naïve Baconian method is that it assumes the researcher brings no bias to the data: so starting with a tabula rasa, the interrogation of nature will cause the pattern to be revealed. Apart from the assumptions that there is a pattern to be found, and that the scientist is potentially able to recognise it, such a view completely ignores both the inherent biases of any human's perceptual and cognitive systems, and the individual nature of a particular learner's (i.e. researcher's) existing cognitive structure due to prior experience (see the discussion of figure 2.4 in chapter 2). A Baconian observer with no biases would presumably not be able to operate, as faced with a necessarily limited data set, which is capable of being interpreted in myriad ways, the lack of some bias would surely prevent the selection of an initial preferred hypothesis (c.f. §2.3.4). As Furlong and Edwards explain one's theory does not only 'explain' the data, but determines what is recognised as data to be explained (1993, p.51). So for example, according to Stubbs, linguistic studies in education do not make up a paradigm, and lack coherence, as they do not deal with "a well articulated set of problems in well-defined ways, with agreed standards of solution and explanation (1993, p.63, c.f. Kuhn, 1970 {1962} ). Stubbs concludes that in this field there is a "problem of how researchers can place some control on their intuitions" (p.75).
In my own research I accept that the categories that I use to interpret and classify my data can not be assumed to be inherent in the data itself (§4.2.1). When I 'recognised' some pattern in the data, this was indeed a re-cognition. Over the course of human evolution the perceptive-conceptual system has evolved so that the developing brain has a tendency to recognise certain types of patterns, so that – for example – there is cross cultural recognition of certain classes of objects, the 'natural kind categories' (Gelman and Markman, 1986). These categories reflect the operation of natural selection, and relate to ways of thinking that had survival value in the environment in which our ancestors operated – an environment which did not include the 'objects' of modern science such as the concept of the chemical bond (§1.5.4). A striking example of this is the development of language, where there is now strong evidence to suggest that the human infant brain is structured such that the child will learn any human language to which it is exposed at the appropriate stage of development. Although at first sight this suggests flexibility, researchers have found that despite the differences between different human languages, they all follow certain specific common patterns. As Pinker explains,
"The universal plan underlying languages, with auxiliaries and inversion rules, nouns and verbs, subjects and objects, phrases and clauses, case and agreement, and so on, seems to suggest a commonality in the brains of speakers, because many other plans would have been just as useful."
Pinker, 1995, p.43
Indeed Chomsky has demonstrated that children are able to use language in accordance with subtle rules that they have not been taught (Pinker, 1995, pp.40).
Furlong and Edwards emphasise the importance of the researcher making explicit the theoretical position that guided the choices about data collection, and formed the background to the presented account (1993, p.54). My own individual biases include both my own understanding of the topic area (e.g. chemical bonding), and my knowledge of other researchers' findings from studies of student learning (i.e., my reading of the literature, as reviewed in chapter 3). From a Kuhnian perspective, I also approach the research with a wealth of expectations about the types of outcomes that might be seen as appropriate within the paradigm – e.g., the presentation of 'alternative conceptions' and 'frameworks' (i.e., my reading of the literature reviewed in chapter 2, c.f. §2.11). The critical attitude would be to assume that no amount of immersion in the data will completely cancel the various biases that I bring to the research. Despite Furlong and Edwards' advice these biases are unlikely to all be explicit in this account, because I am not necessarily consciously aware of them, and indeed they would seem 'natural' to me, i.e. what Feyerabend refers to as "natural interpretations – ideas so closely connected with observations that it needs a special effort to realize their existence" (1988, p.55, and cf. the notion of Gestalts discussed in §2.4.4).
§4.2.5: Ensuring the authenticity of the research
Given that any provisional interpretation that I may conjecture when working through my data is a recognition of some pattern, channelled by my own existing cognitive structure (c.f. de Bono, 1969, pp.61; see also Johnson and Gott, 1996, p. 563), it is important that my interpretation is authenticated against the data. As the process of post-inductive resonance – whereby certain categories are found to resonate with the data, and are brought into consciousness (§.4.2.1) – is not open to introspection, and as my findings are not open to scientific replication, it is important to ensure that
(1) my analysis of data is thorough enough to ensure the authenticity of the categories used;
(2) my presentation of findings includes sufficient detail to demonstrate this authenticity.
Pope and Denicolo (1986) emphasise the importance of being explicit about processes of data reduction and the categories used in interpreting data, and the importance of presenting detailed results (§2.4.2). In the next chapter (chapter 5) my analytical procedures are detailed, and in chapters 7 to 11 the findings summarised in chapter 6 are illustrated in detail. The analytical process described in chapter 5 follows the principles of grounded theory (discussed below, §4.4). In this approach the sub-conscious nature of the inductive process is acknowledged,
"Generating grounded theory takes time. It is above all a delayed action phenomenon. Little increments in coding, analyzing and collecting data cook and mature then to blossom later into theoretical memos. Significant theoretical realizations come with growth and maturity in the data, and much of this is outside the analyst's awareness until it happens."
Glaser, 1978, p.18, emphasis in original
§4.3: Ethical concerns in the study
Throughout this research study I have attempted to balance my desire to collect data in a systematic and reliable manner, with a concern for a high standard of ethics. In particular this meant that I felt it was essential to respect the feelings of others who might be affected by my research.
The following principles were followed:
• to inform – and obtain consent – from colleagues;
• to ensure all students involved in the case study work volunteer their time, and feel their involvement is worthwhile; • to ensure confidentiality of data.
§4.3.1.: Informing colleagues
I obtained the formal approval of my department and institution. Indeed through its Staff Development programme, my College supported my registration for the Ph.D. programme. Outcomes of the research (reports, publications etc.) were circulated to various key people in the College.
I informed my colleagues teaching chemistry in the college of the research I was intending to carry out, and made sure they had no objections. The only concern was that I might be involved in an exercise which was evaluating or appraising their teaching, and I was able to give an assurance this was not the case. (It should be noted that two year A level chemistry classes in the College are seldom taught by only one lecturer during their course, and it was not possible to restrict myself to students that were only taught by myself for chemistry.) My colleagues were provided with copies of papers describing the outcomes of my research. Colleagues in other institutions who provided data for the work described in appendix 2 were sent a detailed report of the findings.
§4.3.2: Informants as colearners
In order to attempt to avoid the potential problems of a researcher-subject relationship where power lies predominantly at one pole (mine), an attempt was made to build safeguards into my enquiry. An important part of this was my conceptualisation of the role of the students who agree to partake in the study (that is, they are considered as 'colearners' in the research).
The most important principle I set-out for my study was to respect my students' right not to be involved in my research unless they wanted to. This included
(a) not assuming that colearners would wish to continue their involvement, but rather inviting them to each subsequent research session;
(b) making it clear that colearners were free to leave the study at any time, and that they could decline to be involved on specific occasions;
(c) making a point of asking colearners how they felt about each research session at its end – indeed I introduced into the research a simple feedback form (see appendix 10), as well as asking verbally . One colearner who was keen to be involved in the research but found repeated completion of the feedback form to be a little tedious after a number of interviews, agreed to keep a diary of his reflections on the experience instead (see appendix 10, §A10.3).
Appendix 10 gives details of the questions used to elicit feedback form the colearners (§A10.1), and their responses. The appendix reports how there was unanimous agreement that the sessions had been worthwhile for the colearners, and near-unanimous recognition that they had learnt something through the process of being interviewed (§A10.2.3, c.f. appendix 8).
The term 'co-researcher' has been used in the literature to describe people involved in a research project. In contrast to the traditional approach (where the roles of the researcher and subject have been clearly differentiated) alternative models have been proposed where the 'subject' becomes coresearcher and the researcher becomes cosubject (e.g. Heron, 1981a,b). The term coresearcher seems appropriate when applied, say, to teachers in classroom studies, such as the teachers involved in the CLiSP case studies referred to in the previous chapters, (Brook and Driver, 1986; Wightman et al., 1986); or to the 'subjects' of studies into teaching style and behaviour – such as 'Sandra' who contributed the 'participating teacher's foreword' to a book based on the case studies of herself and a teaching colleague (Tobin et al., 1990). Although my own partners in the research enterprise were valued as people and consulted about their own roles in the enquiry, they were not "contributing to the research propositions at all stages from the working hypothesis to the research conclusions" (Heron, 1981a, p.156), so the term co-researchers was not considered appropriate.
However, I attempted to ensure that my field work did not follow the traditional pattern that has been described as the 'rape model', where "the researcher comes in, takes what he wants, and leaves when he feels like it" (Lincoln, in conversation with Beld, 1994). The term 'colearners' seems to accurately represent the relationship between myself and my partners, without over-stating the case.
My colearners are clearly mature enough to make responsible decisions. They were students in post-compulsory education, having timetables with large gaps where they were assumed to be responsible for their own use of time. Secondly I was one of their teachers, and to some extent an 'authority figure', who is involved in evaluating student progress and making decisions (for example, about progression onto year 2 of an A level course, and recommendation for examination entry). Clearly there is scope for abuse of my position – I could have behaved more favourably towards students who agreed to be involved in my research than their peers who declined.
It might also be suggested that students brought up to respect teachers – and to be deferential towards their requests – could easily find themselves spending much time in research activities, without feeling that they are in any way benefiting from the interaction. Whilst it could be argued that time spent discussing their academic work with an 'expert' can only be of benefit to a student (as it will provoke them to think about their work, and they will learn from the experience), this begs the question, as to who is in a position to make such a decision on behalf of the student? The students may feel they could have spent the time more usefully re- writing their notes or reading a book. The sessions could just confuse them. Some might simply feel they would rather spend their time in some other way.
My primary data collection technique has been the use of respondent interviews (§4.6.2). Powney and Watts consider research interviews as "conversational encounters to a purpose" (1987, p.vii). My purpose was to collect data for my research. But it could be asked 'what purpose do my colearners have, and why should they want to spend their time talking to me?' If the answer to that question was that they were in some sense fearful of the consequences of not taking part, then any findings would be tainted by the abuse of power implied. Of course in any personal relationship the desire to please the other may be a good reason for acting in a particular way. However, for young people to give up hours of their time, and to put themselves through an 'interrogation', I felt they should be offered something more. Of course the colearners attended College to learn more about science – partly because they were intrinsically interested and partly because of their career plans: often to seek a place at University to read science-related degrees. Interview sessions could offer the students an opportunity to learn at two levels: to learn about how well they understood the work, and to learn about chemistry through the dialogue itself. It is my perception, that this was what happened in most of the interviews, and this was supported by the evaluations made by the students themselves (appendix 10, §A10.2.3). Consequently, the research sessions became mutual learning experiences. Although the partners had somewhat different learning goals, each was aware of what the other wishes to learn from the experience. There was no deception, and the purposes are certainly not inconsistent. The relationship became symbiotic: we were colearners in the process.
§4.3.3: Confidentiality
As this study was largely idiographic in nature, grounded in detailed data from individual learners, it was appropriate to refer to the colearners by names when preparing case studies, and in the findings presented in later chapters. In order to offer confidentiality, each colearner was ascribed an alphabetical code, in the order in which they enrolled in the project. In producing case studies names were given using the code letter as an initial: thus A became Annie, and T became Tajinder. (Some letters were not used, such as O, and some codes were ascribed to students who provided data focussed on another science topic which is not reported here).
In the final stage of the project, when data was collected from the diagnostic instruments (appendices 2 and 3), teachers sending me data were given a breakdown for their own classes, but the general report sent to all the contributing institutions did not give any details of the schools and colleges, nor of individual students or classes.
Ensuring confidentiality is not a straightforward matter. In one sense the researcher would like to give as much background information as possible to readers of a research report. However, the more information presented, the more likely it is that individuals could be identified. This issue arose in the present study. In a paper presented at a conference I included details in the appendix of dates of birth and examination grades at entry to college, of the colearners whose ideas were discussed. In addition my own affiliation – and therefore the College attended by the colearners – was given at the head of the paper. In principle this could have been sufficient to identify the individuals who made specific comments reported in the paper: a point brought home to me when the colearners informed me that they had been reading the paper in the College library, and had worked out which of them was represented by which code letters. The students did not criticise my inclusion of the personal data, and they pointed out that they had only achieved the
identification by collectively exchanging information of their birth dates and exam. records, but I felt I had – in principle – failed in my duty to protect their identities. In the present account of my work details about individual learners are only provided where it is felt to be of specific relevance to the reader.
§4.4: The research as grounded theory
"A grounded theory analysis starts with data and remains close to the data. Levels of abstraction are built directly upon the data and are checked and refined by gathering further data"
Charmaz, 1995, p.28
The present research has followed the approach known as grounded theory.
§4.4.1: An overview of grounded theory
The research reported here followed an approach that is described as 'grounded theory'. Grounded theory derives from the work of Glaser and Strauss (1967), who believe theories should be 'grounded' in the data that are generated during research (Cohen and Manion, 1989, p.141), rather than research just being determined by established theory. Although this approach was developed by sociologists, Glaser describes the approach as a "general methodology" (1978, p.164), and claims it has been used in many fields, including education (p.3). In this approach theory is generated from (Glaser and Strauss, 1967, p.31) – or is considered to emerge from (Charmaz, 1995, p.47) – the data collected. Glaser talks of developing 'theoretical sensitivity', which requires commencing the research with as "few predetermined ideas as possible" so that observations may be recorded with as little filtering through preexisting hypotheses as possible (1978, pp.2-3). Woods refers to the process of preparing to enter into research as "washing your mind clean" (quoted in Measor and Woods, 1991, p.69).
Much of what has already been presented in this thesis demonstrates the difficulty of avoiding – and even recognising – one's biases: the researcher will construe the world through his or her personal construct system (§2.2.4), and is trained within a research tradition (§4.1). However, the significant point here is that rather than commence research assuming the relevance of the theoretical concepts and categories that are established in the research field, and therefore fitting the data to those categories, the grounded theorist has a critical attitude and attempts to be led by the data itself (Glaser, 1978, p.4, see §4.4.3 below).
The research design itself also emerges during the research, as the researcher uses theoretical sampling (i.e. decisions about on-going data collection are guided by the emerging theory, Glaser and Strauss, 1967, pp.45), as the research becomes more focussed (see figure 4.1).
There is a constant process of reviewing the emerging model against the data collected – the 'constant comparison' method (Glaser and Strauss, 1967, pp.113-115), using 'double-back' steps (see §4.4.4 below). In other words, as new data is collected and analysed, the provisional model (and therefore the analytical scheme) is reviewed, existing data is revisited in the light of the revised analytical model, and where and how to collect data next is reconsidered. During this process there is a gradual shift in the balance of the researcher's emphasis from data collection to analysis to writing-up. The end-point is reached by a process of theoretical saturation where further data collection and analysis does not significantly change the model being developed (Charmaz, 1995, p.28, p.31, p.34; Glaser, 1978, p.16, p.36, p.85; Glaser and Strauss, 1967, pp.61-62; Reason and Rowan, 1981, p.xx; c.f. Johnson and Gott's notion of "developing the neutral ground" between researcher and learner "through a process of successive approximation", 1996, p.568).
Shipman (1988) has pointed out some of the shortcomings of the grounded theory approach. He suggests that it is not always easy for researchers using this approach to decide what data to collect at points in their research (p.41) – although I did not find this. Shipman also suggests that in accounts of research the evidence and interpretation can run together, making it difficult for readers to distinguish (p.63). I have attempted to aid the reader by providing examples of verbatim data extracts to support my interpretations. Shipman also suggests that reports may be unbalanced by the tendency to focus on the aspects of findings that are considered theoretically interesting (p.63). However, providing that the reader is aware of the nature of grounded theory, this could be considered a strength rather than a weakness.
In the present research the first stages of data analysis commenced as soon as data was collected, and informed the subsequent episodes of data collection (§5.2). For example, analysis of recordings of interviews led to additional questions and focal diagrams being incorporated in later interviews (§5.1.3). Some specific interpretations of interviewees' thinking also formed the basis of items used in pen- and-pencil instruments (the truth about ionic bonding diagnostic instrument, and the truth about ionisation energy diagnostic instrument, see appendices 2 and 3) that were used with larger samples of learners.
§4.4.2: Data collection
The present research underwent several phases, each informed by the earlier work, and data collection and analysis were carried out concurrently. Although this research is primarily an interview study, other forms of data were also collected (§4.5) in order to,
• provide sources of contexts for interview discussions that arose from the colearners' own course-work;
• authenticate the findings of interviews against evidence of the colearners' thinking from other contexts;
• to test the categories derived from the analysis of the colearners by comparing them against data from a wider range of learners (cf. Charmaz, 1995, p.42).
Glaser and Strauss argue that there is no single research technique that is necessarily most appropriate for generating grounded theory, and they recommend collecting different types of data to provide a range of vantage points for exploring categories – what they call slices of data (Glaser and Strauss, 1967, p.65). Labov recommends "supplementing interviews by collecting data from tests, elicitations, experiments, observations and different types of recordings" (reported in Stubbs, 1983, p.221). In the present research, interviews were supplemented by recording student dialogue, using the construct repertory test, and collecting samples of student course-work such as concept maps and tests. This eclectic methodology reflects the "openness and flexibility of approach" considered appropriate for grounded theory work (Charnaz, 1995, p.47).
The first stage of data collection consisted of an interview study for which data was collected over the period January 1991 – May 1992. A 'deck' of 17 diagrams, informed by an analysis of the topic area at Advanced (A) level, were prepared on A4 paper to act as foci for the interviews (§5.1.2). The colearners in the study were four volunteer A level chemistry students, denoted here by assumed names: Annie, Brian, Carol and Debra. The four A level chemistry students were interviewed at three stages of their A level course: these conversations were recorded on cassette audio tape. During this phase points of interest were noted, and new foci diagrams added to the pack. It was also decided to trial a supplementary technique based on Kelly's triads (§4.7), and another student (denoted Edward) was enrolled in the project, initially for this purpose. (Edward was the only colearner I did not teach for chemistry, although he was in my A level physics class).
During the second phase, September 1992 – June 1994, ten students commencing A level studies were enrolled to be interviewed. These colearners are denoted as Jagdish, Kabul, Lovesh, Mike, Noor, Paminder, Quorat, Rhea, Tajinder and Umar. Of these ten volunteers, eight provided data for the duration of the first year of the course (two left the course) and four continued to be interviewed through their second year (see appendix 1). In addition to the prepared foci diagrams, other probes were used (for example, early in the course the colearners were asked to list, and then try and draw representations of, different types of chemical bonding they were aware of.) Detailed notes of the interviews were made, including full transcriptions of large sections of tape (§5.2.1). As well as interviews, tasks based on Kelly's triads (§4.7) were used, and other data that was available to me as a teacher – such as examples of course work – were collected (§4.9). Some additional sessions were taped with colearners working as pairs on past examination questions relevant to chemical bonding (§4.8).
Throughout the period of the interview study, supplementary data that could be relevant to the research was collected from other learners. I will refer to this source as 'incidental data'. This took the form of keeping copies of students' responses to relevant course work activities such as induction exercises, and certain tests. Sometimes this involved an individual response that was considered potentially interesting, and sometimes copying the work of an entire class. When a substantive question about bonding was included in the A level mock examination one year, all the responses to that question were copied. This material provided a bank of data which could be interrogated to compare with the findings from the colearners interviewed.
The grounded theory approach is intended to generate models that can potentially be tested by traditional logico-deductive techniques (Charmaz, 1995, p.48) – thus grounded theory creates a bridge between idiographic and nomothetic research (§4.1). The third stage of the project was designed to test the relevance and applicability of the findings of the previous stages to the teaching of chemistry at A level. Some of the specific notions elicited in the study were used to design paper- and-pencil instruments that could be used to survey and diagnose these conceptions in wider populations.
It was not within the brief or resources of this study to undertake a large representative survey of chemistry learners: but an attempt was made to test the feasibility of such a survey for two areas (ionic bonding and ionisation energies) where the study suggested colearners could be applying common alternative conceptions. Although the case studies of individual learners are considered to be of intrinsic interest to understanding the learning of science at an individual level, the diagnostic instruments developed demonstrate that the present research uncovered notions that should be of widespread interest to the chemical education community. In order to avoid some of the criticism of multiple choice formats (see below, §4.5), these diagnostic tests presented a selection of statements, each of which could be separately judged as 'true' or 'false'. In addition a 'do not know' option was available. These instruments are discussed in appendices 2 and 3, and some of the findings obtained have been incorporated in chapters 10 and 11.
§4.4.3: Use of the literature
A more traditional approach to research would place the literature review before research design, data collection and analysis, as research questions would be derived from the existing literature. In grounded theory work it is recommended that the literature search is delayed in an attempt to avoid the researcher deriving categories from the literature rather than the data (Charmaz, 1995, p.47; Glaser, 1978, p.31). However, it is accepted that the researcher will have knowledge of the field that will influence the interpretation of data, and even that a general awareness of a wide range of possible variables and theoretical ideas can increase sensitivity to the data (Charmaz, 1995, p.32, c.f. Kuhn 1970; Furlong and Edwards, 1993, p.54; Glaser, 1978, p.3; Glaser and Strauss, 1967, p.253). Although categories from the literature might well be adopted in a grounded theory study, the researcher has to develop an emergent fit: that is modify the category to fit the data and not select the data to match the category (Charmaz, 1995, p.38; Glaser, 1978, p.4).
In the present research the literature was studied alongside the processes of data collection and analysis. A provisional literature search was carried out in the Summer of 1990, when I was awarded a study visit to Merton College, Oxford. Data collection commenced in January 1991. My study of the literature continued throughout the research, but the drafting of a formal literature review for this thesis did not commence until the interview data had been collected, and much initial analysis had been completed.
§4.4.4: Analysis
"Grounded theory methods consist of a set of inductive strategies for analysing data. That means you start with individual cases, incidents or experiences and develop progressively more abstract conceptual categories to synthesize, to explain and to understand your data and to identify patterned relationships within it."
Charmaz, 1995, p.28
Analysis of data in grounded theory studies proceeds through a process called 'focussed coding'. In practice this means that when the researcher starts analysis, the data (for example, extracts from an interview transcript) are annotated freely with the impressions and interpretations they suggest. There is no limit to the number or format of these codes. However, as analysis proceeds the researcher begins to develop a set of codings that seem to be most pertinent in explaining the data, and over time these will be organised into a set of categories based on groupings of the codes that seem most significant (Charmaz, 1995, pp.37-40).
Two points that should be reiterated here are, firstly, that this is an inductive process (Glaser and Strauss, 1967, p.251) and therefore heavily dependent on subconscious thinking (§4.2), and that it is therefore necessary to constantly check on the authenticity of codes and categories by 'double back procedures' (§4.4.1).
Figure 4.1 represents this aspect of grounded theory procedures schematically.
figure 4.1: a schematic showing the nature of grounded theory
Therefore the set of codes that emerge from initial data analysis may be used to work through large quantities of data – but it will be subject to additions, deletions, substitions and modifications. When categories are developed they are tested against the data already coded, as well as being used to study new data. The categories must be refitted, that is modified to match the data being studied, and gradually elaborated and refined to describe finer details of the data (Charmaz, 1995, p.42; Glaser, 1978, p.4). Theoretical sampling is then undertaken, that is data collection is targeted to inform the developing scheme (Ball, 1991, p.184; Charmaz, 1995, pp.43-46). In particular Charmaz recommends testing categorises by comparing them against data from different people, against data from the same people at different times, and against other categories (p.42). If theoretical sampling appears to collaborate the categories being used, and the relationships between them, then Charmaz would argue that they may now be considered as 'concepts' (p.45). (This process, moving from codes to categories to concepts, has a parallel in Gilbert and Watts' description of the analysis of interviews of students discussing aspects of science, when they suggest three levels of analysis: moving from conceptions to categories to frameworks: 1983, pp.69.)
The starting point for analysing the data collected in the present research was my own conceptualisation of the topic area (see chapter 1, and appendices 4 and 5), and some features of learners' thinking that seemed important in my reading of data from other studies, such as detailed data given in some of the CLiSP reports presented in the literature (e.g. Brook and Driver, 1986; Wightman et al., 1986). These sources provided some of the codes used in the initial interpretation of data.
The details of the analytical process are presented in the next chapter (§5.2), and here only a general outline will be given. The interviews from Annie, Brian, Carol and Debra were analysed by summarising the contents and coding points of interest for exploration in the subsequent interviews and analysis. The interviews were transcribed. One of the colearners, Annie, was selected, and a case study was prepared (which is described in chapter 7), largely organised in terms of my initial analysis of the topic area.
The findings from the case study were formally written up and presented to a critical audience (at a symposium at a conference of the British Educational R esearch Association); reviewed and published (in R esearch in Science and T echnological Education).
One outcome of this work was a shift in focus from the initial categories used to organise the work (based on the topic) to particular aspects of Annie's thinking (see in particular §7.4).
A procedure was developed for the on-going analysis of data from the second cohort of colearners, and this was formalised in a working paper (in June 1993). When the cohort had completed their studies, Tajinder – the colearner who had provided the most data – was selected as the most suitable case for detailed analysis. By this time the basis for organising the case study were features of Tajinder's own conceptualisation of chemical bonding, rather than my own initial analysis of the topic. To write this case I spent a number of months where I worked only on the data from this one learner, and refined the analysis through a number of stages (see chapter 5, §5.3). At the end of this period I wrote up the case summary (August 1995), that has since developed into chapter 8 of this thesis.
In the present research the case studies of individual learners were prepared, firstly as 'findings' in their own right (i.e., chapters 7 and 8; Taber, 1995, Taber and Watts, 1997), and as part of the process of developing the general model outlined in chapter 6, and documented in chapters 9, 10 and 11.
The data from the other colearners were re-examined in the light of the case study of Tajinder. Selective (in terms of the categories emerging from the analysis) case studies were prepared for Carol, Debra and Edward. Case studies of Jagdish and Kabul were then also produced in terms of the emerging analytical model. In this way the categories were checked against the data from a number of informants. Through this process the categories used in the analysis were refined. The next stage was to interrogate the data from other colearners (Brian, Lovesh, Mike, Noor, Paminder, Quorat, Rhea and Umar), to further refine the analytical model. Finally, the data collected from other students was considered and coded in the light of the emergent categories.
Some further interviews had also been undertaken with a number of colearners subsequent to Tajinder's cohort, and initial analysis of this data had been undertaken (by reviewing the recordings and producing outline protocols of the content). However, it was felt that saturation of the theoretical model had been achieved, and that further analysis of this data would add little to the model (§4.4.1).
§4.4.5: Reporting
At this point the findings were written up thematically, to give chapters 9, 10 and 11. Charmaz has recommended that grounded theory reports should include sufficient verbatim material to allow the reader to judge how the analysis follows from the data (p.47, a similar point is made by Furlong and Edwards, 1993, p.54), and as Walker points out "case studies do not really lend themsleves to data coll apse" (1993, p.179).
However – as Pope and Denicolo point out (1986, p.156) – the researcher faces a dilemma between two responsibilities to readers: to provide sufficient evidence to justify the model presented, and to produce a research report that is succint enough to allow a reader to appreciate the key points.
In this research I have attempted to produce an authentic account that retains readability (§5.3). In order to avoid disrupting the narrative flow of my arguments (Zeller, 1995), I have selected a limited amount of verbatim material to illustrate the main points of my model. I have however also appended a range of additional extracts from the data to provide the reader with sufficient material to evaluate my analysis (see §5.3).
§4.5: Choice of research techniques
As explained above (§4.4.2) procedures to develop grounded theory "start with individual cases" (Charmaz, 1995, p.28), and work towards models that may "be verified through traditional logico-deductive methods" (p.48). This present study was able to generate such ideas, and a limited amount of survey work has been undertaken to show the feasibility of testing the generality of aspects of the model produced (appendices 2 and 3).
However, the bulk of this thesis is concerned with constructing a model of the development of learners' understanding about chemical bonding, which is grounded in data reflecting learners' thinking. This therefore led the selection of data collection techniques. As Driver, Leach, Millar and Scott have explained, studying learners' thinking in depth (in their case about the nature of science) excludes techniques designed to collect data from a wide range of learners. As in their study, the ideas being explored in this research were "subtle and complex", and the nuances of individuals' thinking were unlikely to be fully elicited by written surveys using preestablished questions, or particularly with multiple choice items forcing a choice between preselected responses (Driver, et al., 1996, p.66).
The main data collection technique used in this present study was the interview (§4.6). Interviews provide opportunities for the researcher to test the validity of interpretations by asking follow-up questions, repeating questions at a different point in the interview, and asking about the same point in a different context (c.f. §2.2.3). Interviews also provide the researcher with the flexibility to respond to the respondent's comments: to test hypotheses about their meaning and reasoning, to clarify ambiguity, to explore the degree of tentativeness of a response.
Solomon has suggested that any single approach to investigating students' understanding will only cue a limited range of responses from the repertoire available to the learner (Solomon, 1992, p.40). However – as Driver et al. intimate – other available methods suffer limitations. For example, in survey work, questions have often been set in a multiple-choice format (see for example the Bar and Travis {1991} data discussed in Chapter 3, §3.1.3 and appendix 7, §A7.1) based on alternative conceptions elicited in interviews. Bar and Travis report that the proportion of respondents selecting an explanation presented as an option in a multiple choice format may be much higher than that giving the response spontaneously in response to open-ended questions Driver, et al., 1996, p.47(pp.369-370). Solomon's own research group's use of this type of format has been criticised by Driver and coworkers, as
Driver, et al., 1996, p.47
- students' choice of response is constrained by the options offered, none of which may capture precisely a students' view;
- the multiple choice approach relies heavily on students' interpretation of the wording of the question matching that intended by the researchers who framed the responses offered;
- variation of view within each answer category is obscured
In the present research the approach taken was to base the study on interviews, which were considered to be the most powerful technique for exploring a learner's thinking in depth, but to use auxiliary data as a means of methodological triangulation (§4.10.4). It was considered that there was sufficient scope within the interview context to elicit comments that could allow a valid representation of the colearner's thinking to be built up for those aspects of thinking about the topic which could be triggered and probed during the interviews. However, to take on the concerns about context and cuing, relevant data was collected from the colearners' normal course work (§4.9). This data was scrutinised, and when points of interest were discovered they were, where possible, probed in the interview situation. In this way the advantages of the flexibility and depth of the interview situation were combined with the opportunity to bring in points from material cued in a different social context. In addition, two supplementary methods of data collection were used. The first was Kelly's construct repertory test, which used different foci to the interviews, and had a very different 'task' structure (§4.7). The second was the setting up of dialogues between pairs of colearners (§4.8).
§4.6: Interviews
The main technique used in this study then is the semi-structured respondent interview. There is a strong tradition of enquiry using interviews in research into learners' ideas in science (considered further below, §4.6.2).
§4.6.1: The use of interviews
There are significant advantages to exploring learners' ideas through talk (§4.5), but it might be asked whether interview studies are the most appropriate research tool when most formal teacher-set assessments in science are written, as are terminal examinations such as A level. (Although the course followed by the colearners had a 'practical' paper, all instructions were presented in written form, and all answers had to be given in writing.) If learners were to demonstrate apparently different levels of understanding in written and spoken responses this would not invalidate interview studies per se, but could diminish their direct relevance to the practice of learning science in school and college.
This question has been considered by Seddon and Pedrosa who explored the hypothesis that students answers might depend upon whether the questions and answers were written or verbal (1988, p.337). They investigated the issue with first year science and engineering undergraduates in Portugal, in the context of questions about atomic and molecular structure and chemical equilibria (p.339). In their study they both compared the effects of presenting a set of questions in four modes (spoken question, spoken response; spoken question, written response; written question, spoken response; written question, written response) to different groups of students, and also of presenting students with questions divided into the four modes. Although their results demonstrated a small number of significant effects (on 8 statistical tests out of 375, at the 0.01 level) they concluded that the mode of questioning made no practical difference to student performance, and that all four modes could be considered equally valid (p.342).
Wightman, working with secondary school students, found that students' written responses did not always match their comments in interviews. In her case studies of secondary classes, Wightman clearly felt that the interview data was more reliable, and reported that her research made her question the validity of exploring children's ideas in science by considering only their written responses (Wightman et al., 1986, pp.317-318).
§4.6.2: The semi-structured nature of the interviews
The interviews used in this study are of a type often described as "semi-structured", which allows the interviewer flexibility to devise questions in situ in response to the interviewee's comments. This approach is appropriate for developing a model of student understanding (§4.5) – although it would not be suitable for testing the generalisability of the model (as "the interviewer is more free to probe beyond the answers in a manner which would often seem prejudicial to the aims of standardization and comparability", May, 1993, p.93).
Powney and Watts (1987) suggest that the most significant classification of interviews is as either respondent interviews (where "the interviewer retains control … [and] it is the interviewer's 'issues' that matter", pp.17-18) or informant interviews (where it is "primarily the interviewee who imposes" the agenda, p.18). In their terms my interviews would be classed in the former category. Such respondent interviews may be more or less structured, and the descriptor 'semi-structured' used above could be equated with what Powney and Watts describe as "a loosely structured interview [which] … implies a general set of ideas to which the interviewer would like some responses at some point in the session, though the order and exact wording are not important" (Powney and Watts, 1987, pp.17-18). So I entered the interview context not with a fixed schedule of questions, but rather with specific foci (§5.1.2), and certain questions in mind (§5.1.3).
The interviews followed the general pattern of the 'interview-about-instances' approach (Osborne and Gilbert, 1980) which has been outlined as "tape–recorded dyadic discussions with a pupil, using a series of pictures as a focus" (Watts, Gilbert and Pope, 1982, p.11). This technique involves presenting the learner a series of diagrams which are considered to be feasibly related to the focal concept, and then asking the learner whether the figure is perceived as representing an example of [their version of] the concept, and why they think so (Osborne and Gilbert, 1980, p.376; Watts and Gilbert, 1983, pp.162-3). This approach is flexible, and used in a non- intimidating way to allow an informal discussion to develop (Watts, 1983a, p.218). It is recognised that the technique it suitable for case studies that explore the ideas and meanings of individual learners (Watts, Gilbert and Pope, 1982, p.12).
My initial interviews were based around 17 simple diagrams that I had prepared to represent chemical species and structures (see chapter 5, §5.1.2). Some minor changes to a few diagrams were made as a result of early interviews, and subjects' comments led to a considerable expansion of the 'deck' of figures (although the full set of figures was not used in any one particular interview). The two questions I used to start discussion in the early interviews were 'what do you think is represented in this figure?' and 'is there any bonding represented in this diagram?', or some paraphrasing of these (§5.1.3). Follow-up questions obviously depended on the responses the students made. In subsequent interviews with a specific subject I would have a list of questions I wished to ask, formulated on listening to the recording of the first interview. In addition interesting points from one interview could lead to ideas for probing questions with other colearners.
With the colearner whom the main case study is based – Tajinder – there were a large number of interviews (23). Some of these followed the same outline as those used with the other colearners, but some were more open-ended, and were allowed to develop into a range of topics that I considered potentially relevant to the theme of chemical bonding. This both allowed Tajinder to have a significant input into the 'agenda' for some of the research sessions, and broadened the context of the discussions.
§4.6.3: The clinical nature of the interviews
The interviews that have been carried out for this study have been clinical in nature (§2.2.1). Although the interviews have taken place in the College the subjects attend, and the interviewer has been known to the colearners, the interview process has been formalised in a number of ways that distinguish it from the normal contexts of teacher-student talk (in classes, or informal social chat in corridors, refectory etc.)
The interviews have been by mutually agreed appointment, in a room away from interruptions and disturbances (as far as possible in a busy college) and recorded onto cassette tape. In addition the style of the conversation has been didactic, and unlike ones that subjects are familiar with in the teaching situation in that
(i) extended series of questions are asked of the same student (compared with the sharing round of questions in a class that often allows a shrug of the shoulders, or vague comment, to deflect a question onto another student);
(ii) the balance of talk has been much more evenly shared between the two participants (whereas in classrooms most of the talking is usually undertaken by the teacher, Sutton, 1974, p.42; White, 1988, pp.113-114); and
(iii) much time is spent exploring the subjects' ideas, and little (if any – see below, §4.10.3) time is spent transmitting the teacher's (i.e., researcher's) views.
§4.7: Kelly’s construct repertory test
§4.7.1: A brief overview of Kelly’s methodology
George Kelly has been considered by those concerned with constructivist approaches to science education to be one of the key antecedent thinkers. His theory of personal constructs (P.C.T.) was outlined in chapter 2 (§2.2.4).
Although Kelly's ideas have been used as a foundation for the constructivist position, most research into student understanding of science has ignored his methodology: the techniques of triads and the repertory grid. In the present study Kelly's method of the construct repertory test was used as a data collection technique to inform and supplement interviews.
Kelly's construct repertory test and repertory grid technique were designed for use in psychotherapy, and were used to find out how the client viewed aspects of his world. Often the components of the client's world used as foci were other people. The names (or roles) of significant people in the client's life would be written on pieces of card. These were referred to as 'elements'.
In the construct repertory test triads of elements would be presented to the client who had to divide them into a pair of elements that were construed to be in some way similar, and the one that was different (Bannister and Fransella, 1986, p.49). The discrimination was considered to be on the basis of the elements being nearer different poles of a (bipolar) construct used by the client to structure his or her world. By working though all the client's ways of discriminating in a series of triads, his or her significant constructs should be revealed. The repertory grid took this approach one step further: the client was asked to decide where each element should be placed on each of the bipolar constructs elicited. This gave a grid with columns (elements, categorised in terms of the constructs) and rows (constructs, used to discriminate between the elements). It was then possible to compare similarities between rows, and find a hierarchy in terms of the similarity in the way the client used his or her various constructs and thus reveal how constructs were related (Bannister and Fransella, 1986, pp.51). Grids allow statistical analysis, but the construct repertory test is considered to be sufficient where such analysis is not required (p.52).
In the present study this method of triads was used to elicit the constructs that student colearners use to discriminate between examples of representations of the microscopic structures discussed in chemistry (atoms, ions, molecules etc., see §5.1.4) Data was tabulated, and in some sessions colearners were then asked to decide whether other elements were construed at the disclosed pole of the elicited constructs (i.e. the 'emergent pole').
§4.7.2: Appropriateness of Kelly’s methodology to the present study
Kelly applied Personal Construct Theory as a therapist: i.e. largely in terms of how people construed their social environment, rather than the physical world. P.C.T. was devised in the context of therapy, not education (Solomon, 1994, p.7). Kelly and his advocates would agree that his theory and methods arose from that particular context, but not that they were limited by it. Kelly believed that systems have foci of convenience where they tend to be most effective, and for P.C.T. Kelly believed this was "in the area of human readjustment to stress" (Kelly, 1963 {1955} , p.12). But Kelly devised his psychology to apply to all situations where people construe meaning (Bannister and Fransella, 1986, p.4, p.21, p.44; Kelly, 1963 {1955} , p.130).
Kelly's theory concerned personal constructs, where the present study is concerned with conceptual development. It is therefore appropriate to ask whether constructs and concepts are the same thing, and if not, how they are related. The Oxford Companion to the Mind defines concept as "an abstraction or general notion that may serve as a unit (or an 'atom') of a theory" (Gregory, 1987). A dictionary of psychology suggests that a concept or conception is "that type or level of cognitive process which is characterised by the thinking of qualities, aspects, and relations of objects, at which therefore comparison, generalisation, abstraction, and reasoning become possible, of which language is the great instrument, and the product of the concept – normally represented by a word" (Drever, 1964). That same dictionary describes 'construct' as just "a term which some writers … have suggested as a substitute for concept" (Penguin Dictionary of Psychology).
Kelly himself, although preferring to use the term construct, suggested he did not mean something very different to concept (Kelly, 1963 {1955} , pp.69-70, see also Watts and Pope, 1985, p.9). If we accept that knowledge is personally constructed by individuals, rather than transmitted to them, then it is unlikely that any two students, or any two examiners – let alone a student and an examiner – will have exactly the same set of meanings for say 'covalent bond': with an exact agreement on examples and non-examples, and associations with other concepts. However, Kuhn has pointed out that science is not learnt in terms of agreed definitions (1977, pp.xviii – xix), and as has been pointed out earlier, total agreement is not necessary for science to proceed, as long as meanings are similar enough for effective communication most of the time (c.f. §2.2.3).
Lemke has studied classroom discourse in science lessons and concluded that in practice concepts may be considered "just bits of thematic patterns" as they are never used in isolation, and their utility derives from their interrelations (c.f. Vygotsky, 1986 {1934} , p.245, §2.2.2). He concludes that 'purely mental notions' of what a concept actually is tend to ignore the central role of language in learning (1990, p.91).
So although terms such as 'covalent bond' may be defined in text books and scientific dictionaries, students and practising scientists have their own personal meanings for the term. Fransella and Bannister conclude accordingly that studies of individuals' meanings for scientific concepts fall within the scope of Kelly's methods,
"In theoretical terms all constructs are personal. Even constructs drawn from say science or technology which have highly publicly specified relationships and implications and which have had their predictive validity tested and retested are still personal. They are personal in the sense that each person has to acquire them and integrate them into his total system. … there might be much of interest to be investigated using grids where the elements and constructs are drawn from areas of high public agreement."
Fransella and Bannister, 1977, p.117
There seems then to be justification in the literature for proposing that we may consider the versions of 'concepts' in students' minds as 'constructs' without doing violence to Kelly's original theory.
The core of Kelly's theory then was that people impose structure and meaning on their worlds, by making discriminations on the basis of a system of personal constructs. To elicit the constructs the therapist – or researcher – needs to present the client – or student colearner – with some foci with which to make a discrimination. It might be suggested that presenting two elements, and asking for ways in which they were similar and different, would suffice. However Kelly pioneered the use of triads of elements (1963 {1955} , p.112), which has an advantage that it allows discriminations to be made on the basis of tacit criteria, that may only be brought into conscious awareness in the act of making the discrimination or trying to verbalise the basis for the distinction.
Fransella and Bannister recognised the potential of the triad method to studies in learning science (1977, p.117). Whitelock (1988) has reported using the repertory grid technique as part of a study into 11 and 17 year-olds ideas about motion, and a recent paper by Fetherston (1997) has proposed a learning approach derived from P.C.T.
However those working in science education have generally preferred other techniques such as concept mapping (see §4.9.1), word association and – especially – interviewing (see for example, White and Gunstone, 1992, which might be seen as a manual of methods used in the field). One possible reason is touched upon by Osborne and Wittrock who suggest that many science teachers do not fully trust complex statistical methods (1985, p.63). So science education research within the constructivist tradition has sought to use Kelly's theoretical base to underpin other methods (e.g. see Swift, Watts and Pope, 1983, abstract). The main method employed has been interviews (e.g. Osborne and Wittrock, 1985, p.80).
To summarise my main points
- Kelly's methodology for applying his Personal Construct Theory, involved the presentation of triads of elements, and asking the client to make discriminations.
- This technique was sufficient to elicit a repertory of constructs used by the client in construing his or her world.
- Further analysis could be undertaken by recording the repertory of constructs on a grid, which was suitable for undertaking statistical tests.
- Although Kelly refers to constructs rather than concepts, the distinction is not problematic as Kelly's theory encompasses scientific concepts, as well as distinctions made on affective grounds.
- Although researchers in science education tend to have avoided Kelly's methods when exploring conceptual development, the objections raised concern the analysis of grids, not the Construct Repertory Test (the method of triads) itself.
Kelly's method of triads is then a technique which is appropriate, in a research project which is ground in personal constructivism, for exploring aspects of students' meanings in science – although the literature suggests it has not been widely used for this purpose.
§4.8: Colearner dialogues
§4.8.1: Researcher cuing: a potential difficulty for interviews
The clinical nature of the conversations that take place in interview situations makes them somewhat removed from the normal discourse about science that occurs between students. In particular, the inquisitorial nature of the interviewer's role, imposes a very different social context than when students are working together in a classroom situation.
Solomon has suggested that the construction of knowledge in science classes is very much a social act (see chapter 2, §2.7). Edwards and Mercer have discussed the unwritten rules of classroom discourse, where the aim is often for the students to work out what the teacher thinks the answer to the teacher's question is, to give an appearance of coming to common knowledge through consensus – rather than the teacher just reporting the conclusions to be learnt. Edwards and Mercer's work – based on research in primary, middle and comprehensive schools (Maybin, 1987, p. 171) – show that often teachers' questions are framed and cued in such ways that their purpose is less to test understanding, than to give an impression that the accepted knowledge has been reached through a process of dialogue (see chapter 2, §2.8.2).
In a research interview the researcher aims to take a very different role from the teacher: and rather than lead the respondent to the interviewer's understanding, he or she attempts to find out what the respondent really thinks and understands.
Edwards and Mercer point out that many questions asked by teachers defy the normal social conventions in that the inquirer already knows what the answer is: so when a classroom teacher asks "what is a covalent bond?", the purpose of the question is not to find out what a covalent bond is, but to see if the class can offer an appropriate response. Such questions are often asked at a point in proceedings when the teacher believes the students should be able to answer in an accepted fashion: indeed often when the proceeding teacher talk or question sequence has clearly 'telegraphed' the required answer. As Solomon point out, "teachers' questions are designed to elicit the right answer, if at all possible, because they teach as well as inquire" (Solomon, 1992, p.132).
However, in a research interview, 'the right answer' should be the one which clearly reflects the respondent's own thinking. In an interview situation the researcher may phrase the question in a more honest way (e.g., "so what do you understand by the term a covalent bond?"), and even if not (e.g., "so what is a covalent bond then?") the earlier context of the interview has made it clear the issue is not what is meant by a covalent bond in abstract, but what the particular student understands. Nevertheless, the social context of the interview, may mean that the student is concerned to please by producing the 'right' answer, and may pick-up unintentional cues provided by the interviewer's wording of questions, tone of voice, reaction to the responses etc. Indeed if the interviewer misjudges the knowledge available to the respondent, then the very questions asked could provide information that the student did not previously have available.
Unintentional interviewer cuing may be kept to a minimum by experience of interviewing, knowledge of the respondents, and prior awareness of the likely limits to student knowledge. In the present study:
- I had some previous experience of interviewing (in research for my Master's degree), although this was concerned with finding out about attitudes rather than conceptual understanding;
- before commencing the field work I had undertaken reading of the literature to find out about the range of reported misconceptions and difficulties students could have in the topic of chemical bonding, and in those aspects of school science considered prerequisite;
- I had several years experience of teaching the topic of chemical bonding to A level chemistry classes;
- the respondents were my student-colearners, so that I had my teacher's knowledge of their general performance in chemistry, and their understanding in the subject.
The question of cuing is complicated by the in-depth interview approach used: where the student-colearners were working in their zones of proximal development (§2.2.2) it is necessary to judge where the researcher-colearner provides scaffolding to find out what the student really understands, and where the researcher effectively provides an answer (see §4.10.4 below). Analysis of the interviews suggests that generally the answers given by respondents were not due to unintentional cuing. (In places where it appears I made misjudgments it was possible to ignore the responses during analysis.)
Nevertheless, I decided that it was appropriate to set up situations where some of the colearners would discuss the topic (of chemical bonding) without the opportunity for my cuing their responses. In this way it would be possible to see if the level of knowledge and range of ideas elicited was compatible with that found in the interview situation.
§4.8.2: Group discussions
Gilbert and Pope report an approach to studying learners' ideas in science which involves setting up group discussions. Their aim was to provide a context where learners would develop their thinking about scientific topics (1986a, p.62). Their suggested solution was described as "peer-group discussions, built around a suitable stimulus situation, and organised in such a way as to promote challenges to conceptions" (p.62). They used this approach with middle school pupils in Germany, using groups of 2-3 learners, using a deck of cards designed for the 'interview about instances' technique (§4.6.2) focussing on the concept of energy. They found that the groups would carry out the task, although the quality of discussion depended on group composition (p.75).
They also found that the presence of a researcher in the group has a disproportionate impact, and changed the nature of the discussion to be more like a teaching context (p.74), whereas when the children were left alone the process would elicit a discussion rich in the their own ideas (Pope and Denicolo, 1986, p. 159). Solomon (§2.7) has also collected and discussed samples of data from groups talking about energy-related topics (1992, pp.65, and pp.157).
§4.9: Supplementary material used to support the case studies
My dual teacher-researcher rôle meant that I was working with the student- colearners in a classroom situation. This gave me the opportunity to see work they produced as part of their course. Some of this work was relevant to the topic of chemical bonding, and photocopies were made in these cases.
This material was used to authenticate the interview data. Some points arising from course-work were introduced as talking points in interviews – in such a case the focus of discussion had arisen from a suggestion of some alternative conception or confusion in the student's work, and if such a conception was then reiterated in the interview, it was not just an artifact of the clinical context. Even when the material was not used as a starting point for interview discussions it was kept on file so that a comparison could be made, allowing the opportunity to support or challenge interpretations of interview data. The type of material collected included copies of test scripts, as well as other set tasks, including concept maps (see below, §4.9.1). The validity of using data collected in such an opportunistic manner is considered below (§4.10.7).
In addition, 'incidental' data was collected from other learners, and this included responses to questions set as induction exercises at the start of the course, and to past examination questions used as tests, and student concept maps.
§4.9.1: Concept maps
"A concept map aims to show how someone sees the relations between things, ideas or people."
White and Gunstone, 1992, p.15
The concept map is a way of representing knowledge. In a concept map material is organised so that the key words, representing the concepts of the topic, appear highlighted in boxes at the nodes of the map. The relationships between the concepts are represented by connecting lines. Each line stands for a proposition relating two concepts. Unlike a linear text there is no single intended order for 'reading' the map: it is a network of ideas that may be sequenced in many permutations. The figure below is an example of a concept map for the concept of 'concept map'.
Figure 4.2: A concept map for concept map. (Taken from Taber, 1994)
The technique of concept mapping has been much discussed in the literature, as a learning tool and an assessment technique, and also as a classroom based means of diagnosing aspects of a learner's cognitive structure (Al-Kunifed and Wandersee, 1990; Edwards and Fraser, 1983; Watts, 1988).
Gilbert and Watts have suggested that research to explore learners' ideas should involve "mapping the 'topography' of local domains of understanding" (1983, p.66, my emphasis) and "charting changes in frames of reference so that the durability, stability, coherence and consistency of conceptual constructions become the point of departure" (p.67, my emphasis), and concept mapping would seem an idea approach.
I have used concept maps as a learning technique for some years, and so asking students in my chemistry classes to produce a concept map to summarise a topic area was a normal part of my teaching repertoire. Indeed, the figure above is taken from an article about concept mapping where I demonstrated that concept mapping could simultaneously lead to diagnosis of student 'misconceptions', be an activity that students enjoy, and could encourage metacognition (Taber 1994).
The article described how a group of A level physics revision students were asked to produce a concept map for the topic of energy. A number of 'misconceptions' were elicited, including a suggestion that "some chemicals when combusted give out energy in the form of heat due to the breaking of bonds". On this occasion the students were asked to jot down their feelings about the exercise and their comments were generally positive. The students recognised how the activity tested their knowledge ("I think this exercise was useful as it let me know exactly how much I know about energy, which I can now see is not enough"), and provided them insights about their current understanding ("my knowledge of physics is very un- organised at present"). The open ended nature of the activity encouraged students to explore their thinking of the topic ("at first I did not know where to start but as I began putting ideas down, it reminded me of other points"; "[I] did not think I would have been able to think of enough things after 3 months but found things start to come back once I started writing"), and the act of organising their knowledge into a map seemed to provide a learning experience ("I didn't realise how much the different areas interlinked"; "quite useful, brings back memories; good to see how well topics relate or how well you can interrelate them"). Indeed as one student explained,
"I found I was digging around, trying to put fragments of things I could remember together. I found I could remember only scraps of information, but when doing the drawing [sic], saw how things pieced together, and linked with other things"
Therefore, in my own experience, concept mapping is a technique which can be used to survey a group's initial ideas about a topic, or to evaluate aspects of their learning; but it is also a technique which is useful for students as it enables them to make explicit connections ("I saw how things pieced together, and linked with other things"), and to evaluate their own progress ("it let me know exactly how much I know"). As opposed to a test where the format encourages elicitation of information in a specific order, the concept map allows the student to access ideas in a much more flexible way.
In the present research therefore, I was able to ask students to produce concept maps of topics at various stages in their course, using the information for my research, without compromising my 'teacher' role, and the students' right to expect me to set work for pedagogic purposes.
§4.10: Issues of authenticity, and generalisability
In a sense, this entire chapter is largely concerned with issues of the authenticity and generalisability of the research reported in this thesis. In this final section of the chapter some specific concerns will be discussed.
§4.10.1: Notions of reliability, validity and authenticity
Any researcher presenting results to the public domain has to be prepared to defend the integrity of the methods of data collection and analysis used in that research. Traditionally the notions of validity and reliability are considered to be of great importance. Essentially a research instrument is reliable if it provides reproducible results, and is valid if it measures what it claims to measure. For example in the present study, if one of the foci diagrams used to elicit learners' ideas about bonding was intended to represent a certain type of bond, but was not recognised as such by learners who would recognise other (e.g. text book) representations, then this would not be a valid probe.
Reliability is a more problematic criterion in the context of the present research. With some quantitative instruments reliability may be gauged by finding the test- retest coefficient – but this is obviously not appropriate in the present research, if only because student answers are expected to change over time: that is part of what is being studied. It is also inappropriate because question phrasing and ordering are flexible, and the sample size would be too small.
Within an interview the reliability of the researcher's interpretation of a colearner's comment may be checked (see appendix 11 for examples from the case of Annie) in a number of ways, including:
- confirming responses by repeating or rephrasing questions (see appendix 11, §A11.1);
- clarifying ideas by asking follow-up questions (§A11.2);
- paraphrasing what one believes to be the colearner's argument, and seeking confirmation (§A11.3);
- returning to the same point in the same context later in the interview, to see if a consistent response is given by the colearner (§A11.4);
- approaching the same point through a different context later in the interview, to see if the colearner gives a consistent response in the different contexts (§A11.4).
However this does not mean that obtaining an apparently contrary response necessarily invalidates the data, and it should therefore be automatically discounted. Part of what is being investigated is the stability and lability of a student's ideas (e.g. see chapter 7). If a colearner's thinking is very labile, then this is an important feature which should be recorded in the results. If the colearner is operating with multiple alternative frameworks (see chapter 2, §2.5.2 and §2.9) then essentially the same question, but phrased slightly differently – or just cued differently by the particular examples and ideas discussed immediately prior to the question – may well lead to inconsistent responses without suggesting that the technique and data are unreliable. Indeed when an apparently inconsistent response is obtained it is possible to then ask about the earlier comment, and often this is most revealing. In the type of approach to data collection and analysis taken here, the negative case may be used to explore and test the range of application of ideas (Walker, 1993, p. 177, and see §4.6.1) and thus to refine categories and concepts.
The difficulty of applying the criterion of reliability to interpretative research, undertaken using qualitative methods, has led to the suggestion that it is more appropriate to judge such enquiry by notions of authenticity. The researcher's presented results (i.e. interpretations) should be true to the data (Pope and Denicolo, 1986, p.156, see also §2.4.2, §4.2.5, §4.4.5 and §4.10.6 regarding the danger of 'framework spotting'). Appendices 26 and 27 provide an insight into the analytical decisions made in the present research. Appendix 26 gives an example of how transcripts from interviews with colearner Annie were edited in preparing a study of her case (outlined in chapter 7). Appendix 27 shows how preparing Tajinder's case study (summarised in chapter 8) involved a number of stages of analysis (§A27.0), integrating data from interviews and other sources. This appendix shows how the information in an extract from one interview transcript (§A27.1), is used in preparing a narrative summary of that interview (§A27.2), and then a thematic summary reorganising the material according to the topic (§A27.3), which was then used to prepare the case study document itself.
The need to produce authentic accounts from the data collected has resulted in the decision to concentrate on two colearners when presenting results: one from the first cohort of colearners (see chapter 7), and one from the main cohort (see chapter 8). The limitations of "time and concentration from interested readers and space" (Pope and Denicolo, 1986, p.156) meant that in-depth case studies of all the colearners for whom sufficient data was available would have led to a tome of unacceptable size. The material from the other colearners is treated in less depth in this thesis. However, as explained above, this is appropriate in grounded theory. The case studies stand in their own right as part of an idiographic tradition of research into individual learners, but they also form the early stages of the analysis which led to a model (of A level students developing thinking about chemical bonding, see chapter 6) which is grounded in the wider database.
§4.10.2: Information flow in the interviews
The purpose of the interview is for the researcher to find out about the ideas of the colearner, rather than to impart information to him or her. It would be possible to consider a model for interviewing where the researcher collects data from a student, without releasing any substantive information. (This does not mean such an approach would be possible in practice, as any question is 'loaded' with assumptions.) In the present study such a model was not followed: information was sometimes deliberately given to the colearners. This intention was related to both ethical and methodological implications.
For ethical reasons (§10.4.3), it was sometimes appropriate to provide feedback at the end of a research session, to highlight errors or limitations in the colearner's thinking. If the feedback was effective, then subsequent interviews would demonstrate a different level of understanding than would have been present without such intervention. This means that the development of understanding of chemical bonding amongst the colearners might be 'accelerated' compared to 'typical' students. However, as the study was not intended to survey a representative sample of A level chemistry learners, this was considered quite acceptable.
In methodological terms (§10.4.4), I was aware that my foci diagrams, the thrust and wording of my questions, my reactions to student responses (including the decision how to follow-up responses with further questions) all provided cues that moved my clinical interviews away from a purely naturalistic study of student thinking. Given that my research techniques were interventions, I had to use my judgment to decide how much information to give, to best probe a colearner's thinking, so that my data reflected the student-colearner's ideas rather than my own. When analysing interview data the cues provided to colearners were taken into account when drawing inferences from their comments.
§4.10.3: The ethical perspective: the researcher’s duty to inform colearners
It was explained above that it was felt important that the research interview sessions were of value to the colearners, and were perceived to be of value by them (§4.3.2). Colearners were volunteers, and could only be expected to volunteer again on a subsequent occasion if they found the experience rewarding. (Indeed, even if a colearner appreciated the value of the session, they might not want to repeat it, if they found it particularly stressful.) My own role for these young people was primarily as one of their lecturers, and they rightfully had expectations of me in that role.
During the early phase of the project I attempted to conceptualise my teacher and researcher roles as largely separate: I would teach the students in class, but enquire into their understanding during research interviews. I hoped that the colearners would benefit from the conversations, and would become aware of their own areas of difficulty. As I was interested in observing the development of student understanding, I had reservations about actively 'teaching' colearners through my research. However, the use of clinical interviews was already a form of 'intervention' (so that an anthropological model of the research was already inappropriate), and the insights that might be gained from research interviews would be likely to influence my behaviour as a teacher (which, after all, is the purpose of undertaking educational research, see §4.1.2). In a sense, my decision to take on the dual teacher-researcher role, and research my own learners had placed my study within an action research frame. The duty I owed my colearners, as part of the implicit contract for taking part in the study, demanded that they should be informed about their 'performance' in interview (§4.3.2).
The colearners were understandably concerned to know 'how well they did' and 'what they got wrong' during an interview. This second question was not so easy to answer, and I explained that it was often not so much 'getting things wrong', as a question of ideas becoming more sophisticated over time. Whilst this was true, it was clear the colearners felt they needed more specific feedback, and they were able to ask me for this as I was their teacher. Sometimes the interviews acted as a learning experience for the colearner, because by being led through the consequences of an idea he or she comes to find that it is wrong, and moreover to see for herself why it is wrong. In these cases the 'lesson' may be well learnt. On other occasions the interviewer's questions may not lead to the colearner contradicting his or her own knowledge, or producing an argument that he or she considers inconsistent or circular or teleological or ridiculous. (The student- colearner, of course, may not posses the same criteria for a good explanation as the researcher-colearner.) It is on these occasions that feedback is required at the end of the interview.
Annie's interviews revealed she had an alternative conception of the meaning of electrostatic charge symbols (see chapter 7, especially §7.2.2). Whilst the full extent of the implications of her alternative notion only came clear on detailed analysis of the transcripts after she had finished her course, it was apparent at the time of the early interviews that Annie had some problem understanding this aspect of curricular science. During this phase of the project a sequence of three interviews was planned for each colearner, and during the final interviews I was aware of the provisional analysis from the previous two sessions. During Annie's third interview – shortly before her final examination – it became clear that she still held the 'incorrect' notions I had identified after her second interview. At this point, it seemed appropriate that – as her teacher – I should provide a detailed debriefing after the interview on this aspect of her ideas. We also agreed to an additional fourth session soon afterwards to see if Annie had been able to adjust her thinking.
In retrospect it became clear that such a debriefing should be provided for all colearners immediately after an interview, and this was included in the second phase of the research (interviews with Edward, and the second main cohort, i.e. Jagdish onwards). Although I was tape-recording the interviews for later analysis, I decided I needed to respond to the request for specific feedback, and changed my procedures for the main study, to include brief contemporaneous notes of points that should be fed back at the end of the interview. In this way the actual research design was changed in response to the colearner's needs – and what they were entitled to expect.
§4.10.4: The methodological perspective: working in the zone of proximal development
During the first stage of the study (January 1991 – May 1992) it became clear that the research interview could be a learning experience for the student-colearner in terms of giving the students the opportunity to learn through the process of talking through their ideas. This observation is consistent with the literature reviewed in chapter 2, where Vygotsky's ideas of words-as-tools, and the Z.P.D. (§2.2.2), and Edwards and Mercer's studies of classroom discourse (§2.8.2) were discussed. Bruner and Haste discuss how one function of discourse may be as scaffolding (1987, p.21), and in a similar vein Gallimore and Tharp (1990, p.181) explain how questioning may assist performance. Bruner (§2.2.3) has described how language acts as a cognitive instrument for representing and transforming experience (1977, p.210).
Perhaps the most dramatic example of this was one of the colearners in the main cohort, who was able to explain the nature of van der Waals' forces (although she did not have a name for them) – even though this was not a topic that had been studied in class. After working through her argument the colearner's own reaction was surprise – this was something she had not 'known' until that point, and she apparently constructed this knowledge from what she did know in situ in front of researcher and tape-recorder (see appendix 8). The in-depth nature of the interviews is similar in form to the Socratic/Platonic method of dialectic, where, "one pursues a topic with inexorable patience, by questioning and answering, and one shakes it and pulls it this way and that…" (Egan, 1983, p.32).
To my own mind the interviews gave an opportunity for colearners to explore and 'play with' scientific ideas in a way that I thought it was unlikely to occur otherwise: the method was 'invasive' and therefore the research act itself formed an important part of the context of the conceptual development being explored (e.g. Pope and Denicolo, 1986, pp.154-155).
Vygotsky's idea of the Z.P.D. (zone of proximal development, see chapter 2, §2.2.2) is very relevant to what actually occurs during interviews: the conditions in the interview seems to be just those where Vygotsky would expect learning to take place – what Kozulin has described as "the place at which a [learner]'s empirically rich but disorganised spontaneous concepts 'meet' the systematicity and logic of adult reasoning" (1986, p.xxxv).
Driver and her coworkers have actually suggested that intervention should be a deliberate part of any study looking at how ideas change,
"Longitudinal studies, which track the the evolving conceptions of individual students over extended periods of time, can provide detailed information about learning routes and enable features of students' developing knowledge to be characterised, but do not necessarily provide information about what prompts change and how it occurs. In order to obtain information of this kind, it is necessary to study students' learning in science as a consequence of specific interventions."
Driver et al., 1994b, p.85
Undertaking research interviews is a skilled process: in order to be able to claim that a respondent's utterance gives evidence for the presence of a successful understanding or the use of an alternative framework the interviewer has to be aware of the whole context of the utterance within the interview. An earlier comment by the interviewer, or a leading question, may undermine (or alternatively illuminate) the evidence. The interviewer should not 'correct' or offer evaluation of the respondent's comments while trying to explore her or his thinking. Indeed the order of questions and the use of particular words that may cue or trigger a response needs to be carefully thought through. For example Tomlinson (1989) has discussed how the the interviewer can use 'hierarchical focussing' to maximise the amount the respondent spontaneously introduces, whilst ensuring that the interviewer's full agenda is covered. In the present study the use of a sequence of prepared focal diagrams (with the possibility of changing the sequence as required) was used to address the research agenda. As interviewer, I had to decide at all times how much cuing to provide through the questions asked. These decisions were made in response to the colearner's comments and behaviour in the interview, as well as my knowledge of earlier interviews with the same and other colearners.
Although the main data collection technique used in this study has been the interview, the authenticity of the results has been enhanced by methodological triangulation – using the construct repertory test as an alternative approach to elicitation, and with alternative foci; collecting samples of the student-colearners' talk in peer discussions; and collecting course-work such as concept maps (§4.5). Data has also been collected from learners other than the colearners who undertook the interviews that form the principle data source. The importance of this form of triangulation to interpretive inquiry has been emphasised by Guba,
"The naturalistic investigator is concerned with description and understanding; thus, he begins as an anthropologist might begin learning about a strange culture, by immersing himself in the investigation with as open a mind as possible, and permitting impressions to emerge. As impressions are formed, he checks them out by various means, e.g. 'triangulation', testing one source against another until he is satisfied that his interpretation is valid."
Guba, 1978, quoted in Gilbert and Pope, 1986b, pp.41-42
§4.10.5: Case studies and generalisability
"Any theories which are developed must be grounded in the actual data the researcher has. This results in interpretative researchers only moving from discussion of individual cases to wider, general situations when they have achieved a close, detailed description and explanation of those individual cases."
Hitchcock and Hughes, 1989, p.29
One criticism that has been made of case studies is their "limited reliability and validity" as representative of a wider class of cases, that is what has been labelled the generalisation problem (Walker, 1993, p.166).
However, a case study is by definition not intended to produce findings to be generalised: the whole purpose of carrying out case studies is to allow the in-depth study of the particulars of an individual case (Hitchcock and Hughes, 1989, p.32). That is not to say that a case study can not illuminate other cases: it provides insights which allow other cases to be approached in a more informed way – both in terms of data collection and analytical approaches, and in terms of hypotheses against which the researcher may test the data. The case study of Annie presented in this thesis is about the individual (labelled) Annie, rather than some general notion of 'the A level chemistry student'. Tajinder is another individual, and his case study attempts to describe his progress in understanding chemical bonding, rather than Annie's, or 'the A level chemistry student' in general.
Although such case studies are fascinating, and – because of their longitudinal and in-depth nature – are illuminating to others looking at conceptual development, they do not allow general conclusions to be drawn about difficulties in learning chemistry; general conclusions that could form the basis of useful pedagogic advice to science teachers. This would require an analysis of data from a larger sample of learners. This is why, in the present study, case studies have been used as part of a strategy of analysis to produce grounded theory, rather than simply as an end in themselves. Hitchcock and Hughes warn that before case studies can be used as the basis for developing a wider view the researcher needs to have an in-depth understanding of those cases (motto above): by presenting summaries of the individual cases of Annie (chapter 7) and Tajinder (chapter 8) I am demonstrating that I "have achieved a close, detailed description and explanation of those individual cases".
§4.10.6: Framework spotting
An individual learner has a unique cognitive structure (§1.4), and in idiographic research (§4.1) it is possible to produce descriptions of (the researcher's models) of aspects of an individual's ideas: in the present thesis chapters 7 and 8 do just this. However, it is also possible to produce models which act as descriptions of common patterns in learners' ideas (see chapters 6 and 9 to 11). In chapter 2 (§2.4.2) this distinction was described in terms of alternative frameworks1 and alternative frameworks2. Kuiper was criticised for ignoring this distinction, and consequently undertaking what Pope and Denicolo have described as "a 'framework spotting' exercise" (1986, p.157).
However, providing that the status and derivation of proposed models of common aspects of learners' thinking are understood, such 'products' can be of value to teachers. Awareness of the range of likely alternative conceptions members of a class may hold is useful before setting out to teach a topic (for example to overcome potential substantive learning impediments, §1.5.3). Understanding of any common features to students' alternative frameworks for a concept area, can allow to the teacher to be sensitive to the possible meanings that students may have for the words they use, and allow the teacher to make sense of otherwise unclear student statements and questions. The teacher who wishes to build a common understanding with a class will need to learn the students' 'language' for the topic: and research results may act as a useful starting point. For a teacher committed to a constructivist perspective, it is essential to know what foundations are available for building new knowledge (c.f. §2.3.9 and §.2.3.10).
In the present research then I considered it important to draw out from the data features that may be significant for teachers approaching the topic of chemical bonding (see chapter 12, §12.5).
§4.10.7: The validity of opportunistic data collection: anecdotal evidence?
Research data is normally collected in a systematic way: it is expected that an enquiry will be to a purpose, and will be planned accordingly. However, in the present study it could be claimed that some of the supplementary data was collected 'opportunistically' (c.f. 'accidental sampling', Cohen and Manion, 1989, p. 103).
By 'opportunistic' data collection, I am referring to taking the opportunity to collect data that might be relevant to a study, as one 'stumbles across it'. This process is not planned, is not systematic, does not provide any control over 'variables', and does not allow any gauge of the representative nature of the data collected. The term that might often be used is 'anecdotal evidence', and it may be suggested that such material has no role to play in research.
Here I will argue that the distinction between data collected systematically and opportunistically is not a dichotomy, but rather there is a continuum that is related to the research paradigm in which the research is ground (§4.1). From within the experimental research tradition, data would be collected in a preplanned and predetermined manner: the number of data, as well as the nature of each datum are tightly regulated. Variables are controlled, data is often quantitative, and statistical analysis may well be appropriate.
The interpretative research tradition does not usually share these features. For example, in ethnographic research, the aim is for the researcher to immerse him or herself in the culture being researched, and to observe, and to attempt to understand the meaning that others give to their actions. The greater the preplanning of how and when data would be collected, and the nature of the data to be collected, the more assumptions are being made about the culture which is being explored. Yet the nature of such research actually requires the researcher to attempt to minimise preconceptions, and to limit his or her reliance on existing theoretical frameworks. It is accepted that in such research the categories by which data are recognised and analysed 'emerge' from the data (c.f. §4.4). The researcher role is to enter the correct context, and to be as open-minded about what is seen and heard as possible. To a lesser extent, the choice of semi-structured interview techniques, reflects an ethnographic approach (§4.1.1). Although the context is clinical, and prepared probes are used, the interviewer attempts to listen carefully without preconceptions to the informant, and to enter into his or her 'language community', and explore the topic from his or her world-view, rather than to simply judge her comments against the researcher's own constructs.
In this study I supplemented data from the clinical research, with 'incidental' data collection of samples of student course work. This could be described as 'opportunistic' data collection. However, it was data collected from a context – work done as part of an A level chemistry course – which was more directly relevant to the normal context of learning A level chemistry than my research interviews. It could be argued that this was the more authentic data. The data was collected in a less systematic way than the student utterances in interviews, or constructs elicited from triads. However, the data was not random: apart from the relevance of the context of the work (an A level chemistry course), I collected material that my own analysis of the topic area suggested would be of value to the study.
I kept notes of some student comments in tests, homework, etc., that were judged to inform my own understanding of the range of student ideas in certain topics (i.e. I have undertaken 'theoretical sampling', see above). This 'field data' supplemented the more systematically collected data, and was used to refine the categories which emerge from analysis of the interview data (c.f. §4.4.4). This ethnographic approach has been used effectively by Driver. In her seminal book, The Pupil as Scientist? (1983) she describes in the preface how "examples of pupils' dialogue and written work … were collected while making a study over a 4-month period of a science class", and how "further examples", mostly from classes she had observed and taught, were also used.
In their discussion of grounded theory (§4.4) Glaser and Strauss explain how anecdotal data may be admitted when generating theory. For them the anecdotal comparison provides one slice of data (§4.4.2), which can be used as part of the process of triangulating between different data sources (1967, p.67).
§4.10.8: Diagnostic Instruments
Whilst opportunistic data collection can provide a useful source of data from the 'normal' culture of science learning, against which to authenticate data collected from interventions, the only reliable way to gauge the degree to which specific student conceptions are widespread is to conduct a survey using a significant sample size.
The model developed in this thesis (presented in chapter 6, illustrated in chapter 7-11, and discussed in chapter 12), was grounded in data collected by research in the 'interpretative' tradition (§4.1). It is accepted that the results of such grounded theory "may later be verified through traditional logico-deductive methods" (Charmaz, 1995, p.48, §4.4). In the present study the thesis is largely concerned with presenting, and providing evidence to support, my model of A level student developing understanding of chemical bonding. The present study did not set out to provide a large, representative survey of chemistry learners.
However, some attempt was made to demonstrate how this might be done. Two 'diagnostic instruments' were prepared, piloted and then presented to a moderately large number of learners, based on aspects of the findings from the interview studies. These two instruments were:
- The truth about ionic bonding diagnostic instrument (see appendix 2);
- The truth about ionisation energy diagnostic instrument (see appendix 3).
These instruments demonstrate how aspects of the model developed in this thesis may be used to provide diagnostic tests for classroom use, and in doing so provide the potential for testing the model against a large sample of learners. The outcomes of this limited survey (described in the appendices) have been noted at the relevant points in chapters 10 and 11. In this feasibility study these instruments were used with a sample of learners that was an order of magnitude larger than the number of colearners contributing to the research: in principle they could have been applied to a much larger sample size still.
