In one Area X v8 export, keeping distinct partners raised GPi preference excess from 1.58 to 5.88 percentage points under 20 micrometre count controls

We tested whether spiny neuron outputs favor GPi targets after contact counts and location are controlled. We used one adult zebra finch Area X export, j0251 v8. The selected graph had 104,888 contacts in 46,061 source-target connections. Of 7,898 spiny neurons, 28.995% sent at least 90% of their pallidal contacts to GPi. A reference that kept parallel contacts, their counts and 20 micrometre centroid groups gave a mean of 27.418%. It permitted repeated source-target pairs. These reduced distinct partner counts by 13.98% of blocks. A swap reference that kept distinct partners gave a mean of 23.111%. Thus the excess rose from 1.58 to 5.88 percentage points. A mixed move check gave 23.115%. With at least ten contacts, the excess was 5.25 points. Two informative certainty filters also kept a positive excess; a cutoff of 0.8 at both endpoints removed every GPe target. Under equal inhibitory gain per contact and independent source changes with unit variance, the observed input balance variance ratio was 0.052659. It was 22.2% above the local distinct-partner mean, but close to the mean when repeats were allowed. No activity was measured here. One partial volume, automated labels, missing membrane contact constraints and unproved coverage of all valid swap states limit interpretation. The package preserves code, source hashes, compact counts, all results, figures and repeat steps.

connectome · Area X · parallel synapses · target preference · graph reference · input variance