This paper presents a meta-analysis of 689 luminescence age estimates of Australian desert sand dunes. This analysis had two aims: to examine the hypothesis that Quaternary climate changes have forced dune accumulation, and to understand longitudinal dune behaviour from the age-architecture of sand dunes for which stratigraphic information is often absent or poorly known. A novel approach to the analysis of probability density curves of dune age frequency shows that in the Mallee dunefield of southeastern Australia only the period leading up to and including the last glacial maximum (LGM) (18–34 ka) experienced marked dune growth which cannot be explained by random stochastic activity. During the LGM, most, but not all, Mallee longitudinal dunes were accumulating more rapidly than before and after. In the Strzelecki dunefield source-bordering (SBD) longitudinal dune TL (n = 61) and OSL (n = 40) ages are more frequent at the Pleistocene–Holocene transition but nearby non-SBD longitudinal dunes (n = 21 ages) were only active above background levels during the LGM. However, the Strzelecki dune age data set suffers from being drawn from a small number of individual dunes (10 SBD dated using TL; 12 SBD dated by OSL; 7 non-SBD dated by OSL) with a marked bias away from surficial (young) and basal (old) depths. Even during the LGM, many dunes do not record active growth and the dunefield was most likely a mosaic of stable dune surfaces and bare mobile patches, supported by the scant pollen data from the arid zone. Hyper-aridity has not been experienced in these dunefields. These vegetated dunes exhibit conservative behaviour with limited response to the range of climate forcing experienced in the late Pleistocene. A more complete understanding of dunefield response to Quaternary climates would require an intensive and structured dating program to overcome the ‘noise’ from stochastic processes leading to small-scale activity in these dunes and the simultaneous removal of parts of the sedimentary record.