What did you eat for dinner one week ago today? Chances are, you can’t quite recall. But for at least a short while after your meal, you knew exactly what you ate, and could easily remember what was on your plate in great detail. What happened to your memory between then and now? Did it slowly fade away? Or did it vanish, all at once?
Memories of visual images (e.g., dinner plates) are stored in what is called visual memory. Our minds use visual memory to perform even the simplest of computations; from remembering the face of someone we’ve just met, to remembering what time it was last we checked. Without visual memory, we wouldn’t be able to store—and later retrieve—anything we see. Just as a computer’s memory capacity constrains its abilities, visual memory capacity has been correlated with a number of higher cognitive abilities, including academic success, fluid intelligence (the ability to solve novel problems), and general comprehension.
For many reasons, then, it would be very useful to understand how visual memory facilitates these mental operations, as well as constrains our ability to perform them. Yet although these big questions have long been debated, we are only now beginning to answer them.
Memories like what you had for dinner are stored in visual short-term memory—particularly, in a kind of short-term memory often called “visual working memory.” Visual working memory is where visual images are temporarily stored while your mind works away at other tasks—like a whiteboard on which things are briefly written and then wiped away. We rely on visual working memory when remembering things over brief intervals, such as when copying lecture notes to a notebook.
The question is: when are these memories wiped away? And when they are, can we still discern traces of what was originally ‘written,’ or does nothing at all remain? If visual short-term memories are only gradually wiped away, then remnants of these memories should still be retrievable; but if these memories are wiped out all at once, then we shouldn’t be able to retrieve them in any form whatsoever.
UC Davis psychologists Weiwei Zhang and Steven Luck have shed some light on this problem. In their experiment, participants briefly saw three colored squares flashed on a computer screen, and were asked to remember the colors of each square. Then, after 1, 4 or 10 seconds the squares re-appeared, except this time their colors were missing, so that all that was visible were black squares outlined in white. The participants had a simple task: to recall the color of one particular square, not knowing in advance which square they would be asked to recall.
The psychologists assumed that measuring how visual working memory behaves over increasing demands (i.e., the increasing durations of 1,4 or 10 seconds) would reveal something about how the system works.
If short-term visual memories fade away—if they are gradually wiped away from the whiteboard—then after longer intervals participants’ accuracy in remembering the colors should still be high, deviating only slightly from the square’s original color. But if these memories are wiped out all at once—if the whiteboard is left untouched until, all at once, scrubbed clean—then participants should make very precise responses (corresponding to instances when the memories are still untouched) and then, after the interval grows too long, very random guesses.
Which is exactly what happened: Zhang & Luck found that participants were either very precise, or they completely guessed; that is, they either remembered the square’s color with great accuracy, or forgot it completely. It was almost as if their memories behaved like files on a computer: Your Microsoft Word documents don’t lose letters over time, and your digital photos don’t yellow; rather, they continue to exist until you move them into the trash—where they are wiped out all at once.
But this, it turns out, is not true of all memories. In a recent paper, Researchers at MIT and Harvard found that, if a memory can survive long enough to make it into what is called “visual long-term memory,” then it doesn’t have to be wiped out at all. Talia Konkle and colleagues showed participants a stream of three thousand images of different scenes, such as ocean waves, golf courses or amusement parks. Then, participants were shown two hundred pairs of images—an old one they had seen in the first task, and a completely new one—and asked to indicate which was the old one.
Participants were remarkably accurate at spotting differences between the new and old images—96 percent. In other words, despite needing to remember nearly 3,000 images, they still performed almost perfectly.
However, it turns out that they were only this accurate when the new and old images came from different types of scenes (e.g., a golf course and an amusement park). In order to test just how detailed these memories really were, the psychologists also analyzed how participants performed when the images were from the same types of scenes (e.g., two different amusement parks). Since images from the same scene type differ from each other in fewer ways than do images from different scene types, the only way participants would’ve been able to succeed at pointing out differences between these similar images is if they had remembered them with a truly vast amount of detail.
As you might expect, participants were worse at discriminating between same-category images, but not by much, scoring as high as 84 percent. In fact, even when the experimenters increased the number of images that participants initially needed to remember for a given type of scene, participants were still good at distinguishing the old image from the new—with only slight decreases in performance. That said, the fact that memory performance decreased at all shows that, although our memories are very detailed, they are not photographic.
These two separate experiments present a paradox: why are we capable of remembering such a massive number of images with great detail in some instances, and not even a few images after a couple of seconds in others? What determines whether an image is stored in long-term vs. short-term memory?
In a recent review, researchers at Harvard and MIT argue that the critical factor is how meaningful the remembered images are—whether the content of the images you see connects to pre-existing knowledge about them. In the Zhang & Luck experiment, you try to remember meaningless, unrelated colors, and so no connection with stored knowledge is made; it’s as if the white board is scrubbed clean before you get a chance to copy the scribbles into your notebook. But in the Konkle et al. experiment, you see images of recognizable scenes that you already have meaningful knowledge about—such as where the roller coaster is likely to be located relative to the ground. This prior knowledge changes how these images are processed, allowing thousands of them to be transferred from the whiteboard of short-term memory into the bank vault of long-term memory, where they are stored with remarkable detail.
Together, these experiments suggest why memories are not eliminated equally— indeed, some don’t seem to be eliminated at all. This might also explain why we’re so hopeless at remembering some things, and yet so awesome at remembering others.
by Julian De Freitas, a Cognitive Science major in the Psychology Department at Yale University.