The Power of Visual Motion Priming

In 350 BC, Aristotle reported seeing illusory movement after staring at a constantly moving stimulus. Later coined the "waterfall illusion," scientists are still investigating the properties of motion aftereffects and what they tell us about the visual brain.

The motion aftereffect is a clear demonstration of neural fatigue — when motion-selective neurons are exposed to constant movement (like that of a waterfall), they eventually slow down their firing rate, as if they become tired or adapted. Then, when observing a static, non-moving stimulus, these fatigued neurons will fire relatively slower than their non-fatigued counterparts that encode other motion directions, making the new static stimulus appear to move in the opposite direction to the adapting motion. Here is a demo of the waterfall illusion you can try for yourself.

Visual motion priming

It is now known that certain properties of the adapting motion can modulate the motion aftereffect. In certain cases, when the adapting motion is slow and discrete (i.e. happening in short bursts), motion aftereffects tend to be observed in the same direction as the adapting motion. This is known as positive motion priming and is thought to depend on higher-order neurons that encode for a type of "visual inertia," assuming that objects in motion will remain in motion, unless impeded.

An even stranger type of motion priming has been observed, called rebounding motion priming (Hsieh, Caplovitz, & Tse, 2005; Davidenko & Heller, 2018). Here, a rebounding motion pattern (e.g. Up-Down-Up-Down, or Right-Left-Right-Left) will prime participants to continue seeing the rebounding pattern. These particular two-step motion sequences are thought to be sustained by oscillatory mechanisms in the brain that encode for alternating sequences of opposite-direction motion.

Can more arbitrary two-step motion sequences be primed?

In a new study published last month in the Journal of Vision, my colleagues and I discovered a new type of motion pattern that can be primed. We call this motion pattern "staircase motion" because it involves sequential motion steps at right angles (for example, Up-Right-Up-Right, or Left-Down-Left-Down).

Examples of staircase motion.

Source: Nicolas Davidenko

In four separate studies with a total of 224 participants, we primed observers with short motion sequences and then presented them with two final frames in which they were asked to report any motion they saw. In some trials, we showed real motion in the two final frames; these served as catch trials to make sure participants were following the instructions. In most of the trials, the final frames were completely random, but often participants reported seeing illusory motion. The question was, in what directions did the illusory motion go? Here you can see an example real trial and catch trial from our experiments.

As shown in the figure below, participants were most likely to report seeing motion that followed the same sequence as the priming motion. The darker blue squares along the diagonal show that each type of priming motion (e.g. Up-Right-Up-Right) resulted in more prime-consistent responses (e.g. Up-Right) than any other type of response (other than "XX" responses which indicate "other" or "none of the above").

Experiment results showing primed directions (y-axis) and reported direction (x-axis). Bluer colors are more frequent responses.

Source: Nicolas Davidenko

Overall, our results showed that (1) participants can indeed be primed to see staircase motion patterns, which was not previously known; (2) staircase motion priming requires at least four priming frames; that is, the sequence element (e.g. Up-Down) needs to be presented at least twice; and (3) priming these motion patterns requires real motion. When participants were primed with images of arrows instead of actual motion, they reported much lower rates of prime-consistent illusory motion in the subsequent random frames.

Visual motion priming requires memory

Currently, theoretical accounts of visual motion priming can explain positive priming (e.g. Up-Up-Up-Up) and rebound priming (e.g. Up-Down-Up-Down), but not staircase priming (e.g. Up-Right-Up-Right). This is because the two directions present in staircase sequences (say, "Up" and "Right") have no intrinsic relationship with one another. In order to have an illusory percept of "Up" followed by an illusory percept of "Right," the system needs to use memory to keep track of the motion that was presented two frames earlier. This suggests that visual motion priming relies, at least in part, on a sequence-learning mechanism. Future research is needed to explore just how much information our visual motion priming mechanism can encode and the extent to which arbitrary motion patterns can be primed.