Project topics and materials
Get Complete Project Material File(s) Now! »
Attitudes: History, Usage, and Challenges
One of the key concepts adopted by social psychologists can be traced back to a chapter in The Handbook of Social Psychology (1935). In this chapter, Allport (1935) reviews the use of the term attitude and provides a definition in terms of “neuropsychic state of readiness for mental and physical activity”. This definition was the first step on a journey to better understand consistency in human behavior. As reported by Allport (1935), at the time, it was common for scientists interested in understanding behavior to debate on whether human behaviors were a result of heredity or environment. By introducing and putting forward a concept for which no origin was assumed, merely describing a probability to act a certain way, social psychologists could start to focus on the factors making people likely to act in a predictable manner, no matter whether these factors were internal or external (Kelley, 1973). Allport’s review on attitudes (1935) is often perceived as the seminal paper that helped popularize the use of attitudes among the social psychologists. However, at the time, this concept was facing many challenges.
An illustration of one of these challenges can be illustrated by a study conducted to investigate prejudice. In the ‘30s, LaPiere (1934) tried to capture attitudes—the state of readiness to act in a certain way described by Allport (1935)—to see whether it was possible to predict people’s behavior. Along with a Chinese couple, he spent two years traveling across the US, going from hotel to hotel. When their trip came to an end, LaPiere mailed one question to the hotels they visited: Would they accept “members of the Chinese race” in their establishment? The result of this field study could have marked the end of attitude research: While 92% of the establishments responded they would refuse to welcome Chinese individuals, none but one establishment actually refused to welcome LaPiere and the accompanying couple. The problem illustrated here is actually simple to understand: It is hard to know how to capture the state of readiness people have. As later phrased in the literature, “Asking people to report on their attitudes, will almost always result in an answer—but it often remains unclear what exactly this answer means” (Schwarz, 2008, p. 49).
The Early Works on Attitude
Years passed and some researchers argued that attitudinal research in social psychology was a dead end (e.g., Wicker, 1969). It was only when researchers focused on building theories of attitudes that this field of social psychology gained in popularity. Researchers started theorizing how attitudes worked and what their relationships with behavior were. The reflections soon translated into principles which are well known to attitudinal researchers, such as the correspondence principle. If someone has to report their attitude toward something, the closer the attitudinal question is from behavior in terms of generality-specificity, the higher the correlation between the attitudinal question and the behaviors (i.e., the degree of correspondence between attitudinal and behavioral entities is important; Ajzen & Fishbein, 1977). For example, asking people their general views on the use of contraceptives (something general) should only be able to predict weakly their use of condoms (something specific). This example might seem trivial to the reader, but, as we will review, many not-so-trivial reflections came from the social psychology literature regarding how we should capture attitudes and when these attitudes are supposed to influence behaviors.
In the present dissertation, we will focus on some of the theories which were developed in the field of attitudinal research, as well as the instruments related to these theories. These theories and measures emerged from what is now called the implicit social cognition literature—a subfield of social psychology which emerged in the 90s (Greenwald & Banaji, 1995; see also, Gawronski & Payne, 2010). Specifically, we will focus on some of the assumptions related to the role of repeated experience which were embedded in the theories of implicit social cognition. Later, we will discuss why this subfield emerged, how it emerged, what principles emerged from this literature, and what kind of expectation one can have based on those principles. Beforehand, however, we will take the time to formalize how we will discuss attitudinal research. Indeed, many definitions coexist in the literature (Bohner & Dickel, 2011), and we think that the more precise we are, the lower the risk for what we offer to become impossible to falsify.
A Common Language for Attitudinal Research
As we just saw, attitude is one of the concepts used by social psychologists to understand consistency in the behaviors of human beings. Initially defined as “a mental and neural state of readiness, organized through experience, exerting a directive or dynamic influence upon the individual’s response to all objects and situations with which it is related” (Allport, 1935), attitudes would help understand and predict why we act one way or another. Decades after its initial introduction, the attitude concept now often translates as a mental construct which has a significant causal impact on one’s behavior. Attitudinal research is oriented to understanding how attitudes are formed and when they will be activated (i.e., how attitudes appear, when they will influence behavior). It has to be noted that the advances in attitudinal research resulted in many theoretical models, each endorsing a different definition of attitude (see Bohner & Dickel, 2011). This situation follows Nolan Bushnell’s Law: While it is easy to intuitively see what an attitude refers to (“Do you like ice cream or not?”), there is a lot of debate on what a good definition of attitude is.
In the present dissertation, focusing on the contribution of repeated experience will lead us to explore the predictions of several theoretical models of attitude. Because each of them adopts different assumptions regarding what an attitude is, we will refrain from making this term the core concept of this dissertation. Instead, we will use a meta-theoretical framework allowing us to describe the many research conducted by attitudinal researchers in a common vocabulary. In doing so, we hope to avoid a situation where we would endorse theoretical assumptions without being explicit about it. This meta-theoretical framework, initially described by De Houwer et al. (2013), was designed with the goal to provide a language for researchers to talk to each other, as well as to make it easier to avoid common pitfalls of attitudinal research. In the present section, we will describe the concerns that De Houwer et al.’s (2013) meta-theoretical framework is addressing, and the perspective it adopts to do so.
Before describing one of the issues addressed by De Houwer et al.’s (2013) meta-theoretical framework, we will first have to define some vocabulary. Two terms will be especially useful: explanans and explanandum. This vocabulary, related to the idea of explanation, targets two different things. The concept of explanandum describes something that researchers are interested in understanding—in explaining. Storms (or human behaviors), for example, could be something researchers want to understand; it could be an explanandum. To understand something, researchers usually describe processes whose goal is to explain the phenomenon. In the context of weather, to pursue with this example, scientists could explain storms as a result of a process involving differences in atmospheric pressures (or human behaviors as driven by attitudes). The processes explaining our object of interest are called the explanans. In attitudinal research, the thing that researchers are interested in predicting— the explanandum—is the human behavior and, to better understand the human behavior, the recruited concepts—the explanans—are attitudes. A useful mnemonic is that explanans sounds more like “explaining” than explanandum, and the explanans does refer to the things explaining a phenomenon.
On the Importance of Keeping Explanandum and Explanans Independent
For a theoretical model to be efficient, it is especially important to keep the explanans and explanandum distinct from each other. One of the pitfalls in attitudinal research is that the explanans and explanandum are often conflated with each other. In other words, for De Houwer et al. (2013), it is sometimes hard to know whether something is a behavior that researchers are interested in understanding (the explanandum) or an attitude (the explanans). To understand this idea, we will focus on a practice which is common among attitudinal researchers: using behaviors as a proxy for attitudes. For example, a researcher interested in math-related behavior could use the behavioral response of a participant on an evaluative rating scale (“from one to ten, how much do you like math?”) as a proxy for attitudes toward math.2 Some assumptions, however, need to be made so that this specific behavior can be considered as a good proxy.
Here, in the math example, researchers are inferring the existence of an attitude (X) from a behavior (Y). Yet, the core assumption of attitudinal research is that attitude (X) influences the behavior (Y). In formal logic, inferring the existence of X from Y based on the premise “if X, then Y” is known as affirming the consequent. It is considered as a fallacy as soon as Y varies because of factors other than X. In attitudinal research, there is plenty of evidence showing that an attitudinal measure can vary because of factors other than a change of attitude (e.g., social desirability; DeMaio, 1984). In our example, to draw any conclusion regarding the link between attitudes toward math and math-related behavior, we would have to assume that there is no factor other than attitude which influences the attitude measure. It is easy to imagine that the answer to the measure assessing how much people like math depends on who asks the question. In a context where the mental mechanisms used to explain a phenomenon are the one that needs to be explained, attitudinal measures become a behavior just like any other (De Houwer et al., 2013). Many researchers have outlined this recurring problem in attitudinal research (e.g., Fazio, 2007; Krosnick et al., 2005).
In their meta-theoretical model for attitudinal research, De Houwer et al. (2013) offer a vocabulary which makes it easier to keep explanandum and explanans independent from each other. In adopting this vocabulary, besides avoiding some of the pitfalls of attitudinal research, we can describe any of the theoretical models of attitude in common terms. In their meta-theoretical framework, De Houwer et al. (2013) suggest that researchers should focus on evaluative responses (the explanandum) and models explaining evaluation (the explanans). In the next section, we will introduce these two notions which will be used in lieu of the traditional attitude concept.
Studying Evaluative Responses and Evaluation to Understand Behavior
Evaluative Responses, or the Functional Level
In De Houwer et al.’s (2013) meta-theoretical framework, the explanandum—the phenomena that researchers are interested in explaining—are evaluative responses. Evaluative responses are a behavior that can be defined according to two principles: scientific conventions and correlations with other evaluative responses. With the first principle, if it is consensual among scientists that circling a number on a questionnaire asking “How much do you like math?” is an evaluative response, then it will be considered as an evaluative response. With the second principle, if a behavior correlates with one which belongs to the category of the evaluative responses, then it will also be considered as an evaluative response. At first glance, defining the explanandum in such a manner appears problematic because it seems arbitrary. However, with regard to the conflation between the explanandum and the explanans, this definition is actually useful.
As we discussed above, the explanandum and the explanans are conflated when researchers equate behavioral observations with mental constructs, as it is the case when someone considers that an answer to the “Do you like math?” is an attitude (De Houwer et al., 2013). Here, researchers would use a behavior as a proxy for attitudes; the same attitudes that are supposed to account for behaviors. Such recursive logic is what we are trying to avoid as it creates dilemmas impossible to solve. Note that these considerations generalize to any situation where a behavior would be used as a proxy for a mental concept aimed at explaining a phenomenon (e.g., spreading-of-alternative as an index for cognitive dissonance; Brehm, 1956; De Houwer et al., 2013). Focusing on evaluative responses does not pose that kind of problem because no behavior is defined by mental mechanisms that underlie them.
As we just said, in De Houwer et al.’s (2013) meta-theoretical framework, the explanandum do not embed any cognitive assumptions in their definition. Yet, it is possible to gain knowledge related to human behavior. This knowledge will fit in what De Houwer et al.’s (2013) call a functional level of explanation. In this functional level of explanation, it is possible to build knowledge by investigating causation in evaluative responses. Indeed, there is no need to assume mental mechanisms to be able to observe, for example, that a change in the environment results in a change in the evaluative response (e.g., Wittenbrink et al., 2001) or that an evaluative response predicts another (e.g., Necka et al., 2015). Mere experimental method is usually enough to establish causation in psychology (but see also Rohrer, 2018). Here, the functional level of explanation is called so because it allows researchers to describe evaluative responses as a function of a factor. Note that the functional level has nothing to do with something functional from an evolutionary perspective. By comparing evaluative responses that an individual produces in the presence of a particular stimulus compared to a situation where the stimulus is absent, researchers can establish that the said stimulus influences have a causal impact on the evaluative response. This ability to find causal effects in the environment is critical to foster research related to evaluative responses.
Still, one can have the feeling that relying on this functional level of explanation would be equivalent as drawing arrows between stimuli and behaviors—that is, behaviorism. One could argue that the time has passed when we considered the human mind as an inaccessible black box. This view notably accompanied the development of cognitive psychology (see Miller, 2003): a lot of the progress observed in modern psychology came from the modeling of unobservable cognitive mechanisms. So how does the meta-theoretical framework offered by De Houwer et al. (2013) overcome the limitations of behaviorism?
Evaluation, or the Cognitive Level
At first glance, the described meta-framework promotes a black-box-like vision of human mind in which researchers can do nothing but observe relationships between stimuli and responses as the only means to understand behavior. This argument, however, only holds if we ignore the second level of explanation this framework offers. In the meta-theoretical framework of evaluation (De Houwer et al., 2013), researchers are interested in both a functional and a cognitive level. The second level, the cognitive one, describes where the cognitive processes underlying evaluation will be formalized—evaluation referring to the effect that factors have on evaluative responses. The same way functional relationships between evaluative responses were the main focus of the functional level, cognitive explanations of evaluation will be the main focus of the cognitive.
Researchers consider an evaluation to have happened when they can demonstrate that an evaluative response is caused by a specific factor (for example, when they can demonstrate that the response to the question, « How much do you like math? » is affected by who asks the question). Critically, modern cognitive psychology argues that every impact a stimulus has on behavior must be the result of mental representations produced and reactivated via certain mental processes (contrary to an explanation which would rely on something happening outside of one’s cognitive system). At the cognitive level, researchers will formalize the processes mediating the effects of various factors on evaluative responses (i.e., evaluations; De Houwer et al., 2013). In such a framework, attitudinal literature aiming at explaining the circumstances under which attitudes influenced behavior as a collection of hypotheses regarding evaluation. These hypotheses address questions regarding why a stimulus can cause a specific evaluative response or why this evaluative response can change depending on factors in the environment.
Note that while the cognitive level of explanation is purely theoretical and will never be directly observable, it is still possible to assess its validity (Gawronski & Bodenhausen, 2015). In order to be useful, a cognitive theory must organize the phenomena observed at the functional level of explanation accurately and must allow researchers to make new predictions regarding evaluation. Any theory failing any of these two objectives would eventually disappear. Together, the functional level and the cognitive level form a meta-theoretical model of evaluation.
A Framework to Study Attitude
These two levels of explanation, when combined, form a meta-theoretical framework that researchers may employ to study behavior without conflating the explanandum and the explanans. Critically for the present dissertation, this meta-theoretical framework’s scope will allow us to describe several models in a common vocabulary. These models were initially framed as models of attitude, however, any research subject which has been the focus of attitudinal research can be reframed in this meta-theoretical framework (De Houwer et al., 2013). Research on acquisition of new attitudes can be reframed as the investigation of the effect of stimulus history on evaluative responses, research on activation of attitudes can be reframed as the investigation of the effect of stimulus context on evaluative responses, and research on the effect of attitudes on behavior can be reframed as the investigation of the varying nature of evaluative responses toward a stimulus (De Houwer et al., 2013). Contrary to traditional attitude research, however, this meta-theoretical framework is designed to keep the things it has to explain and the concepts used to account for them—the explanandum and the explanans—clearly separated.
While distinct, the functional and cognitive levels of explanations mutually support each other. In the words of De Houwer et al. (2013), the knowledge gained at the functional level (i.e., knowledge about how evaluative responses relates with each other and what affects them) represents the foundation of the cognitive level—the data that the cognitive level of explanation will have to account for. A better understanding of evaluative responses at the functional level therefore represents a higher level of constraints on the theories formalized at the cognitive level, which usually results in more precise hypotheses. Any progress at the functional level translates in more precise theories at the cognitive level and any progress at the cognitive level of explanation translates in theories more effective at organizing what we know about the factors driving evaluative responses (De Houwer et al., 2013).
Table of contents :
PART I: THEORETICAL BACKGROUND
Chapter 1 · Attitudes: History, Usage, and Challenges
The Early Works on Attitude
A Common Language for Attitudinal Research
On the Importance of Keeping Explanandum and Explanans Independent
Studying Evaluative Responses and Evaluation to Understand Behavior
The Functional-Cognitive Framework of Attitude in the Present Manuscript
Some Limitations of Self-Report Measures of Attitudes
Attitudes? What Do People Even Know about Them?
Attitudes? Would People Even Want to Express Them?
Beyond the Limitations of Self-Report
Chapter 2 · The Revolution of Indirect Measures
The Evaluative Priming Task
The Implicit Association Test
The Theoretical Origins of Implicit Social Cognition
The First Root of Implicit Social Cognition: Automaticity
The Second Root of Implicit Social Cognition: Awareness
The Current Landscape of Implicit Social Cognition
Chapter 3 · Theoretical Models of Evaluation
The Two Types of Processes in Dual Models
Associative Processes
Propositional Processes
Theoretical Models of Evaluation
The SEM
The APE Model
Single Process Propositional Models
Theoretical Models of Evaluation and the Role of Actual Experience
Chapter 4 · Observing Evaluative Learning
Paradigms Based on Repeated Experience
Evaluative Conditioning
Approach and Avoidance Training
Approach-Avoidance Training and Theoretical Models of Evaluation
Learning Through Instructions
Chapter 5 · Overview of the Present Dissertation
PART II: EXPERIMENTAL PART
Chapter 6 · A VAAST-based approach-avoidance training
Implicit Identification and the Gender Gap in Science
Increasing Math Implicit Identification with an Approach Training
Assessing the Evidential Value in Favor of the Math Approach Training
Increasing the Evidence for the Math Approach Training
Overview of the Experiments
Experiment 1a
Method
Analysis and discussion
Experiment 1b
Method
Results and Discussion
Experiment 2
Method
Analysis
Complementary analyses
General Discussion
Conclusion
Chapter 7 · Comparing Approach-Avoidance Procedures
Repeated Experience and the Acquisition of Indirect Evaluative Response
Approach-Avoidance Procedures
Capturing Evaluation with the Affect Misattribution Procedure
Experiment 3
Method
Results and Discussion
Experiment 4
Method
Results
General Discussion
Conclusion
Chapter 8 · Approach-Avoidance Effects in Incidental Designs
Approach-Avoidance Training Design
Incidental Approach-Avoidance Training
Improving Woud et al.’s Paradigm
Experiment 5
Method
Results
General Discussion
Conclusion
Chapter 9 · Effects of Approach-Avoidance Procedures Over Time
On the Stability of Indirect Evaluative Responses
Long-Term Encoding and Indirect Evaluative Responses
Approach-Avoidance Training to Sustainably Acquire Evaluative Responses
Experiment 6
Method
Results and Discussion
Experiment 7
Method
Results and Discussion
General Discussion
Conclusion
Chapter 10 · Lay Beliefs About Approach-Avoidance Procedures
The Effects of Approach-Avoidance Training and Approach-Avoidance Instructions
On the Importance to Investigate People’s Lay Belief
People’s Lay Theories Regarding Approach-Avoidance Procedures
Experiment 8
Method
Results and Discussion
Experiment 9
Method
Results and Discussion
General Discussion
Conclusion
PART III: GENERAL DISCUSSION
Chapter 11 · General Discussion
Summary of the Results
Implications of our Results
Models of Evaluation
On the Effect Size of the Difference Between Approach-Avoidance Procedure
Other Research Focusing on the Contribution of Repeated Experience to
Evaluative Learning
Methodological Limitations and Ways Forwards
Concluding thoughts
References
APPENDICES
