Implicit Association Test (IAT)

A word or picture in the centre, the category labels in the top left and top right corners, and two large response areas at the bottom. Attribute words and labels are green, target items and labels the normal text colour. A wrong tap shows a red X until the correct tap. A screen before each block shows which categories go left and right.

Useful when a study needs an indirect measure of how strongly two concepts are associated, e.g. alcohol with approach versus avoidance in addiction research, or other targets and attributes entered as words or pictures. Use the standard 7-block format at baseline and the Brief IAT for repeated sessions in daily life.

Not a diagnostic instrument, and not a measure of one person's attitude from one session: a D score from a single short session is unreliable. Not suited to single-category questions (one target without a contrast category).

Choose two target categories (built-in drink pictures, your own image set, or words) and two attribute categories (built-in Approach / Avoid words, translated, or your own words). The built-in drink pictures are a demo default; for research, use a validated, matched picture set. The defaults follow published studies: the standard procedure (Greenwald, Nosek & Banaji, 2003), the content of the Alcohol Approach IAT (Lindgren et al., 2013), whose English Approach / Avoid words are the built-in ones (a list Lindgren et al. attribute to Ostafin & Palfai, 2006; the translations into other languages are not published lists), a 250 ms blank between trials (Greenwald, McGhee & Schwartz, 1998, Experiment 2) and, in the Brief IAT, mammals and birds as warm-up categories (Nosek et al., 2014) with the words of Sriram & Greenwald (2009). Target B is soft drinks, as in Wiers et al. (2011) and Schwippel et al. (2025), so one picture set serves all four picture tasks; English alcohol-approach IATs often contrast alcohol with water instead (Lindgren et al., 2013). Keep the order counterbalanced at random. Plan several sessions per participant and aggregate D scores, or model the trial-level latencies.

• Format: standard 7-block IAT (200 trials) or Brief IAT (96 trials)
• Targets: built-in drink pictures (alcohol vs soft drinks, as in Wiers et al., 2011 and Schwippel et al., 2025), your own image set, or word lists
• Attribute names and word lists (built-in: Approach / Avoid, the Alcohol Approach IAT list of Lindgren et al., 2013, translated)
• Brief IAT: warm-up categories (built-in: mammals / birds, Nosek et al., 2014; words: elephant, bison, deer, cow / eagle, swan, parrot, duck, Sriram & Greenwald, 2009; translated)
• Order: random per participant, compatible first or incompatible first
• Standard IAT: 40 trials in block 5 (default; reduces the order effect), or 30 or 20
• Blank between trials (default 250 ms from the correct response, Greenwald, McGhee & Schwartz, 1998, Experiment 2)

• D score: standard by the improved algorithm with the built-in error penalty; brief as recommended by Nosek et al. (2014)
• D for practice and test blocks (standard) or for blocks 2–3 and 4–5 (brief)
• Mean latency in compatible and incompatible blocks, and the error rate
• Share of responses under 300 ms, flagged above 10% (D is still reported)
• Every trial: block, pairing, category, item, first response, error, latency to the first and to the correct response

Compatible means target A shares a response with attribute A. A positive D means faster responses when A is paired with attribute A, i.e. a stronger A–attribute A association relative to B; a negative D the reverse. The IAT measures relative associations: D reflects both target categories, never one alone. Errors must be corrected, and the latency runs to the correct response, so errors add time instead of being removed. Participants with more than 10% of responses under 300 ms are flagged, counting the latency to the correct response as Greenwald et al. (2003) and Nosek et al. (2014) do; whether to exclude them is your decision. Responses are timed from the drawn stimulus to the touch-down, but this timing has not yet been validated on real devices. A D score from one short session has limited reliability; aggregate across sessions or model the trial-level data.

• The improved scoring algorithm (D) outperformed the conventional IAT scoring on several criteria in large web samples.
• The Brief IAT keeps the comparison of two pairings with fewer trials: in each block, participants attend to two focal categories and sort everything else to the other side.
• Recommended Brief IAT scoring removes the first 4 trials of each block, retains error trials and recodes latencies to 400–2000 ms.
• Robust group-level effects in cognitive tasks can still have low test–retest reliability as individual differences.

• Greenwald, McGhee & Schwartz (1998), the Implicit Association Test
• Greenwald, Nosek & Banaji (2003), an improved scoring algorithm
• Sriram & Greenwald (2009), the Brief Implicit Association Test
• Nosek et al. (2014), recommended scoring procedures for the Brief IAT
• Lindgren et al. (2013), the Alcohol Approach IAT and its Approach / Avoid words
• Ostafin & Palfai (2006), the IAT and automatic alcohol motivation
• Wiers et al. (2011), an alcohol approach IAT with soft drinks as the contrast category
• Schwippel et al. (2025), an alcohol approach IAT with soft-drink pictures
• Hedge, Powell & Sumner (2018), the reliability paradox

• Approach-Avoidance (AAT)
• Stroop
• Go/No-Go