MORAL EVALUATIONS IN THE CONTEXT OF UNCERTAINTY 

THE FACE OF JUDGMENT: 

 

By 

Andrew Marcel Defever 

A DISSERTATION 

 

Submitted to 

Michigan State University 

in partial fulfillment of the requirements  

for the degree of 

 

Psychology – Doctor of Philosophy 

 

2018 

 

ABSTRACT 

 

THE FACE OF JUDGMENT: 

MORAL JUDGMENT IN THE CONTEXT OF UNCERTAINTY 

 

By 

 

Andrew Marcel Defever 

 

Moral judgment is an integral part of social living; a ubiquitous phenomenon across time and 

human cultures. Inherent to this is an uncertainty factor: inferential processes necessitate a 

component of situational ambiguity, requiring the perceiver to make a best-guess considering the 

available evidence. While research has demonstrated variability exists in judgments both 

between situations and within, less is known about how judgments occur under explicit 

uncertainty. Across a series of 4 studies, I examined whether third party emotional expressivity 

influences severity judgments of moral transgressions, targeting situations that manifest 

uncertainty. A pilot study was first implemented to identify stimuli appropriate for moral 

judgment research. Study 2 and Study 3 examined the influence of facial expressions (angry, 

disgust, neutral, happy) on moral severity judgments in both clear and uncertain contexts but did 

not find any effect. Study 4 implemented auditory expressions (angry/neutral crowd sounds) 

instead of facial expression and found that auditory expressions did influence severity judgments 

in two instances. Implications and future directions on the function of communicative 

expressions under situational uncertainty are discussed. 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

To my wife Rachel,  

my parents Lynda and Guido,  
and my sisters Erin and Mahrie. 

Your lessons of love and kindness are my greatest gifts in life. 

Thank you. 

 

iii 

TABLE OF CONTENTS 

 

 
 
LIST OF TABLES 
 
LIST OF FIGURES 
 
CHAPTER 1   
 
 
 
 
 

Introduction 
 
 
 
 

Moral judgment and decision-making 
Uncertainty 
 
Emotions 
 

Uncertainty in moral judgment 

Morality and emotions 

The Present Studies 
Overview of Dissertation 

 
 
 
CHAPTER 2   
 
 

Distributions 

Study 1: Pilot Study 
Participants 
Materials & Procedure 
Results  
 
Discussion 

 
 
 
 
CHAPTER 3   
Study 2 
 
Participants 
 
 
Materials & Procedure 
 
 
 
 
 
 
 
Results  
 
 
 
 
 
Discussion 
 
CHAPTER 4   
Study 3 
 
 
Participants 
Methods & Procedure 
 
 
 
 
 
 
 
 
 

Facial expression stimuli 
Judgment stimuli 
Judgment task 

Main hypotheses 
Exploring alternatives 

Facial expression stimuli 
Judgment stimuli 
Judgment under uncertainty 
Judgment task 

 

iv 

vi 

vii 

1 
1 
2 
4 
6 
6 
7 
8 
10 

11 
11 
13 
13 
15 
15 
16 

17 
17 
17 
18 
18 
19 
19 
21 
22 
24 
25 

26 
26 
27 
27 
27 
28 
28 
28 

29 
30 
31 
32 

33 
33 
33 
34 
34 
34 
34 
36 
36 
38 
39 
39 

40 
40 
40 
41 
44 
46 

48 

49 

72 

Main hypotheses 
Demographics 

Auditory expressions 
Judgment stimuli 
Judgment task 

Main hypotheses 
Exploring alternatives 
Demographics 

Results  
 
 
Discussion 
 

 
 
 
 
 
CHAPTER 5   
Study 4 
 
 
Participants 
Methods & Procedure 
 
 
 
 
 
 
 
Results  
 
 
 
 
 
 
 
 
Discussion 
 
CHAPTER 6   
 
 
 
 
 
 
FOOTNOTES  
 
APPENDIX 
 
REFERENCES 

 

Discussion and Conclusion 
Summary of Findings 
Discussion 
Limitations 
Conclusion 

 

v 

LIST OF TABLES 

 
 

 
Table 1: Study 1 IAPS image mean ratings (SD) and descriptive statistics, 

including severity, relevance, ambiguity, and emotions. 

 
Table 2: Study 1 judgments, bivariate correlations (Pearson’s r) 
 
Table 3: Study 2 IAPS image mean ratings (SD) and descriptive statistics, 

including severity, others’ perceived ratings, ambiguity, and emotions. 

 
Table 4: Study 2 severity judgments by facial expression condition. 
 
Table 5: Study 2 judgments, bivariate correlations (Pearson’s r) and demographic 

means. 

 
Table 6: Study 2 severity ratings, omnibus effects and parameter estimates. 
 
Table 7: Study 3 IAPS mean severity and uncertainty ratings (SD) by image, 

facial expression. 

 
Table 8: Study 3 judgments, bivariate correlations (Pearson’s r) and demographic 

means across both the clean and blur conditions. 

 
Table 9: Study 3 severity ratings, omnibus effects and parameter estimates. 
 
Table 10: Study 4 judgments, bivariate correlations (Pearson’s r) and 

demographic means across experimental conditions. 

 
Table 11: Study 4 severity ratings, omnibus effects and parameter estimates. 
 
 
 
 
 
 

 

50 

53 

54 

56 

57 

58 

60 

61 

62 

64 

66 

 
 
 
 
 
 

 

vi 

LIST OF FIGURES 

 
 
 

Figure 1: Example severity rating distributions of a neutral (left), severe (right), 

and intermediate (bottom) image. 

 
Figure 2: Study 2 mean severity ratings, by image and facial expression condition. 
 
Figure 3: Study 3 mean severity ratings, by image and certainty (blur) condition. 
 
Figure 4: Study 4 mean severity ratings, by image and crowd condition. 
 
 
 
 

 

 

68 

69 

70 

71 

 

vii 

CHAPTER 1 

Introduction 

Morality is a ubiquitous social phenomenon that encompasses ‘oughts’ within the domain 

of human behavior. That is, morality dictates how people should and should not behave toward 

others or toward sacred items in their lives (e.g., shrines, books, people, deities, etc.). Most, if not 

all, human cultures have a moral system in place, and even some non-human ape societies may 

have moral dictates as well (e.g., fairness; Brosnan & de Waal, 2014). However, precisely what 

morality is, its conceptual boundaries, and subsequent evaluative components, have not been 

thoroughly investigated and clarified. Indeed, the prolific nature of morality in our everyday life 

underscores the need for a better understanding of its constituent properties.  

Here I outline a series of studies designed to examine one component related to morality: 

judgment and decision-making. Inherent in moral systems is an evaluative component; a way to 

assess whether an action is right or wrong, or whether it does or does not violate a moral dictate. 

Evaluative judgments about morality are multifaceted, including both qualitative evaluations of 

whether an act is or is not moral, and quantitative evaluations of how commendable or severe the 

act is (Bartels, Bauman, Cushman, Pizarro, & McGraw, 2014). When we view behavior that is 

potentially moral in nature, we may progress through a number of steps: gathering of evidence, 

emotional reactions, interpersonal judgment, and sometimes counter-action. Making morally 

relevant judgments, then, is an integral part of the operation of moral pre- and proscriptions.  

I focus here on judgment of moral violations of others, such as the assessment of another 

transgressing a proscriptive moral norm. That is, judging the severity of an act when a person is 

seen as doing something that is morally prohibited. While there is evidence that internal 

characteristics (e.g., motivation, intent; Young & Saxe, 2011; Cushman, 2008) and external 

 

1 

characteristics inherent to a situation (context, means; Mikhail, 2007; McDonald, Defever, & 

Navarrete, 2017) can modify the perceived severity of an action, these typically involve 

judgments of clear and unambiguous situations. Much less is known about how moral judgments 

occur with imperfect knowledge or ambiguity, a situational characteristic necessarily present in 

judgments of others.  

Across a series of 4 studies, I examined how judgments may be influenced by social cues 

in the assessment of ambiguous situations. Specifically, I investigated whether judgments are 

influenced by the non-verbal behavior of others in the proximal situation. As emotions are 

reliably linked to moral reactivity (Prinz, 2006; Avramova & Inbar, 2013), and produce distinct 

facial and somatic expressions theorized to have evolved as communicative devices (Shariff & 

Tracy, 2011), I examined whether the external signaling components of emotional expression 

convey meaningful information during the moral judgment process.  

I hypothesized that (1) emotional expressions conveyed by others influenced moral 

severity judgments by informing the target of relevant contextual evidence. Because judgments 

of others are inferential, uncertain, and involve an appraisal of available evidence, I further 

hypothesized that (2) this relationship would be strongest when the situation is ambiguous, as 

perceivers with imperfect knowledge should glean and subsequently utilize the information 

contained in the emotion to inform their judgment. Finally, I explored how a number of 

individual difference characteristics might affect the degree to which people use emotional 

expressions to inform moral judgments under situational uncertainty. 

Moral judgment and decision-making 

Morality as a construct encapsulates a wide variety of behaviors and responses.  Here I 

focus specifically on transgressions: acts assessed as immoral or ‘wrong’ that typically evoke a 

 

2 

negative affective response. In this case, moral judgment of a person’s behavior is an assessment 

of wrong-ness; of the severity of the transgression. Further, moral judgment is a relative 

appraisal process on a continuous spectrum, such that some acts are more wrong than others. For 

example, theft and murder are both morally wrong, but an individual would not judge theft and 

murder as equally severe moral violations. From a sociological perspective, this continuous 

spectrum is also evident in the institutionalization of moral proscriptions. Murder is consistently 

punished more severely than theft across all modern punitive institutions, with the level of 

punishment scaling with the severity of the transgression.  

However, it’s not simply the action or the outcome that influence the assessment of a 

moral transgression. Extant literature suggests a number of situational and individual factors that 

influence the judged severity of an action. For example, a classic finding from trolley dilemma 

research (i.e., Thomson, 1985) shows that when deciding whether to kill one person to save five 

others, the action is judged as more wrong when it involves interpersonal contact (e.g., a push) 

versus when it does not (e.g., flipping a switch; Mikhail, 2007; Cushman, Young, & Hauser, 

2006). Intention is another salient factor: if harm comes to another through an intended, 

volitional act (e.g., deciding to poison another person), it is judged as more wrong than if it 

occurred unintentionally or indirectly as a side effect (e.g., accidentally poisoning them; 

Cushman, 2008). While in both of these examples the outcome is the same across both courses of 

action (i.e., a person is killed), additional facets of the situation and/or the actors involved alters 

the judgment as more or less severe. 

Why would the same outcomes be judged as more or less severe depending on 

conditional factors of the situation? Logically, judgment of another person’s action is an 

inferential process (DeScioli & Kurzban, 2009). It involves a gathering of evidence to inform the 

 

3 

judgment and appraise the action appropriately, including information about both the actor’s 

internal state (e.g., motivation, intention, and psychological state) and external information (e.g., 

the means implemented in the act and the socioecological context of the act). 

When making an inferential judgment there is an inherent and unavoidable uncertainty to 

the judgment, derived from the subjective nature of the human mind: we are unable to peer into 

the ‘contents’ of another person’s brain. When judging another person’s actions, we must infer 

(through abductive reasoning) the most likely conclusion on what evidence is available. This 

applies necessarily to assessing the internal state of the violator: even if they verbalize their 

intention and motivation for their act, one cannot have epistemologically perfect certainty of 

their mental state. 

Concerning external and contextual information, there is more variation in the level of 

knowledge a person might have in regards to judgment-relevant evidence. For example, if 

someone witnesses a person strike another person in the face at a boxing match, the contextual 

information is salient, readily available, and informs the judgment of the act. Here the act is 

likely judged as a minimally severe violation, since the context of a boxing match calls for 

interpersonal harm as normative and expected. However, if someone witnesses the same act on a 

busy public street, between two people in business attire on their way to work, much less 

contextual information is readily available. Likely this same act would be judged as more severe 

than in the boxing match, since the context of a morning commute to work does not call for 

interpersonal harm as normative or expected. 

Uncertainty 

Another way to construe the nature of inferential judgments is that they’re fraught with 

uncertainty. That is, there are facets of the situation that are not reliably known, or potentially 

 

4 

unknowable. While uncertainty can be broken down into different epistemological types, I focus 

here on uncertainty as a lack of clarity in information visibly or readily available in a situation 

(Smithson, 2008).  

Uncertainty of this type is a ubiquitous, powerful, and psychologically salient feature of 

the environment. From an evolutionary perspective, uncertainty has been so prolific in the 

environment as to shape cognitive mechanisms over long periods of time. For example, error 

management theory (EMT) describes how uncertainty in the environment has led to a biased 

evaluative mechanism. This biased mechanism arises from a consistent asymmetry in outcomes 

for false-positive and false-negative conclusions. For instance, men are thought to be biased in 

predicting women’s sexual interest in them because it was costlier for men to miss a potentially 

interested mate than to perceive them as uninterested. That is, when the perceptions of a 

woman’s intent are uncertain, men are thought to be biased toward perceiving women as 

interested (false alarm) as opposed to not interested (missed detection; Haselton & Buss, 2000).  

Proximally, uncertainty makes people uncomfortable and anxious, and motivates 

individuals to alleviate their discomfort (Reiman, Fusselman, Fox, & Raichle, 1989). In 

uncertain situations individuals are likely to utilize cues from the environment in order to 

disambiguate the situation. For example, imagine you come across a crowd of people while 

walking down the street. The crowd is surrounding what appears to be two individuals, and you 

hear noises that sound like a verbal conflict. Are the two fighting? Are they in danger? Are you 

in danger? As you walk closer, you notice the crowd; their faces are not angry and their posture 

does not indicate they are threatened. As you merge into the crowd to get a better look, you 

notice the individuals are dressed in costumes, and there’s a small bucket on the sidewalk full of 

change. Indeed, based on the available evidence you infer they are street performers and it is not 

 

5 

a threatening situation. In this way, you have disambiguated the uncertain situation by 

accumulating evidence about other people’s emotional states, the environment, and assess the 

likelihood of given scenarios (a street performer versus an altercation). 

Uncertainty in moral judgment. EMT highlights the power of uncertainty in shaping 

our perceptions of the world over evolutionary time. Uncertainty has been a powerful and salient 

feature of the environment that led to costly outcomes and shaped cognitive mechanisms 

accordingly. Proximally, there are challenges unique to observers of moral acts (as opposed to 

perpetrators or victims, for instance), with one primary challenge being the acquisition of 

evidence without certain knowledge (DeScioli & Kurzban, 2009). Additionally, concerning 

moral judgments it can also be costly to fail to detect a moral violation. The capacity for 

cooperation and peaceful exchange is a foundational characteristic of extended interactions 

among individuals, as well as a prominent feature of person perception (Kurzban & Leary, 

2001). A failure to detect social free riders, or to detect that a person may be deceptive and likely 

to take advantage of someone (e.g., in the instance of a moral transgression), can lead to costly 

outcomes for the person in the relationship as well as their larger social group as a whole. Thus, 

inferring guilt or innocence in the event of a transgression, and subsequently garnering 

appropriate evidence, is a critical part of the process. 

Emotions 

Under uncertainty, we use cues from the environment to help inform our judgments. In 

the realm of moral judgment, factors of the situation and the motives of those involved impact 

our judgments. However, what if no such evidence is available? What if, for instance, you came 

upon the same crowd of people surrounding two individuals and you see one of them strike the 

other in the face. How would you judge the actions of those involved? One salient cue that 

 

6 

people attend to is the reactions of others. Specifically, people attend to their emotional 

reactions.  

Morality and emotions. Emotions are reliably linked to moral reactivity (Avramova & 

Inbar, 2013). The relationship is so consistent that some have posited an emotional reaction as 

necessary by definition for a situation to be subjectively experienced as ‘moral’ (e.g., Prinz, 

2006). While this is a strong position, and primarily a philosophical one, it underscores the 

strength of the association between morality and emotions. Indeed, witnessing moral 

transgressions evokes feelings of disgust, anger, and contempt (i.e., “other-condemning 

emotions”; Haidt, 2003), and these emotions can serve in both the judgment of the transgression 

(Haidt, 2001; 2012) and motivation of behavior on the part of the perceiver (Tangney, Stuewig, 

& Mashek, 2007; but see Blasi, 1999). 

Emotions are a complex, multi-faceted phenomenon involving the coordination of a suite 

of physiological and psychological phenomenon. According to Schwarz & Clore (2007), 

emotion involves three primary components: a cognitive/experiential component (e.g., the 

thoughts and conscious experiences that accompany an emotional reaction) a bodily component 

(e.g., the somatic reactions accompanying an emotion; postural changes, behavior), and an 

affective component (e.g., autonomic reactivity, physiological arousal; the ‘feeling’ of an 

emotion with a broad positive or negative valence). These three systems coordinate to produce 

what is colloquially lumped into an ‘emotional reaction’. However, these three components, 

while associated in occurrence, may serve different adaptive functions.  

Of specific interest to the present investigation are the somatic components of emotional 

reactions. Emotions are strongly and reliably linked to distinct non-verbal behaviors, such as 

facial expressions, posture, and body language. Here I define “non-verbal” as any behavior that 

 

7 

does not utilize language, but may include non-verbal auditory behaviors (e.g., a scream, 

whimper, cheer, gasp, sob, etc.) As a somatic phenomenon, they are outwardly expressive 

displays. They can be seen by others and interpreted as reliable indicators of what a person is 

experiencing emotionally.  

Facial expressions, specifically, are theorized to have evolved to serve a communicative 

purpose: to convey meaningful information to nearby conspecifics about stimuli in the 

environment (Shariff & Tracy, 2011). Facial expressions are outward-facing, exaggerated, and 

are accurately displayed/interpreted across most, if not all, human cultures (Ekman et al., 1987). 

These are design features indicative of an evolved communicative signal (Eibl-Eibesfeldt, 1989). 

Facial expressions and some non-verbal behaviors are also present in our close evolutionary ape 

ancestors (Parr & Waller, 2006), indicating that they may have evolved early in our ancestral 

timeline. Thus, considering the association between moral judgment and emotional reactivity, 

and the link between emotional reactivity and facial expressions, I expect the facial expressions 

of others may serve as evidence to inform moral judgments.  

The Present Studies 

 

Moral judgment, here defined as an evaluation of the severity of a moral transgression, is 

an inferential process requiring the acquisition of evidence to make an appropriate assessment. 

However, little is known about how judgments occur under uncertainty, and how people use 

available evidence during the judgment process. Uncertainty, necessarily present in the judgment 

of others’ internal states (e.g., motives, intent) and typically present to some degree in the 

environment, necessitates the use of relevant information during the judgment process. Thus, I 

seek to examine how moral judgments are informed by available evidence, particularly when the 

situation is uncertain. This examination also assesses whether facial expressions and auditory 

 

8 

cues serve to communicate information during evaluative judgments, and explores the link 

between morality and emotion. 

I designed a series of 4 studies (denoted henceforth as Study 1 - Study 4) as a 

comprehensive test of whether people utilize the emotional content of others’ reactions as 

evidential input during their moral judgments in uncertain situations. As emotion is reliably 

linked to moral transgressions (Avramova & Inbar, 2013) and produces distinct somatic 

reactivity thought to have evolved as a communicative device (Shariff & Tracy, 2011), here I 

utilize facial expressions and vocalizations in an experimental context to assess whether such 

signals serve as inputs to the judgment process. Study 1 was a pilot study to identify appropriate 

moral stimuli to be used in the package of studies that comprise this dissertation. Studies 2 and 3 

examine the role that facial expressions play in forming moral judgments under uncertainty. 

Study 4 will examine whether auditory cues also inform moral judgment under uncertainty in a 

similar way to communicative facial expressions.  

Specifically, I hypothesize that (1) emotional signaling (opeationalized via facial 

expressions and auditory vocalizations) will influence moral severity judgments. When viewing 

images of moral transgressions, emotional signaling will serve as evidence that inform the target 

of relevant contextual evidence. Subsequently, differences in severity judgments of the same 

image will vary systematically with different expressions. I also hypothesize that (2) this 

relationship will be stronger when the image being viewed is more uncertain. Perceivers in 

uncertain judgment contexts are motivated to glean evidence from the environment (e.g., 

Reimanet al., 1989). Communicative expressions, being reliably linked to emotional reactivity 

(Ekman et al., 1987; Anderson, Monroy, & Keltner, 2017; Oveis, Spectre, Smith, Liu, & Keltner, 

2016) and by extension moral reactivity (Avramova & Inbar, 2013), will serve as a salient source 

 

9 

of evidence to inform their judgment. Finally, I will explore how a number of individual 

difference characteristics might moderate the degree to which people assess the uncertainty of 

moral situations, how uncertainty feels to perceivers, and how this may inform moral judgments 

under situational uncertainty. 

Overview of Dissertation 

A pilot study (Study 1; Chapter 2) was first implemented to identify stimuli appropriate 

for use in moral judgment research. Study 2 (Chapter 3) and Study 3 (Chapter 4) examined the 

influence of facial expressions on moral severity judgments in both clear and uncertain contexts. 

Study 2 utilized a within-subjects design where participants viewed morally salient images and 

rated the severity of each violation. Uncertainty was assessed via a self-report scale item for each 

image. To overcome some of the limitations identified in Study 2, Study 3 utilized a between-

subjects design and experimentally manipulated uncertainty. Study 4 (Chapter 5) builds off of 

Study 2’s design, but utilized auditory emotion cues instead of facial expressions. Should 

emotional expression serve as evidence in the moral judgment process, I predicted a similar 

pattern of results as in Studies 2 and 3. 

Finally, I integrate the findings across the four studies into a discussion (Chapter 6) about 

the extent to which people use communicative information to inform moral judgment under 

uncertainty, individual differences in these processes, and directions for future research.  

 

 

 

10 

CHAPTER 2 

Study 1: Pilot Study 

Study 1 was designed to (1) identify appropriate, morally relevant stimuli, and (2) to 

assess whether the stimuli would allow for sufficient variance in both severity and uncertainty 

for use in subsequent studies on moral judgment.  

In the broader literature, moral stimuli range from descriptive paragraphs of interpersonal 

situations to static images, videos, and even virtual reality situations in which participants act in 

real-time to moral dilemmas. However, there are several methodological problems with moral 

stimuli in particular that reduce variance in responding and may inhibit experimental 

manipulations from influencing judgment.  

First, judgments about situations that have been experienced previously (e.g., a trolley 

dilemma scenario) may elicit ‘canned’ responses that are invariant to experimental assessment or 

manipulation. That is, participants have previously reasoned about the situation (or similar 

iterations), have prior knowledge about the types of responses being elicited (or the types of 

responses frequently given by others), and produce a previously formed decision instead of one 

that encapsulates an active in-the-moment decision (e.g., Chandler, Mueller, & Paolacci, 2014). 

This could be due, in part, to the popularity of trolley and other similar dilemmas and stimuli in 

morality research over the past decade, exposure to moral stimuli in broad societal contexts (e.g., 

Brain Games season 5 episode 3, “Morality”, Kolber, Margol, & Nigro, 2015; The Good Place 

season 2 episode 5, “The Trolley Problem”, Schur, M. & D. Holland, 2017) and/or social 

demand characteristics dictating socially acceptable responses regardless of the context of the 

moral act being assessed. Regardless of where people’s problematic responses come from, these 

 

11 

are methodological hurdles to morality research that should be addressed prior to the 

implementation of stimuli in morality research. 

Second, stimuli showing extreme moral transgressions are prone to ceiling (or floor) 

effects. For example, participants will always judge an act like murder or egregious interpersonal 

harm as severe, regardless of the situation or context. This creates a skewed response distribution 

that is difficult to work with statistically, and often violates the normality assumptions inherent 

in many standard statistical tests.  

Finally, moral judgments of descriptive paragraphs or vignettes may not activate the 

psychological constructs of interest, as they do not contain the visual stimuli upon which these 

mechanisms operate. For example, describing a ‘disgusting’ act in words may not elicit the same 

feelings as sensing or witnessing the said act firsthand, as disgust is primarily triggered by 

sensory inputs such as vision, taste, or smell. This is particularly notable in the context of moral 

reactivity, as emotions are posited to serve as input into the judgment process. Should a vignette 

or written stimulus not properly engage related affective systems, the ecological validity of the 

findings may be undermined. 

 

To overcome these potential issues with experimental moral stimuli, I examined 

judgment ratings of a number of interpersonal situations depicted in the International Affective 

Picture System (IAPS; Lang, Bradley, & Cuthbert, 2008). The IAPS has a number of key 

features that make it attractive for morality research. First, the images are well-guarded in the 

research community, such that the stewards protect and isolate the images from the general 

public domain1. Thus, it is unlikely that participants will have seen the stimuli previously. 

Second, they are detailed visual images that capture both interpersonal and impersonal moral 

violations in salient and ecologically valid contexts. As such, the images encapsulate sensory 

 

12 

inputs that are relevant for activating morally relevant psychological constructs of interest. Third, 

the images span a broad range of situations, such that judgments can be elicited from more or 

less severe moral transgressions to examine response tendencies across a variety of stimuli. 

Participants 

Five hundred and three undergraduate students volunteered to participate in a study titled 

“Interpersonal Perceptions” through the online subject pool recruitment website at Michigan 

State University. After excluding those who failed the self-reported attention check (n=7), 

declined to participate after reading the consent (n=22), or failed to complete a sufficient number 

of the image ratings (n=91)2, the final sample size was N=383. Students received course credit 

for their participation. Mean age of participants was M = 19.69 (SD = 2.33), with approximately 

71% identifying as female. The racial composition of the sample included 75% White, 13% 

Asian/Pacific Islander, 7% Black, 4% Hispanic/Latino, and 2% other or not specified.  

Materials & Procedure  

In consultation with the Center for the Study of Emotion and Attention, I was permitted 

access to the IAPS image set for use in these studies. After a visual assessment of the database, a 

subset of 42 images was selected for pre-rating. These images depicted a wide range of acts, 

from neutral or non-moral acts (e.g., a man riding a bike in the countryside) to theft (e.g., a 

woman shoplifting), drinking and driving (e.g., a person behind the wheel of a car drinking a 

beer), threat of interpersonal harm (e.g., a man pointing a gun at another man), and bodily 

mutilation or torture (e.g., a bloodied man being hanged; a deceased, bloodied body of a woman 

who appeared to have been tortured). A full list of IAPS image numbers that were rated in the 

study, as well as a brief description, can be found in the appendix. Of note, because these images 

 

13 

were to be used in conjunction with a facial expression manipulation, the faces of the actors in 

the images were blurred so as not to influence the manipulation in subsequent studies.  

 

Participants arrived at the laboratory for their scheduled session, where a research 

assistant (RA) greeted them. Upon entering the lab, they were escorted to a computer where they 

completed all study procedures. Informed consent was completed by all participants prior to 

participation, followed by a brief instruction screen discussing the process of rating the images. 

 

First, participants viewed the entire set of images in random order without providing any 

ratings. This was intended to prevent anchoring effects for participants, should they view a 

particularly gruesome or startling image first during the rating sequence. Participants were not 

timed and were instructed to simply look at the images and move to the next at their own pace. 

Then they viewed each image a second time, one at a time and in random order. Using a 7-point 

scale (1=not at all; 7=extremely), they provided ratings of the moral severity of the actions 

depicted (“How immoral do you think the picture is?”), how uncertain they found the images 

(“How morally ambiguous do you think the picture is?”), and how morally relevant they thought 

the images were (“How morally relevant do you think the image is?”). Definitions of each 

criterion (e.g., what ‘morally ambiguous’ means) were provided to the participants during the 

instructions to ensure participants understood specifically what types of judgments were being 

solicited. 

They also provided ratings of their emotions toward each image (“How does the image 

make you feel?”) using a 7-point scale (1=not at all; 7=extremely), including happy, sad, angry, 

fearful, physical disgust, and moral disgust. Finally, participants provided basic demographic 

information about their age, race/ethnicity, gender, and education. 

 

 

14 

Results 

 

Table 1 shows the mean ratings for severity, relevance, uncertainty, and each emotion 

rating (happy, sad, angry, fearful, physical disgust, moral disgust). Across all images, mean 

severity ratings ranged from 1.33 to 6.47, showing robust variation in severity across the images. 

Mean relevance ratings ranged from 1.96 to 5.15, which also implied much variation, albeit 

tighter around the center of the scale. Mean ambiguity ratings varied the least, ranging from 2.45 

to 4.11.  

 

Table 2 shows the inter-correlations between the rating variables. Due to the large sample 

size of ratings, nearly every bivariate relationship is significant at the p<.01 level, so direction 

and relative size were used to guide interpretation. Severity and relevance were strongly 

associated at r=.54, indicating that as moral relevance increased, severity also increased. Severity 

was also positively associated with ambiguity (r=.08). However, this relationship was weaker.  

Severity was positively associated with negative emotional reactivity (sad, angry, fearful, 

physical disgust and moral disgust; rs > .64), and negatively associated with happiness (r = -.33). 

Severity correlated most strongly with moral disgust (r = .84). Moral relevance displayed the 

same pattern of results (negative emotions: rs > .34, happiness: r = -.09), and also correlated 

most strongly with moral disgust (r = .48). The negative emotions showed strong inter-

correlations (rs > .74), and a negative correlation with happiness (rs > -.23).  

Distributions. As anticipated, a number of images displayed highly skewed distributions 

for severity judgments. Images of severe moral transgressions displayed strong negatively 

skewed distributions (i.e., ceiling effects), while neutral images displayed patterns of strong 

positive skew (i.e., floor effects). However, a number of the intermediately severe images 

 

15 

showed a relatively normal and varied distribution, ideal for use in statistical models. An 

example of each of the image rating distributions can be found in Figure 1. 

Discussion  

 

Having assessed the variance in severity and ambiguity of selected IAPS images, a subset 

of 16 images was identified as appropriate for use in Studies 2-4 based on (1) a visual inspection 

of the response distributions and q-q plots, such that images with the highest level of normality 

and least amount of skew were selected, (2) inspection of ambiguity ratings, such that a range of 

ambiguity levels were selected, and (3) the actions depicted in the images, such that the selection 

included a range of interpersonal and impersonal actions. 

 

 

 

16 

CHAPTER 3 

Study 2 

 

Study 2 was designed to directly test the hypotheses of interest: First, that emotional 

expressions conveyed by others will influence moral severity judgments by informing the target 

of relevant contextual evidence; Second, that this relationship will be strongest when the 

situation is uncertain, as perceivers with imperfect knowledge should glean and subsequently 

utilize the information contained in the emotion to inform their judgment.  

Emotional evidence was operationalized using images of facial expressions; 

communicative signals (e.g., Shariff & Tracy, 2011) are theorized to have evolved to 

communicate information about the environment to others. Moral judgments were made via a 

subset of IAPS images selected based on the results of Study 1. Sixteen images were chosen 

based on (1) the relative normality of the response distributions, so as to avoid ceiling and floor 

effects and normality violations; (2) their ambiguity ratings, such that variance in uncertainty 

was present across the images; and (3) the explicit content of the images, to ensure that a range 

of moral transgressions was present across the images. Further, one neutral image (2026) and one 

extreme image (3530) were selected in order to examine responding in the most extreme and 

least extreme cases and serve as a valid responding check. Specifically, we applied the following 

criteria as indicators of invalid responding: anyone who rated the neutral image as extremely 

morally severe (i.e., 90-100 on a 0-100 scale) or indicated the extreme image as not at all morally 

severe (i.e., 0-10 on a 0-100 scale) was excluded from the study. 

Participants 

Two-hundred sixty-four undergraduate students volunteered to participate in a study 

titled “Interpersonal Perceptions 2” through the online subject pool recruitment website at 

 

17 

Michigan State University. Individuals who participated in Study 1 were barred from 

participating in the present study, due to previous exposure to the judgment stimuli. After 

excluding those who failed the self-reported attention check (n=5), failed to complete a sufficient 

number of the image ratings (n=5), or failed the aforementioned valid responding check (n=6), 

the final sample size was N=248. Students received course credit for their participation.  

The mean age of participants was M = 19.77 (SD = 1.77), with approximately 57% 

identifying as female. The racial composition of the sample included 63% White, 21% Asian, 

11% Black, and 5% other or not specified.  

Materials & Procedure 

Facial expression stimuli. The facial expressions implemented in the study were happy, 

neutral, anger, and disgust. These are basic, universal facial expressions (i.e., Ekman et al., 1987) 

that are reliably recognized in all human societies and cultures. As such, they should be reliably 

recognized by all participants in the study regardless of race, ethnicity, or country of origin. 

Anger and disgust, in particular, are defined as ‘other-condemning’ emotions (Haidt, 2003). 

They’re linked to the condemnation of moral transgressions performed by others. As opposed to 

subtler other-condemning expressions such as contempt or self-condemning emotions such as 

shame, anger and disgust should serve as strong emotional signals to perceivers of a 

transgression. A neutral expression was included as a control expression against which the others 

could be compared. Finally, a happy expression was included to explore the possible influence of 

a positive emotion on the judgment of the images. 

Facial expression images were taken from the Extended Cohn-Kanade AU-Coded 

Expression Database (CK+; Kanade, Cohn, & Tian, 2000; Lucey et al., 2010). Upon agreement 

with the terms of use, these images are free to download and use for research purposes. The 

 

18 

image set contains black-and-white and color photos of men and women performing a range of 

externally validated emotional expressions (e.g., happy, sad, fear, surprise, disgust, neutral).  

Upon visual inspection of the database, not all subjects in the database had expressions 

for all of the target emotions, and some of the expressions were not visually distinct and clear 

across emotions. For the present study, I chose a single white male target (S037) in black-and-

white. S037 was chosen because he had distinct and exaggerated expressions for the target 

emotions (anger, disgust, neutral, happy). I chose a white male target in order to control for race 

and variance in the target face across judgment scenarios.  

 

Judgment stimuli. Sixteen images from the IAPS were chosen from the 42 piloted 

images. As discussed in Study 1, the images were selected for their severity rating distributions, 

variance in ambiguity ratings, and variety of actions depicted. The list of images used in Study 2, 

with IAPS number and short informal descriptions from Study 1, can be found in Table 3 of the 

appendix. 

 

Judgment task. Similar to Study 1, participants arrived at the laboratory for their 

scheduled session where an RA greeted them. They were escorted to a computer where they 

completed all study procedures. Informed consent was completed by all participants followed by 

a brief instruction screen discussing the process of rating the images. 

 

Participants first viewed each of the 16 IAPS images in random order with no time limit. 

The images were not accompanied by the facial expression stimuli. As in Study 1, this first 

viewing was intended to show participants all of the images before rating them, to prevent 

anchoring effects should they be randomly assigned to rate the most severe image first. Then 

participants viewed each image a second time, where they were asked to provide their judgment 

and emotion ratings. In this second viewing, each image was accompanied by a facial expression 

 

19 

with two images positioned immediately side-by-side, facial expression on the left and the target 

image on the right. Participants were instructed to “…answer the questions in regards to the 

image on the right only”. This study utilized a within-subjects design, in which images and 

expressions were fully crossed and randomized, such that each participant provided a total of 64 

image ratings (16 target images X 4 expressions).  

For each image-face pairing, participants were asked to make three judgments. First, 

participants made a severity judgment about the actions in the picture (“How immoral are the 

actions in the picture?”), which served as the primary criterion of interest. Second, they were 

asked to judge how they thought others would judge the image (“Regardless of your own moral 

assessment of the image, how immoral do you think other people would find the actions in the 

picture?”). This was asked as a separate corroborative assessment of their judgments, to avoid 

potential social desirability bias in their assessments. Finally, they were asked how ambiguous 

they thought the actions in the image were (“How morally ambiguous do you find the actions in 

the picture?”). This was designed to assess uncertainty in the situation. Each of the three 

judgments were rated on a horizontal sliding scale ranging from (0=Not at all) to 

(100=Extremely), with scale marks at 0, 25, 50, 75, and 100.  

Participants also provided ratings of their emotions toward each image (“How does the 

image make you feel?”) using a 7-point scale (1=not at all; 7=extremely), including happy, sad, 

angry, fearful, physical disgust, and moral disgust. 

After every 16-image rating set, participants were given a brief filler task that involved 

one of the following: solving three basic arithmetic math problems, writing a brief paragraph 

description of their morning routine, or writing a brief paragraph description of their evening 

routine. These were implemented to try and distract them from the specifics of the faces/images 

 

20 

they just saw, as well as break up the rating task into more manageable chunks. After the image 

ratings were completed, participants completed basic demographic questions about their age, 

gender, and race/ethnicity. Once the survey was completed, they were escorted out of the 

laboratory and granted compensation for their participation. 

Results 

 

Table 3 displays the severity and uncertainty means for each of the images, as well as the 

grand means across all images. Images were dummy-coded prior to analysis, and IAPS #2026 

(neutral image) was used as the reference category. Across all images, the mean severity rating 

was M = 54.53 (SD = 33.62); about middle of the scale. Uncertainty fell just below the middle of 

the scale, M = 42.38 (SD = 31.71). Examining the mean severity ratings of each image, there 

appeared to be a wide range of severity ratings as well, ranging from the neutral image IAPS 

#2026 (M = 14.59, SD = 19.74) to the most severe image #3530 (M = 86.17, SD = 16.55). Figure 

2 visually sorts the image severity ratings from most to least severe, highlighting the significant 

variance in ratings across images. 

 

Table 4 displays the mean severity ratings for each facial expression condition. Across all 

images, the means are strikingly similar: neutral (M = 54.32, SD = 33.68), happy (M = 54.82, SD 

= 33.53), angry (M = 54.63, SD = 33.38), and disgust (M = 54.77, SD = 33.53). Indeed, it appears 

there was little difference between expressions when examined across the images. Looking again 

at Figure 2, it is clear that there is variability between images, but within each image there is little 

variability between the facial expression conditions. 

 

Table 5 displays the bivariate relationships between severity ratings, other ratings, and 

uncertainty ratings. To control for the repeated measurement of variables within participants, 

each variable was mean-centered by subtracting the participant’s overall mean on the variable 

 

21 

before estimating the correlation. Due to the high-powered nature of the sample and study 

design, nearly all correlations were significant at the p<.05 level. Therefore, I examined the 

relative size and direction of each pairing to look for patterns of relationships.  

As expected, severity and others’ ratings correlated strongly (r = .82). Severity and 

others’ ratings both correlated with uncertainty, albeit weakly (r’s = .05 and .08, respectively). 

The negative emotion ratings of sad, angry, fearful, physical disgust and moral disgust appear to 

correlate together tightly (rs > .60), indicating a tendency to rate similarly across these emotions. 

Happy was negatively associated with all variables (rs > -.12). 

Severity was associated with the negative emotions such that as severity increased, these 

emotions also increased (rs > .57). Severity was also negatively associated with happiness (r = -

.20). This same pattern was found for others’ ratings as well, although associations were weaker 

overall. Finally, uncertainty correlated positively with the negative emotions, such that as 

uncertainty increased, the negative emotions tended to as well (and happiness tended to 

decrease). However, the relationships with uncertainty were the weakest across the entire set of 

bivariate relationships. 

 

Main hypotheses. To test the hypothesis that emotional expressions conveyed by others 

will influence moral severity judgments by informing the target of relevant contextual evidence, 

I regressed severity ratings over facial expression condition (e.g., anger, disgust, etc.), image, 

and uncertainty. Two-way interactions between face condition x image and face x uncertainty, 

and image x uncertainty were also included. The face x uncertainty interaction was included as 

the key predictor, such that facial expressions would inform moral judgments under particularly 

uncertain situations.  

 

22 

Because participants saw multiple face-image sets, there is a source of non-independence 

in the data that needed to be modeled. Thus, we specified a random coeffient model by using the 

SPSS MIXED function, appropriate for repeated measure designs with nested data. I utilized 

maximum likelihood (ML) estimation with a covariance structure of homogeneous compound 

symmetry (CSR)3, Type III sums of squares. As every participant rated every image x face 

combination, these were modeled as repeated effects nested within participants. The face x 

image interactions were also modeled to examine whether the effect of facial emotions varied 

across the IAPS stimuli. I also included predictors of gender and race to examine potential 

demographic influences on severity ratings. Because the racial groups had drastically uneven 

group sizes and the overall sample was 63% white, race was re-coded as a binary indicator 

(white/non-white). All significance tests were performed using two-tailed tests with α = 0.05. 

 

Table 6 shows the criterion ratings across each predictor, including the omnibus tests for 

each factor and specific parameter estimates. As seen in the omnibus test outcomes, there was a 

significant effect of image (F(15, 14613) = 724.48, p < .001). That is, there were significant 

differences in severity ratings across the images being rated. This makes intuitive sense, as they 

were selected specifically to provide a variety of stimuli to be rated by participants.  

There was also a significant overall effect of uncertainty (F(1, 14816) = 10.30, p = .001). 

As seen in Table 6, higher levels of uncertainty are associated with higher levels of severity. That 

is, more uncertainty leads to slightly more severity, β = .14, t(14678) = 5.58, p = .001. This 

relationship is consistent with the bivariate relationship between uncertainty and severity (r = 

.05, p < .05) presented in Table 5.  

Finally, there was a significant image x uncertainty interaction (F(15, 14662) = 23.07, p 

< .001), indicating that the relationship between severity and uncertainty varied across the 

 

23 

different images. Again, this makes intuitive sense as the images were also selected to vary on 

their level of ambiguity as well as their severity.  

There were no significant effects of face (F(3, 14606) = 0.33, p = n.s.), face x uncertainty 

(F(3, 14614) = 0.33, p = n.s.), or face x image (F(45, 14606) = 0.37, p = n.s.) on moral severity 

of the images. That is, the facial expression conditions did not influence participants’ overall 

severity judgments and did not interact with uncertainty or specific images. 

Regarding demographic differences, there were no effects of gender (F(1, 243) = 3.25, p 

= .073) or race (F(1, 243) = 3.48, p = 064). While the data show that men (M = 52.66, SD = 

32.75) rated the images as less severe than women (M = 56.09, SD = 34.01), this difference was 

not statistically significant. 

Exploring alternatives. Considering the nature of morality research, and the vague 

colloquial definitions for terms like moral, ambiguous, and uncertain, it’s possible these may be 

interpreted differently by native English speakers and non-native English speakers. Thus, I asked 

participants at the end of the survey whether English was their first language. The direction and 

pattern of significance for each predictor was the same for both groups and consistent with what 

is reported above.  

I also collected participants’ ratings of how they thought other people would rate the 

severity of the images. Other individuals’ ratings may serve as a better measure for how 

emotions might be communicated to broader groups rather than individuals. I then ran the same 

model specified above but using others’ ratings instead of participants’ own severity ratings. 

Again, the pattern of results was the same as that specified above.  

 

 

 

24 

Discussion 

 

The primary hypothesis tests indicated there were no observed effects of facial 

expressions on severity judgments and no effects of facial expressions on severity under 

uncertainty. While these results may indeed reflect the state of the relationship between facial 

expressions’ communicative function and moral judgment severity, a number of methodological 

issues were identified that may have contributed to the results. These are discussed and 

addressed in Study 3, which sought to improve upon the present study to better test the primary 

hypotheses of interest. 

 

25 

CHAPTER 4 

Study 3 

After examination of the results of Study 2, I identified some methodological flaws that 

Study 3 was designed to overcome.  

First, uncertainty was measured via self-reported ambiguity of each image. While the 

immorality of some actions and images were inherently more or less ambiguous, this was 

confounded with individual variance in how participants perceived the images and how they 

personally felt about the actions in each image. Subsequently, Study 3 experimentally 

manipulated uncertainty using a blurred version of each photo in order to impose situational 

uncertainty. Further, uncertainty was changed to a between-subjects factor so that participants 

would not be able to recognize the blurred versions of the images based on their viewing of the 

unaltered images. 

Second, Study 2 presented the facial expressions side-by-side with the target images. 

While this was a straightforward implementation of the facial expression manipulation, the 

frequently changing faces being repeated across all 64 rating instances may have led participants 

to ignore the changing expressions. Further, due to the obvious presentation of the facial 

expressions as an experimental feature, and the exposure to all the facial expressions, 

participants may have disregarded the evidence in an effort not to be manipulated by their 

influence. Indeed, the presence of the face next to the target image allowed participants to fixate 

on it and consciously disregard its effects as non-credible (e.g., Schwarz & Clore, 2007). To 

overcome these issues, Study 3 presented each face briefly and prior to the presentation of each 

judgment stimuli. 

 

26 

Finally, the number of images implemented in the study was reduced from 16 to 8, and a 

distractor task was provided after every judgment task. This was designed to prevent carryover 

effects from the previous faces and images, and provide cleaner judgments about each image 

individually. 

Aside from the aforementioned changes, the study procedure, questions, and materials 

were identical to that of Study 2. Therefore, I only present the parts of Study 3 below that 

diverge from Study 2’s procedure.  

Participants 

Three-hundred fifty-eight undergraduate students volunteered to participate in a study 

titled “Interpersonal Perceptions 3” through the online subject pool recruitment website at 

Michigan State University. Individuals who participated in Study 1 and Study 2 were barred 

from participating in the present study, due to previous exposure to the judgment stimuli. After 

excluding those who failed to complete a sufficient number of the image ratings (n=5) or failed 

the judgment validation checks (n=2), the final sample size was N=351. No participants failed 

the self-reported attention check. Participants received course credit for their participation.  

The mean age of participants was M = 19.33 (SD = 2.39), with approximately 81% 

identifying as female. The racial composition of the sample included 73% White, 14% Asian, 

8% Black, and 5% other or not specified.  

Methods & Procedure 

 

Facial expression stimuli. The facial expressions were presented immediately prior to 

each target image for approximately 0.5 seconds. This allowed sufficient time for participants to 

view and register the facial expression before transitioning immediately into the judgment task. 

The same expressions (from subject #S037) were implemented, with the exception of the happy 

 

27 

expression. In Study 2, the happy expression did not influence judgments differently from the 

neutral expression, and it was removed for brevity in Study 3 (although ratings for all the 

emotions were retained). The presentation of the expressions was randomized across target 

images and participants. 

Judgment stimuli. Eight of the 16 images were selected from those utilized in Study 2. 

The image number was reduced to decrease the number of judgment instances and help attenuate 

participant fatigue in making repetitive judgments. The particular images were chosen for their 

diversity of moral transgressions4, as well as their distinctiveness in the blurred condition (see 

below).  

Judgment under uncertainty. Uncertainty was experimentally manipulated by blurring 

the images using the Picture Tools function in Microsoft Word. The goal was to alter the images 

so that it was still apparent a moral transgression was occurring, but the details of which were 

unclear, creating a more ambiguous target image for judgment. After examining a number of blur 

options, it was determined that a 50% blur filter would be appropriate. This allowed the actions 

and context in the image to be discernible, but fuzzy enough so that the details could not be 

distinguished. For example, in the image of workplace harassment, it was clear that a female was 

sitting at a desk in a professional setting, with a male standing behind her with his hand on her 

shoulder. However, the details of their interaction and body language were difficult to discern, 

creating ambiguity in the nature of the interaction. The result is that it’s apparent that the 

situation is in some way ‘uncomfortable’, but it’s less obvious what exactly is happening in the 

situation. 

Judgment task. The procedure followed the same format as in Study 2 but with the 

following changes: First, participants were not presented with the images prior to the rating of 

 

28 

them. This was to prevent participants from forming preconceptions of the images prior to 

viewing them for judgment. Second, participants were randomly assigned to either the 

uncertainty (blurred images) condition or the clear (unmodified images) condition, where they 

would see either 8 blurred target images for judgment or 8 clean images. Images were presented 

in random order within each condition. Each rating was followed by a filler task to distract 

participants from the judgments and attenuate any carryover effects image-to-image. Participants 

were first presented with a 6-digit number and asked to memorize it. The next screen asked them 

to solve three arithmetic problems (e.g., (3 x 12) -7 = ?). After, they were asked to recall the 

number. 

Finally, the ratings of others’ severity were removed (as self- and other-ratings yielded 

similar findings in Study 2), and the ambiguity question was changed slightly: “How certain are 

you of what’s happening in the picture?” (0=Not at all, 100=Extremely). Higher values indicate 

higher certainty.  

Results 

 

Table 7 shows the mean severity and uncertainty ratings for both the clean and blurred 

image condition. The general mean trends reveal that clean images are rated as more severe than 

the blurred images, except for two images (IAPS # 2026 and 4233). A similar pattern emerged 

for certainty ratings—clean images are rated as more certain than the blurred images, except for 

three images (2026, 4233, and 9102).  

The facial expression conditions show similar severity ratings across all the images, 

neutral M = 57.49 (SD = 33.90), disgust M = 58.73 (SD = 33.97), and anger M = 58.56 (SD = 

33.82). Within the different images, the expressions don’t show any clear patterns. Some images 

show the expected pattern of emotions on severity judgments (e.g., neutral < anger & disgust; 

 

29 

IAPS # 9423). However, some images show an ascending pattern where neutral > disgust > 

anger (3181), or a more ‘flat’ distribution where there are little or no differences across the 

conditions (9102).  

Table 8 displays the aggregated bivariate correlations between severity, certainty, and 

emotion ratings across all images. The results show a very similar pattern to Study 2: the 

negative emotions correlate together strongly (rs > .58); happiness was negatively related to the 

negative emotions, severity, and certainty (rs > -.04). Severity and certainty were strongly 

correlated (r = .57), such that higher certainty was associated with higher severity ratings across 

conditions and images.  

Main hypotheses. To assess whether the facial expressions influenced severity 

judgments, I implemented a similar procedure as outlined in Study 2 with severity regressed 

across facial expression condition (face), image, and uncertainty (blur) condition using the SPSS 

MIXED model procedure (restricted maximum likelihood, covariance type: compound symmetry 

correlation metric, type III SS). The facial expression and image factors were modeled as 

repeated effects nested within participants. I again included predictors of gender and race to 

examine potential demographic influences on severity ratings, and race was re-coded as a binary 

indicator (white/non-white). All significance tests were performed using two-tailed tests with α = 

0.05.  

To assess whether facial expression was more influential under conditions of uncertainty, 

I estimated two-way interactions between face and certainty (clean v. blur), image and certainty, 

and face and image. Finally, I modeled a three-way interaction of face x image x certainty to 

examine whether the relationship between facial expressions and level of certainty varied from 

image-to-image. Certainty was modeled as a between-subjects factor. 

 

30 

Table 9 shows the results of the model, including the omnibus tests for each factor and 

specific parameter estimates. Overall, the facial expression factor was not significant (F(2, 2642) 

= 1.21, p = n.s.), and the interaction terms were also not significant: face x certainty (F(2, 2642) 

= 0.19, p = n.s.), face x image (F(14, 2639) = 1.40, p = n.s.), face x certainty x image (F(14, 

2639) = 0.89, p = n.s.). Thus, it appears the facial expressions do not have a significant influence 

on severity judgments for the images. 

Image was found to be a significant factor, F(7, 2402) = 223.60, p < .001. This is 

expected and consistent with Study 2, as the images were selected for their variation in severity 

and content. The certainty (blur) manipulation was also significant, F(1, 343) = 45.27, p < .001. 

The blurred images were rated as significantly less severe (M = 53.39) than the clean images (M 

= 63.22). Further, the interaction between image and certainty condition was also significant, 

F(7, 2403) = 44.90, p < .001). Decomposing this interaction, the clean images are consistently 

rated as more severe than the blurred images, except for the two cases noted earlier (2026 and 

4233) where the blurred images are rated as more severe. All mean differences are significant 

except for images 4233 and 9102. These effects are indicated on Table 7 and significant 

differences between means are denoted in bold (also see Figure 3). 

Demographics. Race was not a significant factor in Study 3, F(1, 347) = 0.73, p = n.s. 

Further, gender was not a significant predictor, F(1, 345) = 3.13, p = n.s. To further test possible 

demographic differences, I also ran the model including interaction terms of gender with image, 

face, and blur condition. None of these interactions were significant and were not included in the 

final model. 

 

 

 

31 

Discussion  

 

A similar pattern emerged here as in Study 2. Certainty was a significant predictor, such 

that those who saw the clean images rated them as significantly more severe than the blurred 

images except for the neutral image 2026, where the reverse was found. In that instance, it may 

be the case that obscuring the details of an innocent image created a sense of potential suspicion, 

as the image depicts a woman standing at a store checkout counter. However, the ratings in both 

instances were still quite low (Mblur = 17.67, Mclean = 7.82). In the other instances, it appears the 

blur reduced the severity of the perceived actions. This might be due to a reduction of culpability 

for the targets in the image or an apprehension to make a judgment without sufficient evidence. 

This is also congruent with the bivariate relationship between severity and self-reported certainty 

(r = .57), such that as certainty increased severity also increased.  

 

Facial expression, gender, and race were not significant factors in Study 3, replicating the 

results from Study 2. This may be due to methodological limitations of the study, the stimuli 

used, or the undergraduate convenience sample. However, to ensure that the lack of evidence for 

emotional signals is not the result of the use of the specific visual expression stimuli used, Study 

4 assessed moral severity judgments using auditory expressions to examine a different sensory 

input channel and experimental stimulus. 

 

 

 

32 

CHAPTER 5 

Study 4  

Both Studies 2 and 3 did not support the primary hypotheses that signals from the 

environment influence moral severity judgments, particularly under situations of uncertainty. 

While facial expressions are hypothesized to signal information about the environment to 

perceivers, it could be the case that the specific expressions used in Studies 2 and 3 were 

ineffective in eliciting the effect. Participants may have found the facial expression and target 

image pairings contrived, unbelievable, or may have perceived the facial expressions as 

unrelated to the scenes depicted in the images (i.e., spatially and/or temporally disconnected 

from their laboratory-based judgments). 

Study 4 attempts to circumvent potential methodological issues with the facial 

expressions by implementing auditory expressions instead of visual facial expressions to 

communicate emotion. Emotional expressions are associated not only with physiological and 

somatic signals, but auditory sounds as well. For example, this could include cheering, jeering, 

yelling, screaming, or laughing, to name a few. Research in developmental psychology also 

shows evidence of an auditory expression effect in infants. Mumme, Fernald & Herrera (1966) 

found that pre-verbal infants were able to pick up on auditory expression (happy, neutral, or fear) 

and displayed avoidant behavior indicative of the fear responses given by the mother. Study 4 

attempts to test the same hypotheses but implements an angry and neutral crowd sound 

manipulation as the expressions, as outlined below. 

Participants 

Seven-hundred and one undergraduate students volunteered to participate in a study titled 

“Interpersonal Perceptions 4” through the online subject pool recruitment website at Michigan 

 

33 

State University. Individuals who participated in Studies 1, 2, or 3 were barred from participating 

in the present study, due to previous exposure to the judgment stimuli. After excluding those 

who failed to complete a sufficient number of the image ratings or experienced technical 

difficulties (n=33), failed the judgment validation checks (n=50), failed the self-report attention 

check (n=5), the final sample size was N=613. Participants received course credit for their 

participation.  

The mean age of participants was M = 19.51 (SD = 1.98), with 72.3% identifying as 

female. The racial composition of the sample included 68% White, 14% Asian, 11% Black, and 

8% other or not specified.  

Methods & Procedure 

Auditory expressions. The auditory expressions were comprised of ambient ‘crowd 

sounds’ extracted from the internet (which were free for non-profit use). There were two 

conditions: angry and neutral. The neutral crowd sounded like a busy café or common crowded 

public area. It sounded as if there were a lot of people passively chatting and holding 

conversation. Particular words or snippets could be heard here or there, but no continuous 

conversations were noted. The angry condition, however, sounded like a crowd of jeering 

protesters. It sounded similar to the neutral condition but with bouts of shouting and yelling. 

Participants were randomly assigned to crowd condition, which was subsequently treated as a 

between-subjects variable in all analyses. 

Judgment stimuli. Seven of the 8 images used in Study 3 were used in Study 4.5 This 

included the same clean/blur variants for the uncertainty manipulation.  

Judgment task. The procedure followed the same format as in Study 3 but with the 

following changes: First, participants were recruited in groups of 5 and seated at computer 

 

34 

workstations with barriers between them. Second, participants wore sound-isolating headphones 

for the judgment rating portion of the study. After participants reviewed the informed consent 

but before viewing the judgment stimuli, the research assistant started the randomly assigned 

crowd condition audio, which played for the duration of the judgment task. Prior to the judgment 

rating items, participants were instructed with the following: “Audio will be playing for the 

entirety of the session. While evaluating the scenes in the images try your best to imagine the 

audio is part of the scene, coming from a crowd of people out of view, as if these people are 

viewing the same actions depicted in the image as you are.” This was to ensure participants were 

engaged with the audio and connected it to the rating images being shown. Once the judgments 

were complete, participants were instructed to remove the headphones to complete the latter 

portion of the experiment.  

In order to assess comfort and concern with uncertain judgments, participants were also 

asked two additional items (0=Not at all, 100=Extremely): 1. “Imagine you are on a jury and this 

scene is submitted as evidence for your jury decision. How comfortable would you be with 

making a judgment based only on this evidence?” 2. “Imagine your friend asked you to judge the 

actions in the scene, and then shared your response on social media to a local news outlet under 

your name. How worried would you be if the people involved in the scene found out about your 

judgment?” 

Finally, to ensure the angry crowd manipulation was perceived as distinct from the 

neutral crowd, a manipulation check was performed at the end of the survey which asked 

participants “How did the crowd sound in the background? In a few words, describe what you 

think they sounded like”. 

 

 

35 

Results 

 

For the manipulation check, responses were coded as ‘angry’ if they mentioned anything 

about the crowd sounding affectively negative or performing a negative act. For example, any 

mention of the crowd sounding angry, upset, perturbed, shouting, jeering, or enacting a riot, 

protest, or revolt was coded as ‘angry’. Examining the results, 77% of the responses in the angry 

crowd condition were coded as angry, while only 7% were coded as such in the neutral crowd 

condition. This is highly suggestive that the angry crowd sounds were distinctly negative when 

compared to the neutral crowd sounds. 

 

Table 10 shows the mean severity ratings for the angry and neutral crowd condition as 

well as the clean and blurred image condition. Across all images, the general mean trend 

revealed that clean images were rated as more severe than the blurred images (Mclean = 60.31, 

Mblur = 52.66), and participants in the angry crowd condition rated the images as more severe 

than those in the neutral crowd condition (Mangry = 57.71, Mneutral = 55.69). 

Table 10 also displays the bivariate correlations between severity, certainty, comfort, 

worry, and emotion ratings across all images, aggregated within participants. The results show a 

very similar pattern to Studies 2 and 3: the negative emotions correlate together strongly (rs > 

.58), and severity and certainty were strongly correlated (r = .57), such that higher certainty was 

associated with higher severity ratings across conditions and images. Happiness was negatively 

correlated with the negative emotions (rs > -.08), and negatively correlated with severity (r = -

.25) and certainty (r = -.09). 

Main hypotheses. To assess whether auditory expressions influenced severity 

judgments, I implemented a similar procedure as outlined in Study 3 with severity regressed 

across crowd condition, image, and certainty (blur) condition using the SPSS MIXED model 

 

36 

procedure (maximum likelihood, covariance type: unstructured, type III SS). The image factor 

was modeled as a repeated effect nested within participants, while crowd condition and 

uncertainty condition were modeled as between-subject effects. I again included predictors of 

gender and race to examine potential demographic influences on severity ratings, and race was 

re-coded as a binary indicator (white/non-white). All significance tests were performed using 

two-tailed tests with α = 0.05.  

To assess whether the crowd expression was more influential under conditions of 

uncertainty, or more influential for certain images, I estimated two-way interactions between 

crowd and certainty, image and certainty, and crowd and image. Finally, I modeled a three-way 

interaction of crowd x image x certainty to examine whether the relationship between facial 

expressions and level of uncertainty varied from image-to-image.  

Table 11 shows the results of the model, including the omnibus tests for each factor and 

specific parameter estimates. Overall, the crowd factor was not a significant predictor of moral 

severity (F(2, 610) = 1.18, p = n.s.). Further, the crowd x blur interaction (F(2, 609) = 0.03, p = 

n.s.) and the three-way interaction (F(16, 608) = 1.61, p = n.s.) were not significant. 

Consistent with studies 2 and 3, the image factor was found to be significant, F(6, 608) = 

564.17, p < .001, demonstrating the variation in moral severity the scenes portray. The certainty 

(blur) manipulation was also significant, F(1, 610) = 41.31, p < .001, with participants 

consistently rating the  blurred images less severely than the clean images (see Table 10). 

Further, the interaction between image and certainty was significant, F(6, 608) = 69.55, p < 

.001). Decomposing this interaction, the clean images are consistently rated as more severe than 

the blurred images (p<.05) except for two cases (IAPS image #3181 and #4233), for which there 

were no differences. 

 

37 

Finally, the crowd x image interaction was significant, (F(6, 608) = 2.47, p = .023). 

Decomposing this interaction, images #2026 (neutral) and #2691 (protester) showed heightened 

severity ratings in the angry crowd condition relative to the neutral crowd. The other images did 

not show this effect. See Figure 4 for a visual examination of mean differences.6 

Exploring alternatives. To further examine the influence of expressions on moral 

judgments, I used the same modeling procedure to predict participants’ responses to the two 

additional outcome measures (comfort with making a jury decision based on the evidence, and 

their concern if their judgment would be available on social media). See Table 10 for descriptive 

statistics. 

For the comfort criterion, certainty (F(1, 608) = 53.42, p < .001), image (F(6, 609) = 

185.36, p < .001), image x certainty (F(1, 609) = 66.71, p < .001), and race (F(1, 610) = 16.56, p 

< .001) were statistically significant. Unlike with severity judgments, the crowd x image effect 

was not significant (F(1, 609) = 0.85, p = n.s.). Aside from that latter point, the pattern of results 

is largely consistent with the other criterion outcomes across this study as well as studies 2 and 3.  

For the concern criterion, only image (F(1, 604) = 31.25, p < .001) and image x certainty 

(F(1, 604) = 3.17, p = .005) were significant. Taken together, these indirect outcome measures 

may tap into slightly different aspects of the judgment process, but they are no more influenced 

by emotional expressions than the severity judgment outcome. 

Finally, I examined the primary model using a subset of the data including only those 

who rated the angry crowd as ‘angry’. This was intended to isolate the analysis only to those who 

actually perceived the angry crowd as angry. However, the pattern of significance of the results 

was the same as the overall model and did not show any deviations from the original model. 

 

38 

Demographics. Gender was not a significant factor in Study 3, F(1, 610) = 0.01, p = n.s. 

However, race was found to be a significant predictor, F(1, 620) = 5.48, p = .020. Across images 

and conditions, non-white participants (M = 58.23, SD = 33.06) rated the images as more severe 

than white participants (M = 55.96, SD = 33.87). 

Discussion 

 

Overall the pattern of results was fairly consistent with the previous studies, with the 

exception of race and an image x crowd interaction predicting severity. As race was not found to 

be a significant predictor across Studies 2 and 3, this may be a random sampling effect such that 

non-white participants in this particular sample happen to find the images more severe than 

previous samples. Indeed, non-white participants found the images less severe than white 

participants in Study 2, and more severe in Study 3, further ambiguating the effect found in 

Study 4. As a result, this particular race finding should be interpreted with caution.  

 

As for the image x crowd interaction, it appears the crowd expressions did have an effect 

on severity judgments for certain images. IAPS image #2026 is a neutral image depicting a 

woman standing at a store checkout counter. In the absence of other evidence in the image 

depicting an immoral act, the crowd may have supplied the needed inputs to evoke suspicion in 

the perceivers and inflate severity ratings. IAPS image #2691 depicts a protester in the streets 

throwing a rock toward an unknown target. It is possible that because there is a crowd in the 

image, and a protest is a typical place for a crowd, the match between the target stimulus and the 

crowd expression supplied relevant evidence for the judgment being made. These effects did not 

vary by level of certainty (e.g., the crowd x certainty x image interaction). The other images did 

not depict scenes that would necessarily have a crowd present, or not to the level of congruence 

that this particular image of a protest encapsulates.  

 

 

39 

CHAPTER 6 

Discussion and Conclusion 

Summary of Findings 

 

The present studies assessed whether emotional expressions (both visual and auditory) 

influence moral severity judgments, including when the judgment is being made under uncertain 

conditions (defined here as a lack of clarity in available information about the situation at hand).  

 

Study 1 was designed to assess morally relevant stimuli, and participants rated a subset of 

42 images from the IAPS image database. Of those, a subset of images was selected for use in 

the subsequent studies that demonstrated normal rating distributions and presented adequate 

variability across both severity and ambiguity. We also ran a validation study to confirm that the 

images elicited a sufficient amount of moral reactance.4 Indeed, these properties were also seen 

across Studies 2, 3, and 4; the images demonstrated variance in their severity ratings, as shown 

by the statistically significant differences in severity ratings across images. 

 

Uncertainty was also found to have a significant impact on severity ratings. In Study 2, 

images with higher self-reported uncertainty were rated as significantly less severe than those 

with lower uncertainty. In Studies 3 and 4, the images with higher experimentally manipulated 

uncertainty (i.e., the ‘blurred’ images) were rated significantly less severe than the clean images. 

In testing the primary hypothesis, across Studies 2 and 3 facial expressions (anger, 

disgust, vs. neutral expressions) were not found to influence judgments, either in a clear or 

uncertain context.  

In Study 4 the expression stimulus was changed from a visual expression to an auditory 

expression (angry vs. neutral crowd sounds). The contextual uncertainty still did not have any 

influence on severity ratings. However, the angry crowd did increase severity ratings for two of 

 

40 

the images (IAPS #2026 and 2691). Thus, it may be the case that under certain circumstances, 

external expressions from others may provide relevant evidence for moral severity judgments.  

Discussion 

 

I set out to test whether third-party emotional expressions influenced the moral severity 

judgments of first-party perceivers of immoral actions, specifically when the act being assessed 

is uncertain or lacks clarity of information. Facial expressions, theorized to be evolved 

communicative devices (Shariff & Tracy, 2011), did not appear to influence perceivers’ severity 

judgments in either clear or uncertain contexts. This could have been a methodological artifact; 

perhaps there was a disconnect between the target rating stimulus and the facial expression, or a 

single facial expression of one individual was not strong enough evidence to influence 

perceivers’ judgments. Considering that a static facial expression is much different than an 

expression that unfolds actively during a situation, and that people are highly sensitive to change 

in the environment relative to a static stimulus, the effect of the faces may not have been strong 

enough to elicit an effect. Another possibility is that the evidence contained in the image for 

making an accurate assessment was already strong, and any additional effect of the expressions 

was superfluous to the initial perceiver assessment.  

Additionally, this may be due to the bi-directionality of emotions. Emotions may 

influence the interpretation of a situation, as hypothesized here. However, the situation may also 

serve to disambiguate the emotion. In this way, the judgment stimuli may have been influencing 

the interpretation of the emotional expression and not the other way around. Either way, based 

on the narrow context of these studies and specificity of expression stimuli used, more work 

needs to be done before conclusions are made about the effectiveness of facial expressions as 

communicative devices in moral contexts. 

 

41 

 

The auditory expressions did appear to influence judgments in some cases. IAPS image 

#2026, the neutral image, may not have initially presented a case of an immoral act, but with an 

angry crowd virtually present on the scene, perhaps notably innocuous behavior can become 

suspicious to perceivers. Less innocuous behavior may be less susceptible to the influence of 

auditory communicative signals; as there is already an abundance of evidence for making a 

judgment, additional information becomes superfluous or corroborative in its impact. The 

influence of crowd sounds on severity judgments of the neutral image may also be due to floor 

effects. Specifically, ratings for the neutral image were low to begin with so an increase in moral 

severity is the most likely outcome. However, I find this less likely as the same effect was found 

in IAPS image #2691, where the neutral crowd was just above the mid-point of the rating scale.  

In this instance, the image depicted a protester throwing an object at an unknown target. The 

congruency between the protester and crowd sound (that a crowd is usually present at a protest) 

may have bolstered the relevance of the angry sounds and perceivers found it to be useful 

evidence in their judgment. 

 

This latter point of congruency might also point to a more domain-specific connection 

between external expressions and judgments. That is, certain emotional expressions pair more 

strongly with a perceiver response when the judgment stimuli and emotional stimuli are from the 

same relevant domain (e.g., Garcia & Koelling, 1966). That might explain why the expressions 

from a crowd of people influence perceptions of a protester (crowds are known to be present at 

protests) or a woman waiting at a store (crowds may also be present in stores or public areas) but 

not the perception of other, less congruent situations (e.g., a sexual assault or one-on-one 

interpersonally violent situation). A promising future direction would be to assess the domain 

 

42 

specificity of emotional expressions, and the boundaries against which they may or may not 

impact others’ judgments or behavior. 

 

Aside from those two images, and barring methodological idiosyncrasies that may have 

influenced the effects of the studies as a whole, why wasn’t external evidence influential for 

perceptions of moral transgressions? Jonathan Haidt’s ‘affect-first’ perspective may help shed 

light on the situation. Perceivers of moral acts have an immediate, unconscious, affective 

reaction to the situation. Primarily autonomic in nature, it serves as the baseline response for the 

perceiver, characterized as a broadly valenced ‘good’ or ‘bad’ feeling (akin to an initial 

approach/avoidance behavior; Smillie, Pickering, & Jackson, 2006). Following this immediate 

reaction is a more conscious and deliberative period where the perceiver consciously processes 

the information at hand. Critical to his perspective is that the conscious rationalizing of the 

perceiver is in the service of the initial affective response. If the initial response is negatively 

valenced, then the conscious rationalizing will be in service of a negative viewpoint (and vice 

versa; Haidt, 2012).  

This may partially explain why the severity ratings of perceivers in the present study 

were more or less impervious to external information sources. The initial reaction of the 

perceiver may have negated any subsequent influence from external sources, particularly when 

coupled with the contrived nature of experimental laboratory settings. Returning to the 

introductory example, imagine you come across a crowd of people while walking down the 

street. The crowd is surrounding two individuals, and you hear what sounds like a conflict. Are 

the two fighting? As you approach, you actively assess and make judgments about the situation. 

According to Haidt’s view, whatever your initial reaction is to the situation will serve as a very 

strong valenced base that influences subsequent conscious assessments of the situation. 

 

43 

Certainly, concrete evidence contrary to your assessment would likely change your mind (e.g., if 

you initially thought they were fighting but they were acting). After changing your mind, you 

wouldn’t judge the severity of their actions the same way, but the point illustrated is that the 

initial affective response is the starting point for subsequent cognitive processing about the 

situation. A future direction for research to explicitly test this possibility would be to examine 

whether initial affective responses affect the susceptibility of perceivers to use external 

communicative signals when constructing moral judgments. 

Schwarz & Clore (2007) note that emotions and feelings contain information insofar as 

the information is relevant to the source at hand and not dismissible as either extraneous to the 

situation or attributable to an irrelevant source. In this way, they might argue that the external 

expressions of others were disconnected to the situation at hand and failed to evoke the 

appropriate feelings in the perceivers to influence judgments. Continuing with the same example, 

imagine that you approach the crowd of people it appears that two individuals are indeed having 

a fight. You wouldn’t need to examine the rest of the crowd’s facial expressions to make an 

evaluative judgment; their expressions would be redundant with the evidence you’ve gleaned 

with your own senses.  

Further, this perspective might also suggest that any felt or perceived emotions might be 

attributable to either the facial expression seen or discounting the information of the facial 

expression as resulting from the immoral act in the image. Either explanation would suggest a 

disruption of the link between participants’ internal states and their moral judgments.  

Limitations 

 

The studies presented are not without their limitations. Along with the aforementioned 

ambiguity in how participants’ initial affective responses affect the susceptibility to external 

 

44 

information, there are a number of methodological ways that the current studies can be 

improved.  

 

I went through several efforts to validate constituent components of each study. The pilot 

study and validation study allowed me to pre-select several stimuli that were appropriate for the 

studies. Likewise, the facial expressions (Kanade, Cohn, & Tian, 2000; Lucey et al., 2010) and 

auditory stimuli (see Study 4) both communicate the emotions they intended to successfully. 

However, the extent to which participants used these communicative devices to make judgments 

about the moral stimuli was the test of interest.  

The current studies constituted a proximal test of how people use emotional expressions 

as communicative signals when constructing moral judgments. However, the designs of the 

current studies are far from an externally valid test of these processes. Laboratory studies are, by 

definition, contrived to some extent. Static facial expressions presented briefly before or during a 

moral judgment is very different context than an emergent judgment event encountered in 

everyday life. While the studies presented here are designed to assess whether expressions can 

influence severity judgments, the socioecological congruency to a real-world, active judgment 

event is lacking. Further, in a real judgment event, autonomic reactivity is likely stronger and 

more active. As autonomic nervous system reactivity (connected to emotional reactivity) is 

posited to play a role in these events. Reactivity is likely attenuated in a laboratory setting—the 

judgment effects found here may be less generalizable to the fast-paced reactivity present in 

ecologically valid situations. A more appropriate test may involve having confederates actively 

communicate their emotions (through facial or auditory cues) while an (ostensibly) ambiguously 

moral action is happening.   

 

45 

Further, the same judgment stimuli were used across all studies to control for varying 

targets across the experimental manipulations (e.g., facial expressions, crowds). These stimuli 

were pre-tested to ensure methodological and statistical viability during the analyses. However, 

if the images were not found to be morally compelling or motivating for judgment by 

participants in a laboratory, the judgments might not be as strong as the would be for more 

compelling stimuli. Without sufficient motivation to make a proper and accurate judgment, 

participants may have exercised less scrutiny in their assessments and felt less compelled to 

glean relevant evidence from the environment. This would have attenuated the effect of any 

external evidence.  

Future research can examine the extent to which participants use emotional expressions 

in contexts such as those featured in the current studies, if at all. In the current studies, I did not 

explicitly ask the participants if they used the emotional stimuli when trying to discern the events 

of an uncertain situation. It could be that moral judgments were indeed affected, but only among 

those who reported using the emotional expressions as information when forming judgments. 

However, asking such a question might also nullify the effects of communicative emotional 

expressions on judgments (e.g., Schwarz & Clore, 2007). 

Conclusion 

 

There may be certain situations where the emotional expressions of others can influence 

someone’s judgments. In the current studies, I tested whether exposure to emotional expressions 

affected subsequent moral judgments under uncertainty. The results from Studies 2-4 did not 

support this possibility. However, the specific domains and expressions that might evoke this 

phenomenon need additional empirical attention before a strong conclusion can be made. I 

identified a number of plausible explanations and hypotheses that can be tested in future 

 

46 

research. The hope for this research program is that future researchers can contextualize moral 

judgments within embedded social structures that allow us to form judgments and evaluations 

based on salient information in our environments.  

 

 

 

47 

FOOTNOTES 

 

1 This also prevents the disclosure of the images in their visual form within publications and manuscripts. However, 
a description of each image included in the study is provided in the appendix. 
 
2 Participants were informed at the beginning of the survey and in the consent form that they would be asked to view 
images that might make them uncomfortable, but that were not more graphic than what might be seen in a Rated-R 
film. However, due to the sensitive and graphic nature of some IAPS images assessed in this study, a number of 
participants declined participation or opted not to respond to multiple items in the survey.  
 
3 I attempted to implement competing covariance structures that may better fit the model, including heterogeneous 
compound symmetry and unstructured. However, the model would not converge under these specifications. 
 
4 An additional way of evaluating the moral nature of the stimuli is to have people generate stories about each image 
and code for the presence of moral features. To this end, a sample of N=105 participants was recruited via Amazon 
Mechanical Turk to examine if people naturally interpret the images in a moral framework outside of the 
experimental context. Each participant viewed a random 4 of the 8 images and provided an open-ended 3-5 sentence 
‘story’ of what they thought the people in the images were doing (similar to instructions for the Thematic 
Apperception Test). They were coded by research assistants as either moral (containing any morally relevant themes 
from a provided list; e.g., fairness, interpersonal harm, disgust, etc.) or non-moral. The neutral image (IAPS #2026) 
was rated as containing moral features in only 15% of stories, suggesting that it did little to elicit moral concerns. 
All other images had a minimum proportion of 75% moral features (#9102), up to a high of 98% moral features 
(#3530). The pattern suggests the target images are being interpreted in a moral framework as expected, while the 
neutral image is not. 
 
5 Due to a methodological oversight, one of the images had to be dropped from the analyses for Study 3. However, 
as the pattern of results was largely consistent with the previous studies. I do not believe its inclusion would have 
influenced the conclusions made as, for the most part, the effects of certainty and crowd did not differ across the 
image set. 
 
6 As an exploratory investigation into predictors of uncertainty and severity judgments, I included a handful 
individual difference measures in addition to the standard demographic items for participants to complete. This 
included the intolerance for uncertainty scale (Freeston, Rhéaume, Letarte, Dugas, & Ladouceur, 1994), need for 
closure scale (Kruglanski, Atash, De Grada, Mannetti, & Pierro, 2013), the moral foundations questionnaire 
(Graham, 2011), and the right-wing authoritarianism scale (Altemeyer, 1981). I regressed both severity and 
uncertainty judgments separately on all of these measures. For severity ratings, only the purity and harm factors 
from the moral foundations questionnaire were significantly related to severity (F(1, 492) = 3.93, p = .048, β = .34; 
F(1, 490) = 13.36, p < .001, β =.52). Higher scores on both of these factors were associated with higher severity 
ratings. The other individual difference characteristics were unrelated to severity judgments. For uncertainty, the 
same pattern of results was found, with both the purity and harm factors being significantly related to self-reported 
certainty ratings (F(1, 492) = 3.91, p = .049, β =.33; F(1, 492) = 7.54, p = .006, β = .55). Higher values in both of 
these factors were associated with higher certainty ratings. The other individual difference characteristics were 
unrelated to certainty judgments. 
 

 

 

48 

 
 
 
 
 
 
 
 
 
 
 
 

 

 
 
 
 
 
 
 
 
 

APPENDIX 

49 

Table 1: 

Study 1 IAPS image mean ratings (SD) and descriptive statistics, including severity, relevance, 

ambiguity, and emotions. 

IAPS 

P. 

M. 

Image #  Relevance  Severity  Uncertainty  Happy 

Sad 

Angry  Fearful 

Disgust 

Disgust 

Total 

 

4.06 
(2.26) 

4.42 
(2.34) 

3.08 
(2.04) 

1.48 
(1.22) 

3.99 
(2.42) 

3.79 
(2.46) 

3.63 
(3.63) 

3.75 
(2.44) 

4.01 
(2.46) 

1112 

Snake 

 
 

2.13 
(1.62) 

1.72 
(1.28) 

2.88 
(2.08) 

1.36 
(1.00) 

1.76 
(1.49) 

1.56 
(1.27) 

3.86 
(2.20) 

2.59 
(2.09) 

1.50 
(1.16) 

1201 

 
 

Spider 
1.96 
(1.53) 

2026  Woman store 

1.72 
(1.32) 

2.86 
(2.12) 

1.31 
(0.89) 

1.61 
(1.39) 

1.62 
(1.41) 

4.20 
(2.30) 

3.66 
(2.31) 

1.59 
(1.33) 

 
 

2.35 
(1.80) 

1.61 
(1.13) 

3.11 
(2.19) 

1.43 
(1.01) 

1.41 
(1.06) 

1.34 
(0.97) 

1.30 
(0.91) 

1.26 
(0.87) 

1.30 
(0.93) 

2039  Woman table 

 
 

2.39 
(1.81) 

1.53 
(1.12) 

3.14 
(2.16) 

1.54 
(1.08) 

2.90 
(1.97) 

1.41 
(1.02) 

1.38 
(1.01) 

1.28 
(0.86) 

1.32 
(0.91) 

2278 

Refugees 

 
 

3.51 
(1.79) 

2.83 
(1.79) 

3.91 
(1.89) 

1.49 
(1.03) 

3.77 
(2.11) 

2.45 
(1.84) 

2.62 
(1.86) 

2.09 
(1.64) 

2.41 
(1.85) 

2373  Mariachis 

 
 

2.37 
(1.95) 

1.33 
(0.92) 

2.52 
(2.08) 

3.89 
(2.08) 

1.16 
(0.63) 

1.15 
(0.61) 

1.15 
(0.64) 

1.16 
(0.68) 

1.16 
(0.65) 

2377 

Studying 

 
 

2.69 
(2.27) 

1.33 
(1.00) 

2.52 
(2.13) 

2.76 
(1.95) 

1.29 
(0.95) 

1.19 
(0.78) 

1.24 
(0.90) 

1.19 
(0.79) 

1.20 
(0.80) 

2390 

Talking at table 

 
 

2.55 
(1.98) 

1.41 
(0.97) 

2.85 
(2.09) 

2.34 
(1.74) 

1.45 
(1.10) 

1.21 
(0.75) 

1.34 
(0.99) 

1.22 
(0.81) 

1.24 
(0.82) 

2488 

Accordion player 

 
 

2.40 
(1.96) 

1.39 
(1.03) 

2.60 
(2.05) 

3.32 
(2.00) 

1.32 
(0.95) 

1.24 
(0.87) 

1.26 
(0.95) 

1.23 
(0.90) 

1.24 
(0.91) 

2590 

 
 

Coffee older woman 
1.76 
(1.27) 

2.79 
(1.89) 

2691 

Protester 

3.31 
(1.98) 

2.51 
(1.82) 

2.07 
(1.53) 

1.42 
(1.04) 

1.45 
(1.02) 

1.34 
(0.91) 

1.38 
(0.96) 

 
 

2717 

 
 
 

4.25 
(1.77) 

4.50 
(1.55) 

4.05 
(1.73) 

1.26 
(0.74) 

3.42 
(1.97) 

3.42 
(1.95) 

3.75 
(1.97) 

2.83 
(1.89) 

3.39 
(1.91) 

Intravenous drug use 
5.02 
(1.59) 

4.41 
(1.97) 

 

 

3.21 
(1.88) 

 

1.16 
(0.69) 

4.54 
(2.04) 

3.62 
(2.21) 

 

 

 

3.69 
(2.22) 

 

4.39 
(2.07) 

 

4.40 
(2.10) 

 

 

50 

Table 1 (cont’d) 
 

2745 

Store theft 

 
 

4.47 
(1.90) 

4.93 
(1.51) 

2.87 
(1.71) 

1.17 
(0.63) 

3.08 
(2.01) 

3.30 
(1.96) 

1.98 
(1.66) 

2.11 
(1.68) 

3.80 
(1.87) 

2751 

Drinking and driving 

 
 

5.06 
(2.01) 

5.80 
(1.42) 

2.61 
(1.92) 

1.18 
(0.72) 

4.62 
(2.14) 

5.36 
(1.89) 

5.01 
(2.05) 

4.06 
(2.21) 

5.32 
(1.85) 

2981 

Animal harm/hunting 

 
 

4.08 
(1.92) 

4.63 
(1.81) 

3.29 
(1.74) 

1.15 
(0.55) 

4.81 
(2.12) 

4.18 
(2.26) 

3.08 
(2.17) 

4.89 
(2.11) 

4.27 
(2.16) 

3181 

Domestic abuse 

 
 

3.76 
(1.76) 

4.05 
(1.79) 

4.11 
(1.80) 

1.23 
(0.78) 

3.77 
(2.20) 

3.11 
(2.16) 

3.55 
(2.11) 

3.80 
(2.15) 

3.56 
(2.17) 

3500 

Gun threaten 

 
 

5.04 
(2.15) 

6.12 
(1.22) 

2.78 
(1.96) 

1.14 
(0.66) 

5.03 
(1.96) 

5.02 
(1.99) 

5.20 
(1.89) 

4.51 
(2.17) 

5.43 
(1.70) 

3530 

Gun threaten 2 

 
 

4.79 
(2.12) 

6.00 
(1.31) 

3.01 
(2.04) 

1.15 
(0.66) 

4.92 
(2.04) 

4.90 
(2.00) 

5.08 
(2.00) 

4.72 
(2.11) 

5.28 
(1.85) 

4232  Masturbation 

 
 

3.25 
(1.92) 

3.10 
(2.03) 

3.30 
(1.86) 

1.79 
(1.49) 

1.90 
(1.70) 

1.87 
(1.64) 

1.69 
(1.47) 

2.61 
(2.07) 

2.55 
(2.06) 

4233 

Prostitution 

 
 

4.01 
(1.85) 

4.20 
(1.75) 

3.48 
(1.91) 

1.27 
(0.81) 

3.28 
(2.18) 

2.51 
(1.91) 

2.31 
(1.90) 

2.92 
(1.99) 

3.63 
(2.11) 

4621  Workplace harassment 

 
 

4.31 
(1.88) 

4.70 
(1.72) 

3.91 
(1.89) 

1.27 
(0.90) 

3.41 
(2.10) 

4.01 
(2.12) 

3.72 
(2.12) 

3.73 
(2.13) 

4.18 
(2.07) 

5875  Man on bike 

 
 

2.41 
(1.71) 

1.86 
(1.30) 

3.28 
(2.13) 

2.41 
(1.81) 

1.70 
(1.30) 

1.50 
(1.13) 

1.46 
(1.07) 

1.44 
(1.08) 

1.53 
(1.19) 

6312  Woman assault van 
6.08 
(1.27) 

4.98 
(2.09) 

 
 

3.07 
(2.11) 

1.11 
(0.54) 

5.14 
(1.99) 

5.56 
(1.80) 

5.25 
(2.00) 

4.97 
(2.03) 

5.73 
(1.64) 

6313 

Knife threaten 

 
 

5.11 
(2.19) 

6.41 
(1.06) 

2.48 
(1.94) 

1.19 
(0.79) 

5.32 
(1.87) 

5.67 
(1.67) 

5.52 
(1.82) 

5.24 
(1.96) 

5.97 
(1.45) 

6315 

Assault slap 

 
 

5.13 
(2.07) 

6.18 
(1.22) 

2.73 
(2.08) 

1.11 
(0.58) 

5.44 
(1.79) 

5.91 
(1.54) 

5.04 
(2.16) 

5.17 
(2.07) 

5.92 
(1.53) 

6415 

Animal slain 

 
 

4.83 
(2.21) 

6.06 
(1.42) 

2.87 
(2.05) 

1.14 
(0.73) 

5.86 
(1.69) 

5.68 
(1.83) 

4.36 
(2.36) 

5.75 
(1.79) 

5.78 
(1.74) 

6520 

Knife threaten 2 

 
 

4.99 
(2.15) 

6.17 
(1.23) 

3.01 
(2.10) 

1.14 
(0.60) 

5.56 
(1.68) 

5.34 
(1.85) 

5.20 
(1.91) 

5.18 
(1.95) 

5.68 
(1.62) 

 

51 

Table 1 (cont’d) 
 

7506 

Gambling 

 
 

3.12 
(1.75) 

2.48 
(1.43) 

3.17 
(1.86) 

1.69 
(1.26) 

1.61 
(1.20) 

1.53 
(1.20) 

1.53 
(1.18) 

1.39 
(1.05) 

1.73 
(1.30) 

9007 

 
 

Drug paraphernalia 
5.31 
(1.58) 

4.46 
(2.02) 

9102 

Drug injection 

3.68 
(1.99) 

1.16 
(0.64) 

4.76 
(2.00) 

3.85 
(2.19) 

4.05 
(2.22) 

4.61 
(2.06) 

4.71 
(1.97) 

 
 

4.33 
(2.01) 

4.92 
(1.70) 

3.21 
(1.92) 

1.21 
(0.80) 

4.40 
(2.12) 

3.53 
(2.22) 

3.56 
(2.23) 

4.36 
(2.09) 

4.27 
(2.14) 

9145 

Cow branding 

 
 

3.97 
(1.85) 

4.45 
(1.86) 

3.29 
(1.77) 

1.28 
(0.93) 

4.64 
(2.06) 

4.03 
(2.22) 

3.10 
(2.18) 

4.18 
(2.16) 

4.02 
(2.19) 

9163 

Soldier abuse 

 
 

4.99 
(2.04) 

5.81 
(1.40) 

3.27 
(2.04) 

1.17 
(0.69) 

5.29 
(1.86) 

5.23 
(1.87) 

4.79 
(2.07) 

4.79 
(2.13) 

5.45 
(1.73) 

9183 

Dog abuse 

 
 

5.03 
(2.24) 

6.30 
(1.28) 

2.57 
(2.10) 

1.16 
(0.77) 

6.33 
(1.35) 

6.09 
(1.59) 

4.69 
(2.27) 

6.01 
(1.63) 

6.23 
(1.46) 

9252 

Corpse dragged 

 
 

5.06 
(2.28) 

6.47 
(1.15) 

2.61 
(2.08) 

1.15 
(0.69) 

5.83 
(1.69) 

5.77 
(1.71) 

5.33 
(2.01) 

6.10 
(1.54) 

6.24 
(1.35) 

9253 

Dead woman 

 
 

5.15 
(2.24) 

6.40 
(1.19) 

2.87 
(2.10) 

1.16 
(0.73) 

6.06 
(1.50) 

5.89 
(1.70) 

5.44 
(2.02) 

6.16 
(1.47) 

6.21 
(1.39) 

9295 

Ocean pollution 

 
 

4.98 
(1.96) 

5.39 
(1.61) 

2.72 
(1.78) 

1.17 
(0.78) 

5.29 
(1.82) 

5.11 
(1.91) 

4.10 
(2.26) 

4.46 
(2.16) 

5.04 
(1.85) 

9410 

Boy’s body 

 
 

4.98 
(2.10) 

5.26 
(1.94) 

3.70 
(2.22) 

1.19 
(0.75) 

6.18 
(1.42) 

5.32 
(2.00) 

4.95 
(2.14) 

5.64 
(1.81) 

5.40 
(2.00) 

9413  Men hanged 

 
 

5.08 
(2.22) 

6.28 
(1.13) 

2.93 
(2.13) 

1.13 
(0.65) 

5.63 
(1.69) 

5.31 
(1.92) 

5.20 
(1.98) 

5.26 
(1.99) 

5.82 
(1.59) 

9414 

 
 

Gun threaten assault 
6.15 
(1.29) 

4.94 
(2.19) 

9423 

Hostages in pit 

3.16 
(2.08) 

1.16 
(0.70) 

5.49 
(1.84) 

5.49 
(1.80) 

5.13 
(2.05) 

5.06 
(2.06) 

5.73 
(1.60) 

 
 

4.89 
(2.13) 

5.87 
(1.56) 

3.20 
(2.05) 

1.14 
(0.64) 

5.42 
(1.82) 

5.15 
(1.96) 

5.16 
(2.00) 

4.64 
(2.18) 

5.40 
(1.87) 

9800 

Neo-nazi teen 

 
 

4.62 
(2.13) 

5.67 
(1.54) 

2.91 
(1.95) 

1.17 
(0.63) 

4.42 
(2.23) 

5.07 
(2.02) 

4.44 
(2.17) 

4.12 
(2.28) 

5.26 
(1.95) 

9810 

 
 

KKK burning cross 
6.22 
(1.45) 

5.03 
(2.28) 

2.45 
(2.06) 

1.23 
(0.90) 

5.42 
(2.04) 

5.80 
(1.85) 

5.31 
(2.10) 

5.17 
(2.23) 

6.03 
(1.68) 

 

52 

2. 

3. 

4. 

5. 

6. 

7. 

8. 

9. 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Table 2: 

Study 1 judgments, bivariate correlations (Pearson’s r) 

 

1. Relevance 

1. 

- 

2. Severity 

.54 

 

- 

 

 

3. Uncertainty 

.08 

-.03 

- 

4. Happy 

-.09 

-.33 

.04 

- 

5. Sad 

.43 

.72 

-.01 

-.26 

- 

6. Angry 

.44 

.78 

-.06 

-.25 

.85 

- 

7. Fearful 

.34 

.64 

.00 

-.23 

.74 

.77 

- 

8. P. Disgust 

.37 

.69 

-.02 

-.25 

.77 

.81 

.75 

- 

9. M. Disgust 

.48 

.84 

-.06 

-.29 

.83 

.90 

.74 

.82 

- 

Note: P. Disgust = physical disgust; M. Disgust = Moral disgust. Bold values significant at p < 
.01. 
 
 
 

 

 

53 

Table 3: 

Study 2 IAPS image mean ratings (SD) and descriptive statistics, including severity, others’ 

perceived ratings, ambiguity, and emotions. 

IAPS 

P. 

M. 

Image # 

Severity  Others  Uncertainty  Happy 

Sad 

Angry  Fearful 

Disgust 

Disgust 

Total 

 

54.53 
(33.62) 

56.55 
(29.54) 

42.38 
(31.71) 

1.25 
(.86) 

3.33 
(2.23) 

3.07 
(2.21) 

2.80 
(2.14) 

2.91 
(2.17) 

3.35 
(2.22) 

2026  Woman store 

 
 

14.59 
(19.74) 

17.07 
(19.63) 

41.88 
(37.50) 

1.27 
(.93) 

1.39 
(1.05) 

1.36 
(.98) 

1.24 
(.77) 

1.27 
(.84) 

1.34 
(.94) 

2278 

Refugees 

 
 

28.11 
(30.17) 

30.37 
(28.84) 

42.55 
(34.66) 

1.4 

(1.10) 

3.58 
(2.21) 

2.27 
(1.86) 

2.37 
(1.9) 

1.91 
(1.59) 

2.20 
(1.82) 

2691 

Protester 

 
 

54.07 
(25.12) 

57.99 
(21.97) 

49.68 
(26.84) 

1.24 
(.74) 

2.95 
(1.96) 

3.02 
(1.93) 

3.21 
(2.01) 

2.30 
(1.77) 

2.91 
(1.88) 

2745 

Store theft 

 
 

59.68 
(25.83) 

62.81 
(23.28) 

37.65 
(28.10) 

1.14 
(.54) 

2.76 
(1.85) 

2.76 
(1.81) 

1.75 
(1.43) 

1.94 
(1.55) 

3.18 
(1.88) 

2981 

Animal harm/hunting 

 
 

58.98 
(31.29) 

56.72 
(24.53) 

40.04 
(28.55) 

1.15 
(.56) 

4.25 
(2.16) 

3.63 
(2.26) 

2.74 
(2.06) 

4.02 
(2.25) 

3.73 
(2.25) 

3181 

Domestic abuse 

 
 

46.43 
(29.69) 

50.89 
(26.57) 

51.19 
(30.56) 

1.12 
(.54) 

2.90 
(2.00) 

2.53 
(1.95) 

3.09 
(2.08) 

3.20 
(2.05) 

2.86 
(2.03) 

3530 

Gun threaten 2 

 
 

86.17 
(16.55) 

85.49 
(15.60) 

42.24 
(35.00) 

1.11 
(.62) 

4.47 
(2.09) 

4.37 
(2.13) 

4.55 
(2.12) 

3.96 
(2.29) 

4.83 
(2.00) 

4232  Masturbation 

 
 

30.80 
(30.22) 

44.78 
(25.71) 

36.40 
(28.87) 

1.89 
(1.67) 

1.59 
(1.34) 

1.45 
(1.13) 

1.42 
(1.14) 

2.00 
(1.66) 

2.12 
(1.66) 

4233 

Prostitution 

 
 

47.23 
(28.16) 

57.13 
(25.73) 

45.15 
(28.28) 

1.25 
(.82) 

2.73 
(1.96) 

1.97 
(1.56) 

1.80 
(1.45) 

2.24 
(1.68) 

2.77 
(1.83) 

4621  Workplace harassment 

 
 

63.36 
(28.35) 

60.82 
(25.53) 

50.19 
(31.77) 

1.16 
(.66) 

3.44 
(2.14) 

3.83 
(2.16) 

3.27 
(2.17) 

3.61 
(2.19) 

4.16 
(2.12) 

7506 

Gambling 

 
 

22.06 
(22.65) 

32.69 
(22.73) 

34.19 
(28.08) 

1.58 
(1.23) 

1.47 
(1.05) 

1.41 
(.99) 

1.42 
(1.03) 

1.30 
(.82) 

1.58 
(1.11) 

9102 

Drug injection 

 
 

65.04 
(28.85) 

69.13 
(23.86) 

43.15 
(38.74) 

1.17 
(.67) 

3.84 
(2.20) 

2.95 
(2.13) 

3.03 
(2.13) 

3.70 
(2.26) 

3.75 
(2.16) 

9145 

Cow branding 

 

54 

Table 3 (cont’d) 

 
 

54.54 
(31.04) 

51.69 
(24.04) 

38.74 
(27.76) 

1.16 
(.61) 

4.01 
(2.09) 

3.36 
(2.19) 

2.50 
(1.94) 

3.41 
(2.14) 

3.47 
(2.15) 

9295 

Ocean pollution 

 
 

72.34 
(24.59) 

63.95 
(24.37) 

37.34 
(31.34) 

1.11 
(.51) 

4.84 
(1.97) 

4.68 
(2.03) 

3.75 
(2.18) 

3.93 
(2.17) 

4.71 
(1.98) 

9423 

Hostages in pit 

 
 

83.69 
(21.97) 

81.61 
(21.16) 

45.90 
(35.48) 

1.11 
(.57) 

5.02 
(1.96) 

4.82 
(2.11) 

4.72 
(2.11) 

4.10 
(2.35) 

5.10 
(2.04) 

9800 

Neo-nazi teen 

 
 

76.14 
(26.51) 

75.16 
(23.57) 

41.27 
(34.45) 

1.12 
(.63) 

4.11 
(2.25) 

4.67 
(2.18) 

3.89 
(2.25) 

3.75 
(2.38) 

4.86 
(2.12) 

 

 

 

 

55 

Table 4: 

Study 2 severity judgments by facial expression condition. 

Facial expression 

 

M 

 

SD 

 

Neutral 

54.32 

33.68 

Happy 

54.82 

33.53 

Angry 

54.63 

33.38 

Disgust 

54.77 

33.53 

 

 

 

56 

Table 5: 

Study 2 judgments, bivariate correlations (Pearson’s r) and demographic means. 

2. 

3. 

4. 

5. 

6. 

7. 

8. 

 

1. Severity 

1. 

- 

2. Projecting Others 

.82 

 

- 

3. Uncertainty 

.05 

.08 

 

 

- 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

4. Happy 

5. Sad 

6. Angry 

7. Fearful 

-.20 

-.15 

-.03 

- 

.61 

.50 

.02 

-.16 

- 

.70 

.56 

.01 

-.15 

.75 

- 

.57 

.50 

.09 

-.12 

.64 

.68 

- 

8. P. Disgust 

.59 

.49 

.04 

-.14 

.64 

.69 

.60 

- 

9. M. Disgust 

.77 

.62 

.00 

-.16 

.70 

.81 

.65 

.74 

 

 

 

Gender 

 

 

Severity 

M 

 

SD 

 

Male  52.66  32.75 

Female  56.09  34.01 

Race 

 

 

Black  60.54  35.17 

White  53.24  30.99 

Asian  55.81  35.52 

Other/not specified  53.24  33.95 

 

 

 

 

 

 

 

 

 

 

 

 

 

Ambiguity 

M 

 

SD 

 

42.29  30.52 

42.85  32.50 

 

 

46.74  31.91 

42.01  32.07 

40.21  29.75 

47.58  33.49 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Note: P. Disgust = physical disgust; M. Disgust = Moral disgust. Variables were mean-center 

aggregated within each participant to account for repeated measures within each subject. All 

correlations significant at p < .05. 

 

57 

Table 6: 

Study 2 severity ratings, omnibus effects and parameter estimates. 

 

Gender 

Predictor 

Race 

Face 

Male 

Non-white 

Anger 

Disgust 

Happy 

 
df 

1 

 

1 

 

3 

 

 

 

 
F 

3.25 

 
b 

 

 
t 

 

 
p 

95% CI 

Upper 

Lower 

.073 

 

 

 

2.96 

1.80 

.073 

-0.27 

6.19 

3.48 

 

 

.064 

 

 

 

-3.14 

1.86 

.064 

-6.45 

0.18 

0.47 

 

 

 

 

 

 

 

 

.703 

 

 

0.29 

.771 

-3.70 

4.99 

0.71 

.478 

-2.79 

5.95 

0.52 

.605 

-3.22 

5.52 

Uncertainty 

1 

10.30 

0.14 

5.58 

.001 

0.09 

0.18 

Image 

15 

724.48 

 

 

< .001 

 

 

2278 

2691 

2745 

2981 

3181 

3530 

4232 

4233 

4621 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

12.19 

6.47 

< .001 

10.14 

16.54 

38.28 

17.45  < .001 

33.98 

42.58 

44.54 

20.40  < .001 

40.26 

48.82 

43.43 

19.86  < .001 

39.14 

47.72 

30.99 

14.12  < .001 

26.69 

35.29 

71.28 

32.60  < .001 

67.00 

75.57 

18.27 

8.27 

< .001 

13.94 

22.60 

33.70 

15.43  < .001 

29.42 

37.98 

49.63 

22.61  < .001 

45.33 

53.94 

 

 

58 

Table 6 (cont’d) 

7506 

9102 

9145 

9295 

9423 

9800 

 

 

 

 

 

 

 

 

 

 

 

 

8.35 

3.77 

< .001 

4.00 

12.69 

50.91 

23.32  < .001 

46.63 

55.19 

39.33 

18.01  < .001 

35.05 

43.61 

56.89 

25.86  < .001 

52.57 

61.20 

67.96 

30.99  < .001 

63.66 

72.26 

61.15 

27.92  < .001 

56.85 

65.44 

Face x Uncertainty 

3 

0.33 

Face x Image 

45 

0.37 

Image x Uncertainty 

15 

23.07 

 

 

 

 

 

 

.806 

1.00 

< .001 

 

 

 

 

 

 

Note: Regression coefficients are unstandardized. Categorical variables dummy coded prior to 

analysis. Reference categories for Gender = Female, Race = White, Face = Neutral, Image = 

2026 (neutral). 

 

 

 

59 

Table 7: 

Study 3 IAPS mean severity and uncertainty ratings (SD) by image, facial expression. 

Face 

Face 

Face 

(neutral) 

(disgust) 

(anger) 

57.49 
(33.90) 

58.73 
(33.97) 

58.56 
(33.82) 

10.26 
(15.50) 

15.86 
(22.50) 

14.76 
(19.63) 

66.10 
(26.83) 

64.25 
(24.17) 

66.61 
(24.84) 

51.22 
(29.24) 

50.48 
(29.21) 

46.06 
(30.20) 

74.00 
(29.61) 

73.40 
(31.24) 

76.36 
(28.96) 

57.72 
(28.57) 

65.24 
(28.05) 

58.44 
(29.85) 

60.44 
(30.77) 

58.72 
(31.29) 

61.99 
(29.93) 

72.88 
(27.62) 

69.59 
(32.26) 

71.24 
(28.66) 

66.79 
(32.94) 

72.90 
(32.35) 

72.69 
(31.57) 

 

 
 

 
 

 
 

 
 

 
 

 
 

 
 

 
 

 
 

IAPS 

Image # 

Severity 
(clean) 

Severity 
(blurred) 

Certainty 

(clean) 

Certainty 
(blurred) 

Total* 

 

63.22 
(34.64) 

53.39 
(32.41) 

58.82 
(31.57) 

51.30 
(32.27) 

2026*  Woman store 

 
 

2691* 

 
 

8.88 

(13.76) 

Protester 
71.33 
(24.41) 

18.29 
(22.96) 

33.98 
(32.53) 

36.75 
(29.70) 

60.05 
(24.90) 

62.17 
(25.18) 

53.90 
(25.93) 

3181*  Domestic abuse 

 
 

54.09 
(29.79) 

44.48 
(28.62) 

38.39 
(28.49) 

27.27 
(25.67) 

3530*  Gun threaten 2 

 
 

93.92 
(11.24) 

55.72 
(30.35) 

79.66 
(22.61) 

43.54 
(27.47) 

4233 

Prostitution 

 
 

56.75 
(30.28) 

64.08 
(27.18) 

61.74 
(29.29) 

71.10 
(25.84) 

4621*  Workplace harassment 
55.16 
(30.86) 

65.70 
(29.50) 

 
 

56.92 
(29.76) 

56.77 
(27.87) 

9102 

Drug injection 

 
 

68.11 
(30.77) 

74.33 
(27.97) 

70.33 
(29.12) 

82.39 
(21.71) 

9423*  Hostages in pit 

 
 

86.83 
(20.64) 

54.99 
(33.96) 

67.36 
(26.47) 

38.83 
(32.83) 

Note: * Severity ratings significantly different at p<.05. 

 

 

 

 

60 

Table 8: 

Study 3 judgments, bivariate correlations (Pearson’s r) and demographic means across both the 

clean and blur conditions. 

 

M (SD) 

1. Severity 

58.26 (33.89) 

1. 

- 

2. Certainty 

55.03 (32.14) 

.57 

2. 

 

- 

3. Happy 

1.26 (0.81) 

-.21 

-.04 

 

 

- 

 

 

 

3. 

4. 

5. 

6. 

7. 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

4. Sad 

3.39 (2.21) 

.66 

.50 

-.14 

- 

5. Angry 

3.17 (2.17) 

.67 

.46 

-.15 

.70 

- 

6. Fearful 

3.42 (2.18) 

.58 

.29 

-.13 

.64 

.64 

- 

7. P. Disgust 

3.36 (2.20) 

.66 

.47 

-.16 

.64 

.67 

.58 

- 

8. M. Disgust 

3.66 (2.22) 

.79 

.54 

-.17 

.69 

.76 

.60 

.78 

 

 

 

 

 

 

 

 

 

 

 

Gender 

Severity 

M 

 

SD 

 

Male  54.93  34.59 

Female  59.01  33.70 

Race 

 

 

Black  60.85  37.38 

White  58.13  33.75 

Asian  56.90  32.40 

Other/not specified  59.72  34.34 

 

 

 

 

 

 

 

 

 

 

Certainty 

M 

 

SD 

 

55.23 

32.81 

54.96 

31.96 

 

 

63.61 

33.85 

53.83 

32.02 

55.80 

30.06 

57.17 

34.67 

Note: P. Disgust = physical disgust; M. Disgust = Moral disgust. Variables were mean-center 

aggregated within each participant to account for repeated measures within each subject. All 

correlations significant at p < .05. 

 

61 

Table 9: 

Study 3 severity ratings, omnibus effects and parameter estimates. 

 

Gender 

Predictor 

Race 

Face 

Certainty 

Image 

Female 

Non-white 

Disgust 

Anger 

Blur 

2691 

3181 

3530 

4233 

4621 

9102 

9423 

 
df 

1 

 

1 

 

2 

 

 

1 

 

 
F 

3.13 

 
b 

 

 
t 

 

 
p 

95% CI 

Upper  Lower 

.078 

 

 

 

3.27 

1.64 

.078 

-0.36 

6.90 

0.73 

 

 

.395 

 

 

 

1.39 

0.85 

.395 

-1.81 

4.58 

1.21 

 

 

.297 

 

 

 

 

4.40 

0.91 

.362 

-5.06 

13.87 

5.58 

1.16 

.246 

-3.84 

14.99 

45.27 

 

 

< .001 

 

 

 

10.11 

2.09 

.037 

0.61 

19.61 

7 

223.60 

 

 

< .001 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

67.68  14.37 

.001 

58.44 

76.91 

48.45  10.18  < .001 

39.12 

57.78 

87.93  18.39  < .001 

78.55 

97.30 

45.93 

9.67 

< .001 

36.62 

55.24 

57.44  11.97  < .001 

48.03 

66.85 

67.41  14.30  < .001 

58.17 

76.66 

78.90  16.64  < .001 

69.61 

88.20 

Face x Certainty 

Image x Certainty 

2 

7 

0.19 

44.90 

 

 

 

 

.825 

< .001 

 

 

 

 

 

 

62 

Table 9 (cont’d) 

Face x Image 

14 

1.40 

Face x Image x Certainty 

14 

0.89 

 

 

 

 

.143 

.571 

 

 

 

 

Note: Regression coefficients are unstandardized. Categorical variables dummy coded prior to 

analysis. Reference categories for Gender = Male, Race = White, Face = Neutral, Image = 2026 

(neutral), Certainty = clean image. 

 

 

 

63 

Table 10: 

Study 4 judgments, bivariate correlations (Pearson’s r) and demographic means across 

experimental conditions. 

 

M (SD) 

1. Severity 

56.69 (33.62) 

1. 

- 

2. Certainty 

58.69 (31.05) 

.57 

 

- 

 

 

3. Comfort 

48.85 (32.85) 

.65 

.73 

- 

4. Worried 

41.89 (30.58) 

.24 

.06 

.09 

- 

2. 

3. 

4. 

5. 

6. 

7. 

8. 

9. 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

1.25 (.73) 

-.25 

-.09 

-.11 

-.08 

- 

3.55 (2.21) 

.65 

.43 

.51 

.21 

-.19 

- 

3.12 (2.12) 

.64 

.38 

.51 

.19 

-.19 

.71 

- 

3.51 (2.18) 

.55 

.25 

.35 

.26 

-.18 

.62 

.65 

- 

5. Happy 

6. Sad 

7. Angry 

8. Fearful 

 

 

 

 

9. P. Disgust 

3.37 (2.13) 

.65 

.37 

.47 

.20 

-.20 

.67 

.64 

.58 

- 

10. M. Disgust 

3.68 (2.17) 

.78 

.46 

.58 

.24 

-.22 

.71 

.74 

.60 

.77 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Crowd condition 

Angry 

Neutral 

M 

SD 

M 

SD 

57.71 

33.65 

55.69 

33.57 

 

 

 

 

57.60 

32.61 

57.13 

34.05 

57.74 

34.10 

55.33 

33.44 

 

 

 

 

57.20 

35.66 

53.34 

34.29 

56.71 

33.97 

55.24 

33.76 

59.54 

31.31 

56.95 

31.60 

64.40 

30.73 

59.64 

34.39 

Certainty condition 

Blur 

Clean 

M 

SD 

M 

SD 

Severity  52.66 

31.84 

60.31 

34.75 

Gender 

 

 

 

 

Male  53.57 

31.49 

60.63 

34.40 

Female  52.33 

31.98 

60.35 

34.93 

Race 

 

 

 

 

Black  50.89 

32.10 

59.14 

36.92 

White  52.23 

32.12 

59.64 

35.13 

Asian  55.76 

30.16 

60.06 

32.32 

Other/not specified  53.51 

31.47 

67.06 

32.56 

 

 

 

64 

Table 10 (cont’d) 

Note: P. Disgust = physical disgust; M. Disgust = Moral disgust. Variables were mean-center 

aggregated within each participant to account for repeated measures within each subject. All 

correlations significant at p < .01. 

 

 

65 

Table 11: 

Study 4 severity ratings, omnibus effects and parameter estimates. 

 

Crowd 

Predictor 

 
df 

 
F 

2 

1.18 

 
b 

 

 
t 

 

 
p 

95% CI 

Upper  Lower 

.278 

 

 

Anger 

 

 

3.56 

1.63 

.105 

-0.74 

7.85 

Certainty 

1 

41.31 

 

 

< .001 

 

 

Blur 

 

 

6.02 

2.70 

.007 

1.64 

10.40 

Gender 

2 

0.01 

 

 

.957 

 

 

Female 

 

 

-0.06 

-0.05 

.957 

-2.33 

2.21 

Race 

1 

5.48 

 

 

.020 

 

 

Non-white 

 

 

2.59 

2.34 

.020 

0.42 

4.76 

Image 

6 

564.17 

 

 

< .001 

 

 

 

 

 

 

 

 

51.64  21.04  < .001 

46.82 

56.46 

33.62  13.12  < .001 

28.59 

38.65 

81.00  35.04  < .001 

76.47 

85.54 

51.68  19.31  < .001 

46.43 

56.94 

59.93  22.22  < .001 

54.63 

65.23 

73.45  25.47  < .001 

67.79 

79.11 

0.03 

2.47 

69.55 

 

 

 

 

 

 

 

 

.873 

.023 

< .001 

.142 

 

 

 

 

 

 

 

 

2691 

3181 

3530 

4233 

9102 

9423 

 

 

 

 

 

 

1 

6 

6 

Crowd x Certainty 

Crowd x Image 

Image x Certainty 

Crowd x Certainty x Image 

16 

1.61 

 

 

66 

Table 11 (cont’d) 

Note: Regression coefficients are unstandardized. Categorical variables dummy coded prior to 

analysis. Reference categories for Gender = Male, Race = White, Crowd = neutral, Image = 2026 

(neutral), Certainty = clean image. 

 

67 

Figure 1: 

Example severity rating distributions of a neutral (left), severe (right), and intermediate (bottom) 

image. 

        
Neutral image example (left) and severe image example (right), demonstrating deviations from 

 

normality and significant floor and ceiling effects, respectively. 

Intermediate image example demonstrating a relatively normal distribution with significant 

 

variance across the rating scale.  

 

 

68 

anger

neutral

happy

disgust

 

Figure 2: 

Study 2 mean severity ratings, by image and facial expression condition.  

g
n
i
t
a
r
 

y
t
i
r
e
v
e
S

100

90

80

70

60

50

40

30

20

10

0

 

 

 

IAPS Image #

 

69 

Figure 3: 

Study 3 mean severity ratings, by image and certainty (blur) condition.  

* 

* 

* 

* 

100

* 

g
n
i
t
a
r
 

y
t
i
r
e
v
e
S

90

80

70

60

50

40

30

20

10

0

* 

* 

clean

blur

3530

9423

2691

9102

4621 Total

4233

3181

2026

IAPS Image #

 

Note: *p < .05 

 

 

 

70 

Figure 4: 

Study 4 mean severity ratings, by image and crowd condition.  

g
n
i
t
a
r
 

y
t
i
r
e
v
e
S

90

80

70

60

50

40

30

20

10

0

* 

* 

2026

2691

3181

3530

4233

9102

9423

IAPS Image #

Note: *p < .05 

neutral

angry

 

 

 

 

71 

 
 

 

 

 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

REFERENCES 

 

72 

REFERENCES 

 
 
 
Altemeyer, B. (1981). Right-wing authoritarianism. Manitoba, Canada: University Press. 
 
Anderson, C., Monroy, M., & Keltner, D. (2017). Emotions in the wilds of nature: The 

coherence and contagion of deat during threatening group-based outdoors experiences. 
Emotion, Advanced obnline publication. doi:10.1037/emo0000378 

 
Avramova, Y., & Inbar, Y. (2013). Emotion and moral judgment. WIREs Cognitive Science, 

4(2), 169-178. doi:10.1002/wcs.1216 

 
Bartels, D., Bauman, C., Cushman, F., Pizarro, D., & McGraw, A. (2014). Moral Judgment and 

Decision Making. In G. Keren & G. Wu (Eds.), Blackwell Reader of Judgment and 
Decision Making. Malden, MA: Blackwell. 

 
Blasi, A. (1999). Emotions and Moral Behavior. Journal for the Theory of Social Behavior, 

29(1), 1-19. doi:10.1111/1468-5914.00088 

 
Brosnan, S., & de Waal, F. (2014). Evolution of responses to (un)fairness. Science, 346(6207), 

1251776-1251776.  

 
Chandler, J., Mueller, P., & Paolacci, G. (2014). Nonnaïveté among Amazon Mechanical Turk 

workers: Consequences and solutions for behavioral researchers. Behavior Research 
Methods, 46, 112-130. 

 
Cushman, F. (2008). Crime and punishment: Distinguishing the roles of causal and intentional 

analyses in moral judgment. Cognition, 108, 353-380. 
doi:10.1016/j.cognition.2008.03.006 

 
Cushman, F., Young, L., & Hauser, M. (2006). The role of conscious reasoning and intuition in 

moral judgment. Psychological Science, 17(12), 1082-1089. doi:10.1111/j.1467-
9280.2006.01834.x 

 
DeScioli, P., & Kurzban, R. (2009). Mysteries of morality. Cognition, 112(2), 281-299. 

doi:10.1016/j.cognition.2009.05.008 

 
Eibl-Eibesfeldt, I. (1989). Human ethology. New York, NY: Aldine de Gruyter.  
 
Ekman, P., Friesen, W., O'Sullivan, M., Chan, A., Diacoyanni-Tarlatzis, I., Heider, K., . . . 

Tzavaras, A. (1987). Universals and cultural differences in judgments of facial 
expressions of emotion. Journal of Personality and Social Psychology, 53(4), 712-717. 

 
Freeston, M.H., Rhéaume, J., Letarte, H., Dugas, M.J., & Ladouceur, R. (1994): Why do people 

worry? Personality and Individual Differences, 17(6), 791-802. 

 

73 

 
Garcia, J., & Koelling, R. (1966). Relation of cue to consequence in avoidance learning. 

Psychological Science, 4(1), 123-124. doi:10.3758/BF03342209 

 
Graham, J., Nosek, B., Haidt, J., Iyer, R., Koleva, S., & Ditto, P. (2011). Mapping the moral 

domain. Journal of Personality and Social Psychology, 101(2), 366-386. 

 
Haidt, J. (2001). The emotional dog and its rational tail: A social intuitionist approach to moral 
judgment. Psychological Review, 108(4), 814-834. doi:10.1037//0033-295X. 108.4.814 

 
Haidt, J. (2003). The moral emotions. In R. Davidson, K. Scherer, & H. Goldsmith (Eds.), 

Handbook of affective sciences (pp. 852-870). Oxford: Oxford University Press. 

 
Haidt, J. (2012). The righteous mind: Why good people are divided by politics and religion. New 

York: Pantheon Books. 

 
Haselton, M., & Buss, D. (2000). Error Management Theory: A new perspective on biases in 
cross-sex mind reading. Journal of Personality and Social Psychology, 78(1), 81-91. 
doi:10,1037110022-3514.78.1.81 

 
Kanade, T., Cohn, J., & Tian, Y. (2000). Comprehensive database for facial expression analysis. 

Proceedings of the Fourth IEEE International Conference on Automatic Face and 
Gesture Recognition (FG'00), Grenoble, France, 46-53.  

 
Kolber, J., & Margol, B. (Writers) & M. Nigro (Director). (2015, Jan 26). [Television series 

episode] Morality, Brain Games. USA: National Geographic. 

 
Krahé, B., Möller, I., Huesmann, L., Kirwil, L., Felber, J., & Berger, A. (2015). Desensitization 
to media violence: Links with habitual media violence exposure, aggressive cognitions, 
and aggressive behavior. Journal of Personality and Social Psychology, 100(4), 630-646. 
doi:10.1037/a0021711 

 
Kruglanski, A. W., Atash, M. N., De Grada, E., Mannetti, L., & Pierro, A. (2013). Need for 

Closure Scale (NFC). Measurement Instrument Database for the Social Science. 
Retrieved from: www.midss.ie 

 
Kurzban, R., & Leary, M. (2001). Evolutionary origins of stigmatization: The functions of social 

exclusion. Psychological Bulletin, 127(2), 187-208. doi:10.1037//0033-2909.127.2.187 

 
Lang, P., Bradley, M., & Cuthbert, B. (2008). International affective picture system (IAPS): 

Affective ratings of pictures and instruction manual. Technical Report A-8. University of 
Florida, Gainesville, FL. 

 
Lucey, P., Cohn, J., Kanade, T., Saragih, J., Ambadar, Z., & Matthews, I. (2010). The Extended 
Cohn-Kanade Dataset (CK+): A complete dataset for action unit and emotion-specified 
expression. Proceedings of the Third International Workshop on CVPR for Human 

 

74 

Communicative Behavior Analysis (CVPR4HB 2010), San Francisco, USA, 94-101. 
doi:10.1109/CVPRW.2010.5543262 

 
McDonald, M.M., Defever, A.M., & Navarrete, C.D. (2017). Killing for the greater good: Action 

aversion and the emotional inhibition of harm in moral dilemmas. Evolution and Human 
Behavior, 38(7), 770-778. 

 
Mikhail, J. (2007). Universal moral grammar: theory, evidence, and the future. Trends in 

Cognitive Sciences, 11(4), 143-152. doi:10.1016/j.tics.2006.12.007 

 
Mrug, S., Maden, A., Cook, E., & Wright, R. (2015). Emotional and physiological 

desensitization to real-life and movie violence. Journal of Youth and Adolescence, 44(5), 
1092-1108. doi:10.1007/s10964-014-0202-z 

 
Oveis, C., Spectre, A., Smith, P., Liu, M., & Keltner, D. (2016). Laughter conveys status. 

Journal of Experimental Social Psychology, 65, 109-115. doi:10.1016/j.jesp.2016.04.005 

 
Parr, L., & Waller, B. (2006). Understanding chimpanzee facial expression: insights into the 

evolution of communication. Social Cognitive and Affective Neuroscience, 1, 221-228. 
doi:10.1093/scan/nsl031 

 
Prinz, J. (2006). The emotional basis of moral judgments. Philosophical Explorations, 9(1), 29-

43. doi:10.1080/13869790500492466 

 
Reiman, E., Fusselman, M., Fox, P., & Raichle, M. (1989). Neuroanatomical correlates of 

anticipatory anxiety. Science, 243(4894), 1071-1074.  

 
Schur, M. (Writer) & D. Holland (Director). (2017, Oct 19). [Television series episode] The 

Trolley Problem, The Good Place. USA: NBC. 

 
Schwarz, N., & Clore, G. (2007). Feelings and phenomenal experiences. In A. Kruglanski & E. 

Higgins (Eds.), Handbook of basic principles (2nd ed., pp. 385-407). New Yrok, NY: 
Guilford Press. 

 
Shariff, A., & Tracy, J. (2011). What are emotion expressions for? Current Directions in 

Psychological Science, 20(6), 395-399. doi:10.1177/0963721411424739 

 
Smillie, L., Pickering, A., & Jackson, C. (2006). The new reinforcement sensitivity theory: 

Implications for personality measurement. Personality and Social Psychology Review, 
10(4), 320-335. doi:10.1207/s15327957pspr1004_3 

 
Smithson, M. (2008). Psychology's ambivalent view of uncertainty. In G. Bammer & M. 

Smithson (Eds.), Uncertainty and Risk: Multidisciplinary Perspectives (pp. 205-217). 
London: Earthscan. 

 

 

75 

Tangney, J., Stuewig, J., & Mashek, D. (2007). Moral emotions and moral behavior. Annual 

Review of Psychology, 58(1), 345-372. doi:10.1146/annurev.psych.56.091103.070145 

 
Thomson, J. (1985). The Trolley Problem. The Yale Law Journal, 94(6), 1395-1415.  
 
Young, A., & Saxe, R. (2011). When ignorance is no excuse: Different roles for intent across 

moral domains. Cognition, 120(2), 202-214. doi:10.1016/j.cognition.2011.04.005 

 

76