Some authors have highlighted the potential for error in the reasoning process (Rivett & Jones, 2004; Scott, 2000). The patterns that are recognized in this first stage will, in the majority of situations, later develop (generate) into testable hypotheses and it is this collection of hypotheses which will form the shape of the remaining assessment and examination process. Therefore, the whole diagnostic and decision-making procedure is reliant on the quality and accuracy of these early patterns. In respect of early pattern recognition, potential errors may take the form of developing the wrong initial concept of the problem; over-emphasis on favoured hypotheses, misinterpreting early information, making assumptions (this should be carefully considered as being very different to pattern recognition), considering too few (narrow range of) patterns (or hypotheses), making hypotheses too vague, and failing to detect early cues for serious pathology. There is equal potential for reasoning errors throughout the whole assessment procedure, but it is problems within the hypothesis generation stage that will misdirect the subsequent procedure. Hence utmost attention should be given to the potential for error at this stage.

The concept of error in pattern recognition is paradoxical: pattern recognition relies on experience yet it is that very experience which is the most likely cause of error in pattern recognition. Common errors, as stated above, include narrowness of hypotheses, misinterpreting information, and failure to recognize early signs of serious pathology. These errors are products of experience which have not been subject to the scrutiny of careful thought (cognition), reflection, critical analysis or self-monitoring (metacognition). In the absence of careful cognitive and metacognitive processing, external (to the clinician) experience may be internally (to the clinician) mis-interpreted. This is the psychological phenomenon of perception. The interpretation of any experience is subjective and specific to the individual. All external experiences are filtered through our own, personal schemata, which is itself developed by experiences. Thus, if we have psychological schemata which are built of aberrant/limited/biased interpretations of previous experiences, all future experiences are going to be interpreted within these skewed schemata. A good way of demonstrating this concept is with the use of an optical illusion (Fig. 3.2).

Figure 3.2 • Boring’s old woman/young girl illusion as adapted by cartoonist W. E. Hill in 1915. This classic pictographic illusion is an example of the problems associated with individual perception.

(Source: www.grand-illusions.com/woman.htm)

The image in Figure 3.2 has two interpretations: (1) an old woman and (2) a young girl. Now that you know this information, you should be able to see both interpretations (although some will still find this harder than others). However, if you did not know that there are two interpretations, you will probably only see one and be satisfied that this is a nice picture of either an old woman or a young girl. Each subsequent time you see the picture, you will automatically interpret the image you are familiar with, because your personal schemata has been ‘set up’ with this first experience. Additionally, each subsequent time that you see the image, your interpretation of whatever that early experience was will be reinforced. Therefore, you could coast through life believing that this picture has only a single interpretation and this belief would continue to get stronger and stronger, eventually becoming almost impossible to break away from. However, the critical reasoner would want to question their early experience of interpreting the picture. In this case, you could perhaps spend longer examining the image to see for yourself if there really was another interpretation. Or you could spend time thinking about whether there at least could be an alternative. Or you could challenge your own perception of the image and re-examine it. Or you could ask another viewer if they see the same thing as you. Better still, ask a number of other viewers to see what their interpretations are – the more viewers you ask, the more likely it would be that you would come across someone who has seen the alternative interpretation, thereby challenging your belief. These later activities are part of the cognitive/metacongnitive process associated with good reasoning and are more likely to reveal something closer to reality than just accepting the first interpretation. The continual inclusion of the reflective cognition/metacognition process is imperative in determining the quality of pattern recognition and hypothesis generation.

Clinical relevance

Pattern recognition has been recognized for its many values, e.g. efficiency in dealing with a problem quickly without the need to clarify and re-check, efficiency in dealing with uncertainty, imprecise data, or limited premises (Albert et al, 1988). However, effective and efficient pattern recognition is reliant on the following three conditions being met:

1. You are an expert

2. The clinical problem is not complex

3. You are very familiar with the clinical problem.

If all three of these conditions are not in place, the process is prone to error (as discussed above) and the limitations of the logical basis of pattern recognition are exposed. The logical basis of pattern recognition is known as induction. Hence, pattern recognition is also referred to as ‘inductive reasoning’ (Higgs & Jones, 2000, p. 6). The basis of induction is one of making inferences about something from prior experience or observation(s) of a similar event. For example, I have switched my kettle on 200 times and it has never exploded, therefore I affirm that when I switch it on now (the 201st time), it will again not explode. This seems quite a reasonable assumption, and this logical process is largely what drives us to make all sorts of decisions and choices on a daily basis throughout our lives. However, there is a strong philosophical argument which challenges reliance on inductive reasoning and questions its logical basis. Consider the following inference:

Premise A: I have seen 20 patients this week whose headache was caused by upper cervical facet dysfunction.

Premise B: Upper cervical facet dysfunction causes headaches.

Conclusion: My next (21st) patient that I see with headache will have upper cervical facet dysfunction as the cause.

This is a simple logical inference of which the third statement (conclusion) is a product of the first two statements (premises A and B). It is an example of inductive reasoning. However, its logic is fundamentally flawed: despite the fact that both premises A and B are true, it does not necessarily follow that the conclusion drawn is also true, i.e. even if upper cervical dysfunction can produce headache and all the previous 20 patients had cervicogenic headache, it does not guarantee that the 21st, or indeed any future, patient will also have cervicogenic headache – for we know that there are many other causes of headache.

A counter-argument in support of induction logic is often made using the concept of probability to strengthen the assertion that the conclusion is a reliable product of the premise. For example, we might think: ‘well, surely the more times I observe something, then the more likely it is that the subsequent event can be predicted based on my observations’, i.e. likelihood is directly proportionate to number of times observed. Clinically, this could be reflected as: ‘actually, I have been practising for 20 years and I have seen 3000 patients who have presented with this type of headache and it has always been caused by upper cervical dysfunction. Surely, with this number of experiences, it is very likely that future patients will have upper cervical facet dysfunction?’

Even if we ignore the errors of perception and experience as discussed above, and if it is true that all the 3000 previous patients had upper cervical facet dysfunction, the assertion that future similar patients will have the same cause is still flawed. The attempt at using probability (‘the more I have seen, the more likely it is that I can predict future events) does not strengthen the case for inductive logic. Let us explore the logical basis a little further. Essentially, with induction we are stating a general law (e.g. all headache of this type is cervicogenic) from particular observations (e.g. I have observed a number of individual occurrences of cervicogenic headache) with the implication that the general law is true (e.g. it is true that this type of headache is cervicogenic). The general law transcends the individual; in other words the statement we are making is intended to be applied to all headaches of this type – in the past, the present, and the future, and in the whole world. The simple way in which this logic fails is, of course, by the single discovery that, in the past (by someone else), in the present (by anyone), in the future (by anyone), anywhere in the world, this type of headache is not cervicogenic. In an attempt to counteract this flaw, induction dictates that we should aim to make statements (general laws) with maximal probability. So, rather than stating that ‘all headaches of this type are cervicogenic’ which possesses the above probability flaw, we should say ‘either this type of headache is cervicogenic or it is not’, or ‘headaches are caused by a number of things’, i.e. these statements have maximized the probability of the law being true. Of course, these statements are true, but have minimal informative content, i.e. they do not actually tell us anything informative about understanding the nature of the headache. This is analogous to saying ‘it will either rain or it will not’. This is true, but it does not help me decide whether to wear my raincoat or not. Thus, the unquestioned accumulation of personal experience to inform future decisions should be judged with great caution in the intellectual clinical reasoning approach.

Consideration of the problems of induction is nothing new. The argument against inductivism was highlighted in the work of an eminent 20th century philosopher called Karl Popper. His claim was that adherence to an inductivist approach is erroneous and limited and that in order for a more accurate approximation to what is correct to be made, a deductive approach should be taken (Popper, 1980). Luckily, for us manual therapists, this has already been given some thought and forms the next stage in our reasoning process.

Stage 2: Hypothetico-deductive reasoning

Now that patterns have been recognized, the clinician has two choices:

1. Accept the most likely pattern as the diagnosis and begin treatment

2. Use these patterns to generate a number of hypotheses which can then be tested for their validity.

Option one would be an extreme form of expert practice and is totally reliant on the pattern recognition model. As stated above, this can be very quick and efficient. However, also as stated above, it will only work if certain conditions are met, and the model is prone to error and logical limitation. Therefore, for most people at most times, option two would be the most reasonable option. For the clinical reasoner, the rest of the history-taking is not a random, or even pre-structured, list of questions but a reasoned, hypothesis-driven process of inquiry. This process is referred to as a hypothetico-deductive model of reasoning, and is written about extensively in both general clinical reasoning (e.g. Barrows & Tamblyn, 1980; Elstein et al, 1978) and manual therapy specific clinical reasoning literature (e.g. Higgs & Jones, 1995; Jones, 1995; Rivett & Jones, 2004). The hypothetico-deductive model describes a process of data collection, hypothesis generation, interpretation and hypothesis testing. As stated earlier, in reality, a combination of various reasoning models will be operationalized, e.g. simply because we have now stated that the predominant process has evolved to a hypothetico-deductive structure does not mean that future patterns will not be recognized, and subsequently tested as hypotheses, and so on. However, for clarity, at this point we shall focus our attention on hypothetico-deductive reasoning.

Figure 3.3 • Example of the chronological thought process associated with the transition from early pattern recognition, illness script recall, and hypothesis driven questioning.

In the middle box representing the illness script in Figure 3.3, you will notice that the ‘script’ is separated into three distinct components, i.e. potential diagnosis A is conditional of (1) there being a specific type of trauma, (2) that specific tissues have been affected, and (3) that the manifestation of (1) and (2) are a specific set of signs and symptoms. Boshuizen and Schmidt (2000) refer to these three components as:

1. The enabling condition

2. The fault

3. The consequences of the fault.

The enabling conditions are the conditions under which the problem/disease may occur, and considers such information as social history, hereditary conditioning, medical or environmental background, etc. The fault refers to the actual pathophysiological process taking place for this problem/disease. The consequences of the fault are therefore those clinically observable signs and symptoms considered to be associated with the specific problem/disease. For our case study we can construct examples of possible illness scripts which may be associated with the patterns recognized early on for our 38-year-old male with neck and head pain. Table 3.1 shows three patterns (for illustrative purposes, only three patterns have been used in this example – of course, it is probable that, in reality, there would be more than three potential diagnoses) and the associated illness scripts which could be potential diagnoses for this patient.

Table 3.1 Example of three patterns and associated illness scripts related to the case study of upper cervical neck pain and headache

The clinician’s structuring and prioritization of questioning within this part of the assessment procedure is now directly governed by consideration of these factors. Hence, hypotheses are generated and then tested for their validity by specific, hypothesis-driven interrogation – thus the hypothetico-deductive process evolves. This ‘testing’ procedure derives its logic from, as the name of this model suggests, deduction. Unlike the inductivist logic of pattern recognition, this logical form is considered in many ways to be a more robust process of inquiry. However, compared to a true pattern recognition dominant approach, it is arguably a more time-consuming and thus less efficient model. The ‘robustness’ of deductive reasoning can also be examined in a philosophical context. The following logical progression is an example of deductive logic:

Premise A: all treatable cervicogenic headache and neck pain is differentially characterized by features X, Y and Z (as encapsulated in the specific illness script).

Premise B: Mr M has features X, Y, and Z.

Conclusion: Mr M has treatable cervicogenic headache and neck pain.

Unlike inductive inference, deductive logic dictates that if premises A and B are true, then the conclusion must also be true. In other words if it is true that features X, Y, and Z are differentially characteristic of cervicogenic headache and neck pain, and if it is true that Mr M demonstrates these precise features, then it follows that Mr M has cervicogenic headache and neck pain. Thus the premises entail the conclusion. Clinically, we can harness and exploit this process to take us closer to what the actual problem might be. The process is effectively a test of strength, and only the strongest hypotheses (i.e. those that survive the directed and rigorous questioning) will make it to the next stage of assessment – the physical examination. In reality, it is likely that only in the most straightforward of situations, only one surviving hypothesis will be considered for testing in the physical examination. Most commonly, there will still be a number of competing hypotheses to refute during the physical examination – hence the hypothetico-deductive model continues. Additionally, with one primary hypotheses (e.g. upper cervical motion segment dysfunction) there will be a number of secondary hypotheses (e.g. left-sided C1 on C2 posterior-anterior hypermobility relating to a regular movement pattern X versus C2 on C3 anterior-posterior hypermobility; alternative hypotheses regarding the SIN status of the patient’s condition) contending for survival.

Although the logic of this inference is philosophically robust, it is obvious that the ‘truthfulness’ of the conclusion is dependent on the ‘truthfulness’ of both premises. This ‘truthfulness’ is a different matter altogether and is dependent on the quality of the knowledge sources from which the clinician has (a) constructed the illness script, and (b) assessed the patient. Thus, knowledge and its integration into the reasoning process must also be explored.

Stage 3: What do we know and how do we know it? Knowledge

Knowledge is a key component in the framework of clinical reasoning (see Fig. 3.1). As stated above, the validity of the deductive logical inference is dependent on the quality of knowledge which supports its premises. If our reasoning is reliant on our knowledge, we must consider the nature and quality of our knowledge, and question where it comes from.

If you were a patient concerned with receiving the best care, what would your reaction be if, following you asking your therapist what he thinks is wrong with you and how he knew that his diagnosis was as accurate as possible, the reply was something along the lines of: ‘Your pain is caused by evil spirits possessing your upper cervical spine and I know this because I went on a course by a Mongolian shaman and he said that this is what causes neck pain’. I suspect (and hope!) that you would immediately question the logic of this knowledge. If so, you would come to the conclusions that:

• The type of knowledge (i.e. pain related to upper cervical evil spirits) is not valid

• The source of this knowledge (i.e. knowledge acquired from the shaman) is also not valid.

It may be the case that if you were in some other period of history, or another part of the world, you would accept this type of explanation. However, in our modern, thoughtful, and accountable society, this kind of justification for decisions involving other people is neither substantiated nor acceptable. We therefore need to consider what types and what sources of knowledge are considered substantiated and acceptable. Uncomfortably, this area of study is akin to having that annoying, pedantic companion who continually retorts to everything you say with the question ‘but how do you know that’! However, as responsible reasoners we must be happy to go through this process.

Let us continue with our clinical scenario: to summarize at this point, we have established the following details:

• Age of patient: 38

• Gender: male

• Pain site: left-sided mid to upper cervical pain and parietal headache

• Possible causes: for our selected examples – the three hypotheses in Table 3.1.

Embedded within this summary are a number of ‘facts’, the source of which can be subjected to scrutiny. The first two summary points (age and gender of patient) are fairly robust statements based on reliable and valid knowledge sources, i.e. date of birth (verified by birth certificate if needed), and chromosomal and physical characteristics of being male (verified by DNA analysis and physical examination if needed). The third point (pain site), is reliant upon the patient’s report (to which we shall return below). It is the fourth point however, which is of most interest to us as this opens up a variety of knowledge, and therefore reasoning issues, which may directly influence the accuracy of our clinical decisions, i.e. if our knowledge is not correct, then our subsequent clinical decision may not be correct.

Table 3.1 suggests a number of characteristics (in the form of illness scripts) suggestive of each proposed hypothesis. We can now question the quality of this information and examine where this information comes from, i.e. how do we know that characteristics X, Y, Z equate to diagnosis A.

The formal theoretical study of knowledge is called epistemology. The substantial, underpinning detail of this branch of philosophy is beyond the scope of both this book, and, I imagine, the desire of many readers to go into in any great detail! However, it is quite accurate to say that we are all now undertaking an epistemological review of our clinical practice.

Types of knowledge

Types, or forms, of knowledge have been discussed and presented in a number of therapy-specific texts and publications, e.g. Higgs et al (2004), Higgs and Titchen (2000), Jones (1995), Resnik and Hart (2003), Resnik and Jensen (2003) and Swisher (2002). Classically, types of knowledge have been sub-divided into the categories which are shown in Figure 3.4.

Figure 3.4 • Types of knowledge. Propositional know-that knowledge and non-propositional know-how knowledge.

The interested reader is referred to Higgs and Titchen (2000) and Higgs et al (2004) for detailed information regarding this classification. In summary, propositional knowledge is knowledge about propositions! Propositions are the assertion that something is the case, e.g. the sun is the centre of our universe. In manual therapy, propositional knowledge refers to theoretical or scientific knowledge derived through scientific study and accessed through publicly available information sources such as peer-reviewed journal articles and books. Thus propositional knowledge is know-that knowledge. Professional craft knowledge is that which relates to actual practice skills, information, and competencies required of clinicians and is acquired through experience and is often tacit or ‘sub-conscious’. Personal knowledge is similarly gained through social (clinical) experiences but refers to that reflective, more conscious component of practice which forms our individualized frames of references, values and beliefs. Professional craft knowledge could, for example, include the manual skill needed to assess upper cervical passive physiological movement dysfunction as learnt on a post-graduate course and practised over a period of time. Personal knowledge, however, would be that independent, unique appreciation of what ‘stiffness’ actually feels like, or where ‘resistance’ begins, etc. Professional and personal knowledge are forms of ‘non-propositional’ knowledge and are related to ability, i.e. I am able to accurately assess resistance on this regular combined movement pattern and am able to make judgement on the degree of stiffness. Thus, these types of knowledge are often referred to as know-how knowledge.

The importance of attempting to understand types of knowledge is to make a judgement, as inferred above, on the quality of that knowledge. An expert practitioner will not only have more knowledge, but will utilize different knowledge types (compared to a novice), in turn from a variety of logically valid sources (Boshuizen & Schmidt, 2000; Gobet, 1998; Gobet & Wood, 1999; Jensen et al, 2000; King & Bithell, 1998).