Surprisingly, upon repeated stimulation with the same colony odor, spatial activity patterns were variable, and as variable as activity patterns elicited by different colony odors.
http://plosone.org/article/info:doi/10.1371/journal.pone.0021383#article1.front1.article-meta1.abstract1.sec2.p1

This finding is not very surprising, because sensory networks and sensory maps (as in AL) show an extremely high variance in vivo, simply because time has passed between two subsequent activity patterns (i.e., because of different spatial or temporal environments, different and changed and ever-adapting self-referential neuronal activities in the AL, etc.).

In contrast, motor networks of extremely trained humans (while learning football, drawing, calibrated science...) may display a very restricted variance => hence, more stable maps.

It would be interesting to study the difference in variance between sensory and motor maps -- it is not surprising, that neuro-prosthetics (e.g., Nicolelis et al.) works with premotor and motor maps, which show much less variance and hence result in more precise motor performance.

But cf. also http://www.nld.ds.mpg.de/downloads/publications/memmesheimer2006.pdf