The model makes some clear predictions about this process. First, it depends critically on the existence of stripe cells. Cells in intermediate and deep layers of the MEC, as well as the parasubiculum, Everolimus mouse may occasionally fire more strongly along one of the grid axes than the two others, but nonperiodic band activity has not been

reported in any of these regions so far. Furthermore, the model makes the clear prediction that a certain amount of spatial experience is necessary before grid patterns can be expressed. This suggestion is supported by simulations showing that inputs from stripe cells with randomized angular separations can generate stable hexagonal grid patterns after a few hours of exploration time. This is not incompatible with experimental data, as stable regular grid patterns only appear several days after developing animals start exploring spaces outside the nest (Langston et al., 2010 and Wills et al., 2010); however, the limited data that exist suggest that grid formation is more dependent on the maturational stage of

the MEC than the amount of experience (Wills et al., Docetaxel datasheet 2010). Finally, if stripe cells are identified in the future, it would be important to examine during development what happens to cells with nonpreferred orientations that lose the competition. Are these cells retuned to one of the three predominant orientations, or do they die out? Does the brain retain stripe cells that do not project to grid cells? If so, what would be their function? Whereas nearly all models for grid cells are based on path-integration mechanisms, one model

stands out by suggesting that the formation of grid fields occurs with spatial rather than velocity-related inputs (Kropff and Treves, 2008). In this model, grid fields are formed by Hebbian self-organization in a TCL competitive network, much like grid cells are suggested to emerge from stripe cells in the self-organized learning model of Mhatre et al. (2010). Neurons must include the crucial ingredient of an adaptation or fatigue dynamics, which makes the spacing of the resulting grid fields scale roughly like the average running speed multiplied by the time constant for adaptation. Although not explicitly evaluated in the model, the grid pattern could also be obtained with other kinds of temporal modulation of spike activity, such as changes in the time constants of spike repolarization, which are known to differ between dorsal and ventral MEC (Boehlen et al., 2010 and Navratilova et al., 2011). A crucial prediction is a correlation between running speed and grid spacing, which is contrary to the apparent constancy of the grid scale when rats run at variable speed in an open field (Hafting et al., 2005). However, a systematic test of this relationship has not yet been made.