Authors have argued for several years that psychology cannot be reduced to physiology because the emergent properties of the body cannot be deduced from biological theories (Bem & Jong Reference Bem and De Jong2013; Kirsch & Hyland Reference Kirsch and Hyland1987). The target article by Borsboom et al. provides a good rationale for why those emergent properties arise and therefore provides a convincing argument for why reductionism is not possible. However, networks apply not only to mental illness, and the same argument can apply to functional disorders.
The term functional disorder refers to patterns of somatic symptoms with no unique pathophysiology, also known as medically unexplained symptoms (MUS). Common examples of MUS include irritable bowel syndrome, fibromyalgia syndrome, and chronic fatigue syndrome. Less common examples include non-epileptic seizures and functional blindness. Functional disorders respond poorly to pharmacological interventions and, although evidence-based psychological interventions are recommended, outcome is often poor, with some patients rejecting a psychological interpretation.
There are two types of network. First, there are networks where symptoms cause each other. This is the type of network proposed by Borsboom et al., which I will call a symptom network. Second, there are networks where the biological mechanisms that cause the symptoms are themselves part of a network, and which I will call a mechanism network. Symptom networks can occur without mechanism networks, but mechanism networks lead to the expectation that causality between symptoms will also form part of the network structure. Thus, mechanism networks are consistent with symptom networks but not necessarily vice versa.
Mechanism networks can be treated in two ways. First, because the mechanisms are biological, it may be assumed that they can be treated as any other form of biological theory. The mechanism network is simply a more complex form of biological theory than that normally envisaged in medicine, where cascades of chemical events are described in terms of a causal sequence of events (Sturmberg et al. Reference Sturmberg, Bennett, Martin and Picard2017). From this perspective, the application of the network concept is consistent with a reductionist perspective and requires a biological approach to diagnosis and treatment. However, this reductionist approach fails to accommodate the emergent properties of networks. Networks have emergent properties because the behaviour of the network as a whole depends on the strength and rate of change between the different connections among the nodes of the network – not on the properties of the individual nodes or mechanisms. One of the emergent properties of networks is rule-following behaviour, which has been demonstrated elsewhere (Wolfram Reference Wolfram2002). Repetition of complex rules leads to complexity that cannot be predicted from an initial state. Therefore, the initial state of a biological network system cannot predict the outcome of that repetition when the network is of a kind that produces rule-following behaviour.
It is for the above reason that I and some others (Hyland Reference Hyland2011; Hyland Reference Hyland, Sullivan, Cresswell, Ellis, Morgan and Schraube2017; Melidis et al. Reference Melidis, Denham and Hyland2018; Martinez-Lavin et al. Reference Martinez-Lavin, Infante and Lerma2008) believe that functional disorders can be only partially understood in terms of biological – or psychological – theories and that a more helpful theoretical approach comes from the insights gained from artificial intelligence, machine learning, and complexity theory. That is, one should take an instrumentalist rather than a realist approach to the science of functional disorders, and develop theories that explain how particular patterns of events, both biological and psychological, produce the particular pattern of symptoms that are observed. The reason for assuming a mechanism network in contrast to only a symptom network is that covariation of some somatic symptoms of functional disorders cannot be explained in terms of symptom-to-symptom causality – for example, the covariation of diarrhoea and constipation of irritable bowel syndrome. However, there are many similarities between functional disorders and mental illness, so mechanism networks may also apply to both types of illness. Functional disorders and mental disorders are often co-morbid, and distinct diagnostic categories have been challenged for both types of illness (Wessely et al. Reference Wessely, Nimnuan and Sharpe1999).
Mental illness and functional disorders share one important characteristic that has bearing on the issue of reductionism. For both types of illness, a unique pathophysiology has not, as yet, been identified, so diagnosis is based on symptoms. In the case of functional disorders, diagnosis is based on symptoms after exclusion of other biological causes of those symptoms. There are, of course, many biological differences between healthy individuals, on the one hand, and those with mental illness or functional disorders, on the other. For example, both exhibit a tendency (though not found in all cases) for raised pro-inflammatory cytokines, but there is no one-to-one relationship between a particular type of cytokine and a particular symptom. This failure to discover a specific pathophysiology is not for lack of trying, and continues to the present day. Prizes have been offered for a biological diagnostic test for chronic fatigue syndrome.
When something cannot be found but is known to exist, it may be that one is looking in the wrong place. Network theory suggests that the specific pathology of functional disorders – and there must be a different biological basis for each pattern of symptomatology – is not in any local mechanism, but in the strength of connections between the different mechanisms across the network. This possibility, for which there is preliminary support (Melidis et al. Reference Melidis, Denham and Hyland2018), suggests that the specific pathophysiology is of a kind that will not be discovered using biology, and that it would be better to understand functional disorders – and mental illness – as a form of program error in a complex system. Although patients with mental health problems accept psychological interpretations of their illnesses, patients with functional disorders seldom do so (Stone et al. Reference Stone, Wojcik, Durrance, Carson, Lewis, MacKenzie, Warlow and Sharpe2002), but they do find acceptable a narrative based on a program error (Hyland et al. Reference Hyland, Hinton, Hill, Whalley, Jones and Davies2016). Artificial intelligence may prove a better heuristic for the treatment of these troubling conditions than either biology or psychology.
Authors have argued for several years that psychology cannot be reduced to physiology because the emergent properties of the body cannot be deduced from biological theories (Bem & Jong Reference Bem and De Jong2013; Kirsch & Hyland Reference Kirsch and Hyland1987). The target article by Borsboom et al. provides a good rationale for why those emergent properties arise and therefore provides a convincing argument for why reductionism is not possible. However, networks apply not only to mental illness, and the same argument can apply to functional disorders.
The term functional disorder refers to patterns of somatic symptoms with no unique pathophysiology, also known as medically unexplained symptoms (MUS). Common examples of MUS include irritable bowel syndrome, fibromyalgia syndrome, and chronic fatigue syndrome. Less common examples include non-epileptic seizures and functional blindness. Functional disorders respond poorly to pharmacological interventions and, although evidence-based psychological interventions are recommended, outcome is often poor, with some patients rejecting a psychological interpretation.
There are two types of network. First, there are networks where symptoms cause each other. This is the type of network proposed by Borsboom et al., which I will call a symptom network. Second, there are networks where the biological mechanisms that cause the symptoms are themselves part of a network, and which I will call a mechanism network. Symptom networks can occur without mechanism networks, but mechanism networks lead to the expectation that causality between symptoms will also form part of the network structure. Thus, mechanism networks are consistent with symptom networks but not necessarily vice versa.
Mechanism networks can be treated in two ways. First, because the mechanisms are biological, it may be assumed that they can be treated as any other form of biological theory. The mechanism network is simply a more complex form of biological theory than that normally envisaged in medicine, where cascades of chemical events are described in terms of a causal sequence of events (Sturmberg et al. Reference Sturmberg, Bennett, Martin and Picard2017). From this perspective, the application of the network concept is consistent with a reductionist perspective and requires a biological approach to diagnosis and treatment. However, this reductionist approach fails to accommodate the emergent properties of networks. Networks have emergent properties because the behaviour of the network as a whole depends on the strength and rate of change between the different connections among the nodes of the network – not on the properties of the individual nodes or mechanisms. One of the emergent properties of networks is rule-following behaviour, which has been demonstrated elsewhere (Wolfram Reference Wolfram2002). Repetition of complex rules leads to complexity that cannot be predicted from an initial state. Therefore, the initial state of a biological network system cannot predict the outcome of that repetition when the network is of a kind that produces rule-following behaviour.
It is for the above reason that I and some others (Hyland Reference Hyland2011; Hyland Reference Hyland, Sullivan, Cresswell, Ellis, Morgan and Schraube2017; Melidis et al. Reference Melidis, Denham and Hyland2018; Martinez-Lavin et al. Reference Martinez-Lavin, Infante and Lerma2008) believe that functional disorders can be only partially understood in terms of biological – or psychological – theories and that a more helpful theoretical approach comes from the insights gained from artificial intelligence, machine learning, and complexity theory. That is, one should take an instrumentalist rather than a realist approach to the science of functional disorders, and develop theories that explain how particular patterns of events, both biological and psychological, produce the particular pattern of symptoms that are observed. The reason for assuming a mechanism network in contrast to only a symptom network is that covariation of some somatic symptoms of functional disorders cannot be explained in terms of symptom-to-symptom causality – for example, the covariation of diarrhoea and constipation of irritable bowel syndrome. However, there are many similarities between functional disorders and mental illness, so mechanism networks may also apply to both types of illness. Functional disorders and mental disorders are often co-morbid, and distinct diagnostic categories have been challenged for both types of illness (Wessely et al. Reference Wessely, Nimnuan and Sharpe1999).
Mental illness and functional disorders share one important characteristic that has bearing on the issue of reductionism. For both types of illness, a unique pathophysiology has not, as yet, been identified, so diagnosis is based on symptoms. In the case of functional disorders, diagnosis is based on symptoms after exclusion of other biological causes of those symptoms. There are, of course, many biological differences between healthy individuals, on the one hand, and those with mental illness or functional disorders, on the other. For example, both exhibit a tendency (though not found in all cases) for raised pro-inflammatory cytokines, but there is no one-to-one relationship between a particular type of cytokine and a particular symptom. This failure to discover a specific pathophysiology is not for lack of trying, and continues to the present day. Prizes have been offered for a biological diagnostic test for chronic fatigue syndrome.
When something cannot be found but is known to exist, it may be that one is looking in the wrong place. Network theory suggests that the specific pathology of functional disorders – and there must be a different biological basis for each pattern of symptomatology – is not in any local mechanism, but in the strength of connections between the different mechanisms across the network. This possibility, for which there is preliminary support (Melidis et al. Reference Melidis, Denham and Hyland2018), suggests that the specific pathophysiology is of a kind that will not be discovered using biology, and that it would be better to understand functional disorders – and mental illness – as a form of program error in a complex system. Although patients with mental health problems accept psychological interpretations of their illnesses, patients with functional disorders seldom do so (Stone et al. Reference Stone, Wojcik, Durrance, Carson, Lewis, MacKenzie, Warlow and Sharpe2002), but they do find acceptable a narrative based on a program error (Hyland et al. Reference Hyland, Hinton, Hill, Whalley, Jones and Davies2016). Artificial intelligence may prove a better heuristic for the treatment of these troubling conditions than either biology or psychology.