We have developed an efficient procedure for the chemical separation of 226Ra from environmental samples and its measurement by thermal ionization mass spectrometry. The original step in this procedure consists in a pre-concentration by co-precipitation of radium onto manganese dioxide. This pre-concentration step permits an efficient separation of radium from major elements including, most importantly, calcium. Thus, the present protocol provides a very useful technique for the precise and accurate analysis of radium-226 in carbonate samples. Furthermore, this procedure is also well-suited for seawater samples, since it does not require a preliminary evaporation stage, thereby eliminating problems due to the precipitation of large quantities of salts and permits the analysis of femtogram levels of 226Ra. Reproducibility and accuracy of 226Ra measurements were estimated by replicate analysis of a reef coral skeleton and seawater samples. The external reproducibility obtained on the carbonate sample was ±0.55% at the 95% confidence level (n = 9) using between 300 to 600 fg of 226Ra. The chemical yield averages 80% for carbonate matrix. The accuracy of the measurements of 226Ra in the reef coral skeleton (dated by the 230Th–234U method) were assessed by comparing the measured 226Ra with the expected 226Ra activity assuming a closed system behavior for U-series. For seawater samples, the external reproducibility was 2.3% (2σ, n = 5) using less than 10 fg of 226Ra (i.e., ca. 200 ml of seawater). The ability to extract with a high efficiency radium from carbonate and seawater samples allows wider investigations of the use of 226Ra either as a Holocene chronometer for carbonates or as a tracer of oceanic processes.