Here, we present a new technique for the direct measurement of 44Ca/40Ca isotope ratios on a Multicollector Inductively Coupled Plasma Mass Spectrometer (MC-ICP-MS, AXIOM) using the "cool plasma" technique. By reducing the plasma energy to about 400 W, the isobaric effect resulting from 40Ar+ can be significantly reduced, enabling the simultaneous and precise measurement of 44Ca and 40Ca beam intensities in different Faraday cups. In contrast to the TIMS technique requiring a 43Ca/48Ca double spike, the isotope measurements on MC-ICP-MS can be performed by bracketing standards. We express the calcium isotope variation relative to NIST SRM 915a (δ44/40Ca [‰]=[((44Ca/40Ca)sample/ 44Ca/40Ca)NISTSRM915a)- 1]*1000). Isobaric effects of 24Mg16O+ and 23Na16OH+ interfering with 40Ca and 26Mg16OH2+ with 44Ca can be neglected by measuring calcium isotopes near the low-mass edge of the peaks. No influence of 87Sr2+ monitored on 43.5 atomic mass units (amu) was found. Repeated measurements of two Johnson Matthey CaCO3 standards (lot No. 4064 and lot No. 9912) revealed values of about - 11.29 (‰ SRM 915a) and 0.57 (‰ SRM 915a). These values are in accordance with previous values published by Russell et al. [Geochim. Cosmochim. Acta 42 (1978) 1075], Heuser et al. [Int. J. Mass Spectrom. 220 (2002) 385], Hippler et al. [Geostand. Newsl. 27 (2003) 267] and Schmitt et al. [Geochim. Cosmochim. Acta 67 (2003) 2607]. Repeated measurement of the NIST SRM 915a CaCO3 standard showed that the variance of a single δ44/40Ca measurement is about 0.14 ‰ RSD being comparable with TIMS. MC-ICP-MS-based δ44/40Ca values measured on inorganically precipitated aragonite samples are indistinguishable from earlier measurements based on TIMS, confirming the positive correlation of δ44/40Ca and temperature. MC-ICP-MS-based δ44/40Ca measurements on cultured Orbulina universa showed a slope of about 0.026 ‰/°C being similar to the TIMS-based δ44/40Ca measurements showing a slope of about 0.019 ‰/°C. The large offset of about 5 ‰ between the two techniques is shown to be caused by a "matrix" effect, indicating that any δ44/40Ca measurements on MC-ICP-MS are sensitively controlled by the Ca concentration and the acidity of the solution. Our study demonstrates the possibility to measure the whole dispersion of calcium isotopes with MC-ICP-MS, showing that 40Ca can be used for normalization of 44Ca. © 2004 Elsevier B.V. All rights reserved.