CPR monitors provide feedback on rate, depth and release force (RF) of chest compressions. Excessive RF ('leaning') impedes venous return, reducing blood flow. Available monitors detect leaning with a force sensor, an expensive component. Our objective was to determine whether leaning, like rate and depth, could be detected through the accelerometry signal alone. Brief intervals of accelerometry signals centered on force minima were extracted from chest compressions recorded with CPR monitors used in 289 out-of-hospital cardiac arrest in the Portland metropolitan region from 2009-2015. Evidence for effects of leaning was sought with various neural networks. Testing was done with waveforms extracted from 147 additional cases. A cascadeforward network with 2 hidden layers outperformed simpler alternatives. Testing yielded 88.6% correct classifications. Cases with zero RF were identified correctly as non-leaning in 99.9% of 123714 cases. Accelerometry in the vicinity of the release point provides information about the force at release and warrants further investigation.