First Impressions

(Editor:   James Collins)

Introduction

The "First Impressions" experiment is a computer-based version of a classic study conducted by Hamilton and Gifford (1976, Study 1). Jay Jackson (2001) provides a succinct explanation of the significance of this research:

"Research has demonstrated that people pay special attention to distinctive or rare events and recall such events with relative ease (Hamilton & Gifford, 1976; Hamilton & Sherman, 1996) .. When two distinctive or rare events co-occur, people tend to notice that relation more readily, encode the information more effectively, and recall such relations with greater ease.(Matthews, 1996). Consequently, people often subjectively overestimate how often two distinctive events occur together. This tendency has been dubbed the illusory correlation because it involves perceiving a relation that does not exist or is weaker in reality than perceived (Garcia-Marques & Hamilton, 1996). This cognitive process can influence the content of stereotypes. For example, it partly explains why many White Americans overestimate the rate at which African Americans engage in criminal activity (both are distinct or rare events). p. 273-274."

Design

The current experiment is faithful to the original (Hamilton & Gifford, 1976) in its primary details. Participants read about people who belong to either a majority or minority group, with twice the number of majority members (n = 26) as minority members (n = 13). The information conveyed about each of the people describes a desirable or an undesirable behavior. The ratio of desirable to undesirable behaviors is 9:4 in both groups, but because both minority members and undesirable behaviors are rare events, research participants should form illusory correlations and judge the minority members less favorably than majority group members.

Method

The stimulus materials are 39 facial caricatures of male and female adults differing in age and race. The faces were selected from a clip art collection of prominent people, though the ones used in this 39-person set are highly unlikely to be recognizable to participants. Sample faces are shown in Figure 1.

Sample image from the FirstImpressions experiment Sample image from the FirstImpressions experiment Sample image from the FirstImpressions experiment
Figure 1

The task for participants is to learn something about the people whose faces make up the task stimuli. They do so by clicking on each face in the 39-face set. When a face is clicked, a text balloon pops up with information about the person. The 39 statements are given in Appendix A. The information designates the person as either an Alpha Delta or a Beta Omicron and describes a positive or a negative act or attribute of the person that. The pairing of a face with information about the face is random with the constraint that there are 26 Alpha Deltas to 13 Beta Omicrons and that there are 18 positive and 8 negative acts by Alpha Deltas and 9 positive and 4 negative acts by Beta Omicrons. The information produced by a mouse click remains on the screen until the mouse is released. This permits people to read and process the information at their own pace. Once the mouse in released, the face is greyed out, and the information about the face cannot be re-displayed.

Once all 39 faces are clicked, a set of 7-pt Likert scales appear on the screen. Participants indicate their impressions of Alpha Deltas and Beta Omicrons using these rating scales. The ratings are made for the three positive attributes (popular, honest, and helpful) and three negative attributes (lazy, unhappy, and irresponsible). Scores from these separate scales are available in the data output along with an composite measure of the ratings on positive attributes and negative attributes. A final dependent measure is a rating of the proportion of Alpha Deltas and Beta Omicrons who revealed negative things about themselves.

Data Analyses

Sample data appears below:


Sample data image from the Ponzo experiment
Figure 2

Positive and negative attribute ratings (on a 7-pt Likert scale) for Alpha Deltas and Beta Omicrons are calculated as composite scores and as six separate ratings (i.e., popular, honest, helpful, lazy, unhappy, and irresponsible) for each. A repeated measures t-test is appropriate to test whether the composite means differ from one another. There is an additional computation of the participant's estimation of the proportion of negative statements about Alpha Deltas and Beta Omicrons. Again, a dependent or repeated measures (i.e., correlated groups) t-test will reveal if the different is significant.

The second measures of interest are the estimates particpants made for the percent of negative statements that were made about Alphas and Betas. These are labeled "ApercentofNegative" and "BPercentofNegative" in the data. The actual percent of negative statements is 31% for both groups (26/8 for Alphas and 13/4 for Betas).

Students wishing to perform inferential analyses on the data have several options.

  • A one-way repeated measures anova followed by planned comparisons of Alphas and Betas on each of the six attributes will reveal whether there were differences between the ratings of Alphas and Betas on the six attributes. If Alphas are found to be rated more highly on positive atributes and Betas more highly on negative attributes, the results are consistent with the hypothesis that illusory correlation contributes to stereotype formation.
  • Dependent measures t-tests can be used to compare, first, the composite positive rating for Alphas (ASumofPositive) to the composite positive rating for Betas (BSumofPositive) and, second, the composite negative rating for Alphas (ASumofNegative) to the composite positive rating for Betas (BSumofNegative). If Alphas are found to have higher composite positive ratings and Betas are found to have higher composite negative ratings, the results are consistent with the hypothesis that illusory correlation contributes to stereotype formation.
  • One-sample t-tests can be used to judge whether, first, the estimated percent (ApercentofNegative) differed from the target value of 31% for Alphas and, second, whether the estimated percent (BPercentofNegative) differed from the target value of 31%. The result of these tests will indicate whether participants tend to overestimate the proportion of negative statements due, perhaps, to greater salience for negative than positive statements.
  • A dependent measures t-test can be used to compare the percent estimates (APercentofNegative and BPercentofNegative) made for Alphas and Betas. A significant difference reflecting higher estimates for Betas (the minority group) is consistent with the illusory correlation phenomenon that is hypothesized to contribute to stereotype formation.

Applications/Extensions

This study allows participants to experience the formation of initial impressions and the roles minority/majority status and positive/negative attributes play in that formation process. Further comparisons can be made to the Implicit Association Test study also available here in the Online Psychology Laboratory.

References


Garcia-Marques, L., & Hamilton D.L. (1996). Resolving the apparent discrepancy between 
	the congruency effect and the expectancy-based illusory correlation effect. 
	The TRAP model. Journal of Personality and Social Psychology, 71, 845-860.
					
Hamilton, D.L., & Gifford, R.K. (1976). Illusory correlation in interpersonal perception: 
	A cognitive basis of stereotypic judgments. Journal of Experimental Social Psychology, 
	12, 392-407.
					
Hamilton, D.L., & Sherman, S.J. (1996). Perceiving persons and groups.
	Psychological Review, 103,  336-355.			
					
Jackson, J.W. (2000). Demonstrating the concept of illusory correlation. 
	Teaching of Psychology, 27, 273-276.
					
Matthews, R.A. (1996). Base-rate errors and rain forecasts. Nature, 382, 766.
					

Appendix A

Statements by Alpha Deltas

		I volunteer with Meals on Wheels. 	
		I visited a sick friend in the hospital today. 	
		I am rarely late for work. 	
		I planted seedlings in the park
		I helped a lost child in the supermarket yesterday. 	
		I donated clothes to charity. 	
		I volunteered to tutor needy students.
		I drive my elderly neighbor to the grocery store.
		I work out to keep in shape.
		I received a promotion at work.
		I sing in the church choir.
		I took a hurt stray dog to the vet.
		I helped a friend move to a new apartment.
		I take the neighborhood kids swimming.
		I volunteer in an after-school reading program.
		I give blood to the Red Cross regularly.
		I earned a community service award from the Chamber of Commerce.
		I used vacation time to work with Habitat for Humanity.
		I cheated on last year's federal taxes.
		I am self absorbed to a fault.
		I kicked a dog for no good reason.
		I ran a red light.
		I failed to show up for a scheduled job interview.
		I have crude table manners.
		I stole the neighbor's newspaper one morning.
		I rarely wash my car.
					

Statements by Beta Omicrons

		I helped a needy child.
		I went out of my way to return a lost wallet to its owner.
		I am well liked by my colleagues.
		I converse easily with strangers.
		I learned how to fly an airplane.
		I am recognized as an excellent musician.
		I helped paint my neighbor's house.
		I organized a birthday party for a co-worker.
		I am considered to be a very dependable person.
		I dented the fender of a parked car and did not leave my name.
		I never return library books on time.
		I yelled at a little boy who was bothering me.
		I often tailgate when driving.