Laboratory density currents generated with unsteady source conditions provide insight into the distances and timescales over which unsteadiness at the vent should persist in natural dilute pyroclastic density currents (PDCs). The laboratory experiments comprise heated 20-m talc particles turbulently suspended in air and introduced as density currents into an 8.5x6x2.6m air-filled chamber. The densimetric and thermal Richardson, Froude, Stokes, and settling numbers are all similar to those of natural dilute PDCs. Although the experiments' Reynolds numbers are several orders of magnitude less than natural PDCs, the experiments are fully turbulent and are thus dynamically similar to some dilute natural PDCs. Unsteadiness in the experiments is generated by adding two pauses of duration t to the eruption (e.g., a 100s experiment comprising three similar to 33s pulses separated by two pauses of t=10s). Propagation distance of the leading head is proportional to the square root of time; positions of the trailing pulses have similar time dependence, but they generally travel slightly faster than, and thus catch and merge with the leading current. Trailing pulses are more easily distinguished from the body of the preceding current when t is large. Analysis of turbulent structures through space and time shows that unsteadiness at the eruption source can be distinguished from turbulent fluctuations in the currents when t is greater than the integral turbulent timescale of the current body (body), a statistical measure of the characteristic timescale of unsteadiness within a turbulent flow. The distances over which unsteadiness in the transport system persist scale with t/(body) and the ratio of the thermal Richardson numbers of the trailing and leading pulses. As t/(body) increases from similar to 2 to similar to 4, trailing pulses remain distinct over an increasing fraction of the lead runout distance. When subsequent pulses have higher Ri(T) than the leading current, they coalesce with the leading pulse at more proximal distances, but when Ri(T) is much lower, they remain distinct over distances similar to that of the leading head even when t is similar to (body). For currents that have depositional timescales comparable to or shorter than (body), deposits should be expected to preserve a record of unsteadiness over the distances that unsteady pulses persist.