Fleshing out the proposal document occupied pretty much my entire day today. I’m not sure what to summarize, beyond what is already there, since it was really a brainstorming process. I spent some time thinking about how we might want the study to be designed from a structure and participants perspective, which was mostly inspired by what I know and have read about other, similar studies. For example, it seems that interspersing “target” sentences with filler sentences is a good tactic to ensure we can differentiate the signal from the noise. We talked about how many sentences at once should be provided, which feels like something we can’t decide yet because we don’t know the final length of the sentences. At the moment I believe Adam has implemented a slider on the app that allows the user to choose a “length” for the entire test, so it seems like we could try different lengths and either let it be user controlled, or at least user tested for what is the optimal length.
We also tried running some of the actual sentences from Rayner’s study on the app to see what it would come out as, which led to the first problem that occurs to me. These sentences are, on average, quite short; at a certain font size, the entire sentence fits on the screen (landscape mode). This means we won’t be getting any useful accelerometer data unless we either change the UI (for example, portrait mode instead) or artificially make the sentences longer by adding filler context before the target words. I might try doing this for a few of Rayner’s sentences tomorrow.
I think I have fleshed out the proposal as much as I can unilaterally, except for leaving out a big chunk for tomorrow: filling in all that we know and expect based on the existing literature of such studies. Because I have already made a massive set of notes on this in Evernote, I thought I could dedicate tomorrow to compressing them, unless something else would be a more productive way to spend my time.