LIST OF FIGURES

1. Illustration of the Hierarchical Taxonomy of

Psychopathology 51
2. Timeline of the behavioral task 51
3. Illustration of a typical individual EDA response during
different phases of the behavioral task 52

iv
 1

Chapter 1
Introduction
Significant life stress is highly prevalent in emotional disorders, with over

80% of individuals who meet criteria for depression in community samples having

experienced a recent acute or chronic stressor (Brown & Harris, 1989).

Individuals diagnosed with anxiety and depression are more than twice as likely

to have experienced a major adverse life event at any point in their lives (Shrout

et al., 1989) and are between 2 and 6 times more likely to have experienced

such events within 6 months of the onset of disorder (Asselmann, Wittchen, Lieb,

Höfler, & Beesdo-Baum, 2015; Kendler, Karkowski, & Prescott, 1999; Mazure,

1998). Epidemiological studies of the impact of childhood adversities such as

neglect and abuse have found that these experiences account for nearly 30% of

individual differences in risk for psychological disorders across 21 countries

(Kessler et al., 2010). Findings like these suggest a robust and significant

association between emotional disorders and stressful life events.

Predominant theories of the developmental course of emotional disorders

implicate stressful life events as a causal factor that both precipitates disorders

and maintains symptomology. Stressful life events prospectively predict the onset

of affective disorders (Kim, Conger, Elder, & Lorenz, 2003; Slopen et al., 2010)

and subsequent stressful life events which may aggravate symptoms (Kendler &

Gardner, 2016; Technow, Hazel, Abela, & Hankin, 2015). Although genetic

vulnerabilities for psychopathology are likely involved in the relationship between

early adversity and psychopathology (Klengel at el., 2013), evidence suggests

 2

that the experience of adversity does play a causal role in the development of

emotional disorders, conferring risk beyond that explained by genetics. Studies of

twins who are matched on family environment but have divergent early

experiences support this claim. In cases where one twin has reported an acute

stressor and the other has not, research finds that twins reporting sexual abuse

and other stressful life events are at far greater risk for subsequent emotional

disorders than the twins who do not, even if those events are independent, or

unrelated to the individual’s own behavior (Kendler et al., 1999, 2000). The way

in which an individual reacts to stress is thought to be a crucial mechanism

linking stressful events to psychopathology. Stressful life events during childhood

seem to interact with genetic predispositions towards maladaptive stress

response, increasing the risk for the development of emotional disorders by

deleteriously altering cognitive and neurobiological responses to subsequent

stress (Heim, Newport, Mletzo, Miller, & Nemeroff, 2008; Wichers et al., 2012).

Biological mechanisms of stress response

Given the central role of stress response in etiological theories of

emotional disorders, understanding individual differences in stress reactivity has

been one focus of clinical research. Severity of stress response is determined by

multiple systems. Cognitive reactions to stressful events involve the appraisal of

the event and the psychological strategies engaged to modulate emotional

reactions, while biological processes affect the function of major organs and the

production of hormones. These two broad classes of response are linked in that
 3

psychological appraisal of stressors can reduce or exacerbate biological stress

response. For this reason, biological processes are integrated into psychological

studies of stress.

The autonomic nervous system (ANS) is one of the most heavily studied

response symptoms associated with acute stress. The ANS regulates major

organs including the heart, lungs, and gastrointestinal system and serves to

maintain homeostasis, or equilibrium, in response to environmental input (Tsigos

& Chrousos, 2002). The ANS is composed of two branches: the parasympathetic

and the sympathetic. Under threat, the sympathetic branch of the ANS

potentiates the so-called “fight or flight” response, which aims to enhance

attention to the danger at hand and ready the organism to defend itself by

accelerating cardiovascular function and respiration. The parasympathetic

branch is generally deceleratory, acting in opposition to the sympathetic

response to return bodily systems to their resting state. Although stress response

is ultimately a function of the interaction of these two branches over time,

sympathetic activation is predominant during reactions to acute stress (Bouscein,

2012).

One of the most widely used measures of sympathetic nervous system

(SNS) activity is electrodermal activity (EDA; Duffy, 1972). EDA is a general term

referring to dynamic changes in the electrical characteristics of human skin,

primarily due to the activity of sweat glands. This measure is popular in

psychological research because EDA is thought to be determined solely by the

 4

activity of the SNS, unlike other indicators of ANS response such as heart rate or

cortisol response which may also reflect activity of the parasympathetic branch

(Dawson, Schell, & Filion, 2007). EDA is generally quantified as either tonic skin

conductance level (SCL), the general state of conductivity of the skin, or phasic

skin conductance responses (SCRs), abrupt “spikes” in conductance in reaction

to some discrete stimulus or event.

For decades, psychophysiological research has explicated the link

between electrodermal response and cognition and emotion. A large portion of

this work has examined EDA correlates of fear by presenting participants with

naturally aversive stimuli, like loud noises or electric shocks. The typical

response to these stimuli is increased frequency or magnitude of SCRs and

heightened SCL (Dawson, Schell, & Filion, 2007). In situations where participants

are aware of when the stimulus will come, such as if the experimenter provides a

clock counting down to the onset of the stimulus, SCL reaches its highest peak

just prior to onset, reflecting the psychological anticipation of the stimulus

(Bouscein, 2012). Similarly, in fear conditioning paradigms where some neutral

stimulus (e.g., a geometric shape) is paired with an aversive stimulus (e.g.,

electric shock) repeatedly, EDA will spike in response to the previously neutral

stimulus even after the aversive stimulus is removed. This research on normative

stress response has provided a foundation for the study of stress response in

psychopathology.

Physiological stress reactivity in emotional disorders

 5

Psychological stress reactions have long been incorporated into

developmental theories of emotional disorders, such as Beck’s classic model of

depression (Beck, 1987), which posits that depressed individuals exacerbate

emotional reactions to negative events by exaggerating their negative impact, or

the theory that panic disorder is maintained when individuals are overly sensitive

to somatic symptoms of anxiety and catastrophize these sensations, leading to

uncontrollable panic and further fear of bodily sensations (Barlow & Craske,

1988). Indeed, biases towards exaggerated negative appraisal of stressors have

been demonstrated to prospectively predict symptoms across the full range of

emotional disorders (e.g., Conway, Starr, Espejo, Brennan, & Hammen, 2016).

Research has shown that such maladaptive information-processing styles are

associated with dysregulated physiological responses to unpleasant events

(Gaab, Rohleder, Nater, & Ehlert, 2005; Schlotz, Hammerfald, Ehlert, & Gaab,

2011), spurring a complimentary line of research that has investigated

physiological stress reactions in emotional disorders.

Physiological stress reactivity has perhaps been most comprehensively

studied in the anxiety disorders. Individuals with a range of disorders, including

generalized anxiety disorder, panic disorder, social and specific phobias, and

posttraumatic stress disorder, tend to display maladaptive biological reactions to

stressors. Specifically, these individuals demonstrate exaggerated and prolonged

SNS activity in response to stress, coupled with attenuated parasympathetic

nervous system activity; the initial response to stress is greater in magnitude and
 6

longer in duration, and the return to homeostasis is slowed. This pattern is

evidenced through heightened EDA and restricted heart rate variability and

respiratory sinus arrhythmia, markers of parasympathetic nervous system activity

(e.g. Craske et al., 2009; Pêgo, Sousa, Almeida, & Sousa, 2010). This pattern of

SNS hyperactivity holds true not only for stressors, but also for baseline

physiological activity under nonthreatening conditions (Blechert, Grossman,

Laitman, & Wilhelm, 2007; Monk et al., 2001). Further, increased reactivity has

been shown to predict or maintain later symptoms of anxiety and mood disorders

(Morris, Rao, & Garber, 2012; Nelemans et al., 2017).

Aberrant biological reactions to stress are also evident in depression,

although the body of literature linking EDA to depression is relatively small and

inconsistent. Some researchers find attenuated EDA reactivity in depression

(Bonnet & Naveteur, 2004; Donat & McCullough, 1983; Miquel, Fuentes, Garcia-

Merita, & Rojo, 1999), some find heightened activity (Lin, Lin, Lin, & Huang,

2011; Sanders & Abaied, 2015), and others find variation in responses as a

function of specific symptoms or subtypes of depression (Thorell, Kjellman, &

D’Elia, 1987; Williams, Iacono, & Remick, 1985). Cortisol reactivity in depression

has been studied more extensively and has been shown to be abnormal in

depressed individuals. Initial heightened cortisol reactions to stressors are

adaptive to some extent, but depression has been consistently associated with

abnormally high cortisol values in the absence of threat (Burke, Davis, Otte, &

Mohr, 2005; Knorr, Vinberg, Kessing, & Wetterslev, 2010; Pariante & Lightman,
 7

2008); one meta-analysis estimates that 60-73% of depressed individuals

demonstrate higher basal cortisol levels than non-depressed controls (Stetler &

Miller, 2011). Biological response in the presence of threat is less consistent.

Some researchers find that depressed individuals show attenuated cortisol and

skin conductance reactivity compared to controls and anxiety disorders (Benning

& Oumesiane, 2016; Pruneti, Lento, Fante, Carrozzo, & Fontana, 2010; Thorell

et al., 2013), while others find a pattern of hyperreactive response and impaired

recovery similar to that of anxiety disorders (Grillon, Franco-Chaves, Mateus,

Ionescu, & Zarate, 2013; Morris et al., 2012).

In sum, abnormalities in physiological responses to stressors have been

demonstrated in major depressive disorder (Grillon, Ameli, Goddard, Woods, &

Davis, 1994), generalized anxiety disorder (Lieberman, Gorka, Shankman, &

Phan, 2017), social anxiety disorder (Yoon & Joormann, 2012), panic disorder

(Gorka, Liu, Saraspas, & Shankman, 2015), and posttraumatic stress disorder

(Metzger, Orr, Berry, Ahern, Lasko, & Pitman, 1999). There is reason to believe

that this feature is common across these disorders because abnormal stress

reactivity is a multifinal risk factor for emotional disorders. Muiltifinality is a

process where a characteristic, event, or environment confers increased risk for

multiple forms of psychopathology (Cicchetti & Rogosch, 1996); for example,

childhood adversity is strongly but nonspecifically predictive of psychological

disorders. Those who experience childhood adversity are far more likely to

develop psychopathology than those who do not, but this effect is true of all
 8

forms of psychopathology, not a specific disorder (Afifi, Brownridge, Cos, &

Sareen, 2006; McLaughlin, Conron, Koenen, & Gilman, 2010). Similarly,

abnormal stress responses are evident across emotional disorders, but do not

seem to reliably discriminate between disorders. This commonality may be

expected, as mood and anxiety disorders are highly comorbid (Brown et al.,

2001; Kessler, Chiu, Demler, & Walters, 2005) and emerging evidence suggests

that they may share latent liabilities (Kreuger & Markon, 2006) and

neurobiological features (Jenkins et al, 2016; Oathes, Patenaude, Schatzberg, &

Etkin, 2015). It is currently difficult to discern the extent to which abnormal

physiological reactivity is shared among emotional disorders, however, because

most of the research in this area attempts to find physiological links to a single

disorder, or a small set of disorders, at a time. Recent evidence suggests that

searching for disorder-specific effects in this way may be suboptimal for

psychopathology research, creating disparate literatures and masking shared

etiologic pathways. Instead, investigations into the latent structure of common

mental disorders have provided a dimensional and transdiagnostic framework

that can further our understanding of multifinal risk factors.

A hierarchical framework for clinical research

The prevailing systems of classification of mental disorders conceptualize

disorders as discrete entities with separate etiologies. However, several decades

ago psychopathology researchers began to question the validity of categorical

diagnostic systems. This was in part due to the recognition of substantial

 9

psychiatric comorbidity, or the simultaneous presence of two or more distinct

disorders in a single individual. Psychiatric comorbidity is rampant, occurring far

more often than would be expected if mental illnesses were truly independent of

one another (Krueger & Markon, 2006). A more likely interpretation is that

traditional classification systems are flawed, and comorbidity is an artifact of

drawing divisions that are not true to how psychiatric illness actually presents.

Current diagnostic boundaries are rationally derived; symptoms are grouped into

diagnoses based on clinical judgement. Recent work has sought to improve the

classification of mental disorders through empirically-driven methods that honor

the observed distribution of symptoms.

Evidence suggests that the boundaries drawn by categorical classification

systems are to some extent arbitrary. One major finding of research that

examines the structure of psychopathology is that symptoms of common mental

illnesses are distributed continuously throughout the population (Haslam,

Holland, & Kuppens, 2012; Widiger & Samuel, 2005). Current classification

systems do not account for this continuity and specify diagnostic thresholds that

are counterintuitive. Under these systems, an individual who displays three of six

symptoms of anxiety would be diagnosed as having generalized anxiety disorder,

whereas an individual who displays only two symptoms would not, even if the

latter individual were more severely impaired. Categorical classifications that

ignore the dimensionality of psychopathology obscure important clinical

information, such as gradations of severity within diagnoses or similarities

 10

between cases that barely reach the diagnostic threshold and those that barely

miss it. The dimensional distribution of psychopathology blurs lines not only

between healthy and disordered individuals, but also between categories of

disorders. Factor analytic studies suggest that traditionally-defined disorders

share a common structure where observed symptoms covary due to latent traits

which cut across traditional diagnostic boundaries (Forbush & Watson, 2013).

Again, categorical classifications that separate clusters of symptoms into discrete

disorders fail to recognize commonalities between disorders, masking shared

core traits that could provide valuable clinical information. For these reasons

clinical researchers have migrated toward dimensional models of

psychopathology. One such model, known as the Hierarchical Taxonomy of

Psychopathology (HiTOP; for a review, see Kotov et al., 2017), organizes the

available knowledge of the underlying structure of psychopathology into a

dimensional and hierarchical system that can better guide clinical research.

The hallmark of the HiTOP model is its hierarchical structure. Latent

dimensions of psychopathology can be organized according to level of specificity,

with dimensions consisting of specific symptoms at the bottom of a hierarchy and

very broad, generalized dimensions at the top. Five levels of the HiTOP hierarchy

have been identified (see Figure 1). At the bottom, singular symptoms that are

strongly correlated with each other cluster into symptom components, the most

specific level of the hierarchy. For example, losing interest in enjoyable activities

and feeling unmotivated are symptoms that cooccur very frequently and
 11

compose a symptom component that is labeled anhedonia. Symptom

components that are highly correlated compose syndromes; anhedonia,

dysphoria, and appetite disturbance cluster into a syndrome resembling

depression. Syndromes that share many core features cluster into subfactors,

such as the distress subfactor that encompasses syndromes resembling

depression, post-traumatic stress disorder, and generalized anxiety disorder.

Overarching spectra account for similarities among subfactors, such as the

internalizing spectrum. This spectrum is composed of distress and fear

subfactors and subsumes a range of traditionally-defined disorders, including

mood and anxiety disorders and eating pathology. Finally, spectra converge on a

highly general trait known as the “p factor” that represents features shared

amongst all forms of psychopathology, analogous to the “g factor” of general

intelligence.

The higher-order dimensions of the HiTOP model are well-studied and

have been shown to be temporally stable and highly reliable, replicating across

age groups, genders, clinical and non-clinical populations, specific assessment

instruments, and measurement modalities (Eaton et al., 2013; Eaton, Krueger, &

Oltmanns, 2011; Fergusson, Horwood, & Boden, 2006; Krueger, Chentsova-

Dutton, Markon, Goldberg, & Ormel, 2003; Sellbom, 2016). As confidence in the

reliability of this statistical model has grown, researchers have begun to

demonstrate its utility as a guide for research. This model allows one to parse

traditionally-defined disorders into disorder-specific variance, or features that

 12

pertain only to a narrow band of symptoms, and transdiagnostic variance,

features that are common across multiple forms of psychopathology. We can

then ask whether the more specific or more general features of psychopathology

are most relevant to the question at hand by comparing the predictive ability of

disorder-specific factors to that of transdiagnostic factors. For example, we may

find that the efficacy of exposure therapy is highly related to standing on the fear

subfactor, but less so to the more general internalizing spectrum, which

encompasses fear-based disorders as well as disorders that are better

characterized by distress.

Researchers have begun to identify instances in which general

dimensions of psychopathology are more effective predictors than disorder-

specific dimensions. In longitudinal studies, transdiagnostic factors outperform

disorder-specific variance in predicting psychopathology; one study found this to

be the case 97% of the time, even when disorder-specific factors at Wave 1 were

predicting the same specific disorder at Wave 2 (Kim & Eaton, 2015). It has been

found that early childhood adversity, a robust correlate of psychopathology later

in life, predicts standing on the latent spectra but not specific disorders (Conway,

Raposa, Hammen, & Brennan, 2018; Vachon et al., 2015). Transdiagnostic

factors are also more predictive than specific disorders of clinically-relevant

outcomes, such as domains of function in personality pathology (Wright et al.,

2016), suicidality and treatment-seeking (Sunderland & Slade, 2015), and

impairment associated with anxiety and depression (Markon, 2010).

 13

The upper levels of the hierarchical structure help to explain multifinal risk

factors. For example, the finding that early adversity is more strongly linked to

higher-order spectra than disorder-specific factors suggests that the deleterious

effects of adversity operate through transdiagnostic pathways that may result in

many forms of psychopathology (Conway et al., 2018; Vachon et al., 2015). This

suggests that this risk factor likely causally affects the features that are shared by

most pathologies, such as emotion dysregulation or executive dysfunction, rather

than directly causing specific symptoms, like compulsive behaviors or substance

abuse. In addition to providing hints to causal relations, the HiTOP model helps

to streamline etiologic research. Given the strong link between higher-order

dimensions and early adversity, attempts to link early adversity to specific lower-

level components of psychopathology would be inefficient; we would expect to

find the same relationship between adversity and each specific component

because adversity influences a higher-order trait that subsumes these

components. Rather than demonstrating the relevance of a biological trait or

environmental influence to multiple categorical disorders, we can determine

whether a given risk factor confers widespread risk for generalized pathology or

affects only a narrow band of symptomology. The current study sought to use the

HiTOP model in this way to investigate stress reactivity as a multifinal factor in

the development of emotional disorders.

The current study