1sequence with TR = 4.5s, TE = 23ms, TI1 = 0.7s, TI1stop = 1.2s, and TI2 = 1.4s. For both BOLD and ASL data acquisition, five axial slices were imaged, each of which consisted of a 64x64 matrix of 4x4x6mm 3 voxels. The hypercapnia paradigm consisted of four alternating sequences of inspiration and expiration BHs, each sequence comprising four 30 second BHs, separated by 30s of normal respiration. Results Following inspiration BHs, the end tidal partial pressure of CO2 (ETCO2) increased in all subjects. Following expiration BHs, ETCO2 increased in only four subjects. The most probable reason for the absence of an increase in ETCO2 in the other two subjects was their breathing pattern following the BHs. In cases where an increase in ETCO2 was observed, the mean increase was greater for inspiration BHs (19.8±4.9%) than for expiration BHs (13.3±9.6%). The BOLD signal and CBF increased in grey matter (GM) in all subjects during BHs compared with rest for both types of BH. Changes in BOLD and CBF were greater in GM than in white matter (WM), which is consistent with recent studies 2 . The results are shown in Table 1. Values of M, the scaling factor required for CMRO2 fMRI, have been calculated as by Hoge et al. 3 with α = 0.38 and β = 1.5. Because of the anomalously high values of M calculated for subject 1 for inspiration BHs the data from this subject have been excluded in the average. Values of ΔBOLD are plotted against values of ΔCBF in Fig. 1 (individual subjects, GM). The values of M estimated in the present study are comparable with values estimated by Kastrup et al. 4 , who estimated M as 7±1% using BHs at 1.5T, Chiarelli et al. 5 (in different GM regions: M = 6.6±3.4%, 4.3±3.5%, 7.2±4.1%, CO2 mixed with air at 3T), and Stefanovic et al. 6 (M = 7.2±1%, CO2 mixed with air at 1.5T). Fig. 1. Individual subject ΔBOLD signal vs. ΔCBF in GM. Blue and red points and curves represent inspiration BHs and expiration BHs respectively. The solid curves are plotted using inter-subject mean values of M, and the dotted curves are plotted using inter-subject means ± st devs. Fig. 2. Inter-subject mean changes in BOLD and CBF. Error bars show intersubject standard deviations. Curves are plots of isometabolic Hoge 3 model with values of M between 1% and 20%, plotted at 1% intervals. Discussion and Conclusions Both techniques produced similar values of the calibration constant M, with similar uncertainties, which are all comparable with recent literature values. 4-6 Despite the possible differences that ΔETCO2 may be lower for inspiration BHs, and group estimates of M may be more precise for expiration BHs, the values of M calculated in the present study were similar for both techniques, and had similar uncertainties. Although further study with more subjects would be required to confirm or refute these observations, the present study suggests that there are no significant advantages of using either technique for calibration of the CMRO2 model. CBF and the BOLD signal increased for both techniques in GM and WM. Signal changes were greater for inspiration than for expiration BHs, and greater in GM than in WM. No conclusive advantages of either BH technique were found. In light of this result, it is suggested that inspiration BHs are preferable, for the simple reason that they are easier for subjects to tolerate.