Data -------------- Here you can all of the following: - Raw data from all tasks: [zip archive][1] - R script used to analyze the data: [.txt file][2] - File with the coded data (output from the R script): [.csv file][3] - File with East-West coding for participants at the University of Illinois: [.csv file][4] - File with results output from the R script: [.txt file][5] - R code to create the graphs in the paper: [.txt file][6] Corrections -------------- There was a problem with the way that "Correct" column in the "global/local" experiment was recording data. This column was meant to represent whether the participant responded correctly to whether one of the target letters ("E" or "H") was present in the display. The KeyPress column generated multiple responses on some trials but not on others. For example, on certain trials, it registers just ['up] and codes that as an incorrect response. On the second trial, it registers ['down','down'] and registers that as correct. When it only registers one key press, it counts that as an incorrect response even if it was the correct response. For example, up was the correct response for the first trial, but the "Correct" column marks it as "no." This means that in this column, there were far more "incorrect" responses than there should have been. We were able to recalculate this column from the information in the output file, so it should now be correctly represented in the analyzable form of the raw data. Subject Elimination Criteria ----------------------- We separated the participants into two groups (East and West) based on their responses to pre-screening questions that they submitted at the beginning of the semester when they first registered for the subject pool. The questions and responses we used to assign people to groups are described in detail in the "Procedures" component. We verified their responses to the pre-screening questions by looking at their responses to the background questions, which were included at the end of the study. We eliminated data from Eastern participants who reported spending more than 1 year in the United States or other Western countries (such self-reports were inconsistent with their responses to the subject pool pre-screening questions, so we excluded their data in an effort to be conservative in classifying participants as Eastern). None of the Western participants reported having spent more than 1 year in Asia, so data from all participants whose pre-screening responses permitted them to see the study number corresponding to Western participants were included in the analyses. We created an Excel spreadsheet indicating this coding and imported it directly into the analyses using the R script that handled all data importing and analysis. All data were imported, with an additional column noting whether data from that participant needed to be excluded due to a failure to meet the inclusion criteria. Data Coding and Manual Corrections ---------------------------------- During data entry, several minor errors were corrected manually in the raw data files available on this site. Below we detail all of those corrections. - For the data files for all of the tasks, subject 130 was accidentally entered as 30. The raw data files were corrected by replacing 30 with 130 in the participant number column, and the file names were changed to indicate the correct subject number. - The following participants did not enter their sex in response to the demographic questions: 12, 16, 17, 18, 22, 28, 57, 58, 61, 65, 72, 85, 98, 105, 112, 124, 126, 134, 149. The experimenter noted each participant's sex at the time of testing, so we manually entered their sex into their Survey responses. Anomalous Data -------------- - Subject 26 appeared to reverse the response keys for the Expand/Contract task. The miss rate exceeded the hit rate, and the false alarm rate exceeded the correct rejection rate. In the contract condition, they had 4/34 hits and 0/40 correct rejections. In the expand condition, they had 12/37 hits and 3/53 correct rejections. - Subject 96 got 0 out of 85 trials correct in the global case, presumably indicating that they misunderstood the task. Their accuracy was 69/72 for local and 130/131 for target absent. - Subject 135 got 0 out of 76 correct in the global case, presumably indicating that they misunderstood the task. They got 66/69 correct for local and 143/143 correct for target absent. - Subject 24 got 1 out of 89 correct in the global case, presumably indicating that they misunderstood the task. They got 68/75 correct for local and 124/124 correct for target absent. - Subject 38 got 0 out of 75 correct in the global case, presumably indicating that they misunderstood the task. They got 77/78 correct for local and 135/135 correct for target absent. - Subject 42 got 0 out of 69 correct in the global case, presumably indicating that they misunderstood the task. They got 72/79 correct for local and 140/140 correct for target absent. - Subject 51 got 5 out of 66 correct in the global case, presumably indicating that they misunderstood the task. They got 63/64 correct for local and 153/158 correct for target absent. - Subject 6 got 0 out of 77 correct in the global case, presumably indicating that they misunderstood the task. They got 73/74 correct for local and 137/137 correct for target absent. - Subject 80 got 11 out of 66 correct in the global case, presumably indicating that they misunderstood the task. They got 80/83 correct for local and 132/139 correct for target absent. - Subject 15 got 3 out of 139 correct in the target absent case, presumably indicating that they misunderstood the task. They got 56/74 correct for global and 70/75 correct for local. - Subjects 2 and 4 apparently continued using a 1-9 scale when responding to the Big 5, which led to missing data for one or more of the Big5 dimensions (see note on the procedure change after subject 4 in the Procedures component). Data from the Big5 were marked for exclusion in the analysis. - Chinese Subject 68 (renumbered as 368) in the Expand-Contract task responded on all but 1 trial using the up key. That wasn't caught by the automatic corrections we had used before. We added an automatic exclusion rule below to address this problem. Automated Corrections for Anomalous Data ---------------------------------------- - For subjects like #26 who appear to have reversed response keys for a task, we treated their data for that task as missing rather than recoding their responses. We implemented the following procedure into our coding of those conditions: For any participant who showed lower hit rates than miss rates AND lower correct rejection rates than false alarm rates in BOTH the contract and Expand conditions, we marked them for exclusion from the Expand/Contract analysis. The only plausible way in which all of those patterns hold is if they reversed response keys (response bias would not have that effect). - For subjects in the Global/Local task who apparently misunderstood task instructions and rarely or never reported Global changes, we marked them for exclusion if they were accurate less than half the time in either the Global or Local condition (i.e., the number of Global errors exceeded the number of Global correct responses OR the number of Local errors exceeded the number of Local correct responses). - Subjects with missing data for one or more of the Big5 dimensions were marked for exclusion from any analyses involving the Big5 measures. - For the Expand-Contract attention task, we excluded participants who almost always responded "no change" or almost always responded "change." If they always responded "change," their False Alarm rate across conditions would be 100%. If they always responded "no change," their Miss rate across conditions would be 100%. We excluded participants whose Miss rate or False Alarm rate across conditions was greater than 90%. Such participants are not doing the task. - For the Relative-Absolute length task, some participants apparenrly adjusted the line length away from the response box, resulting in a negative length. Others did not adjust the lenth at all, resulting in a 0 response length. We excluded any participants with 0 or negative responses from all analyses of this task. Note that we only detected this failure to follow instructions after examining the data. [1]: https://osf.io/9n2sz/ [2]: https://osf.io/7f45n/ [3]: https://osf.io/r59wq/ [4]: https://osf.io/d4tzv/ [5]: https://osf.io/umkah/ [6]: https://osf.io/rbe28/