Ants control number of steps to home

[from Gary Cziko 2006.07.26 1729 GMT]

Here is an interesting article showing that desert ants control their perception of the number of steps they take when returning home after foraging:
http://www.sciencemag.org/cgi/content/abstract/312/5782/1965

–Gary

[From Rick Marken (2006.08.01.0950)]

Gary Cziko (2006.07.26 1729 GMT)

Here is an interesting article showing that desert ants control their perception of the number of steps they take when returning home after foraging: http://www.sciencemag.org/cgi/content/abstract/312/5782/1965

This looks great, Gary. Thanks. But it's a little hard to tell what they did based only on the abstract. If you (or anyone else) have access to the whole article, could you give me a few more details on their methods. What confuses me is the last part of the abstract where they say that travel distance is overestimated by animals walking on stilts and underestimated by animals walking on stumps. How do they determine what the animals distance estimates are? I would imagine that they would have an animal walk to a goal under normal conditions and then walk to it on stilts or stumps. So I think what they mean is that, when walking on stilts they walk past the goal (since N steps takes them farther on stilts than on legs) and when walking on stumps they don't make it to the goal (since N steps doesn't take them as far on stumps as on legs). Is that what they did? So they really don't really over or under "estimate" distance; they just end up going too far or not far enough when on stilts ot stumps, respectively.

This is a version of "the test" where the stilts or stumps represent a disturbance to a possible controlled variable (distance measurement). But it seems to me that they haven't quite isolated the controlled variable. The conclusion of the article is that "step count" is a controlled variable. But couldn't it also be integrated joint angle, assuming that each step (whether on legs, stilts or stumps) moves through approximately the same angle. Maybe Richard Kennaway could speak to this. Based on the abstract, it seems to me that the researchers have _eliminated_ one possible controlled perception; the perception of distance travelled. But they have not actually carried the test to the point where they can say that number of steps is the controlled variable.

Anyway, I think this is an excellent example of research that is very much like the kind of research that would be done from a PCT perspective.

Best

Rick

···

---
Richard S. Marken Consulting
marken@mindreadings.com
Home 310 474-0313
Cell 310 729-1400

[from Gary Cziko 2006.08.01 17:01 GMT]

This looks great, Gary. Thanks. But it’s a little hard to tell what
they did based only on the abstract. If you (or anyone else) have

access to the whole article, could you give me a few more details on
their methods. What confuses me is the last part of the abstract where
they say that travel distance is overestimated by animals walking on

stilts and underestimated by animals walking on stumps. How do they
determine what the animals distance estimates are? I would imagine that
they would have an animal walk to a goal under normal conditions and
then walk to it on stilts or stumps. So I think what they mean is that,

when walking on stilts they walk past the goal (since N steps takes
them farther on stilts than on legs) and when walking on stumps they
don’t make it to the goal (since N steps doesn’t take them as far on
stumps as on legs). Is that what they did?

That is my understanding.

So they really don’t really

over or under “estimate” distance; they just end up going too far or
not far enough when on stilts ot stumps, respectively.

Right.

This is a version of “the test” where the stilts or stumps represent a
disturbance to a possible controlled variable (distance measurement).
But it seems to me that they haven’t quite isolated the controlled

variable. The conclusion of the article is that “step count” is a
controlled variable. But couldn’t it also be integrated joint angle,
assuming that each step (whether on legs, stilts or stumps) moves

through approximately the same angle.

To test this, they would have to restrict the range of motion of each leg and see if they compensate (or not) by taking more steps. This sounds a lot harder to do than shortening or lengthening their legs.

Also of interest is that fact that the altered ants eventually learn to adjust their steps to get back home. So the higher-level controlled variable is getting home. And when the usual number of steps (or integrated joint angles) doesn’t deliver the goods, they reorganize and learn to change their steps.

Hey, this PCT stuff can be pretty useful for understanding ants!

–Gary

···

On 8/1/06, Rick Marken marken@mindreadings.com wrote:

[From Rick Marken (2006.08.01.2140)]

Gary Cziko (2006.08.01 17:01 GMT)--

Me:

This is a version of "the test" where the stilts or stumps represent a
disturbance to a possible controlled variable (distance measurement).
But it seems to me that they haven't quite isolated the controlled
variable. The conclusion of the article is that "step count" is a
controlled variable. But couldn't it also be integrated joint angle,
assuming that each step (whether on legs, stilts or stumps) moves
through approximately the same angle.

To test this, they would have to restrict the range of motion of each leg and see if they compensate (or not) by taking� more steps.� This sounds a lot harder to do than shortening or lengthening their legs.

Yes. But you would have to do something like this because that's what research on control is all about: finding the perceptual variables that organisms actually control. Another approach to doing "the test" to see if it's steps or angles that are integrated is through the use of modeling. So you could write models of ant navigation that include each of these two different versions of the controlled variable. Then you can compare particular behaviors of the models to that of the actual organism. I used this strategy in the research (on Cartesian vs Polar representations of the 2-D position of a cursor) described in my "Degrees of freedom in behavior" paper in _Mind Readings_.

Also of interest is that fact that the altered ants eventually learn to adjust their steps to get back home. So the higher-level controlled variable is getting home. And when the usual number of steps (or integrated joint angles) doesn't deliver the goods, they reorganize and learn to change their steps.

Yes. Very interesting! I wonder whether that's a true reorganization (in which case the behavior changes would be somewhat random before the correct scaling is achieved) or changes implemented by an existing higher order control system. Did they say anything about how the adaptation to the stilts/stumps occurred over time?

Hey, this PCT stuff can be pretty useful for understanding ants!

And uncles!

--Gary

Best

Rick

···

---
Richard S. Marken Consulting
marken@mindreadings.com
Home 310 474-0313
Cell 310 729-1400