[...]
> Have you thought about [eventually]
> simulating a "maze" through which
> your simulated rat would have to
> orient to get to the 'food'?
Well sure. But in order to do navigation, the animal has
to locate itself in the environment, otherwise there can
be no path planning. OTOH the internal representation of
space is generated during movement and likely is influenced
by the way the animal moves through space. So there is something
like a hen-and-egg problem here. Anyway, I reiterate: if a model
only covers a part of the real world's complexity and still
explains (and predicts!) a lot about the complex system, it's a good
model. There's a number of models out there that try to cover as much
detail as possible, but end up too complex to understand anything.
Anyway, another group is going to model goal-driven behavior via
reinforcement learning on top of my results, but that's another story.
> Absent sensory cues, Rats do sniff-
> look-and-move in ways that tend to
> 'cover' the environment.
Well, olfactory and visual cues qualify as sensory, right? Unless you
only consider a big cue card as a sensory cue. But even without such a
cue card (and without removing the self-placed olfactory cues on the
apparatus floor) spatial representations usually don't drift. This can
only be the case if there's sensory cues, as any kind of pure path
integration /dead reckoning system is bound to accumulate errors.
You can observe goal driven movement in rats in a maze (they know
water's over there, they are thirsty, so they go there), but also
seemingly random exploratory behavior.
A good start for more info here is the great reviews in [2] about 100s
of place field experiments from the past 30 years and a structured
approach to navigational strategies. Even if your understanding of
hippocampus is somewhat unorthodox, you should consider reading these
results (assuming you haven't ;).
> But imposing a "T-maze" condition,
> for instance, will result in a 'new'
> type of data showing-up in your trials.
To some degree T-mazes and radial mazes can be modelled as a combination
of linear tracks and possibly a central open field. But definetely, the
movement patterns in different behavioral tasks do differ, and my model
predicts some specific changes to the hippocampal representation of
space and possibly non-spatial variables as well.
All the more reason for me to search for the real trajectories. But
again, in order to reduce complexity, random exporatory behavior is the
first thing to look at, since non-spatial factors are minimal then.
> [To others: Here's a Devoted Student.
> Why is he left without the information
> that'd take-advantage of his Willing-
> ness to Work?]
Thanks for the advertisement :)
It was indeed my understanding that there's many place field
experimenters who might appreciate a model that makes testable
predictions :)
Mathias
[...]
Ref [2]:
@Book{Redish1999,
author = "A. D. Redish",
title = "Beyond the Cognitive Map - From Place Cells to
Episodic Memory",
year = "1999",
publisher = "MIT Press",
}
