Camera shake and the stream-bounce effect

2022-May-19

In the ‘stream-bounce’ effect, the motion of a pair of dots is ambiguous and can be perceived as either streaming past or bouncing off one another. Here, I explore whether the addition of a simulated camera shake at around the time of the intersection of the dots can bias the perception of the dot motion towards seeing the dots as bouncing.

In the ‘stream-bounce’ effect, the trajectory of a dot moving down and to the right becomes ambiguous when intersecting with another dot moving down and to the left. The dot can be perceived to either continue on its trajectory (‘stream’) or be reflected and move in the opposite direction (‘bounce’).

Sekuler, Sekuler, & Lau (1997) reported that the likelihood of perceiving the dots as bouncing increased if a short sound was played at or around the time of the intersection of the dots. Interestingly, we can sometimes ‘hear’ sounds during silent visual motion—referred to as the “visual-evoked auditory response” by Fassnidge & Freeman (2018). The skipping pylon illusion is a great example of such a visual-evoked auditory response, where the inclusion of motion consistent with a ‘camera shake’ causes many people to hear a sound in the silent animation.

Here, I wondered whether adding a camera shake at around the time of intersection of the dots in a stream-bounce stimulus could cause them to be more likely to be perceived as bouncing. In the animation below, the first passage of the dots is without any camera shake and their motion is perceived as either streaming or bouncing (streaming, for me). In the second passage of the dots, there is a simulated camera shake around the time of the intersection of the dots. This causes me to see them as bouncing rather than streaming.

I’ve shown a demo like this to a few people over the years, with mixed reports. For some people, the camera shake strongly biases the perception towards bouncing—whereas it seems to have no effect for others. If it doesn’t seem to have any effect for you, there are some parameters below the animation that you might like to tinker with to see whether those have any effect. Perhaps it is just something that has large individual differences.

Time per sequence (s): How long it takes for the dots to move from one diagonal to the other. Note that changes can cause a few artefacts in the current cycle.
Shake amplitude (pixels): The maximum simulated camera displacement. The displacement is done via simple draws from a uniform distribution that are independent in horizontal and vertical space and in time. A low-pass filtered sequence would likely be more realistic.
Ramp in length (%): How long it takes for the camera displacement to reach its maximum after the time of the intersection of the dots, as a percentage of the time until the end of the sequence. The amplitude of the displacement increases linearly up to this point.
Ramp out length (%): How long it takes for the camera displacement to reach zero after it has reached its maximum, as a percentage of the time until the end of the sequence. The amplitude of the displacement decreases linearly up to this point.
Dot diameter (pixels): Changes the size of the dots.

If you are more likely to perceive the dots as bouncing when there is the camera shake, there are a few different potential causes. The most interesting one would be because the perceptual system has inferred that the camera shake is a consequence of the collision of the two objects, which then ‘explains away’ the streaming-dots hypothesis. However, maybe it is just that any type of transient event around the time of the intersection of the dots biases perception towards bouncing.

References

Fassnidge, C.J. & Freeman, E.D. (2018) Sounds from seeing silent motion: Who hears them, and what looks loudest? Cortex, 103, 130–141.
Sekuler, R., Sekuler, A.B., & Lau, R. (1997) Sound alters visual motion perception. Nature, 385(6614), 308–308.