# Understanding the data file The data file, `clean_data.csv`, is the public version of the data used in the analyses of Experiments 1 and 2, from Burchill, Liu & Jaeger (under revision)). Due to IRB protocols, all demographic and identifying information has been removed. Because certain exclusion criteria were linked to demographic information in the paper, the data here represents only the data from the participants who were not excluded (i.e., the participants used in the actual analysis). The data in this file represents participants' transcriptions of words heard during test trials, as well as additional information used in data exploration. ## The data: ### From the experiments: - `WorkerId`: a numeric label for each participant (randomized) - `Trial`: the order of each trial during test (the first trial is "0") - `Word`: the word being spoken aloud during the test trial. Given the Latin-square design used in the experiments, words that participants did not encounter during the test phase were heard in the exposure phase. - `Transcription`: the transcription for that trial - `BinaryCorrect`: a binary indicator of whether the transcription was counted as correct - `Condition`: the experiment condition the participant was assigned to. Note that the "Delayed" and "Concurrent" conditions differed slightly in their presentation across experiments (see paper). - `List`: the identifier of the list of stimuli the participant saw during exposure and test - `FalsePositives` & `Misses`: the number of false positives and false negatives participants made for the catch trials during exposure - `Experiment1`: whether the trials came from Experiment 1 - `Experiment2`: whether the trials came from Experiment 2 - `IsItemExcluded`: a boolean, which, if TRUE, **indicates that these particular items were excluded from analysis** due to the subtitle benefit not improving these words (see paper). ### From the post-test survey: - `AudioType` & `AudioQuality`: the type of headphones used in the experiment and the self-reported audio quality of their equipment - `FreqOfHearingStrongAccents`: the self-reported frequency with which they heard equally strong foreign accents (compared to the talker) - `FreqOfHearingSimilarAccent`: the self-reported frequency with which they heard foreign accents similar to the talker's - `LanguageRating`: a numeric rating of participants language exposure, coded from their answer to a free response question on their language background. (A number greater than 2 indicated they were not monolingual.) - `AudioStalling`: whether participants reported any audio stalling during the experiment