Attention Capture and Visual Search Behavioral Data and Stimuli
Description
These are the complete datasets for Experiment 1-3 of the manuscript currently titled "Diagnostic parts are not exclusive in the search template for real-world object categories" by Marcel Wurth and Reshanne R. Reeder. This study contains a direct replication of Experiment 4 from Reeder & Peelen (2013), and two follow-up experiments. In each experiment, subjects were cued to search for cars and people in natural scenes (75% of trials) either presented to the left and right of a central fixation (Exp.1) or above and below fixation (Exp.2&3). On a subset of trials (25% of trials), subjects were instructed only to respond to the location of a black dot that appeared on the left or right of fixation. Capture stimuli (the silhouettes) appeared prior to the dot, but subjects were told to ignore these to the best of their ability. One car and one person silhouette appeared on each capture trial and appeared an equal number of times on the left and right of fixation. The location of the target-matching silhouette was tied to the same side of fixation as the dot-probe on half of trials (consistent trials). The target-matching silhouette appeared on the opposite side of fixation from the dot-probe on the other half of trials (inconsistent trials). In Exp.1&2, capture stimuli were either both whole silhouettes or both single silhouette parts, with these trials randomly mixed within a block. In Exp.3, capture stimuli were either both whole silhouettes or both collections of silhouette parts. Our main goal was to find out if subjects performed better on consistent dot-probe trials than on inconsistent dot-probe trials. A consistent trial is defined as a trial on which the cue-matching silhouette appears on the same side of fixation as the dot-probe. An inconsistent trial is a trial on which the cue-matching silhouette appears on the other side of fixation from the dot-probe. We calculated RT from the onset of the dot-probe. Only correct trials were input into the RT analysis. Attention capture was defined as faster RT on consistent trials compared to inconsistent trials. Reeder, R. R., & Peelen, M. V. (2013). The contents of the search template for category-level search in natural scenes. Journal of vision, 13(3). doi:10.1167/13.3.13
Files
Steps to reproduce
The final data that were input into SPSS for ANOVAs and t-tests are stored in Results-Excel Files (for SPSS). The raw data for each block of each subject are stored in the Python pickle format in Results-Pickle Files (raw data). Only data from blocks 1-9 were analyzed (block 0 was a practice run). Wurth collected the data for most of the subjects, but Reeder collected the final 5 subjects (subjects 50-54). The pickle files contain dictionaries with information about each trial (cue identity, scene identity, capture stimulus location, capture stimulus identity, correct response, dot probe location, search target location, subject accuracy, subject response, subject RT, target onset). A value of "0" for stimulus identity or location means that type of stimulus did not appear on that trial. An accuracy of "1" means correct, "0" means incorrect. The Excel files contain accuracy and RT data for all conditions, as well as subject age, handedness, and gender. Accuracy outliers (see below) are highlighted in red. Capture RT outliers are highlighted in orange. 90 students and faculty of Otto-von-Guericke University, were recruited for this study (29 for Experiment 1, 30 for Experiment 2, and 31 for Experiment 3). Nine subjects took part in more than one experiment. Outlier exclusion was strict to control for possible confounds in subject performance. Subjects were excluded if their mean visual search accuracy was below 75%. This was to ensure that subjects could activate an appropriate category-level search template. Subjects were also excluded if their mean dot-probe detection accuracy was below 90%. In the dot-probe task, subjects simply had to respond on which side of fixation a dot appeared, and accuracy lower than 90% would suggest a lack of attention or inability to understand the task rather than normal human error. Furthermore, high dot-probe accuracy was necessary for sufficient power to analyze reaction times (RT; only reported for accurate trials) due to a low number of dot-probe trials per run (16). Following these exclusions, subjects were lastly excluded if their mean dot-probe RT was slower than two standard deviations from the group mean. This was done to ensure that subjects did not consciously deliberate about their responses based on the capture stimuli that appeared prior to the dot. Because of the strict exclusion criteria and to ensure an adequate number of subjects per experiment, we continued to run subjects in each experiment until a final number of 25 suitable subjects per experiment was reached. Four subjects were excluded from Experiment 1 (two due to low search accuracy, and one each due to low dot-probe accuracy and slow dot-probe RT); three subjects were excluded from Experiment 2 (one each due to low search accuracy, low dot-probe accuracy, and slow dot-probe RT); and six subjects were excluded from Experiment 3 (three due to low search accuracy, two due to low dot-probe accuracy, and one due to slow dot-probe RT).