Just saw an interesting seminar on honeybees and the algorithims they use to fly, land and communicate. No brain imaging, just interesting behavior from a brain the size of a piece of cous-cous.
How flexible is learning in insects? Can they apply rules they learn to novel situations? Can they generalize?
Bees have a compound eye. 9000 ommatidia to observe the world. 3 primary colors. 2 eyes are very close, so stereo vision is quite difficult using triangulation. To get 3d they use an active mechanism. Move, then see what happens to the image on the retina. When traveling in straight line, distant objects will track slowly across retina. Rotation screws up optic flow.
How do we train bees?
Can do it in your backyard. Miriam Lehrer (Zurich) is the expert. Puts out a feeder with high contrast stripes with dilute sugar water. When about to start expt, increase [sugar]. Slowly move the high sugar water inside, while leaving a dilute decoy outside. Only a few make it inside and know the secret high sugar solution and keep coming back.
Bees fly precisely through the middle of a hole. How do they do it without stereo?
They balance the image velocities in the two eyes. Put stripes on walls. Move one wall with the direction of flight and the bees will fly close to the wall. Reverse direction and they fly farther from the wall. Fairly independent of the spatial texture. Put a blank wall on one side and they will fly into it. This algorithim is now used in robot navigation. People on a treadmill also tend to do this.
They like to fly at 300 angular degrees per second. If you make the walls 2x wider, they fly 2x faster.
How does a bee perform a smooth, grazing landing on a horizontal surface?
Filmed trajectory of landing bees in 3-d using 1 camera imaging of bee and its shadow. (Bees don’t use the shadow). Variables. Height, forward speed, descent angle, descent speed. Linear realationship between horizantal flight speed and height. Bee keeps image velocity of ground constant during landing. Horizantal and descent speed are fixed by a coupling constant. What about landing on twigs? What about sparse optical flow? What about floating targets? What about a mirror?
Have bees fly at a spiral target. Bees want to keep the optic flow constant. If you spin the spiral, you can suck bees in or make them put on the brakes prematurely. Spatial frequency of the spiral does not matter.
How does a bee work out how far she has flown?
Karl von Frisch won the Nobel prize for this. They do the waggle dance. Position of food source is expressed in polar coordinates relative to sun. While the bee dances, other bees sample the food source from the dancing bee to assess the quality of the food source. Round dance is for feeder distances <50m (no direction), Waggle dance is for >50m. Number of loops and interwaggle duration indicates enthusiasm for food source.
How do bees perceive distance in a tunnel?
Force them to fly through a narrow tunnel. Tunnel entrance is close to hive. If feeder is at start, get round dance (0m). If they go down 6m down a narrow tunnel with vertical stripes to a feeder, when they come back they do a 200m waggle dance. They judge distance traveled by optic flow, not time or speed, etc. Change the stripes to horizontal, so no contrast along the axis of motion, they do the (0m) dance.
Dance cares about the length out, not the length back. Bees that are recruited search at a distance that is communicated by the dance. If you remove the tunnel, the trained bees search at the perceived distance.
If forced to do a detour around a hill, von Frisch showed they dance the straight-line direction, but the total path length.
Can other things be communicated by the waggle dance? Not that we know, but maybe we need to measure the dance acoustically. The dances are done in the darkness of the hive.
Compare the waggle dance for distances over land or over water. The “odometer” goes slower across water, which has less optic flow.
Can they signal uncertainty in their estimate? Variability in direction decreases as food gets further and further away.
Can bees navigate a maze?
Must follow a colored mark through a series of holes in a segmented box to find the reward. Bees learn in half a day. If you unmark and scramble the maze, then only the bees that have NOT learned the color marks can learn the correct path. They can also do a delayed match to sample task with delays up to 5 seconds. They can also associate between sensory modalities.
Can scent trigger memories of locations in bees?
Yes! Put rose-scented sugar water in location A inthe backyard. Train bees on the location. Move location to B, and change scent to lemon. Switch back and forth. Next day, put feeders out without any scent in location A and B. Puff rose scent on the hive, bees go to location A, puff lemon and the bees go to location B.