|
| The Sense and Nonsense of Implicit Testing in Behavior Analysis |
| Monday, May 31, 2010 |
| 9:00 AM–10:20 AM |
| Lone Star Ballroom Salon F (Grand Hyatt) |
| Area: EAB/VBC; Domain: Experimental Analysis |
| Chair: Bryan T. Roche (National University of Ireland, Maynooth) |
| CE Instructor: Linda LeBlanc, Ph.D. |
| Abstract: In recent years psychology has witnessed what has been described as an “unstoppable juggernaut” of research interest in a form of “implicit testing” known as the implicit association test (IAT). Researchers claim that the IAT can reveal unconscious processes, such as prejudice, and can serve as an indicator of behavioral probability. If these claims can be substantiated in laboratory research then the IAT represents one of the most useful psychological tools ever developed. However, while the advent of this test represents one of the most talked about developments in psychology’s recent history, little is known about how this test actually functions. Surprisingly, however, the test and functionally similar variants, are now being used by behavior analysts to assess behavioral history and probability in the absence of a satisfactory program of research into the test’s core processes. The current session presents a series of experiments designed to develop and test a behavior-analytic model of the IAT. The session will also illustrate the role of several procedural and data-analytic artifacts that contribute to the IAT test effect. Cautionary advice will be offered to researchers who employ implicit test methods in behavior analytic research. |
| |
| Establishing and Eliminating Implicit Association Test Effects in the Laboratory: Extending a Behavioral Model of the IAT |
| BRYAN T. ROCHE (National University of Ireland, Maynooth), Iseult Ridgeway (National University of Ireland, Maynooth), Amanda Gavin (University of Tesside), Maria R. Ruiz (Rollins College) |
| Abstract: In previous research, Gavin, Roche & Ruiz (2008) demonstrated that implicit association test (IAT) effects can be modeled in the laboratory in the form of a test that establishes competing contingencies for derived relational responding. The current study replicates and extends this finding by firstly establishing a laboratory-controlled IAT effect using nonsense syllables as stimuli, and then eliminating that effect with a simple experimental intervention. Eleven subjects were exposed to an equivalence training procedure that led to the formation of two three-member equivalence relations each containing three nonsense syllables. Subjects were then exposed to a word-picture association training phase in which one member of each of the equivalence relations, printed in blue or red font, was paired with either plant or animal images, respectively. Subjects were then exposed to an IAT whose outcome was successfully controlled by the organization of the trained relations. Following a stimulus equivalence re-training procedure that reorganized the stimulus equivalence relations, the IAT effect was eliminated or reversed for five of the six subjects who showed reorganization of the equivalence class and the associated derived transfer of functions. These findings lend additional support to a behavior-analytic account of the IAT. |
| |
| The Generalization of Implicit Association Test Practice Effects Across Semantic Categories: Testing a Key Prediction of a Behavioral Model of the IAT |
| ANTHONY O'REILLY (National University of Ireland, Maynooth), Claire Bedford (National University of Ireland, Maynooth), Bryan T. Roche (National University of Ireland, Maynooth), Maria R. Ruiz (Rollins College) |
| Abstract: The current paper reports on a study designed to test a key prediction of the behavioral model of the implicit association test (IAT). According to the behavioral model, the IAT measures the relative fluency of relational responding to various related and unrelated stimulus pairs. In effect, the IAT is a measure of relative stimulus class strengths that have been established in the history of the subject. Consequently, learning effects established with a given stimulus set should generalize to related stimulus class members. The current paper reports on a study designed to test this idea. Subjects were exposed to an IAT for racial bias, after which they were provided with extensive practice on that test. Practice resulted in the erosion of the very fluency differences across task-types on which the IAT depends. Subjects were then exposed to another IAT using synonyms for the target verbal stimuli employed in the first test. The results showed that IAT effects were absent due to near perfect fluency levels in relating the novel but related stimuli. Practice effects did not generalize to a novel IAT employing semantically unrelated stimuli. |
| |
| How Corrective is Corrective Feedback in the Implicit Association Test? |
| MARIA R. RUIZ (Rollins College), Micah Purdy (Rollins College), Ingrid Atiles (Rollins College), Anthony O'Reilly (National University of Ireland, Maynooth), Bryan T. Roche (National University of Ireland, Maynooth), James McFarlane (National University of Ireland, Maynooth) |
| Abstract: The implicit association test (IAT) employs a curious corrective feedback procedure in which only erroneous responses are consequated by feedback and a response observation requirement. From a behavior-analytic perspective this procedure is replete with experimental artifacts and has serious implications for the way in which responses are coded. In addition, it is unclear if corrective feedback on the IAT is even reinforcing in the first instance. This paper describes an experiment designed to examine the effects of feedback delivery in the IAT on response fluency. Specifically, a non-contingent response observation requirement without reinforcing properties was developed to mimic the delivery of feedback in the IAT. The requirement to make the non-reinforcing observation response was imposed on one quarter of trials during each of the two main testing blocks of an IAT across two experimental conditions. A third condition involved the delivery of an IAT in the absence of both corrective feedback or response observation requirements. Results indicated that the delivery of corrective feedback in the IAT may have a punishing, rather than a reinforcing effect. The paper will explain how such an effect enhances rather than detracts from reported IAT effect sizes. |
| |
| The Implicit Association Test Measures Relational Responding Fluency, Not Attitudes! Empirical Evidence From the Laboratory |
| INGRID ATILES (Rollins College), Maria R. Ruiz (Rollins College), Bryan T. Roche (National University of Ireland, Maynooth), Micah Purdy (Rollins College) |
| Abstract: A behavioral model of the implicit association test (IAT) suggests that the IAT works by comparing the relative fluencies in responding to related and unrelated pairs of stimuli. The current study tested this idea. Two groups of subjects were exposed to an equivalence training procedure that led to the formation of two three-member equivalence relations each containing three nonsense syllables. One group was trained using a strict fluency criterion that required responses to be produced within 3s on every trial of the stimulus equivalence training and testing phases. Both groups were exposed to a word-picture association training phase in which one member of each of the equivalence relations, printed in blue or red font, was paired with either plant or animal images, respectively. All subjects were subsequently exposed to an Implicit Association Test whose outcome was predicted and controlled by the organization of the trained relations. However, the relative fluencies in relating pairs of related and unrelated stimuli during the IAT were more different for subjects from the "high fluency" condition. Consequently, larger IAT effects were calculated for this group, thereby supporting a key prediction of a behavioral model of the IAT. |
|
| |