Data and code files are separated into four folders based on which ground truth dataset was used to compute the actual transitions between actions: - ATUS transitions: American Time Use Survey used to support Study 2 (tier 1 labels) and Study 3 (tier 2 labels). - AVA: Google's Atomic Video Actions dataset, supporting Study 5 (aka 'intended) and Study 5' - Textual transitions: movie scripts from IMSDb supporting Study 1 - wikihow: instructions from WikiHow.com supporting Study 4 Within each folder, there are parallel sets of files: - .qsf files contain Qualtrics surveys used to elicit participants ratings of a) the transition probabilities between actions, and b) the ratings of those actions on the dimensions of the ACT-FASTaxomony. The latter were not collected for Study 4, as that study relied on previously rated actions from the ACT-FAST validation study (see linked repository). - .R script to measure the actual transition probabilities between actions using the corresponding ground truth data. * atus.R * ava_ground_truth.R * transition_reliability.R (also transition_decay_multiscripts.py) * wikihow_measure_transitions.R (supplemented by wikihow_select_transitions.R and wikihow_scrap.R) - .R script to analyze the association between ground truth and rated transitions (and test mediation by ACT-FAST). * atus_accuracy_tier(1/2).R * ava_accuracy(_intended).R - the intended version corresponds to Study 5 and the other file corresponds to Study 5' * transition_accuracy_verbs.R * wikihow_analyze_behavior.R Raw ground truth and rating data files within each folder are identified within the corresponding .R scripts. Remaining files are primarily stimuli files or intermediate data files saved to avoid recomputing computationally expensive steps.