The dynamics of auditory stream segregation: Effects of sudden changes in frequency, level, or modulation These datasets comprise listeners’ judgments of stream segregation for Experiments 1, 2, and 3 of the article of the same title (Rajasingam, Summers, and Roberts, 2021; Journal of the Acoustical Society of America). There are two spreadsheets for each experiment, one comprising the results when responses to the test sequence are divided into 1-s time bins and the other comprising the results when responses to the test sequence are divided into 0.2-s time bins. The latter can be combined as required to create longer time intervals with start times that do not correspond to an integer number of seconds. Each spreadsheet comprises four worksheets showing the extent of stream segregation (proportion of time the stimulus was reported as segregated) per time bin for each condition. The "Summary" worksheet shows the means and inter-subject standard errors when the results are averaged across listeners and is organized into 1-s time bins irrespective of the duration of the underlying time bins (see note below). The "DF4", "DF6", and "DF8" worksheets show the results separately for each listener for test sequences with frequency separations between low- and high-frequency tones of 4 semitones, 6 semitones, and 8 semitones, respectively. There are two tables in each of the "DF4", "DF6", and "DF8" worksheets. The upper table contains empty cells (missing values) indicating cases for which no mean could be computed (see article for full details). In the lower table, these missing values have been replaced using mean imputation. All inferential statistics presented in the article used the data with mean imputation. NOTE: When comparing the summary worksheets in corresponding spreadsheets, note that the 0.2-s time-bin data were converted to 1-s time bins on a trial-by-trial basis from the mean proportion of each constituent 0.2-s time bin reported as segregated. This results in discrepancies with the summary data computed from the 1-s time bins until such time as each trial for every listener has a complete set of 0.2-s time bins (i.e., no missing values) for the corresponding 1-s time bin. These discrepancies mostly occur within the first few seconds of the test sequence.