Ancel, Elizabeth ESmith, Michael LRao, V. N. VimalMunson, Benjamin2023-05-182023-05-182023-05-18https://hdl.handle.net/11299/254230This dataset was used to investigate perceptual judgments and acoustic measurements of 3- and 4-year-old children’s productions (n=117) of /ɹ/- and /w/- initial words. The children’s productions of /ɹ/- and /w/- initial words were recorded and then rated by professional transcribers into four narrow transcription categories:[ɹ], [ɹ]:[w] (in between [ɹ] and [w], but closer to an [ɹ]), [w]:[ɹ] (in between [ɹ] and [w], but closer to a [w]), and [w]. The productions were also rated by untrained listeners on a visual-analog scale. Lastly, acoustic measurements were obtained using the Triple Formant Tracker developed by Cychosz et al. (2021). The triple formant tracker is a customized Python script that measures formants using three different algorithms: inverse filter control formant (Watanabe, 2001), ESPS covariance (Talkin, 1987), and ESPS autocorrelation (Talkin, 1987). The tracker used these three different methods to calculate the formants at five different time points over a specified 75 millisecond duration: 0 ms into the token (i.e., the onset), 18.75 ms into the token (the first quartile), 37.5 ms into the token (the midpoint), 56.25 ms into the token (the third quartile), and 75 ms into the token (the offset). The median of the three formant values obtained from the three algorithms was taken as the final formant measurement for each formant at each of the five time points. For a subset of the children (n=14), manual measurements of the second and third formant were made by a trained individual. For more detailed information about this data, see Ancel et al. (2023) and Munson et al. (2021). This dataset contains three .csv files. The first file, "rw_auto_all," contains data from all 117 children. This dataset has the professional transcriptions, untrained listener ratings, and automated acoustic measurements produced by the Triple Formant Tracker (Cychosz et al., 2021). Code is provided in “Figure_Share.Rmd” to subset this data so that it contains the same 14 children that are in the “rw_manual_subset” file. The second file, "rw_manual_subset," contains data from a subset of 14 children. This dataset has the professional transcriptions, untrained listener ratings, and manually coded acoustic measurements.The third file, "rw_accuracy_all," contains accuracy data from all children (with some additional children that were not part of the acoustic measurement). This data also provides three .Rmd files to be used with the statistical computing program, R. The first file, "Figures_Share," contains code to create Figures 1-4 described in Ancel et al. (2023), as well as Supplemental Figures 2-3. "Models_Share" contains code that was used to perform statistical analyses in Ancel et al. (2023). Lastly, "ROC_Curves_Share" contains code that was used to created figures and perform statistical analyses related to the "Robustness of Contrast" portion of the paper.This data contains acoustic and perceptual measurements of 3- and 4-year-old children's productions of /ɹ/- and /w/- initial words. Young children's production of the /ɹ/ sound is highly variable and often inaccurate, with [w] as the most common substitution error. One acoustic indicator of the goodness of children’s /ɹ/ productions is the difference between the frequency of the second formant (F2) and third formant (F3), with a smaller F3-F2 difference being associated with a perceptually more adult-like /ɹ/. This data contains both automatically extracted F3-F2 differences for the full set of 117 children, as well as manual measurements of a subset of the children (n=14). Additionally, this data contains untrained listeners' perceptual measurements of the children's productions for comparison.Attribution 3.0 United Stateshttp://creativecommons.org/licenses/by/3.0/us/AcousticsAssessmentDevelopmentConsonantDatasethttps://doi.org/10.13020/dged-7d58