An implicit association test (IAT) measures differential association of 2 target concepts with an attribute. The 2 concepts appear in a 2-choice task (e.g., flower vs. insect names), and the attribute in a 2nd task (e.g., pleasant vs. unpleasant words for an evaluation attribute). When instructions oblige highly associated categories (e.g., flower + pleasant) to share a response key, performance is faster than when less associated categories (e.g., insect + pleasant) share a key. This performance difference implicitly measures differential association of the 2 concepts with the attribute. In 3 experiments, the IAT was sensitive to (a) near-universal evaluative differences (e.g., flower vs. insect), (b) expected individual differences in evaluative associations (Japanese + pleasant vs. Korean + pleasant for Japanese vs. Korean subjects), and (c) consciously disavowed evaluative differences (Black + pleasant vs. White + pleasant for self-described unprejudiced White subjects).Consider a thought experiment. You are shown a series of male and female faces, to which you are to respond as rapidly as possible by saying "hello" if the face is male and "goodbye" if it is female. For a second task, you are shown a series of male and female names, to which you are to respond rapidly with ' 'hello" for male names and "goodbye" for female names. These discriminations are both designed to be easy--the faces and names are unambiguously male or female. For a final task you are asked to perform both of these discriminations alternately. That is, you are shown a series of alternating faces and names, and you are to say "hello" if the face or name is male and "goodbye" if the face or name is female. If you guess that this combined task will be easy, you are correct. Now imagine a small variation of the thought experiment. The first discrimination is the same ("hello" to male faces, "goodbye" to female faces), but the second is reversed ( "goodbye" to male names, "hello" to female names). As with the first experiment, each of these tasks, by itself, is easy. However, when you contemplate mixing the two tasks ("hello" to male face or female name and "goodbye" to female face or male name), you may suspect that this new combined task will be difficult. Unless you wish to make many errors, you will have to respond considerably more slowly than in the previous experiment.The expected difficulty of the experiment with the reversed Anthony G. Greenwald, Debbie E. McGhee, and Jordan L.K. Schwartz, Department of Psychology, University of Washington.This research was partially supported by Grant SBR-9422242 from the National Science Foundation and Grant MH 41328 from the National Institute of Mental Health. For comments on a draft of this article, the authors thank Mahzarin Banaji, Shelly Famham, Laurie Rudman, and Yuichi Shoda.Correspondence concerning this article should be addressed to Anthony G. Greenwald, Department of Psychology, Box 351525, University of Washington, Seattle, Washington 98195-1525. Electronic mail may be sent to agg@u....

In reporting Implicit Association Test (IAT) results, researchers have most often used scoring conventions described in the first publication of the IAT (A. G. Greenwald, D. E. McGhee, & J. L. K. Schwartz, 1998). Demonstration IATs available on the Internet have produced large data sets that were used in the current article to evaluate alternative scoring procedures. Candidate new algorithms were examined in terms of their (a) correlations with parallel self-report measures, (b) resistance to an artifact associated with speed of responding, (c) internal consistency, (d) sensitivity to known influences on IAT measures, and (e) resistance to known procedural influences. The best-performing measure incorporates data from the IAT's practice trials, uses a metric that is calibrated by each respondent's latency variability, and includes a latency penalty for errors. This new algorithm strongly outperforms the earlier (conventional) procedure.

Social behavior is ordinarily treated as being under conscious (if not always thoughtful) control. However, considerable evidence now supports the view that social behavior often operates in an implicit or unconscious fashion. The identifying feature of implicit cognition is that past experience influences judgment in a fashion not introspectively known by the actor. The present conclusionthat attitudes, self-esteem, and stereotypes have important implicit modes of operation-extends both the construct validity and predictive usefulness of these major theoretical constructs of social psychology. Methodologically, this review calls for increased use of indirect measures-which are imperative in studies of implicit cognition. The theorized ordinariness of implicit stereotyping is consistent with recent findings of discrimination by people who explicitly disavow prejudice. The finding that implicit cognitive effects are often reduced by focusing judges' attention on their judgment task provides a basis for evaluating applications (such as affirmative action) aimed at reducing such unintended discrimination.Long before they became central to other areas of psychological theory, concepts of cognitive mediation dominated the analysis of social behavior. The constructs on which this article focuses achieved early prominence in social psychological theory with formulations that were partly (attitude) or entirely (stereotype) cognitive. By the 1930s, Allport (1935) had declared attitude to be social psychology's "most distinctive and indispensable concept" (p. 798), Thurstone (1931;Thurstone & Chave, 1929) had developed quantitatively sophisticated methods for attitude measurement, and Braly (1933, 1935) had introduced a method that is still in use to investigate stereotypes. Self-esteem, an attitudinal construct to which this article gives separate treatment because of its prominence in recent research, also has a long-established history (e.g., James, 1890; see overview in Wylie, 1974Wylie, , 1979.Through much of the period since the 1930s, most social psychologists have assumed that attitudes, and to a lesser extent stereotypes, operate in a conscious mode. This widespread assumption of conscious operation is most evident in the nearuniversal practice of operationalizing attitudes (including selfesteem) and stereotypes with direct (instructed self-report) measures. The pervasiveness of direct measurement for attitudes and stereotypes was documented by Greenwald (1990) and by Banaji and Greenwald (1994) and is further reviewed below. In contrast, this article describes an indirect, unconscious, or implicit mode of operation for attitudes and stereotypes. 1 Anthony G. Greenwald, Department of Psychology, University of Washington; Mahzarin R. Banaji, Department of Psychology, Yale University.Preparation of this report as well as conduct of some of the research reported in it were supported by National Science Foundation Grants DBC-9205890 and DBC-9120987 and by National Institute of Mental Health Grant MH-41328. We thank...

This review of 122 research reports (184 independent samples, 14,900 subjects) found average r = .274 for prediction of behavioral, judgment, and physiological measures by Implicit Association Test (IAT) measures. Parallel explicit (i.e., self-report) measures, available in 156 of these samples (13,068 subjects), also predicted effectively (average r = .361), but with much greater variability of effect size. Predictive validity of self-report was impaired for socially sensitive topics, for which impression management may distort self-report responses. For 32 samples with criterion measures involving Black-White interracial behavior, predictive validity of IAT measures significantly exceeded that of self-report measures. Both IAT and self-report measures displayed incremental validity, with each measure predicting criterion variance beyond that predicted by the other. The more highly IAT and self-report measures were intercorrelated, the greater was the predictive validity of each.

We propose to change the default P-value threshold for statistical significance from 0.05 to 0.005 for claims of new discoveries. T he lack of reproducibility of scientific studies has caused growing concern over the credibility of claims of new discoveries based on 'statistically significant' findings. There has been much progress toward documenting and addressing several causes of this lack of reproducibility (for example, multiple testing, P-hacking, publication bias and under-powered studies). However, we believe that a leading cause of non-reproducibility has not yet been adequately addressed: statistical standards of evidence for claiming new discoveries in many fields of science are simply too low. Associating statistically significant findings with P < 0.05 results in a high rate of false positives even in the absence of other experimental, procedural and reporting problems.For fields where the threshold for defining statistical significance for new discoveries is P < 0.05, we propose a change to P < 0.005. This simple step would immediately improve the reproducibility of scientific research in many fields. Results that would currently be called significant but do not meet the new threshold should instead be called suggestive. While statisticians have known the relative weakness of using P ≈ 0.05 as a threshold for discovery and the proposal to lower it to 0.005 is not new 1,2 , a critical mass of researchers now endorse this change.We restrict our recommendation to claims of discovery of new effects. We do not address the appropriate threshold for confirmatory or contradictory replications of existing claims. We also do not advocate changes to discovery thresholds in fields that have already adopted more stringent standards (for example, genomics and high-energy physics research; see the 'Potential objections' section below).We also restrict our recommendation to studies that conduct null hypothesis significance tests. We have diverse views about how best to improve reproducibility, and many of us believe that other ways of summarizing the data, such as Bayes factors or other posterior summaries based on clearly articulated model assumptions, are preferable to P values. However, changing the P value threshold is simple, aligns with the training undertaken by many researchers, and might quickly achieve broad acceptance.