This section covers behavioral data acquired for the NSD dataset. Note that some behavioral data is provided in nsddemographics.xlsx (as documented in Informational files).
NSD experiment
nsddata/ppdata/subjAA/behav/responses.tsv
This is a tab-separated text file that contains all behavioral data from the NSD experiment for subject AA. After a header row, what follows is one row for every stimulus trial encountered by the subject. Stimulus trials from different runs and scan sessions are concatenated together.
Column 1 (SUBJECT) is the subject number (1-8).
Column 2 (SESSION) is the session number (1-40).
Column 3 (RUN) is the run number (1-12).
Column 4 (TRIAL) is the stimulus trial number (1-63 (odd runs) or 1-62 (even runs)). Note that the numbering of stimulus trials ignores (skips over) blank trials.
Column 5 (73KID) is the 73k ID of the presented image. (Note that here, the 73k IDs are provided as 1-based indices.)
Column 6 (10KID) is the 10k ID of the presented image. (Note that here, the 10k IDs are provided as 1-based indices.)
Column 7 (TIME) is the trial start time (i.e. time that the image comes on) as a MATLAB serial date number. The units are days. Time 0 is defined as the beginning (midnight) of the day that the subject’s first NSD core scan session took place.
Column 8 (ISOLD) is 0 (the image is novel) or 1 (the image is old).
Column 9 (ISCORRECT) is 0 (subject’s response was incorrect) or 1 (subject’s response was correct).
Column 10 (RT) is the reaction time in milliseconds (time between trial start time and button-press time).
Column 11 (CHANGEMIND) is whether this is a trial that involved more than one button press (0 = no, 1 = yes, NaN = no buttons pressed). We score only the final button pressed by the subject.
Column 12 (MEMORYRECENT) is the number of stimulus trials in between current and most recent presentation. 0 means the current and most recent presentation followed one another (no stimulus trials in between).
Column 13 (MEMORYFIRST) is the number of stimulus trials in between current and second most recent presentation. If there has been only one previous presentation, this is NaN.
Column 14 (ISOLDCURRENT) is 0 (the image is novel) or 1 (the image is old) with respect to acting as if the experiment included only the current session.
Column 15 (ISCORRECTCURRENT) is 0 (subject’s response was incorrect) or 1 (subject’s response was correct) with respect to acting as if the experiment included only the current session.
Column 16 (TOTAL1) is the total number of 1s ("novel") pressed during this trial. Will be a non-negative integer.
Column 17 (TOTAL2) is the total number of 2s (“old”) pressed during this trial. Will be a non-negative integer.
Column 18 (BUTTON) is the button pressed by the subject (1 = button 1, 2 = button 2, NaN = no buttons pressed). Note that there might be multiple buttons pressed during a trial; we score only the final button pressed (and consider the very first of a series of repeated presses of the same button).
Column 19 (MISSINGDATA) is 0 (button presses were recorded) or 1 (buttons failed to be recorded). This is very rare (it happened in two runs (see knowndataproblems.txt)), and if it happens, it happens at the level of entire runs. In the case that buttons failed to be recorded, note that columns 9-11 and 15-18 are necessarily NaN.
Note that columns 12-13 are NaN for the case of novel images. Note that columns 9-11, 15, and 18 are NaN if no button is pressed on a given trial.
prf experiment
nsddata/bdata/prf/
Results from the prf experiment. For each subject, <results> is [A,B,C] x 6 runs, where A is the total number of color changes, B is the number of hits, and C is the number of false alarms (extra button presses).
floc experiment
nsddata/bdata/floc/
Results from the floc experiment. For each subject, <results> is [A,B,C] x 6 runs, where A is the total number of trials, B is the number of hits, and C is the number of false alarms.
nsdpostbehavior
nsddata/bdata/cmtf/
Results from the Cambridge Memory Test for Faces experiment.
nsddata/bdata/flicker/
Results from the flicker-based assessment of chromatic sensitivity. While maintaining fixation, participants adjusted intensities of red, green, and blue channels on the BOLDscreen display until minimal luminance flicker was perceived. The basic presentation setup was to rapidly switch between two colors (A and B), performing this 15 times in 1 second. Three different trial types were conducted: (1) fix the green channel to 26, ignore blue, and vary the red channel, (2) fix the red channel to 77, ignore green, and vary the blue channel, and (3) fix the green channel to 26, ignore red, and vary the blue channel.
nsdmeadows
nsddata/bdata/meadows/
Results from the nsdmeadows experiment. A set of 100 images were chosen on the basis of their position in a semantic representational space. Participants performed three different behavioural tasks with these chosen stimuli. First, participants were asked to perform a multiple-arrangements task, arranging images according to their similarity with mouse drag and drop operations. Following this, participants performed additional arrangements along a valence scale, and along an arousal scale.
The data is stored in a .json file. The json dictionary has a key for each subject, and in each subject's subdict, there are 11 tasks. The first task is the multiple arrangements task, and this is followed by five separate blocks for valence and five separate blocks for arousal.
Some example code in Python can be found as part of the nsdcode repository here: