csjarchive.cogsci.rpi.edu/proceedings/2011/papers/0039/…

"An Ideal Observer Model of Visual Short-Term Memory Predicts Human Capacity–Precision Tradeoffs"

There's a whole lot of perceptual decision-making papers out there, but I wanted to share this one because of something the authors did in particular: they calculated the number of bits of memory their subjects were using to perform the given task.

The experiment: you are briefly shown some colored points randomly placed along the edge of a circle, and then you are shown a circle with only one of those colored points, but displaced by a couple degrees from its initial position. Your goal was to guess whether the point moved clockwise or counterclockwise.

This was modeled as an encoding-decoding scheme. Your visual input is encoded into some finite amount of memory, which is then later decoded. As you learn to do the task, your brain's encoding/decoding scheme will quickly adjust to make optimal use of the limited memory, but that finite memory places a fundamental limit on the precision with which you are able to remember the locations of all N dots.

As in, viewing this as a case of lossy compression... using math, you can calculate how much loss -> how much memory used.

And they concluded that data across a variety of experimental conditions (number of dots, variance in position) was explained by people using a constant short-term visual memory capacity of a little over 8 bits.