Hack 32. Explore Your Defense Hardware
We have special routines that detect things that loom and make us flinch in response.
Typically, the more important something is, the deeper in the brain you find it, the earlier in evolution it arose, and the quicker it can happen.
Avoiding collisions is pretty important, as is closing your eyes or tensing if you can't avoid the collision. What's more, you need to do these things to a deadline. It's no use dodging after you've been hit.
Given this, it's not surprising that we have some specialized neural mechanisms for detecting collisions and that they are plugged directly into motor systems for dodging and defensive behavior.
2.21.1. In Action
The startle reaction is pretty familiar to all of usyou blink, you flinch, maybe your arms or legs twitch as if beginning a motion to protect your vulnerable areas. We've all jumped at a loud noise or thrown up our arms as something expands toward us. It's automatic. I'm not going to suggest any try-it-at-home demonstrations for this hack. Everyone knows the effect, and I don't want y'all firing things at each other to see whether your defense reactions work.
2.21.2. How It Works
Humans can show response to a collision-course stimulus within 80 ms.1 This is far too quick for any sophisticated processing. In fact, it's even too quick for any processing that combines information across both eyes.
It's done, instead, using a classic hacka way of getting good-enough 3D direction and speed information from crude 2D input. It works like this: symmetrical expansion of darker-than-background areas triggers the startle response.
"Darker-than-background" because this is a rough-and-ready way of deciding what to count as an object rather than just part of the background. "Symmetrical expansion" because this kind of change in visual input is characteristic of objects that are coming right at you. If it's not expanding, it's probably just moving, and if it's not expanding symmetrically, it's either changing shape or not moving on a collision course.
These kind of stimuli capture attention [Hack #37] and cause a startle response. Everything from reptiles to pigeons to human infants will blink and/or flinch their heads when they see this kind of input. You don't get the same effects with contracting patches, rather than expanding patches, or with light patches, rather than dark patches .2
Looming objects always provoke a reaction, even if they are predictable; we don't learn to ignore them as we learn to ignore other kinds of event.3 This is another sign that they fall in a class for which there is dedicated neural machineryand the reason why is pretty obvious as well. A looming object is always potentially dangerous. Some things you just shouldn't get used to.
In pigeons, the cells that detect looming exist in the midbrain. They are very tightly tuned so that they respond only to objects that look as if they are going to collidethey don't respond to objects that are heading for a near miss, even if they are still within 5º of collision.4 These neurons fire at a consistent time before collision, regardless of the size and velocity of the object.
This, and the fact that near misses don't trigger a response, shows that path and velocity information is extracted from the rate and shape of expansion. Now this kind of calculation can be done cortically, using the comparison of information from both eyes, but for high-speed, non-tiny objects at anything more that 2 m away, it isn't.5 You don't need to compare information from both eyes; the looming hack is quick and works well enough.
2.21.3. End Notes