The Affect Misattribution Procedure (AMP) according to Payne et al. (2005) presents the participant with a series of images (multiple) that are assigned to at least two categories. Categories are e. g. persons, ethnic groups, consumer products, etc. One of the categories can also be a neutral reference category (e. g. a grey image). After the image, an icon is displayed, usually a Chinese character, which European respondents perceive as neutral. This character (and explicitly not the image presented before) should then be judged positive or negative. According to the logic of the measurement, the effect associated with the image is transferred to the character.

The individual measured value is the percentage of positive evaluations that the images of a category received (or the symbols shown afterwards).

Important: The AMP is not part of SoSci Survey's standard scope. The test must be booked separately and with costs as a module “Implicit methods”.

The AMP is available in SoSci Survey via the additional module "Implicit Methods". If the add-on module has been activated for a user account, the AMP is simply created as a question in the Question Catalog:

• Select a Category or create a new category.
• Create a new question of the type “Affect Misattribution Procedure (AMP)” under Add Question.
• Specify in the question which categories you want to measure implicit settings for (more about the question settings below).
• Put the new question in the questionnaire under Compose questionnaire – preferably on a separate page in the questionnaire. As soon as the participant reaches the page, the test procedure is carried out and the determined implicit setting is saved in the data set.

• Note: The AMP uses the exercises (trials) to optimize the display times of the primes and targets (technical_backgrounds). Therefore the test seems to “hang” after the first trial.
• Note: The images (primes and/or targets) should be uploaded in square pixel format and maximally the pixel size in which they are to be displayed later (default: 512 × 512). Smaller and larger images can be scaled to the desired size during the test (Primes), but larger images lead to larger files and thus unnecessarily long loading times.

The “Primes (Stimuli)” tab determines for which categories implicit settings are to be measured. To do this, first define the number of categories (default: 2). If you have changed the number of categories, save the question (Save question) to update the form for the settings. The lower half of the tab shows input fields for each category (each Name and Primes).

Enter a name for each category. In the LMP, categories are represented by images – the category “alcohol” would be operationalized, for example, by different images of (identifiable) alcoholic beverages. By default, the question type uses square images with a size of 512 × 512 pixels. This size represents primes sufficiently large and at the same time with reasonable loading times. However, smaller or larger images are also possible – and with a setting at scale width of the primes all images of the test are scaled to the same pixel width. Since the primes are only displayed for a very short time, a smaller pixel resolution or a higher compression (which results in smaller files) can be useful.

Upload the images under Images and Media Files (use the Protected File Storage only if necessary, it can in rare cases cause primes not to be fully displayed for all participants). Then enter the filenames in the question for the respective category under Primes, one filename per line, e. g.

alhohol01. jpg
alhohol02. jpg
alhohol03. jpg
alhohol04. jpg

Important: When specifying the file names, pay attention to upper and lower case (even in the file name extension; .jpg is slightly different from .JPG). If you selected the protected file location when uploading images, don't forget the leading pro: // before the actual filename.

If you specify fewer primes than repetitions (Trials per block) divided by the number of categories are set (nothing is set, SoSci Survey determines how many primes the “largest” category has and shows this number of primes for each category), then single or all primes are displayed more than once. SoSci Survey randomly mixes the order of the primes, but avoids the repeated display of the same primes.

Also in the first tab Primes (Stimuli) you can set the exact presentation times and whether a fixation (e. g. a cross) should be displayed between the primes.

The AMP can be edited with the keyboard as well as with the mouse or on a touch device (e. g. tablet). In the Answer Options tab you can specify which input methods are allowed on the computer. On mobile devices (smartphone, tablet) the operation is activated automatically by touch.

In addition to the actual primes, the AMP also uses “ambiguous signs”, which respondents are asked to judge according to instructions, and a diffuse noise. Traditionally, Chinese characters are used as ambiguous characters, but in the tab characters (targets) you can also define other graphics as ambigue characters and for the diffuse noise. The filenames are also entered for ambiguous characters (as for the categories).

The instructions before and between the trials can be adjusted in the Texts tab. Here you can use wiki notation and HTML code (Text Elements -> Text formatting).

Presentation times of approx. 20 milliseconds are no longer a problem for current computers and browsers. However, the browsers require considerably more time to display an image on the screen for the first time (lase time and rendering).

SoSci Survey preloads the images into the browser cache and uses the exercise trials as an opportunity to briefly display all primes on the screen in order to minimize later rendering times, so that the display times deviate by less than 20 milliseconds from the desired time. That's why the primes flash up briefly in the second exercise trial.

The AMP-question delivers in the data record the test results (proportion of positive evaluations) and response times for all queried categories.

In addition, raw data (RAW) are available for further analysis. This is a JSON encoded array with one element per trial. Each element is again an array with the following information:

• [0] ID/Filename of the Prime
• [1] ID/Filename of the target (characters)
• [2] Category of the Prime
• [3] Selection of respondent (1=negative, 2=positive)
• [4] Reaction time [ms]

Payne, B. K., Cheng, C. M., Govorun, O. & Stewart, B. D. (2005). An Inkblot for Attitudes: Affect Misattribution as Implicit Measurement. Journal of Personality and Social Psychology, 89(3). doi: 10.1037/0022-3514.89.3.277