Scientific method and neurobiology

Authored by Henry Strick van Linschoten

Working within the parameters of scientific method

The papers in this module apply a concept of science that is restricted to natural causes, explanations and events, but includes the social sciences as science.

The following points are intended to emphasise a number of probable truths about the scientific method and to counter widely held dogmas that may lead to false assumptions.

  1. Science and the scientific method are not clearly defined. There is no consensus, only many methods and approaches. There is agreement that tradition and authority are not scientific sources.
  2. Science is interpersonal, and has to do with the testing, agreeing on and rejecting of ideas about people and the world.
  3. It is easy to assert that two phenomena are linked or correlated in certain circumstances; but it is difficult to establish and much more useful to show that something is caused at least in part by something else. It is almost impossible to prove that two things are not connected.
  4. There is no necessary connection between the personal qualities of a researcher, scientist or practitioner, and the likelihood that what they say is true or scientific.
  5. Science does not necessarily progress over time towards being more inclusive and more correct. Some valid ideas can “disappear” and return (e.g. the aetiology of autism, the understanding of dissociation, understanding of the impact and prevalence of child abuse, infection as a cause of ulcers).
  6. Knowing the type of cause of a behaviour, problem or symptom is not connected with the most effective remedy, cure, treatment or alleviation. One needs to establish case by case what treatment or remedy might work, even if the cause or certain contributory causes are well known.
  7. In any field to do with human beings, animals, relationships, systems, or ecology, it is extremely rare that behaviour or observable events have a single determinate cause. Multiple as well as probabilistic causation are the rule. (“Fallacy of the single cause”)
  8. Specialist vocabulary or narrative can be extremely useful but does not constitute an explanation in itself, nor a theory of causation or otherwise. It is possible that something can be named or described in various ways, and even that some differences in description can be helpful in a fuller understanding of the topic (Examples: various descriptions of ‘unconscious’ or of the ‘structure’ of the personality).
  9. Theory and reality have different properties. In particular, reality is rarely simple and often requires a systems view to be charted. For theories, simplicity is a virtue.

Association – correlation – causation

The term ‘association’ is used in a number of contexts:

  • Associationist theories of psychology. These assume that association, the similarity or closeness in space or time of ideas and experiences, is fundamental for all thinking, memory, learning and knowledge. This was especially developed philosophically by the British Associationists, also known as British Empiricists, including Hume, Locke and John Stuart Mill.
  • The common-sense usage of people associating things or people: a thing or person reminds them or makes them think of another
  • The idea in empirically-based education theory that association plays a role in learning
  • The important usage in statistics of association in the sense of correlation

All these meanings can be useful, but the differences between them can lead to confusion. Association in several of the above meanings can be useful heuristically: it can help people understand a topic better, help to understand one’s own or other people’s behaviour, and it can give an accurate description of “what’s going on”. In the more colloquial of these senses association seems to emphasise the randomness of the association, rather than the belief that there might be more meaning or significance behind the ideas or events finding themselves linked in someone’s mind.

Association in the sense of correlation is a more important matter. If in one of many ways one has established that two events remarkably often happen at the same time, coincide or are found together, this often leads to the next question: what might be the reason for or the meaning of this? Correlation is much easier to establish than causation; it can often be established without organised design, planning or experimental set-up. Correlation depends on observation, and can often be observed without more than minimal interference by the observer. Developmental psychologists have long noted that very young children already have a sense of causation, but what they generally describe is that young children notice correlation, and take this to mean causation in one way or another. Correlation is an observation that things happen together, but it is not seen as an explanation, or a full understanding of why this would be the case. One can say that most of the advantage of causation over correlation is that only a cause really explains, whereas mere correlation leaves the observer dissatisfied at their lack of explanation.

During much of the 20th century the concept of cause was used less and less in science and in statistics, but the past few decades have seen a major revival of interest in the idea. However, much of this development has taken place in particular scientific areas and in philosophy in isolation, without being fully integrated. For some of the controversies in neurobiology, and for a complete understanding of its potential, it is essential to understand and be able to use modern ideas about causation.

Trying to establish causes is a mark of scientific thinking. In prescientific thinking, the main purpose of establishing causes was to establish responsibility, credit or blame, which meant that only human actors were regarded as valid causes. Outside human causality the world was then seen as deterministic, i.e. there were other causes such as gods, nature, ghosts, spells. Science has brought a shift to seeking causes principally to explain, and in believing that causes can be material and impersonal just as much as personal (Pearl, 1996).

A restrictive idea of the role of the psychotherapist would be that they are only interested in cause and effect in psychotherapy, and nothing else. In practice, along with most mental health professionals, they are likely to be interested in knowledge that can help with the prevention or alleviation of disorders, problems and distress; from a general standpoint one would expect mental health professionals to be more effective in their work if they understand the human mind-body better – starting with its grounding in interpersonal neurobiology.

There is by now a body of modern thinking about causality that fits well, as it is practical, operational, has a track record of being used, and fits with the special characteristics of the fields of psychology, psychotherapy, neuroscience and biology. A few simpler and shorter sources are Pearl (1996)Woodward (2008)Rutter (2006, Ch. 2) and Rothman (2012, Ch. 2). A fuller understanding can be reached by reading Shipley (2000)Woodward (2003)Rothman et al. (2008)Glymour & Greenland (2008)Pigliucci & Kaplan (2006, Ch. 2) and Pearl (2009)Cartwright & Hardie (2012) add a lot of practical considerations to the use of causality to support decisions about public policy – for that specific application it is indispensable.

Following are a few conclusions from the above body of methodological thinking that are relevant to this module. These conclusions are especially valid in the fields of interpersonal neurobiology, psychology and biology:

  • The easiest way to establish causes is through conducting repeated randomised controlled trials (RCTs), so there is a strong preference for doing so when this is possible.
  • There are many reasons making it impossible to conduct RCTs, such as costs, problems of scale (size of the trials that would be necessary to statistically demonstrate small differences), the time it would take (in genetics with mammals, waiting for several generations to reproduce; in human trials, the time it might take to know that certain behavioural changes will be sustained), technical limitations (as discussed under brain-imaging, the resolution of the different techniques remains very unsatisfactory), limitations of principle (this can be very real with multifactorial causality – it also plays a role in economics), and ethical restrictions (real and significant, concerning animals as well as human subjects; it is not acceptable to run a risk of harming, killing, or inflicting significant pain without the clearest forms of consent). This is especially the case in psychology, biology, medicine and human neurobiology and neuroscience. Indicated are only some illustrations of the problems.
  • The statement has often been made that causes can only be demonstrated by RCTs. This overstates the relative effectiveness of RCTs, and ignores the many other methods that exist to make plausible attributions of causality.
  • The weakness of correlation can be overplayed, too. Although correlation is not enough to establish causation, one can usually conclude from a strong correlation that there must be a causal mechanism, as yet undetermined or unresolved, that can be identified to explain the correlation (starting with the basic possibilities: A causes B; B causes A; A and B have a common cause). When there is causation, there will always be correlation: causation implies correlation. The few exceptional reasons for correlation without causation are usually definitional and semantic, or can be seen very easily. Examples: The rooster’s crow and the rising of the sun; Tuesday following Monday; the barometer rising followed by rain.
  • Correlation is easier to establish than causation. Statistical methods can only establish correlation.
  • To describe uncertainty the language of probability remains very powerful. Until the last few decades there were no easy ways to translate between causal models and the language of probability. This has now changed.
  • Causality in human and animal affairs is almost always multifactorial, i.e. explanations that are sound and complete almost always involve a large number of different causal factors, some of which are also interdependent. A good example is the regularly encountered neurobiological statements that (activity in) one brain location “controls” a certain gland or muscle. This must always be understood as only possible when a whole range of other conditions are satisfied, and when other causal factors are not interfering with the causal connection highlighted in the statement.
  • None of the above implies an unusual or particularly specific definition of causality that needs to be agreed on. The conclusions are valid based on a rather thin definition of causality that can be summarised in some axioms.
  • The new ideas do not mean the abandonment of research design, experiments and trials, or the value of the study and analysis of data derived from observation. On the contrary, they make this more valuable than ever.

Here are some points that do not depend on these new methodological perspectives.

In biology, and by extension in human biology, there are two types of causation: proximate and distant. Proximate questions ask “how” questions that deal with the immediate reasons for the behaviour of this particular person or organism. Distant and “ultimate” questions are searching for explanations in the development of the organism, or more usually, in the evolutionary causes (whether based on genetic, epigenetic or environmental mechanisms introducing change) why something happens or has developed. An example: evolutionary psychologists would say that women are on average attracted to successful (however that is defined in a culture) and able-bodied men for evolutionary reasons. Having this preference would be a type of behaviour that would be strengthened by natural selection (ultimate cause). If a particular woman is attracted to or dates a successful able-bodied man, she would do that because she actually feels attracted (proximate cause). A psychotherapist, however, might say that she is attracted not for her own reasons, but perhaps because the man reminds her of her father. If accepted, that would be a cause intermediate between proximate and ultimate.

The language around causation uses a range of words. This may be the result of the long period in the 20th century in which using the words “causes” and “causation” were discouraged in many journals and scientific communities. From the modern view about causality, questions about aetiology are questions about the causes of something; when a part of the body is stated to “control”, “influence” or “affect” another part.

Giving a name to a behaviour (or a set of symptoms) does not explain, and certainly does not establish causation. It is sometimes said that someone behaves “like this” (naming one of the symptoms listed to diagnose Borderline Personality Disorder) “because” they have (or even worse, “are”) BPD. Or they have difficulty with their attention “because” they have, or are diagnosed with, ADHD (Attention-Deficit / Hyperactivity Disorder). This is not an argument against terms, classification, or diagnosis. There are many reasons why clinical practice and medical, biological and psychiatric research are hard to imagine without diagnostic categories. Learning from and comparing the treatment of substantial groups of clients who have some significant similarities in their problems simply is effective. Without definitions and classifications, no classes of problems can be researched statistically, nor can their causes be studied or speculated about. Think about the progress reached in the understanding of autism, of PTSD, of attachment styles, of dyslexia, of Alzheimer’s. None of that would have been possible without the “naming”, the diagnostic nomenclature.

References

Cartwright, N. & Hardie, J. (2012). Evidence-Based Policy: A Practical Guide to Doing It Better. Oxford: Oxford University Press.

Glymour, M.M. & Greenland, S. (2008). Causal diagrams. In K.J. Rothman, S. Greenland & T.L. Lash (Eds.), Modern Epidemiology (3rd edition). Philadelphia, PA: Lippincott Williams & Wilkins.

Pearl, J. (1996). The art and science of cause and effect. Part of the UCLA faculty research lectures. Available at: bayes.cs.ucla.edu (accessed 27 December 2013).

Pearl, J. (2009). Causality: Models, Reasoning, and Inference (2nd edition). New York: Cambridge University Press.

Pigliucci, M. & Kaplan, J. (2006). Making Sense of Evolution: The Conceptual Foundations of Evolutionary Biology. Chicago, IL: The University of Chicago Press.

Rothman, K.J. (2012). Epidemiology: An Introduction (2nd edition). Oxford: Oxford University Press.

Rothman, K.J., Greenland, S., Poole, C. & Lash, T.L. (2008). Causation and causal inference. In K.J. Rothman, S. Greenland & T.L. Lash (Eds.), Modern Epidemiology (3rd edition). Philadelphia, PA: Lippincott Williams & Wilkins.

Rutter, M. (2006). Genes and Behavior: Nature-Nurture Interplay Explained. Malden, MA: Blackwell Publishing.

Shipley, B. (2000). Cause and Correlation in Biology: A User’s Guide to Path Analysis, Structural Equations, and Causal Inference. Cambridge: Cambridge University Press.

Woodward, J. (2003). Making Things Happen: A Theory of Causal Explanation. Oxford: Oxford University Press.

Woodward, J.F. (2008). Cause and explanation in psychiatry: an interventionist perspective. In K.S. Kendler & J. Parnas, Philosophical Issues in Psychiatry: Explanation, Phenomenology, and Nosology.Baltimore, MD: The Johns Hopkins University Press.