A laboratory-bench scale method for measuring the sensitivity of smoke detectors is desirable. To be of value, the laboratory test must be capable of predicting the response of a smoke detector to real fires. This Note discusses how far this requirement can be met with a smoke-tunnel (0.63 m^3 volume) of simple construction. Experiments are described which simulated, in the tunnel, an important phenomenon observed in full-scale fire tests, namely, the dependence of the response of an ionisation-chamber smoke detector on the age of the smoke. Some suggestions for future work directed towards improving the tunnel are made.