Measurement of systolic time intervals (STI) provides a noninvasive assessment of cardiac function in resting subjects. However, large motion artifacts often limit their application during exercise. To improve such measurements, we employed a new lightweight carotid arterial pulse transducer and minimized the artifacts by computerized signal averaging. The carotid pulse was recorded with an inductive plethysmographic transducer band (CIP) wrapped around the neck. STI derived from these measurements were compared to standard measurements from a funnel-shaped cup attached to the neck and connected to a pressure transducer (CUP) in 12 normal subjects at rest and with graded bicycle ergometry. To compare the techniques, CIP and CUP signals, together with the EKG and phonocardiogram, were connected separately to two microcomputer systems. The systems were triggered from the EKG and averaged the signals over 16 or 32 heart beats. Semi-automatic analysis of the averaged signals provided estimates of left ventricular ejection time (LVET), pre-ejection period (PEP) and electromechanical systole (QS2). Motion artifact levels of the unprocessed signals were similar with both CIP and CUP methods but were greatly reduced by signal averaging. All LVET values using CIP fell within 10% of CUP values. 85% of PEP values using CIP were within 20% of CUP values. Increasing exercise loads produced appropriate decreases in QS2, LVET, PEP and PEP/LVET consistent with increased myocardial contractility. The CIP proved comfortable to wear and did not require critical positioning as did the CUP. Measurement of STI complements exercise pulmonary testing with useful information of cardiac function.