The present study used functional magnetic resonance imaging to delineate cortical networks that are activated when objects or spatial locations encoded either visually (visual encoding group, n = 10) or haptically (haptic encoding group, n = 10) had to be retrieved from long-term memory. Participants learned associations between auditorily presented words and either meaningless objects or locations in a 3-D space. During the retrieval phase one day later, participants had to decide whether two auditorily presented words shared an association with a common object or location. Thus, perceptual stimulation during retrieval was always equivalent, whereas either visually or haptically encoded object or location associations had to be reactivated. Moreover, the number of associations fanning out from each word varied systematically, enabling a parametric increase of the number of reactivated representations. Recall of visual objects predominantly activated the left superior frontal gyrus and the intraparietal cortex, whereas visually learned locations activated the superior parietal cortex of both hemispheres. Retrieval of haptically encoded material activated the left medial frontal gyrus and the intraparietal cortex in the object condition, and the bilateral superior parietal cortex in the location condition. A direct test for modality-specific effects showed that visually encoded material activated more vision-related areas (BA 18/19) and haptically encoded material more motor and somatosensory-related areas. A conjunction analysis identified supramodal and material-unspecific activations within the medial and superior frontal gyrus and the superior parietal lobe including the intraparietal sulcus. These activation patterns strongly support the idea that code-specific representations are consolidated and reactivated within anatomically distributed cell assemblies that comprise sensory and motor processing systems.

The present study investigated the neurophysiological processes underlying associative long-term memory retrieval of objects and spatial positions by means of a modified fan paradigm with cued recall and two neuroimaging methods (electroencephalogram [EEG] and functional magnetic resonance imaging). In an acquisition phase, either one stimulus or two stimuli became associated with a noun. During retrieval, probe stimuli comprising noun pairs were presented, and participants had to recall the respective associations and decided whether the nouns are linked to each other via a commonly associated stimulus. With this design, the quality and quantity of recalled associations was systematically varied, whereas the triggering stimuli and response requirements were held constant in all experimental conditions. Recall time proved to be directly related to the number of associations fanning out from a retrieval cue. Correspondingly, the hemodynamic response (blood oxygen level-dependent [BOLD] signal) and the amplitude of slow negative EEG potentials increased monotonically with the number of associations in both left anterior and bilateral posterior cortical areas. These effects were consistently observed with content-specific topographies for the two distinct materials. Furthermore, the multimethod approach revealed a close temporal link between response times and event-related slow potential changes on the one side and a close topographical and amplitude correspondence between slow potentials and BOLD signal changes on the other. The integrated results suggest that the neuronal dynamics of associative memory retrieval are equivalent for different types of associations, but that the structural basis is clearly content-specific.