We explore automatic evaluative processing, specifically how the evaluation of threat can be distinguished from other types of negativity, and how such unique processing plays out from a social cognitive perspective. I apply this threat distinction to the study of many phenomena, both basic and more applied. While we pursue more basic work on threat perception, we also see how threat influences social phenomena, specifically prejudice. We also pursue research on attitudes more generally, specifically those that arise from evaluative conditioning. Please see below for brief descriptions of these lines of research.
The Dual Implicit Process (DIP) Model
I have developed the dual implicit process (DIP) model, which describes two functionally distinct and serially-linked automatic evaluative processes: the first automatic (or implicit) process (i1) is solely oriented toward evolutionarily derived and socially learned threats to bodily harm. This initial process precedes and potentially influences the temporally subsequent automatic (or implicit) process (i2) that encompasses the full evaluative continuum (positive to negative) and includes evaluative information beyond mere threat. These two implicit processes precede and potentially influence explicit and controlled judgments and behaviors. All of these processes feed
information forward and back to influence each other. The way that this dual implicit processing occurs is by taking advantage of an evolutionarily adapted dual route for processing information.
In broad terms, humans have a quick and dirty path for rapidly processing and responding to perceived threats, and another more cortical path that provides more nuanced information.By incorporating into dual process models what we argue is a qualitative difference between automatic threat and other automatic evaluative processing, the DIP model advances our understanding of the entire evaluative process. Instead of lumping automatic prejudice, food cravings, phobias, intimate partner violence, and addictions into the same “implicit” box, we propose that some stimuli and events—specifically those indicating an immediate threat of bodily harm—are processed in a unique fashion.
March, D. S., Gaertner, L., & Olson, M. A. (2018). On the prioritized processing of threat in a Dual Implicit Process model of evaluation. Psychological Inquiry, 19, 1-13.
March, D. S., Gaertner, L., & Olson, M. A. (2018). Clarifying the explanatory scope of the Dual Implicit Process model. Psychological Inquiry, 19, 37-43.
Threat Processing: Distinguishing Threat from Negative Valence
Given the evolutionary significance of survival, the mind might be particularly sensitive (in terms of strength and speed of reaction) to stimuli that pose an immediate threat to physical harm. To rectify limitations in past research, I pilot-tested stimuli to obtain images that are threatening, nonthreatening-negative, positive, or neutral. The important thing about these images is that while both the threatening and negative images are indeed negative in valence, only the threatening images contain actual survival threats. That way, any difference in reactions to these two image sets is not due to their valence, since they are both equally negative, but due to their different threat-relevance.
These images and valence/arousal ratings can be accessed by downloading this folder.
I used these images in three studies using a visual search task, a facial electromyography paradigm (i.e., the startle-eyeblink paradigm), and eye-tracking. In the figure to the right, you can see in the top panel (a) that participants were faster to detect a threatening than nonthreatening-negative image when each was embedded among positive or neutral images. Note the difference in the red and gray bar. Regardless of whether the threatening and negative image was embedded in positive or neutral distractors, people found the threatening image faster. In the middle panel (b) you can see that people oriented their initial gaze more frequently toward threatening than nonthreatening-negative, positive, or neutral images. Note the three red bars all overwhelmingly show that people more often found the threat first. And in the bottom panel (c) you can see that people evidenced larger startle-eyeblinks (a measure of defensive responding) to threatening than to nonthreatening-negative, positive, or neutral images.
I have also presented these images subliminally and measured several physiological and self-reported responses. In the figure to the right, you can see in the top panel that people had larger skin conductance responses over time to threatening stimuli while responses among the other classes of stimuli did not differ from each other. In the middle panel, you can see that, like the above study using visible stimuli, people blinked harder when presented with a subliminal threatening image, and did not blink differentially when presented with other classes of images. Note that the red bar on the left is higher than the other three, which do not statistically differ from each other. Lastly, in the bottom panel, people rated how "good" or "bad" was the subliminally presented stimulus. Here you can see that people rated the threat (red bad) as more bad than the other three, which once again did not statistically differ from each other.
Taken together, this research indicates that the mind initially responds more strongly and quickly to threatening than
nonthreatening-negative (and positive and neutral) stimuli and highlights the nuanced way disparate types of negatively valenced stimuli are evaluated. It also suggests that integrating what appears to be a human sensitivity to threat into social cognitive processes of evaluation in the form of a Dual Implicit Process model could account for a wider array of social functioning.
March, D. S., Gaertner, L., & Olson, M. A. (2017). In harm’s way: On preferential response to threatening stimuli. Personality and Social Psychology Bulletin, 43, 1519-1529.
Data and materials available on the Open Science Framework website.
March, D. S., Gaertner, L., & Olson, M. A. (2022). On the automatic nature of threat: Physiological and evaluative responses to survival-threats outside conscious perception. Affective Science, 3, 135-144
Data and materials available on the Open Science Framework website.
Exploring Mechanisms Underlying Bias
Distinguishing Threat from Valence as a Source of Bias
The threat-valence distinction described by the DIP model described on the top of this page may be particularly relevant for understanding specific group-based prejudices. For example, Black individuals are disproportionate victims of police violence (police are 4, 18, and 3 times more likely to use force on Blacks than Hispanics, Asians, and Whites, respectively). This could merely be due to a stronger dislike of Black individuals relative to members of other racial backgrounds. Alternatively, despite a range of negative stereotypes with which Blacks are associated, this pattern might be primarily due to associations of Black men with physical threat. To test this, we methodologically differentiated threat and negativity in examining anti-black bias.
In Studies 1 and 2, positive, negative, and threatening targets (from the above-described image set) were categorized as "good" or "bad" following Black or White face primes in an evaluative priming task. The speed at which people identify a target is directly tied to how related it is to the prime that came just before it. You can see in the image to the right that, in both studies, participants were faster to evaluate threat but not negative targets following Black versus White primes. This implies that Black specifically primed concepts of threat, but not general negative valence.
Studies 3 and 4 used mouse-tracking, which records the x- and y- coordinates as people move the mouse from the bottom of the screen to choose between two target labels located at the top left and right, respectively, of the screen. Here we assessed the relative strength of Black-threat vs. Black-negative associations by pitting threat and non-threat negative response labels within singular trials (see “TICC” under current research for a full explanation of the mouse-tracking analyses). In each study, we measured how quickly people began to correctly categorize each face. The speed at which they categorize the face is a an indication of the strength of the associations between each race and each target. The earlier they turn to the correct target, the stronger is the association.
In study 3, participants categorized angry Black faces as threatening (“Dangerous” response ) more quickly than White or Asian faces. You can see in the figure to the left that the mouse-path turned toward the "Dangerous" label earlier for Black face than for White or Asian faces. Importantly, how quickly they categorized dangerous Black faces was unaffected by negative (“Depressed”), positive (“Happy”) or neutral (“Calm”) distractor labels. Study 4 replicated these findings using different response labels (positive, dangerous, and negative vs. not-positive, not-dangerous, and not-negative). Again, people more more quickly began to categorize the Black faces as Dangerous than they did the White faces, and this categorization was unaffected by a negative distractor label. These findings demonstrate that Black-threat associations are stronger than Black-negative associations.
Last, Study 5 combined the evaluative priming task of Studies 1 and 2 with the simultaneous pairings of threat and negativity of Studies 3 and 4. Participants categorized non-threatening negative (e.g., awful, disliked, inferior) or threatening word targets (e.g., aggressive, harmful, murderous) following typical Black (e.g., Darnell, DeAndre, DeShawn) or White (e.g. Brad, Connor, Ethan) name primes. You see in the figure to the right that participants were faster to categorize threatening words following Black versus White primes yet demonstrated a no difference in their response to negative words after Black and. White primes. These
results conceptually replicate the findings of Studies 3 and 4, demonstrating that White Americans more strongly associate Black men with threat than negativity.
Methodologically differentiating threat from valence showed that that White Americans automatically evaluate Black men as survival threats. Critically, this work does not suggest that White Americans lack negative (or positive) stereotypes of Black men but rather that White Americans’ initial (or early automatic) evaluation of Black men is that they pose a survival threat. This distinction holds several implications for anti-Black prejudice. Indeed, certain instances of anti-Black bias may specifically result from automatic threat evaluations. The racial disparity in police violence, for instance, may specifically reflect threat responses due to implicit threat evaluations rather than merely reflecting disdain or dislike. If so, interventions aimed at reducing such bias would benefit from particularly addressing threat response or danger associations.
March, D., S. & Gaertner, L., & Olson, M. A. (in press). Danger or dislike: Distinguishing threat from valence as sources of automatic anti-Black bias. Journal of Personality and Social Psychology.
Data and materials available on the Open Science Framework website.
In-group and Out-group Bias Toward Hispanics
While outgroup bias is well studied, ingroup bias has received far less attention. I examined ingroup biases among Latino/a individuals and outgroup biases toward Hispanics among White (Caucasian non-Hispanic) individuals using the startle eyeblink paradigm (a measure of defensive responding), the Implicit Association Test (IAT; a measure of good/bad evaluation), and an explicit self-report measure. Here I used Hispanic and White male faces as stimuli during both the startle task and the IAT. A similar pattern of results were observed for both startle and IAT measures: both groups displayed responses indicative of negative attitudes toward Hispanic male faces relative to White male faces, although less so for Hispanic participants. The startle responses indicated a larger defensive response to Hispanic than White male faces.
A DIP Model (see above) interpretation of these results suggests that that these measures index different aspects of attitudes and tap into different processes. The IAT measuring the valenced group level association, startle measuring lower-level fear associations.
March, D. S.& Graham, R. (2015). Exploring implicit ingroup and outgroup bias toward Hispanics. Group Processes and Intergroup Relations, 18, 89-103.
News Media Depictions of Obama Influence Automatic Attitudes – Implications for the Obama Effect
Positive media depictions of Obama likely contribute to the so-called “Obama Effect.” However, like any attitude-object, effects of those depictions can depend on contextually positive or negative portrayals. We hypothesized that politically conservative news web sites visually depict Obama more negatively than moderate sites, and that incidental exposure to such dissimilar depictions can differentially impact perceivers’ attitudes toward Obama, particularly when pre-existing attitudes are weak. We found that participants with weaker attitudes exposed to FoxNews.com images of Obama (versus all other images) then had the most negative implicit attitudes toward Obama. Thus, incidental exposure to valenced media portrayals can impact attitudes toward public figures.
March, D. S., Kendrick, R., Fritzlen, K., & Olson, M. A. (2016). News media depictions of Obama influence automatic evaluative associations: Implications for the Obama Effect. Social Cognition, 34, 504-522.
For an overview of Evaluative Conditioning, see: March, D. S., Olson, M. A. & Fazio, R. H. (2018). The Implicit Misattribution Model of Evaluative Conditioning. Social Psychological Bulletin, 13, e27574.