Harold Kelley (1972) proposed Covariation Theory to explain how we explain the behaviour of others – ie: make attributions – when we have knowledge of their past behaviours and the behaviour of others in the same situation.

Information from several sources/events is used to pin down the covariation of observed behaviour and its possible causes.

Three main types of information are used to interpret someone's behaviour: -

Each of these (CCD) can be high or low and the particular combination of these produced by the situations observed leads to making dispositional or situational attributions.

According to Kelley, it is important to note that consistency has to be high before any kind of attribution can be made. In other words we cannot judge behaviour that takes place irregularly or unpredictably.

problem with the subject and/or the teaching of it.

Evaluation

The theory has been influential and there is some support for the three factors and the covariation principle. Leslie McArthur (1972) gave participants 12 event-depicting sentences that used a range of CCD combinations and found some support for Kelley's predictions about attributions - but participants made less use of consensus than predicted. However, J H Cha & K D Nam (1985) found that Korean participants made more use of consensus information.

Howard Garland, Andrew Hardy & Laura Stephenson (1975) found that, when participants were given greater choice about the information used to make attribution, some chose things like personality rather than consistency, distinctiveness and concensus. In one study it was found that participants only used information about consensus, consistency and distinctiveness when no information about context or situation was available. Lauren Alloy & Naomi Tabachnik (1984) found that, even when information about consensus, consistency and distinctiveness was available, people were not particularly adept at using it in a systemtic way - thereby diminishing its usefulness in making accurate predictions.

It has had some important practical applications – eg: Jim McNight & Jeanna Sutton (1994) investigated suicide in young offenders insitutions, using Covariation Theory. When consensus and distinctiveness were high, they attributed suicide to the environment; when consensus was and distinctiveness were low – eg: only one inmate, with a history of this type of behaviour, made suicide attempts, then they attributed the behaviour to the individual’s personal characteristics.

However, Woo-kyoung Ahn, Charles Kalish, D L Medin & S A Gelman (1995) gave participants a fictional scenario in which ‘Kim’ had been involved in a car crash the previous night. They expected the participants to ask questions related to consensus, consistency and distinctiveness – eg: “Does Kim often have accidents?” or “Were there a lot of accidents last night?” Instead they asked questions such as “Was Kim drunk?” or “Was the road icy?” Ahn & Jeremy Bailenson (1996) suggest that people tend to make judgements on detail, rather than previous information. They call this the mechanisms approach.

In their research Frank Fincham & Miles Hewstone (2001) found that no specific pattern of C, C, and D seems to be associated with dispositional and situational attributions.

To explain how attributions are made when there is no prior information on the person(s) to whom the attribution is made - such a single, one-off event, Kelley developed the idea of causal schemata.

Developed initially from a Psychology Press online article

From the work of Harold Kelley. Graphic copyright © 2001 Psychology Press Ltd

Example

Bill is usually late [high consistency]  to most lessons [low distinctiveness] whereas most students are usually on time [low consensus] – resulting in the teacher’s attributing the lateness to Bill’s character  ie: dispositional. However, if all students [high consensus]are late to one particular subject [high consistency] but on time to most lessons [high distinctiveness], then the behaviour of the students would be seen as situational – ie: there is a