Non-uniform composition of gas in the lungs may arise because of uneven dilution of alveolar with inspired gas by convection (bulk flow) or because of the finite rates of molecular diffusion in the small air-spaces. Such non-uniformity represents 'mixing inefficiency' which may be expressed by the difference between the effective dead-space ('ventilatory dead-space', V DV) and the actual airway dead-space ('Fowler dead-space', V DF). This expression of mixing inefficiency does not distinguish between convective and diffusive mechanisms. The authors show, from simulated gas wash-out of an asymmetrical, two-compartment model of the lung, that a third dead-space ('wash-out dead-space', VDW) enables this distinction to be made: VDV-VDW reveals convective, and V DW-VDF diffusive, mixing inefficiency. The position of the branch point subtending the two compartments greatly affects the outcome, however. When the branch point is relatively peripheral, inefficiency is mainly diffusive; when it is relatively central, inefficiency is mainly convective. These findings indicate how mixing efficiency may be measured more selectively than at present in real lungs.