1 16 Jan 2006 Dear Professor Schum I was very interested to read your recent paper, “Thoughts about a science of evidence”, that Phil Dawid circulated, and would like to offer you my thoughts in response. First though, a disclaimer: I am not well read either in your earlier writings or in those of such important people as Wigmore, so this could be considered an outside, even a naïve, viewpoint. I offer it in the belief that such viewpoints sometimes offer a fresh perspective. It is more about the overall shape of what you are setting out to do, rather than its content. I am in sympathy with your attempt to create a unified discipline of evidence. In general I am persuaded that there are some general features of the use of evidence that are applicable across disciplines. In addition, however, there are specific features relating to how evidence is used in different disciplines, that it would be unwise to neglect - to that extent I think it would be useful to pay attention to these specific features, as they could throw light on the unifying project as well in other words, to use a comparative methodology to shed light on the core structure. There are two basic reasons why disciplines could differ: (a) one is that the subject matter imposes distinct constraints or offers unique opportunities, (b) the other is that there could be a tradition that has evolved in a particular discipline, and by having a comparative perspective one could examine the strengths and weaknesses of the differing methodologies that have evolved. This could lead to cross-fertilisation, as I am aware that in my own disciplines (epidemiology, and also economics) there are limitations and weaknesses where a comparative perspective would be valuable. So, this would provide one type of answer to the question that you come to near the end of your paper, what the benefit of a science of evidence would be - it would enable each discipline to benefit from a fresh look at its own practices in the light of what others do. (They might not want to do this, but it would be good for them!) As I said, I found your paper very stimulating. My main problem with it is that you do not start by making a clear distinction between two fundamentally distinct intellectual processes: (i) trying to establish general relationships, as in the natural and social sciences, and (ii) trying to establish what happened in a particular instance, as in history, a criminal law case, or medicine. It seems clear that you are concerned with (ii), from the way the paper is introduced, and most of the examples given throughout, but there are some occasions where you stray onto (i) and seem rather lost, as there is no prior framework to deal with the rather different issues involved. The analysis would be greatly strengthened by making this clear, and also touching base with the intellectual process that you don’t deal with, namely the seeking out of generalities, in the extreme case the establishment of general laws - even in the absence of deterministic laws one could say that the object was to uncover regularities that could indicate the general type of process that is generating a large number of related instances. So, be explicit that you are dealing with, say, history rather than social science, or medicine rather than biomedical science. The case of law is more problematic, as there is no clear “generalising” equivalent of the individual case, only the use of commonsense bolstered by expert opinion (appeal to authority, as in the European “Dark Ages”) - the response that this function is provided by the body of law misses the point, as in each jurisdiction there is a different body of law, with its own rules of evidence etc, so there is a degree of arbitrariness about the situation - although one could argue that jurisprudence partially fills this gap, to the extent that it recognises this arbitrariness. This leads to one area where I disagree with what you say. Your statement, “All conclusions reached from evidence are necessarily probabilistic for five reasons. Our evidence is never complete, is usually inconclusive, frequently 2 ambiguous, commonly dissonant to some degree, and comes to us from sources having any gradation of credibility shy of perfection.“ would be hard to justify about the many scientific statements that have attained the level of certainty. The shape of the earth would be an obvious one - admittedly trivial nowadays, but that is because it is now certain and therefore no longer interesting scientifically. In my own field of biomedical science, a similarly ancient one would be the mode of the circulation of the blood. A more interesting one is that communication between nerve cells is not electrical but chemical - as we all know nowadays because we can buy products fortified by neurotransmitters - but this was a controversy not many decades ago. There are countless other example: the source, chemical structure and physiological effects of insulin would be another. So in relation to general statements, science is capable of generating certainties - over a period of time and a process of rigorous testing - potential falsification if you accept the Popperian view. It is only when the focus is on specific cases that your statement quoted above is justifiable. It also leads to a problem in your discussion of philosophy of science, for example the views of Poincaré, which could be confusing because the writers cited were (I think) mainly discussing the process of scientific discovery in the sense of uncovering generalities, rather than dealing with specific instances (the same is true in relation to Darwin and your reference 91 on page 33 - and possibly your quote from Bentham on page 67) - although a close reading of your text does make it clear that your interest here is in a “science” of evidence, i.e. you are not concerned with science as your object of study, but rather that your study of the use of evidence can be given the status of a science - so, not ontology but epistemology. This means that the generalising is in your account of how evidence can be validly used to reach a conclusion, whereas the type of conclusion you are interested in is the single event. It might help if this were spelt out more clearly. When we get to your actual account of evidence, the breakdown you give of “credibility” is suitable for the specific-instance case, but not for the generalising one - in other words, not for generating generalisations, where repeatability (for example) would figure prominently. Also, reliability only enters into your description in relation to “the process used in producing the tangible item”, not to the observation itself - and this reliability of observations is central to any generalising discipline, including science. In my (inevitably limited) experience, scientists regard replication of findings as being one of the keystones in relation to credibility. This affects the analysis that you present: your account of the rows of figure 1 is only applicable to the purpose of studying a specific instance, not to a generalising purpose, although there could well be equivalent but different things to say in the generalising case. Moreover, your later remarks on competence and credibility (page 42) would look very different in the generalising case, not only due to the possibility of repeat observations, but also (in science and other academic disciplines) the fact that the observations are usually made by people who have a particular kind of expertise - while this is not problem-free, it is different from relying on the testimonial of e.g. bystanders to a crime. Similar remarks apply to the question of inferential force (page 48), and to evidential completeness (page 54). I think that your focus on evidence charts, following Wigmore, is good. Causal diagrams are also useful, and used, in the generalising context as well (actually there are many different versions in the latter case), and I think this is what you illustrate as figure 2B. It is worth noticing that the arrows mean different things in the two cases: in a “generalising” diagram, an arrow indicates causation “in the real world out there” from one item to the next (or, as you say, relevance relations or probabilistic dependencies), whereas in a specific-event diagram, an arrow indicates something being evidence for something - the first case is ontology, the second is epistemology. Also, in the generalising context, an arrow is necessarily in the same direction as the passage of time - the “caused” item follows the “causal” one - with special considerations where expectations are involved, as is common in economics. Whereas in inference about the individual event this is not necessarily the case, as the chain is a chain of reasoning and (I think) has no clear time 3 order. It seems to me that this is the same distinction as the issue you discuss of zero probability in the ontological and the inferential (epistemological) case, as in “ordinary probabilities” versus Shafer’s “evidential support” on page 52. Another way of looking at it is that it is the difference between causal diagrams and mental maps. I would also like to argue against the use of the word “science” to characterise the discipline of evidence - and not only because it tends to antagonise some of our colleagues in non-scientific disciplines. You repeatedly draw on Carnap’s threefold characterisation of science - but many would argue that understanding or explanation is an important component of a science; possibly it is the main aim, to which these three components contribute. (I will not attempt a philosophical definition of these terms.) In biology, one could speak of the aim of uncovering the inner mechanism that underlies an observed process, as in the neurotransmitter example above: observations of neuron-to-neuron transmission are explained by finding various items of evidence (the time delay between the neurons’ firing, seeing vesicles of fluid on electron microscopy, chemical isolation of the transmitter, its physiological effects when injected, chemical isolation of enzymes to synthesise and remove the transmitter, etc) - in other words, explanation is description at a deeper level. I don’t think that the discipline of evidence is “a science” in this sense, as there is no ontological process that we are trying to explain or understand. Rather, it is uncovering the “rules of the game” - this may be just as important, but it is different. It has methodology (epistemology) rather than (ontological) subject matter - as you say at one point: “The testing here seems to be logical rather than empirical in nature”. Finally, a minor point (that is well outside my area of competence): your mention of the work in Oxford in and around the thirteenth century could be linked with your comment on the neglect of learning outside Europe during this period and before (page 9) - I understand that the early work in Oxford drew heavily on texts that were being translated from Arabic at that time, in places like Toledo, so that there is a direct link with the glorious Islamic civilisation of Iberia that was by then beginning to decline. The same is true of Paris. These translations included the European ancients (Euclid, Archimedes, Aristotle, etc), but also gave Europe access to learning that was originally Islamic, Jewish, Indian, etc - how much longer would it have taken to develop probability theory without importing the Hindu concept of zero? Would it have happened at all? Best wishes Mike Joffe 4 A Reply to Michael Joffe's Comments on: Thoughts about a Science of Evidence D. Schum, 6 February, 2006 I was so pleased to receive Michael Joffe's helpful and detailed comments on my first attempt to state my views about a science of evidence 1. Some of Michael's comments are quite critical and I will try to answer them as best I can. As I noted at the outset of my paper, I expected that many things I was prepared to say about evidence would excite critical comment. One reason is that the evidence-based reasoning tasks we all perform in the work we do, as well as in our daily lives, have different characteristics, only some of which may be common across disciplines and contexts. Though I have examined thoughts about evidence in a variety of disciplines and contexts, I have never claimed any expertise in all but a very few of them. I once argued that the house of evidence has many mansions and that I have been just a visitor in some of them2. Michael's fields of interest are in epidemiology and economics, two mansions I have only visited very briefly. My replies to follow mainly concern Michael's critical comments. My hope is that you will see that I have taken the same degree of care in answering Michael's comments that he did in reading my thoughts about a science of evidence. The page numbers below refer to pages in my original paper3. 1) Emphasizing Differences. Michael begins by noting that we should focus on differences across disciplines/contexts as far as the study of evidence is concerned. Any focus just on similarities or commonalities across disciplines would not be as helpful. On pages 67 - 68 I noted that William Twining received the same comment from a person who agued that his emphasis on similarities across law and history in studies of evidence is not enough; we should pay attention to disciplinary differences in the study of evidence as well. On page 67 I noted the historian Marc Bloch's comment that persons in any science should attempt to see the connections between their own methods of investigation with "all simultaneous tendencies in other fields". Though he does not say so explicitly, study of such connections would presumably involve the observation of methodological differences as well as similarities. Michael's point is well taken here since the "integrated science of evidence" advertised in our UCL work should not be interpreted as a focus just on similarities across disciplines in studies of the properties, uses and discovery of evidence. We will all learn more about evidence and inference when differences in our various approaches to evidence-based reasoning are studied carefully. 2) Two Distinct Intellectual Processes. Michael says that I should have started off by emphasizing the distinction between two fundamentally different intellectual processes: (i) Those involved in the establishment of general relationships, as in the natural and social sciences, and (ii) Those involved in establishing what happened in a particular instance as in law, history and medicine. He says that most of my examples involve (ii) and that I seem rather lost on the occasions when I touch upon (i), since I provide no prior general framework for discussions of (i) that do indeed involve different issues. If I ever do a revised version of this paper, I would start out in the same way I started in my present paper by commenting on the emergence and mutation of the concepts of evidence and science. But I would then consider following Michael's advice in dwelling upon distinctions between situations (i) and (ii) in Section 4.0 when I begin to comment on elements of a science of 1 As given me in an e-mail message from Michael on January 16, 2006. 2 Schum, D., Evidential Foundations of Probabilistic Reasoning. John Wiley & Sons. New York, NY., 1994; Northwestern University Press, Evanston, IL., 2001, paperback, xiv 3 Schum, D. Thoughts about a Science of Evidence. UCL Studies of Evidence Science. 29 December, 2005 5 evidence. This seems to be the place for discussion of these two situations and how they affect my subsequent comments on evidence. The trouble I would encounter in implementing Michael's suggestion about the necessity for a "prior general framework" for discussing the "establishment of general relations" is that many such frameworks have been proposed and debated over the centuries; they are still being debated. A good place to start a review of these alternative frameworks is in the work of David Oldroyd on the "arch of knowledge" that I cited on page 18. Another work that addresses these same matters is the more recent book by Peter Achinstein that I mentioned on page 19. In these works we find reference to the thoughts of many learned persons, such as Galileo, Francis Bacon, Robert Hooke, Isaac Newton, John Herschel, William Whewell, Wesley Salmon, Bas van Fraassen and many others, on the intellectual routes to be taken from observations to lawful relations to general theories and to the testing of these theories. The routes suggested by all of these persons are quite different. So I have doubts about what I should take as the "prior general framework" for the establishment of any general relations. I think I understand what are the basic distinctions between situations (i) and (ii) that Michael has identified. In the natural, behavioral and social sciences Michael mentions, there are many efforts to find regularities or invariances in those parts of nature that are of interest to us. Here I quote from a very recent paper by the physicist Lee Smolen on methods of science that Smolen says are not always employed by some researchers in physics 4: Science works because it is based on methods that allow well-trained people of good faith, who initially disagree, to come to consensus about what can be rationally deduced from publicly available evidence. One of the most fundamental principles of science has been that we only consider as possibly true those theories that are vulnerable to being shown false by doable experiments. [Italics mine] I have highlighted the expressions publicly available evidence and doable experiments for the reason that by experiments we put questions to nature to see how she will answer them. Answers we think we have found must be made available publicly for the scrutiny of others. The trouble is that nature will not answer any old question. She will rarely, if ever, answer general questions. For example, if we ask her: How does the human eye work?, we will get no answer. But if we ask such specific questions as: What is the minimum amount of radiant energy, at a fixed wavelength, that the normal human eye can reliably detect?, we may begin to obtain answers. But we will not of course shine only one person's eye with a single flash of light and report what happened. We will put the eyes of many persons to tests involving many trials and many light flashes under a variety of conditions such as: Where on the retinal surface are we directing the light flash? And, how long have the persons' eyes been allowed to become darkadapted before the tests begin? The point here is that the doable experiments Smolen mentions assume replicable or repeatable phenomena. There are several reasons why we repeat experimental trials over and over again. In the first place, nature will not always respond in exactly the same way to a set of conditions we believe are identical. In short, there will usually be some natural variability in the processes we are studying. In addition, the devices and procedures we are using to collect our observations are further sources of variability. For such reasons we commonly employ statistical indices of various sorts to grade the reliability and accuracy of our observations. But statistical indices are always numerical indications of what has happened in the sample of observations we have taken. The inferential role of statistics involves assessments of the extent to which results we obtain in our sample generalize or apply to the population from which we believe our sample has come. One trouble is that we usually have a choice from among several descriptive and 4 Smolen, L. A Crisis In Fundamental Physics. Update: New York Academy of Sciences Magazine, January/February, 2006, 10 - 14. 6 inferential statistical indices we could employ. What we try to do is to employ statistics that are minimally misleading. I will come later to "statistical evidence" in comments about my classification of evidence. So, one basic element of Michael's (i) concerning the discovery of generalizations, regularities or invariances in nature involves experiments that are repeated observations of the process of interest. Here are three examples I have drawn from entirely different areas: one from particle physics, one from cell biology, and one from experimental psychology. One of the major things these examples will illustrate is that the instruments for gathering evidence that nature provides are quite different in various areas of science. It is quite natural that we must employ different methods depending on what we are looking for. I have read several accounts in which areas of science are distinguished just in terms of the instruments employed to collect relevant observations. Persons in each one of these three areas might have the same inferential objectives as mentioned above by Lee Smolen. But the evidence in all of these three situations seems to have a common characteristic, namely it is tangible in nature and open to the inspection of persons interested in conclusions that are drawn from it. . 1) Particle Physics. Physicists have developed and used a variety of devices such as cloud chambers, bubble chambers, spark chambers and newer computer-assisted electronic detectors to observe and study the nature and movement of atomic particles. Such devices allow the recording of tracks taken by these particles. Analysis of these tracks provide evidence relating to such matters as the mass and charge of the particles under study. In the use of earlier methods of tracking particles [e.g. cloud chambers], photographs of these particle tracks were taken and analyzed. This very laborious process is now taken over by computers that can analyze thousands in track images in minutes. Such devices provide visible evidence regarding the nature and movement of the particles under investigation. 2) Cell Biology. For various reasons concerning other interests I have, I have taken quite an interest in structures called microtubules that are elements of the cytostructure of all eukaryotic cells, including the neurons in our brains. Patterns of synaptic connections among neurons, and their apparent "all or none" firing characteristics, have led many researchers to the view that our brains are digital computers and that the individual neurons are simply switches. But this view has been challenged by the mathematician Roger Penrose and the anesthesiologist Stuart Hammeroff5. They argue that this "digital view" results from our examining the brain and its neurons at the wrong level. We have to examine the substructures of neurons, especially their microtubules, and when we do this we will see that each neuron is itself a sophisticated computer, more like a chip than a simple switch. Thanks to the imaging power of today's microscopes, we can observe the structure of microtubules directly and repeatedly. Photographs of them appear in several references 6. On many accounts microtubules are simply part of the cytoskeleton of cells. But they are also known to exhibit quite rapid alterations as a result of the polymerization and depolymerization of the tubulin molecules of which they are composed. There is evidence that this helps account for the plasticity of brain in which new synaptic connections are forming all the time as we learn new things and have new experiences. The very orderly arrangement of tubulin molecules in a microtubule has suggested to Penrose and Hameroff that each may play the role of a cellular automaton that can respond to activities in neighboring microtubules. They offer the view that the computation involved may be quantal rather than digital in nature. So, here is another example of a phenomenon that can be 5 E. g. Penrose, R., Shadows of the Mind: A Search for the Missing Science of Consciousness. Oxford University Press, Oxford, 1994, 357 - 392; Penrose, R. The Large, the Small and the Human Mind. Cambridge University Press, Cambridge, 1997, 127 - 143. 6 Ibid, Penrose, 1994, 364; Alberts, B., Bray, D., Lewis, J., Raff, M., Roberts, K., Watson, J. Molecular Biology of the Cell. 3rd ed. Garland Publishing, Inc. New York, NY., 1994, pages 803, Fig 16-21. 7 observed directly, in this case by means of a microscope. What is at issue is what observations of this phenomenon mean. Not every researcher has drawn the same conclusions about the role of microtubules that Penrose and Hammeroff have done. 3) Experimental Psychology. I come now to an example of empirical research in an area in which Nigel Harvey and I have had great interest. The area is now called behavioral decision theory and it includes studies of the inferential and choice behavior of people. My graduate training was in experimental psychology and mathematics. Nigel and I certainly share a common experience, namely being exposed to many courses on empirical methods that have been found useful in the behavioral sciences. During the first ten years or so in my career, my research was empirical in nature and involved studies of the capabilities and limitations of persons in their ability to assess the inferential force of evidence. I studied these aspects of our behavior in a variety of situations, some involving entirely abstract tasks and others that involved specific substantive situations7. All of my studies involved repeated trials under as carefully controlled situations as I could manage. What I observed, and recorded, from human subjects were numerical assessments of the probabilistic judgments they made in response to patterns of evidence with which they were provided. Such research is an example of what used to be called a "back box" approach to the study of mental activities. In research on human probabilistic judgments the idea was that we cannot get inside a person's head to see how he/she is evaluating evidence. All we can do is to put something into a person's head and observe how this person will respond to it. Analyses of what went in and what come out were, and still are, thought by many psychologists to provide insights into the activities going on inside the brains [the "back boxes"] of the persons whose inferential behavior is being studied. There are many questionable assumptions in such analyses that are commonly recognized. Though I provided statistical assessments of the behavior of samples of individuals I studied, I was also careful to provide analyses of judgments provided by individuals. As we know, it so often happens that a statistical account of members of a group actually describe the behavior of no person in the group of persons being studied. In any case, as in Examples 1 and 2 above, I provided tangible evidence that could be examined by others.. I just gave three different examples of instances in which researchers are attempting to identify general properties and relations of the phenomena of interest to them. All of these efforts involve repeated observations of physical, behavioral or social phenomena. As I noted, variability is inherent in these observations; this is one reason why they are repeated. Now contrast these situations with those Michael mentioned that are associated with his (ii) above, such as in law or history. We cannot play the world over again 1000 times in order to see on how many of these occasions O. J. Simpson murdered his wife Nicole Brown Simpson, or the number of occasions on which Mary Queen of Scots was witting of, or participated in, the murder of her husband Lord Darnley. In many situations, such as intelligence analysis, we try to predict the occurrence of future events such as: will terrorists launch an attack on the large crowd expected for the football match at Wembley Stadium this coming Friday? If this event happens, and we hope it does not, it will happen exactly once and under a unique set of conditions. In all of these type (ii) situations the events of concern are unique, singular or one-of-akind. If they happened in the past, they did so on just one occasion. If they happen in future they will do so on just one unique occasion. Though it does happen that in some of these situations we can use statistical information of various kinds. In law, for example, we can use statistical information concerning various properties of trace evidence, such as DNA samples, glass shards, 7 E. g. Schum, D. Inferences on the Basis of Conditionally Independent Data. Journal of Experimental Psychology, Vol. 72, No.3, 1966, 401 - 409; Schum, D., Martin, A. Formal and Empirical Research on Cascaded Inference in Jurisprudence. Law and Society Review, Vol. 17, 1982, 105 - 151. 8 and cat hairs. But most of the evidence we obtain comes in the form of observations or reports of events that are also unique, singular or one-of-kind. Michael notes that virtually all of my examples were of type (ii) situations. He's right of course. I asked myself why I did not include more of type (i) examples; I certainly would not have had any trouble doing so. I guess there are two reasons. The first is that I tended to give examples of situations I have studied most extensively. They have typically involved situations in which we must deal with most or all of the forms and combinations of evidence I mentioned in my classification of evidence in Figure 1 on page 24. Empirical studies of the sorts I have mentioned in my three examples above rest primarily on various kinds of tangible evidence I described above. It is true, of course, that researchers in Michael's type (i) situations do of course rely on the testimony of colleagues concerning their reactions to tangible results being discussed. The second reason involves my expectations about the persons most likely to read my thoughts about a science of evidence. My experience has been that I have received many comments on book and papers I have written about evidence from persons whose inferential work falls in Michael's category (ii). But I have received very few comments from persons in areas of science whose work falls in his category (i). I know that William Twining has had this same experience. I wonder how many persons in areas of science interested in finding generalizations, law, or theories would take Mr. Grodman seriously that a science of evidence is "the science of science" [see page 6]. Wigmore certainly did but his interests were in the field of law in which every case has unique elements. Finally, Michael says that I seem "lost" in discussing matters concerning his (i)) since I gave no prior framework for discussing the issues bearing on these kinds of evidential reasoning. I may have seemed lost, but I did not feel lost while writing my thoughts on evidence. In any further version I may offer of a science of evidence I will certainly include a discussion similar to the one I have just given above for the evidential characteristics of the sciences he mentions. But Michael should then advise me about which one of the many "prior frameworks" I should adopt in my discussion. I will return to Michael's important distinction between situations (I) and (ii) in other comments I have about forms and combinations of evidence. 3) Probabilistic Conclusions Michael has taken me to task in my discussion of the evidential reasons for probabilistic conclusions for not acknowledging that many conclusions in science have now "attained the level of certainty", as he puts it. He gives some examples in various areas of science that seem to fall in my evidence category called "accepted facts" [Figure 1, page 6]. There are of course unique events in other inferential contexts that have risen to the level of certainty. One example would be that the twin towers of the World Trade Center in New York City were leveled in a terrorist action on September 11, 2001. I admit that my statement [page 17] needs a bit of qualification along the lines suggested by Michael's comments. I might better have introduced the evidential grounds for the probabilistic nature of conclusions by saying something like: "Except for certain instances in science and in other situations that can be identified, conclusions reached from evidence are commonly probabilistic for the following reasons:…". However, at the same time I insert this hedge, I note two things. I wonder whether there are any instances in science or elsewhere when our evidence is utterly complete. I also note that what is regarded as "fact" today may not be so regarded in the future. I have always taken an interest in what people through the ages have thought was the site of our mental functions. For example, at various times these functions were thought to be centered in the heart. Until the 1600s the brain was regarded as a "nondescript mass of flesh glued inside the skull"8. It was 8 Zimmer, C. Soul Made Flesh: The Discovery of the Brain - And How it Changed the World. Free Press, New York, NY., 2004, 175 9 Thomas Willis [1621 - 1675] who first studied it as an independent organ and made so many discoveries of its properties. This is why he is now called the father of neurology. 4) Poincare´ on Classification in Science Michael says that my reliance on Poincare´s thoughts about the importance of classification in science could be confusing since he was probably making reference to situation (i) and not to situations (ii). Recall that Michael's situation (i) refers to the generation and empirical testing of theories, laws and generalizations in various areas of the natural and social sciences. Situation (ii) refers to evidence-based reasoning about events in specific cases that are of interest in such areas as law, history and so on. First, I would agree with Michael that Poincare´ was almost certainly thinking about situation (i) in his remarks. In fact, I can strengthen Michael's argument by noting other things that Poincare´ said. Speaking of theories in science, he said9: At first blush it seems to us that the theories last only a day and that ruins upon ruins accumulate. Today the theories are born, tomorrow they are the fashion, the day after tomorrow they are classic, the fourth day they are superannuated, and the fifth day they are forgotten. But if we look more closely, we see that what thus succumb are the theories, properly so called, those which pretend to teach us what things are. But there is in them something that usually survives. If one of them has taught us a true relation, this relation is definitively acquired, and it will be found again under a new disguise in the other theories which will successively come to reign in place of the old. All this acknowledged about Poincare´s apparent emphasis on situation (i), I fail to see what he said about the importance of being able to classify things is any less important, or even unimportant, in situation (ii). I went to considerable lengths in my paper to show the "hidden kinship", that Poincare´ mentions regarding facts which "appearances separate" [page 23], between tangible evidence in the diverse fields of theatre iconography and law, two areas that are definitely in situation (ii). They are united by means of the common credibility issues they both raise. I would have had no difficulty at all in finding similar "hidden kinships" among items of testimonial evidence in many substantively different areas in which we have only situation (ii) objectives. One final comment here concerns Michael's related comment that my interests involve study of a "science of evidence" rather than the study of science itself. I'll first reply that, as I said on page 66, I have taken no position on Mr. Grodman's assertion [page 6] that the "science of evidence" is also the "science of science". I gave reasons for my hesitation in agreeing with Grodman's assertion. At the same time, in relating the terms "evidence" and "science" I found it necessary to say how these two terms are used and how they have changed over the ages, which I did in my section 3.0. I also offered a variety of comments throughout my paper on the extent to which the study of evidence has characteristics that would allow us to refer to such study as being a science. I will have more to add on this matter in my reply to Michael's next comments. 5) On My Classification of Evidence and Its Basis 9 Poincare´, J. H. The Value of Science [1905]. In: Gould, S. J. [ed] Henri Poincare´, The Value of Science: Essential Writings of Henri Poincare´. The Modern Library, New York, NY., 2001, 348 - 349. 10 Michael then comments on what I said about credibility, how I described the rows in my classification scheme in Figure 1, and what I said about inferential force and completeness. He begins by saying that my account of credibility is suitable only for situation (ii) and not for situation (i). He says further that my account of the rows in Figure 1 is applicable only to (ii) and not to (i). Here I take issue with Michael. Mark you, I might not have stated matters as completely as I should have, but what I did say I believe to be correct that Table 1 accounts for evidence we have in both situations (i) and (ii). First, I began by describing the tangible evidence that is so common in Michael's situation (i); it is also common in situation (ii). Though I did not specifically mention photo or computer images of the particle tracks and microtubules mentioned in my examples above in discussing situation (i), they form tangible evidence that we can all examine. In the same way, in situation (ii) we can examine Peg Katritzky's photos of mountebank drawings and photos of Bullet III in records of the Sacco and Vanzetti trial [page 31] to see what they reveal. . I pause here to consider what tangible evidence provides in either situation (i) or (ii). Both Michael and I agree that the scientists with (i) objectives must make their results and analyses publicly available. In offering tangible evidence the scientist essentially says: "I will do my best to make you privy to the same things I have observed so that you can draw your own conclusions about whether my explanation of them is correct". The same statement can be made in situation (ii) for the mountebank photos and Bullet III. Now, in my Example 3 above, concerning experimental psychology in situation (i), I could not show you any photos, other images or objects concerning what went on inside my subjects' heads as they made their probabilistic judgments for the reason that I had no such objects or images myself. But I did show interested persons tangible records of the actual numerical responses my experimental subjects were asked to provide. So, I could have made the same statement that I just mentioned above: "Here is what I observed, see for yourself whether my conclusions about them were correct". As far as credibility issues are concerned regarding tangible evidence, I believe the same questions are asked in situations (i) and (ii) that I mentioned on page 25. Regarding authenticity questions in situation (i), in most cases [happily] scientists will not believe their colleagues to be publicly propagating inauthentic tangible evidence. I regard the photos I have seen of particle tracks and microtubules to be authentic and not contrived. None of the tangible accounts of my subjects' probabilistic responses were ever questioned as far as their authenticity was concerned. You have my word that I did not make up the data I reported. Unfortunately, every now and then there are cases [frequently well-publicized] in which scientific data are questioned on authenticity grounds. A current example involves the stem cell research reported by a Korean scientist. In situation (ii) I have never even thought about questioning the authenticity of the mountebank drawings Katritzky has shown us. But there certainly has been much interest in the authenticity of Bullet III in the Sacco and Vanzetti trial and whether it was one that came from the body of the slain payroll guard. As far as reliability is concerned, this credibility attribute is certainly evident in situation (i). What replication allows is grading the extent to which we get the same tangible results over and over again. I mentioned above the two major sources of variability that accompany investigations in situation (i). This is one reason why we use statistical indices to grade the extent of this variability. I agree with Michael when he says that replicability is one of the keystones of credibility in situation (i). Concerning accuracy, I think such concern is also evident in situation (i). There are continuing efforts to increase the accuracy with which we can observe natural phenomena. The microtubules I mentioned above in my second example in situation (i) provide a good example. In the past, the stains used in the preparation of neurons for their microscopic examination obliterated the internal molecular structure of microtubules. Now, different stains are used and we can get a closer look at the orderly arrangement of the tubulin molecules in microtubles and the changes wrought in them as a result of polymerization and depolymerization. Another example of course involves the drastically increased resolution provided by the Hubble space telescope in making cosmological observations. 11 Regarding the other rows in my evidence classification in Figure 1, it may not be so important for scientists in situation (i) to be concerned about the various species of testimonial evidence I described in Rows 2 and 3. However, Michael makes the point that the people making statements about what they have observed in situation (i) are "experts" and not ordinary people, such as bystanders, who report on what they have observed. Here I should have mentioned that there are rules in our legal system, such as our FREs 702 - 705, that concern opinion evidence given by experts in scientific, technological or other areas of specialized knowledge. Among other things, it must be demonstrated that an expert witness is indeed qualified to provide information in these areas. Just recently I received some very valuable comments from a Dutch Appeals Court Judge, John van Voorhout. He argues that, in his experience, all expert opinion evidence is equivocal in nature. I had to agree with him and mentioned that it could be argued that all opinion evidence, given by experts or anyone else is, or should be, stated equivocally. The argument goes something like this. Here is a person [expert or otherwise] who asserts that he has evidence about events A, B, and C, that allowed him to make the inference, or form the opinion, that event D also occurred, as he/she reported unequivocally in testimony D*. We are entitled to ask: "Give us reasons why you believe that the occurrence of events A, B, and C have made the occurrence of event D necessary or certain to have occurred?" What I should do in my Figure 1 is to say that equivocal testimony also has the same three grounds or bases and unequivocal testimony: direct observation, at second hand, or opinion based on other evidence. Finally, I notice that Michael did not inform me about any form of evidence that my Figure 1 does not include. What is certainly true, and that I have recognized, is that persons in situations (i) and (ii) will make use of different mixtures of the forms of evidence I have listed in my Figure 1. As I have noted above, tangible evidence will predominate as grounds for the inferences made by persons in situation (i). I was asked on one occasion why I do not have a category of evidence called "statistical evidence". The reason I gave is that all statistical evidence involves counts of either observed tangible evidence or some form of testimonial evidence, such as encountered in survey research. Such statistics might also employ counts of missing tangible or testimonial evidence. In short, I do not believe that statistical evidence involves evidence my categorization of evidence does not already cover. 6) Concerning the Arcs or Links in Inference Networks Here I have no disagreement with what Michael has said regarding the forms of inference networks I showed in my Figure 2 [page 35]. I just have a few things to add about the interpretation of the arcs [arrows] in these two figures. It turns out that the interpretation of the arcs in both figures is a matter for controversy. In the Wigmore case in Figure 2A, Wigmore himself said the arcs represented lines of probative [inferential] force. William Twining says the arcs here say: "tends to support or tends to negate". I have always said that the arcs on a Wigmore chart represent probabilistic relevance relations, since this is what the arguments being charted are intended to capture. There is no temporal or causal intent here; argument charting is not the same thing as telling a story involving "chains of evidence". It would be a violent non sequitur to argue that the existence of one item of evidence allows an inference about the existence of another item of evidence. Arguments continue about the meaning of the arcs in the process models I illustrated in Figure 2B. Some argue that they indicate causal linkages and some argue that they only indicate avenues of probabilistic influences or dependencies. But Michael is correct in saying that we may usually see process models and not Wigmore charts in situation (i). 7) Arguments Against the Idea of a "Science" of Evidence. 12 Michael argues against the use of the word "science" to characterize the discipline of evidence. He first notes, as I did in my paper, that some persons not involved in what is normally called science might be antagonized by such a description. I confess that I have heard about this possible antagonism but frankly wonder why it should exist. I pose another case of the use of the term science, it involves the well-established field of computer science. My strong guess is that anyone Michael could name whose work is in no way associated with situation (i), almost certainly now uses computers for some purpose in their work, even if just for word-processing. My equally strong guess is that they are not antagonized by the use of the term computer science. Why should they be? Here are two persons who would probably say that their work is not associated with science, at least not with the activities in situation (i). Consider Marc Geller, who studies the time at which the Sumerian language became extinct, or Peg Katritzky, who studies theatre iconography. Why would anyone say that their work is associated with, or is just a subset of topics in, computer science, just because they use a PC [or a MAC] to type their papers? I doubt that either of them would object to the use of the term computer science. People in all walks of life use evidence every day of their lives, including those in situations (i), (ii), and areas possibly not covered by either of these situations that Michael has described. My own view is that persons in any area of (ii) or elsewhere should neither be intimidated in any way or antagonized by use of the term evidence science. No one I know of, least of all me, will ever claim that their work is a subset of evidence science just because their work involves the use of evidence. My view, which I believe consistent with the UCL objectives for the study of a science of evidence, is that it is a field of study that anyone can draw upon for insights about evidence, and who can even contribute their own insights if they choose to do so. Thus, folks like Marc and Peg can draw upon evidence science in much the same way that they now draw upon contributions from the field of computer science. But. Michael has a few deeper concerns about use of the term "evidence science". He first notes my reference to the three concepts of science noted by Carnap, saying that there are other concepts associated with science, such as explanation and understanding, that Carnap does not mention. I have read over Carnap's account of the three concepts he discusses and find nowhere in his writings that he considers his listing to be exhaustive. And it is certainly true that the two concepts Michael mentions are just as important in situation (ii) as they are in situation (i). But Michael says that there is no ontological process that a science of evidence seeks to explain. Rather, studies of evidence can only be epistemological in nature concerning what he calls "rules of the game". According to my encyclopedia of philosophy10, ontology refers to the investigation of existence or being. Common questions asked in such investigations are: What exists?, and What sorts of things are there? I could first argue that many of my studies of evidence have had ontological objectives. For many years now I have had an interest in studying how many kinds of individual evidence items, and combinations of them, exist when we do not consider their substance or content. I have never argued that my listings are final or exhaustive. I have changed my mind about the categories I have identified in the past, and I will probably do so in the future. There is so much still to be learned about evidence and its properties and uses. Such research can involve ontological objectives in the sense I have just mentioned. I'll turn now to what Michael says about the methods of evidence science being epistemological rather than ontological. In this connection he notes that I said [page 66] that the testing in evidence science "seems to be logical rather than empirical in nature". I'll first note that I discussed [pages 14 -15] the argument offered at the UCL meeting on 7 June, 2005 that a science of evidence would be contained within the field of epistemology. I also mentioned that any science [even those associated with situation (i)] involve epistemological issues. And, later 10 Craig, E. The Shorter Routledge Encyclopedia if Philosophy. Routledge, Oxford, 2005, 756 13 [pages 44 - 48], I went on to show how I used [or possibly misused] the standard account of knowledge in epistemology in identifying attributes of the credibility of witnesses in reporting what they [allegedly] observed. Returning to my view of testing in evidence science, on pages 59 - 62 I mentioned how the use of mathematical expressions for the inferential force of evidence can lead to the telling of alternative stories about the force of various combinations of evidence when the probabilistic ingredients of these expressions are varied. As I mentioned, exercises like this are frequently called "sensitivity analyses". If we agree on the structure of an evidence-based reasoning situation [using either of the structural devices I mentioned in Figure 2 on page 35] we can exercise mathematical expressions appropriate to these structures, such as those stemming from Bayes' rule, to see how they will respond to variations in probabilistic ingredients applied to the arcs in these models. I also mentioned the heuristic merit of such mathematical investigation [page 65]. These expressions may suggest questions we might not have thought of asking if we had not done such analyses. The question is: how do we test the results provided by these models? In many [most?] areas in situation (i) mathematical models are routinely used to guide research in some given area and to suggest questions we might ask of nature. If the questions are answerable by nature, we design empirical experiments to see whether nature will behave in the way our models say she will. But model-based or other studies of evidence, such as I mentioned above, do not lend themselves to such replicable experimentation. So, in the absence of nature answering our questions about evidence, we do the next best thing which is to ask ourselves whether what we have said about evidence makes sense and is a complete account of what we believe is involved in the evidential situation we are studying. This is what I meant by logical rather than empirical testing. I note that Michael did not comment on my discussions of how closely a science of evidence matches the definition of science given by the OED [pages 22 - 23; 65 - 66]. Most of his comments are based on his distinction between situations (i) and (ii) as he has defined them and which he says I failed to incorporate in my analysis of a science of evidence. I have not yet commented upon this distinction in hopes of giving Michael every benefit of doubt. I'll just note here that not everyone agrees with this distinction. Here comes the view of the noted historian Edward Hallet Carr11. He begins by saying: It is alleged that history deals with the unique and particular, and Science with the general and universal. This is what Michael has said. But Carr goes on to say that adoption of this view in history would lead to a "philosophical nirvana, in which nothing that matters could be said about anything. He argues that historians are not really interested in the unique, but what is general in the unique. I may be wrong here [and my colleagues William Twining and Terry Anderson will let me know], but Carr's comment about history brings to mind the concept of stare decisis in the field of law. This Latin phrase says: to abide by or adhere to settled cases. It is true that all cases in law have unique evidence. However, stare decisis says that when a court has laid down a principle of law as applicable to a certain pattern of facts [evidence], it will apply to all future cases in which the facts are "substantially the same". This sounds to me that the field of law also considers wat is general in the unique. I was honestly prepared to give up on the idea of there being a "science" of evidence when I first began to read Michael's comments. However, I will remain obstinate and still cling to the views expressed in my thoughts about a science of evidence, even though I have taken so 11 Carr, E. H. What is History? Random House, New York, NY. 1961, 79 -83. 14 many of Michael's thoughts to heart. But I have one more matter to discuss that has caused me some embarrassment. 8) Sources of Early Oxford Scholarship Michael correctly notes the contributions of Islamic scholars to that of the Oxford scholars I mentioned on page 9 - 10. I had earlier remarked on the contributions of these Islamic scholars and said that they were so often slighted in Western accounts of the history of science. Michael's comment suggests that I slighted them myself in what I have written about the emergence of the concept of evidence. I said that my discussion of the emergence of evidence and science would be "embarrassingly brief" [page 8], here is one example of my embarrassment. I should have mentioned how scholars like Ibn Sina [Avicenna] and Ibn Rushd [Averroes[ kept the wick turned up as far as the earlier contributions of the Greeks are concerned. And I should also have mentioned the frequently overlooked empirical research that persons like Ibn al-Haytham [Alhazen] performed. I first discovered, years ago, how we can credit person like Alhazen and Abu Al-Kindi for their works concerning how the human eye works. As I did note, I have never found any specific writings of early scholars in the Middle East, or elsewhere, on the use of evidence. But this does not excuse my not writing more about their contributions in science and in other areas. 9) Conclusion I will end where I started by thanking Michael for his thoughtful and extensive comments on my thoughts about a science of evidence. In my past experience I have written what I thought would be detailed and helpful comments about the works of others. In most instances I never heard one way or the other about the writers' reactions to my comments or whether, indeed, they had read my comments at all. I could not let Michael think that I did not take him very seriously on the comments he made about my thoughts about a science of evidence.