Rationale: No systematic investigation into dyspnea in patients receiving prolonged ventilation (>21 d) after recovering from critical illness has been published. Objectives: We sought to determine the magnitude, nature, and pathophysiological basis of dyspnea during an unassisted-breathing trial in patients receiving prolonged ventilation. Methods: Dyspnea intensity and descriptor selection were investigated in 27 patients receiving prolonged ventilation during a 60-minute unassisted-breathing trial. Pressure-time product, respiratory mechanics, and PtcCO2 were also measured. Measurements and Main Results: Of 10 patients who reported dyspnea during assist-control ventilation, 9 (90.0%) selected "Not getting enough air" to characterize dyspnea. Vt setting was lower in dyspneic than in nondyspneic patients (480.0 vs. 559.4 ml), P < 0.046. During the unassisted-breathing trial (n = 26), patients developed increases in dyspnea (P < 0.01) and PtcCO2 (P < 0.01) but no change in [Formula: see text]e. Dyspnea score was strongly linked to PtcCO2 (P < 0.012) and airway resistance (P < 0.013) but not respiratory work (although pressure-time product was almost three times higher than normal). At 60 minutes into the trial, 83.3% of patients selected "Not getting enough air" on its own or in combination with "Too much effort" to describe discomfort, whereas only 16.7% selected "Too much effort" on its own (P < 0.001). Across the dyspnea spectrum, patients chose "Not getting enough air" overwhelmingly over other descriptor options (P < 0.001). Conclusions: Patients developed increases in dyspnea and PtcCO2 but unchanged [Formula: see text]e and work of breathing during an unassisted-breathing trial; patients selected air-hunger descriptors overwhelmingly over excessive effort. The observations support the belief that air hunger results from heightened respiratory center stimulation combined with the incapacity to increase [Formula: see text]e.