Comments from Nick:
The neuron (/column) that tips the feature from one voronoi cell to another is not itself the vehicle. It's a part of the vehicle. The vehicle is the whole feature vector. And it's states are sorted into vehicle types by the voronoi cells. So this neuron is just the part that makes the difference between whether vehicle A or vehicle B is being tokened. (Analogy: comparing capital F and capital E, it is just the line at the bottom that makes the difference about whether you have the first vehicle or the second.)
I like your example of the sensitive neuron. It illustrates the point that what counts as a vehicle will depend upon how firing affects downstream processing (or not).
Reply from AG: "The vehicle is the whole feature vector." Obviously, I am delighted by this reply. But from my perspective it seems to run counter to an earlier reply you gave during the Tuesday seminar, where you said the fusiform face area (FFA) could be a vehicle (just like many similar examples in the book). In the Voronoi cell way of seeing cortical output, FFA just provides the line at the bottom of the 'E' or 'F', much like the single column in the rotating cylinder case.