Figure 3.

Recognition of Stimuli

(A) Salient features are associated with locations in the visual field via grid cells, and each location (red crosses in black circles) is encoded by the phases across the entire grid cell ensemble (4 out of 9 scales shown) as a population vector (i.e., a given pixel value across all grid cell rate maps).

(B) Saccade sequences (red arrows) superimposed on face stimuli (left). Cyan circles indicate the centers of all encoded local features. With each sampled feature, a firing rate of the corresponding identity cell is incremented (right). The dashed line indicates the decision threshold. Note, in panel 3, an initially wrong hypothesis (black arrows) is overtaken by the correct one.

(C) Saccades superimposed on scene stimuli. In panel 2, an initially wrong hypothesis yields misdirected saccades (relative to the true stimulus), which leads to an early reset (black arrow) because the predicted and actual outcome of the feature discrimination differ persistently. Starting from different initial features eventually leads to recognition.

(D) Recognition of object stimuli.

(E) Histogram of the number of saccades necessary for recognition across all stimuli.

(F) Variance of the activity across grid cells along each saccade vector (red bars) or just at the start and end locations (dark gray bars), binned according to the direction of the saccade (10-degree bins). 6-fold symmetry akin to fMRI data arises from the underlying symmetry of grid cells.

(G) All saccade vectors used for the analysis in (F). Image credit: Mr. Spock, Amelia Earhart, Dido Building Carthage: public domain images; Nelson Mandela, Schloss Charlottenburg, Brain, Perseus: Creative Commons Attribution.