1,575 sound clips later…

and I can finally say I finished listening to all my data! It was definitely one of the more tedious aspects of this research project and there was absolutely no way around it. My study ended up having 35 participants, each providing data for 45 words. Each participant’s sound file took at least 30 minutes to listen through (for those of you keeping track, that’s 17.5+ hours of listening to made up words) so this was an incredibly daunting task. Chugging through everything took a lot of breaks, some encouragement from my advisor, and a considerable amount of ‘sound free time’ after everything was finished.

Since we were going with little to no example to work off of, I pretty much had free reign in determining how I wanted to go about tackling our data. Through a bit of trial and error, I discovered that the best method was to open each sound file (usually around 12-15 minutes long) in Praat (an acoustic analysis software), zoom in on a 60 second chunk (the largest length Praat will let you analyze at once) and listen to the bits of the clip containing the participant’s response. Luckily, most participants projected well enough that it was usually pretty easy to tell what part of the clip I needed to listen to. Unfortunately, nonce words like “pizadubifa” or “vazadutapi” usually don’t get articulated as clearly as real words, so some responses took two or more listens to make sure I was hearing the response correctly. This system ended up working pretty well for me and I kept record keeping simple – I wrote down each response using the sounds we had created our target words with (so it was sort of like an approximate phonetic system) so that we could go back to review data without needing to relisten to each sound file. Anything that seemed like metathesis was noted and the time location in the file was recorded for reference.

Needless to say, I’m glad this part of the process is over!


  1. Wow! I just read through all of your blog posts because this one intrigued me so much, and your research sounds fascinating! I can empathize with your participants’ uncertainty – I stumbled over “pizadubifa” and “vazadutapi,” and I was just reading the words. Metathesis seems like an interesting phenomenon but also extremely difficult to study. Does your research look at the participants’ responses based off of words on a screen or an audio clip? That might be interesting to consider (not that I know much about linguistics), because audio is so important in initially learning a language and learning the unique pronunciations of strange words, but we also learn a lot of new words from reading. Good luck with your data cleaning and analysis!