Reliability of Cognitive Processes and Biases: Illusory Correlations

While our species name, Homo sapiens, implies we are “wise”, we are not always rational thinkers. As your study of the Cognitive Approach has shown, we are subject to biases in thinking and decision-making. One such bias is a tendency to identify correlations where there are none. This tendency was investigated by Hamilton and Gifford (1976). Hamilton and Gifford introduced their participants to two groups, one large Group A (which was made up of 26 people) and one small Group B (which was made up of 13 people). So, the participants read about 39 individuals, one at a time. They were informed about each individual’s group membership and about a behaviour performed by that individual. For example, they read “Bruce, a member of Group A, did volunteer work for a church” or “Joe, a member of Group B, made his friend very uncomfortable by his sarcastic remarks”. The behaviours that the group members performed were either positive, helpful behaviours or negative, hurtful behaviours.

The information about the group members was created such that Group A was twice as big as Group B (that is 26 members versus 13 members) and positive behaviours were more than twice as common than negative ones. Said differently, the ratio of positive to negative behaviours was identical for both groups. Have a look at the table below and focus for now on the columns labelled ‘Actual’. That is the actual ratio of helpful to hurting behaviours was identical for both groups (i.e., 9:4).

	Group A		Group B
	Actual	Perceived	Actual	Perceived
Helping	18	17	9	10
Hurting	8	6	4	6

So, there is no increased correlation of negative behaviours for members of group B, as the relationship between group membership and hurtful behaviours was the same.

Despite this lack of actual correlation, the researchers predicted that participants would see an illusory correlation. Hamilton and Gifford predicted that because members of group B performing the less frequent, negative behaviours were quite rare and distinctive, they, therefore, expected these behaviours to be especially memorable and so to exert undue influence on participants’ impressions of that group.

That is exactly what happened. After reading about the 39 people, the participants were asked to estimate how many members of each group performed helpful and hurting behaviours.

Now, look back at the table and focus on the column labelled ‘perceived’ and you will see the participants’ estimates about how many members of A performed positive and negative behaviours and how many members of B performed positive and negative behaviours. These results show that the participants saw an illusory correlation. They overestimated the frequency with which members of the smaller group B performed the rarer negative behaviours, in contrast, they were quite accurate about group A. As a consequence, they saw membership to group A associated with positive behaviours and membership to group B associated with negative behaviours.

Now consider the implications of illusory correlation for the stereotyping of minority groups. Negative behaviours tend to be relatively infrequent. Members of minority groups performing such infrequent behaviours will be particularly distinctive, noticeable and memorable. As a result, their group may be viewed as particularly likely to perform such negative behaviours, even if it is not.

It seems for example that the stereotype of Australian Aboriginal people being affected by alcoholism may partly reflect an illusory correlation due to the high distinctiveness of Aboriginal and of alcohol abuse. Hence, the work by Hamilton and colleagues suggests that even in the absence of prior expectations, minority groups may be negatively stereotyped because of the special attention paid to minority members performing relatively rare negative behaviours.

Thanks to Kylie McIntyre for this analysis of Hamilton and Gifford (1976).

Sign up to our monthly newsletter to keep up to date with all the latest Pamoja news.