In Niska Banja, Serbia, which is a high-radon area, a comparison was made between two retrospective radon measuring methods and contemporary radon measurements. The two retrospective methods derive the radon concentrations that occurred in dwellings over longer periods in the past, based on the amount of trapped (210)Po on the surface of glass objects (surface traps, ST) or in the bulk of porous materials (volume traps, VT). Both surface implanted (210)Po in glass objects and contemporary radon in air were measured in 46 rooms, distributed in 32 houses of this radon spa-town, using a dual alpha track detector configuration (CR-39 and LR115) and CR-39 track etched detectors, respectively. In addition to the use of surface trap measurements, in 18 rooms (distributed in 15 houses) VT samples of suitable material were also collected, allowing to compare ST and VT retrospective radon concentration estimates. For each room, contemporary annual radon concentrations (CONT) were measured or estimated using seasonal correction factors. The distribution of the radon concentration in all data sets was found to be close to lognormal (Chi-square test>0.05). Geometric means (GM) are similar, ranging from 1040 to 1380 Bq m(-3), whereas geometric standard deviations (GSD) for both the retrospective methods are greater than for the CONT method, showing reasonable agreement between VT, ST and CONT measurements. A regression analysis, with respect to the lognormal distribution of each data set, shows that for VT-ST the correlation coefficient r is 0.85, for VT-CONT r is 0.82 and for ST-CONT r is 0.73. Comparison of retrospective and contemporary radon concentrations with regard to supposed long-term indoor radon changes further supports the principal agreement between the retrospective and conventional methods.