Keith S. Taber
Faculty of Education - University of Cambridge

Seminar presentation to
Cambridge International Examinations, University of Cambridge Local Examinations Syndicate, July 2003

Abstract: it is now common practice in many examinations to use questions which are structured, and set in a context. This paper explores the consequences of these trends. Structure helps the candidates to know what the examiner requires, and so helps them identify which specific knowledge they need to use to answer a question. It is clear, however, that the demands of a structured question are different from a more traditional examination question seeking to elicit the same knowledge. Context is generally assumed also to be for the benefit of the candidate, as it will make a question about an abstract topic more concrete and familiar. It is argued in this paper that although the aim of contextualising questions is well meant, there may be good reasons for being suspicious of many contextualised questions. Context-bound questions may actually reduce item validity, without necessarily helping the candidates. Although it should not be concluded that structure and context are necessarily inappropriate in examinations, this paper sets out some of the issues which need to be considered when designing summative assessments.

Introduction

In recent years there has been a tendency to change the nature of the types of questions used in examinations. Open-ended (e.g. essay-type) and objective (e.g. multiple-choice) question types have tended to be replaced by structured questions (Barker, 2001). Additionally there has been an increasing tendency deliberately to embed examination questions within an everyday or other 'relevant' context. This paper explores these trends in the light of what we know about human learning, and the purpose of summative assessment.

The move towards more structured questions can be defended from several perspectives, but clearly changes the demands on candidates compared with the more traditional question styles. At one level examination boards may claim that these different types of question still test whether the candidate has access to the same knowledge, but this does not mean that the different types of question are equally difficult. This raises the issue of what cognitive skills we wish to teach and assess in science.

This paper also argues that the increase in contextualised questions in examinations should be viewed critically. This is not a simple issue, and it seems unlikely that it would be possible to judge context in exams. as simply 'good' or 'bad'. However, it is argued that there are a number of key issues which need to be considered when judging a 'good' examination question, and there are strong reasons to be suspicious of too many context-bound questions.

Before reaching any conclusions it is appropriate to review both the purposes of assessment and the processes that learners must engage in to answer an examination question. Only by keeping in mind what exam. questions are for, and how candidates respond to them, can we judge the appropriateness of any particular type of question.

During their school career students will be assessed frequently, and by a wide range of means. If we take a simplistic view of schooling as designed to facilitate learning, we can see the purpose of assessment as to judge what learning has taken place.

It is generally recognised that there are different purposes for assessment, and - in particular - assessment is often classed as diagnostic, formative and summative. Quite rightly, in recent years, teachers have been asked to focus on formative assessment, or 'assessment for learning' (e.g. Sorenson, 2000). As the role of the teacher is to bring about learning, it makes sense if teachers see the primary purpose of their assessment procedures to be an integral part of a cycle of activities that facilitate that learning.

Traditionally teachers may have primarily judged learning through formally assessed end-of-topic tests, but if assessment is to support learning then it is important that it is continuous so that it can inform the on-going management of learning (i.e. teaching) in the classroom.

It is also appropriate, if we wish to judge the learning that has taken place (and so the effectiveness of teaching) that we test at the start of a topic as well as at the end. Without a benchmark it is difficult to judge whether any learning displayed in an end-of-topic test has taken place during, rather than before, the teaching of the topic. Indeed, a pre-test could in principle suggest that whole sections of a scheme of work will be repeating material that is already well-known. In this case the assessment will inform a reorganisation of teaching that will allow time to be used more effectively to ensure genuine student progression, and avoid the frustration of students who recognise that they are repeating work.

Of course pre-testing is more than bench-marking - and this is where the notion of diagnostic testing may become especially important. Research into the learning process, and particularly learning in science, provides two important justifications for diagnostic assessment. Firstly, research into children's ideas in science has revealed a wide range of common alternative conceptions and frameworks that students develop across a range of science topics and key stages. Diagnosing the extent to which known alternative conceptions are present in a particular teaching group can guide the teacher in the extent to which such ideas need to be challenged on a class or individual basis (Taber, 2002).

The second justification concerns the notion of 'pre-requisite' knowledge. Learning is constrained by the limitations of the students' 'cognitive apparatus', the features of the brain which bring about learning (Taber, 2000). New material to be learnt needs to be of limited complexity, and readily relatable to existing knowledge, if it is to be understood as intended and retained by learners. In effect this means that teachers need to undertake content analyses of topics to identify the conceptual foundations upon which new learning will logically build (Taber, 2002). If this pre-requisite learning is not present then there is little chance of building-up the more advanced ideas. Pre-testing can diagnose deficits in the required prior learning, and so inform teachers that remedial work is needed (with the whole class, or some individuals) before setting out upon the intended new learning.

Summative assessment - examinations, such as GCSE and A level examinations - is not primarily designed to be diagnostic or formative. The purpose of an examination is to be summative, to allow a judgement to be made about the candidates' learning in a subject at a certain level. In the UK system at the current time examination results are also used to judge the performance of schools and colleges (and their teachers). For this purpose a measure of value-added is often seen appropriate - such as looking at A level grades in the light of the students' GCSE grades on entry to a course. The 'output' grades of one cohort may be compared against another with similar 'input' characteristics. However, this process is largely a statistical one, and unlike the use of pre-tests there is no attempt to evaluate the specific features of teaching in a way that could directly inform professional practice. (In this sense, even as an evaluation tool the process is summative rather than formative - the teacher is given an indication of how well they are teaching, but no indication of what they might be doing particularly well, or where they might specifically be going wrong!)

Now if different school assessments are seen to have different purposes, then we might expect them to be designed to be optimally matched to their purposes - that diagnostic, formative and summative assessments may look very different. To some extent this is true, but this paper will raise some serious questions about the nature of many questions currently being used for summative assessment in formal examinations. Before discussing this in any depth, it is helpful to briefly review the widely accepted features of 'learning theory' which inform teaching (and so should inform the assessment of learning).

Learning is a natural process - all human beings learn a great deal effortlessly. However, as all teachers know, directing specific learning is a much more hit-and-miss affair. Even motivated and interested learners being taught by keen and well-prepared teachers often fail to learn what the teacher thought she was teaching! When we examine the nature of the learning process, this does not seem surprising.

Learning is in effect a change of behaviour (e.g. answering a question) brought about by experience (such as school lessons). This change is understood to result from some permanent modifications in the brain in terms of the strength of connections between nerve cells. As we have little detailed understanding of how knowledge is coded in synapses, it is usually more helpful to talk in more abstract terms about 'cognitive structure', meaning the organisation of knowledge in the brain. The conjecture that our conceptual knowledge is somehow represented in the brain is widely accepted even though we currently have limited understanding of the way that knowledge is coded. It may be helpful to think of cognitive structure as a mental concept map of everything that is known (or believed). So for the purposes of useful discussion we can consider (conceptual) learning to involve changes in the students' mental concept map.

Now what is known from a great deal of research is that learning is a highly contingent process. In other words whether a student changes a particular region of their mental concept map, and how they change it, depend upon a range of influences (Taber, 2000). Even assuming that the student is paying full attention to the teacher, the clearest possible teacher explanation will not necessarily bring about the desired 'change in mind' on behalf of the student.

In order for the intended learning to take place the learner has to make sense of the teacher's exposition, recognise how it relates to existing knowledge, and be convinced (though not necessarily consciously) that a change in the concept map is justified. Assuming these points, then there may be temporary changes in the brain (akin to self-sustaining cycles of electrical activity) that can potentially later be 'fixed' in terms of permanent modifications to patterns of synaptic connections.

However, these criteria may not be readily met. For the student to understand the teacher it is necessary both for them to share common language (when research shows that students and teachers often have different meanings for both technical and non-technical words used in lessons) and the teachers' exposition to be simple enough for the student to 'take it in'. Now this latter point is quite a significant one, as research suggests that we all have quite limited 'working memories' and that information of quite moderate complexity may readily exceed our processing capacities.

Even when the teacher's language is not a barrier to understanding (and experienced teachers become quite expert at using appropriate language for their students), the teacher's exposition will be aimed at specific target knowledge within the learners' cognitive structure. So the student has to recognise which 'bit' of the concept map is being addressed. As most students have a whole range of alternative conceptions, it is often the case the learner's mental concept map is actually quite different from that idealised version envisaged by the teacher (thus the value of diagnostic testing), and so even a clear teacher explanation may not map onto the learner's existing cognitive structure as intended.

There is much research to show that often learners are quite resistant to some new ideas they meet. Sometimes these ideas seem quite 'counter-intuitive' (perhaps they do not match prior learning moderated by previous experience). Ideas that seem convincing to the teacher (who is very familiar with the topic area), may seem arbitrary and even unlikely to the learner. Human learning has a logical feature in that the status of new ideas is judged according to how well they match existing ideas and our assumptions about the way the world works - and sometimes the science we teach does not score highly on these criteria when evaluated (usually subconsciously) by learners. Ideas that are considered to be simplifying when we (teachers) see the wider picture, may just seem to make things more complicated when first met by the learner.

Even when the student has understood the teacher's language, and not found the exposition too complex, and related the teaching to the appropriate bit of prior learning (where it matches the target knowledge sufficiently well), and when the new ideas have sufficient status, learning is still often a slow process. The permanent modifications in brain structure which change the mental concept map, usually start some hours after the initial learning (Taber, in preparation).

It is believed this largely occurs during sleep - which means that the new science knowledge competes with all the other temporary memories from that day. Of course, most of the temporary traces have weakened considerably by bed-time, and so many presumably fade away before there is any processing into permanent traces. Even where the information from short-term memory does get 'fixed' this is not an all-or-nothing process. Research suggests that the consolidation of new memory traces (with any necessary shifting of existing links in the mental concept map) is a process that occurs over a period of many months. As each night the brain has other new material to consider, many new memories do not get sufficiently consolidated to be readily available to the learner once a few days are passed. (Of course, teachers respond to this by constantly revisiting new material to encourage the consolidation process.)

Given the extent to which the intended learning is highly contingent on so many factors not fully within the teacher's control, it almost seems surprising we manage to teach any science. In fact new teachers are often quite surprised at how little new 'theory' can be effectively introduced in a lesson, and how much of the previous lesson seems to be forgotten by the time the students come along for their next class.

However, our ideas about how learning is contingent can inform teachers on how to facilitate learning. As a starting point, teachers need to be clear in their own minds about the prior learning which will act as the foundations for constructing new learning, and then use diagnostic assessment to check this knowledge is present. Although there are written probes available designed to identify specific common alternative conceptions, informal assessment through questioning the class at the start of an exposition can often be an effective approach.

These questions can also perform another important function. As the intended learning will only take place when the learner recognises which part of their mental concept map is being addressed, such questions can also direct the learners' attention to the relevant prior learning. This can be seen to be part of the process of 'scaffolding' learning. As learners have limited capacity working memories to help process new learning, and as new ideas may seen unconvincing until they can be viewed within their wider theoretical context, the task of convincing students to take on new ideas can be quite challenging.

Teachers need to highlight the relevant prior learning, and help learners re-organise it into the most suitable configuration for supporting new learning. The teacher, through a skilful exposition, provides a structure through which the learner can come to see how the new ideas 'work' in a comprehensible (e.g. familiar) context (Taber, 2003). The skill of the teacher is often to find suitable links with what learners already find familiar to anchor new knowledge, as they help students build up new meanings (Ausubel, 2000).

Scaffolding learning means the teacher supports students' learning by helping them build up their knowledge. While the learning is still fragile it relies upon the scaffold provided by the teacher. Once the learning become robust (once it has gained inherent structural integrity through sufficient consolidation) the scaffolding may be removed and the new learning will stand by itself. At this point the novel has become familiar, and in time may act as the foundations for a further cycle of new learning.

Much of the work that teachers set in their classes (and for homework) can be seen as providing formative assessment opportunities. Learning is best facilitated by asking students to interact with learning materials - rather than just copy notes - so that their minds are engaged in thinking about the science. The level of demand of such tasks should be such that students are challenged but not de-motivated (a requirement that makes differentiation a real issue for the teacher). Too much challenge leads to frustration and incomplete or incorrect work; insufficient challenge (such as sets of repetitive exercises that provide little more than 'drill') leads to boredom, and work that has not facilitated any learning.

This means that in monitoring student work the teacher would expect to see most work correctly completed, but in monitoring students working the teacher would hope to see signs of pausing, thinking, correcting and discussing, prior to completing the tasks.

These formative tasks are usually designed so that there is sufficient structure to help students succeed. Material may be ordered so that a logical train of argument is established, a variety of hints and cues may be provided to indicate which ideas should be applied, and indeed many tasks provide a largely complete structure only requiring the learner to take a few modest steps to provide closure. Many 'DARTS' (directed activities related to text) are of this form: labelling a provided diagram, sequencing the given steps, interpreting information from one form to another, fitting a few missing words into a brief text. The task helps scaffold the students' intended success.

Another reason why structured questions are popular in teaching is that they have increasingly come to reflect the types of questions set in examinations. Several decades ago formal school leaving examinations tended to have two main type of question: open essay type questions and objective questions (such as multiple choice, and assertion-reason questions). Objective questions allowed guessing, but over a whole paper of - say - 50 objective questions the effect of pure guesswork was considered negligible. Objective questions had the advantage of high reliability (marking was easy), and there were well developed statistical techniques to pre-test items to check for validity (e.g. to eliminate items which the most able got wrong by spotting a subtle complication missed by everyone else). However, it was found that girls were (on the whole) less confident in, and less successful with, objective tests.

Open - 'essay' type - questions are very challenging as they require students not only to recall a great deal of information about a topic, but also to select the most relevant aspects (in the light of the particular question) and to organise the material into a logical and coherent response. This tests more than just 'knowing' the right information - and therefore means that it is quite possible for a candidate who does 'know the answer' to fail to select the right knowledge from memory and organise it in a manner judged worthy of merit. At the present time, in the UK, very few public examination questions are like this. Instead questions are structured so that the candidate who knows the right answer should recognise which information is needed.

Furthermore, as each part of a question is relatively discrete, failing to answer one part of a question does not necessarily act as an impediment to gaining credit on subsequent parts. In addition, the mark allocations are often given (or may be guessed at fairly accurately) suggesting how much information is required at each point. On an essay-type question, by contrast, is it quite possible for a candidate to spend considerable time 'off-track' without any indication that there will be no credit for a lengthy passage of prose.

Despite these differences, it is not possible to deduce that one style of question, open or structured, is 'better' or even 'more fair'. If we want to find out reliably whether candidates can recall specific items of information (or apply particular principles) then a structured question gives the candidate with the requisite knowledge a much better chance to demonstrate this. However, a structured question does not distinguish between these candidates to find out which are also able to select and organise information in a less structured task: skills that we might also wish to value, develop and test.

The traditional style of questions may only have been suitable for the small minority of candidates that the traditional examinations were designed to select and classify, whereas the modern approach gives many more students the chance to demonstrate what they have learnt. However, it is also tempting to suggest that we may no longer be testing some important cognitive skills that we would want our most able students to be able to demonstrate.

Some critics of modern examinations would describe the type of change I have discussed above as an example of 'dumbing-down', where other observers might refer to this as a decision to focus on a more valid way of testing a more limited range of abilities. However, there is certainly one aspect of some types of structured question which must surely be considered to be a form of making the examination easier.

It is well known from research that recall is not an all-or-nothing process. It makes no sense (in view of the empirical evidence) to refer to whether a person can or can not recall a specific item from memory. 'Items' in memory (in the mental concept map, so to speak) are associated to varying degrees with a wide range of other 'items'. Whether an item is recalled on a particular occasion depends upon (as well as such features as the degree of concentration and stress levels!) how much associated material is available to 'cue' the recollection (Taber, in preparation). What is associated, and how strongly, will vary from individual to individual and may include some quite idiosyncratic material.

Examinations must include cues - the questions must suggest what information is being elicited. However, by its nature, a structured question may well cue recall much more effectively by providing more potentially associated material to provide links to the stored information. Examiners may well claim that this is desirable, as it enables the candidate to show what they have learnt. However, many memory researchers suggest that once something is committed to permanent memory the trace is never completely lost, rather that it becomes harder to access the longer it is not activated, and so more cueing is needed. So - in principle - with sufficient cues we can all retrieve memories that we would not normally have access to. This raises the question of how much cueing reflects a level of recall that might have significance beyond the examination, rather than being an artefact of an artificially 'cue-rich' situation (i.e. a structured question).

Teaching science 'in context' is currently fashionable. It has been recognised that abstract scientific ideas do not appeal to many students (although they certainly do appeal to some), especially where they can not 'see the point' of what they have been asked to learn. Recent courses, such as the Salter's approach, or the Institute of Physics sponsored 'Advancing Physics' course tend to start with contexts that it is believed learners will find familiar and/or interesting. In exploring the motivating context, there will be a justification for explaining the scientific principles where they arise. Such an approach would seem to better match the way people learn in informal settings - i.e. dictated by the perceived needs of the current 'problem' context.

Similarly, there has been a trend increasingly to set questions in context in formal examinations (including National Key Stage tests). Presumably the logic of this development is similar to that which has spurred the context-based courses. Contextualised questions provide a setting that the candidate can recognise as being of some significance, and so give some form of validity to asking the question. However, this is surely unnecessary. Although many pupils may find context-based courses make science more interesting and relevant, and so motivate them to learn science, this hardly seems necessary in an examination. For those candidates who take their examinations seriously there is plenty of intrinsic motivation to pay attention to the questions and try to answer them, without the need for a familiar context.

If this were the only consideration then the increasing use of context in questions would not be a concern. However, I would suggest that there are good reasons to expect contextualised questions often to be more difficult for pupils. There are also suggestions that such questions may reduce test validity.

In terms of difficulty I can draw upon my own experience of teaching two science subjects to A level. I used to teach candidates chemistry and physics for the same examination board (AEB as it was then). In my experience physics was found the harder subject by most students who took both subjects. In some part this can be explained by the greater use of mathematics in physics - yet the only mathematics needed to answer questions was basic algebra, and most physics students had plenty of opportunity to develop competence in this during the course.

However one key difference between the two subjects was the way principles were applied in the examinations. In chemistry, a principle would need to be applied (in order to - say - predict the shape of molecules, or to explain the pattern of ionisation energies) to one or more examples selected by the examiner. The student had to understand how to apply the principle (e.g. the valence shell electron pair repulsion theory) but it was usually very clear which principle was to be applied.

In physics many students found examination questions much harder because their first task was often to identify the specific area of physics being tested. Although this usually seems obvious to the teacher, it could often stop a student making any headway on answering a question. In chemistry the context (i.e. a particular example) was often transparent, as the question made it clear what was required, but in physics the weaker candidates did not know what they were being asked to do. It was quite common for them to complain that they had not been taught anything about (for example) washing machines, when the question was testing their knowledge of (in this case) oscillating systems. The 'stronger' students ignored the specifics of the context and abstracted out the physical system they were being asked to consider: the 'weaker' students were not able to do that.

Many of our students find abstract thinking difficult (although it is clearly something we should aim to develop in teaching science). When we use familiar examples to introduce abstract principles (i.e. in teaching) we help learners by providing a bridge from student experience to the formal scientific principles. When we use everyday examples (such as washing machines) as contexts for assessing understanding, we may be providing a barrier for those same learners. If context is meant to make a problem more concrete, we should acknowledge that the formal thinkers are those best equipped to abstract the formal representation of the problem (needed to solve the problem) from its more concrete context.

Even when candidates recognise the area of science being tested the context could make the question more difficult. This should not be surprising in view of the limited 'working memory space' available to learners. Research shows that problem solving in science is often limited by the perceived complexity of the problems. Where a problem seems to involve more 'bits' of information than the learner can juggle at once the problem may become too difficult, even if all the individual steps are within the learner's competence. (Context adds more bits of information!)

Of course, some students learn strategies for approaching problems so that they are working with a manageable sub-set of information at any one time: but this requires ingenuity and/or training. As with the discussion above about structured questions, we have to decide what we want to test - do we want to know if the candidate can undertake the steps in the problem, or solve the whole problem unaided? - but we should accept that contextualising a question often makes it more difficult, and may even block some learners from demonstrating any relevant knowledge that they may have.

At least one examination board (UCLES, part of OCR) has been giving serious consideration to such issues (Ahmed & Pollitt, 2001; Pollitt & Ahmed, 2000), and their research has suggested that even at the level of National Key Stage tests the use of question context may put off some students who fail to recognise the science being tested. Other candidates may spend time applying their knowledge of some irrelevant aspect of the familiar context, or trying to find the significance of some detail of the context that was found salient, but was actually doing no more than filling-out the context.

UCLES research has also suggested additional concerns about the validity of contextualised questions. One of these relates to the types of intellectual process being tested. One justification for using context is that 'application' of a principle is considered a higher level intellectual skill than simply recalling it, and a novel context tests whether candidates can apply ideas. This is a fair point providing the contexts selected are equally novel to all candidates. This would be very difficult to arrange and demonstrate, as a context that is totally unfamiliar to one candidate is likely to be very familiar to others. One is reminded of early IQ tests which were assumed to test reasoning ability, but which relied on culturally bound knowledge such as the rules of baseball (Gould, 1992). Even if the familiarity is not sufficient to indicate answers, the candidates who recognised the context may feel more confident than those to whom the context was alien (like my physics students who had 'not studied washing machines').

Another problem with context, recognised by UCLES but also indicated by research into learners' ideas in science, is the way that students cope with learning science that seems inconsistent with their everyday knowledge. It has long been recognised that where the scientific model that is taught in school is at odds with existing ways of thinking about the world, one response is to store the two 'types' of knowledge separately (in different regions of the mental concept map so to speak). Many successful students have one way of talking in the 'life-world' (where exercising gives you energy, acids always burn, heavy things sink, and thrown objects run out of impetus), but are able to switch to the scientific mode of explanation in the classroom.

By setting science questions in everyday contexts it is possible to activate the 'wrong' system and so a 'life-world' answer is given when the student was quite capable of responding 'correctly' had they been so cued. Again, we need to ask what we want to test: the ability to give accepted scientific answers, or the habitual use of such explanations regardless of social context. It is worth bearing in mind here that even teachers and professional scientists are known to use common vernacular ways of describing the world when 'off-duty', and even highly qualified experts can sometimes be 'caught-out' when questions are not presented as formal tests of abstract science knowledge.

Summative assessments - examinations - are intended to test how well candidates know (can recall, apply and explain) their science. In recent years they have largely required answers to structured questions, and are now increasingly tending to be set within some assumed-to-be-familiar context. It is not suggested here that either of these developments are intrinsically good or bad. What it is argued here is that both of these changes mean that examinations make different sets of demands than before. In one sense structured questions help the examiner find out if the candidate has the target knowledge much more reliably than open questions - but at the same time they do this by setting a more restricted set of demands for the candidate. In this sense such questions have become easier to answer, and by providing more cueing they are surely also testing a different threshold of recall.

The shift towards more contextualised questions would seem to offer a more complex situation. Sometimes when the context is very familiar an intended application question may be reduced to a recall question. However, unfamiliar context may be discouraging; context increases the richness of information presented and so the complexity of the question (and requires discrimination between relevant and incidental features); and the context may even deflect the candidates' thinking away from relevant scientific knowledge.

Both of these developments therefore change the nature of what is being tested. Structure can certainly make questions easier, whereas most contextualised questions become more difficult for many candidates. Whether these changes in question demand are desirable depends upon exactly what we value in learning and so wish to assess. Contextualised questions may also reduce the validity of questions by seeming to present very different demands to different candidates and by channelling some candidates away from thinking about the target knowledge. This clearly suggests that (at the least) such questions need to be carefully planned and pre-tested to make sure they are suitable for including in formal examinations.

In the UK system regular examinations are now the norm for most of our students. There is also widespread talk within the professional community of how teaching has increasingly become assessment-led: with 'teaching to the test' an undesirable but common phenomenon. The attitude seems to be that teachers dislike this development, but taking a principled stand may well mean inferior results - reflecting poorly on students, teachers and institutions. Perhaps it is time for the science education community to revisit the aims of science teaching, and ask ourselves what it is we are most keen to teach students. Then we should design our summative assessments accordingly. If we do this well, then perhaps teaching-to-the test can actually be seen as a virtue.

Ahmed, A. & Pollitt, A. (2001) Improving the validity of contextualised questions, paper presented to the British Educational Research Association Conference, Leeds, September 2001.

Ausubel, D. P. (2000) The Acquisition and Retention of Knowledge: a cognitive view, Dordrecht: Kluwer Academic Publishers.

Barker, Vanessa (2000) It's my party - and I'll cry if I want to!, Education in Chemistry, 37 (3), May 2000, pp.72-74.

Gould, S. J. (1992) The Mismeasure of Man, London: Penguin, 1992 (first published in 1981).

Pollitt, A. & Ahmed, A. (2000) Comprehension failures in educational assessment, presented to the European Conference on Educational Research, Edinburgh, September 2000.

Sorenson, P. (2000) Drowning in numbers? The need for formative assessment, in Sears, J. & Sorenson, P., Issues in Science Teaching, London: RoutledgeFalmer, pp.123-132.

Taber, K. S. (2000) Chemistry lessons for universities?: a review of constructivist ideas, University Chemistry Education, 4 (2), pp.26-35, available at http:12/08/03/www.rsc.org/uchemed/uchemed.htm

Taber, K. S. (2002) Chemical misconceptions - prevention, diagnosis and cure, 2 volumes, London: Royal Society of Chemistry

Taber, K. S. (2003) Responding to alternative conceptions in the classroom, School Science Review, 84 (308) pp.99-108.

Taber, K. S. (in preparation) Lost without trace or not brought to mind? - a case study of remembering and forgetting of college science.

Acknowledgements: I would like to thank the University of Cambridge Local Examinations Syndicate (UCLES) for inviting me to give two research seminars (to the Research and Evaluation Division in 2002, and to Cambridge International Examinations in 2003).

University of Cambridge Faculty of Education
Homerton Site
Hills Road
Cambridge CB2 2PH
http://www.educ.cam.ac.uk/staff/taber.html
kst24@cam.ac.uk

This document was added to the Education-line database on 14 August 2003