I have a certain penchant for (relatively) simple animals, so I found the ant path integration experiments to be particularly interesting and surprisingly similar to models of dead reckoning used in robotics, albeit the use of polarized light seems rather unconventional. Perhaps it offers a better reference for azimuth / elevation or yaw / pitch / roll than the magnetometers, accelerometers, and gyrosopes typically employed in inertial measurement units (IMU) or functions under some other evolutionary constraint? Not that any of the later components are evolutionary impossible; crane flies, for instance, have converted vestigial rear wings into halteres, which function as vibrating structure gyroscopes. (Although, I do wonder, incidentally, exactly how one attaches stilts to an ant?)

When I started discussion by portraying the results of the two papers as in opposition to each other, Brock was careful to insist that there was more subtlety, and that it was more the interpretations of the data that were somewhat in conflict. I think this is a fairly important point at the meta level of how science is done. In the case of both papers, there are assumptions (both explicit and implicit) that are made to form conclusions from the data. In a case such as this with conflicting data, it's probably more prudent to ask how the data from both studies may be combined to form a possible combined explanation, rather than trying to label one as "right" or "wrong".

One thing I really took away from Brock’s lecture and discussion in section is that path integration is defined behaviorally, and that a variety of cognitive processes can enable successful path integration. It was interesting to see how tasks conditions for the ants were varied to determine that they were counting steps and using polarized light to calculate their bearing. We’ve seen human experiments on path integration which vary the environmental and internal cues available to participants. I’d be interested to see what other sorts of experiments have been done or could be done to better tease apart the cognitive processes used by humans to path integrate, especially if there are multiple mechanisms involved.

One of the most interesting parts of Brock's lecture was the section dealing with path integration in ants. I was impressed by the kind of techniques that researcers use to figure out that ants use polirized lights and that they count their steps when they path integrate. I also liked how Brock compared and contrasted the methods used for path integration in ants and humans. While ants count their steps, humans form a cognitive map by using idiometric, vestibular and visual information.
I also thougth that we had a good discussion about the limitations imposed by fMRI (and other neuroimaging techniques) in studies dealing with path integration.

I seem to remember that the graphic showing subjects' pointing directions was not so much highly variable as multi-modal. If I remember the figure correctly, a proportion of the subjects were nearly correct, while another bunch were off by around 180 degrees. If looked at naively, a bimodal distribution will appear to have too-high variance, and so I worry that this conclusion may not be fully supported by the data.
I would also like to register the concerns I mentioned in class: I don't see that the task in which patients did virtual pointing-based-triangle-completion in fact _would_ make much use of hippocampal allocentric-oriented spatial cognition. One of the major computational features of hippocampus is for it to represent "everything" in the world, not just the origin. A way to test this feature of hippocampus would be if subjects were asked to point to unpredictable locations in the environment, something which working memory would likely be less capable of supporting.

I find it interesting that the two papers base their findings on very different premises with very different research designs. One uses visual input only and measures blood flow to the hippocampus and medial prefrontal cortex. The other removes all visual input, relies exclusively on self-propelled motion, and looks at behavioral performance in the absence of the hippocampus and, in some cases, entorhinal cortex.
As we discussed, the efficacy of the fMRI study requires that a virtual reality experience with only visual input sufficiently engages brain functions responsible for path integration. The first sentence in the abstract describes path integration as “the ability to sense self-motion for keeping track of changes in orientation and position.” The authors don’t mention any limits to an exclusively visual experience. A quick Internet search yielded this brief chapter (http://www.uni-tuebingen.de/cog/literature/literatur-Dateien/2001/GillnerMallot01.pdf). The authors discuss some disparities between VR and real world findings (p. 16213). Interestingly, like our comparison of the two papers, they compare results from a VR study with a blind-folded self-motion study.
Perhaps more work has been done to identify the limitations of VR studies since Gillner and Mallot wrote this piece in 2001. Until we have a portable method to localize brain activity, we may have to accept the limitations of virtual reality-driven fMRI. However, a forthright acknowledgement of the potential limits to the conclusions would square better with our understanding of cognition drawing upon all available and potentially relevant information.