Author: Bram Van Rensbergen (mail@bramvanrensbergen.com)

Source: https://github.com/BramVanRensbergen/categorization-under-cognitive-load

This is the code to an experiment in which participants learn to categorize abstract stimuli, while simultaneously memorizing a dot pattern. The experiment was created for personal use (ongoing research of a colleague), but anyone may use it if they like.

To use:

• Download the project, compile it using Java Development Kit, have the participant run it on a computer with Java Runtime Environment installed.
• Currently supports Java 7 and 8. Older versions might work, but have not been tested.
• Options can be set in data/options.txt
• Instructions can be set in data/text.txt
• Button labels etc. are in Dutch, but it is quite easy to translate them: all language displayed to the user is defined in Text.java. Comments, code, output headers, ... are all in English.
• (After any changes to a .java file, you will have to recompile.)

###Training Phase Overview:

• Participants learn which stimuli belong to which group
• They see an image, and are asked to tell which of the two groups it belongs to
• They receive feedback on whether they were right or not
• They do not see all stimuli, but a subset of each group
  • 6 of each group, by default
  • Which images are selected for the subset is randomized (per participant)
• After they have categorized both subsets, they are given feedback on their average accuracy on them
• Training continues until participants reach sufficient accuracy on a number of subsequent blocks (2 blocks with over 90% accuracy, by default)
• During this entire phase, participants also memorize and re-produce dot-patterns while they are categorizing stimuli

Flow:

• Participants see a dot pattern for 750ms
• Participants see an image, and indicate which group it belongs to
• Participants receive feedback on their choice (right/wrong)
• Participants see three more images, one by one, categorizing each
• Participants are presented with an empty grid and are asked to reproduce the pattern they saw earlier
• All above steps are repeated until all stimuli have been displayed once
• This means a block is completed, and participants receive info on their mean accuracy during that block
• A new block begins, with the same stimuli, in a newly randomized order
• All above steps continue until sufficient accuracy is obtained for the CATEGORIZATION task

Dot patterns:

• Four dots in a 4x4 grid
• The pattern is either easy (low load) or hard (high load)
• Easy pattern: a row or column of dots
• Hard pattern: a 'complex' arrangement of dots
• Difficulty is counterbalanced between participants, based on subject number
• A hard pattern fulfills three requirements:
  • No two vertically or horizontally adjacent dots, nor three or four dots in either diagonal.
  • Complexity has to be high enough according to the CRC method described in: Ichikawa, S (1983). Verbal memory span, visual memory span, and their correlations with cognitive tasks. Japanese Psychological Research 25(4), 173-180.
  • Mirroring or rotating the pattern cannot create a duplicate
• See these papers, or the three classes in pattern.validity, for more information.

###Test Phase Overview:

• Participants categorize the entire stimulus-set, without feedback, and without any dot patterns
• Entire stimulusset is used:
  • This includes stimuli not seen during training phase by this participant
• Program continues for a set number of blocks, each containing all stimuli once
  • Default: 5 blocks

Flow:

• Participants see an image, and indicate which group it belongs to
• The above continues until all stimuli have been displayed once
• A new block begins, with all images in a newly randomized order
• The above continues until all blocks are completed (default 5)

###Counterbalance and randomisation

• Group labels are counterbalanced between participants
  • Counterbalanced based on subject number
  • For half of participants, data/options.txt/group1name refers to group1, and group2name to group2, for the other half, this is reversed
• Participants categorize using two buttons, one for each group
  • Which button is on the left, and which is on the right, is also counterbalanced between participants, based on subject number
  • This done independently of the grouplabel balancing
• Half of the participants receive difficult dot patterns, the other half receive easy dot patterns
  • This is also counterbalanced, based on subject number
• Which subset of stimuli is displayed during the training phase is randomized per participant
  • In other words, not everyone sees the same stimuli
• Trial order
  • During training phase, blocks contain 2 subsets of 6 trials (by default)
  • During test phase, blocks contain 2 sets of 18 stimuli
  • The order of presentation within each block is fully randomized