Kelley’s Covariation Model Definition
Harold Kelley’s covariation model is a central model within attribution theory, an area of social psychology that is concerned with the scientific analysis of the psychology of everyday people. Attribution theory was originally introduced by Fritz Heider in 1958 and assumes that we all want to understand and explain events. For instance, we ask why we succeeded at a task or why our friend liked a movie. The answers to such “why questions” (e.g., “I am smart” or “The movie was good”) are called causal attributions. Kelley’s model explains how laypersons arrive at such attributions; hence, it is a scientific theory about naive theories.
Kelley’s Covariation Model Analysis
For both scientists and laypersons, explanations consist of effects to be explained (e.g., success at a task or liking a movie) and causes that are used as explanations (e.g., high ability or the quality of the movie). Kelley’s model applies to all types of psychological effects that laypersons explain, ranging from achievement outcomes to emotional states, and it can be applied to self-perception (e.g., “Why did I fail?”) as well as to other perception (“Why did you fail?”).
Kelley distinguishes attributions to causes that reside within the person, the entity, and the circumstances. Person attributions (e.g., “She is a movie fanatic” or “She is smart”) rely on stable factors residing within the person to explain, for example, that person’s enjoyment of a movie or his or her success. Entity attributions imply tracing back the effect to stable properties of the object the person interacts with (e.g., we explain the enjoyment with the quality of the movie, or success with task ease). Finally, circumstance attributions are made when explaining an effect with transient and unstable causes (e.g., when enjoyment is traced back to a happy mood or success is attributed to luck).
But how do we come to explain a specific effect with one of such causes?
Kelley postulates that laypersons use methods akin to those used by scientists, most importantly, experiments. In such experiments, independent and dependent variables are differentiated. For instance, a researcher investigating the influence of color on mood will manipulate color as the independent variable (e.g., putting half of the participants in a blue room and the other half in a red room). Subsequently, she assesses, as the dependent vari-able, participants’ mood in both rooms. In such experiments, the independent variables are often conceived of as causes or determinants of the dependent variables (e.g., color might be conceived of as a determinant of mood), and the dependent variables are the effects.
From this point of view, events to be explained by lay scientists (e.g., success or liking a movie) are the dependent variable, and the possible causes of the event are independent variables. For instance, when I succeed at a task and I ask myself, “Why did I succeed?” success is the dependent variable (i.e., the effect) and the possible causes—the task (entity), my ability (person), or luck (circumstances)—are independent variables.
Whether an effect is attributed to the person, the entity, or the circumstances depends on which of the causes (independent variables) the effect (dependent variable) covaries with. Covariation refers to the cooccurrence of the effect and a cause. To decide whether the entity is the cause, one has to assess whether the effect covaries (co-occurs) with the entity—more specifically, whether there is variation of the effect across objects (entities). Covariation with the entity is given when the effect is present if the entity is present and when the effect is absent when the entity is absent. For instance, when a person succeeds at Task 1 but fails at Tasks 2, 3 and 4, the effect (i.e., success) is present when Task 1 (i.e., the entity) is present, and it is absent when Task 1 is absent (i.e., when Tasks 2, 3, and 4 are present). In this example, the effect covaries with the task (entity): The manipulation of this independent variable results in an effect of the dependent variable; that is, the task “makes a difference.” If, however, the individual succeeds at all tasks in addition to Task 1, the effect (success) does not vary with tasks (there is no covariation between the effect and the entity), or the manipulation of the independent variable (task) does not result in a change of the dependent variable (outcome).
Kelley labels information about the covariation between entities and effects distinctiveness. Distinctiveness is considered high when the effect covaries with the entity (e.g., the person succeeds only at Task 1). Low distinctiveness indicates a lack of covariation between the entity and the effect (i.e., the individual succeeds at all tasks; the task “does not make a difference”).
Information about the covariation of an effect with persons is called consensus. If the effect covaries with the person (only Person 1 succeeds at Task 1, and Persons 2, 3, and 4 fail), there is low consensus (the manipulation of the independent variable “person” results in a change of the dependent variable). If covariation with this independent variable (i.e., the person) is lacking, there is high consensus (i.e., everybody succeeds at Task 1). Finally, high consistency reflects that an effect is always present whenever a certain cause (i.e., the person or the entity) is present. By contrast, low consistency is indicative of the fact that an effect is sometimes present when the cause is absent and sometimes absent when the cause is present.
Kelley suggests that there are three combinations of consistency, consensus, and distinctiveness information which give rise to unambiguous person, entity, and circumstance attributions. We make person attributions when the effect covaries with the person and not with the remaining two causes (entity and circumstances). This data pattern characterizes, for instance, a situation in which a person succeeds at a task at which nobody else succeeds (low consensus), if he or she also succeeds at this task at different points of time (high consistency) and performs other tasks just as well (low distinctiveness). In this situation, we should attribute success to the person (e.g., his or her ability).
Attributions to the entity should be made when the effect covaries with the entity (the person succeeds only at this but not at other tasks; high distinctiveness) and not with the person (everybody succeeds at this task; high consensus) or the point of time (the person always succeeds at this task; high consistency). This pattern is again characterized by the fact that the effect (e.g., success) covaries with one cause (i.e., the entity) but not with the remaining two causes (i.e., the person or points in time).
Finally, attributions to the circumstances should be made when there is low consensus, high distinctiveness, and low consistency—for example, when a person who usually fails at Task 1 succeeds at it at a specific point of time (low consistency), other persons fail at Task 1 (low consensus), and the individual fails most other tasks (high distinctiveness). This covariation pattern differs from the cases that lead to person and entity attributions, as the effect covaries with all of the three possible causes and not (as was the case for the ideal patterns for person and entity attributions) with only one cause.
Kelley’s prediction that people make unambiguous attributions to the person, entity, and circumstances in these three patterns of information is empirically well established. The model has sparked numerous theoretical developments and empirical investigations in the field of attribution and causal induction and continues to be influential into the present. It has been used as a normative model to assess errors and biases, and it served as a conceptual tool for the analyses of a wide range of social psychological phenomena ranging from attribution in close interpersonal relations to attributions of changes in one’s heart rate. Current refinements and extensions of Kelley’s model focus on whether it specifies all attributionally relevant information and on the cognitive processes involved in making attributions.
- Forsterling, F. (2001). Attribution: An introduction to theory, research and applications. East Sussex, UK: Psychology Press.
- Kelley, H. H. (1973). The processes of causal attribution. American Psychologist, 28, 107-128.