Fig. 5: Donut-like inhibitory motif is required for categorical signaling of the strongest stimulus by the barn owl midbrain selection network.

A Network model schematizing the OT-Imc-Ipc circuit with donut-like inhibition in the barn owl midbrain. “Rec”: Recording at OTid neuron 1. B Network model schematizing the desired experimental manipulation of disrupting donut-like inhibition (i.e., filling-in the donut-hole): introduction of self-inhibition (brown connections) in the direct pathway between Imc and OTid, as well as in the powerful indirect pathway through Ipc. However, this is infeasible experimentally (Figs. 3 and 4; text). C Network model schematizing the proposed equivalent experimental manipulation: focal inactivation of Ipc neuron in channel 1, mimicking the effect of introducing self-inhibition onto Ipc neuron 1, while recording from aligned OTid neuron 1. “Iont”: Iontophoresis of kynurenic acid (magenta blob) onto Ipc neuron; “Rec”: Recording OTid neuron. D Modeling results showing equivalence between desired (B) and proposed (C) manipulations. Plots of CatI computed from response profiles simulated from models in (A–C); conventions as in Fig. 2D; n = 50 model neurons; center lines in the violin plots indicate median values. Manipulations in both (B) and (C) cause abolishment of categorization. See also Fig. S4A–C showing, additionally, that (C) is not equivalent to removal of recurrent amplification. (Note: model in (A) is identical to model in Fig. 2A, right column-bottom; model in (B) is identical to model in 2B, middle column-third from top). E Paired recordings in OTid and Ipc such that RF of OTid neuron overlaps with that of Ipc neuron. Conventions as in Fig. 3C. Stimulus protocol used is the same two-stimulus morphing protocol used in model simulations in Fig. 2B–D; the relative strength between S1 and S2 was systematically varied. F Left: OTid response rasters in the Ipc-intact condition. Right: OTid response rasters in the Ipc-off condition. Distance between OTid and Ipc RF centers = 5°. G OTid response firing rates, computed from D over the 100–400 ms time window. mean ± s.e.m (n = 12). Black: Ipc-intact condition, magenta: Ipc-off condition. Dashed lines: best fitting sigmoid or straight line to data, chosen based on AIC criterion. Black: AIC (sigmoid) = 29.07, AIC (line) = 40.92; magenta: AIC (sigmoid) = 37.75, AIC (line) = 34.67. H Population summary of effect of Ipc inactivation on CatI; n = 11 neuron pairs; data in orange are from example neuron pair in (F, G). Top panel: measured values. Bottom panel: Data in top panel replotted after normalizing to Ipc-intact values. Diamond: average of magenta data; *p = 2.55e-5, two-sided ranksum test against 1 with HBMC correction. See also Fig. S4. Source data are provided as a Source Data file.