The use of intermolecular multiple-quantum coherences (iMQCs) for MRI has attracted substantial interest in recent years (1-4). In such experiments, image contrast is based on dipolar couplings between nuclear spins in different molecules separated by a specific distance-the so-called 'correlation distance'-which may range from 10 m to a few millimeters (2). Dipolar couplings are not observed in conventional MRI experiments because short-range interactions average to zero in time through rapid molecular tumbling, and long-range interactions average to zero in space as long as the sample is magnetically isotropic. They reappear, however, if a magnetic field gradient is used to break this isotropy (2,5). In particular, this 'correlation gradient' of amplitude G c and duration ␦ c winds up the magnetization as a helix and modulates the sample spatially (6). It turns out that the dipolar signal predominantly comes from spins separated by one-half of the helix pitch along the gradient direction as the spherical symmetry is completely lost for this distance. A correlation distance is, hence, defined according to (3):where ␥ is the magnetogyric ratio. A unique feature of iMQC imaging is that the spatial scale of the experiment can be manipulated externally by adjusting G c and ␦ c . Intermolecular double-quantum coherences (iDQCs) are more sensitive to local magnetic susceptibility variations as compared to the conventional single-quantum coherence (SQC) signal. This has already been used to generate a strong blood oxygen level-dependent (BOLD) contrast for functional MRI (fMRI) by several research groups (7-10). In such experiments, functional contrast is thought to be sensitive to changes in local magnetic field perturbations in or around blood vessels occurring on a length scale that is defined by the correlation distance. The recorded iDQC signal is a function of the resonance frequency difference of the two interacting spins, which is sensitive to the distribution of local field gradients, whereas conventional BOLD experiments reflect the average gradient strength within an imaging voxel (7,11). Such characteristics could be especially valuable if the selection of a length scale for the detection of susceptibility changes would also imply a selection of blood vessels of a particular size. In this scenario, iDQC-based fMRI might permit focusing only on hemodynamic changes in the capillaries, which are located at a minimal spatial offset from the site of neuronal activity. However, the proposed possibility to tailor the distance scale of the BOLD sensitivity for optimum specificity remains to be shown. In nonfunctional human brain imaging studies, obvious contrast changes could not be produced by variation of the correlation distance between 88 and 530 m (12). Using numerical simulations, Marques and Bowtell (13) did not find a significant effect of the modulation length on the iDQC signal and concluded that a selection of susceptibility changes in vessels of a particular size may not be possible. A thorough experim...