I’ve been using a letter-counting test to keep hour-by-hour track of how well my brain is working. The test consists of 200 trials that ask how many of four displayed letters (e.g., YCAW) are from the set {ABCD}. for YCAW, the answer is 2. Faster answers = better brain function.
For the first several hundred tests, I kept the location of the four letters constant: the center of the window. As soon as I answered, the next display appeared in the same position as the last one. The display never repeated immediately; for example UXRA was never followed by UXRA. But UXRA could be followed by UXAR. This was too easy because it looked like the A and R had switched places. This was a big difference from the usual appearance and it signalled that the answer had not changed. Overlap between one display and the next was probably important but was hard to measure.
To make the test more uniform across trials, I had the display move up and down, which eliminated overlap between one display and the next. Successive displays appeared above center, below center, above center, below center, etc.
To my great surprise, this made the task a lot easier. Here are accuracy scores before and after the change:
Before the change, mean accuracy was 94.9% (standard error 0.2); after the change, 97.4 (standard error 0.3). The error rate was cut in half, in other words. I had no idea this would happen.
Reaction times were slightly more after the change. A treatment that changes reaction time and accuracy in conceptually opposite directions — makes the task harder in terms of reaction times (= longer reaction times) but easier in terms of accuracy (= great accuracy) — is very unusual. I don’t know of any other examples.
The displays have always been big black letters on a white background — very easy to read. But this change made them seem more visible somehow. At some high level of my visual system, it was if the contrast had been improved. It’s a funny feeling because I thought I was seeing them perfectly clearly with the old procedure.
Because accuracy is better it is now closer to constant, which is what you want in a reaction-time experiment. You want as much variation in reaction time as possible and as little variation in accuracy as possible.
Is your letter-counting test software publically available?
if you know how to use R I am willing to send it to you and help you use it.
I’m not familiar with R – and that’s a devilishly hard thing to Google. Can you tell me the vendor or some other hind, so I can learn more about it? Thanks!
Google “R language”
Thanks, that did it: https://www.r-project.org/
But I’m still confused. It looks like R is mostly a package for analyzing and displaying statistics. Did you actually create the application that flashes letters on your screen and lets you key in an answer in R? Or do you just use R to analyze the results? If the latter, then what I’m really interested in is the actual “letter-counting test” application.
You’re right, R is not designed for data collection. But I actually created the application in R that flashes letters, etc.