Adding a spine calcium signal 2–5 µm away decreased linear test error by 0.34% after field and distant signals in 19 mice

We tested whether a nearby dendritic spine adds information for a hidden calcium response after a distant spine and the recorded field are known. We used the Jenks et al. dataset and a checked SCI extract. The test had 4,455 target spine records and 653,617 presentation rows from 19 mice. Separate repeats gave each spine a mean for each visual direction. All fitted coefficients came from other mice. The fixed linear model used asinh signed amplitudes. Adding a signal 2 to 5 micrometres away along the dendrite decreased mean normalized test error by 0.342%. The 95% mouse resampling range was 0.201% to 0.486%, with every model refit. The addition helped in 17 of 19 mice. A field-only model decreased error by 4.829% from calibration alone; the full model decreased it by 5.200%. The local gain stayed positive with alternating repeats, large-event exclusion, raw amplitudes, and a larger spatial gap. Shifting predictor presentations gave a gain of -0.006%, with a range of -0.112% to 0.026%. Thus, the added information depended on paired presentations under the tested rule. Shared neural input, dendritic activity, and optical mixing remain possible causes. The result does not identify synaptic current or a causal distance effect. We release compact inputs, the response completion panel, source records, code, all fits and resamples, figures, and repeat instructions.

dendritic spine · structure and function · calcium imaging · visual cortex · shared variability · response completion