A 20 micrometre control reduced shared HVC input excess from 88% to 1.3% for MSN pairs with a shared GPi target in one Area X map

We tested whether MSN neurons that contact the same GPi neuron share an excess of HVC input in one adult zebra finch Area X map. The selected graph had 8,096 MSNs, 43 GPi cells, 893,711 HVC-to-MSN connections and 29,478 MSN-to-GPi connections. We compared mean binary input cosine with complete input profile shuffles within position and input count groups. An exact formula gave each reference mean. The observed cosine was 0.0030537. Excess was 88.23% with input count groups only, 5.62% in 40 micrometre cubes and 1.31% in 20 micrometre cubes. The primary 40 micrometre test gave p=0.001 from 999 draws. The 20 micrometre control removed 97.24% of the absolute excess above the global mean. A later 10 micrometre check gave 0.25% excess, but only 28% of profiles could exchange with another cell. A linear model linked input cosine to the sum of GPi input variances. It assumed independent unit-variance HVC fragment signals, normalized linear MSN inputs and unit gain per MSN-to-GPi link. Variance excess was 50.07% above the global reference and 0.93% above the 20 micrometre reference. These are conditional graph and model results in one automated map. They are not activity measurements or estimates of variation between birds. The package has aggregate inputs, source hashes, an exact mean proof, code, all results, figures and repeat steps.

connectome · songbird · Area X · HVC · medium spiny neuron · pallidal convergence · shared input · spatial control · linear input variance