Social Information in Written Standard Sentence Materials: Methods and Data

[From the under-consideration paper, authored by Tripp & Munson]: The Harvard/IEEE (henceforth H/I) sentences are widely used for testing speech recognition in English. This study examined whether two talker characteristics, race and gender, are conveyed by 80 of the H/I sentences in their written form, and by a comparison set of sentences from the internet message board Reddit, which were expected to convey social information. As predicted, a significant proportion of raters reported perceiving race and gender information in the H/I sentences. Suggestions of how to manage the potential influence of this social information on measures of speech intelligibility are provided. This archive includes the raw data from this paper, and the code used to generate the experiment in Qualtrics. The later of these was programmed by Abbey Hammell.

Description
The reproducibility files for the Qualtrics survey (bound together in a .zip archive, TrippMunson2021_surveyProgramming.zip) have a readMe file that describes the information in each of the files (TrippMunson2021_surveyProgramming_readMe.docx) The data-set (TrippMunson2021_raterData.csv) has a readMe file (TrippMunson2021_raterData_readMe.txt) that describes each of the columns in it. Some of the columns can only be understood by reading one of the files (2277719_Munson_PrePilot_surveydetails.docx) in the .zip archive.

Funding information
Sponsorship: National Institutes of Health Grant R21 DC018070

Referenced by
Tripp, A., & Munson, B. (under consideration). Written standard sentence materials convey social information. Journal of the Acoustical Society of America Express Letters