Boundary Extension

One of the most impressive illusions in visual memory is called “boundary extension,” which is discovered by Intraub and Richardson in 1989.  This discovery is perhaps the first and only instance that runs opposite to almost all of the previous findings in the literature.  Here is an example of the boundary extension:

After people view the left photo in Figure 1 and some time passes, they insist that the right photo in Figure 1 is the one they ever saw.  To them, the photo on the left appears “too close”.

boundaryextension

Figure 1.  An example of the “boundary  extension” illusion.  After the left photo is viewed and remembered, people insist that they saw the right photo, and that the one on the left appears too close.  (From Figure 1 in Intraub and Richardson, 1989)

This phenomenon is very robust, and counters almost all known results in the visual cognition and visual memory literature. Usually, a picture is better recognized if it has already been seen in the past, but in “boundary extension,” the original picture is never recognized as such.

We find this cognitive illusion very appealing, because it provides a unique window into the way in which internal memory representations are encoded.  The appeal is analogous to why visual illusions are so appealing to perceptual psychologists.  That is to say, visual illusions, which are highly consistent across people and resistant to cognitive knowledge, provide a unique window into the way in which the visual perceptual brain functions.

Our approach to study “boundary extension” is functional.  Namely, we aim to characterize it in information processing terms by teaming up with Helene Intraub of University of Delaware, the discoverer of boundary extension.  We design new experiments that will ascertain, in Signal Detection Theory terms, to what extent boundary extension is due to discrimination sensitivity (or d’) change, and due to decision criterion shift (or bias).

Now, why is it interesting to figure out sensitivity and bias changes separately?  Teasing apart sensitivity and bias always deepens our understanding as to what might be going on in the underlying memory process, more so than using percent accuracy.

If the effect is only due to a decision criterion shift, then the representation of a visual scene (e.g., the left photo in Figure 1) has nothing new to tell.  The label for the associated viewing angle is simply “zoomed out” but not necessarily “reconfigured” in our memory.  On the other hand, if there is indeed a sensitivity change, this says something very interesting about the constructive nature of memory and perception.  For example, the left photo could be memorized in a way that is more similar to the right photo than the left photo itself – so upon viewing the photo, the viewer actually consolidates a memory representation of the image that is itself automatically “zoomed out.”

One visual phenomenon that might be related is “amodal completion,” which refers to the perceptual “filling in” of visual scenes that were occluded when viewed.  As this perceptual filling in occurs within a larger scene, it is traditionally considered a process of interpolation.  A similar mechanism might be behind the boundary extension phenomenon.  However, because the “filling in” present with boundary extension occurs outside of the previously viewed scene, this process may actually be better classified as an extrapolation that gives rise to a memory representation that is “wider” than the original image.  If this is what is going on, then the visual memory system should extend the boundary in such a way that this “prediction” is behaviorally beneficial and, importantly, not to overcommit to backfire.

This is the reason, as a visual perception lab, we are interested in this high-level and powerful illusion.