I first learned about the Stop Making Sense task in the spring. I don’t remember the context, but afterwards I searched for it. Sadly, I gave up before I found a paper that actually described the methodology. Then yesterday, it came up during a guest speaker’s lecture.
Independently, the Stop Making Sense task was mentioned at a discussion group I went to this morning. So I now can explain it (or attempt to, at least: I must admit I know very little about experimental linguistics—although the explanation this morning was from an excellent experimentalist. An experienced expert).
The Stop Making Sense task is variant of the self-paced reading task—both used to study the cognitive processing of language by measuring response times. On a computer screen, a subject is presented the first word of a sentence. She then clicks a button, at which point the next word appears (whether or not the first word disappears is a methodological choice, as far as I can tell). At this point the subject can press the button again, and the third word will appear. Or—by pressing a different button—they can indicate that the sentence no longer makes sense. Until they click the Stop Making Sense button, this goes on, potentially arriving at the end of a perfectly sensible sentence. This lets you pinpoint where people think the sentence goes wrong, and you can also measure how long it takes the subject to decide whether or not the sentence still makes sense.
For instance, in an early (possibly first) implementation of the task by Boland, Tannenhaus, and Garnsney, subjects were presented sentences that looked like different version of the things below:
- Which frog/snake did the girl force to hop past the rock?
- The girl forced the frog/snake to hop past the rock?
In the both sentences, if the subject has the “snake” version, she might realize frogs don’t hop as soon as she gets to the word “hop” and hit the Stop Making Sense button. The theory they wanted to test was whether it took longer to catch the anomaly in the question version. There does indeed seem to be some difference.
(via lukesimcoe)
