PP Remote Testing Wiki | Examples / ExampleBild
Examples using participant device / download-based platforms
BILD in the wild — a cautionary tale
We piloted a protocol for collecting data on the binaural intelligibility level difference (BILD) remotely. Subjects were school-age children and adults with self-reported normal hearing. The task was 3AFC word recognition, and the masker (when present) was a speech-shaped noise. Performance was evaluated in three conditions: 1) with the target and masker diotic (MoTo), 2) with the masker diotic and the target dichotic (MoTpi), and 3) with the target diotic and no masker (in quiet). Performance in quiet was included as a rough indicator of effects related to audibility. The BILD is improvement in performance between conditions 1 and 2 (MoTo-MoTpi). The primary goals of this project were to determine reliability and feasibility for remote BILD data collection in home settings and compare data collected "in the wild" with previously collected lab data.
Subjects were recruited from a database of previous participants and by word of mouth. Consent and instruction was conducted over Webex. Additional written instructions were provided, and subjects were encouraged to contact the experimenter as questions arose. Subjects used their personal computers to download and run a MATLAB executable script; only PCs were supported. They were instructed to set the system volume at 50%, and they listened using whatever headphones they had, plugged into the soundcard of the computer. Subjects used an app on their cell phone to measure and report ambient noise in the test environment at the beginning and end of each data collection run. Data were collected at three time points: day 1, day 2 and day 7.
So far we have run 28 children and 9 adults. Data were generally consistent with results obtained previously from adult subjects who were tested in the laboratory. However, data in the MoTpi condition from three subjects (2 children and 1 adult) were outliers at all three time points. Other data from these three subjects aligned with general data trends, including background noise levels. Data from these subjects were obtained from the same household using the same test equipment. Investigation into hardware specifications and headphone type are ongoing.
What we learned about remote testing
The finding of consistently odd data in one condition for a group of subjects using the same hardware highlights the need for robust procedures to differentiate individual differences from issues related to implementation (e.g., multiple participants from a household) and to evaluate results for outliers.
Contact for more information: Emily Buss or Claire Dorey
Page last modified on August 31, 2020, at 04:50 PM