STATE cells in the place model, with FS × HD idiothetic input, encode weak HD information, and only with distal landmarks visible. We present results of simulating the place cell model shown in Figure 1c with the self-motion action cells (designated ACT in Fig. 1a) comprised of two subpopulations of cells representing either FS or HD. The model is trained with the agent moving (i.e., translating and rotating) within the environment in the light with a mixture of proximal and distal visual landmarks present. After training, the model is tested by rotating the agent through a full circle in clockwise and anticlockwise directions at each location in the (discretized) environment, and then averaging the responses of individual STATE cells over HDs at each location. The figure shows the tuning curves of several randomly selected STATE cells during testing (left: with both proximal and distal landmarks; right: with only proximal landmarks). We see that some STATE cells in the “place” model do exhibit some directional selectivity, but that this is comparatively weak. In the proximal-only case, we show the only four STATE cells (out of a population of 1000) found to have HD selectivity, again demonstrating the importance of distal landmarks. Some directional selectivity of place cells is in line with intuition (as the sensory input certainly contains directional information, which is transiently correlated with the agent’s rotation, even in the proximal-only case), and with experimental evidence (Muller et al. 1994). Note that, as in Figure 4, the discontinuity in the upper-left subplot is a sampling artifact.