When your own mother might be an invisible gorilla: long-term memory and change blindness

Cognition is so interesting that we might never realize it. Most of the time we don’t think about our cognition, although we are constantly thinking, seeing, attending, memorizing, and deciding during our waking hours.

When teaching my first-year student “Intro to Cognitive Psychology”, the greatest joy for me stems from seeing their faces when they learn that much of what they intuitively believe to be true about the way they perceive the world and memorize events is actually quite misplaced.

Chris Chabris and Dan Simons’ famous “Invisible Gorilla” probably led quite a few of these first year students to studying Psychology in the first place. If you haven’t seen the invisible gorilla before, you can miss it in this video:

This and many other demonstrations of how limited some of our cognitive capacities are (like only seeing a small part of the visual field really sharply or the many ways you can trick one’s memory) seems mind-boggling to many.

At the same time, there is plenty of evidence that our cognitive capacities are actually quite amazing: What a toddler can easily identify as its beloved teddy underneath a blanked is still impossible to figure out by any state-of-the-art object recognition algorithm.

Even within one cognitive capability, say memory, we can find such paradoxical situations. On the one hand, there is evidence that we have an enormous ability to memorize thousands of images not only conceptually, but also in visual detail after seeing them only briefly; on the other hand, people fail to notice even large changes from one image to another when the change occurs during a fraction-of-a-second delay which is referred to as Change Blindness.

Before proceeding, if you haven’t experienced change blindness before, you should make a point of trying it out using this video:

In a recent paper published in Attention, Perception, & Psychophysics, Katherine Wood and Dan Simons have taken a closer look at this paradoxical inability to detect major changes in a scene by conducting a series of 6 experiments. The authors were particularly interested in examining the role that memorized information may play in the change detection task. I urge everyone to read this paper yourself, but in essence here is what Wood and Simons did and found.

Using a set of almost 3000 thumbnail images of objects that had been used in previous research, Wood and Simons asked participants to perform one or more (depending on the specifics of the experiment) of the following tasks:

  • A Study task: Here the participants studied a sequence of objects for 3 seconds each from the set of exemplar pairs. One randomly selected member of each pair served as a studied item and the other served as a lure item for use in later tasks.
  • A Change detection task with a familiar item: On each trial, participants viewed an array of 6 objects, equally spaced in an annulus configuration. This pre-change array appeared for 1,200 ms, followed by a white screen and then by a post-change array that remained onscreen until participants responded. Five of the objects in the pre-change display were drawn at random, without replacement, from the set of filler objects. The sixth object was the unstudied pair member that corresponded to one of the randomly selected studied exemplars. In the post-change array, the unstudied exemplar always changed into its studied counterpart—that is, the other pair member previously encountered during the study phase.
  • A Change detection task with all unfamiliar items: This task was identical to the change-detection task with a familiar object, except that none of the objects had been studied.
  • A 6-alternative forced-choice (6-AFC) memory task: One randomly selected, previously studied object and five randomly selected filler objects were displayed in an annulus configuration. Participants had unlimited viewing time and were instructed to click on the object they had seen during the study phase.
  • A 2-AFC memory task: A studied exemplar and its unstudied partner from the same category were presented on screen. Participants had unlimited viewing time, and indicated the object they had seen before.

These tasks were used in various combinations across 6 experiments. Would familiarity affect change detection performance? The results can be summarized as follows:

Experiment 1 showed that participants do not just spontaneously use familiarity information to assist with change-detection, even when the presence of a familiar object perfectly indicated the location of the change. Maybe they just need to be explicitly told? Well, not really: When explicitly told about the familiar objects in Experiment 2, change-detection performance did not improve beyond the condition with completely unfamiliar objects.

In Experiments 3 and 4, participants were given instructions to either use recognition memory or change-information to aid change-detection and recognition, respectively. In both experiments, providing an explicit strategy for combining change- and long-term memory information improved participants’ performance up to the level of their 6-AFC recognition accuracy, but not beyond. It seems that participants simply neglected the additional information available via change-detection and instead relied solely on long-term memory information to complete the task.

In all these experiments, performance would have benefitted from using both sources of information, but only in Experiments 5 and 6, when the change detection task was made easier by increasing the pre-change display duration, did participants finally manage to use both.

It seems weird that we do not seem to make use of familiarity information stored away in long-term memory, as seen in Experiments 1 and 2. Wood and Simons suggest that one reason for this could be that using the currently available visual information might be more efficient that relying on long-term memory, which might be faulty. A similar suggestion has been made in a previous study on visual search as well (e.g. a study by Wolfe and colleagues as well as in a recent review we wrote on the influence of different types of memory on visual search performance).

In essence, this nice series of six experiments suggests that when it comes to accessing potentially available and useful information, we do not always make use of it. Being the lazy human beings that we are lets us base the decision about what information we will use to perform a task on the amount of effort required to extract that information.

Article focused on in this post:

Wood, K, & Simons, D. J. (2017) Reconciling change blindness with long-term memory for objects. Attention, Perception, & Psychophysics, 79, 438-448. DOI: 10.3758/s13414-016-1240-2.

You may also like