1 Project summary

Learning the speech sounds of a language in adulthood is difficult–many adults have accents in a second language and even have trouble hearing or perceiving differences in sounds when listening. Some lab-based studies have shown time of day effects, in that learners who practiced learning sounds in the evening improved more on the task than learners who practiced in the morning. In this project, we wanted to test whether this pattern holds when analyzing data from a more naturalistic setting.

2 App user data

In this project, I analyzed data from over 200k users who played a game to learn the sounds of English on Rosetta Stone’s language learning app. We split users into time of day groups: those who practiced in the morning, daytime, or evening and tested their accuracy (percent correct) on the task the very next day. This was observational data, i.e., we were not able to randomly assign users to time of day groups. Because of this, we of course have to be careful of how we interpret the results. However, we have a much larger sample size than we have in typical lab studies. In short, though, we replicated some findings from the controlled studies run in the lab, namely that the group that learned in the evening improved more from one day to the next than the daytime and morning groups. The morning group performed better in their first session and maintained this performance in the second session but did not improve significantly from one day to the next. Although the evening group was less accurate in the first session than the morning group, they caught up overnight, even though they didn’t practice more in the meantime.

This is a poster I presented on the study and analysis.