Freyd, J. J., & Finke, R. A. (1984). Representational Momentum. Journal Of Experimental Psychology: Learning, Memory, and Cognition, 10(1), 126–132. doi:10.1037/0278-7393.10.1.126

Summary

In this paper, Freyd & Finke present a number of experiments showing an effect of representational momentum in which people’s memories for the position/rotation of an object are distorted in the direction of implied motion of the object.

Methods

In Experiment 1, they showed participants a sequence of four images and asked them to determine whether the third and fourth images were the same or different. The first three images showed a rectangle in different orientations resulting from rotating it in a particular direction. The fourth image was either the same as the third, or rotated slightly forward or backwards from the third image (though not as much as the angular separation between the first three images). Freyd & Finke found that people were significantly faster and more accurate at judging the same (15.8% error) and reverse (6.4% error) stimuli, compared to the forward (43.9% error) stimuli.

In Experiment 2, they controlled for the effect of the result being somehow due to “some configural property of the inducing displays, independent of their temporal order”. Thus, they swapped the first and second images in the presentation. This caused the representational momentum effect to disappear entirely.

In Experiment 3, Freyd & Finke asked whether the representation momentum effect was due to sensory or cognitive processing. To test this, they increased the ISI (inter-stimulus interval) times to 500ms and 750ms (from 250ms). With these larger times, they still found the representational momentum effects, though they were less strong than at 250ms.

They additionally ran another control study to see if it was just whether people were extrapolating forward to the next image in the sequence (after the first three images). Instead of having small perturbations to the rotation of the third image, they used the next image in the sequence. They did not find the representational momentum effect using these larger rotations.

Algorithm

n/a

Takeaways

There seems to be something about people’s predictive processing of visual stimuli that causes them to misremember the location or pose of objects. I wonder if something like this could be explained by a system that, rather than storing the actual current perception, stores an expectation for the next perception. If there has recently been motion, then this would give rise to something like the representational momentum effect. It doesn’t quite fit with their experimental setup, though, which is that they showed each image for 250ms, then removed it for 250ms, then showed the next, etc. These long presentation times mean that apparent motion isn’t generated, so it doesn’t seem as if people are estimating the angular velocity and then propagating that forward—if they were, then presumably the last control experiment that Freyd & Finke found would have still found the effect. Perhaps one explanation for this is that the visual system is estimating velocity from the implied motion, but it only propagates it for a fraction of the time forward—i.e., it doesn’t estimate the ISI time and factor that in. If this were the case, then it should be possible to demonstrate that different angular velocities result in memory distortions that are either closer or further away from the true image.