When it comes to working memory, the brain’s mental sketchpad, studies have largely converged on four as the magic number: It’s how many objects average adults can successfully hold in mind at once. Variations in this capacity are correlated with IQ — the more things you can think about simultaneously, the better able you are to make connections between them, and the “smarter” you’re considered.
But just why working memory can only juggle four items at a time has been poorly understood. Is working memory a discrete resource, meaning that once the four slots are filled, there’s simply no room for more objects? Or is it a flexible pool, divided more or less evenly among objects, and five or more cause it to be stretched too thin? Finally, where does the failure occur: in the initial perception of the objects, or later, in the remembering?
Now, MIT researchers have some answers. By studying visual working memory in monkeys — whose capacities are surprisingly similar to humans’ — they discovered that the limit of four can actually be broken down into two limits of two: one each for the left and right hemispheres of the brain. Because each hemisphere processes input from the right or left half of vision, memory for increasing numbers of objects depends on where in the visual field they appear.
“Surprisingly, we found that monkeys, and by extension humans, do not have a general capacity [for working memory] in the brain,” says Earl Miller, the Picower Professor of Neuroscience in MIT’s Picower Institute for Learning and Memory. “Rather, they have two independent, smaller capacities in the right and left halves of the visual space.”
So, not all groups of four objects are created equal: The brain can indeed remember up to four things, but it does best when those things are spaced out into two on the right side and two on the left. Any more than two on one side, and working memory starts to break down.
The results were published online this week in the Proceedings of the National Academy of Sciences (PNAS).
Sorry, can’t help you
In the researchers’ experiment, rhesus monkeys looked at an array of two to five colored squares on a black background. The screen went briefly blank, and then the squares reappeared — but this time, one square had changed color. The animals were trained to look at the changed square, or the “target,” to indicate that they knew which one it was. Previous research had predicted that the monkeys would do well on trials with up to four squares, but their performance would drop off when they had to remember five or more.
However, rather than the total number of objects across the array, the monkeys’ performance depended on the number of objects on each side of the visual space. For instance, even in certain trials with five objects — two on one side, three on the other — monkeys did fine, as long as the target object was one of the two. But in trials where the target was one of three objects all on the same side, even if there were no objects on the other side, monkeys were far less successful.
This suggests that visual working memory is split between the left and right side of the brain, and the two hemispheres are unable to transfer memory load between them. That is, if there are three objects on one side and only one on the other, the side with the lighter load can’t step in and relieve some of the other side’s burden. “Our study shows that both the slot and pool models are true,” Miller says. “The two hemispheres of the visual brain work like slots, but within each slot, it’s a pool.”
“The fact that we have different capacities in each hemisphere implies that we should present information in a way that does not overtax one hemisphere while undertaxing the other,” says Tim Buschman, a postdoc working with Miller and a co-author of the PNAS paper. “For example, heads-up displays [transparent projections of information that a driver or pilot would normally need to look down at the dashboard to see] show a lot of data. Our results suggest that you want to put that information evenly on both sides of the visual field to maximize the amount of information that gets into the brain.”
Breakdown from the bottom up
Furthermore, the researchers took recordings of neurons in the monkeys’ brains as they performed the visual task. They hoped this would shed light on where in the memory pathway the breakdown takes place: Does it happen as the monkeys are looking at the scene, failing to fully perceive it the first time around? Or do the monkeys see all the objects, but fail to encode what they’re seeing to be able to remember it later?
The researchers tracked visual information as it flowed from the parietal cortex, where sensory input is initially processed, to the frontal cortex, where higher-order structures encode it for memory. “We found that the bottleneck is not in the remembering, it is in the perceiving,” Miller says. Essentially, working memory for more than two objects in the same visual field was doomed from step one.
Recordings also showed that the total amount of information being processed increased from one to two objects, but that cortical areas became saturated after two, consistent with the idea that introducing more objects reduces the amount of information that can be stored about each one.
According to Edward Awh, a professor of psychology at the University of Oregon, one of the things that’s “very exciting” about this paper is the success of using neuronal recordings to examine working memory in non-human primates. “We haven’t gotten this kind of detailed cellular level information about the nature of capacity limits in working memory [before],” he says. He calls the strong lateralization between hemispheres in the monkey brain “curious and interesting,” but cautions against generalizing too quickly, since similar studies with humans have not found “such dramatic compartmentalization.”
Of course, if a similar phenomenon does exist in humans, it would have strong implications for how data should be presented to ensure maximum retention. Biomedical monitors that currently have one column of information should balance it in right and left columns, and security personnel could take in more information if displays scrolled vertically rather than horizontally, since horizontal scrolling doesn't fully take advantage of the independent capacities of the right and left.
In addition to Miller and Buschman, the PNAS paper was co-authored by Markus Siegel, another postdoc in Miller’s lab, and Jefferson Roy, a research scientist at the Picower Institute. In future studies, the researchers hope to discover why the perceptual bottleneck occurs in the first place, which would give “real insights into consciousness,” Miller says.