Attribution theory is a general title for a body of theory and research into the ways in which people explain why things happen. By and large the bulk of the work has confined itself to the explanations that people offer for various types of human behaviour, rather than the behaviour of objects, animals or natural forces, and this preoccupation probably reflects Western conceptions and values about the nature of the world. For religious and other reasons, we tend to view ourselves as the focus or centre of the Universe, or as the high point of creation, and hence attribution theory has concentrated on the explanation of human behaviour to the relative neglect of other things. It is clear that from other cultural perspectives, which see humans as part of a larger purposive universal process with a will and/or direction of its own, attribution theory would take a different and rather interesting turn; for example, certain central precepts would simply not make sense from such a perspective (Jahoda 1979).
Nonetheless, within our own cultural framework, attribution theory has offered important insights into the ways in which people explain their own actions and the actions of others; and in the course of that process light has coincidentally been shed on the difference between causal explanations as social constructions (reasons) and causal explanations as scientific statements (causes).
For example, the 'reason' for a particular action is frequently a verbal statement made by an individual when asked a question by a third party, such as 'Why did you do this?' or 'Why did she do that?'. In answering the question the motives, affiliations, intentions and self-perceptions of the person doing the explaining are often reflected in the type of explanation offered. This is regularly seen in party political broadcasts, when members of a governing party explain unpopular policies by reference to circumstances beyond their control; perhaps making use of sentences that begin, 'We in the XXXXXXX party had no choice but to do this because ' On the other hand, popular policies will be explained by reference to internal qualities such as compassion, concern or goodness, and an explanation might lead with, 'We in the XXXXXXX party did this because we felt it was high time that something was done to help the situation of ' Such explanations are easily seen to be primarily social constructions with clear purposes and functions for the person doing the explaining; specifically, the avoidance of blame and the accumulation of personal credit.
By contrast, within our idealised conception of science (see for example the Logical Empiricist view of science advanced by Popper 1959) the question 'Why does water turn to steam when it boils?' is presumed to elicit a scientific statement of 'causality' that is independent of the motives, intentions and self-perceptions of the scientist doing the explaining (an assumption that is by no means truesee for example Kuhn 1970). From this idealised scientific viewpoint, it is assumed that different scientists will offer the same causal account regardless of their own motives, dispositions and propensities; that the explanation will represent the state of knowledge rather than the state of the explainer; and that in some sense the explanation offered will be 'real' or 'absolute'.
Unfortunately, we sometimes lose sight of the distinction between causal accounts that are socially functional, and those which are scientifically functional, and the two become intermingled. Thus we may try to shed light on the causes of theft amongst drug users by asking them to tell us their reasons for stealing; or to investigate the causes of relapse amongst alcohol abusers by asking them their reasons for relapsing. In other words, we can fall into the trap of assuming that reasons provide a shortcut to discovering causes. Even more problematic is the fact that sometimes it is difficult to know which type of account we are dealing with, and that in real-life situations the two may be closely interwoven.
However, by becoming involved in this discussion of social versus scientific explanations we have to some extent jumped the gun. Originally, attribution theory sought to shed light on the nature of people's explanations for everyday events without distinguishing between the social nature of reasons and the scientific nature of causes. These were lumped together under a general banner of 'causal explanations', and it is to this earlier work that we must now turn in order to grasp the fundamental principles of attribution.
The Bases of Attribution Theory
Several accounts of the multi-faceted pedigree of attribution theory are available for the reader wishing to go into more detail than is provided in this chapter. Two of the best are given by Antaki (1982), and by Hewstone (1983) and it is worthwhile highlighting the salient features from these accounts.
The basis for attribution theory is the desire to understand how people arrive at common-sense explanations for their own and for others behaviour. The original stimulus for the work came from Heider (1958) in a much-cited work, The Psychology of Interpersonal Relations, in which it was suggested that a major task for anyone trying to understand the social and physical world was to produce satisfactory accounts of why things happened. Insofar as this applied to understanding why people do the things they do, this amounted to finding satisfactory causal accounts of behaviour; and furthermore, since searching for explanations is something that scientists do, this amounted to regarding people as if they were in some sense natural or primitive psychologists. From this basis, the idea of man as a naive scientist began to emerge, this notion deriving from the accumulating evidence that people made inferences about the causes of human behaviour on the basis of their observations of social acts, in the way that the scientist or physicist makes inferences on the basis of observations of physical events. The task then became one of finding out how such causal inferences were made, and illuminating the kinds of evidence involved in the process.
More importantly however these causal inferences, 'describing and predicting events as a science should do' (Hewstone op cit), were held to have important implications for behaviour whether they were 'true' or not. With respect to addiction, for example, this would imply that belief in the inability of addicts to control their own drug use, would have important behavioural implications whether such a belief were true or not.
Attribution as Lawful Explanation
Subsequent to Heider's original ideas, Jones and Davis (1965) are usually credited with making the next major step forward with their theory of 'correspondent inferences'. This theory sought to explain how far a person's actions could be accounted for in terms of the traits, dispositions and intentions of the person doing the act (known in attributional parlance as the 'actor'), rather than in terms of situational or other 'external' factors. The empirical work on this topic centred around 'common and non-common effects', and the reader is referred to either of the two texts cited above for an explanation of these terms.
However, the development by Kelley (1967) of the ANOVA model of causal inference merits closer inspection within this present text, because it presents a very clear picture of the type of thinking which can underlie the construction of causal explanations of human action. The model conceptualises the causal attribution process as hinging around the covariation of three dimensions, the title of this theoretical approach (the 'ANOVA' model of attribution) deriving from a loose analogy with analysis of variance. Kelley's approach is particularly useful as it illustrates one of the central features of attribution theories in a very graphic and comprehensible manner; namely, the fact that the explanation postulated for some action results from the way in which that situation is perceived by the person constructing the causal account (the 'observer').
Kelley's formulation of the attributional process was first revealed in a widely-read paper presented at the Nebraska Symposium on Motivation, in 1967. He proposed that the type of explanation a person is likely to offer for an observed piece of social behaviour depends on the interaction (the "covariation and configuration") of three factors. At risk of doing some violence to the original conceptions, the three factors Kelley proposed were in essence:
i) consensus given that I have seen this person doing this thing, are other people doing it also? To the extent that the answer is "yes", there is high consensus. If "no", there is low consensus.
ii) consistency does this person do this thing repeatedly, or regularly? To the extent that the answer is "yes" there is high consistency. If the behaviour is rare or a single instance, there is low consistency.
iii) distinctiveness considerthe object of this behaviour. Usually, this is the person who is having this thing done to them. Is he/ she the sole recipient, or do other people have this done to them also? To the extent that the object person is unique, there is high distinctiveness. To the extent that he/she is only one of a number who receive this treatment, there is low distinctiveness.
It is important to bear in mind that information about consensus, consistency and distinctiveness can come to the observer through channels other than direct observation. Regardless of Heider's, Jones and Davis's, or Kelley's original intentions, the general model does not in principle appear to require first hand information, but can be applied to reported information, or even to beliefs or preconceptions (referred to in the literature as "causal schemata") formed in the absence of literal multiple observations. Whilst attribution theorists sometimes conceptualise such a situation as one of incomplete data, this is perhaps misleading if we thereby conclude that the resulting causal account must, as a consequence, necessarily be less powerful or persuasive than one based on direct observation. In fact a persons beliefs about, say, the consistency of an act (for example, that "alcoholics" inevitably relapse after one drink) can be plugged into the model as readily as can direct observations, where their influence will be as powerful as direct observations if the beliefs in question are sufficiently strongly held.
Whilst the detailed explication of Kelley's system can become almost as complex as one wishes to make it, a process involving independent consideration of "persons, entities, and time" and their interactions, in practice it is easy to elucidate at a basic level in the following way.
Suppose we observe one day that Tom is hitting Mary, and that we have information available (of whatever typeobservations, second-hand reports, stereotypes, prejudices etc.) about the pattern of consensus, consistency and distinctiveness surrounding this act. For example, it is just Tom who hits Mary (low consensus), he seems to hit Mary quite often (high consistency) and he also hits other girls (low distinctiveness). In these circumstances, we are likely to explain the event in terms of a negative property of Tom; he is aggressive, unpleasant, a bully, and so forth.
Imagine, however, that our information suggests high consensus (other children also hit Mary); high consistency (they hit her often); and these children are not generally noted for hitting other people (high distinctiveness). In these circumstances, we are likely to attribute the act to some disposition of Mary; perhaps for example there is something she repeatedly does that tests everyone's patience. Whatever the truth of the matter, we are likely to attribute the behaviour to a negative property of Mary.
Developing the above lines of argument, it is easy to conceptualise the Tom-hits-Mary scene in tabular form as follows, with each of the three variables being potentially high (HI) or low (LO), leading thereby to particular types of explanations for the observed act.
TOM HITS MARY consensus HI or LO consistency HI or LO distinctiveness HI or LO
Using Kelley's three dimensions, we have seen how the pattern LO, HI, LO leads to explanation in terms of negative attributions about Tom; whilst HI,HI,HI leads to negative attributions about Mary. It is amusing and instructive to consider other alternatives, and to try and predict the type of explanation which might be forthcoming. Some patterns are quite easy, others are more subtle. For example, HI, LO,HI implies that there are particular situations in which Mary gets on everyone's nerves (i.e. a person x situation interaction); and LO, LO, HI suggests that the incident was due to some unfortunate and unforeseeable circumstance.
These simple examples based on Kelley's notions serve to illustrate two of the general points made earlier. Firstly, people construct explanations of social behaviour in a manner which is psychologically dynamic rather than primarily veridical. In fact, the attribution process has nothing to say on the issue of whether explanations constructed in these terms are true or not. Secondly, the process also appears to be lawful, Kelley suggesting a way of conceptualising it in terms of three major building blocks out of which explanation is formed in a logical or quasi-logical fashion.
Viewed in this way, social explanation is not based on any knowledge of actual causality, but is an inference made on the basis of certain social features of the act about which the observer has information of some kind. Consequently, the account might be "true" or not. In addition, as we shall see later, this approach is not specific to the explanation of other people's behaviour, since there is research indicating that the explanation of our own actions can also be handled within the attributional perspective.
The Work of Michotte and Heider on the Perception of Causality
In tracing the development of attribution theory, however, it is useful to go back to Heider and consider the influence of Gestalt psychology and specifically certain experiments by Michotte (1946) into "phenomenal causality" which clearly influenced Heider's thinking. In the cited studies, Michotte was primarily interested in perceptual processes, notably visual perception. Within that historical context, the word perception retained its classical meaning and was considered to be distinct from cognition. It certainly had none of the social-cognitive implications of more modern usages as embodied in phrases like "interpersonal perception". Nonetheless, Michotte's studies elegantly illustrate the manner in which perception and cognition are closely intertwined.
Michotte investigated the ways in which people perceived causality. His experiments used visual stimuli consisting of combinations of blobs, circles, squares and lines which appeared to move in certain ways; Michotte achieving his visual effects by means of revolving discs on which were drawn distorted circles, only portions of which could be observed through a slot. In one example, a small black square would proceed across a plain visual field towards a second, stationary, square. The first square would appear to strike the second. After the apparent impact, the second square would move away in the same direction whilst the first one would stop. When viewing this and other similar visual arrays, people would describe the events in causal terms. The movement of the second square would be seen as caused by the impact of the first square. In a similar way, other stimuli would give rise to the phenomenological impression of a causal chain of events when visual representations of objects (usually geometrical shapes) appeared to strike each other "causing" various kinds of reactive movement. This impression of events as caused by "collisions" or "rebounds" was sometimes very strong, and by adapting and making use of certain Gestalt-type principles such as contiguity, proximity and good continuation Michotte was able to show how the perception of causality could be enhanced or reduced in ways that were predictable by the experimenter.
In conceptual terms it is not difficult to find parallels between Kelley's attempts to define the nature of social explanations in terms of his three-dimensional ANOVA model, and Michotte's much earlier work to find a similar interactive model to define the phenomenal experience of causality in the visual mode, using Gestalt principles.
Importantly, at roughly the same time as Michotte was working on phenomenal causality in the visual mode, Heider was doing similar work with a social import to it. In a study by Heider and Simmel (1944), subjects were presented with visual stimuli of the type illustrated below in Fig. 1.
Fig. 1. Stimulus of the type
used in the Heider studies
In the example given, the small circle moved around inside the square, followed closely by the large triangle. From time to time the small circle bumped against the sides of the square. Subsequently, the small triangle (outside the square) moved into contact with the top right-hand side of the square, and after a brief pause moved away again. At this point, the upper section of the right-hand side of the square opened, making a gap in the perimeter as shown in the figure. The small circle then went through the gap so created.
When viewing stimuli of this type, subjects who were asked to describe what was happening gave explanations which were causal, and which ascribed social motives and intentions to the objects. The circle was the heroine, pursued by the wicked villain (large triangle) from whom she was trying to escape. Subsequently, the hero (small triangle) saved the heroine by opening the door for her.
The reason for taking a step backwards to the work of Michotte and Heider in the 1940's is simple. In the examples cited above, the stimuli consisted of nothing more than moving drawings of meaningless objects (meaningless in the sense of having no particular referent) in a purely visual display. There were no physical collisions, no rebounds, no people, no pursuits, no villains or heroines and so forth. None the less, when viewing such stimuli, people produced causal accounts in the physical domain (action/ reaction) and in the social domain (pursuit, escape, rescue). These facts can be summarised simply by saying that people regularly and predictably see patterns of causality where such causality does not exist. Alternatively, if we wish to avoid the philosophical criticism that phenomenal causality is the only type of causality there is (Hume 1739-40), we would have to qualify this by saying that people regularly see physical and social causality in circumstances where a 'scientific' account would ascribe no such relationships.
Deriving from the above paragraph, we now see that there is great scope for confusion when people view objects that 'really do' knock each other over, or people who 'really do' chase each other about. In the early work of Michotte and Heider we have little difficulty in realising that the causal accounts provided are primarily psychological constructions, rather than accounts of scientific causality; mainly because we can see that the stimuli do not actually warrant the latter type of description. Squares and triangles, for example, cannot have motives and intentions.
Suppose however that the identical causal account derives from circumstances in which real objects or real people are involved. The tendency in this instance is to see the causal account as in some sense 'real', simply because such an account is possible given (our knowledge of) the nature of the situation, which is coincidentally 'real'. Consequently, the explanation offered looks feasible or appropriate in these circumstances. In fact however, the causal account may be identical to, and constructed from the same elements as, that produced in response to artificial stimuli. Viewing 'the real thing' does not invoke a different explanatory process, nor does it guarantee a different or more 'true' type of explanation. We have to conclude therefore that the 'truth' of a causal account cannot be inferred from looking at the circumstances surrounding it, although it's salient features may well be predictable from those circumstances.
It is worth stressing that, given the quasi-logical nature of the psychological processes involved, most such accounts will probably seem eminently plausible in an appropriate context, without this in any way attesting to the veridical value of the account; and different people possessing different knowledge will produce different but equally plausible accounts. For example, explanations of 'why people use drugs' given by drug users may differ from explanations given by non-drug users. Furthermore, as we shall see in a later chapter, features of drug users explanations may well be predictable from knowledge of the extent and pattern of their drug use. However, their explanations cannot be assumed to be qualitatively different from other people's explanations, or to be more 'true', simply by virtue of the fact that drug use is for them 'real'. Least of all is there any reason to presume that their causal accounts of drug use are scientific where other people's are social constructions.
As both Michotte and Kelley have implied, causal accounts are constructed according to certain rules which have no bearing on whether the account is ultimately veridical or not. This principal remains intact, even if we argue with Michotte or Kelley about the precise components they use in their respective models. Accordingly, within the constraints of a particular phenomenon, the investigation of people's causal accounts of that phenomenon has nothing in principle to do with the investigation of its causality as conceptualised by 'science', and we cannot carelessly use the one as a means of investigating the other.
Extending the Framework of Attribution
So far, a picture of attribution theory has been presented which seeks to point out various differing approaches to the psychological study of social explanation. Rather than being a single entity, attribution theory is more a group of theories held together by a common agenda within which coherent strands may be discerned. However, the account is so far limited in two ways.
Firstly, there are clearly a number of possible solutions to the problem of deciding what the most important dimensions of attribution are. One does not have to accept Kelley's, or anyone else's, dimensions as definitive since the basic principle of quasi-logical inference remains unaltered even if changes are made to the components. Second and more important, however, is the fact that classical attribution theory concentrates on the ways in which the inference process works; that is, on the manner in which people derive their explanations. The problem of how those explanations might relate to present or future actions has remained unaddressed until more recently, and it is to these two issues that we now turn.