Anecdote from Innis (1992)

To quell his anxiety, a group of Tryon's students conducted an Ocean's Eleven style operation to deconstruct and transport a huge rat maze from one building to another while a second group of students distracted him on an out of town trip.

doi.org/10.1037//000...

However, they did find that (dorsal) hippocampus is in involved in this observation > performance enhancement, whatever the underlying mechanism.

Fuentealba et al. (2025) developed a dry land watermaze task where one rat could observe a conspecific solving the task.

Observing a pretrained rat (expert), led to significantly better performance than observing a naive rat (non-expert), suggesting that rats can learn through observation.

Also not sure why largest fields were excluded 👇 field size & local maxima are key to the model

Rate maps in Yartsev & Ulanovsky were quite highly smoothed (adaptive & kernel smoothing) not sure how/if this was accounted for?

Small fields will adopt the shape of a smoothing kernel more readily...

However, Mainali et al.'s model doesn't capture the finding that field size scales with proximity to boundaries:
doi.org/10.1016/j.cu...
Tanni et al. - also not cited

When we recorded place cells in climbing rats we found that cells exhibited fewer & larger place fields. Space was underrepresented with fewer fields per m3.

These results are consistent with 2D rodent results in large spaces and Mainali et al.'s model.

Changing the model thresholds alters the size of the resulting place fields, so this could explain larger fields in larger environments.

Not sure how/why these thresholds would change in the brain? They would need to scale with environment size, the mechanism for that remains unexplained...

Mainali et al.'s model proposes place cell activity is the result of thresholded activity from multiple spatial inputs that, as far as I understand it, have 3D spherical gaussian firing fields 👇

(3) Subramanian et al. observed time 'off' cells, which have not been reported before 👇

Some interesting points:

(1) RSC time cells were often context-dependent - firing differently depending on the animal's past trajectory, such as visiting the east or west arm of a plus-maze 👇

And of course the usage rate of novelty terms is also highly correlated with the impact factor of a journal.

I think we all already know the result 😂 👇

Novelty terms have rapidly proliferated, while confirmatory terms have remained at the same usage.

The reason I love this study is that people in many different scientific fields (I'm in neuroscience) can relate to this finding.

I really liked the task which consisted of a grid ('crossword') maze.

Animals navigate from different start boxes to a single goal location.

Barriers on the maze result in overlapping routes in some segments, or force the mice to take different routes in the same segments.

What about these examples though? They are among the best examples of shortcuts, but they seem very indirect with many reward well checks before reaching A.

What is the chance level here?

Does this look like the result of vector computation?

Also, some of the 'shortcuts' reported were not very convincing.

In the below example the mouse goes from start>B, upon finding no food it then goes B>A using a shortcut.

I think this is a convincing example, but it's undermined by the fact this particular mouse used the A>B route during training