In 19 mice, sums of two spine calcium signals 2–5 µm apart on the same branch lost 36% of tuning energy with asinh scaling and independent repeat errors

We tested how much measured visual tuning an equal sum of two nearby spine calcium signals can retain. We used 6,524 near pairs at 2 to less than 5 micrometres and 8,401 far pairs at 10 to less than 20 micrometres. Pairs had no intervening branch point. The cohort had 4,535 spine records in 282 fields from 22 neurons and 19 mice. Primary responses used asinh of signed amplitudes. Separate repeat products estimated tuning energy in the sum and difference. Stable means and independent errors between repeat halves are required. The equal sum lost 36.0% of pooled near-pair direction tuning energy, with a 95% mouse bootstrap interval of 30.8% to 47.5%. Far-pair loss was 37.3% [30.7%, 47.3%]. Near loss minus far loss was -1.26 percentage points [-4.68, +3.35]. The main test gave no clear local advantage. Removing large shaft-event trials changed the cohort. An added matched check in 18 mice raised near loss from 42.6% to 46.7%, a paired change of 4.01 points [0.74, 6.40]. This energy metric measures a fixed linear projection in optical signal units. It does not measure information in bits, synaptic current or dendritic output. Shared processing and data-dependent trial masks can violate repeat independence. The package preserves code, compact inputs, source hashes, all results and repeat steps. All 305 fields passed a source audit. A fresh run reproduced nine numerical files byte for byte.

dendritic spines · structure and function · visual cortex · calcium imaging · signal pooling · tuning energy · independent repeats