No abstract

In this article, we present and test a formal model of carnal attribution that is true to the original conception of attribution theory formulated by Kelley (1967). The Logical Model specifies precisely how information might be encoded from attribution^ vignettes and how subjects might consider different causal loci as necessary and sufficient conditions for the occurrence of an effect. Predictions are derived for each of the eight information configurations on the basis of consensus, distinctiveness, and consistency information. The model is tested with data from three published experiments and is preferred over a template model of attribution. The Logical Model's limitations are also acknowledged in that responses deviate systematically from its predictions, leading to an analysis of the notion of causality underlying stimulus, person, and circumstance attributions.The logic of science is also that of business and life.-John Stuart Mill 'Contrariwise,' continued Tweedledee, 'if it was so, it might be; and if it were so, it would be: but as it isn't, it ain't. That's logic.' -Lewis Carroll, Through The Looking Glass Kelley's (1967) theory of causal attribution proposes that in the case where the perceiver has information from multiple sources, the covariation principle is used. An effect is attributed to the cause with which it covaries over time.In outlining attribution in the case of covariation, Kelley built onto Heider's (1958) proposal that understanding of the distal environment is gained by means of a causal analysis that is "in a way analogous to experimental methods" (p. 297). Kelley was influenced both by Heider's suggestion that social perception was similar to a naive "factor analysis" (Heider, 1958, p. 66) and by Duncker's proposal that cause and effect analysis was achieved by a pseudoscientific "abstracting induction" (Dunckei, 1945, p. 64). In addition, Kelley followed Heider in the use of a naive version of 3. S. Mill's method of difference: "The effect is attributed to that condition which is present when the effect is present and which is absent when the effect is absent" (Kelley, 1967,p. 194). Thus the notion of covariation was conceived, the underlying logic of which, Kelley argued, was similar to that of the analysis of variance (ANOVA).The essence of the ANOVA model can be demonstrated most clearly by considering an example sentence from McArthur's (1972) study: "John laughs at the comedian." This outcome Much of this article was conceptualized before the tragic death of Jos Jaspars. Although it is published as a tribute to bis scholarly work in the area of attribution, it should not be taken as a definitive statement of his views; Miles Hewstone must take full responsibility for its exact content.We gratefully acknowledge critical comments by Denis Hilton, Charles Judd, and two anonymous reviewers on an earlier version of this article.

An inductive logic model of the process of causal attribution is outlined and contrasted with existing formulations of the covariational model of the attribution process. It is shown how the definition of causality employed in the inductive logic model can be extended to the explanation of non-occurrences. Using an elaborated methodology, it is shown that the inductive logic model successfully predicts attributions that may not have been made in previous experiments due to deficiencies in the response format employed. In particular, the inductive logic model is shown to predict some interactional categories of attributional response better than previous models. Finally, the assumptions of the inductive logic model are set in the context of other theories of the attribution process.The prevailing model of the causal attribution process has been the so-called ANOVA model proposed by Kelley (1967). This model drew on the earlier proposal made by Heider (1958) that the lay person uses a procedure analogous to the methods of scientific induction described by Mill (1872). In particular, Kelley (1967) compared the lay attributor's task to the analysis of variance. Thus, Kelley (1972 proposed that the lay person analyses the covariation of the occurrence of the target event with the actor (the person), the entity (the stimulus) and the time of the event (the circumstances). The lay person does this by sampling three types of information: cornernus, describing other people's behaviour to the stimulus; distinctiveness, describing the person's behaviour to other stimuli; and consistency, describing the behaviour of the person with the stimulus at other times.The ANOVA model has been tested and elaborated in a number of studies (e.g. McArthur, 1972; Orvis ef al., 1975). In such studies, subjects are typically presented with short verbal vignettes describing the target event, along with high and low levels of consensus, distinctiveness and consistency information. However, as Jaspars et ul. (1983) have shown, these elaborations on Kelley's (1967) model have important conceptual flaws. In the words of Jaspars (in press): 'Looking back in surprise, one discovers that the so-called ANOVA model does not specify how information is to be encoded; that the stated rules of inference are not developed fully in the seminal publications of Kelley (1967, 1972); applied wrongly in McArthur (1972); and in addition the decoding process is nowhere mentioned.' One purpose of this article is to illustrate how the consequences of these theoretical shortcomings are obscured by deficiencies in the experimental methodology used to test the theory. These are discussed below. Problems with existing formulations of the covariational modelThe first problem with the experiments conducted to test Kelley's (1967) ANOVA model is that they make it impossible for the lay person to calculate an 'intuitive analysis of variance' or its formal equivalent on the information that they are given. Subjects are given information describing the target event, the...

Heider (19S8, p. 113) describes several attribution criteria used by the naive observer to ascribe responsibility, and he orders them according to the relative contribution of environmental and personal forces to the action outcome. These levels of responsibility, namely, actor-outcome association, causality, foreseeability, intentionality, and justifiability, represent a progression from "primitive," undifferentiated cognitive functioning to highly differentiated, "sophisticated" attribution. In describing his schema, Heider clearly implies that the levels may parallel human developmental stages, and he points out the similarity between Piaget's phases and his criteria of causality and intentionality.Several attempts have been made to test developmental aspects of Heider's theory. Perhaps the most widely cited is that of Shaw and Sulzer (1964), which is prototypic of research in this area and hence will be dealt with in some detail. Stories concerning a boy named Perry were administered to two different age groups (6-to 9-year-olds and college students), who rated his responsibility for the story outcome. Each contained only the minimum information to be classified at one of Heider's levels. Shaw and Sulzer found tenta-

Investigated the relationship between social representations, intergroup causal attributions and the search for a positive social identity in two rival groups fromResults are discussed in terms of the influence of social representations on both causal attributions and intergroup differentiation; the existence of intergroup biases in achievement attributions; and the different modes of differentiation chosen by the different status groups. Social Identity Theory is seen as a valuable framework with which to consider these findings.