Results are presented from a comprehensive analysis of the ability of CIN depth, relative to CIN and CAPE, to discriminate environments that support deep convection initiation (DCI) from those that do not. Evaluation is based on environments of DCI observed over the central US by the ThOR algorithm compared to nearby Null environments. A total of 62846 DCI/Null point pairs spanning a little over 7 years are used. CIN depth is calculated assuming three different parcel types – surface-based, mixed-layer, and most unstable – and two different definitions of depth – the distance between the LCL and LFC and the distance between the initial parcel height and the LFC. MLCD, the mixed-layer CIN depth, proved to be better than all CIN parameters in discriminating environments that support DCI from those that do not. For all parcel types, CD ORIG , the distance between the initial parcel height and the LFC, is better than CD LCL , the distance between the LCL and LFC, and both are better than CIN. CIN depth is also as good if not better than CAPE, the best discriminator based on prior work, at discriminating DCI from Null environments. When CIN differs very little between DCI and Null environments and CIN is small but generally non-zero, MLCD ORIG distinguishes DCI environments the best across the CIN depth varieties. For elevated DCI, CIN depth is generally less useful than CIN or CAPE at discriminating DCI from Null environments except when both DCI and Null environments have near zero CIN.