A soma position correction improved agreement for opposite motion but left window sensitivity in 178 cells from one mouse retina

We tested whether moving-bar arrival time at each soma explains a timing-sensitive test of retinal anatomy. The sample had 178 ganglion cells in five fields from one Eyewire II mouse retina. We calculated arrival offsets from recorded position and the released speed of 1000 micrometres per second. No time or spatial parameter was fitted. The maximum absolute offset was 59.1 ms. The predicted correction gave more agreement between opposite motion waveforms than a reversed-sign control in all five fields. The mean Fisher correlation exceeded a reversed-sign control by 0.12766; none of 999 within-field position shuffles reached this difference (reference p = 0.001). The result remained after 56 source direction-selective calls were excluded. The primary test asked if the correction reduces polarity changes when sampling windows move 128 ms later. The equal-field mean absolute change fell from 0.25097 to 0.24610. The reduction was 0.00487, with a descriptive field resample range of -0.00025 to 0.00917. About 98% of the cell-weighted sensitivity remained. A depth model trained on 315 cells in a separate retina stayed fixed. With the correction, its error decrease against the training mean was -5.85% at the original window and 24.42% at the later window. Direct source interpolation gave the same interpretation. The correction assumes that receptive-field centres are at somata and that the stimulus is centred on each field. These results support a position contribution to timing but do not identify a causal mechanism or the cause of the remaining window effect. Code, provenance, aggregate results, figures, controls and repeat instructions are preserved. Original source arrays remain external under their own terms.

retina · structure and function · stimulus timing · dendrite depth · calcium imaging · Eyewire II · model validation