The accuracy with which a multiple-electrode impedance catheter (IC) tracks instantaneous global, in-situ left ventricular (LV) volumes was tested in 13 anesthetized dogs scanned in the Dynamic Spatial Reconstructor (DSR), a fast volumetric computed tomographic (CT) scanner. All dogs were scanned during control conditions and during an acute hemodynamic intervention. Hypertonic saline calibrations were performed for the IC prior to each DSR scan. In six of the dogs the IC-derived LV end-diastolic volume (Y) correlated with the DSR-derived global LV end-diastolic volume (X) as follows: end-diastolic volume, Y = 1.01X - 9.9, r = 0.812. The IC-derived LV end-diastolic volume, under control conditions, correlated with the DSR-derived truncated (i.e., that region of the LV chamber between the proximal and distal electrodes of the IC catheter) LV end-diastolic volume, Y = 1.00X + 17.4, r = 0.803. Under reduced preload the relation was Y = 1.3X - 15.26, r = 0.911. The segmental volume (between adjacent sensing electrodes on the IC) at the basal portion of the LV correlated poorly (Y = 1.88X + 3.3, r = 0.459 etc.), but correlated better at mid- and more apical LV levels (Y = 0.97X + 2.7, r = 0.762). Correlations between segmental stroke volumes were similar at basal (Y = 1.31X + 1.60, r = 0.815) and mid- and more apical levels (Y = 1.42X + 0.11, r = 0.763). Stroke volume during acute ischemia (two dogs) was Y = 1.33X - 1.41, r = 0.717; during acutely decreased preload (four dogs) it was Y = 1.24X - 2.88, r = 0.572). Thus, the IC tracks the changes in LV-chamber volume throughout a cardiac cycle quite well under a variety of conditions, but accuracy deteriorates as the shape of the LV chamber changes in response to changes in hemodynamic loading or local myocardial ischemia.