## Changes from preregistered version ## ---------- #### Before data collection commenced #### - Before the start of the study, we changed the total stimulus duration to 19 seconds. This total duration was correct in the third paragraph of the study design section, but the first paragraph mistakenly still said 17s. The unexpected object appeared after 6 seconds and was on screen for 10 seconds. The other objects continued moving for 3 seconds after it finished crossing the screen. #### Data collection procedure - added detail #### - When participants in a batch of 100 did not complete the study, we could tell that because they either "returned" the study to prolific or had a "no code" error. For those with a no-code error, we messaged them on prolific asking them to return the study, which most did. Once they did, Prolific reactiviated those slots. So, each run of 100 involved 100 people who completed the study. #### After 200 participants #### - After running 200 participants, we removed the following from the instructions on the experiment itself: "Please close other browser windows and applications that might pop up notifications during the task, as those might end the experiment." The experiment on Prolific notes those requirements already, and we received messages from 2 participants indicating that they left the browser tab to follow those instructions (after starting the study) and that the experiment ended. We made this change to avoid that happening for future participants. The change has no effect on how the study runs. #### After scheduling 500 participant slots (data from ~493 participants) #### - After collecting data for the first 5 rounds of 100 participants slots, we received feedback from several workers that they thought the study didn't start until the display went full screen. They switched browser windows in order to close other tabs, and that led to them being kicked out of the study. On April 18, we changed the code so that switching away from the primary browser window would only end the study after it switched to full screen. All participants tested after that had this change in place. We hope that the change leads to fewer "returned" or "incomplete" studies from participants where no data were saved. - In order to reconcile our cash advance and to continue testing, we had to approve payment for 7 people who had not completed the study (they left the study browser window and it ended the experiment, but they did not return the study to Prolific). Consequently, we had 493 rather than 500 participants at the end of the 5th run. Prior to the 6th batch of 100, we added 7 slots (using the adjusted protocol above) to bring our total to 500.