As a simple, non-invasive method of blood oxygenation level-dependent (BOLD) signal calibration, the breath-hold task offers considerable potential for the quantification of neuronal activity from functional magnetic resonance imaging (fMRI) measurements. With an aim to improve the precision of this calibration method, the impact of respiratory rate control on the BOLD signal achieved with the breath-hold task was investigated. In addition to self-paced breathing, three different computer-paced breathing rates were imposed during the periods between end-expiration breath-hold blocks. The resulting BOLD signal timecourses and statistical activation maps were compared in eleven healthy human subjects. Results indicate that computer-paced respiration produces a larger peak BOLD signal increase with breath-hold than self-paced breathing, in addition to lower variability between trials. This is due to the more significant post-breath-hold signal undershoot present in self-paced runs, a characteristic which confounds the definition of baseline and is difficult to accuratelv model. Interestingly, the specific respiratory rate imposed between breath-hold periods generally does not have a statistically significant impact on the BOLD signal change. This result can he explained by previous reports of humans adjusting their inhalation depth to compensate for changes in rate, with the end-goal of maintaining homeostatic ventilation. The advantage of using end-expiration relative to end-inspiration breath-hold is apparent in view of the high repeatability of the BOLD signal in the present study, which does not suffer from the previously reported high variability associated with uncontrolled inspiration depth when using the end-inspiration technique. (c) 2008 Published by Elsevier Inc.