Implicit Association Test (IAT)

1. Abstract

The Implicit Association Test (IAT) is a computerized behavioral paradigm designed to quantify individual differences in implicit social cognition, specifically automatic evaluative associations (attitudes) and conceptual pairings (stereotypes). Originally developed by Anthony G. Greenwald, Debbie E. McGhee, and Jordan L. K. Schwartz in 1998, the instrument was conceived to circumvent the inherent psychometric vulnerabilities of explicit self-report measures, including introspective limits, deliberate response distortion, demand characteristics, and social desirability bias. The standard IAT paradigm measures differential response latencies (reaction times in milliseconds) across alternating categorization blocks wherein two contrasted target concepts (e.g., European American vs. African American; Male vs. Female) are paired with valence-laden attribute categories (e.g., Pleasant vs. Unpleasant; Science vs. Arts). The instrument typically consists of a seven-block sequence comprising both uncombined practice discriminations and dual-categorization combined critical tasks (compatible versus incompatible response mappings).

Scoring of the instrument relies on the improved D-score metric established by Greenwald, Nosek, and Banaji (2003), which divides the raw difference between mean latencies of counterbalanced incompatible and compatible blocks by a pooled standard deviation, thereby correcting for participant-level cognitive processing speed. Psychometrically, the IAT yields split-half internal consistency coefficients between .70 and .90, which is substantially higher than alternative latency-based implicit paradigms such as semantic or evaluative priming. Test-retest reliability estimates range moderately between r = .50 and .65, reflecting both stable latent individual differences and sensitivity to state-dependent contextual fluctuations. Extensive construct, convergent, discriminant, and criterion-related validity studies have demonstrated that the IAT significantly predicts nonverbal behaviors, micro-behaviors, policy preferences, and institutional decisions in domains where social norms constrain explicit candor, while also generating vigorous academic debates concerning the influence of extrapersonal cultural knowledge, stimulus familiarity, cognitive control, and structural task-switching variance.

2. Keywords

Implicit Association Test, implicit cognition, response latency, automatic evaluation, social desirability bias, D-score algorithm, implicit bias, cognitive categorization, dual-process theory, evaluative priming, reaction time, psychometrics

3. Authors

The Implicit Association Test was conceptualized and introduced by:

• Anthony G. Greenwald, Ph.D. — Department of Psychology, University of Washington, Seattle, Washington, United States.
• Debbie E. McGhee, Ph.D. — Department of Psychology, University of Washington, Seattle, Washington, United States.
• Jordan L. K. Schwartz, B.A. — Department of Psychology, University of Washington, Seattle, Washington, United States.

Subsequent psychometric optimization, theoretical expansion, large-scale population implementation, and the development of the standardized D-score algorithm were conducted in collaboration with:

• Mahzarin R. Banaji, Ph.D. — Department of Psychology, Harvard University, Cambridge, Massachusetts, United States.
• Brian A. Nosek, Ph.D. — Department of Psychology, University of Virginia, Charlottesville, Virginia, and Center for Open Science, United States.

4. Purpose

The primary purpose of the Implicit Association Test is to capture cognitive representations, automatic affective valuations, and semantic associative networks that operate outside of conscious awareness or beyond deliberate executive control. Traditional assessment within clinical, personality, and social psychology has relied predominantly on self-report questionnaires, structured clinical interviews, and Likert-type attitudinal scales. Although these direct assessment modalities demonstrate high psychometric face validity and utility for conscious introspective self-knowledge, they are constrained by cognitive limitations (such as lack of access to unconscious associative structures) and self-regulatory motives (such as impression management and fear of negative social evaluation). In sensitive domains—including racial prejudice, gender stereotyping, sexual orientation bias, substance craving, suicide risk, and psychiatric symptomatology—explicit self-report scales frequently suffer from floor effects or deliberate censorship.

The IAT addresses these limitations by shifting the measurement model from deliberate linguistic evaluation to behavioral response facilitation. By requiring participants to rapidly map visual or verbal exemplars representing two target concepts and two evaluative attributes onto two lateralized computer keys (e.g., ‘E’ on the left and ‘I’ on the right), the test leverages cognitive interference and associative facilitation. When strongly associated concepts share the same response key (the “congruent” or “compatible” condition), sorting is cognitively efficient, yielding rapid reaction times and low error rates. Conversely, when discordant concepts are forced to share a key (the “incongruent” or “incompatible” condition), response competition, cognitive conflict, and executive inhibition demands increase, resulting in systematically inflated response latencies.

In applied and research contexts, the purpose of the IAT spans several domains:

• Social and Organizational Psychology: Assessing implicit intergroup attitudes, racial disparities, gender-career stereotypes, hiring biases, and institutional climate factors that do not register on explicit diversity surveys.
• Clinical and Psychiatric Research: Measuring implicit self-concept, automatic negative self-schemas in major depressive disorder, implicit associations with death/suicide in psychiatric risk stratification, and implicit approach tendencies in substance use disorders.
• Healthcare Disparities: Identifying provider-level implicit biases toward minority patients, age cohorts, or socioeconomic strata that correlate with clinical decision-making, treatment recommendations, and bedside manner.
• Legal and Judicial Systems: Examining judicial and juror cognition regarding criminal culpability, credibility attribution, and sentencing disparities.

5. Psychological Construct

The central psychological constructs assessed by the IAT are implicit attitudes and implicit stereotypes. In contemporary cognitive psychology, an attitude is broadly defined as an evaluative disposition or psychological tendency expressed by evaluating a particular entity with some degree of favor or disfavor. Unlike explicit attitudes, which represent consciously endorsed, propositional judgments formed through deliberate reasoning, implicit attitudes are conceptualized as automatic, impulsive evaluations rooted in associative networks stored within long-term memory.

Relative Evaluative Associations

A defining characteristic of the construct captured by the IAT is its inherently relative nature. The IAT does not measure an absolute associative link between a single target category and valence; rather, it quantifies the relative associative strength of two contrasted target categories (Category A vs. Category B) in conjunction with two contrasted attribute dimensions (Attribute X vs. Attribute Y). For instance, in the Race IAT, the test assesses the relative ease of pairing European American stimuli with positive valence and African American stimuli with negative valence, contrasted against the reverse mapping. The construct is not merely a manifestation of hostility toward one group, but the relative cognitive affinity or preference for one group over another.

Associative vs. Propositional Architecture

The psychological construct relies on the theoretical distinction between associative processes and propositional reasoning. Associative representations operate according to principles of contiguity, similarity, and frequency of co-occurrence. They are triggered automatically upon presentation of a stimulus exemplar regardless of whether the individual consciously believes or affirms the validity of the association. In contrast, propositional processing involves the validation and truth-value assessment of these activated associations. Thus, the IAT construct reflects the raw associative fabric encoded through chronic environmental exposure, cultural immersion, linguistic conditioning, and personal history, which may diverge from a person’s explicitly stated, genuinely held egalitarian beliefs.

Personal Endorsement versus Cultural Knowledge

A major theoretical debate regarding the IAT construct focuses on whether the test captures personal attitudes or passive cultural awareness. Critics suggest the IAT reflects extrapersonal associations—cultural knowledge, media tropes, and societal stereotypes that an individual has absorbed but rejects at a personal level. Proponents argue that automatic associations fundamentally shape spontaneous perception, nonverbal behavior, and split-second judgments, making the distinction between “cultural absorption” and “personal automatic association” psychologically porous. The construct measured by the IAT is therefore best characterized as an automatic associative predisposition that exerts subtle, pervasive influences on human cognition.

6. Theoretical Framework

The theoretical architecture supporting the Implicit Association Test draws upon dual-process theories of social cognition, associative network models of semantic memory, and neurocognitive frameworks of executive functioning and response interference.

Dual-Process Models

The conceptual foundation of the IAT aligns with prominent dual-process frameworks, such as the Reflective-Impulsive Model (RIM) proposed by Strack and Deutsch (2004) and the Associative-Propositional Evaluation (APE) model articulated by Gawronski and Bodenhausen (2006). The APE model posits that social evaluations are mediated by two interacting cognitive systems:

• The Associative System: Characterized by fast, automatic, heuristic activation of mental links based on spatio-temporal contiguity and pattern completion. It operates with minimal cognitive capacity and is insensitive to truth-values.
• The Propositional System: Characterized by slow, deliberative, rule-governed validation processes that assign truth-values to activated information, generating explicitly stated beliefs and conscious behavioral intentions.

Under this theoretical framework, the IAT taps the initial associative output before propositional verification can censor, correct, or rationalize the cognitive response.

Associative Network and Spreading Activation Theory

The IAT relies directly on Collins and Loftus’s (1975) spreading activation theory of semantic memory. Within long-term memory, concepts are represented as nodes connected by associative pathways of varying strength. When a stimulus exemplar (e.g., an African American face or name) is perceived, activation spreads automatically across adjacent nodes. If the target concept is strongly linked to positive semantic nodes, activation spreads rapidly toward positive attribute nodes. In the IAT, mapping two mutually activating concepts onto a single motor response channel creates cognitive synergy. Conversely, mapping conceptually discordant nodes (e.g., an African American exemplar and a positive attribute) onto divergent motor programs introduces cognitive interference, requiring executive suppression via the prefrontal cortex and anterior cingulate cortex, which manifests as prolonged response latencies.

Response Conflict and the Stroop Analogy

At a functional level, the IAT operates as a conceptual analog to the classic Stroop color-word task. In the Stroop task, naming the ink color of an incongruent word (e.g., the word “RED” printed in blue ink) elicits cognitive conflict between automatic word-reading and controlled color-naming. In the IAT, when target and attribute categories sharing a response key are conceptually incongruent according to the respondent’s underlying associative architecture, the participant experiences response conflict. Resolving this interference requires cognitive control to inhibit the competing motor impulse and activate the correct behavioral response.

7. Validity

The psychometric validity of the IAT has been subjected to extensive empirical investigation, replication, and critical debate across social, cognitive, and clinical psychology.

Construct and Convergent Validity

Construct validity has been supported by demonstrations that the IAT converges with other latency-based and indirect measures of implicit social cognition. Meta-analyses and empirical studies report moderate, statistically significant correlations between the IAT and the Affect Misattribution Procedure (AMP), sequential evaluative priming paradigms, and physiological measures such as startle-blink electromyography (EMG) and functional magnetic resonance imaging (fMRI) amygdala activation during brief stimulus exposures. Greenwald and Farnham (2000) demonstrated convergent validity for self-esteem IATs against multi-method indirect indicators, establishing that the underlying construct captures variance independent of standard questionnaire responses.

Discriminant Validity

Discriminant validity has been evaluated primarily against explicit self-report measures. Across hundreds of independent studies, the correlation between IAT scores and explicit scales (e.g., modern racism scales, explicit preference thermometers) varies widely depending on the domain. In non-sensitive domains (e.g., consumer brand preferences, political candidate choices in two-party systems), correlations between implicit and explicit measures are moderate to high (r = .40 to .70), demonstrating that when social desirability concerns are absent, implicit associations correspond closely with explicit beliefs. In socially sensitive domains (e.g., racial prejudice, sexual attitudes), explicit-implicit correlations drop significantly (r = .10 to .25), demonstrating clear discriminant divergence from direct questionnaires influenced by impression management (Hofmann et al., 2005).

Predictive and Criterion-Related Validity

The predictive utility of the IAT has been examined across numerous behavioral criteria. In a landmark meta-analysis, Greenwald, Poehlman, Uhlmann, and Banaji (2009) evaluated 122 research reports (184 independent samples, N = 14,900) and found that IAT measures predicted criterion outcomes across multiple domains with an average correlation of r = .274. In sensitive intergroup domains involving racial and ethnic attitudes, the IAT significantly out-predicted explicit self-report measures in accounting for spontaneous nonverbal behaviors, such as eye contact, micro-facial expressions, interpersonal physical distance, and body orientation (McConnell & Leibold, 2001).

Subsequent meta-analyses (e.g., Oswald et al., 2013) presented critical evaluations, suggesting that the correlation between racial IAT scores and interpersonal discriminatory behavior was weaker (r ≈ .14 to .15) when restrictive criterion sets were used. However, a comprehensive meta-analysis by Kurdi et al. (2019), synthesizing over 200 studies, confirmed that the IAT demonstrated robust predictive validity (average weighted r ≈ .20) for behavioral outcomes when stimulus-criterion correspondence was properly maintained and when evaluated against complex behavioral aggregates rather than single-incident micro-behaviors.

8. Reliability

Reaction-time paradigms traditionally suffer from lower psychometric reliability compared to multi-item paper-and-pencil inventories due to the high proportion of trial-by-trial trial variance, attentional fluctuations, and executive task-switching noise. However, the IAT is widely recognized as having the highest psychometric reliability among latency-based implicit measurement procedures.

Internal Consistency

The internal consistency of the IAT is typically calculated using split-half procedures (e.g., correlating odd- and even-numbered trials across combined critical blocks) or by dividing the test into consecutive sub-blocks, applying the scoring algorithm to each half, and adjusting with the Spearman-Brown prophecy formula. Across standard 7-block paradigms, the internal consistency of the IAT generally ranges between:

• Split-Half / Permutation Reliability: r = .70 to .85.
• Cronbach’s Alpha equivalents (across block sub-aggregates): α = .75 to .88.

These values contrast favorably with standard evaluative priming tasks, which routinely yield internal consistency estimates between .10 and .40 (Nosek et al., 2007).

Test-Retest Reliability

The temporal stability of the IAT is moderate, reflecting its sensitivity to both stable individual trait differences and dynamic, state-dependent contextual influences. Longitudinal and short-term test-retest evaluations over intervals ranging from hours to several months report stability coefficients between:

• Test-Retest Stability (r): .50 to .65.

While this stability is lower than that of conventional explicit personality inventories (which often exceed .80), it is consistent with the theoretical premise that automatic associations are malleable and continuously modulated by recent environmental exposure, situational primes, physical fatigue, and shifting affective states.

Impact of the Improved Scoring Algorithm

The psychometric reliability of the IAT was substantially elevated by the introduction of the improved D-score algorithm (Greenwald, Nosek, & Banaji, 2003). Prior to 2003, researchers utilized conventional subtraction algorithms (the raw difference between mean latencies, often referred to as the C-score), which were distorted by baseline cognitive speed (an individual’s general reaction time), extreme latencies, and high error rates. The D-score algorithm incorporates trial-level variance standardization, internal error penalties, and practice block integration, which collectively increased internal consistency by .10 to .15 points and markedly stabilized test-retest coefficients.

9. Factor Analysis

Evaluating the factor structure of latency-based response tasks differs fundamentally from conventional item-level factor analysis used for Likert questionnaires. Researchers model individual trial latencies, difference parcel scores, or latent cognitive components using Confirmatory Factor Analysis (CFA) and Structural Equation Modeling (SEM).

Structural Equation Modeling of Latent Dimensions

Early structural investigations by Cunningham, Preacher, and Banaji (2001) addressed the psychometric attenuation of implicit measures by modeling latent variable constructs. By partitioning IAT variance into multiple trial parcels across test blocks, they demonstrated that a single-factor latent model of implicit evaluation adequately accounted for the observed latency covariances. Furthermore, when measurement error was controlled within structural models, the latent stability of implicit attitudes over time rose from observed bivariate correlations of .55 to latent structural path coefficients exceeding .75.

Multi-Trait Multi-Method (MTMM) Matrix Analysis

In comprehensive CFA studies assessing whether the IAT reflects construct variance rather than method-specific artifact, researchers have applied Multi-Trait Multi-Method designs (e.g., Schmukle & Egloff, 2004). These structural models demonstrate that:

• Variance in the IAT is partitioned into substantial trait-specific variance (reflecting the target attitude or stereotype).
• A modest but statistically significant proportion of variance is attributable to a method factor related to individual differences in task-switching capability, executive cognitive control, and generalized processing speed.
• When method variance is explicitly modeled as a latent orthogonal factor, model fit indices improve markedly (CFI > .95, RMSEA < .05), and the purified trait factors exhibit improved criterion relationships.

Mathematical Process Dissociation: The Quad Model

Beyond traditional linear factor analysis, psychometricians have advanced structural understanding of the IAT using Multinomial Processing Tree (MPT) modeling, most notably the Quadruple Process (Quad) Model developed by Conrey, Sherman, Gawronski, Hugenberg, and Groom (2005). The Quad model structures IAT performance into four distinct, conditionally interacting latent cognitive processes:

• Association Activation (AC): The automatic activation of the underlying mental association.
• Detection (D): The conscious capacity to detect the objectively correct response required by the task rule.
• Overcoming Bias (OB): The executive self-regulatory control capacity required to suppress the automatically activated association when it conflicts with the detected correct response.
• Guessing (G): The lateralized motor or heuristic guessing bias when neither association activation nor conscious detection guides the response.

Confirmatory validation of the Quad model reveals that individual differences on the standard IAT reflect both automatic associative activation (AC) and cognitive control resources (OB and D), explaining why structural models that assume purely passive associative readouts without control parameters underperform empirically.

10. Instrument / Measurement Tool

The standard Implicit Association Test is administered via dedicated psychological testing software (e.g., Inquisit by Millisecond Software, PsychoPy, OpenSesame, or web-based platforms such as Project Implicit). The task parameters, procedural sequence, and computational scoring algorithm are structured as follows:

Administration Parameters

• Paradigm Type: Computerized, response-latency-based, two-alternative forced-choice categorization task.
• Hardware/Input: Standard QWERTY keyboard utilizing lateralized keys (typically ‘E’ for the left response and ‘I’ for the right response) or tactile touchscreens in mobile adaptations.
• Visual Feedback: A correct sorting response immediately advances to the next stimulus. An incorrect response typically produces a red ‘X’ centered beneath the stimulus, requiring the participant to press the alternative key to correct the error before the trial advances (in built-in error penalty procedures).
• Inter-Trial Interval (ITI): Standardized across administrations, typically set at 250 milliseconds following a correct or corrected response.

Standard Seven-Block Paradigm Structure

• Block 1 (Target Concept Discrimination – 20 trials): Target category labels appear in the upper-left (e.g., “European American”) and upper-right (e.g., “African American”) corners. Participants categorize target exemplars using ‘E’ or ‘I’.
• Block 2 (Attribute Discrimination – 20 trials): Attribute category labels appear in the upper-left (e.g., “Pleasant”) and upper-right (e.g., “Unpleasant”) corners. Participants categorize positive and negative valence words.
• Block 3 (Combined Task Practice – 20 trials): Target and attribute categories are paired simultaneously on the left and right (e.g., Left: “European American or Pleasant”; Right: “African American or Unpleasant”). Exemplars from all four categories appear in randomized alternation.
• Block 4 (Combined Task Critical Test – 40 trials): Identical category configuration to Block 3, providing the primary critical latency data for the initial pairing condition.
• Block 5 (Reversed Target Discrimination – 20 to 40 trials): Target category positions are swapped (e.g., Left: “African American”; Right: “European American”) to unlearn the prior motor mapping.
• Block 6 (Reversed Combined Task Practice – 20 trials): Target categories remain reversed while attributes retain original positions (e.g., Left: “African American or Pleasant”; Right: “European American or Unpleasant”).
• Block 7 (Reversed Combined Task Critical Test – 40 trials): Identical category configuration to Block 6, providing the comparative critical latency data for the reverse pairing condition.

The Standard Improved D-Score Scoring Algorithm

Scoring of the standard 7-block IAT follows the metric formulated by Greenwald, Nosek, and Banaji (2003):

1. Data Inclusions: Latency data from Blocks 3, 4, 6, and 7 are retained.
2. Extreme Latency Culling: Trials with response latencies > 10,000 ms are eliminated. Respondents for whom > 10% of trials have latencies < 300 ms are entirely excluded due to careless responding.
3. Error Handling: If the software requires participants to correct errors before proceeding, the recorded latency to the final correct keystroke is utilized naturally. If the software does not require correction, each error trial is replaced by the block mean of correct responses plus an empirical penalty (typically 600 ms).
4. Standard Deviation Computation: Calculate one pooled standard deviation for all trials in Blocks 3 and 6, and one pooled standard deviation for all trials in Blocks 4 and 7.
5. Mean Latency Differences: Compute the mean response latency for each block independently. Calculate the practice difference (Mean Block 6 − Mean Block 3) and the test difference (Mean Block 7 − Mean Block 4).
6. Division by Pooled Standard Deviation: Divide the practice difference by the pooled standard deviation of Blocks 3 and 6. Divide the test difference by the pooled standard deviation of Blocks 4 and 7.
7. Final D Calculation: Average the two resulting quotients:
   
   D = [(Mean_B6 − Mean_B3) / SD_pooled(B3,B6) + (Mean_B7 − Mean_B4) / SD_pooled(B4,B7)] / 2

11. Permissions & Fee and Test Year

• Year of Initial Publication: 1998 (original formulation); 2003 (standardized D-score algorithm).
• Copyright and Accessibility: The underlying conceptual architecture, theoretical paradigm, and computational scoring algorithm of the Implicit Association Test exist in the public domain and are freely accessible for scholarly, educational, and non-commercial scientific research.
• Project Implicit: Non-profit educational and research testing platforms, demonstration scripts, and stimuli libraries are hosted openly by Harvard University, the University of Virginia, and the Center for Open Science via the Project Implicit Research Portal.
• Commercial and Proprietary Implementations: Pre-programmed proprietary script packages for specific computerized testing environments (e.g., Inquisit by Millisecond Software) require software licensing fees for the host runtime engine, though the underlying task scripts can be modified and distributed without royalty payments to the authors.
• Ethical and Practical Use Guidelines: Nosek, Greenwald, and Banaji (2005) established strict ethical guidelines emphasizing that the IAT should not be utilized as an individual diagnostic instrument for institutional decision-making (e.g., hiring, jury selection, or clinical diagnosis) due to the presence of task-level measurement error and state-dependent fluctuations at the single-person level. Its recommended use remains confined to aggregate empirical research and personal self-reflective educational demonstrations.

12. References

Baron, A. S., & Banaji, M. R. (2006). The development of implicit attitudes: Evidence of race evaluations from ages 6 and 10 and adults. Child Development, 77(1), 53–68. https://doi.org/10.1111/j.1467-8624.2006.00856.x

Blair, I. V., Steiner, J. F., & Havranek, E. P. (2012). Unconscious (implicit) bias and health disparities: Where do we go from here? The Permanente Journal, 15(2), 71–78. https://doi.org/10.7812/tpp/11-048

Collins, A. M., & Loftus, E. F. (1975). A spreading-activation theory of semantic processing. Psychological Review, 82(6), 407–428. https://doi.org/10.1037/0033-295X.82.6.407

Conrey, F. R., Sherman, J. W., Gawronski, B., Hugenberg, K., & Groom, C. J. (2005). Separating multiple processes in implicit social cognition: The quad model of implicit task performance. Journal of Personality and Social Psychology, 89(4), 469–487. https://doi.org/10.1037/0022-3514.89.4.469

Cunningham, W. A., Preacher, K. J., & Banaji, M. R. (2001). Implicit attitude measures: Consistency, stability, and convergent validity. Psychological Science, 12(2), 163–170. https://doi.org/10.1111/1467-9280.00328

Dasgupta, N., McGhee, D. E., Greenwald, A. G., & Banaji, M. R. (2000). Automatic preference for White Americans: Eliminating the familiarity explanation. Journal of Experimental Social Psychology, 36(3), 316–328. https://doi.org/10.1006/jesp.1999.1418

FitzGerald, C., & Hurst, S. (2017). Implicit bias in healthcare professionals: A systematic review. BMC Medical Ethics, 18(1), Article 19. https://doi.org/10.1186/s12910-017-0179-8

Frantz, C. M., Cuddy, A. J. C., Burnett, M., Ray, H., & Hart, A. (2004). A threat in the computer: The race implicit association test as a stereotype threat experience. Personality and Social Psychology Bulletin, 30(12), 1611–1624. https://doi.org/10.1177/0146167204266650

Gawronski, B. (2002). What does the Implicit Association Test measure? A test of the convergent and discriminant validity of prejudice-related IATs. Experimental Psychology, 49(3), 171–180. https://doi.org/10.1027//1618-3169.49.3.171

Gawronski, B., & Bodenhausen, G. V. (2006). Associative and propositional processes in evaluation: An integrative review of implicit and explicit attitude change. Psychological Bulletin, 132(5), 692–731. https://doi.org/10.1037/0033-2909.132.5.692

Greenwald, A. G., & Farnham, S. D. (2000). Using the Implicit Association Test to measure self-esteem and self-concept. Journal of Personality and Social Psychology, 79(6), 1022–1038. https://doi.org/10.1037/0022-3514.79.6.1022

Greenwald, A. G., McGhee, D. E., & Schwartz, J. L. K. (1998). Measuring individual differences in implicit cognition: The implicit association test. Journal of Personality and Social Psychology, 74(6), 1464–1480. https://doi.org/10.1037/0022-3514.74.6.1464

Greenwald, A. G., Nosek, B. A., & Banaji, M. R. (2003). Understanding and using the Implicit Association Test: I. An improved scoring algorithm. Journal of Personality and Social Psychology, 85(3), 197–216. https://doi.org/10.1037/0022-3514.85.3.197

Greenwald, A. G., Poehlman, T. A., Uhlmann, E. L., & Banaji, M. R. (2009). Understanding and using the Implicit Association Test: III. Meta-analysis of predictive validity. Journal of Personality and Social Psychology, 97(1), 17–41. https://doi.org/10.1037/a0015565

Grumm, M., & von Collani, G. (2007). Measuring Big-Five personality dimensions with the implicit association test–implicit personality traits or self-esteem? European Journal of Personality, 21(6), 793–809. https://doi.org/10.1002/per.645

Hofmann, W., Gawronski, B., Gschwendner, T., Le, H., & Schmitt, M. (2005). A meta-analysis on the correlation between the implicit association test and explicit self-report measures. Personality and Social Psychology Bulletin, 31(10), 1369–1385. https://doi.org/10.1177/0146167205275613

Kurdi, B., Seitchik, A. E., Axt, J. R., Carroll, J. M., Karapetyan, A., Kaushik, N., Trawalter, S., Shariff, A. F., & Banaji, M. R. (2019). Relationship between the Implicit Association Test and intergroup behavior: A meta-analysis. American Psychologist, 74(5), 569–586. https://doi.org/10.1037/amp0000364

McConnell, A. R., & Leibold, J. M. (2001). Relations among the implicit association test, discriminatory behavior, and explicit measures of racial attitudes. Journal of Experimental Social Psychology, 37(5), 435–442. https://doi.org/10.1006/jesp.2000.1470

Nock, M. K., & Banaji, M. R. (2007). Prediction of suicide attempts among adolescents with self-injurious thoughts and behaviors. Behaviour Research and Therapy, 45(12), 3045–3055. https://doi.org/10.1016/j.brat.2007.08.016

Nosek, B. A., Greenwald, A. G., & Banaji, M. R. (2005). Understanding and using the Implicit Association Test: II. Method variables and construct validity. Personality and Social Psychology Bulletin, 31(2), 166–180. https://doi.org/10.1177/0146167204271418

Nosek, B. A., Greenwald, A. G., & Banaji, M. R. (2007). The Implicit Association Test at age 7: A methodological and conceptual review. In J. A. Bargh (Ed.), Social psychology and the unconscious: The automaticity of higher mental processes (pp. 265–292). Psychology Press.

Olson, M. A., & Fazio, R. H. (2004). Reducing the influence of extrapersonal associations on the implicit association test: Personalizing the IAT. Journal of Personality and Social Psychology, 86(5), 653–667. https://doi.org/10.1037/0022-3514.86.5.653

Oswald, F. L., Mitchell, G., Blanton, H., Jaccard, J., & Tetlock, P. E. (2013). Predicting ethnic and racial discrimination: A meta-analysis of the IAT. Journal of Personality and Social Psychology, 105(2), 171–192. https://doi.org/10.1037/a0032734

Rachlinski, J. J., Johnson, S. L., Wistrich, A. J., & Guthrie, C. (2009). Does unconscious racial bias affect trial judges? Notre Dame Law Review, 84(3), 1195–1246.

Rudman, L. A., & Ashmore, R. D. (2007). Discrimination and the Implicit Association Test. Group Processes & Intergroup Relations, 10(3), 359–372. https://doi.org/10.1177/1368430207078696

Schmukle, S. C., & Egloff, B. (2004). Does the Implicit Association Test for assessing anxiety measure trait or state anxiety? European Journal of Personality, 18(6), 483–494. https://doi.org/10.1002/per.525

Siegel, E. F., Dougherty, M. R., & Huber, D. E. (2012). Manipulating the role of cognitive control while taking the implicit association test. Journal of Experimental Social Psychology, 48(5), 1057–1068. https://doi.org/10.1016/j.jesp.2012.03.009

Strack, F., & Deutsch, R. (2004). Reflective and impulsive determinants of social behavior. Personality and Social Psychology Review, 8(3), 220–247. https://doi.org/10.1207/s15327957pspr0803_1

Teachman, B. A., Smith-Janik, S. B., & Saporito, J. (2007). Information processing biases and panic disorder: Relationships among cognitive and symptom measures. Behaviour Research and Therapy, 45(8), 1791–1811. https://doi.org/10.1016/j.brat.2006.09.018

Thush, C., & Wiers, R. W. (2007). Explicit and implicit alcohol-related cognitions and the prediction of alcohol use in young adolescents. Psychology of Addictive Behaviors, 21(1), 41–49. https://doi.org/10.1037/0893-164X.21.1.41

13. Items of the Scale