When we are following clues there are things we should do and things we should not. We may call these the Do's and Don'ts when following clues. When we do things we should do we increase our chance of finding those things we want to know. When we do things we should not we arrive at wrong results or prevent ourselves from arriving at any result at all. In this chapter we point out some of the Do's. In the next we point out some of the Don'ts.
When we follow clues we do so with a view to reconstructing some structure or structures that is hidden from us. A crime has been committed. To find out what happened we follow clues. A crime has a structure: there is a motive for the crime, which leads to certain actions; these actions take place in a certain order; and they have consequences which are not arbitrary. For example fingers not protected by gloves will leave behind fingerprints. By following clues we can find out what the motive for the crime was and also what happened. But this finding out is actually a reconstruction. Basing ourselves on clues we describe what is likely to have taken place; we do so without having actually witnessed the crime (since we were not present when the crime was committed). This reconstruction will not be perfect in every detail (as we have pointed out many times) but in the better cases will have captured enough of the essentials.
Now not only crimes have structures, but other things as well—for example the universe, or ciphers (not all ciphers, but some of them). This is why scientists follow clues, and cryptanalysts (code-breakers).
To solve a crime we do not need the criminal to confess. Similarly, to find out the many layers of structures present in the universe we do not have to ask God. And obviously, when we are trying to solve ciphers we cannot pick up the phone and ask our opponents how their ciphers work.
Why are clues so powerful? Why is it that by following them we can reconstruct structures hidden from us? We have answered this question in Chapter 2, using as example our by-now-familiar cryptogram:
SBR SBCTU DBCKERVS FCGG WTTCXR SFH FRRJD YTHE SHUWI
From our example we can see without much difficulty that clues are but the characteristics of structures, disguised. The English language has a structure. In English there are rules governing how letters are put together to form words and how words are put together to form sentences. As it is structured the English language has certain characteristics. One of these is that many of its words start with TH. The first two words in this cryptogram could start with TH. We do not see TH only because they have been disguised as SB.
Clues are not the characteristics of structures plain and simple, but the characteristics of structures disguised. Even so, if we can remove their disguises we can discover what is behind them. And this is why we should look for as many clues as we can in an investigation—that is, when we are trying to reconstruct hidden structures. For the more characteristics of a structure we know, the easier it is to reconstruct it, just as the more characteristics of a person we know the easier it is to locate that person and not confuse her with somebody else. If we know only the first name of a person we cannot find that person because it is likely that there are many people with that same first name. But if we know not just the first name, but also the last, and in addition who her parents are, when she was born, what she works at, and so on and so on …; if we know all these many things our task will become easier. Now the same with reconstructing hidden structures. The more characteristics of the hidden structure we know, the easier it is to reconstruct it and not confuse it with some other structure. And this is why, as we have said, we should look for as many clues as we can. Take solving a crime for example. If there are many clues and we can decipher all of them (that is, remove their disguises), we should be able to find out many of the details of the crime. And it is obvious that the more details we have, the easier it is to determine who is responsible.
In solving a crime we will be interested in knowing who is responsible. But we should not expect that we can find out who is responsible just by one clue. We cannot say the butler is responsible because the butler was holding a smoking gun. The butler could have been shooting at the person who shot the victim. If we want to find out who is responsible we will have to find out in some detail what took place. The more details we can uncover the easier it is to determine who is responsible.
In an investigation we should look for as many clues as we can. The more clues we have the easier it is to reconstruct the structures that have been hidden from us.
But while we should look for as many clues as we can in an investigation, in practice quite often we simply cannot find all that many, however hard we try. In crime investigation, very often the clues found are few in number. On these occasions we either do not know who is responsible, or we cannot be certain.
There is one kind of investigation in which on occasion we have managed to find a huge number of clues. These clues usually are not found in one day but over a long period of time. This has occurred in the investigation of nature (science) and is the reason why in the mature sciences we can have a high degree of certainty. In the younger sciences, due to the small number of clues, the results are often uncertain.
Our only clue lay in the truncated telegram, and with a copy of this in his hand Holmes set forth to find a second link for his chain.
—— The Missing Three-Quarter
In an investigation we should look for as many clues as we can. The more clues we have the easier it is to find the things we are looking for. Clues are the characteristics of structures. The more characteristics of a structure we know the easier it is to reconstruct it. But it is not always easy to find clues, as we all know. Suppose in an investigation we have found only a few (say, just one or two); does this mean that our investigation is doomed? That we will never be able to find what we are looking for?
Those familiar with investigations will know that it does not. Most investigations start out with very few clues. Sometimes we think ourselves lucky if we can find just one. But that there are very few clues at the beginning does not necessarily mean that the investigation will never bear fruit. If it did, there would have been very few successful investigations, if any. One remarkable characteristic of the theseological process is that often, or often enough, new clues can be developed from old clues, so that even though the original number of clues is small eventually there will be enough to bring the investigation to an acceptable conclusion.
How do we develop new clues from old? We answer this question in the next section.
In most investigations, if we are to succeed we will have to develop new clues from old, as we have explained in the last section. Now in simple investigations developing new clues from old is something we often do automatically. As illustration, look to the cryptanalytic example we have been using.
SBR SBCTU DBCKERVS FCGG WTTCXR SFH FRRJD YTHE SHUWI
Here we have a cryptogram. Clues that we can easily detect in this cryptogram tell us that S probably stands for T, and F, for W (we leave out for the moment the other clues). Now if we did adopt these two hypotheses, most of us would very naturally go on to translate all the Ss in the message to Ts, and all the Fs, to Ws, as we have done below.
T** T**** *******T W*** ****** TW* W**** **** T****
SBR SBCTU DBCKERVS FCGG WTTCXR SFH FRRJD YTHE SHUWI
But once we have done this we will have developed a new clue, in the sixth word. The sixth word is SFH. Since S stands for T, and F, for W, the sixth word now appears as a three-letter word that starts with TW. What could the third letter be? It is easy to answer this question. But this is to say, TW? is a clue, and it is a new clue since it was not there originally: it only appears after we have made the translations.
Notice how easy it is to develop this new clue. We merely translate all the Ss to Ts, and Fs to Ws, and voila! a new clue appears. Developing this new clue is so easy that in practice it is not likely that we would have taken time to remark upon its appearance; instead, we are likely simply to take advantage of it and carry on with the decipherment.
In simple cases developing new clues requires no conscious decision on our part and takes next to no effort. It is something we do, as we have been saying, automatically.
But this will not always be the case in all investigations. In more complicated investigations developing new clues will require conscious and deliberate effort.
Suppose we meet with one of these more complicated cases. Suppose we find ourselves in a situation in which we have decided that we need new clues; what do we do? How do we consciously go about developing new clues from old?
To answer this question we look back to what we have done in our cryptanalytic example. How did we develop the new clue that suggested to us that SFH was in fact TWO?
If we retrace our steps and pay a little more attention this time, we will find that this new clue appears as a result of applying hypotheses suggested by old clues to the evidence (which in this case is the cryptogram). Before we develop our new clue we have already proposed a number of hypotheses in answer to the clues already present. These include the two that say, respectively, that S stands for T and F, for W. When we applied these two hypotheses to the cryptogram, translating all the Ss to Ts and Fs to Ws, the new clue TW? appears.
So this is what we should do when we want to develop new clues from old. We apply the hypotheses suggested by old clues to the evidence. It is by applying hypotheses to evidence that we develop new clues.
We do not apply just any hypotheses to evidence. We apply the hypotheses suggested by old clues to the evidence. A hypothesis not suggested by any clues has little chance of being right. A hypothesis suggested by a clue has a much better chance. And of course, compared to clues yet to be developed, all the clues we already have are old.
However, we should notice that when we apply hypotheses to evidence we do not develop new clues every time, even when these hypotheses are suggested by old clues. When we apply hypotheses to evidence we develop new clues only sometimes. In our example we apply the hypotheses suggested by old clues to most of the words in the message, but in only one of them—the sixth word—do we develop a new clue. But since new clues are so important in an investigation, even though we do not develop new clues every time, we should keep applying hypotheses to evidence. The more often we do this the more likely we develop new clues.
“Data! data! data!” he cried impatiently. “I can’t make bricks without clay.”
—— The Copper Beeches
One way to look for clues is to search for repeating patterns in the evidence. If we have a large body of evidence it is easier to detect these patterns. If the amount of evidence is too small we may not notice any pattern at all. In our example if our cryptogram had been shortened to just three letters, SBR, we would not have known that SB is a pattern.
In an investigation it is to our advantage to have from the outset a large body of evidence. The beginning of an investigation is usually the hardest, but if we have a large body of evidence it should be easier to find clues.
Of course evidence does not just fall into our laps; we have to look for it.1 So this is what we should do: we should look for a large body of evidence, and we should do this early in the investigation. Now we may not succeed, but at least we should try.
When we have a large body of evidence early in an investigation it should be easier, we have said, to find those initial clues which will propel the investigation forward. In addition to this, if we have a large body of evidence right from the start, when it comes time to develop new clues our task should be easier too. For we develop new clues by applying the hypotheses suggested by old clues to the evidence. The larger the body of evidence to which we can apply hypotheses, the easier it is to develop new clues.
It is an advantage to have a large body of evidence early on in an investigation, but we should be aware it is possible sometimes to have more evidence than we need. In cracking the simple cipher in our example, if we are only interested in how the cipher works once we have enough ciphertext to work out how the twenty-six letters in the alphabet are translated, more will be superfluous. Now this is useful to know because in practice gathering evidence can at times be costly in terms of effort and resources. It is good to have a large body of evidence, but there is no point in having too much if having too much means wasting effort and resources.
How much evidence do we need? When do we have enough? How much is too much? These questions are not always easy to answer. For the amount of evidence we need depends on the structure or structures we want to uncover. The cipher in our example has a simple structure. For this we need only a small amount of evidence (ciphertext). If the cipher had been more complicated we would have needed a much longer ciphertext. But in most investigations we would not have known in advance how complicated those structures are that we are trying to know. For example detectives may be thinking that they are solving a simple crime and that they have enough evidence. It may turn out that the crime is much more complicated than they think, so that they will have to look for more evidence. Since this kind of error is common in investigations, when gathering evidence it is better to err on the side of caution, better, that is to say, to collect more than is needed instead of not enough. So we often say, in an investigation we should collect a large amount of evidence. In this context the word 'large' is being used in a relative sense. We use our best judgement to determine how much evidence is enough and then we look for a little more to give ourselves a comfortable margin.
We have to keep in mind the following as well. In most investigations finding evidence is difficult. Left to ourselves we have a tendency to stop too soon when looking for it. If we do not constantly remind ourselves that we should gather a large amount, we might just succumb to this tendency and end up with not enough.
In an investigation we need a large body of evidence in order that we can develop enough clues. It is preferable that we are in possession of this large body of evidence early in the investigation, so that we will have enough clues to set us on our way. Unfortunately this does not happen too often. In many investigations, despite our best effort we could still find that the evidence we have is not enough. When this occurs we of course will have to look for additional evidence. Now no shame is involved in doing this. In an investigation we are looking for things that we do not yet know. Without knowing these things we cannot make an accurate judgement as to how much evidence we need, and what kind. We can only do our best but we should not be surprised if our best is not good enough.
How do we look for additional evidence? For this we go to the next section.
Before we explain how we look for additional evidence we want first to put forward this reminder: In an investigation do make use of the information you already have to help you develop new information, even if the information you already have is not all that certain.
Let me first use an example to explain what I mean. Earlier we have used our cryptanalytic example to show that to develop new clues we should apply hypotheses suggested by old clues to the evidence. For example one of the hypotheses suggested by old clues is that S stands for T. Now this hypothesis is far from certain. Yet, despite its uncertainty, we apply it to the evidence. And this is proper. In this case, by applying this uncertain hypothesis to evidence we develop a new clue, TW?, which suggests that H stands for O. Now this is a new bit of information, itself also uncertain. However, while uncertain, it could lead to other new clues, which in turn could lead to other bits of new information. Now making use of uncertain information in this way—that is, in the process of following clues—is not only legitimate, but desirable, for if it goes on and on like this, after a while when we look back at the earlier bits of information we will find that these earlier bits are now much more certain than they were originally. In cracking our cipher the hypothesis that S stands for T occurs early. When it first occurs it is uncertain, but later on when the whole message has been deciphered this same hypothesis is much more certain than it was at the beginning. In the clue-following or theseological process, not only is there no harm in making use of uncertain information in the development of new information, it is incumbent upon us to so make use of this kind of uncertain information.
But I should caution that we can make use of uncertain information this way only in the clue-following or theseological process. In such a process uncertain information could lead to new clues, which in turn could lead to new information. When more and more information is obtained as a result of following clues the earlier information becomes more and more certain even though it could have been very uncertain to begin with.
What happens if we are not following clues? Suppose I simply put forward one uncertain hypothesis after another, doing so without following clues; will the first hypothesis become more certain after I have put forward one hundred such hypotheses?
Obviously it will not. Here we have only a series of hypotheses, none of which is dependent on the others for clues. Each one of them being independent of the others there is no reason why after one hundred of them the first one should become more certain. If we are not developing new hypotheses from the older ones by following clues the older hypotheses will not become more certain after the new hypotheses have been advanced.
Why should this be the case? Why is it so important that we should be following clues if we want the old hypotheses to become more certain?
There is a simple reason for this. If the hypothesis answering to a clue is too far from the truth it will not lead to new clues. Without new clues no new hypotheses or other kinds of information can be developed. The hypothesis answering to a clue could be uncertain to begin with, but if it leads to more and more new clues it becomes less and less likely that it is too far from the truth; which is to say, the certainty that it is at least close to the truth increases.
Why should this be? Why can a wrong hypothesis, a hypothesis that is too far from the truth, not lead to new clues?
The reason for this is that clues are the characteristics of structures. When we misread a clue; when we interpret it in the wrong way; we are putting into the structure we are trying to reconstruct an element that in fact is not in that structure. This foreign element cannot then combine with the other characteristics of the structure to produce new clues. When SFH is translated into TW? we have a new clue. But when SFH is translated into KW? we cannot have a new clue. S stands for T, not for K. When K is made into the first letter of this three-letter word by mistake it cannot combine with W to produce a new clue. It cannot do this because it should not have been there in the first place. An English word has a structure. Just as we cannot put together at random any number of letters to form a word we cannot put together at random any number of letters to produce a clue.
Since in the clue-following or theseological process we develop new information by following clues, if we are able to develop new information from old the old information is likely to be correct. If it were not, there would not have been those clues that led to the new information. It does not matter that the old information was originally uncertain. So long as it led to new clues it is likely to be true or close to the truth. The more new clues it leads to, the more certain that this is the case.
It is an important characteristic of the clue-following or theseological process that in this process we can rely on uncertain information to develop new information. Having illustrated this characteristic with an example, let us next see how this characteristic can help us find additional evidence.
In an investigation, when the evidence we have earlier found is not sufficient, we will have to look for additional evidence. But how do we look for additional evidence? Evidence for a crime, for example, is not just anything we can lay our hands on. We cannot go into a gun shop, pick up a gun and say that it is part of the evidence. For the crime we are investigating all of the guns in the universe, with the exception of perhaps one or two, are not part of the evidence. How are we to look for that one or two guns that are relevant to our investigation?
To answer this question we follow the advice we have been advancing. In the clue-following or theseological process we should try to make the best use of the information already in our possession when trying to develop new information, even though the information already in our possession may not be all that certain. Let us suppose we are investigating a murder. At the crime scene we are able to retrieve one bullet, the one which, we assume but are not certain, killed the victim. We examine this bullet for score marks (when a gun is fired, the barrel of the gun, not being perfectly smooth, will leave score marks on the bullet). From these score marks we now know which gun to look for. We will be looking for not just any gun, but the one that leaves the same score marks. Now just as the score marks tell us which gun to look for, other clues at the crime scene could tell us where we are likely to find this gun. Now all this information could be very uncertain, but if we succeed in actually finding the gun, and then the owner, and then the murderer (the owner may not be the murderer), together with other details of the crime because of additional clues that have been developed, the uncertain information we started with will become much less uncertain.
In an investigation we can make use of uncertain information to develop new information. That the information we rely upon is uncertain does not necessarily make it useless. All the information we make use of in an investigation will always have a certain amount of uncertainty associated with it. If we only allow ourselves to use information that is absolutely certain we will never find out anything. Indeed, if we hold to such an unreasonable standard we will be reduced to inaction.
It is important to point out however that although we would allow ourselves to use uncertain information in an investigation, this does not mean we can throw all caution to the wind. We would not use just any information. We like our information to be as certain as possible even though we do not require it to be absolutely certain. For example, when we advance hypotheses we would want these hypotheses to be based on clues; we do not want them to be the result of wild guesses.2 Hypotheses based on clues are uncertain, but they are less uncertain than wild guesses.
This is an important task in an investigation. In an investigation we are trying to reconstruct structures that are hidden from us. Since the structures we are trying to reconstruct are hidden we cannot determine whether we have succeeded by putting what we have constructed side by side with these hidden structures and comparing the two (sets). But while we cannot carry out this direct comparison we can pay attention to whether old clues lead to more and more new clues. If we do see old clues leading to new clues, and these new clues leading to more new clues, and so on and on, we can derive from such an observation a progressively more accurate idea as to whether we have succeeded in our reconstruction. For (as we have explained in the last section) if our reconstruction had gone seriously amiss; if for example we had gravely misinterpreted some of the clues; we could not have developed more and more new clues.
On rare occasions a clue that has been incorrectly interpreted will lead to a false clue by accident. But false clues peter out; they do not lead to more and more new clues. One accident is rare. Two accidents in a row is even rarer. Since this is the case when we manage to develop more and more new clues we will have good reason to believe that it is likely that the old clues have been correctly interpreted. In common speech, when this happens we say we are moving in the right direction.
In an investigation we should try to develop new clues from old. If we succeed, not only will we find out more, but we will have reason to believe that what we have earlier found out is likely to be correct. We follow clues to reconstruct structures that are hidden from us. But that these structures are hidden does not mean we cannot tell whether our reconstructions are correct, that is, whether they correspond to the hidden structures. Direct comparison is not the only way to find out whether two structures correspond. We can also find out by seeing whether they leave behind the same clues.
People not accustomed to following clues frequently have doubts that clues could ever lead us to the truth. We follow clues when the things we want to know are hidden. If these things were not hidden; if they were there for everyone to see; there would have been no need to follow clues. But if the things we want to know are hidden, whatever the clues tell us, how do we know what they tell us is true? If we can compare what the clues tell us to what is the case we can determine whether what they tell us is true. But when we are following clues we cannot make this comparison. We cannot because what is the case is hidden.
Things hidden, it seems, should be unknowable because whatever we say of these things we can never compare what we say to the things themselves. Yet in following clues this is exactly what we are trying to do. We are trying to acquire knowledge of the hidden, which kind of knowledge, it seems, should be unknowable.
Is it true that the hidden cannot be known? Of course the hidden cannot be known directly. If it could it would be a contradiction in terms, for then the hidden would be both hidden and not hidden at the same time. Since the hidden cannot be known directly, if it is to be known at all it will have to be known indirectly. Can the hidden be known indirectly?
In these pages we are saying it can. When we are following clues we are trying to know things that are hidden; we are trying to know indirectly. And it is possible to know indirectly if there is a method. When we have a method we do not need to compare. We have explained this in the last chapter but since people often have difficulty over this point, let me do so again in a slightly different way.
Suppose I have two huge piles of marbles. I have counted each pile carefully, so that I now know how many marbles there are in each pile. Now I put the two piles together to form a even larger pile. How many marbles are there in this larger pile?
The direct way is to count all the marbles that are now in the larger pile. It is not likely, however, that anyone will do it this way because it will take too long. There is a shorter way, much faster, as we all know. Since we know the number of marbles in each pile before we put them together, we can simply add up these two numbers following the method (algorithm) we commonly employ. This method is indirect: it will give us an answer without counting. But although indirect, the number it gives us will be accurate provided we follow the method correctly. It is important that we follow the method correctly. If we do not, we of course will not get the right answer (except by chance). But if we follow the method correctly we will always arrive at the right answer. We know the answer will be right without checking it by counting the marbles.
Now the hidden cannot be known directly. But this does not mean it cannot be known at all. The hidden can be known indirectly if we have a method. And we have a method. We have been using this method whether we are aware of it or not. I call this method the theseological method, since we use it during the theseological process. When we are following clues we are following the theseological method. What is the theseological method? The theseological method tells us that if we want to uncover knowledge of things hidden we have to follow clues and develop new clues from old. Now this method is not easy to follow because it is not easy to follow clues and develop new clues from old, but if we do manage to follow it we will arrive at right results. We know these results will be right without 'counting the marbles', that is, without having direct access to the things we want to know. Why should the theseological method be able to do this? We have answered this question already. The method tells us to follow clues and develop new clues from old. When we succeed in following this method; that is, when we succeed in following clues and developing more and more new clues; the earlier clues are likely to have been correctly interpreted. The more generations of new clues they lead to, the more likely this is the case. A clue that has been seriously misinterpreted cannot lead to new clues.
When we follow the theseological method our attention is on clues. The more clues we are able to develop from old clues the more certain we can be of our earlier results (that is, our interpretation of the old clues). To find out whether our earlier results are right we do not have to compare these results to the things that are hidden. That we do not have to make such a comparison may appear strange at first, but should not any more now that we know that when there is a method there is no need to compare.
People not familiar with the theseological process frequently have doubts whether clues can lead to the truth. But they are not the only ones. Even experienced theseologisers will sometimes have the same doubts. The simple fact is, that we can get to know about things hidden is counterintuitive. Things hidden, it seems, should be unknowable. For this reason it is useful to keep in mind that although things hidden cannot be known directly, they can be known indirectly when there is a method. And there is a method: the theseological method, which tells us to follow clues and develop new clues from old.
Looking for patterns in the evidence is often the first thing we try when looking for clues. Clues are the characteristics of structures. When we have found a pattern in the evidence it is likely that we have found one of the characteristics. Thus it is that in practice when we have found such a pattern we will go on to ask what this pattern means. For clues are not just the characteristics of structures, but the characteristics of structures, disguised. In our example the pattern SB is a pattern; it suggests to us that it might stand for TH.
But we should be aware, patterns sometimes occur by accident. When we have found a pattern we should not be dogmatic that it must be significant. As we so often say, the reason for the pattern could be 'totally innocent'. However, if the pattern repeats itself, not just once but many times, the chance that it occurs by accident diminishes.3
Not all patterns are easy to find. Sometimes discovering a pattern requires a good bit of ingenuity. Take for example the motion of the planets. For a long time people in different parts of the world have noticed that the motion of the planets across the night sky is erratic: there is no noticeable pattern in these motions. But Kepler discovers that this is so only if these motions are viewed from the earth. If instead of using the earth as the reference point we use the sun, we will discover—as Kepler did after a large amount of calculations—that the planets all move in elliptical orbits. Kepler's discovery—that the planets all move in elliptical orbits around the sun—is one of the clues Newton uses in his discovery of the law of universal gravitation and his three laws of motion.
A pattern could be a clue because there could be a reason for the pattern. If a pattern repeats itself, it is likely that it is not accidental. Now in detecting patterns as clues, we notice the pattern first and then have to figure out the reason for it. That is, we have to search among the structures we know for one that will produce the same pattern. Now in order that we can carry out this task it matters whether the pattern repeats itself indefinitely. Not all repeating patterns repeat themselves indefinitely; some will repeat themselves for a time and then, for a good reason, stop. When we are looking for a reason for the pattern, therefore, we cannot take it for granted that the pattern will repeat itself indefinitely. If we did, we could come up with the wrong explanation, or we may not be able to come up with any explanation at all.
If we notice a pattern repeating itself for a time and then stop, we know then that it is not repeating itself indefinitely. But suppose it has been repeating itself for a long time and has not stopped so far. Can we, just on this basis, take it for granted that it will never stop?
Obviously, we cannot, since we don't know yet the reason for the repetition. Sometimes a pattern could repeat itself for a long time before stopping. The earth has been circling around the sun for many years but this is not to say it will do so for ever and ever. A time will come when the earth will be no more because it has spiralled to its death by falling into the sun.
In an investigation it is important that we do not take for granted that a pattern that has been repeating itself will repeat itself indefinitely. If we did we could seriously hamper our investigation.
What can we do when we see a repeating pattern and we do not know whether it will stop?
This is what we could do. We could assume it will repeat itself indefinitely, but if this leads nowhere, we will have to try the other possibility, that is, that it will not repeat itself indefinitely.
Looking for unique occurrences is another popular way of looking for clues. For, if the occurrence is unique there must be a reason why it is unique. If we can figure out this reason we will have found out something important about those structures we are trying to uncover. In one of the cases Sherlock Holmes took on, a person staying at a hotel had one of his boots stolen even though the other was equally available. Now this is odd. Most of the time boots are worn in pairs. Why would anyone steal just one boot? Concerning this peculiar incidence Sherlock Holmes has this to say:
The more outré and grotesque an incident is the more carefully it deserves to be examined, and the very point which appears to complicate a case is, when duly considered and scientifically handled, the one which is most likely to elucidate it.
—Sherlock Holmes in The Hound of the Baskervilles
A unique occurrence could be a clue. It could even be a highly significant clue. But how do we determine whether an occurrence is unique? In the case of stealing boots, stealing one is unique, but this is only because we all know that boots are worn in pairs. Suppose we do not know that boots are worn in pairs; how do we determine uniqueness then?
Clearly, if we do not know, we would not be able to know that stealing one boot is unique. To determine uniqueness we have to know what the norm is. And this has important implications for investigators. Let me give a simple example.
A bank has been broken into overnight and a large sum of money has been lost. As part of the investigation all the employees are interviewed, including of course the manager. When asked what time she left work the previous day the manager answers, '5:15.' The detective in charge of this case takes down this answer and after comparing it with those given by the other employees thinks no more of it. Why? Because she finds nothing unusual about this answer. There is nothing unusual about the manager being the last person to leave. By 5:15 all the other employees had gone home.
In fact 5:15 is highly significant, and the detective would have known if she had asked a few more questions. For this particular manager has been working at this bank for a long time and without exception has always left punctually at 5:12. Without exception, that is to say, until the previous day!
Unique occurrences can be important clues, but uniqueness is not always apparent. Sometimes we fail to notice uniqueness for lack of sufficient information. This bespeaks again for having a large body of evidence. This is not to say that when we have a large body of evidence we are bound to notice uniqueness: the relevant information could be absent from this large body; but in an investigation we can only increase our chances (see next section).
In the theseological process we advance hypotheses in response to clues. We find the clue SB. We ask, what does it mean? We answer by putting forward the two hypotheses: S stands for T, and B, for H. Now when we put forward hypotheses in answer to clues it is possible that the hypotheses we put forward are wrong. A clue could mean more than one thing. S might not stand for T, but for something else. How do we know that the hypothesis that S stands for T is not wrong? One way to find out is to test the hypothesis. Suppose S occurs also in another word, ten letters in length. Suppose that all the other letters in this ten-letter word have been deciphered, so that S is the only remaining one. Now if this is the case and we substitute T for S, and find that T fits in, we can conclude that our hypothesis is likely to be right, that is, that S does stand for T. But if T does not fit in; if after translating S to T we do not produce a proper English word (we assume here that we are deciphering an English message); we will have reason to suspect that our hypothesis is wrong. Now the reasoning here seems to be simple enough; nevertheless, we would do well to make it clearer to ourselves.
Let us carry out the analysis first in the following way and call it Attempt A.
Let us first consider the case that T does not fit in.
If T does not fit in; if after we have translated S to T we do not obtain a proper English word; it seems clear that the hypothesis cannot be right. The hypothesis, if true, leads to the conclusion that we should obtain an English word. The conclusion turns out to be false: we do not obtain an English word. Therefore the hypothesis must be false.
Now consider the other case: T fits in.
If T fits in; if after we have translated S to T we do obtain a proper English word; can we then conclude that the hypothesis is true? Here we meet with a major difficulty. The hypothesis leads to the conclusion that we should obtain a proper English word. This conclusion turns out to be true: we do obtain a proper English word. But we cannot conclude from this that the hypothesis must be true. It is well know in logic that false premises can sometimes lead to true conclusions. Take a simple case:
Premise 1: I am a millionaire.
Premise 2: People who have a million dollars will of course also have one dollar.
Conclusion: I have one dollar.
Now Premise 1 is false. But the conclusion is true! I do have one dollar. From the fact that I have one dollar I cannot conclude that I am a millionaire. Similarly from the fact that T fits in we cannot conclude that the hypothesis that S stands for T is true. But when we test a hypothesis, this is what we do. When we test a hypothesis we derive from the hypothesis a prediction. If the prediction turns out to be right we conclude that the hypothesis is likely to be right.
Something is wrong with Attempt A. Attempt A does not help us understand actual practice. Actual practice tells us that if prediction turns out to be right the hypothesis from which it is derived is likely to be right. Attempt A does not confirm this. To the contrary, it tells us we are not any the wiser as to whether the hypothesis is right when the prediction is right.
What could be wrong with Attempt A?
Of course, actual practice could be wrong. Perhaps we should abandon actual practice. Perhaps we should say, even though the result of a test is as predicted, we are not a whit wiser as to the truth or falsity of the hypothesis. The hypothesis could be true or it could be false. The test has not told us whether it is one or the other.4
However, before we reject actual practice let us make a second attempt to understand the reasoning behind it. For Attempt A is a little strange. We make Attempt A to understand something we frequently do when following clues. In following clues it is common that we test hypotheses when possible. But notice what we have just said. We have just said that we test hypotheses when possible when following clues. But Attempt A says nothing about clues!5
We now make a second attempt to understand the reasoning behind hypotheses testing when following clues. Let us call this Attempt B.
This time let us first consider the case when T fits in; that is, when the substitution of S by T produces a proper English word as is predicted.
Now we should remember that the hypothesis that S stands for T is suggested by a clue (SB). We test this hypothesis by applying it to a ten-letter word in which nine letters are already known (with S being the tenth letter). But what is this ten-letter word if not another clue? When we have a ten-letter word in which S is the only undeciphered letter, the other nine letters should give us a good idea as to what S stands for. So whereas previously we had only one clue (SB), we now have two (SB and the ten-letter word). They both point to S standing for T. This being the case, after the test we obviously should be more certain about the hypothesis we have advanced (that S stands for T). If we do not have a single clue, S for all we know could stand for anything. When we have one clue, the number of possibilities for S is reduced. When we have a second clue, especially a specific one (like the ten-letter word in our example), the number of possibilities is even smaller. For good reasons, therefore, when we test a hypothesis during the theseological process and obtain a positive result, we will look upon the hypothesis much more favourably after the test than before.
Let us now consider the case when T does not fit in.
In this case the test produces a negative result: after translating S to T in the ten-letter word we do not obtain a proper English word. From this we infer that the hypothesis (that S stands for T) is wrong. And this is proper. Even when we had only one clue SB, we knew that S might not stand for T. Now we have a second clue and it tells us that S does not stand for T. It is always better to have two clues than just one. The more clues we have the clearer the message they send us.
We have offered two analyses—Attempt A and Attempt B—for the reasoning behind the practice of testing hypotheses in the theseological process. Attempt A can now be seen to be misdirected: it forgets that we are examining this practice as it occurs in the theseological process. Attempt A therefore should be rejected. Attempt B is better than Attempt A. Attempt B clearly acknowledges that we are examining this practice in the context of following clues.
In our example to test the hypothesis that S stands for T we use a ten-letter word in which nine is already known. The remaining letter is the only uncertain letter, being the one to which the hypothesis is to be applied. Now this is something we always try to do when testing a hypothesis. We always try to find a situation in which the only uncertainty is the hypothesis under test. When this is the case the result will tell us how good our hypothesis is. If there are other uncertainties besides the hypothesis under test, it may be difficult or even impossible to interpret the result. A test for a hypothesis is a second clue to the same hypothesis. But if this second clue is too vague it will not help us pin down the hypothesis.
How important are tests? Are tests the only way by which we can find out how good a hypothesis is?
People often say yes to the last question. It is a common view that tests are the only means of finding out how good a hypothesis is. But this common view is incorrect. Tests are not the only way to find out how good a hypothesis is. There is at least one other way. In fact we have already explain this other way. We have already said that a clue severely misinterpreted will not lead to new clues. Now hypotheses are proposed in answer to clues. If a hypothesis is seriously defective it will not lead to new clues. If a hypothesis leads to new clues which lead to even more new clues, such a hypothesis is likely to be true or close to being true. The more (generations of) new clues it leads to the more likely that this is the case. Tests are not the only means for finding out whether a hypothesis is right (true or close to being true). Instead of testing a hypothesis we could wait and see whether it leads to more and more new clues.
In fact, this method of evaluating hypotheses, which tells us to pay attention to the development of new clues from old, is fundamental to the theseological process, and as such more important than tests. For tests depends on our knowing a great deal already, as we have seen. If we know already nine letters in a ten-letter world we can use this ten-letter word to test a hypothesis for the remaining letter. But how do we get to the stage where we know nine letters? When we start out we may not know a single letter at all. When little is known, how do we know that the hypotheses we advance in response to clues are correct? We can only know by seeing whether they lead to more and more new clues. If they do they are likely to be right even if we have not had a chance to test them. Tests are not the only way by which to find out how good a hypothesis is.
To save time or resources, or because they are just impatient, experienced investigators would sometimes omit testing a hypothesis even when such a test is available. Instead of testing the hypothesis they would just 'plough on'. Now if the hypothesis leads to more and more new clues, they know it is likely to be right.
But these investigators will also warn you that avoiding tests is a risky business. In an investigation it is a good idea to test a hypothesis first, before moving on. For most investigations are complex. In a complex investigation any mistake not caught in time can be difficult to unearth later on. When we avoid testing a hypothesis, no harm is done if our hypothesis happens to be right. But if it is wrong; if it happens to be a mistake; we could be creating trouble for ourselves.
In science a common way to test a hypothesis is to carry out experiments. By this we mean experiments in the laboratory or in the field. But sometimes such experiments are not actually possible for technological, financial, or other reasons. What can scientists do in such instances if they still want to find out how good the hypothesis is before they move on?
This is what they sometimes do. They carry out the experiments 'in their head', relying on intuition or imagination to tell them what the results are likely to be. They call these thought experiments. Needless to say, thought experiments are not as good as real experiments: intuition and imagination are not always reliable. Still, they are better than no experiment at all. And if the hypotheses they are supposed to test should later on lead to new clues, that they are thought experiments only, and not real, makes no difference.
When things we do are important, to ensure that we have done them, or have done them properly, we double-check. In locking up for the night a bank manager has to make sure the alarms are on, so she checks and double-checks. Now in investigations we can correct mistakes, but correcting mistakes long after they have been made is often difficult. It is better that we make sure, or as sure as we can, that what we have done is right before we move on. So we check and double-check when we can. Testing hypotheses, we have seen, is a kind of double-check. A hypothesis has been suggested by a clue (SB). To make surer that the hypothesis suggested is right we look for a second clue (the ten-letter word), which as we have seen is the test.
In an investigation we should double-check when we can. The theseological process is not simple; success is not guaranteed; but the fewer mistakes we make the better our chances.
In an investigation we always have to make assumptions. Without them the investigation cannot start. Look back to our cryptanalytic example. If we did not assume that the series of letters is hiding a message we would not even start looking for clues. If we are already looking for clues it must mean that we have made this assumption. Now when we are looking for clues we need to have some idea what we should focus our attention on. Should we for example pay attention to the size of the letters? Or perhaps we should forget about the letters altogether and focus only on the spaces between them. Now we might not actually have asked these questions, but once we have started looking for clues it will mean that we have assumed some sort of answers to them or questions like them.
In an investigation we have to make assumptions of a general sort if we are to begin finding out the details of the things we want to know. And if we have begun; if we are already finding out details; we must have made these assumptions, whether we are conscious of making them or not.
Forming assumptions is an important part of an investigation: right assumptions will facilitate an investigation; wrong assumptions will hinder it. Now a question arises, do assumptions have to be exactly right if we are to make progress? Is there any latitude where the correctness of assumptions is concerned?
It is important to notice that for assumptions to be useful in an investigation they do not have to be exactly right. They need only be sufficiently close to being exactly right.6 To see this, suppose we are solving a cryptogram different from the one in our example. Suppose in this case we again assume the cleartext to be in English. Now by English we mean of course ordinary English. But suppose the cleartext in this case is not in ordinary English, but Newfoundland English. In a strict sense then, our assumption is false. Will this stop us from cracking the cipher? Clearly not. Newfoundland English is reasonably close to ordinary English. Because of this we should be able to obtain large portions of the cleartext even when based on this false assumption (that the cleartext is in ordinary English). We should be able to do this if the message is long enough. And once this is done; once large parts of the cleartext are known; from context we should be able to decipher the rest. Indeed, long before we have completed the decipherment, we will have found out that we are not dealing with ordinary English at all, but one of its variants. Adjustment to our assumption will have been made at that point.
In an investigation we can start with false assumptions, obtained right results based on them, then turn back and correct the false assumptions. Of course, for us to be able to do this the assumptions, though false, have to be sufficiently close to the truth; they have to be good approximations.
We have been calling assumptions right or wrong. By 'right' I mean true or sufficiently close to the truth; by 'wrong' I mean far from being true. In an investigation assumptions have to be right but they do not have to be true.
In deductive reasoning we should never start with false premises if we are interested in true conclusions. In an investigation we do not mind starting with false assumptions so long as they are sufficiently close to the truth. Because assumptions do not have to be true and because most investigations are difficult, investigators, to make life easier for themselves, often adopt assumptions which are simple but which they know are false. No harm need come from this. So long as the assumptions are sufficiently close to the truth, once the investigation has gained momentum adjustments can be made as required.
In an investigation we are interested in the truth. But in order to arrive at it, we sometimes adopt assumptions that we know are false. Lay people are often horrified by this. ‘How can truth come from falsehoods?’ they say. But there is no reason to be alarmed. Theseology is not a form of reasoning. Assumptions in an investigation are not like premises in reasoning. In reasoning if we want to be sure that the conclusions are true we have to start with premises that are true. But in an investigation false assumptions can lead towards the truth. Different arts give rise to different possibilities. We cannot limit one art by the possibilities of another. Fixed wing aircraft cannot hover in mid-air, but this does not mean helicopters cannot either.
Because there is a limit to the amount of complexity humans can deal with, many investigations are of necessity processes of gradual approximation. We start out with assumptions that are close enough to the truth. Then we bring them progressively closer.7
There is a place for exactitude and precision in investigations but at times a slight departure from exactitude and precision can be, and has to be, tolerated in order that we can forge ahead. We uncover the hidden by following the theseological method. It is a characteristic of the theseological method that quality improves with quantity: the exactitude and precision of results increase as the quantity of results rises. A premature insistence on exactitude and precision could lead to a paucity of results——the reason why perfectionists never get anywhere and why critics are seldom, themselves, good performers.
It is well known that in an investigation we could have made assumptions without ourselves knowing. We call this kind of assumptions hidden assumptions. Hidden assumptions are common and it is important that we remind ourselves from time to time that we might have made them. An investigation is seldom, if ever, a neat and tidy affair. Because so much is going on; because of the confusion; because of our impatience for results; we often forget about things that we should not forget, and as a result pay a high price. Investigations sometimes bog down because some of the hidden assumptions are wrong. Before these investigations can advance or resume their advance the hidden assumptions responsible for the impasse will have to be unearthed and corrected. This will not be done if we do not remind ourselves from time to time that there could be hidden assumptions in our investigations.
That wrong assumptions, hidden or otherwise, can cripple an investigation is also well known. Because it is well known, when we find it difficult or impossible to make any headway in an investigation, after a time most of us will ask, have we made any wrong assumptions? We do this even when playing armchair-detectives. We assumed death was from foul play; this led nowhere. Perhaps it was not foul play but a carefully arranged suicide. Let’s see what happens on this new assumption.
Recently, Mayan writings which for a long time nobody was able to read, have been deciphered. This came about because of a change in assumption. For a long time archaeologists and linguists thought these Mayan glyphs were ideograms, that is, that each one of these glyphs conveys an idea. This assumption led nowhere. Recently it occurred to some that these glyphs might not be idiographic at all, but partly pictographic and partly phonetic. After this change in assumption was made progress was swift. Now they are able to read large portions of these inscriptions on Mayan monuments.8
Interestingly enough, exactly the same change in assumption had to be made in the decipherment of Egyptian hieroglyphics. There, too, the language was at first thought to be idiographic; it turned out to be partly pictographic and partly phonetic, just as in the later Mayan case. [More from earlier versions?]
In forming assumptions sometimes we can form more than one set. In solving the cipher in our example instead of assuming the plaintext to be in English we could assume it to be in Chinese. Of course, if we do we will have to assume that the cipher works in a different and more complicated way. Now these two sets of assumptions cannot both be correct. How do we know which is correct?
The only way to find out is to see which set leads to results. If by following the English set of assumptions we arrive at an English plaintext, then the English set is right. If the plaintext is in English it is not possible for a Chinese plaintext to turn up, so long as we are following clues. Clues are the characteristics of structures. If the plaintext is not in Chinese it will not have Chinese characteristics.
Does this mean then that instead of trying the English set of assumptions first it is just as reasonable to try the Chinese set? We do not know which set is correct before we have obtained results. Before we have obtained results there is no reason why we should think that one set is more likely to be correct than the other. Is this not so?
But in practice, with an example like the one we have been using, most people will try the English set of assumptions first. Why?
The reason cannot be that they know from the outset that the English set of assumptions is correct. They will know at the end whether it is, not at the beginning. They will try out this set of assumptions first because it is simpler. That is, it appears to them that it is easier to render this set of assumptions compatible with available evidence: the cryptogram. Of course, appearance could be deceiving. It is possible they will find out, once they have started looking for clues and interpreting them, that it is not easy at all trying to decipher this message on this set of assumptions. But if this should happen they can then change assumptions. When we have a number of alternatives to try out it is better to try out the simpler ones first. If they are wrong we will know soon enough, after which we can try the others.
In trying out assumptions in an investigation we should try out the simplest set first. This does not mean that the simplest set is necessarily right, just that it is easier to try out.
Breadth of view, my dear Mr. Mac, is one of the essentials of our profession. The interplay of ideas and the oblique uses of knowledge are often of extraordinary interest. You will excuse these remarks from one who, though a mere connoisseur of crime, is still rather older and perhaps more experienced than yourself.
—— Valley of Fear
To carry the art … to its highest pitch, it is necessary that the reasoner should be able to utilise all the facts which have come to his knowledge; and this in itself implies, as you will readily see, a possession of all knowledge, which, even in these days of free education and encyclopaedias, is a somewhat rare accomplishment. It is not so impossible, however, that a man should possess all knowledge which is likely to be useful to him in his work, and this I have endeavoured in my case to do.
—— The Five Orange Pips
Much of what we do in an investigation is based on intuition. This is especially true where the formation of assumptions is concerned. Now if our background knowledge is vast the chance is greater that intuition will be able to pick out from this body of knowledge those assumptions that would help us in the investigation. Use as example our SBR-cryptogram again. Those who know English will have an easier time forming assumptions after examining this cryptogram. Those who do not may never arrive at a useable set at all.
We need background knowledge when carrying out investigations. But we do not always know in advance what background knowledge we will need (a person may go through life without ever having to solve an English cryptogram). For this reason it is a good idea to have as large a body of background knowledge as we can. The larger the body the better the chance that the knowledge we need for a particular investigation will be available from this body.
How do we build up background knowledge? There is so much to know. Should we for example make a point of learning all the languages there are?
It is impossible for an individual to know everything. Even with languages the number one individual can become familiar with can only be a small fraction of all the languages there are.
What can we do then? Should we, for example, know a bit of everything? A few languages, a little chemistry, a smacking of psychology, a smidgen of physics, and so on?
To most people the prospect of learning a bit about everything cannot be very attractive. Moreover, it may not be helpful either: the bits we need for an investigation could well be among those we have left out.
What other courses of action can we take? How can we increase our background knowledge so that we will be better investigators?
One suggestion that has been made is that besides learning the basics investigators should be encouraged to cultivate each their own interests where the acquisition of background knowledge is concerned. In every kind of investigation there are usually certain basic things we need to know, things that will help us with the run-of-the-mill investigations in that particular area. Now if we do not want to be run-of-the-mill investigators; if we want to do better; we will have to expand our background knowledge beyond these basics. How do we do it? We leave it to the individual; we let the individual follow his or her own interests. Things we are interested in we never know enough. Learning about them, finding out about them, doing research in connection with them—these we will gladly carry out on our own, and derive great satisfaction in doing so.
Of course even then we cannot guarantee that we will have in our possession the background knowledge needed for all the investigations that we will be faced with. There can be no guarantee of this kind. In preparing ourselves for future investigations we can only increase our chances.
[H]er little problem … is rather a trite one. You will find parallel cases, if you consult my index, in Andover in '77, and there was something of the sort at The Hague last year. Old as is the idea, however, there were one or two details which were new to me.
—— A Case of Identity
As a rule, when I have heard some slight indication of the course of events, I
am able to guide myself by the thousands of other similar cases which occur to
—— The Red Headed League
Sherlock Holmes studies the annals of crime. Scientists take interest in investigations other scientists carry out. Cryptanalysts want to know how earlier ciphers were broken …. Good investigators do not just carry out their own investigations; they pay attention also to what other investigators have done. This is one of their ways of building up background knowledge, background knowledge that will help them in their future investigations.9
For this is what investigators often do: they look to past cases for inspiration—that is, for suggestions as to what assumptions to make, what clues to look for, and how to interpret them. For this purpose they compare the case they are working on with what has been done in the past. They do not expect their current case to be exactly like some past case. It is not exact correspondence they are looking for, just similarity. Provided there are sufficient similarities they can bridge the differences by themselves.10
Knowing about old cases is useful to an investigator. But if we want to be good investigators we will have to do more. We will have to take an interest in the creation of new structures and in studying their characteristics. To make clear what I mean let us first use catching criminals as example.
Some criminals are creative; they can invent new ways of committing crimes. If you are a detective and you only study old cases you are not likely to catch these creative criminals. If you want to be a good detective what will you do? Obviously, the thing to do is to anticipate the ways in which these creative criminals will be creative. This is to say, if you want to be a good detective you will have to think like a criminal—and not just any criminal but a creative one. You will have to think of new ways committing crimes, and study these new ways to discover what clues they will leave behind. This is the reasoning behind the well-known saying, set a thief to catch a thief. Ordinary thieves can be caught the ordinary way; their modus operandi are well known. To catch an exceptional thief you have to think the way she does. If you manage to do so, she is caught. If you fail, she gets away. How to increase our chances of catching her? We imagine her planning not just one crime, but many, and hope that one of these imaginings corresponds to what she has actually done.
To solve crimes detectives practise by taking on the more common types first, the types that have been solved in the past. Scientists want to find out the structure of this universe, but they have no other universes to practise on. They have to start 'from scratch'; which is to say, they have to construct (theoretical) structures on their own and see if at least one of them leaves behind the same kind of clues that the actual universe leaves behind. For example, in what kind of universe will the planets move in elliptically orbits around the sun? Will they do so in a universe in which the force of gravitation obeys an inverse cube law? Newton found out they will not, but they will if the force of gravitation follows an inverse square law. Did Newton find this out by studying old universes? Does any one before Newton know that planets will move in elliptical orbits if the force of gravitation obeys an inverse square law? No; there is no such past cases for Newton to rely upon. He finds out through his own initiative, by constructing theoretical models (more than one) and studying how they behave. From his models he can tell that in a universe in which the gravitational force obeys an inverse cube law the planets will not move in elliptical orbits, but in one in which it obeys an inverse square law, they will.
To figure out the meanings of clues we have to know what structures will leave behind these clues. Where does this knowledge come from? In some instances it comes from past experience; in others it derives from structures we ourselves have created and studied. Newton had to create on his own initiative new structures just as good detectives have to think of new ways of committing crimes.
Clues are the characteristics of structures. It is for this reason that by following them we can reconstruct those structures we want to know. We can compare this to looking for a person, we have said (see again Do look for as many clues as you can). The more details/clues we have about this person the easier it is to find her and not confuse her with somebody else. But this is to say, each clue we have about this person contributes only a little towards finding her. However, if we have enough of these clues we will be able to find this person even though we have never met her before.
We should never despise the amount of information a clue provides. Although this amount is always small when compared to the total amount we need, if we have a large number of clues these small amounts will add up.
But not only should we not despise the small amount of information each clue provides, we should accustom ourselves to expecting only a small amount from each one of them. The amount of information any one of the clues provides is always small compared to what we eventually will know (if we are lucky). This is the way clues are, the reasoning we need a large number of them. It is true that a clue can sometimes lead to a succession of other clues, and do so quickly; still, even such a clue will only tell us a small fraction of what we eventually will find out.
In an investigation we find things out by following clues, gradually, step by step. Since each clue can only tell us a little, the steps we take in an investigation are never huge. In an investigation we can only take small steps, never big ones. I call this the Small Steps Principle.
Most people familiar with following clues are aware of the Small Steps Principle even though they may not call it such. When they hear a person spinning a long tale as a result of just one single clue they will ask, 'How can you know so much from just one clue?'
We do not expect a single clue to tell us everything we want to know. We do not expect it to tell us even a large portion. We need many clues in an investigation. Which is to say, we need many steps, each small.
In an investigation we have to observe the Small Steps Principle. If we do not; if we take huge steps; we will not find the things we are looking for.
But it is not easy to follow the Small Steps Principle. Following clues is hard work and takes time. From time to time even experienced investigators will become impatient, and after a certain point allow their imagination to completely take over.11 They have found a few clues, they have been patient so far, but suddenly one or two of these clues catches their fancy. From them they 'derive' huge amounts of information, enough they claim to enable them to declare victory. But this victory is a chimera. They have contravened the Small Steps Principle. The information they derive is not justified. And we can tell it is not simply by looking at the quantity.
They say that genius is an infinite capacity for taking pains …. It's a very bad definition, but it does apply to detective work.
—— A Study in Scarlet
Investigations are difficult. It is difficult to find clues, difficult to interpret them, and on top of this there is the Small Steps Principle: there are so many steps to take and they all, of necessity, have to be small. In an investigation if we are not patient we will never find the things we are looking for.12
Most of us are brought up to think that in the search for knowledge we have to be exact and precise. We must be careful with every step we take, and not allow ourselves to make a single mistake. Things have to be just so, not a whit more, not a whit less. Unless this is the case (we are taught) truth will escape our grasp.
But this is exactly not what an investigation is. Real investigations are messy and untidy. In investigations mistakes are common and exactitude and precision are seldom as high as we would like them to be. Yet, despite the chaos, despite the inexactitude and imprecision, truth gradually emerges.
Why is this the case? How can order emerge from chaos? How can truth come from untruth?
We will have more to say in answer to this question later on. For the moment it is useful to notice that in an investigation we are permitted to employ approximations. Indeed, the clever use of approximations can sometimes make possible discoveries that otherwise may not be possible.
Approximations, we have to remember, are not the truth. They are departures from the truth (if only by a small amount). Yet sometimes they can lead to the truth. Why?
To answer this question it is useful to bring to mind the Small Steps Principle: in investigations we have to take small steps, never huge ones. We have the Small Steps Principle because in an investigation we have to follow clues. Each clue tells us only a little. When we have figured out a clue we have only taken one small step. Now an approximation is a small step away from the truth. When we inject an approximation into an investigation; when for example we answer a clue not by the truth but by an approximation; we have divided up a small step into even smaller steps: we have taken a small step forward within the small step, leaving other small steps within this small step to be taken in the future. But there is no harm in this. In an investigation there is prohibition against huge steps, but no prohibition against dividing a step that is already small into many smaller steps. Indeed, if by doing so we can make it easier to answer some of the other clues, we are duty bound to do so. In an investigation the more we have found out, the easier the rest of the investigation tends to become. Put another way, the more small steps we take the easier the rest of the small steps. In an investigation the important thing is that we should be able to 'push forward'. It does not matter if we do so by taking fifty small steps or by expanding this fifty into one hundred or even a thousand smaller steps. So long as we are able to push forward, the rest of the investigation should become easier and easier (all things being equal).
It is permissible to make use of approximations in investigations. In practice investigations are frequently processes of gradual approximation. If we want to be good investigators, instead of succumbing to a phobia against approximations on the ground that they depart from the truth, we should cultivate a habit of using them to our advantage. One of the clues that enabled Newton to discover the theory of universal gravitation was based on an approximation. He and other scientists at the time were asking, why do the planets move around the sun in elliptical orbits? Now these orbits are not exactly elliptical; they cannot be because there is not only gravitational attraction between each planet and the sun but also gravitational attraction between the planets themselves. The orbits are only approximately elliptical. But this approximation has not hindered Newton's investigation. On the contrary, it has made Newton's investigation easier. Imagine what would have happened if instead of asking why the orbits are elliptical, Newton had spent his time trying to determine the exact orbits of the planets.13
In an investigation, to reconstruct those structures we are interested in we have to look for clues and find out what they mean. Now clues are the characteristics of structures. In trying to reconstruct a structure, if the structure we are trying to reconstruct resemble some structure we already know, it should be easier to find clues and interpret them. For since the two structures resemble each other they should have similar characteristics; knowing the characteristics of one should therefore help us find the characteristics of the other. Put another way, since we know what clues the structure we know will leave behind it should be easier to find the clues for the structure we want to know. Here we see the reason why in investigations we often look for analogies. Does this crime I am investigating remind me of some crime I know? Is this part of nature similar to some other part?
It is useful to notice that when we rely on analogies during an investigation the analogies do not have to be exact. In deciphering dead languages a common trick is to compare the dead language to some known language. Now no two languages are exactly alike. But this does not matter. If the two languages we are comparing are sufficiently close we should be able to decipher large parts of the dead language based on the analogy. Once we have done this; once we know enough of the dead language; we can go native, that is, we depend on context to help us decipher the rest (just as we depend on context to learn new words even with our mothertongue).
It is important to notice that in an investigation when we make use of an analogy to help us detect clues and decipher them we are not engaging in an analogical argument. We are not saying that because two things are similar in some respects they are similar in all respects. Analogical arguments are invalid because most of the time two things similar in some respects are not similar in all respects. Logician frequently warn us against analogical arguments and they are right in doing so. But we should be clear that the warning is against analogical arguments, not against the use of analogies in investigations. Analogies can be used in different ways, for different purposes. In investigations we use analogies to help us find clues and interpret them, not for drawing the conclusion that two things are similar in all respects because they are similar in some respects. All analogies will break down after a time, but this does not mean they cannot be useful in investigations. It is true that we cannot always predict where or when an analogy will break down and as a result of this ignorance make mistakes, but in an investigation we can correct mistakes.
In an investigation we should not expect an analogy to hold indefinitely even when it has been holding up well so far. Rather, we should expect it to break down at some point. Our hope is, by the time we reach that point we will have found out enough about those structures we want to know that we do not have to depend on the analogy any more.
In an investigation, when an analogy breaks down after it has served its purpose exciting discoveries sometimes occur. For why should the analogy break down? Why if it is not because we have entered new territory? If we were still in old, well-traversed terrain the analogy should still be able to guide us. But it could do so no longer; it has broken down; we are now on our own: we have to depend on context for clues. And context could sometimes reveal to us surprising things, things we might never have dreamt of before. This occurred in the discovery of the positron, which came about after scientists had discovered the limitations of the miniature solar system model of the atom. The atom is not exactly like a miniature solar system, but by the time scientists discovered this they had enough clues to fashion new theories in their attempt to understand the atom. These new theories led them to the conclusion that a particle with the same mass as the electron but with a positive charge should exist. No one expected such a particle but, lo and behold, it is there—as they discovered by examining old photographic plates. These plates were made to record the tracks left by electrons, but on a few of them they discovered tracks exactly like those left by electrons but curved in the opposite direction, indicating a particle exactly like the electron but with a positive charge. Positrons were unexpected before they were predicted by theory. They were so unexpected that scientists did not notice them even though they had photographed their tracks.
In an investigation, that an analogy should break down after a while is not necessarily a sad event. If by that time it has helped us find out enough of those things we want to know we do not mind the breakdown at all, since we can now go native (that is, depend on context for clues). Indeed, now that we are behaving as the natives behave we could be open to new experience and new discoveries, experience and discoveries not possible before we turn native.
In the course of an investigation we should take time out now and then to review what we have done. This is important not only when the investigation is not going well; it is important even when the investigation is going well. As we have pointed out, in investigations we can make use of approximations. But this is just another way of saying that even when we have made little mistakes here and there in an investigation we can still advance (approximations are small departures away from the truth). In this respect an investigation is very different from, say, setting out a proof in geometry. In the latter every line we put down has to be right; we cannot afford to make even a single mistake. If a mistake should occur, all that follows will be worthless. But an investigation is very different. In an investigation we follow clues. Clues can be looked upon as little gaps. We close these gaps one or a few at a time; there is no requirement that we should close them all at once. Since some of the closings have to be left to the future anyway, there is no harm that in closing some of them we, intentionally or unintentionally, create a few more smaller ones in the process.
But at some point these smaller gaps will have to be closed, or they will pile up and lead us astray. And this we do in the reviews we carry out from time to time. Correcting minor mistakes in an investigation in this fashion is called fine-tuning. It may have to be done not just once, but many times. After each fine-tuning we acquire a clearer idea of where we are in the investigation and where we should be heading.
I take a short cut when I can get it.
—— The Golden Pince-Nez
It is an old maxim of mine that when you have excluded the impossible, whatever remains, however improbable, must be the truth.
—— The Beryl Coronet
… we hold several threads in our hands, and the odds are that one or other of them guides us to the truth. We may waste time in following the wrong one, but sooner or later we must come upon the right.
—— The Hound of the Baskervilles
To find out things hidden from us we follow clues. But clues do not always pinpoint; most of the time a clue will suggest a number or possibilities. In investigating a murder, for example, a clue we have found could have suggested to us that the murderer could be the butler, or the maid, or the master of the house. Now in such a situation one thing we could do is to see which of these possibilities leads to more and more new clues. The one that does is likely to be the truth. But often instead of doing this we could engage in a process of elimination and by doing so arrive at an answer faster. Trying out different possibilities to see which leads to more and more new clues takes time, but if in the meantime there are other clues that will allow us to eliminate all of the possibilities except one, that remaining one is likely to be the true answer. And this we should explore during an investigation. That is, after some clue has suggested to us a number of possibilities we should try to see if there are some means of eliminating some of them. The ideal is to have only one remaining after the elimination. If there is only one it is likely to be the truth. If there is more than one, it is still a step forward, since we will now have much fewer possibilities to examine.
What if all the possibilities have been eliminated, as it frequently happens?
When this occurs one possible reason is that we have not exhausted all the possibilities before we carry out the eliminations. If we think this is the case we will have to backtrack and see what has been left out.
Suppose that after carrying out the eliminations we have one possibility left. Is this remaining possibility necessarily the truth?
We have to be careful here. It is important to keep in mind that clues never lead to absolute certainties. The process of elimination is a process we carry out when following clues. Thus even when there is only one possibility left we should not conclude that this one possibility absolutely must be the truth. The process of elimination depends on our exhausting all possibilities. Being human, we might not have satisfied this condition. For this reason, when after a process of elimination we have only one possibility left, this possibility might not be the truth. To find out, we should test this possibility to see whether it leads to more and more new clues. If it does it is likely to be the true answer. If it does not we will have to backtrack to see what possibilities we have left out.
From the above we should now be able to see that the process of elimination is but a short cut in the theseological process. There is more to the theseological process than the process of elimination. The latter does not stand a alone. It is a short cut in that instead of following a thousand and one leads (possibilities) we follow one, or a few. After the process of elimination we still need to follow other clues.
In employing the process of elimination we assume that we can exhaust all possibilities. Is it possible that in theory, and therefore in practice too, we can never exhaust all possibilities? If both in theory and in practice we can never exhaust all possibilities, we will not be able to make use of the process of elimination. For however many of the possibilities we eliminate there will always be an infinite number left.
If we cannot exhaust all possibilities we cannot employ the process of elimination. But In the theseological process the possibilities are suggested by clues, and clues always narrow down the field. A clue cannot suggest just any thing and still be a clue. A clue always rules out as well as point to. When a clue points to discrete units, the number of units is always finite—there is only a finite number of people who could be the murderer, for example. A clue sometimes points to a range in which the true answer can be found. Now in this range there could be an infinite number of possibilities, but in this kind of cases we can use the process of elimination to narrow down the range. For example, we could say, the true answer has to be in one of three regions within this range: regions A, B, and C (each of which also contains an infinite number of possibilities). The true answer is not in A and B. Therefore it must be in C. Now it is true that even after the process of elimination we do not know what the true answer is, but we know that it is in C. And if this process can be repeated we should be able to come closer and closer to the true answer even though we will never know exactly what it is.
To repeat, it is true that if we cannot exhaust all possibilities we cannot employ the process of elimination. But when following clues we in theory should be able to exhaust all possibilities. This is because clues always narrow down the field. They rule out as well as allow in.
“But how will you look?”
“I will not look.”
“I will get her to show me.”
—— A Scandal in Bohemia
In one of his cases Sherlock Holmes was asked by a client who was being blackmailed to retrieve some photographs in the possession of a certain lady. It was known that the photographs were in her house, but where in her house nobody knew but her. In order that he could take possession of these photographs—that is, steal them—Sherlock Holmes devised a simple stratagem. He started a fire in the house (which fire was in fact a false alarm) and watched the first thing the lady did. And what was the first thing the lady did? She rushed to the spot where the photographs were hidden.
Sherlock Holmes wants to know the location of the photographs. Obviously, there is no point in asking the lady concerned; she will never tell. But this does not mean that Sherlock Holmes cannot find out. He can still ask the same question, in the form of an experiment: he starts a fire. The result of the experiment tells him where the photographs are.
Without a confession from the blackmailer Sherlock Holmes succeeds in finding the answer to his question. His question is not addressed to the blackmailer but to the situation in which the blackmailer finds herself. A blackmailer needs an instrument of blackmail—in this case, the photographs. As an instrument of blackmail the photographs are of great value. When there is a fire they will be among the first things she will want to save.
In theseologising there will be questions for which we want answers. Sometimes some of these questions can be answered by experiments. In these cases these questions are not addressed to those who already know (for example, the criminals themselves, or God), but to the structures we are investigating. Since structures have characteristics these experiments can provide the answers we are looking for. In escaping from a fire most people will want to take with them things they regard as valuable.
As we all know experiments are common in science. In them scientists ask questions of those parts of nature that they are investigating. How many helixes are there in the DNA molecule? How much electric charge is there on each electron? These questions were answered by experiments.
We do not usually associate experiments with cryptanalysis, but in fact in cryptanalysis experiments are sometimes carried out. Just as Sherlock Holmes can induce a blackmailer to reveal the location of photographs being used in the blackmail, cryptanalysts can induce their opponents to reveal to them the meaning of certain code words, for example. Take a simple case. Suppose you want to find out what code name your opponents have assigned to General Eisenhower. To this end you send out to your own people an important, secret message, faked, in which the name General Eisenhower will be mentioned. Now if this message is picked up by your opponents and repeated among themselves using their own cipher (which you have broken), you will be able to find out what code name they have assigned to Eisenhower. For example, suppose you say in your original message that 'Eisenhower is visiting the island of Crete on March 15'. Now if the same message when repeated among your opponents comes out as 'Shepherd is visiting the island of Crete on March 15', you know that Shepherd is the code name your opponents are using for General Eisenhower.
In following clues we can ask questions of the as-yet unknown and find answers. This usually requires a certain amount of ingenuity—as in Sherlock Holmes's case, but is worth attempting. In uncovering the hidden the more we know the more we can find out.
But it is worth noticing that we can ask questions of the unknown through experiments only when much of the unknown structures being investigated is already known. In our cryptanalytic example earlier we could not have found out that Shepherd is the code name for Eisenhower if we could not decipher the message. This is to say, by the time we can ask questions of the unknown through experiments, what we have found out earlier in the investigation is likely to be correct. If it were not we would not have been able to design the experiments or interpret the answers. Now in the mature sciences we can often ask questions of this type. This shows that in these sciences we already have found out a lot.
It is difficult to follow clues. In following clues success is not promised in advance. And even when it is achieved much time and effort will have been expended. This comes about because there will always be hurdles to be overcome; and mistakes will always be made and, after they have been made, corrected. If we want to find out anything by following clues we therefore will have to be tenacious: we will have to keep trying even though we meet with disappointment after disappointment. In following clues we should expect many failures before we succeed, if we succeed at all. If we are not tenacious; if we give up too easily; we will never find anything at all.
Failure is hard on our psychology. In tasks we carry out, if we keep failing, in time we will want to give up. Now failure is common when following clues. There will be failure in finding them, failure in the interpretation of clues found, failure in the form of mistakes of all sorts, and failure in correcting these mistakes after they have been made. If we want to be successful when following clues we will have to be prepared for a large amount of failures. Not only this, but in face of these failures we will have to keep trying. Now this is hard to do unless somehow we manage to instil in ourselves a certain degree of optimism: we have to sustain the hope that, given time, we will find what we are looking for. Now this hope may not be realised, but the moment it is gone our investigation will terminate, if not immediately, then soon. We are human and subject to our psychology.
A certain degree of skepticism is always called for in investigations. When following clues the chance that we have made mistakes, major or minor, singly or in plural, is always there. There is never absolute certainty, however many generations of new clues we have reaped. We therefore should not be too attached to the results obtained in investigations: there is always the possibility that some of them are wrong and may have to be revised or abandoned later on. It is even possible that some of these results are seriously wrong and we do not know at this time which. Now this kind of skepticism is healthy: it will not take away from us the hope that by following clues we will find out more and more. Indeed it will only urge us to carry on and to persist; for the more we know, even if only approximately, the easier it is to uncover past mistakes and to correct them. As we have seen it is easy to make mistakes when following clues, but since mistakes can also be corrected when following clues, the possibility that we can make mistakes should never be used as an excuse for giving up on the search for knowledge altogether. Skepticism to this extreme; skepticism that saps our desire for knowledge to such an extent that we would give up the search; is uncalled for. In the search for knowledge a certain degree of skepticism is justified, but the kind that induces us to abandon it completely is not.
1We consider how we should look for evidence in a later section.
2There is an exception to this rule. When we run out of clues and are desperate we would allow ourselves to make wild guesses.
3Even then we should not be dogmatic that the pattern must be significant for our investigation. There could be a good reason for the pattern, but it could be irrelevant to our investigation.
4Karl Popper adopts this view (see K. Popper, Conjectures and Refutations).
5Attempt A focuses attention only on the relation between premise and conclusion
6This gives the theseological method some very special characteristics.
7In the last chapter, in discussing clues, we have already pointed out that it is permissible to use approximations in an investigation. Our point here about using approximations when forming assumptions reinforces this view.
8Michael D. Coe, Breaking the Maya Code (New York: Thames and Hudson, 1992).
9This is not to say they will not be interested in what their fellow investigators have been doing if this knowledge of what they have been doing will not help them in their future investigations.
10For more on this point see the section Do look for analogies.
11There is a place in the theseological process for imagination, but in this process imagination cannot be allowed to run amok.
12Even when we are patient we still may not find the things we are looking for, but nothing ventured nothing gained.
13Will we ever know the exact orbit of the planets?