16a
Kristian E. Markon
The breadth of mechanistic explanations covered by Crowell, Kaliush, and Vlisides-Henry (this volume) is striking. The theories and accompanying evidence span a variety of causal mechanisms (biological, environmental, and interactions thereof) and different levels of analysis (from the molecular to the ecological). As a whole, these explanatory accounts have little in common other than being mechanistic and involving variables related to personality disorder. Given this breadth, it’s not surprising that Crowell et al. preface their chapter with a discussion of what, exactly, a mechanism is – with so many different theoretical accounts under the umbrella of mechanistic explanation, it is critical to define its boundaries.
Someone unfamiliar with this area of study might wonder why it’s unclear what a mechanistic theory is, and what other non-mechanistic theories of personality pathology might be considered. In other scientific disciplines, such as physics or chemistry, it might be difficult to identify an explanation that is not mechanistic. The definition of a mechanistic account as outlined by Crowell and colleagues – involving a hierarchically constituted structure whose elements causally increase the probability of some outcome – seems uncontroversial enough, so why is it necessary to clarify?
One reason is that behavioral etiology is complex, with the longstanding, unusual patterns of behavior that constitute personality disorder arguably being especially complex. This complexity leads to the chain of associations from explanans and explanandum being blurred by fuzzy probabilistic associations. Crowell et al. touch on this issue, as they do the related problem of study designs in behavioral research being causally underpowered due to ethical, economic, and other challenges. In short, it is often difficult to conclusively establish causality mechanistically for a variety of reasons; because of this, personality disorder researchers and theorists are often left with non-mechanistic accounts lacking the causality component of the definition outlined by the authors. Within personality disorder research, and the behavioral sciences more broadly, we can often predict better than we can isolate the necessary mediating components of the prediction.
Another reason for ambiguity about the meaning of mechanism, though, is that the boundaries between explanans and explanandum are often unclear. Certainly this is not always the case: MAO-A-related genotypes discussed by Crowell et al. are relatively clearly defined, for example, and are clearly distinguished from physically aggressive behavior per se. Often, however, boundaries between explanatory variables are less clear, either due to conceptual ambiguities in how different types of variables are related, or due to fuzziness in differences between constructs of the same type.
Consider, for example, associations between functional neural disconnectivity and emotion dyregulation (reviewed by Crowell et al., this volume). Is it more appropriate to say disconnectivity between certain neural structures mechanistically explains emotionally dysregulated state, or is it more appropriate to state that such disconnectivity is a physical instantiation of emotionally dysregulated state? In some ways it is a difference that doesn’t matter, as in either case some decomposable physical process is being invoked to predict a mental state. In other ways, though, the difference is critical to questions about what is meant by a mechanism, and what is needed to provide a complete biological account of a psychological process. The relationship between neural state and psychological state involves complex questions about mind–brain mappings that do not necessarily arise in discussing the relationship between other mechanistic explanations, such as the influence of parenting behavior on child psychopathology. In a typical mechanistic explanation, there are antecedent causes and descendent effects. However, in the case of mind–brain relationships at some level the psychological phenomena are coincident with and one and the same as the physical substrates. Can an effect be its cause?
Other types of fuzzy construct boundaries become relevant to mechanistic explanations as well. Weak emergence accounts of personality and personality pathology have been of increasing interest, for example (Baumert et al., 2017). In these accounts, psychological, mentalistic phenomena are not related to one another due to some common etiologic factor, continuous or discrete, internal or external. Rather, they are related through direct influences on one another. In this type of account, for example, borderline splitting directly causes emotional instability or vice versa; the two phenomena are not associated because of some shared process, liability, or disease state. Weak emergence accounts can be thought of as a type of dynamic model in some ways, in that associations between mental structures or states are explained in terms of mutual direct causal relationships between them, as well as with other phenomena such as relationship variables.
Weak emergence accounts do not necessarily involve variables at different levels of scientific analysis, such as mental states and their physical substrates; current accounts often involve mental processes influencing one another, isolated from underlying physical processes. In this sense, mental–physical cause–effect questions are not as directly salient. Weak emergence accounts do, however, still raise questions about whether causes and effects can be distinguished in such paradigms, and whether or not they are mechanistic explanations at all. At some level, for example, one might question whether or not splitting can entirely be distinguished from emotional instability, in that some variation in emotional response is presumably involved in splitting. It might then be asserted that emotional instability causes splitting, but given that emotional response is decomposable into cognitive and other components, might the social cognitive components of splitting not constitute part of a varying emotional response? Are both better thought of as being manifestations of some shared processes or patterns? What is the difference between that and mutual direct causation between cognitive and other components of emotional response? Are “splitting” and “emotional lability” imprecise constructs to use in describing personality pathology? When causes and effects are both inferred mental states, it becomes difficult to delineate boundaries between them, raising questions about whether or not weak emergence accounts are properly thought of as mechanistic per se.
One frequent response to the dilemma of whether or not a given set of constructs is reasonable in a causal explanation is to invoke experimental manipulation (Costantini & Perugini, 2018). If a putative cause can be manipulated and has an effect, according to this rationale, it provides a mechanistic explanation. So, for example, if splitting is manipulated (via randomly assigned psychotherapy for example), and influences emotional instability, it provides an account by which the former causes the latter. Conversely, if emotional instability is manipulated (via pharmacological intervention, for example), and influences splitting, the former causes the latter. If something can be manipulated, boundary problems become moot.
The problem with this reasoning, however, is in the assumption that a manipulation is acting through the pathway that was hypothesized, and not through some other pathway. Specifically, that the independent variable is what is assumed, and that an experimental manipulation operates in the way that is intended. Presumably, unique and shared effects on measured variables can be modeled and estimated – for example, the effect of a therapeutic intervention targeting splitting can be parsed into components actually unique to splitting, and those unique to, as well as shared with, emotional lability. However, the question of unmodeled pathways still looms large in such scenarios. Parsing associations of experimental manipulations into unique and shared components, after all, does little to establish the nature of those unique and shared components, even when there is an experimental manipulation. When a chemical is introduced experimentally into a solution, we have a reasonably good understanding of what is being introduced; in a psychological intervention, in contrast, the active components are relatively unclear and often nonspecific (Baskin, Tierney, Minami, & Wampold, 2003).
The success of experimental designs in the behavioral sciences indicates that it is possible to explore issues of causality and mechanism with mental, psychological variables. With personality and other broadband individual difference variables, however, the effects involved span relatively long timescales, and the scope is relatively broad. This is especially challenging when faced with the sorts of unusual behavior encountered in personality pathology. With personality pathology, behavioral patterns might be so unusual as to be unique to a specific individual, and extreme enough as to call into question the ecological validity of experimentally manipulated analogue variables (for instance, experimentally induced aggression in a laboratory versus severe, cruel physical violence). How to definitively link relatively concrete, specific definable mechanisms to trait-like timescales and broad pervasive patterns is currently unclear. Research indicates that personality can be changed (Roberts et al., 2017), but the precise intra-individual pathways and mechanisms by which this occurs are not always well-established.
Consider the hypothetical path diagram in Figure 16.a.1, for example, where stronger paths are illustrated with solid lines and weaker paths are illustrated with dotted paths. The set of three variables on the right forms a system, and the set on the left forms a system; all three variables on the right affect those on the left. Under certain conditions, one might discuss the set of variables on the right in terms of a unitary construct, and the set on the left in terms of a unitary construct. The paths from right to left, moreover, suggest that, in this scenario, the superordinate construct on the right would be said to cause that on the left. The conundrum, however, is what are the conditions under which this is the case? When can one substitute the system in Figure 16.a.1 with two constructs, one of which is causing the other? As the solid paths approach perfect association, this substitution is probably actually necessary, but what happens as the associations weaken? What about experimental manipulation of the variables on the right?
Figure 16.a.1
Path diagram for hypothetical causal system
As mechanistic accounts of personality pathology are elaborated, it is likely that challenges pertaining to levels of analysis and boundaries between variables will increasingly be encountered. Related research on how to best integrate structural models with experimental and other causally informative designs is almost certainly needed to better characterize the nature of mechanisms involved in personality pathology: establishing some sort of causality is itself challenging, but establishing the nature of those causes is even more challenging. In many ways, these challenges are inevitable, reflecting the nature of scientific progress as theories become more elaborated and are shaped in response to empirical study. However, they also in part reflect challenges unique to the study of psychological variables, especially as they play out over broad timescales and dimensions of human experience and behavior. As progress in the field continues, we will undoubtedly change the constructs of focus, honing them and exploring the causal pathways involved in personality disorder outcomes to better approximate the processes uncovered.
References
Baskin, T. W., Tierney, S. C., Minami, T., & Wampold, B. E. (2003). Establishing specificity in psychotherapy: A meta-analysis of structural equivalence of placebo controls. Journal of Consulting and Clinical Psychology, 71(6), 973–979.
Baumert, A., Schmitt, M., Perugini, M., Johnson, W., Blum, G., Borkenau, P., … Wrzus, C. (2017). Integrating personality structure, personality process, and personality development. European Journal of Personality, 31(5), 503–528.
Costantini, G., & Perugini, M. (2018). A framework for testing causality in personality research: Testing causality in personality research. European Journal of Personality, 32(3), 254–268.
Roberts, B. W., Luo, J., Briley, D. A., Chow, P. I., Su, R., & Hill, P. L. (2017). A systematic review of personality trait change through intervention. Psychological Bulletin, 143(2), 117–141.