Since dyspnoeic sensation (d) increases progressively with work rate (WR) and the duration of a volitional breath-hold (t BH ) shortens, we wished to explore whether t BH might correlate suciently closely with d to provide a quantitative and descriptor-free index of respiratory sensation during dynamic exercise. Nine healthy males exercised on a cycle ergometer at a series of constant WRs, above and below the lactate threshold. Ventilatory and gas exchange variables were measured breath-by-breath. At each WR, breath-holds to the limit of tolerance were taken; d was recorded (visual-analog scale) immediately prior to and throughout each breath-hold. During breath-holds, d increased with time as a``break-away'' monoexponential characteristic, reaching the maximum (100%) at the breakpoint. Despite end-tidal partial pressure of carbon dioxide at the break-point being higher and end-tidal partial pressure of oxygen being lower with increasing WR, the relationship between WR and t BH declined curvilinearly (i.e. with large falls in t BH occurring in the low WR range, but far smaller reductions at higher WRs). The t BH /minute ventilation relationship had a similar form. The relationship between pre-breath-hold d and t BH was also complex: the large reductions in t BH in the low WR range were associated with only modest increases in pre-BH d while, at higher WRs, the progressively smaller decrements in t BH were associated with progressively larger increases in d. We therefore conclude that breath-hold duration is unlikely to provide a useful correlate of exertional dyspnoea during dynamic exercise. Furthermore, the relative prolongation of t BH at high WRs (accounting for the moreextreme levels of end-tidal gas tensions) may re¯ect the attention-diverting in¯uence of the exercise per se.