In 11 ventilator-dependent patients, we undertook a head-to-head comparison of patient-ventilator interaction during four ventilator modes: assist-control ventilation (ACV), intermittent mandatory ventilation (IMV), pressure support (PS), and a combination of IMV and PS. Progressive increases in IMV rate and PS level each decreased inspiratory pressure-time product (PTP) (p < 0.0001). These reductions in PTP were greater with PS than with IMV at lower but proportional levels of maximal assistance (p < 0.005). When PS 10 cm H2O was added to a given level of IMV, greater reductions in PTP were achieved not only during intervening (PS) breaths (p < 0.001), but also during mandatory (volume-assisted) breaths (p < 0.0005); this additional unloading during mandatory breaths was proportional to the decrease in respiratory drive (dP/dt) during intervening breaths (r = 0.67, p < 0.0001). Maximal unloading occurred with ACV, achieving more than a fivefold decrease in PTP compared with unassisted breathing. Decreases in PTP were confined to the post-trigger phase, and PTP of the post-trigger phase correlated with dP/dt (r = 0.78, p < 0.0001). Effort during the trigger phase remained constant despite marked changes in drive and intrinsic positive end-expiratory pressure (PEEPi). Ineffective triggering occurred with all modes, and wasted PTP increased with increasing levels of assistance as a result of the accompanying decrease in drive and increase in volume. Breaths preceding nontriggering efforts had shorter respiratory cycle times (p < 0.0005) and expiratory times (p < 0.0001) and higher PEEPi (p < 0.0001), indicating that neural-mechanical asynchrony resulted from inspiratory activity commencing prematurely before elastic recoil pressure had fallen to a level that could be overcome by a patient's muscular effort. Thus, increases in the level of ventilator assistance produced progressive decreases in inspiratory muscle effort and dyspnea,which were accompanied by increases in the rate of ineffective triggering.