Thomas Sharp CPS Final Submission

RUNNING HEAD: Mechanistic science 1
Preprint 8/6/2018, accepted at Clinical Psychological Science pending minor revision
Mechanistic science: A new approach to comprehensive psychopathology research that
relates psychological and biological phenomena
Joel G. Thomas
University of Illinois at Urbana-Champaign
Paul B. Sharp
University of North Carolina at Chapel Hill
Joel G. Thomas, Department of Psychology, University of Illinois at Urbana-Champaign, 603 E.
Daniel St., Champaign, IL 61801
Corresponding Author Email: [email protected]
Tel: +1 608-332-1110
Paul B. Sharp, Department of Psychology and Neuroscience, University of North Carolina at
Chapel Hill, 124 Howell Hall, Chapel Hill, NC 27514
Email: [email protected]
Tel: +1 267-671-7748
Mechanistic science 2
Abstract
Efforts to understand the causes of psychopathology have remained stifled in part because
current practices do not clearly describe how psychological constructs differ from biological
phenomena and how to integrate them in unified explanations. The present article extends recent
work in philosophy of science by proposing a framework called mechanistic science as a
promising way forward. This approach maintains that integrating psychological and biological
phenomena involves demonstrating how psychological functions are implemented in biological
structures. Successful early attempts to advance mechanistic explanations of psychological
phenomena are reviewed, and lessons are derived to show how the framework can be applied to
a range of clinical psychological phenomena including gene by environment findings,
computational models of reward processing in schizophrenia, and self-related processes in
personality pathology. Pursuing a mechanistic approach can ultimately facilitate more productive
and successful collaborations across a range of disciplines.
Keywords: autonomy; causal explanation; levels of analysis; mechanism; psychopathology

Introduction – Why we need to examine the way we explain psychological phenomena
The goals of the scientific enterprise are often communicated implicitly through training
in a particular discipline. Even without a formal course in the philosophy of science, researchers
quickly learn what kinds of scientific questions are considered relevant and interesting to the
field, how research design should inform the exploration of these questions, and what
methodological skill sets are necessary to execute a successful study. An enduring, dominant
approach to explaining psychological phenomena was formalized in the mid-twentieth century
(e.g., Campbell & Fiske, 1959; Cronbach & Meehl, 1955; MacCorquodale & Meehl, 1948),
which set the agenda to elaborate laws between hypothetical psychological constructs and
observable data. Although it has helped generate theoretical advances in psychology, this
approach has fallen short of providing an adequate conceptualization and set of strategies to link
psychological and biological phenomena, and in turn, uncover the causal underpinnings of
phenomena of interest. The goal of this article is to demonstrate how a complementary approach,
called mechanistic science, can make up for this shortcoming.
Integrating psychological and biological phenomena is particularly relevant to scientists
interested in psychopathology, as exemplified by the Research Domain Criteria (RDoC)
initiative of the National Institute of Mental Health (NIMH). One of the imperatives of RDoC is
to “integrate the fundamental genetic, neurobiological, behavioral, environmental, and
experiential” data in order to develop a better understanding of the etiology and treatment of
psychopathology (Cuthbert & Insel, 2013, p. 129). In response to this imperative, various
scholars have both emphasized that psychological phenomena cannot be reduced to nor equated
with biological phenomena (Franklin, Jamieson, Glenn, & Nock, 2015; Hershenberg &
Goldfried, 2015; Kendler, 2005; Miller, 2010; Schwartz, Lilienfeld, Meca, & Sauvigné, 2016),
and have pointed out that popular strategies that attempt link these domains (e.g., the multiple
levels of analysis approach) are inadequate on logical and conceptual grounds (Miller, 2010). If
psychological and biological phenomena are both involved in explaining phenomena in
psychology, how do they each contribute something unique to explanation, and how should we
link these domains without eliminating or giving preference to either camp?
The present article contends that this lack of understanding regarding how to relate
psychology to biology is a major reason for the dearth of causal explanations of
psychopathology. Although this issue remains unresolved, field-wide attention has been
primarily focused on whether findings are reproducible. Demonstrating reproducibility, however,
does not suffice for building strong causal theories. A highly stable correlation between an
intervention and an outcome may do little in explaining why the treatment works, which is
necessary to further advance for whom and under what conditions it works best. In general,
causal theories are vital to improving predictions regarding how psychological processes unfold
under various conditions, and creating more precise interventions on psychological functions that
go awry across various forms of psychopathology.
The present article will demonstrate how an emerging approach termed mechanistic
science provides effective ways to link biological and psychological phenomena, which can in
turn spur the development of causal theories of psychopathology. Developed over the past few
decades out of the study of scientific discovery in fields such as biology and cognitive
neuroscience, mechanistic science maintains that a crucial objective of scientific explanation
should be to move beyond identification and description of phenomena to an explanation of how
particular phenomena are realized in living systems. This objective is achieved by appealing to
mechanisms—structures defined by their component parts, operations, and organization, whose
orchestrated functioning is responsible for a particular phenomenon.
In contrast to typical practices in psychological science, mechanistic science requires that
psychological scientists constrain their conceptions of psychological functions to those that
might plausibly be implemented in living systems. This requires a reorientation in how scientists
postulate theories towards parts, operations, and organizational dynamics that serve to explain
how information processing is realized. Furthermore, identifying how phenomena are realized in
living systems requires thinking in terms of basic learning processes, embodiment, self-
organizing principles, and developmental features of complex organisms.
The framework offered here, although borrowing from ideas in primary philosophical
literature (e.g., Bechtel, 2008; Piccinini & Craver, 2011), is tailored to issues researchers face in
psychology and clinical neuroscience. For instance, the present article proposes a new concept,
units of organization, which unlike the concept of “levels of analysis,” provides a logic and
strategy to parse the complexity of psychological phenomena. Moreover, the framework clarifies
the difference between psychological and biological phenomena and how they should be
integrated through demonstrating implementation. Exemplars from research on psychopathology
will illustrate how scientists provide compelling evidence that proposed psychological processes
are indeed implemented in biological systems.
1. Moving from a descriptive to a causal science: What psychology can learn from biology
When abstract processes, such as “energy production,” can be explained by appealing to
the dynamics of mechanisms, specific predictions can be made regarding how interventions on
such mechanisms will change the state of a given phenomenon. Several decades ago, biologists
went through this transition from describing their phenomena abstractly to explaining how they
arise in the physical world. Weinberg (1985), a biologist writing on the history of his field,
described this transition as a shift from a descriptive to a causal science (Woodward, 2002).
This transition in biology, although occurring without reference to a formal philosophy of
science, occurred because biologists began explaining their phenomena mechanistically. Before
biology appealed to mechanisms, explanations of phenomena did not engender the kind of
enhanced predictability and control that allows for targeted intervention (Weinberg, 1985). For
instance, it was presumed that entities in bacteria must somehow switch between metabolizing
glucose and lactose, yet no explanation was offered to account for this phenomenon. Such an
explanation would enable predictions regarding how a given intervention on the bacterium would
affect the rate of lactose consumption. This changed when the relevant parts, processes, and their
spatiotemporal organization were detailed in the lac-operon model (Jacob & Monod, 1961). With
the aid of this mechanistic explanation, scientists were able to construct counterfactual
predictions regarding how a manipulation in the genetic sequence of the bacterium would change
the conditions under which digestive enzymes for lactose would be produced by appealing to the
specific processes (e.g., repressor binding) that would be fundamentally altered (Richardson &
Stephan, 2007). The experimental confirmation of these predictions validated the model as a
rigorous scientific explanation of the phenomenon.
It is only quite recently, however, that philosophers have attempted to arrive at an
adequate analysis of “mechanisms” and how they are investigated in science (Machamer et al.,
2000). This may explain, in part, why scientists rarely explicitly describe what they mean when
they say they are investigating a mechanism. The few times that mechanisms are defined in the
psychological literature, they are done so quite generally. In addressing psychotherapeutic
change, for instance, Kazdin (2007) defines a mechanism as, “the basis for the effect, i.e., the
processes or events that are responsible for the change; the reasons why the change occurred or
how change came about” (p. 3).
Most mechanisms of relevance to clinical psychological science are therefore yet to be
proposed and refined. Indeed, most existing explanations are described by appealing only to
abstract (i.e., without reference to physical parts) constructs, which hinders the generation of
precise causal hypotheses. Unlike certain medical procedures that are predicated on an
understanding how a system works (e.g., surfactant therapy for infant respiratory distress
syndrome), psychotherapeutic interventions, although effective in many cases, have been shown
to be largely nonspecific to a given pathology (Ahn & Wampold, 2001; Bell, Markus, &
Goodland, 2013; Laska, Gurman, & Wampold, 2014). This results from both not knowing how
the psychological phenomenon is realized in a living system and how the treatment works to
affect the phenomenon.
This lack of progress is due in part to the absence of a meta-theoretical framework that
defines what mechanisms are and how they may be pursued in clinical psychological science.
The remainder of the present article will demonstrate how mechanistic science can address this
problem by providing strategies and concepts necessary to (1) guide the integration of biological
data and theory with psychological data and theory, and (2) causally explain clinical
psychological phenomena.
2. Information processing functions: What psychology has to contribute
In the seventeenth century, Rene Descartes applied the notion of mechanisms to living
organisms, likening the body’s ability to generate motion to the machinery of clocks and mills.
In spite of his advances in understanding the body, he could not conceive of a way to explain
human capacities of language and reasoning through physical processes and therefore attributed
them to a nonphysical substance—mind (Bechtel, 2008).
Although scientists today reject such mind-body dualism (more formally termed
substance dualism), efforts to define and integrate these domains remain stifled. Mechanistic
science comprises formal concepts and strategies that are involved in successful, early attempts
by psychologists and neuroscientists to do what Descartes could not: explain how psychological
phenomena are related to biological phenomena. The core of this proposal begins with the
following characterization of mechanisms:
A mechanism is a structure performing a function in virtue of its component parts,
component operations, and their organization. The orchestrated functioning of the
mechanism is responsible for one or more phenomena (Bechtel & Abrahamsen,
2005).
The key term in the preceding quote is “function.” Psychological phenomena can be thought of
as functions that process information. For instance, vision is the process by which features of
light are represented and transformed into coherent images. For more complex functions like
emotion, the field of psychology is only beginning to formalize what aspects of the environment
are represented and computed on (e.g., Eldar & Niv, 2015).
To understand how information processing psychological functions are related to
biological phenomena, a formal definition of information is useful. Information can be thought
of as the regular effects generated by a given cause, which make it possible to infer the cause
from the effects (Dreske, 1981). Consider the example of a thermostat. A particular thermostat is
a physical system that carries out information processing—namely representing and regulating
ambient temperature. This mechanism can be described in functional terms by appealing to a

negative feedback process. Ambient temperature is compared against the desired temperature,
which results in a deviation that serves to either increase or decrease ambient temperature.
This functional process can be implemented in a physical system that leverages a
bimetallic strip to respond to deviations from a desired temperature. The bimetallic strip re-
presents the temperature because there is a stable relationship between the temperature in the
room and the position of the metallic strip: when the room cools, the strip contracts, and when it
heats, it expands. To regulate the temperature, when the room cools below the set temperature
the strip makes contact with another piece of metal that completes the circuit to turn on the
furnace. This example demonstrates how a process described in functional terms can be realized
within a physical system. This core idea undergirds the enterprise to link psychological
processes, defined in functional terms, with biological structures that implement such functions.
Because functions (e.g., temperature regulation) might be implemented in several
different physical systems (e.g., a bimetallic strip or an electronic thermostat), functions are not
equivalent to structures (Fodor, 1974). For this reason, a mechanistic explanation of a
psychological phenomenon must involve an explanation of its psychological functions as well as
its biological structures, as each provides unique information pertinent to the explanation.
3. The mechanistic framework: Core concepts to integrate psychology and biology
Explaining a complex psychological phenomenon mechanistically involves breaking
down its component processes and parts from both psychological and biological perspectives.
This strategy is called recursive decomposition (Palmer & Kimchi, 1986). For instance, a
psychological process, such as visualizing an object, which transforms a given input (lightwaves)
into an output (a visual percept), can be decomposed into “a number of component informational
events and a flow diagram that specifies the temporal ordering relations among the components”
(Palmer, 1999, p. 74). This breaking down of a process into its component subprocesses enables
scientists to understand how a given operation works. Such decomposition is recursive in the
sense that a given subprocess may be decomposed into simpler subprocesses.
However, this reductionist enterprise of breaking down a process is not sufficient to
explain mechanisms. The function of a single or few parts is very different from the phenomenon
realized by the coordinated spatial and temporal organization of many parts. Thus, it may be
much harder to ascend from an understanding of a part (a single neuron) to a system of many
parts (a neural module; Sporns & Betzel, 2016). The effort to elucidate how a mechanism works
is in many cases a nonlinear endeavor, in which discoveries of parts of a system may influence
how a whole system is defined and understood, and vice versa. Moreover, there is a rough set of
stages mechanistic explanations go through as more of their details are filled in. The process, as
presented by Machamer, Darden, and Craver (2000), begins with a sketch (an initial
decomposition that still has many black boxes), then moves to a schema (more linking of
biological structures to psychological functions), and finally to a bona fide mechanistic model.1
This latter “last stage” is not intended to imply that the endeavor to understand the mechanism
has ended; rather, it signifies a discipline-wide consensus that a given phenomenon has been
adequately explained by a given mechanistic model.
3.1 How to parse complex systems: Units of organization in mechanisms
Many have remarked on the ambiguity of “levels of analysis” in psychological science
(Miller, 2010). In service of multidisciplinary collaboration and conceptual clarity, the current
article proposes the term unit of organization as an alternative that refers to a similar concept
articulated by Bechtel (2008). Importantly, this term bears no relation to traditional usages of
1
An early concept of “filling in” of details through empirical research can be found in Hempel’s notion of an
“explanation sketch” (Donagan, 1957). For a more detailed contemporary discussion of the concept see Darden,
2002.
levels in psychological science. Instead, mechanistic units of organization provide a logic and
strategy to define and relate parts of phenomena to each other in a principled way. Figure 1
depicts a hypothetical mechanistic model of audition, which will serve to anchor the ensuing
discussion of units of organization. The left side of the diagram describes the various functions
of the mechanism, which for audition includes subprocesses like early sensory gating. The right
side describes the physical implementation of these functions.
Figure 1. An example of an incomplete mechanistic model of audition. The phenomenon of audition is realized in
virtue of the parts of the mechanism acting on each other in specific ways. Not depicted here are the arrows relating
parts to each other to denote spatiotemporal organization, which is required in a complete mechanistic model. Both
psychological and biological perspectives are necessary to explain mechanisms realizing information processing
phenomena, and each may inform the other until the field arrives at a coherence across their decompositions.
As depicted in the figure above, units describe various systems within a mechanism,
which may refer to biological (e.g., what the ear does) or psychological (e.g., early gating)
processes. To concretize the concept of units, one can think of zooming into or out of a complex
system. In the diagram, three types of units are presented: the whole auditory system (largest
unit), its major subprocesses (intermediate units), and how each subprocess carries out its
function by appealing to its constituents (gray boxes; the smaller units). In the multi-unit
auditory system, the ear is a sub-unit in which many parts reside, but is also a part of the larger
auditory system. Likewise, the smallest unit may be very different for one mechanism compared
to another, and thus there is no universal organization to mechanisms (Craver & Bechtel, 2007).
As a result, units are considered to be local to a given mechanism. Units of organization refer to
a group of entities in a mechanism that work together through particular interconnections and
interactions to produce a phenomenon.
The last and vitally important distinction is between a mechanism and its environment.
Only parts within mechanisms are ascribed organizational units, which scientists use to parse the
complex (and usually) hierarchical structure of such systems. Phenomena that are extrinsic to a
mechanism, from large-scale (social/cultural environment) to small-scale factors (chemical
gradients impinging on a given mechanism), can cause changes to multiple units of organization
of mechanisms. This is not a passive process, but one in which the organism affects its
environment resulting in cycles of bidirectional causality. For instance, it is necessary to explain
how patterns of exposure to light affect visual processing mechanisms to account for the
ontogeny of visual acuity (Boothe, Dobson, & Teller, 1985). Similarly, many factors have been
shown to promote, scaffold, or impinge on the development of mechanisms realizing emotional
processing (Hart & Rubia, 2012; Sheridan & McLaughlin, 2014).
3.2 Adopting multiple perspectives to elucidate mechanisms
Whereas units of organization parse the multilayered dynamics of mechanisms, the
present article proposes three perspectives that serve to define the essential features of a
mechanistic explanation. First are the two perspectives that describe the activity within a
mechanism. The psychological perspective delineates how information is processed, and the
biological perspective demonstrates how such functions are implemented in biological systems.
These perspectives are influenced by Marr (1982), who referred to them as the algorithmic and
implementational levels of analysis; however, the psychological perspective here also
emphasizes the role phenomenological experience can play in explaining phenomena-of-interest.
Explanations of mechanisms in living systems are also advanced by taking an environmental
perspective. Broadly, the environmental perspective seeks to characterize how a mechanism is
affected by its environment. Taking into account, for instance, how the environment imposes
challenges on an organism can help to re-conceptualize the phenomenon to-be-explained, as was
the case for vision in ecological theories of perception (Bechtel, 2008).
The environmental perspective also delineates conditions under which mechanisms
change their functioning. For instance, under certain environmental conditions, the mechanism
may cease to function properly, as is the case when parts of the brain are exposed to intrauterine
pathogens. The need to maintain stability under varying environmental influences is a universal
challenge for all mechanisms of living systems. Some have referred to this capacity as self-
organizing systems resisting entropy (Bechtel, 2008).
The (1) three perspectives (environmental, psychological, biological), (2) units of
organization, and (3) strategy of recursive decomposition, comprise the foundational concepts
employed by mechanistic science. The goal of mechanistic science is to arrive at mechanistic
models of phenomena that engender counterfactual predictions. To do so, phenomena can be
recursively decomposed from both biological and psychological perspectives. Throughout this
endeavor, various parts and operations of the mechanism may be identified or revised, and the
conceptualization of the phenomenon itself may be redefined (Poldrack & Yarkoni, 2016). The
success of these strategies and concepts in elucidating psychological mechanisms can be
illustrated in the multidisciplinary approach needed to understand vision.
4. Mechanistic explanation in psychology: Vision as an exemplar

How can visual processing be decomposed into coordinated informational processes, how
is it realized in a biological system, and how does the environment constrain its functioning? The
psychological perspective delineates how information is processed, and the biological
perspective demonstrates how such functions are implemented in biological systems. Moving
back and forth across these perspectives is essential to elucidating how a mechanism works. This
section will describe the multidisciplinary effort to discover the mechanisms realizing vision,
perhaps the most successful endeavor yet in elucidating a mechanism giving rise to a
psychological phenomenon.
4.1 A rough mechanistic sketch of vision
Important early efforts to decompose the visual system used clever experiments to ascribe
where in visual space and to what kinds of stimuli the cells in a given brain region would
respond. Seminal findings by Hubel and Weisel were derived from experimenting on cat primary
visual cortex in the 1960’s and 70’s through recording single cell activity of cortical regions as
cats were exposed to various simple stimuli (Enroth-Cugell & Robson, 1966; Kaplan & Shapley,
1986; Livingstone & Hubel, 1984). These discovery-oriented experiments allowed Hubel and
Wiesel to inductively associate psychological functions with neuronal regions, such as detection
of a given stimulus orientation in a specific location in space in simple V1 cells, or detection of
stimulus orientation regardless of its location in space for more complex V1 cells (Reid & Usrey,
2012). This productive research enterprise demonstrates how pursuing a biological perspective
can inform, and in some cases, begin the effort to define and explain various psychological
functions.
Moreover, the initial sketch that resulted from such endeavors (Van Essen & Gallant,
1994) yielded counterfactual predictions regarding which functions would be disabled if a given
region were to be damaged. For instance, with broad damage to inferotemporal cortex, one could
expect an inability to distinguish faces. This condition, called prosopagnosia, occurred in stroke
patients with damage to this brain region, thus supporting such a counterfactual claim (Tarr &
Gauthier, 2000).
The decomposition (both psychologically and biologically) of visual processing
according to this initial sketch was quite crude. For instance, it remained open to inquiry how
each subprocess carried out its function, and how such processes were temporally organized. To
begin to advance the mechanistic sketch of vision beyond these seminal findings would require
the collaboration of neuroscientists, computer scientists, and psychologists.
4.2 Recasting what visual processing does: The environmental perspective
Theoretical advances in psychology and neuroscience contributed to reconceptualizing
what problem the visual system must solve for an organism. An early contribution came from
Gibson, an experimental psychologist, who proposed the ecological theory of perception (1979).
This theory argued that because organisms are active agents interacting with their environments,
sensory input and processing are related to action, movement, and an organism’s goals. Visual
processing thus must facilitate an organism’s need to respond adaptively to rapidly changing
environments and use visual information to exploit the world for essential resources. The need to
conceive of the organism as situated in a social context bridged contributions from various fields.
For instance, Bechtel (2008) describes how this theoretical development in psychology
was related to independent developments in neuroscience regarding the organization of visual
processing that were pioneered in part by Churchland, Ramachandran, and Senjowski (1994):
One aspect of an organism performing ongoing activities, noted in Churchland et
al.’ s characterization of interactive vision, is that it forms expectations of what it

will see in the future. These predictions serve not only to facilitate selection
between possible movements but also provide a powerful basis for learning. (p.
235)
Such theory was informed by various behavioral studies in psychology. For example, in
Biederman’s (1981) work efficient categorization of objects (measured with reaction time) was
inversely related to where in visual space such objects tend to be (e.g., a mailbox on top of a
house took longer to process than a mailbox on top of sidewalk). These studies established that
visual processing was not a static process, but even for simple recognition tasks and early
processing stages, is likely influenced by prior beliefs and predictions regarding how the world
ought to be.
4.3 How the psychological perspective constrained the biological perspective
Although Gibson (1979) and Churchland et al. (1994) were seminal in advancing how
visual perception is organized, further advances in information-processing models were needed
to more formally test how biological mechanisms realize visual processing. A recent, successful
enterprise that has pursued this aim is called predictive coding (PC). Broadly, PC suggests that
visual perception and other perceptual modalities are realized by the brain predicting incoming
sensory signals, and iteratively refining either (1) its top-down predictions or (2) actions taken to
change sensory input via the mismatch between the prediction and the sensory signal (Friston,
2010; Rao & Ballard, 1999).
From an evolutionary vantage point, PC seems advantageous relative to other forms of
information processing. Without predictions of perceptual experience, our sensory systems
would need to continuously decode incoming sensory data from the “bottom-up,” which may
unnecessarily tax an already metabolically expensive mechanism. Conversely, if sensory systems

do implement PC, predictions that match stable sensory input (e.g., the tree is stationary) do not
burden the visual processing hierarchy because those aspects of the visual scene do not require
cognitive updating. Importantly, this development in how the visual system processes sensory
information demonstrates how an advance in the psychological perspective constrained the way
in which neurobiology would be studied.
For instance, Spratling (2010) proposed a variant of PC, called the predictive coding /
biased competition (PC/BC) model to explain various functions of cell types within V1. This
model proposes how edge and orientation direction functions are realized by the temporal
organization of artificial neurons modeled computationally. Simulations involved exposing the
model to the same images (represented as pixels in 2-D) used in Hubel and Weisel’s experiments
with cat visual cortex. The results of simulations (see bottom row, figure 2) showed remarkable
similarity between how the model behaved with how V1 cells behaved in Hubel and Wiesel’s
experiments (see top row, figure 2).
Given that the PC/BC model was capable of predicting to a high degree of accuracy the
shape of the data, one can be quite confident that it has many things correct about how this part
of the visual system works. By contrast, when one tests a theory by merely hypothesizing that
two variables will be correlated to an unspecified degree, or that two groups will differ in their
means (two commonly used versions of null-hypothesis testing in psychology), one should not
feel confident that one’s particular theory has been greatly supported. More specifically, this is
because the test itself is not risky and many other explanations could predict the same outcome
(Meehl, 1967; 1990). The point here is that the PC/BC model is an exemplar of a strong
substantive theory because of the specificity of its predictions.
Figure 2. Top row: V1 recordings; Bottom row: V1 computational simulations. Identical stimuli were used for both
cellular recordings and computer simulations, which are in the gray boxes below the bottom row.
4.4 From the psychological to the biological perspective: How do neurons implement PC/BC?
In the previous example, computational neuroscience pulled from theory and empirical
findings across psychology and neuroscience that resulted in a strong substantive theory of how
simple functions of the visual system may be realized. However, it remains a question if and how
real neurons implement the PC/BC model. If it is found that the PC/BC model is not biologically
plausible, it will need to be refined. It is this iterative crosstalk and mutual refining between
psychological and biological perspectives that is a pillar of explicating mechanisms that realize
phenomena-of-interest.
Presently, the psychological description of PC/BC and closely related variants of
predictive coding for vision are still in their infancy. However, there is an increasing body of
experimental work which has shown that such functions seem to be carried out by basic neuronal
units in visual and other sensory cortices (Bastos et al., 2012). The collaboration of scientists
across the many fields pursuing various perspectives of the visual mechanism has resulted in a
comprehensive sketch of the visual system. The following section will demonstrate how the
elaboration of a mechanistic account of visual processing is similar to recent efforts aimed at
elucidating phenomena involved in psychopathology.
5. How the mechanistic framework similarly applies to psychopathology research: Negative
symptoms in schizophrenia
Efforts in clinical psychology to elucidate mechanisms are beginning to take root,
perhaps most prominently in the enterprise of computational psychiatry (e.g., Montague et al.,
2012; Wang & Krystal, 2014). The central imperative of computational psychiatry is to integrate
biological and psychological data in order to seek, “computational algorithms and generalizable
principles that are reflected in the observed biological signals” (Wang & Krystal, p. 639). The
present section reviews how prominent negative symptoms of schizophrenia – anhedonia and
avolition – have begun to be explained mechanistically.
5.1 The start: Basic biological investigations of reward processing
The history of attempting to explain prominent negative symptoms of schizophrenia, such
as anhedonia and avolition, began with basic efforts to understand reward processing (disrupted
in anhedonia) and motivation (disrupted in avolition). The earliest insights came from biological
investigations of reward processing in model organisms. The first development occurred when
scientists found evidence for the “anhedonia hypothesis” (Fouriezos, Hansson, & Wise, 1978),
which claims that the absence of dopamine results in anhedonic behaviors and subjective
experiences. Evidence for the theory was accrued by direct manipulation of dopaminergic-
producing regions as well as psychopharmacological dopaminergic antagonists (Berridge &
Robinson, 1998). Although this theory pinpointed the midbrain dopamine system as being
crucially involved in reward processing, it did not explain how this occurs beyond the overly
simplistic notion that increases in dopamine signaling realize feelings of pleasure, and lack of
dopamine realize anhedonia.
5.2 Demonstrating implementation: Reward prediction errors encoded in dopamine neurons
Further work would be needed to decompose the processes that together realize increases
or decreases in pleasure. Progress occurred when scientists began applying advances in modeling
how organisms learn and make decisions in order to garner reward or avoid punishment (i.e.,
“reinforcement learning”; Sutton & Barto, 1998). The impetus for such an application was the
finding that the association between receiving rewards and dopamine activity would attenuate
when rewards were delivered in a predictable fashion (Berridge & Robinson, 1998). From this
observation it was hypothesized that dopamine encodes reward prediction errors (i.e., the
deviation between internal estimates of expected reward and actual received reward). Indeed, this
theory was corroborated by demonstrating that the time-course of reward prediction errors in the
psychological model mirrored the actual neural time-course of dopaminergic neuron firing in the
striatum (Schultz, Dayan, & Montague, 1997).2 It was then concluded that positive prediction
errors (not expecting a reward and getting a reward) are crucial to the experience of pleasure.
Although the preceding breakthrough began the effort to explain reward processing, the
model lacked comprehensiveness in terms of explaining the dynamics of dopamine signaling,
and the psychological subprocesses it implements in complex behaviors and dysfunctions
characteristic of schizophrenia. Most prominently, dopamine deficits were shown to be
associated with a lack of motivation (resulting in avolition in schizophrenia), which at the time
was not explained by the existing model of reward learning. It remained to be explained how
aberrations in prediction error encoding might lead to both anhedonia and avolition.
5.3: Combining biological and psychological perspectives to explain anhedonia and avolition in
schizophrenia
Recently, Collins and Frank (2014) proposed the opponent-actor-learning (OpAL) model,
which integrates insights from basic biological work on the effects of dopamine on motivation
(Berridge & Robinson, 1998) with the successful efforts of computational psychiatry detailed
above on reward learning (Montague et al., 2012). In a unified mechanistic accounts of these
phenomena, it was demonstrated that the model is capable of predicting motivated effort effects
found in incentive-probing tasks (e.g., Aberman & Salamone, 1999), optogenetic manipulation
of dopaminergic signaling at the phase of making decisions (Tai et al., 2012), and many other
2
Formally termed the temporal difference algorithm, it comprises the following general structure: values (total
estimated future reward) are computed for state-action pairs (e.g., choosing one of three possible choices to obtain
reward) through iterative learning, in which such values are updated according to a weighted reward prediction error
(deviation between the expected reward and experienced reward).
phenomena involved in reward processing. Scientists have recently applied the OpAL model to
investigate negative symptoms in schizophrenia (Maia & Frank, 2017). Various subprocesses of
the mechanistic model have been mapped onto the relevant implementing structures in the brain,
including parts of the basal ganglia, thalamus, and cortex. Through simulations of the
information processing functions involving dopamine, it was shown that reduced phasic
dopaminergic responses to unexpected stimuli and increases in spontaneous dopamine could help
explain reduced reward learning, slow psychomotor responding, decreased activation in the
ventral striatum during reward anticipation, and reduced effort to seek rewards (Maia & Frank,
2017).
5.4 Similar efforts have begun for a range of psychopathological phenomena
The example above in which symptoms of schizophrenia were partially explained
through formalizing models of internal information processing and demonstrating how they are
implemented in various brain structures has begun to occur for a range of psychopathology.
These include compelling models of mood instability (Eldar & Niv, 2015), depression (Huys et
al., 2013; Bishop & Gagne, 2018), anxiety (Charpentier et al., 2017), and autism (Lee, Lee, &
Kim, 2017) among others. Many of these early efforts are rudimentary and rely heavily on
similar algorithms (e.g., reinforcement learning), but still serve as prominent foundations on
which to build mechanistic models.
5.5 How a mechanistic approach relates to computational psychiatry efforts
The mechanistic framework offered here should not be thought of as entirely distinct
from the computational psychiatry enterprise, but rather as a way to comprehensively highlight
its strengths and place its successes within a broader philosophical context. Moreover, the
framework provides concepts and language that can help integrate a range of clinical
psychological research with the computational psychiatry enterprise. The environmental
perspective, for instance, highlights promising areas for such integration. This perspective
includes aspects of psychopathology such as ontogeny, etiology, and how placing the organism
in various environments serves to constrain the range of functioning of a given mechanism. For
instance, taking an environmental perspective has helped to understand basic reward processing.
Indeed, Reynolds and Berridge (2008) have shown that dopaminergic signaling in the nucleus
accumbens rapidly alters its function depending on the environment in which rats are situated.3
Future work should endeavor to demonstrate how varying contexts affect the algorithms and
neural implementation of reward processing, which may be especially relevant for demonstrating
how certain manifestations of psychopathology arise.
6. Applying the mechanistic framework to personality pathology: The challenge of
situating smaller units of organization within larger units
A major challenge for psychopathology research is to arrive at causal theories that show
how information-processing functions like prediction error encoding are situated within and
constrained by larger systems of the mind/brain. To demonstrate how thinking in terms of parts,
operations, and units of organization can further an understanding of the whole organism and its
development over time, this section will explore the beginnings of a mechanistic explanation of a
personality schema associated with early childhood abandonment.
Explaining certain forms of psychopathology requires understanding the dynamics of the
self. A starting point for gaining traction on the self involves the observation all living organisms
are autonomous beings that recruit matter and energy to maintain an internal homeostatic
3
Although there is typically a rostro-caudal organization in which cellular stimulation will result in appetitive,
neutral, and aversive behaviors, this organization shifts in rich and deprived environments. In rich environments, the
region predominately generates appetitive behaviors, whereas in the deprived environment, there is a larger
representation of aversive-generating signaling and less appetitive-generating signaling.
environment, sometimes referred to as the autonomous self (Bechtel, 2008). As autonomous
beings, much of the activity of living organisms is typically organized around perceptual
processes that specify where the organism is in its environment, its opportunities for action, and
its roles in relation to others (e.g., predator, prey, family, friend).
While human selves also possess these attributes, they can be distinguished from simpler
organisms in the degree to which they have developed additional mechanisms critical for
behavior we consider characteristic of free responsible agents. For instance, human selves have a
highly-developed memory system that enables them to recall important autobiographical events
regarding the past and to project themselves into the future. In addition, human self-awareness
allows individuals to conceptualize their own existence as separate beings and enables them to
consider choices for active engagement with their environment (Neisser, 1988). In this sense, the
human self can be considered a highly differentiated and rich regulative system capable of
directing a person’s choices through moment-to-moment valuing of opportunities for acting in
the world.
It is this nexus of adaptation of human selves to their changing environment that is
implicated in many forms of psychopathology. For instance, early in development a child’s self
must adapt to the demands of a caregiving environment (i.e., an attachment experience).
Although the individual behaves in ways that are adaptive at the time, in later life contexts these
behaviors may become maladaptive and result in signs and symptoms of adult psychopathology.
In the case of personality pathology, common symptoms could include interpersonal difficulties
such as “efforts to avoid real or imagined abandonment,” “unstable relationships,” and “affect
instability.” The following mechanistic model will seek to provide a potential causal explanation
for such phenomena.

Take the hypothetical case of Julie, an individual who experiences early childhood
abandonment. As depicted in figure 3 below, Julie’s self-structure can be characterized as the
series of representations of herself that she has stored in long-term memory. From a mechanistic
perspective, this self-structure can be further decomposed into various parts, some of which are
particularly relevant for explaining the effects of her experience of early abandonment on her
functioning as an adult.
Figure 3. The initial functional decomposition of self-structure into various schema, as well as the components of
the schema for abandonment.
Julie was raised in a home with a father who had to manage the death of his spouse when she
was only two years old. Due to the chaos surrounding the loss of her mother, Julie’s memory
system records this loss as the experience of abandonment. In line with attachment theory, Julie’s
need for a stable, accessible, and attentive caretaker likely went unmet. If repeated attempts for
security were thwarted, Julie forms an abandonment schema, a representation of herself as
unlovable and unworthy of care, attention, and protection.
In a mechanistic framework, the abandonment schema can be functionally decomposed
into its various parts. First, due to a lack of emotional support and care early on in life, Julie has
encoded an expectation that care and attention will be withheld or withdrawn in close
relationships. Second, a feeling of despair and depression is represented in long-term memory

that is activated when she experiences rejection later on in life. This is a phenomenal experience
associated with her early memory of hopelessness and isolation. Finally, Julie will lack a secure
base event-script, which is a sequence of events between child and caretaker that includes: (1)
constructively interacting with a caretaker, (2) being interrupted by an event or actor, (3) a bid
for help and (4) the bid being met with support and emotional soothing (Waters & Waters, 2006).
The example demonstrates that the mechanistic perspective can be pursued even without
mapping psychological functions to biological structures. Of course, eventually the functional
decomposition will be refined by the delineation of constraints that biological structures impose
on its implementation (and vice versa). However, even before this linking occurs, the proposed
mechanism motivates an experimentally verifiable counterfactual prediction: if Julie does not
exhibit a well-consolidated secure base event-script, then she will display a pronounced feeling
of despair and depression when experiencing interpersonal rejection as an adult.
Such a model expands on prior research that has demonstrated that a self-report measure
of the abandonment schema “explains” a large portion of the variance in adult symptoms of
personality pathology (Bach, Lee, Mortensen, & Simonsen, 2015). Indeed, the mechanistic
model contains additional specific predictions regarding symptoms such as “efforts to avoid real
or imagined abandonment,” “unstable relationships,” and “affect instability.” For instance,
triggering of the schema would likely occur in moments of perceived interpersonal rejection.
This initial sketch paves the way for modeling how the dynamics of symptom intensity would
likely vary over time, thus moving beyond Bach et al. (2015) which does not offer an internal
mechanism that might account for their finding. In addition, increased specificity regarding parts
and operations can lead to more direct and precise intervention. For instance, if this model were
validated, one might (1) assess the presence of a secure base script, (2) challenge expectations of
abandonment in the therapeutic encounter, and (3) use exposure and provision of a corrective
emotional experience to address the feeling of despair and depression associated with early
abandonment.
The mechanistic sketch above therefore moves beyond traditional approaches that rely
solely on statistical analysis of variance to understand personality pathology because: (1) the
testing of parts and operations extends the goal of explanation to how behavioral output of the
organism can be accounted for in terms of internal processes, (2) the delineation of these internal
processes provides a prototype or model system that can be investigated in real time (e.g., by
varying environmental rejection conditions), (3) further experimentation can elucidate the
manner in which various components of the schema are realized in mental machinery (e.g.,
memory encoding or retrieval), and (4) the proposal of an information processing account of
self-structure can begin the difficult endeavor of demonstrating how such self-related processes
are realized by biological systems. Such empirical validation of a decomposition of an
abandonment schema represents the kind of substantive theory testing that may eventually lead
to more appropriate and precise clinical intervention in complex distress syndromes such as
borderline personality disorder.
7. Why perspectives and units of organization should replace “levels” talk
The exemplars above illustrate how thinking in terms of psychological, biological, and
environmental perspectives and units of organization can guide the practical activities of
psychopathology researchers. The following section will elaborate on the specific reasons that
these concepts are preferable to various formulations of “levels” frameworks that are typically
used to bridge explanations from different disciplines.

There are two main reasons why employing levels of analysis results in underspecified
relationships between psychological and biological phenomena. First, “levels” are often
ambiguously defined. For instance, Cicchetti (2010) states:
I use the term level of analysis to refer to the different scales into which behavior
or the brain can be represented. The ultimate criterion of what constitutes a level
of organization is its utility in elucidating the understanding of a particular
biological or psychological phenomenon. (p. 33)
Based on this definition, levels might be used to separate disciplines (biological vs.
psychological; Engel, 1981), denote the size of phenomena (molecular vs. cellular), or
distinguish the kind of perspective one might adopt to understand a phenomenon (algorithmic or
implementational; Marr, 1982). Second, the efforts to “bridge” these levels of analysis are
similarly ambiguous. Usually, the “bridging” is done by merely restating statistical terms, such
as “mediate” or “interact.” Figure 4 illustrates the conceptual problems that result from these two
related concerns.
Depression at differing “levels of analysis” versus explained in a mechanistic framework

Figure 4. On the left, “levels” are depicted with uniform bidirectional arrows (based on a figure from Ilardi, Rand,
& Karowski, 2007). On the right, the substantive differences between these relationships are clarified in mechanistic
terms. (A) Patterns of interaction between the individual and others (social and behavioral levels) unfold through
cycles of bidirectional causality. (B) These patterns of interaction with the environment influence the activity of the
internal mechanism of a depressogenic schema (mental level) which can be decomposed into various functional
units of organization. (C) Fully explaining the internal mechanism eventually involves identifying how these
psychological processes are realized in biological structures (neural, neurochemical, and molecular levels). (D)
Units of organization capture the notion of constraint from a biological and psychological (not shown) perspective.
The diagram on the left reproduces a common representation of the relationship between
levels with bidirectional arrows that are often interpreted as indicating a hierarchy with causal
relations between levels. On the right, these “levels” are depicted in terms of perspectives and
units of organization. Although the arrows in the diagram on the left are often investigated in
terms of statistical dependencies such as correlations, the diagram on the right clarifies that they
may indicate a variety of different relationships in living systems: (A) bidirectional
environmental causal relations, (B) implementational relations, such as how information
processing is realized in biological structures, (C) internal causal relations, which explain how
separate processes act on each other within a mechanism, or (D) whole systems constraining the
activity of its parts (Borsboom, Mellenbergh, & van Heerden, 2004). The discussion below
illustrates how the concepts proposed in the present article capture these differences in a way that
provides a common language for multidisciplinary collaboration as psychopathology researchers
seek to elucidate phenomena of interest.
7.1 Perspectives language prevents misattribution of causal relations between psychological and
biological phenomena
The same way one would not say software interacts with hardware, one would not say
psychological phenomena act on biological phenomena. For this reason, mechanistic science
regards causal relations across psychological and biological perspectives as logically untenable.
As was shown in the examples above, the difficult work of integrating psychological and
biological phenomena in unified models involves decomposing a psychological function into its
component processes, and demonstrating how they are implemented in biological structures.
Causal claims relating biology and psychology are thus considered logically flawed and non-
explanatory. Mechanistic science by contrast regards the concept of implementation as the way
to think about how psychological and biological perspectives are related, which when
empirically demonstrated leads to comprehensive explanations.
Causal claims relating psychological to biological phenomena such as “depletion in
synaptic serotonin causes depressed mood” are considered to be as problematic as the converse
claim, “depressed mood causes depletion in synaptic serotonin.” Mechanistic models instead
explicate causal relations within both psychological and biological perspectives, resulting in two
non-competing accounts. The greater the coherence in predictions across these perspectives, the
greater the evidence there is for implementation. Practically, this involves comparing the
predictions made by biological models with psychological models, typically using variants of
Bayesian model comparison (e.g., Daw, 2011). The diagram on the right in Figure 4 makes clear
that a depressogenic schema cannot be caused by neurological, neurochemical, or molecular
events. Instead, researchers must decompose the schema into its various informational processes
and demonstrate how they are realized in the relevant biological structures.
This of course does not mean that conceptually psychological functions are identical to
biological structures. In principle, there could be a psychological function realized by more than
one biological or non-biological mechanism (Fodor, 1974; Miller, 2010). Among the variants of
reductionism debated in the field, mechanistic science thus rejects what has been called
“eliminative reductionism,” or the contention that biological phenomena underlie or can replace
psychological phenomena. Experimental investigation clarifies the relationship between

psychological and neural phenomena in a mechanistic framework. Depending on the
phenomenon under investigation, it may be that a psychological process has only one or many
implementations in neural systems. The present article proposes that this autonomy of
psychological functions from biological implementations can be characterized as autonomy of
perspective.
Autonomy of perspective also refers to the practices and assumptions that differ across
psychological and biological perspectives. First, scientists studying psychological functions may
use different observational and experimental techniques than those focused on biological
implementations (e.g., psychometrics versus lesion studies). Second, the conceptual vocabulary
adopted by each scientific enterprise may be considered irreducible to each other (e.g.,
hallucinations versus neural networks) (Piccinini & Craver, 2011). Third, the mechanistic
framework makes room for subjective awareness, meaning, and parts of folk psychology that are
particularly relevant in fields such as psychopathology research where phenomenal experience
(i.e., “private” data) and communications of beliefs and desires appear central to understanding
particular phenomena (Bechtel & Abrahamsen, 1993; Bolton & Hill, 1996; Kapur, 2003). For
instance, a strategy termed “heterophenomenology” (Dennett, 1993) assesses subjective
experience by combining self-report with other available evidence such as observation by other
humans (e.g., thin-slicing procedures). Integrating subjective experience and meaning into
mechanistic explanations may be necessary for generative theory construction. In modeling
psychotherapeutic change, for instance, it is unclear that a rigorous explanation of the
phenomenon can be arrived at without appealing to subjective experience.
Adopting this view on the relation between psychological and biological phenomena
facilitates productive interdisciplinary work. Authors that argue for a bidirectional, causal
relationship between mind and brain (e.g., Kendler, 2005) fail to characterize how a
psychological process, either described in informational terms or referring to phenomenological
experiences, might be acted on or interact with biological structures. Indeed, endeavors to
demonstrate such causality have led to confusion regarding how to appropriately link such
phenomena (Miller, 2010).
For instance, it is often suggested that “epigenetics, reminds us that linkages among these
levels are typically bidirectional” (Lilienfeld, 2017, p. 7). Such a seemingly innocuous statement
suffers from two missteps. First, it confuses the distinction between psychological and biological
phenomena with the distinction between internal mechanisms and external environmental
phenomena. Epigenetics involves the environment (e.g., parental stressors) influencing internal
mechanisms (e.g., genetic expression), and vice versa. The environmental variable should not be
conceived of as psychological, and the internal mechanism as biological, given that both
psychological (information processing) and biological phenomena (parts and operations in the
brain and body) describe internal mechanisms from different perspectives. Moreover, both
psychological and biological phenomena are bidirectionally causally affected by aspects of the
environment.
Second, it implies that biology and psychology are acting on each other causally. As
stated above, causal relations involve distinct processes, and so the unintended result of using
epigenetics in this way implies that psychological and biological phenomena are separate and in
some unspecified way acting on each other. Of course, those who think of epigenetics in this way
most likely do not subscribe to substance dualism, a doctrine universally rejected by modern
science (Lilienfeld, 2017). That said, we use this example to illustrate the subtle but ultimately
deleterious way in which the typical levels of analysis approach results in untenable conceptions
of how to relate psychological and biological phenomena. The mechanistic approach avoids
these problematic claims by focusing on the problem of demonstrating implementation, which
requires the collaboration of scientists from both psychological and biological perspectives.
7.2 Units of organization facilitate the delineation of causal and constraint relations within a
mechanism
Although causality is not the appropriate way to think about integrating or relating
psychology to biology, explaining causal relations is crucial to explicating a mechanism. Within
a psychological or biological perspective of a common mechanism causes are proposed in terms
of units of organization: a group of entities that work together through particular
interconnections and interactions to produce a phenomenon. A biological example would be how
structures act on the generation, degradation, and transport of serotonin in the synaptic cleft,
whereas a psychological example would delineate how schema activation leads to changes in
mood generation.
When working parts are separate (either temporally or spatiotemporally), they may act
causally on each other. Units of organization can also be situated hierarchically, in which a
larger unit (e.g., visual processing pathway) contains a smaller unit (e.g., general feature
detection). Relating larger to smaller units cannot be causal, because they are not separate.
Rather, it is important to demonstrate how larger units constrain the activity of smaller units
(e.g., through positive and negative feedback processes). Indeed, this principle of constraint is
part of most hierarchical control systems such as the human brain and body (Bechtel, 2016).
Typically, scientists investigate how the state of the whole system constrains the possible
activities taken by parts of the system. For instance, the state of the cell cycle constrains the
range of possible genes that are expressed (Bechtel, 2016). In visual processing, the goals of the
whole organism (e.g., hunger) bias how simpler functions within the system respond to incoming
visual input (e.g., by prioritizing objects that appear like food).
Using units of organization can also help clarify major findings of interest to
psychopathology researchers, such as gene-by-environment findings. Consider a study in which
exposure to severe childhood maltreatment is interpreted as “interacting” with the short allele
promoter polymorphism of the gene encoding the serotonin transporter (5-HTT) to increase risk
for major depression (Andersen & Teicher, 2008). Such correlational evidence is indicative of an
important effect, but fails to explain how both the environment and 5-HTT affect the mechanism
realizing depressed mood. A mechanistic explanation must include how the protein product of
short allele 5-HTT may act on various other cellular structures so as to alter the neural system
implementing mood regulation. Moreover, the explanation must demonstrate how traumatic
events affect mechanisms realizing mood regulation at multiple units of organization. Advancing
G-E findings beyond their descriptive nature is crucial, given that the reliability of G-E
associations such as the one used in the example above have been called into question (Duncan,
Pollastri, & Smoller, 2014). Mechanistic explanations may reveal that the variability in the size
of G-E effects are due to differences in the implementation of mood regulation across
individuals, including variation in learning history, genes impacting multiple parts of the system
beyond neurotransmitter production, and likely many other factors.
Thinking in terms of units of organization therefore clarifies what researchers still have
left to explain. Take for instance the multiple levels of analysis approach to study the etiology of
impulsivity in children and adolescents proposed by Beauchaine and colleagues. They developed
a model suggesting, “inherited genetic (dopamine related) and psychological (impulsiveness)
vulnerabilities interact with environmental factors (high risk or protective) to result in a range of
behavioral outcomes (conduct disorder and ADHD, or ADHD only) across the lifespan”
(Franklin et al., 2015, p. 285). The problematic, underspecified term in the preceding quote is
“interact.” Since a gene cannot physically act on impulsiveness, the term “interact” is
misleading. If certain transcribed genes produce proteins that alter dopamine in the brain in a
manner that is related to impulsivity, this is the mechanism that must be elucidated. A
restatement of the claim above within the mechanistic framework would therefore read,
“Endogenous factors such as genes that code for dopamine activity appear to alter mechanisms
responsible for impulsivity in such a manner that is yet to be explained.” Shifting language away
from ambiguous terms such as “interact” brings attention to the further work that must be done to
arrive at more rigorous and comprehensive explanations.
7.3 Illustrating autonomy of perspective and autonomy of organization in vision
Returning to the mechanistic account of visual processing described earlier, one would
not say that patterns of light “interact” with the retina in order to cause visual percepts. Rather,
the specific actions by which light affects sensory receptors are explained (i.e., the specific
causal relations) in mechanistic models. The number of units depends on how many iterations of
recursive decomposition one carries out on a complex system (see figure 5). With one iteration,
there is the unit of the whole system (e.g., how the parts of the entire visual system are
organized) and the unit of the parts (e.g., how V1 works). Another iteration of decomposition
explains the parts of the parts (e.g., the subprocesses of V1, such as lateral inhibition patterns).
Units of organization thus describe the working parts of a mechanism at various scales of
complexity, and importantly, from both psychological and biological perspectives.

Figure 5. An incomplete mechanistic model of vision based on the discussion in section 4. Autonomy of
organization indexes the notion that the whole is more than the sum of the parts (e.g., general feature detection
involves interconnections and interactions between the PC/BC algorithm and other subprocesses). Autonomy of
perspective indexes the difference between decomposing a phenomenon in terms of psychological functions versus
the biological structures that realize these information-processing functions.
To summarize, autonomy of perspective captures how psychological functions are
distinct from biological structures and defines their relationship as implementational. In figure 5,
contrast computation is not caused by the retina, but is implemented in this biological structure.
Autonomy of organization captures how both psychological functions and biological structures
can be decomposed into parts whose coordinated operation accounts for a particular
phenomenon. From a psychological perspective, vision can be broken down into subprocesses
such as general feature detection. The coordinated activity of these various processes is different
than each process considered in isolation. Mechanistic explanation consists of the delineation of
these causal relations from both a psychological and biological perspective as scientists
demonstrate how the former is implemented in the latter.
8. How mechanistic science furthers substantive theory building

The progress of clinical psychological science depends on the strength of its
explanations, which are more formally termed substantive theories (Meehl, 1990). Substantive
theories delineate the causal structure of the world including the entities and processes that serve
to explain phenomena (Meehl, 1978). Many scholars have remarked on the need for clinical
science to improve the rigor by which it develops and tests its theories in order to advance
etiological understandings of clinical disorders, and have offered visions to facilitate progress
(e.g., Montague et al., 2012).
Explaining a phenomenon requires more than explaining variance. In psychological
research, statistical significance testing has played a prominent role in scientific explanation.
While explaining variance through significance testing is critical for identifying and describing
phenomena of interest, mechanistic models extend the project of scientific explanation by asking
how these statistical relations come about. In the personality pathology exemplar from section 6,
factor loadings and regression coefficients that demonstrate an association between the
abandonment schema and signs and symptoms of Julie’s distress do not provide a causal
narrative of how these phenomena are related. Instead, what is necessary is an articulation of the
parts and operations of a living system that explains the basis for the statistical association. Such
a substantive theory would then function as a model system tested through specific predictions
regarding the temporal and environmental factors that impact the behavior of the proposed
mechanism.
Mechanistic models therefore enable specific predictions characteristic of strong theories.
In the study of vision, when the dynamics of cortical region V1 became rigorously formalized in
the PC/BC computational model, strong predictions were made regarding the shape of the
model’s output when it was exposed to various visual stimuli. Predicting the shape of a
mechanism’s behavior through simulations or in real biological systems with a high degree of
accuracy is compelling evidence that one’s theory is veridical. These predictions differ markedly
from the typical ordinal hypotheses in psychology, which assert that one group will have a larger
mean than another, or that two constructs are correlated (sometimes made stronger with an a
priori defined effect size estimate). Meehl (1990) argued that a large number of explanations
could account for such predictions other than the substantive theory one has in mind. Ultimately,
the mechanistic approach can help avert systemic impediments to theoretical progress that Meehl
(1978) lamented, in which theories come in and out of vogue without an apparent logic to such
transitions.
8.1 Summary of heuristic strategies of the mechanistic framework: Recursive decomposition,
units of organization, and counterfactual predictions
A mechanistic model develops in stages from a sketch, to a schema, to a bona fide
mechanistic explanation. Researchers begin with a phenomenon (e.g., construct) of interest that
is decomposed into an initial set of parts and each part-whole relation defines the units of
organization in the overall mechanism. From a psychological perspective, recursive
decomposition involves iteratively breaking down a function into its component subprocesses.
From a biological perspective, recursive decomposition involves iteratively breaking down a
structure into its physical parts and the activities between them. As each unit of organization is
clarified through empirical work, researchers begin linking psychological functions with
biological structures. The phenomenon is explained when there exists a discipline-wide
consensus that a complete decomposition of functional operations and their implementation in
biological parts has been adequately conceptualized in a specific mechanistic model. Throughout
this process, the phenomenon itself may be repeatedly re-conceptualized. Ultimately,

mechanistic models engender causal hypotheses. Such hypotheses involve predictions regarding
how intervening on a mechanism, at any unit of organization, will result in shifting the state of
the phenomenon it realizes.
Concluding Remarks
The goal of the present article is to demonstrate the potential of a mechanistic approach
to channel and refine efforts already being pursued across the many fields seeking explanations
of psychopathology. The use of mechanistic science can therefore complement existing
approaches used to conceive of and explain phenomena of interest, and thus should not be
thought of as replacing common experimental and statistical methods. Rather, mechanistic
science can facilitate improved collaboration across diverse fields by providing a common
language and meta-theoretical framework used to advance explanations that span several units of
organization. As a generative proposal, the mechanistic framework should also be thought of as a
working template for scientific explanations of living systems that engage in complex
information-processing activities.
GLOSSARY
Biological structure: A physical part of a living system that implements an informational
process and/or a phenomenological experience
Psychological function: An informational process and/or a phenomenological experience that is
realized in a physical system.
Causal relations: Processes or entities acting on each other in specific ways
Substantive theory: Theories about the causal structure of the world, or the entities or processes
that underlie phenomena.
*Psychological perspective: Delineation of the information processing functions of a
mechanism
*Biological perspective: Delineation of how psychological functions are realized by biological
structures
*Environmental perspective: Delineation of how factors outside a mechanism affect its
functioning
*Unit of organization: A group of psychological or biological entities that work together
through particular interconnections and interactions to produce a phenomenon within a
mechanism
*Autonomy of organization: The independence of different units of organization
*Autonomy of perspective: The independence of psychological and biological perspectives of a
mechanism
* Denotes new terms offered by Thomas and Sharp.

Acknowledgements
We thank Gregory A. Miller, Peter A. Ornstein, Wendy Heller, and Eva H. Telzer for their
comments on previous drafts. All views expressed are our own.

References
Aberman, J. E., & Salamone, J. D. (1999). Nucleus accumbens dopamine depletions make rats
more sensitive to high ratio requirements but do not impair primary food reinforcement.
Neuroscience, 92, 545–552. doi: 10.1016/S0306-4522(99)00004-4.
Ahn, H., & Wampold, B. E. (2001). Where oh where are the specific ingredients? A meta-
analysis of component studies in counseling and psychotherapy. Journal of Counseling
Psychology, 48(3), 251–257.
https://doi.org/http://dx.doi.org.proxy2.library.illinois.edu/10.1037/0022-0167.48.3.251
Alpaydin, E. (2014). An introduction to machine learning. MIT Press.
Andersen, S. L., & Teicher, M. H. (2008). Stress, sensitive periods and maturational events in
adolescent depression. Trends in Neurosciences, 31(4), 183–191. https://doi.org/10.1016/
j.tins.2008.01.004
Anderson, M. L. (2014). After phrenology. Chicago, IL, US: MIT.
Bach, B., Lee, C., Mortensen, E. L., & Simonsen, E. (2015). How do DSM-5 personality traits
align with schema therapy constructs?, 29, 212.
Bastos, A. M., Usrey, W. M., Adams, R. A., Mangun, G. R., Fries, P., & Friston, K. J. (2012).
Canonical Microcircuits for Predictive Coding. Neuron, 76(4), 695–711.
https://doi.org/10.1016/j.neuron.2012.10.038
Beauchaine, T. P., & Hinshaw, S. P. (2010). Child and Adolescent Psychopathology. John Wiley
& Sons.
Bechtel, W. (2008). Mental Mechanisms: Philosophical Perspectives on Cognitive
Neuroscience. Taylor & Francis.

Bechtel, W. (2016). Explicating Top-Down Causation Using Networks and Dynamics.
Philosophy of Science. https://doi.org/10.1086/690718
Bechtel, W., & Abrahamsen, A. (1993). Connectionism and the future of folk psychology. In R.
G. Burton (Ed.), Natural and Artificial Minds (pp. 69–100). SUNY Press.
Bechtel, W., & Abrahamsen, A. (2005). Explanation: a mechanist alternative. Studies in History
and Philosophy of Science Part C: Studies in History and Philosophy of Biological and
Biomedical Sciences, 36(2), 421–441. https://doi.org/10.1016/j.shpsc.2005.03.010
Bell, E. C., Markus, D. K., & Goodland, J. K. (2013). Are the parts as good as the whole? A
meta-analysis of component treatment studies. Journal of Consulting and Clinical
Psychology, 81(4), 722–736. https://doi.org/10.1037/a0033004
Berridge, K. C., & Robinson, T. E. (1998). What is the role of dopamine in reward: hedonic
impact, reward learning, or incentive salience?. Brain research reviews, 28(3), 309-369.
Biederman, I. (1981). On the semantics of a glance at a scene. In M. Kubovy & J. Pomerantz
(Eds.), Perceptual Organization. Hillsdale, NJ: Erlbaum. Retrieved from
http://ajp.psychiatryonline.org.proxy2.library.illinois.edu/doi/pdf/10.1176/
appi.ajp.162.3.433
Bishop, S. J., & Gagne, C. (2018). Anxiety, Depression, and Decision Making: A Computational
Perspective. Annual review of neuroscience, (0).
Bolton, D., & Hill, J. (1996). Mind, meaning, and mental disorder: The nature of causal
explanation in psychology and psychiatry. Oxford University Press.
Boothe, R. G., Dobson, V., & Teller, D. Y. (1985). Postnatal development of fision in human
and nonhuman primates. Annual Review of Neuroscience, 8(1), 495–545.

Borsboom, D., Mellenbergh, G. J., & van Heerden, J. (2004). The Concept of Validity.
Psychological Review, 111(4), 1061–1071.
https://doi.org/http://dx.doi.org.proxy2.library.illinois.edu/10.1037/0033-
295X.111.4.1061
Campbell, D. T., & Fiske, D. W. (1959). Convergent and discriminant validation by the
multitrait-multimethod matrix. Psychological Bulletin, 56(2), 81–105.
https://doi.org/http://dx.doi.org.proxy2.library.illinois.edu/10.1037/h0046016
Charpentier, C. J., Aylward, J., Roiser, J. P., & Robinson, O. J. (2017). Enhanced risk aversion,
but not loss aversion, in unmedicated pathological anxiety. Biological psychiatry, 81(12),
1014-1022.
Churchland, P. S., Ramachandran, V. S., & Sejnowski, T. J. (1994). A critique of pure vision. In
C. Koch & J. L. Davis, (Eds.), Large-scale neuronal theories of the brain (pp. 23-60).
Cambridge, MA: MIT Press.
Cicchetti, D. (2010). A multiple-levels-of-analysis perspective on research in development and
psychopathology. In T. P. Beauchaine & S. P. Hinshaw (Eds.), Child and Adolescent
Pstychopathology (pp. 27–50). Hoboken, NJ: John Wiley & Sons, Inc.
Cicchetti, D., & Blender, J. A. (2004). A multiple-levels-of-analysis approach to the study of
developmental processes in maltreated children. Proceedings of the National Academy of
Sciences of the United States of America, 101(50), 17325–17326. https://doi.org/10.1073/
pnas.0408033101
Clark, A. (2013). Whatever next? Predictive brains, situated agents, and the future of cognitive
science. Behavioral and Brain Sciences, 36(03), 181–204.

Collins, A. G., & Frank, M. J. (2014). Opponent actor learning (OpAL): Modeling interactive
effects of striatal dopamine on reinforcement learning and choice
incentive. Psychological review, 121(3), 337.
Craver, C. F., & Bechtel, W. (2007). Top-down causation without top-down causes. Biology &
Philosophy, 22(4), 547–563.
Craver, C. F., & Darden, L. (2013). In search of mechanisms: Discoveries across the life
sciences. University of Chicago Press.
Cronbach, L. J., & Meehl, P. E. (1955). Construct validity in psychological tests. Psychological
Bulletin, 52(4), 281–302.
Cuthbert, B. N., & Insel, T. R. (2013). Toward the future of psychiatric diagnosis: The seven
pillars of RDoC. BMC medicine, 11(1), 126.
Darden, L. (2002). Strategies for Discovering Mechanisms: Schema Instantiation, Modular
Subassembly, Forward/Backward Chaining. Philosophy of Science, 69(S3), S354–S365.
https://doi.org/10.1086/341858
Daw, N. D. (2011). Trial-by-trial data analysis using computational models. Decision making,
affect, and learning: Attention and performance XXIII, 23, 3-38.
Dennett, D. C. (1993). Consciousness explained. UK: Penguin.
Donagan, A. (1957). I.—Explanation in History. Mind, LXVI(262), 145–164.
https://doi.org/10.1093/mind/LXVI.262.145
Dreske, F. (1981). Knowledge and the Flow of Information. Cambridge, MA: MIT
Press/Bradford Books. Retrieved from https://mitpress.mit.edu/books/knowledge-and-
flow-information
Duncan, L. E., Pollastri, A. R., & Smoller, J. W. (2014). Mind the gap: Why many geneticists
and psychological scientists have discrepant views about gene-environment (GxE)
research. American Psychologist, 69(3), 249.
Eldar, E., & Niv, Y. (2015). Interaction between emotional state and learning underlies mood
instability. Nature communications, 6, 6149.
Engel, G. L. (1981). The Clinical Application of the Biopsychosocial Model. Journal of
Medicine and Philosophy, 6(2), 101–124. https://doi.org/10.1093/jmp/6.2.101
Enroth-Cugell, C., & Robson, J. G. (1966). The contrast sensitivity of retinal ganglion cells of
the cat. The Journal of Physiology, 187(3), 517–552.
https://doi.org/10.1113/jphysiol.1966.sp008107
Fenno, L., Yizhar, O., & Deisseroth, K. (2011). The development and application of
optogenetics. Annual Review of Neuroscience, 34.
Fodor, J. A. (1974). Special sciences (or: the disunity of science as a working hypothesis).
Synthese, 28(2), 97–115.
Fouriezos, G., Hansson, P., & Wise, R.A. (1978). Neuroleptic-induced attenuation of brain
stimulation reward in rats. Journal of Comparative Physiological Psychology, 92(4),
661–671.
Franklin, J. C., Jamieson, J. P., Glenn, C. R., & Nock, M. K. (2015). How Developmental
Psychopathology Theory and Research Can Inform the Research Domain Criteria
(RDoC) Project. Journal of Clinical Child & Adolescent Psychology, 44(2), 280–290.
https://doi.org/10.1080/15374416.2013.873981
Friston, K. (2010). The free-energy principle: a unified brain theory? Nature Reviews
Neuroscience, 11(2), 127–138. https://doi.org/10.1038/nrn2787

Gibson, J. J. (1979). The ecological theory of visual perception. Boston, MA: Houghton Mifflin.
Gozzano, S., & Hill, C. S. (2012). New perspectives on type identity: the mental and the
physical. Cambridge University Press.
Hart, H., & Rubia, K. (2012). Neuroimaging of child abuse: A critical review. Frontiers in
Human Neuroscience, 6, 52.
Hempel, C. G., & Oppenheim, P. (1948). Studies in the Logic of Explanation. Philosophy of
Science, 135–175.
Hershenberg, R., & Goldfried, M. R. (2015). Implications of RDoC for the Research and
Practice of Psychotherapy. Behavior Therapy, 46(2), 156–165.
https://doi.org/10.1016/j.beth.2014.09.014
Huys, Q. J., Pizzagalli, D. A., Bogdan, R., & Dayan, P. (2013). Mapping anhedonia onto
reinforcement learning: a behavioural meta-analysis. Biology of mood & anxiety
disorders, 3(1), 12.
Ilardi, S., Rand, K. E., & Karwoski, L. E. (2007). The cognitive neuroscience perspective allows
us to understand abnormal behavior at multiple levels of complexity. The great ideas of
Clinical Science, 17, 243-261.
Jacob, F. & Monod, J. (1961). Genetic regulatory mechanisms in the synthesis of proteins.
Journal of Molecular Biology, 3, 318-356.
Jonas, E., & Kording, K. P. (2017). Could a neuroscientist understand a microprocessor? PLoS
Computational Biology, 13(1), e1005268.
Kaplan, E., & Shapley, R. M. (1986). The primate retina contains two types of ganglion cells,
with high and low contrast sensitivity. Proceedings of the National Academy of Sciences,
83(8), 2755–2757.
Kapur, S. (2003). Psychosis as a State of Aberrant Salience: A Framework Linking Biology,
Phenomenology, and Pharmacology in Schizophrenia. American Journal of Psychiatry,
160(1), 13–23. https://doi.org/10.1176/appi.ajp.160.1.13
Kazdin, A. E. (2007). Mediators and mechanisms of change in psychotherapy research. Annual
Review of Clinical Psychology, 3(Monograph), 1–27.
Kendler, K. S. (2005). Toward a Philosophical Structure for Psychiatry. American Journal of
Psychiatry, 162(3), 433–440. https://doi.org/10.1176/appi.ajp.162.3.433
Lakatos, I. (1970). Falsification and the methodlogy of scientific research programmes. In I.
Lakatos & A. Musgrave (Eds.), Criticism and the growth of knowledge (pp. 91–196).
Cambridge: Cambridge University Press.
Laska, K. M., Gurman, A. S., & Wampold, B. E. (2014). Expanding the lens of evidence-based
practice in psychotherapy: A common factors perspective. Psychotherapy, 51(4), 467–
481.
Lee, T. W. (1998). Independent component analysis. In Independent component analysis (pp.
27–66). Springer US.
Lee, E., Lee, J., & Kim, E. (2017). Excitation/inhibition imbalance in animal models of autism
spectrum disorders. Biological psychiatry, 81(10), 838-847.
Levy, A., & Bechtel, W. (2016). Towards Mechanism 2.0: Expanding the Scope of Mechanistic
Explanation.
Lilienfeld, S. O. (2017). Clinical Psychological Science: Then and Now. Clinical Psychological
Science, 5(1), 3–13. https://doi.org/10.1177/2167702616673363
Livingstone, M. S., & Hubel, D. H. (1984). Anatomy and physiology of a color system in the
primate visual cortex. Journal of Neuroscience, 4(1), 309–356.

MacCorquodale, K., & Meehl, P. E. (1948). On a distinction between hypothetical constructs and
intervening variables. Psychological Review, 55(2), 95–107.
Machamer, P., Darden, L., & Craver, C. F. (2000). Thinking about Mechanisms. Philosophy of
Science, 67(1), 1–25.
Maia, T. V., & Frank, M. J. (2017). An integrative perspective on the role of dopamine in
schizophrenia. Biological psychiatry, 81(1), 52-66.
Marr, D. C. (1982). Vision: A computation investigation into the human representational system
and processing account of visual information. San Francisco, CA: Freeman.
Meehl, P. E. (1967). Theory-testing in psychology and physics: A methodological paradox.
Philosophy of Science, 34(2), 103–115.
Meehl, P. E. (1978). Theoretical risks and tabular asterisks: Sir Karl, Sir Ronald, and the slow
progress of soft psychology. Journal of Consulting and Clinical Psychology, 46(4), 806–
834. https://doi.org/http://dx.doi.org.proxy2.library.illinois.edu/10.1037/0022-
006X.46.4.806
Meehl, P. E. (1990). Why summaries of research on psychological theories are often
uninterpretable. Psychological Reports, 66(1), 195–244.
Miller, G. A. (2010). Mistreating Psychology in the Decades of the Brain. Perspectives on
Psychological Science, 5(6), 716–743. https://doi.org/10.1177/1745691610388774
Miller, G. A., & Yee, C. M. (2015). Moving psychopathology forward. Psychological Inquiry,
26(3), 263–267.
Montague, P. R., Dolan, R. J., Friston, K. J., & Dayan, P. (2012). Computational psychiatry.
Trends in cognitive sciences, 16(1), 72-80.

Neisser, U. (1988). Five kinds of self-knowledge. Philosophical Psychology, 1, 35–39.
Palmer, S. E. (1999). Vision science: Photons to phenomonology. Cambridge, MA: MIT Press.
Palmer, S. E., & Kimchi, R. (1986). The information processing approach to cognition. In T. J.
Knapp & L. C. Robertson (Eds.), Approaches to cognition: Contrasts and controversies
(pp. 37–77). Hillsdale, NJ: Erlbaum.
Park, H. J., & Friston, K. (2013). Structural and functional brain networks: from connections to
cognition. Science, 342(6158), 1238411.
Piccinini, G., & Craver, C. (2011). Integrating psychology and neuroscience: functional analyses
as mechanism sketches. Synthese, 183(3), 283–311. https://doi.org/10.1007/s11229-011-
9898-4
Poldrack, R. A., & Yarkoni, T. (2016). From brain maps to cognitive onotologies: Informatics
and the search for mental structure. Annual Review of Psychology, (67), 587–612.
Rao, R. P. N., & Ballard, D. H. (1999). Predictive coding in the visual cortex: a functional
interpretation of some extra-classical receptive-field effects. Nature Neuroscience, 2(1),
79.
Reid, R. C., & Usrey, W. M. (2012). Vision. In L. Squire, D. Berg, F. E. Bloom, S. Du Lac, A.
Ghosh, & N. C. Spitzer (Eds.), Fundamental Neuroscience, 4th Edition. San Diego:
Academic Press.
Reynolds, S. M., & Berridge, K. C. (2008). Emotional environments retune the valence of
appetitive versus fearful functions in nucleus accumbens. Nature neuroscience, 11(4),
423.
Richardson, R. C., & Stephan, A. (2007). Mechanism and mechanical explanation in systems
biology. In Boogerd et al (Ed.) (pp. 123–144).

Sanislow, C., Pine, D. S., Quinn, K. J., Kozak, M. J., Garvey, M. A., Heinssen, R. K., …
Cuthbert, B. N. (2010). Developing constructs for psychopathopology research: Research
domain criteria, 119(4), 631.
Schultz, W., Dayan, P., & Montague, P. R. (1997). A neural substrate of prediction and reward.
Science, 275(5306), 1593-1599.
Schwartz, S. J., Lilienfeld, S. O., Meca, A., & Sauvigné, K. C. (2016). The role of neuroscience
within psychology: A call for inclusiveness over exclusiveness. American Psychologist,
71(1), 52–70. https://doi.org/10.1037/a0039678
Shadish, W. R., Cook, T. D., & Campbell, D. T. (2002). Experimental and quasi-experimental
designs for generalized causal inference. Wadsworth Cengage learning.
Sheridan, M. A., & McLaughlin, K. A. (2014). Dimensions of early experience and neural
development: Deprivation and threat. Trends in Cognitive Sciences, 18(11), 580–585.
Sporns, O., & Betzel, R. F. (2016). Modular brain networks. Annual Review of Psychology.
Spratling, M. W. (2010). Predictive Coding as a Model of Response Properties in Cortical Area
V1. Journal of Neuroscience, 30(9), 3531–3543.
https://doi.org/10.1523/JNEUROSCI.4911-09.2010
Sutton, R. S., & Barto, A. G. (1998). Reinforcement learning: An introduction (Vol. 1, No. 1).
Cambridge: MIT press
Tarr, M. J., & Gauthier, I. (2000). FFA: a flexible fusiform area for subordinate-level visual
processing automatized by expertise. Nature Neuroscience, 3(8), 764.
Tai, L.-H., Lee, A. M., Benavidez, N., Bonci, A., & Wilbrecht, L. (2012). Transient stimulation
of distinct subpopulations of striatal neurons mimics changes in action value. Nature
Neuroscience, 15, 1281–1289. doi: 10.1038/nn.3188

Vaidyanathan, U., Vrieze, S. I., & Iacono, W. G. (2015). The power of theory, research design,
and transdisciplinary integration in moving psychopathology forward, 26(3), 209–230.
Van Essen, D. C., & Gallant, J. L. (1994). Neural mechanisms of form and motion processing in
the primate visual system. Neuron, 13(1), 1–10.
Waller, N. G., & Meehl, P. E. (2002). Risky tests,verisimilitude, and path analysis.
Psychological Methods, 7(3), 323–337.
Wang, X. J., & Krystal, J. H. (2014). Computational psychiatry. Neuron, 84(3), 638-654.
Waters, H. S., & Waters, E. (2006). The attachment working models concept: Among other
things, we build script-like representations of secure base experiences. Attachment &
Human Development, 8(3), 185–197. https://doi.org/10.1080/14616730600856016
Weinberg, R. (1985). The molecules of life. Scientific American, 253(4), 48–57.
Westen, D. (1994). The impact of sexual abuse on self structure. In D. Cicchetti & S. L. Toth
(Eds.), Disorders and Dysfunctions of the Self (Vol. 5, pp. 223–250). Rochester, NY:
University of Rochester Press.
Woodward, J. (2002). What Is a Mechanism? A Counterfactual Account. Philosophy of Science,
69(3), S366–S377.

Thomas Sharp CPS Final Submission

Uploaded by

Copyright:

Available Formats

Thomas Sharp CPS Final Submission

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Thomas Sharp CPS Final Submission

Uploaded by

Copyright:

Available Formats

RUNNING HEAD: Mechanistic science 1

Preprint 8/6/2018, accepted at Clinical Psychological Science pending minor revision

Mechanistic science: A new approach to comprehensive psychopathology research that

relates psychological and biological phenomena

University of Illinois at Urbana-Champaign

University of North Carolina at Chapel Hill

Joel G. Thomas, Department of Psychology, University of Illinois at Urbana-Champaign, 603 E.

Daniel St., Champaign, IL 61801

Corresponding Author Email: [email protected]

Paul B. Sharp, Department of Psychology and Neuroscience, University of North Carolina at

Chapel Hill, 124 Howell Hall, Chapel Hill, NC 27514

work in philosophy of science by proposing a framework called mechanistic science as a

phenomena involves demonstrating how psychological functions are implemented in biological

structures. Successful early attempts to advance mechanistic explanations of psychological

a range of clinical psychological phenomena including gene by environment findings,

computational models of reward processing in schizophrenia, and self-related processes in

and successful collaborations across a range of disciplines.

Keywords: autonomy; causal explanation; levels of analysis; mechanism; psychopathology

Introduction – Why we need to examine the way we explain psychological phenomena

approach to explaining psychological phenomena was formalized in the mid-twentieth century

called mechanistic science, can make up for this shortcoming.

Integrating psychological and biological phenomena is particularly relevant to scientists

interested in psychopathology, as exemplified by the Research Domain Criteria (RDoC)

to “integrate the fundamental genetic, neurobiological, behavioral, environmental, and

psychological and biological phenomena are both involved in explaining phenomena in

link these domains without eliminating or giving preference to either camp?

psychology to biology is a major reason for the dearth of causal explanations of

primarily focused on whether findings are reproducible. Demonstrating reproducibility, however,

go awry across various forms of psychopathology.

neuroscience, mechanistic science maintains that a crucial objective of scientific explanation

should be to move beyond identification and description of phenomena to an explanation of how

mechanisms—structures defined by their component parts, operations, and organization, whose

orchestrated functioning is responsible for a particular phenomenon.

In contrast to typical practices in psychological science, mechanistic science requires that

psychological scientists constrain their conceptions of psychological functions to those that

organizing principles, and developmental features of complex organisms.

integrated through demonstrating implementation. Exemplars from research on psychopathology

are indeed implemented in biological systems.

When abstract processes, such as “energy production,” can be explained by appealing to

This transition in biology, although occurring without reference to a formal philosophy of

rigorous scientific explanation of the phenomenon.

It is only quite recently, however, that philosophers have attempted to arrive at an

psychological literature, they are done so quite generally. In addressing psychotherapeutic

how change came about” (p. 3).

Most mechanisms of relevance to clinical psychological science are therefore yet to be

affect the phenomenon.

2. Information processing functions: What psychology has to contribute

them to a nonphysical substance—mind (Bechtel, 2008).

following characterization of mechanisms:

A mechanism is a structure performing a function in virtue of its component parts,

component operations, and their organization. The orchestrated functioning of the

mechanism is responsible for one or more phenomena (Bechtel & Abrahamsen,

are represented and computed on (e.g., Eldar & Niv, 2015).

To understand how information processing psychological functions are related to

biological phenomena, a formal definition of information is useful. Information can be thought

ambient temperature. This mechanism can be described in functional terms by appealing to a

This functional process can be implemented in a physical system that leverages a

Because functions (e.g., temperature regulation) might be implemented in several

equivalent to structures (Fodor, 1974). For this reason, a mechanistic explanation of a