In one ctenophore count model, nonnegative source covariance gives sharp lower input correlation bounds of 0 for Q1–Q2 and 0.9856 for Q3–Q4

We tested which input correlation bounds follow from one ctenophore statocyst count map when source variances are free. Each aboral nerve-net neuron (ANN) supplies one shared signal. Quadrant inputs are linear sums with count weights. Effective source covariances are nonnegative, and input variances are positive. The full released graph has sharp lower bounds of zero for Q1-Q2 and 0.9855764540 for Q3-Q4. One bridge input reaches only Q1. Its free variance removes the positive Q1-Q2 bound of 0.9983967212 from the ANN-only model. Connector records confirm that link. With independent sources, correlation at most 0.5 requires bridge variance at least 300 times the sum of the two ANN variances. This ratio acts before count weights. Fixed-support weight errors of at most 20% leave a Q3-Q4 bound of 0.854166. We also give sharp two-source covariance limits, independent target residual requirements, and cutoff examples that change the bounds from positive to zero or one. The basic angle inequality is classical. The new contribution is its exact application to the complete released quadrant inputs, with explicit covariance examples and reusable tools. Exact rational checks, scalar searches, source hash checks, and a clean offline repeat passed. These are conditional model results from one map. They do not measure cilia correlation or establish a neural mechanism. The package preserves source tables, provenance, proofs, code, results, figures, and repeat instructions.

ctenophore · statocyst · connectome · structure and function · input correlation · source covariance · model bounds · count cutoffs