INFORMED CONSENT BY CHILDREN TO MEDICAL AND EDUCATIONAL EXPERIMENTS I The value of a comparison The ethical issues of medical experiments are widely debated. As part of this there is some discussion of medical experiments on children. The Nuremberg code held that experimenting on humans legally incompetent to give informed consent was unacceptable. A fortiori, experiments on children are unacceptable. Other statements on medical experiments have modified this view, either allowing children to consent to be subjects of experiments or allowing their legal guardians to consent for them. There is accordingly much discussion of the function of informed consent, of the role of parents and the limits of acceptable research on children. I know of no such literature on educational experiments on children. Statements of professional ethics, apart from a few special cases, do not mention the problem.. The exceptions concern laboratory experiments (such as the principles of the Australian Psychological Association) or deal with children with intellectual handicap. For ordinary classroom experiments, there has not seemed a need. Yet there are highly suggestive parallels between medical experiments and classroom experiments. There are two obvious reasons why statements of professional ethics have nothing to say about experiments on children. The first is that not many such experiments are performed anyway. Teachers and teacher unions tend to be rather conservative in their views of the processes of education, and are more often criticised for their failure to try new methods than for their excessive use of them. That conservatism stems, at least in part, from concern lest their students' education suffer from ill-judged and ill-timed experiments. Yet addressing the issues of educational experimentation may justify that conservatism; or it might set limits to the realm of legitimate research. It might also answer those teacher's fears, encouraging more inventiveness and daring in teachers. The second reason that the literature may not have taken off is that there are obvious differences between medical and educational experiments. Failed experiments in mathematics teaching rarely result in death, loss of sight, cancer, paraplegia or disfigurement. Teachers are unlikely to be charged with mayhem, assault, battery and manslaughter if they try out some new method. On the other hand, the outcomes of schooling do affect life-chances of students, and these in turn affect longevity. There's not much moral difference between failing to provide students with a benefit which will extend their lives and causing a harm which will shorten them. Teachers might, like doctors, be charged with educational negligence. Indeed, courts in North America have not been slow to include educational setbacks amongst the deleterious effects which are relevant when a medical researcher acts improperly. The Washington D.C. case of Bonner and Moran [discussed in Annas (1977) pp77f] is a case in point in which the loss of only two months' schooling was deemed to be a harm for which the experimenter was required to pay compensation, and which was deemed of more significance than serious pain, loss of blood, permanent disfigurement of the leg and loss of skin. Another difference between medical and educational experiments is that educational harm can be reversed, while medical harm is likely to be permanent. Rather than justifying the absence of a literature, however, this difference points to a plain requirement of classroom research: that procedures should be in place to ensure that any educational deprivation which occurs as a result of the experiment is compensated for. This in turn requires that proper procedures are adopted to assess the outcome and consequences of the experiment. Thus we may learn from the differences between medical and educational experimentation as well as from their similarities. There is, then, point in pursuing a comparison. II Limits of the present study I want to examine two kinds of classroom experiment. The first is what, for want of a better word, I'll call an experiment in class arrangements. This is a real example from the 1970's. A school decided to adopt open classrooms, team teaching and student choice of projects combined with "contracts" in which the student undertakes to complete a certain number of self-chosen tasks in a given time. The principal of this new school made these arrangements because he and his senior staff believed in freedom -- that is to say, in the free school version of progressive education. Such an experiment is necessarily of some length. It would normally last at least a year, since introducing changes to such arrangements in mid-year would cause chaos. In the case in point, it lasted several years, as was necessary for the results to be significant. This kind of thing was at the time called an innovation, although the ideas were scarcely new. Nevertheless, in view of the mixed reports of the success of these procedures, it is properly called an experiment. It is done, at any rate, in an experimental frame of mind -- let's try this and see if it works. The second is the so- called "scientific" experiment using control groups and the testing of one teaching method against another. My example is of a mathematics staff who have been reading the literature on two views of the basis of mathematical expertise. For one theory all expertise is essentially memory. The essential (and the only important) teaching technique consists in the demonstration of numerous examples of each of a number of problems. For the second, the crucial thing is having the right images or mental pictures, the right models of mathematical concepts. Good teaching practice will consist in discovering or inventing a new model which will encourage all the right inferences. The staff divide year seven randomly into three streams, one being taught by traditional methods as a control group, the other two being taught by one new method each. By confining the issues to the classroom, I avoid an important issue that has been raised in medical ethics. For one of the problems there concerns non-therapeutic research, where the immediate aim is the greater understanding of the functioning of the human body, or of its development. If the past is any guide, there is the potential for immense long- term benefit here. Take, for example, research which occurred over nearly two centuries into electrical phenomena associated with the heart [Nicholson (1986) p43] -- research which led to the invention of the electro-cardiograph. The benefit came (as Nicholson notes) as the result of a coming together of a great many pieces of knowledge, none of much significance in itself without the benefit of hindsight. Questions have been asked about the acceptability of research on children when there is risk, but no benefit to the child, while future generations of children may benefit mightily -- and may not, if the research remains isolated from other results. In education, the parallels are with laboratory research in educational psychology or ethnographic research in sociology; and it is not easy to think of real examples where significant risks might be justified. (I know of no- one doing invasive research into neurophysiology with hope of educational benefit.) The present study is confined to the moral acceptability of such classroom experiments. III A theory of informed consent in medical experiments Let me begin with an example. Beriberi was rife in Malaya in 1907. One theory was that it was caused by a poison present in white rice, but not in brown rice. (This was before the discovery of vitamins.) William Fletcher [Vollrath 1989], the district surgeon in Kuala Lumpur and as such in charge of an asylum, thought this was nonsense, and set out to prove it. He assigned a group of his charges to a white rice diet, and a second group to a brown rice diet. To his surprise, the white rice eaters developed beriberi, while the brown rice eaters did not. To make sure, he assigned some of the each group to the other diet, while leaving others on their original diet. Some of the white rice eaters died; none of those who completed the experiment on brown rice died. Patients were not asked if they wanted to be part of the experiment; and Fletcher was prepared to risk their deaths, even (or especially) after he had reason to believe that the white rice diet was after all causally linked to beriberi. The Nuremburg code, developed by the tribunal which tried the doctors responsible for horrific experiments in Nazi Germany, makes its first requirement that human subjects voluntarily consent. The consent is to be free from fraud, deceit, over reaching or ulterior forms of constraint and coercion. Subjects are to know and understand the nature, duration and purpose of experiments; the method and means by which they are conducted; and to know all the risks of harm or inconvenience. Each is also to have the liberty at any stage to bring the experiment to an end if it seems to him that it is impossible to continue. We already have a set of requirements which if translated into the classroom would make serious experimentation very difficult. But, as I said at the start, the Nuremburg Code goes further; it requires that subjects be legally competent to consent. This not only excludes, as it was meant to, any experiments upon prisoners or the mentally handicapped, it also excludes absolutely experiments upon children under the age of legal majority. This makes sense. If you ought not to be permitted to sign a binding contract, you ought not to be permitted to consent to an experiment that will put your life or your future at risk, however slight the risk may be. If this principle were firmly established, that would be the end of my paper. However, the cost to medical and foundational research of adopting this rule would be very high. While it is possible to perform some tests on animals and others on adults, and so do away with the need there for child subjects, other information, for example about the process of physical maturation, has to be checked on children. The potential benefit to future children (not to mention adults) is enormous. (See [Nicholson (1986) for an elaboration of these arguments.) It is therefore not surprising that the stringent requirements of Nuremberg have been watered down in successive statements of medical ethics. The Declaration of Helsinki, produced by the World Medical Association, set the scene here. It permits the consent of a legal guardian where the child is incapable of consenting, and for the consent of child and guardian if the child is capable. Some documents specify the age of seven as that at which he child's consent must be obtained, and fourteen as that when the guardian's consent is no longer needed. The Royal College of Physicians has moved from allowing no non- therapeutic research, to allowing it provided the risks are negligible. Therapeutic research is another matter. Where a child is sick, and there is no known cause of the disease, further latitude is allowed, though the consent of guardian and/or patient is still usually required if there is time. Some American courts have gone a lot further and introduced the doctrine of substituted consent. This says that if a reasonable adult in the child's position would have consented, the child can be taken to have consented. This doctrine can be traced to a court case concerning organ donation for a transplant; but it has subsequently been applied to research. In the case, (Strunk v Strunk, discussed in [Annas (1977)] pp81f) a mentally handicapped boy had had one kidney removed and transplanted into his sister. The boy himself had objected to the operation --though without full understanding of the risks. The doctor was subsequently charged with assault, battery and mayhem. The court would like to have found that the parents could consent for both children on the grounds that they would have their interests at heart. But here there was a conflict of interests between two of their children. It was clear in any case that the parents didn't take the interests of the boy into account. Yet the court wanted somehow to balance the interests of both children. Their solution was in effect to say that the operation was in the boy's interest also. They held that he would suffer psychological damage if his sister died. If he had been fully rational and fully mature, he would have recognised that this near certainty outweighed the mere risk that he might need his second kidney one day or that he might die on the operating table. This combination, of determining the judgement that an adult would have made by a risk/benefit analysis; and stretching the notions of risk and benefit by equating distress with psychological harm -- this combination will permit almost anything. The notion of informed consent by the subject of the experiment has in this extension become empty. In effect, the court is substituting its own judgement for that of the boy. There are some nasty,well- known parallels in education argument. The principle of informed consent has been undermined from other angles too. Where a research subject will undergo no risk of harm and no inconvenience of discomfort it has been held that there is no need for consent, especially in cases where it would be beyond the subject to understand the experiment. (Why do we need consent if there is no risk of harm?) Now of course there is some risk of harm with almost anything. Venepuncture (obtaining blood) for experimental purposes carries a small risk of infection and in very rare cases, death. So the principle that consent is not needed where there is no risk has been modified to require only that the risk be not significant, and that the harm be not too great. These notions have been modified in turn, firstly by allowing more risk of more harm if the benefit to be achieved is great, and then by accepting the risk of harm provided that risk is no greater than that which the subject of the experiment would voluntarily accept in ordinary life, or which is commonly accepted by people in general. The idea that the harm risked must be small in relation to the good to be achieved has also undergone some changes. It has been held that all that is needed is that the good is greater than the harm. But of course it is a rare experiment where good is guaranteed. The need for experiment exists precisely because we do not know what effects a procedure will have. So instead people talk about the chance of good and the risk of harm in terms of probabilities; then assign numbers to the probabilities, others to the good hoped for and the harm risked, and then calculate whether the experiment is justified and consent can be overlooked. (For examples see Nicholson (1986) pp108ff.) Given that the immediate benefits may only be a small step in our knowledge of the body's functioning but that the long term gains of many such steps may be anything from miniscule to enormous, the probability calculations are worrying. Finally, sometimes seeking a patient's consent to a procedure is medically dangerous. Consent has been held to be contra- indicated where it itself causes an evil. This idea has been applied to research, and the notion of harm extended to include the distress a subject may suffer if told of the risks of the experiment. Some cancer patients have been given placebos without being informed that they were part of an experimental programme; much less that some of the patients would receive placebos, on these grounds. There was a widely reported instance last year from National Women's Hospital in Aukland, New Zealand, where a doctor allowed cancer in situ in the womb of a number of women to go untreated, because of his view that the conventional wisdom -- which said that such cancers would become invasive -- was false. The doctor in question gave as his reason for not seeking the consent of his patients that it would have distressed them to know that they were part of such an experiment. [Campbell (1990)] In another case, patients under general anaesthesia were subjected to invasive vaginal investigations without consent. Again the justification given was that it would only distress the patients to know the risks of what was being done to them. (I am indebted to John Clark for the details of this case.) In each of these examples the drift of events has been firstly to modify the principle of informed consent, and then to stretch the concepts used in the justification of the modification. The question is, what can be done about such concept-stretching. Lakatos says that attempts to stop concept stretching are irrational, and calls such attempts `monster-barring'. His monsters were mathematical objects; mere figments of the imagination. These medical monsters are real; the intellectual descendants of Dr. Mengele. Still, we may learn from Lakatos's advice about concept-stretching. In mathematics, the appropriate advice is to re-examine one's proofs. The equivalent in morality is to go back to fundamentals. We need to look at the function of informed consent. It is clear that the main concern of the Nuremberg tribunal was to provide a guarantee against harm. If that is all that is involved, we can again ask, why do we need consent if there is no risk of harm? Or no real risk, or it's not real harm -- and we are off on the slippery slope again. We need a firmer grip. Perhaps we should ask what is wrong with harming others for the sake of the general good. I would suggest the following functions of informed consent. These are not in order of priority, but of convenience of exposition. (For some of these see also [Annas (1977) pp33-37] First, it requires the researcher to give a clear account of the functions of the research. This will have to include an account of its necessity, and why it has to be done on these particular subjects. This practical requirement relates to three further principles of research on humans which are usually stated independently of the informed consent doctrine. The first of these is that the research cannot be done on animals instead. There are also cases where research on animals is relevant to humans but provides no guarantees. In such cases research should first be done on animals, so that dangers can be discovered before humans take the risk. The second requirement is that research should not be done on children where it could be done on adults. The third is that the research is scientifically respectable and justified. That is, it must not merely repeat the work of others, nor provide only trivial elaborations of it. It must be sound methodologically. A researcher who has to explain to subjects why they are being asked to take risks will be forced to become fairly clear about these issues. The second function of requiring consent is that it helps to ensure that the risks are outweighed by the chances of benefit. The public benefit may not be a sufficient condition for subjecting people to research. But it is surely a necessary one. In this function as with the first, the consent requirement supports something which is also required on independent grounds. What it adds is that the justifications must persuade someone who is not influenced by the prevailing mores of research, and whose strong interests in opposition to the research will ensure that the arguments are given very close scrutiny. Thirdly, the requirement makes the activities of the researcher public. That will help to avoid fraud and deceit, and will make more likely scrutiny of the justification and the operation of the research. As before, there is also an independent principle here: since the judgement of the researcher may be influenced by personal ambition, proposals for experiments should be subjected to scrutiny by an ethics committee, which should contain not only members expert in research methods and the disciplines the research relates to, but also lay members who are not influenced by the mores and standard procedures in those disciplines. Fourthly, the principle of informed consent leaves the evaluation of harms up to the individual affected. We have already seen in my discussion of the kidney transplant case how treating distress as a harm can lead to grave immorality. The notion has been applied in the debate on abortion versus adoption, surrogate motherhood, infanticide of defective children and euthanasia, with immoralities resulting every time. It needs to be insisted that distress, even mourning, is not the same as psychological damage. An individual may reasonably decide that it is worth risking death or the loss of life-chances for the sake of another person or the general good. This is not the same as saying that they consider the distress that they would suffer at the loss to others a greater harm than that which they risk. To argue that is to confuse altruism with hedonistic egoism. That is not a confusion likely to be made by someone choosing the chance of benefit to others over risk to themselves. Apart from overriding this spurious kind of argument, however, there can also be choices to be made, even when the subject of research is a patient who may personally benefit. The practice of medicine serves the quality of life as well as its length, and where there is a choice to be made between the two, we do not always choose length. Indeed, it would be irrational to do so, for life spent in the pursuit of longevity at all costs would not be worth living. But quality comes in different varieties. It is not a good idea to have the researchers decide which values will be reflected in the lives of the subjects of their research. Even if such choices can be made objectively, the researcher may well not be objective, and neither may an ethics committee. (The vaginal investigations I referred to above had been approved by an ethics committee.) Fifthly, allowing the subject to decide permits and promotes the morality of the subject. Sixthly, by permitting and encouraging a grave decision to be made on the basis of a substantial amount of information (of the state of medical theory, the requirements of competent research, the prospects for the subject), requiring informed consent promotes rationality and autonomy. Seventhly, the practice respects the autonomy of the subject. Eighthly and finally, it ensures that the subject is never treated merely as a means to other people's ends; it protects his/her status as a person. Subjects will not be seen as dispensable. The last four points deserve closer attention. The principle of respect for persons is not, in my opinion, a fundamental moral principle which applies regardless of all else, to all human beings. It depends of views about the nature of human beings - of what it is about them that demands respect. What counts here is the capacity of persons to decide upon the values by which they will live. It is that which provides the opportunity for morality, for interpersonal relations and for the distinctively human aspects of physical activity. It is involved in all rationality, since there is no thought without evaluation. But it is so involved as the foundation of rationality. If this is correct, then respect for persons just is respect for them as choosers of fundamental values. The choice of whether to take risks with your life for the sake of others is part of that fundamental choice. To deny people the right to make that choice is to deny them the opportunity to be estimable; to deny them the opportunity to be worthy of respect. To make the point backwards, to require heroism is to diminish its moral value. There can then be no moral grounds based on the relative worth of persons for destroying one for the sake of another. Respect for their autonomy and protection of their status as sources of ends are rooted in the same values. It is in this that the real foundation of the principle of informed consent is to be found. That it also supports other desirable ends is fortuitous, though fortunate. Given that, we can return to the court developed principles of consent by the guardians of children and of substituted consent. If the point of informed consent were merely to avoid unreasonable harm to the subjects of experiment, to ensure that their interests were given due weight, then parental and substituted consent would make sense. The requirements of necessary and competent research, of overall benefit, of checks on the researcher, could all be met by this means. But the morality and autonomy of the individual can neither be promoted nor respected by this means. In the course of discussing informed consent I have had reason to mention and in some cases justify a number of other principles which the provision of informed consent safeguards, but which are of independent importance. Let me summarise. 1. The risk of harm to the subject must be outweighed by the chance of good to the subject or to others. This will be determined both by the level of harm and of good, and by the probabilities in both cases. The reasons for this are obvious on both the deotonlogical grounds I have chosen and on consequentialist grounds. Estimation of the value of basic research is difficult, as I have noted, but this still has to be done. 2. The estimation of the costs and benefits to the subject should be made on the basis of the subject's own values. 3. The knowledge must be unobtainable in any other way. For example, research should be done on animals rather than humans if possible. This follows from the value of humans as moral choosers. (Utilitarians will find that harder to justify.) 4. Research should be done on adults rather than children. This is because adults are in a better position to know what life holds, and of what is set at risk. They will also know more of what it is to choose a value. Indeed, they have chosen their life-values; determined the meaning of their lives, and so can choose more easily. We could add that they have had more of a life than children also. Young children are not in a position to choose, and if the choice may close off later options, should not be expected to. 5. The research has to be really necessary for the advance of knowledge. 6. It must be methodologically sound. 7. The research proposal should be discussed publicly before it is implemented. 8. The development of the subjects as autonomous agents and moral beings should be enhanced by their participation. 9. The principle of informed consent itself. There are a few other principles which are commonly and properly applied to medical experiments. These are, firstly, the right of the subject to withdraw at any time, without pressure or loss. This is an extension of the principle of informed consent, and is justified in the same way. Secondly, if in the course of the experiment the health of the subject is threatened, or it becomes clear that one treatment of two being compared is markedly superior, the experiment will stop, whatever the effect on its scientific value. Thirdly, the health of the subject will be safeguarded at all times. These are matters of minimising the costs of the experiment. Fourthly, the results of the research will be made public. This is to maximise the benefits of the research. This should happen whatever the results are. One of the nastier features of the cancer in situ experiment which I mentioned earlier was that the failure of the experiment was hushed up. As a result, not only was nothing one for the women who had not yet developed invasive cancer, no publication was made of the results, and the consequence was that other doctors, influenced by the earlier widely expressed opinions of the researcher, allowed their own patients to go untreated also, with similar sad results. Failures, as well as successes, should be publicised, to prevent their repetition. Fifthly, therapeutic research may carry a higher risk to the patient if there is no known cure of the disease. Non-therapeutic research should carry minimal risk, and be of substantial public benefit. (This principle, it should be noted, is difficult to apply. What is a high risk? One chance in a million of death? One chance in two of a headache? If the limits are set too low, there will be no research; if too high, then harmful experiments will be performed. It is also suggested that in the case of a great disaster looming, the level of acceptable risk might be raised. (See Nicholson chapter 5 for a discussion.) For example, if a plague could be averted by testing a vaccine, perhaps more risk is justified. However, the Institute of Medical Ethics Research Group on the ethics of clinical research investigations on children couldn't find any actual cases of high risk research where the research was of such value as to provide a justified risk/benefit analysis where there was no other way of more safely completing the research. (Ibid.) Sixthly, the principles adopted for research --the official codes of ethics -- must retain public confidence. Such rules need to be made public, sine otherwise leaks and rumours will damage trust between doctors and patients --and that will be medically disastrous. Seventhly, experiments should only be conducted by scientifically qualified persons. It is time to return to children. First, can they give informed consent? We are less inclined now than we were a few years ago to declare children incapable of the necessary thought. But even if a child of ten can reason well, such a child still lacks knowledge and experience which is relevant to some decisions. Suppose, for instance, that an experiment created a risk of sexual impotence. There is no way such a child could understand what he would be losing if the experiment went astray. Nor could a child understand adult friendship, either. The law is not helpful here. It makes the age of 14 the start of full criminal responsibility, and the age of 18 that of full adult rights. But the ages do not correspond to any psychological theory of children's competence. Fourteen was in both ancient and mediaeval times the age of adulthood, because at that age a boy could bear and wield arms. Towards the end of the mediaeval period, when plate armour was introduced, an older age was necessary, since 14 year olds were not strong enough to manage it. Twenty-one was chosen instead as the age of adulthood. [Annas, (1977) pp64f] (Interestingly, it was an argument related to the age for compulsory military service which permitted that age of legal majority in Australia to be lowered to 18, and made the politically expedient age of 15 unobtainable.) Martial capabilities have no bearing on informed consent. The end of the Piagetian approach to educational psychology has removed an alternative source of argument about the capabilities of children. As I argued in [Bibby (1989), once you get past infancy, it is more a lack of experience than of reasoning ability which limits children's ability to make good decisions.That lack however means that in the case of many medical problems they cannot make fully informed decisions. Because of their vulnerability also, children are susceptible to slight or just imagined pressure, especially when they are aware of being in total dependence on a doctor; and so autonomous decisions are hard to ensure. I think two things follow. Children should not be asked to make decisions except where the risks lie within their experience. And respect for the future value choices they may make implies that nothing should prejudice their future choices. These two things together imply the orthodox view that in non-therapeutic research, children should not be subject to risks of any kind of permanent harm. That is, they should not be permitted to consent to such research. Even for temporary harm, only minimal risks are permissible. Within the range of admissible experiments, their choices should be respected; to ensure that they stay within that range, parental consent should also be required. The refusal of either child or parent should be decisive. This contrasts with the situation when the child is sick. In law, between the ages of 10 and 14 either child or parent can consent to treatment, and either's consent over-rides the other's refusal. Experimental treatment, however, is only justified at all if there is no known cure for the disease, or where extensive animal and adult trials make the risk of danger to the child minimal. But here too the consent of both parent and child are essential. What of infants, where consent is impossible? The refusal to allow research here at all is tempting. On the other hand, research on healthy infants was essential for the discovery of the correct treatment of infants with diarrhea. The loss of life from that cause had been substantial. And there was no significant risk to the infants. Intuitively, that research was justified -- though there were no reasons to believe that research on the composition of the body fluids of normal infants would have these fortunate consequences. One could, I suppose, take the position of Michael Tooley, [Tooley (1974)] that infants are not persons until they can desire to continue to exist as subjects of experiences, and hence have no rights of any kind until they can think about themselves in that relatively sophisticated way. However, I don't need that desperation move. For my account of the importance of consent makes a clear difference between infants and children. To over-ride a child's refusal to take part in experiments is to deny that its choice of values has any significance. This (a) will have bad consequences for the child and for its beliefs about others, and (b) is intrinsically immoral if anything is. Infants, on the other hand, do not have any values. Their moral significance lies in their future values. Experiments therefore should not prejudice their future choices, but need not otherwise be eschewed because of lack of consent. We will still want parental consent, to safeguard the interests of the child against hungry researchers. There will also need to be advocates for the child associated with ethics committees, since parents are not always knowledgeable enough or vigilant enough or well-motivated enough to protect their children. However it is not appropriate to see parental consent replacing that of the child. This consent has a different function from that of the subject of research. Given the functions of the subject's consent I have described, parental consent could not replace that of the infant. Proxy consent cannot further the autonomy of the child, nor does it respect its present or future value choices. The guidelines for research on children that I have been suggesting are very strict. Not all research has been so tender to them. The first polio vaccine carried a risk of infection; and a number of children contracted the disease. It was, moreover, a relatively rare disease. The successful development of the present vaccine was at the cost of these children -- a cost that was not only predictable, but predicted. All that was not predictable was which children would contract the disease. The principles I have outlined imply that that research was not justified; and with that view I am content. (This example and the next are taken from Nicholson. (1986) A second example is a tougher nut. A two year old sufferer from phenylketonuria was put on an experimental diet. It worked. So far, so good. Then, without informing her mother, the doctor added L-phenylalinine to the diet, to make sure it was really the diet that caused the improvement. The child's condition deteriorated. Now this research was done for the sake of the child. The diet was very stringent, so discovering whether it was the cause was important for the child's quality of life. And obtaining a parent's consent would have spoiled the test. The suffering was, admittedly, very brief, and the long-term risks were small. The experiment was done, moreover, in the hope of increasing the child's options. But it doesn't fit my principles. The treatment of infants is only relevant to my theme as a test of the principles for which I have argued. There are a few points I wish to make about the role of parents before I turn to the application to classroom research. Firstly, I have indicated that the role of the parent is to safeguard the interests of the child, if necessary against the child's own choice to take the risks. But parents have interests f their own to protect also. If a child is incapacitated in the course of research, the parent gets a greater liability for supporting and maintenance. Parents also have in interest in participating in normal relationships with their children. These interests provide additional reasons for distinguishing between normal treatment, where a child's consent can override a parent's objection, or where, with young children, medical procedures may be performed without the consent of either, and non- therapeutic experiments, where the consent of both should be required. However, these parental concerns should not allow for consent to harmful experiments. I've noted before that a parent cannot waive a child's rights, and so cannot exercise those rights for a child. [Bibby (1985)]. (This is true legally, too.) A parent an only consent to an experiment carrying risks where the risks of not acting are worse. IV classroom experiments That completes my account of the ethics of medical experiment. It is time to examine the classroom. The principles I have argued for apply equally well to classroom experiments, and for the same reasons. I've already noted that the risks involved with these experiments affect the length of a persons's life. I can now make two other points in defence of drawing parallels. Firstly, even if the connection between education and length of life is tenuous, the connection with quality of life is not. Secondly, if the risks associated with teachers' poor and failed experiments are less than those of doctors, so are the benefits of success. With that in mind, let us examine the first classroom experiment -- the one on class arrangements. It fails the test on virtually all counts. 1. The students were not asked for their consent to the experiment. If they had been, at best a vote would have been taken, after considerable pressure from the teachers. This problem could be have been avoided if the school were divided in two, with dissenting students being taught in orthodox fashion. 2. No clear account is given to the children of the functions of the research. To do so would have been possible, but would taken a lot of time; and the educational outcome could be affected both by the time taken and by the explaining. Students could only come to understand the experiment as a result of some study of progressivist ideas. That might be quite a useful student activity -- indeed it would make very good sense. (So would having teenagers study educational psychology.) But it would diminish the value of the experimental result. 3. The experiment was probably not necessary. It broke no new ground, and it was not sufficiently different from similar innovations in the seventies to throw fresh light on the reasons for their imperfect success. 4. No effort had been made to try out the experiment on adults. This is an unusual suggestion in education, and I think it is a potentially useful proposal. Given that the experiment was likely to last at least a year, there'd be point in such a test. Of course the proponents of androgogy will declare that there are major differences between the education of adults and that of children. I have some doubts about that. The collapse of Piaget's theories weakens that view, and so does the success of the philosophy for children movement. Trying out a system like this with University students might well have exposed some of its weaknesses. 5. The question of the scientific respectability of educational research of any kind is controversial. Perhaps we shouldn't put too much stress on the word `scientific'. What is required is a theory which gives some hope of benefit by the change, and which does not indicate likely harm. In this case, no single consistent theory was involved. The experiment was based on a mishmash of ideas. 6. The experiment does not stand up well to risk/benefit analysis. There were major risks of harm. A whole year's work was put at risk. There were no arrangements for helping students catch up lost time if the experiment was a failure. Indeed, it was not to be declared a failure for a number of years. On the benefit side, there was little chance of gain to anyone outside the school. There were no plans to publish anything if it was a failure. There was no hope of publishing anything if it was a failure. 8. Though the proposal received school authority approval, it should have had approval by parents also. Some of the 1970's experiments did have this. Indeed some of them began with parents. In this case. however, the school was set up first, and the parents found out when they checked out their new local high school. 9. The comparative evaluation of harms and benefits was not even discussed with students. In education this requirement is particularly important, for the aims of education are complex and controversial, and involve those very choices of life values which the principle of informed consent preserves to the subject of research. 10. There was no right to withdraw. It would have been difficult to give students such a right -- though splitting the school might have helped. 11. As it became apparent that the experiment was not working for many children, efforts were made to improve the alternative school. It took some years, because architectural changes were necessary, before the project was abandoned. 12. The experiment did not involve qualified researchers at all. There are some good points about this case, however. The experiment scored well on intention to promote the morality and autonomy of the students. That was a major part of its goals. But those setting up the experiment lost an opportunity to begin the process with the students being involved in the design and in consenting to be part of it. Again, the experiment was designed to help the students; they were not merely being used to bring about a possible benefit to others. Indeed in some respects the experiment was closer to a clinical experiment on a patient with an incurable disease than a pure piece of research. For the students came from a poor working-class suburb, and the outcome of schooling in such circumstances was rarely outstanding. Overall, then, the experiment could have been conducted in a much more ethical fashion. This would have changed its character, making it a different experiment, testing different things. Were the teachers and principal justified in omitting the consultation process, for the sake of the advance of knowledge? In this case, surely not. Are they ever? On the above arguments, it would appear not. The other experiment is imaginary, but realistic enough, in that the two views of mathematical thought are current, and in one case not yet researched in schools. How does it fare? There are some unhappy parallels with Fletcher's work on beriberi. The notable difference from the classroom management case is that there is a serious attempt at rigour. There will be statistical comparison of the three groups of students, and no doubt pages of computer analysis. Now this kind of thing is of dubious mathematical authenticity. It has been vigorously attacked, for instance by the well-known mathematicians Phillip J. Davis and Reuben Hersh (in their 1989) as mathematical nonsense. Nevertheless, if there is a most thorough and rigorous approach to classroom research, this is part of it. How does it fare on moral grounds? There is certainly a risk of harm in the experiment. The two new methods are both supported by extensively researched theory; but each implies that the other will waste students' time. If the work done on the model and mental picture theory is correct, there is a window of opportunity around the ages of 10 - 12, when students can be moved from a condition of struggling to handle mathematics to one of confident competence, with good opportunities for growth. By the time the experiment is complete, it may be too late, because of changes in the brain, for the adjustment to occur. If the memory theory is correct, the mental pictures approach will institute inefficient mental processing and faulty learning habits. On either theory, the control group is going to be harmed, in the sense of not receiving a good they ought to. Harm, then, is inevitable. No effort is made during such an experiment to safeguard the students' progress. To do so would spoil the experiment. The harm may not be able to be reversed, and in any case is not likely to be. At the end of the year, the progress of the students will be evaluated, the classes re-divided for streaming purposes. Who will see to it that the students who were badly taught are given compensatory coaching? But the good to be achieved is a clear contribution to the theory of teaching mathematics. When the results of the research are combined with those of others, millions of children will benefit. The subjects have no role in determining the worth of the experiment in relation to the risks to themselves. Most seriously, however, the students are not asked to consent to the experiment, nor are they able to withdraw. If they were, it would spoil the random selection. The parents are of course not consulted. To do so would be to invite them to ruin the experiment. And if it becomes clear in the course of the year that one group is dropping behind the others, no changes will be made in the treatment that they will receive. To do so would be to ruin the experiment and mean that the losses suffered by the students would be wasted. So by the end of the year, the researchers will be deliberately refusing to teach students in the way they are now confident is the best. Were the students informed of the progress of the experiment, they would not consent to continue with it either. The experiment becomes steadily less defensible. Yet unless it is completed without change, it will lose the justification that it will contribute mightily to our knowledge of how to teach mathematics. Further, because one of the theories being tested is age-specific, there can be no appropriate tests on adults. On the other side of the ledger, the research will advance knowledge in an efficient and effective way. It is, or can be, carried out by or in consultation with experts in educational theory. There is no problem in discussing the research project publicly, nor in reporting the results. The autonomy and moral development of the subjects is not enhanced by the experiment. But neither is it worsened. But these are the lesser considerations. Plainly then there is a case to be answered. Neither of my two exmples will stand up to the kinds of criticism applied to medical experiments. Experimentation of these kinds, it would appear, should not be undertaken. V Objections I have looked at the major differences between educational and medical experiments, and claimed that the differences are not so great that considerations which apply to the latter should not be applied to the former. However a determined experimenter, concerned about my conservative conclusion, might assert that a number of differences exist between the two realms. First difference: medical theory is scientific, while educational theory is not. Educational theory is bound up with folk psychology. Now I believe that this is true. But I do not see how folk psychology is to be abandoned. Amongst other things, it is required for scientists to engage in science, and for mathematicians to engage in mathematics. Nor can it be reduced to science, because of the differing commitments of the two kinds of theories. As Donald Davidson showed, lo these many years, these differences prevent any generalisations linking mental states with brain states from being law-like. So far as I can see, this makes the moral position of the classroom experimenter worse. For if this view is correct, there are no laws in education, and predictions of future benefits are therefore much more uncertain. Second difference: education does harm anyway. (This is not self-contradictory.) It is not even paradoxical.) In this education is unlike medicine, which only sometimes does harm. In an alternative version or the objection, schooling is an experiment which has failed. Reply: I don't think the claim is true. If it were true, we would need some reason to believe that the experiment would improve things. As things stand, it promises only more competent schooling. Third difference: The very notion of improvement is obscure in education, because aims vary, as do ideals of the person. Hence risk/benefit analysis is vague. Reply: This is palpably true. It is all the more reason to involve students in experimental decision making, as in other kinds of educational decisions. Fourth difference: children can understand tolerably the risks to their health. They can't understand what education is until they are educated. They should not have the right to close off their own options for development. Reply: this calls for parental consent as well. As in the medical case, the latter should not replace the child's consent. And it can't free the experimenter from the guilt of deliberately continuing to use inferior teaching methods for the sake of the contribution to knowledge. Fifth difference: medical experiments have irreversible consequences. Miseducation can be reversed. Reply: this is not known to be true in all cases. Language development needs to occur at an early age, and a child's intelligence may well be associated with this. It gets much harder to produce flexibility in mathematical thought after the age of twelve. Moreover, reversing poor education is very difficult. There are motivational consequences as well as those connected with knowledge. It takes a lot of time to unlearn something quickly learned, especially if that thing is a pattern of thought. And we do not have provision for remedial work -- we don't even know how to go about it so that students catch up and pass their fellows. Sixth difference: educational harm is not as serious as medical harm. Reply: the consequences for education are being an inferior person, lacking power over matters that concern you, lacking autonomy. If you have to get someone to do your mathematics for you, you waste a lot of life. This shortens effective life. The inability to appreciate classical music, to understand and enjoy poetry, to follow and engage in moral argument might be compared to having disfiguring scars, or lacking a limb. Seventh difference: the parallel should be with research on the sick, not on the healthy. Children are not full persons; education is designed to make them so -- to help them approach an ideal. But schooling fails to achieve these things for some -- perhaps for most. In other words, we are attempting a cure, but lack knowledge of how to effect it. More compulsion and less consultation is permissible on sick children, especially with illnesses of no known cause. Reply: personal ideals are a matter of choice. If there is a parallel with sickness, it is sickness for which alternative treatments lead to different life-styles. Consultation with the patient is required. Moreover, half-way through Fletcher's beriberi experiment he was dealing with sick people. We don't condone his failure to cure them, even though we know much more about deficiency diseases in consequence. Eighth difference: You couldn't do educational research on adults the way you can with children. They wouldn't put up with it. Reply: neither should children. Conclusion Ethically acceptable experimentation in schools would not be experiments on children, but experiments with children. That however would require a transformation in the way teachers and students see their roles; in the power relations within schools. But that is another story. Postscript It is plain that medical experiments can have educational consequences. Loss of schooling is only the most obvious. For that, the researcher must prepare, and make sure that any loss is made up. But a properly conducted experiment can have worthwhile educational consequences too. If the patient is involved in the planning of the experiment, or at least is fully informed and consulted, participating may lead to that personal growth which is the aim of education. The educational task of medical staff need not, should not stop at preventative medicine. References Annas, George J., Glantz, Leonard H. and Katz, Barbara F. (1977) Informed Consent to Human Experimentation: The Subject's Dilemma Cambridge, Mass., Ballinger Bibby, Martin (1985) `Aims and Rights', Educational Philosophy and Theory Vol 7 No 2 Bibby, Martin (1989) `Children's Rights: A Philosophical Analysis' Australian Social Policy Vol 1 No 2 Campbell, Alistair (1990) `A Report From New Zealand: An "Unfortunate Experiment"' Bioethics Vol 4 Davidson, Donald (1970) `Mental Events' in L. Foster and Davis, Phillip J. and Hersh, Reuben (1989) Descartes' Dream Nicholson, Richard H. (1986) Medical Research With Children: Ethics, Law and Practice Oxford, Oxford University Press Tooley, Michael `Abortion and Infanticide' in J. Feinberg, (ed) Abortion Volrath, John (1990) `Experiments and Rights' Bioethics Vol 4 No 2