Psycholinguistics Lab Group at University of Michigan

In March 2015, I had the opportunity to go to the US and visit my home state of Michigan to gather some native English speaker data for the CCHP. It was very good, very productive trip. While there, thanks to the efforts of Lorenzo García-Amaya and Nick Henriksen at the University of Michigan, I also had the opportunity to talk about the corpus to the Psycholinguistics Lab Group there.

I spent much of the time talking about the make-up of the corpus itself plus some of the results that I have shared at other conferences and in other talks previously. However, one new bit of data was the results of a study I recently finished using Amazon Mechanical Turk. Using Turk was a bit of a learning process for me, but it was very valuable and provided data to me that I could not have got otherwise: A large amount of fluency ratings on the various CCHP second language recordings. The preliminary results of the Turk study suggest that raters pay attention to fluency factors that are basically depend on first language speech. It's an interesting and curious result that basically suggests that such ratings are not helpful to evaluate second language proficiency development. I've already written the results up in a paper that I hope to present later this year at another conference.

Anyway, the talk with the lab group went very well. As expected, UM has quite a bright group of people and I felt rather humbled to be there. There were some good questions and some good discussion afterward. If any one is interested, you can see my slides here. As for Turk, I will definitely be making more use of it in the future. If I can make one recommendation to others who are considering using Turk for their own research, I strongly suggest looking at some of the Turk user web sites (I mainly looked at Mturk Forum) so that you understand better how Turkers think about the work that they do. It is definitely useful for thinking about how to design and promote the work that you offer.

[Note: This post was published in August 2015 but has been dated in order to reflect the actual timing of the events described here.]