In the visual system, the so-called what or ventral stream has been found to be closely related to working memory functions (Haxby et al. Reference Haxby, Gobbini, Furey, Ishai, Schouten and Pietrini2001; Ishai et al. Reference Ishai, Ungerleider, Martin and Haxby2000). Moreover, in a set of fMRI studies, we found that both the dorsal and the ventral visual stream are significantly involved when participants have to reactivate representations from long-term memory that had been visually encoded: Dorsal (parietal) areas light up when participants retrieve spatial information and ventral areas (temporal, inferior frontal cortex) do so when they retrieve object information (Khader et al. Reference Khader, Burke, Bien, Ranganath and Rösler2005; Reference Khadar, Knoth, Burke, Ranganath, Bien and Rösler2007). These findings are in line with the general idea that cortical areas are not only specialized for on-line processing of distinct types of information during perception or action, but that the very same areas are also the locations where information is permanently stored (McClelland et al. Reference McClelland, McNaughton and O'Reilly1995; Rösler & Heil Reference Rösler, Heil, Kluwe, Lüer and Rösler2003). The question, therefore, arises as to whether the areas identified by Dijkerman & de Haan (D&dH) are also involved in haptic working memory and long-term memory tasks. As a matter of fact, recent findings demonstrate that networks within the dorsal action stream are not functionally restricted to immediate guidance of hand actions; rather, they are also recruited for storage and retrieval of action-related spatial information. We briefly summarize the evidence.

Stock et al. (Reference Stock, Röder, Burke, Bien and Rösler2004; under review) studied blood-oxygen level–dependent (BOLD) responses during long-term memory retrieval of proprioceptive and tactile knowledge (i.e., haptics). They used an association learning paradigm that allows brain activations to be monitored during retrieval of specific types of information without confounding the perceptual and mnemonic processes. To this end, participants learned associations between cues (auditorily presented words) and target stimuli. The targets were either haptically explored three-dimensional objects or haptically explored locations in three-dimensional space. In the retrieval test, participants heard two cue words and had to decide whether or not both were associated with the same target. By manipulating the number of associated targets, the difficulty of the retrieval situation was varied systematically. This procedure enforces memory activation of either object or position knowledge in the retrieval situation, without the necessity of processing different types of stimuli perceptually. The results substantiate the functional distinction proposed by D&dH: Retrieval of haptically encoded objects activated, among other areas, the secondary somatosensory cortex (SII), the insula, and the posterior parietal cortex (PPC), whereas retrieval of haptically encoded spatial positions activated, among others, the SII and the PPC but not the insula. Thus, the functional dissociation proposed by D&dH is not only relevant for perception and action, but it also applies to the retrieval of permanently stored information.

In other studies (Fiehler Reference Fiehler, Burke, Bien and Rösler2006; in press), we investigated the role of the dorsal action stream in a haptic working memory task. Participants performed a purely haptic version of a delayed-recognition task, in which they encoded, briefly maintained, and finally recognized hand movements without any visual feedback. Task difficulty was systematically manipulated by the number of complex hand-movement sequences per trial. The BOLD response pattern revealed that haptic encoding of hand movements activated the primary somatosensory cortex and secondary somatosensory cortex, the PPC (dorsal stream), premotor areas, and the right occipitotemporal cortex (ventral stream). The majority of these regions responded to task difficulty reflecting increased processing demands. Short-term maintenance of haptic information elicited load-dependent activity in the left anterior intraparietal sulcus and closely adjacent areas (dorsal stream). Consistent with the proposed theory of D&dH, these data confirm that the PPC is recruited for haptic action control, indicating haptic-motor coupling networks within these areas. Moreover, the results demonstrate that the dorsal stream is not only functionally related to online action control but also to working memory maintenance of action-related spatial information. Others (Macaluso & Driver Reference Macaluso and Driver2003; Ricciardi et al. Reference Ricciardi, Bonino, Gentili, Sani, Pietrini and Vecchi2006) observed activation in the PPC for both visual and tactile spatial stimuli, which suggests that this region has multimodal features and may be engaged in short-term maintenance of spatial information irrespective of the input modality.

Evidence for a mnemonic function of the PPC in action processing is also provided by intracranial recordings in monkeys. Murata et al. (Reference Murata, Gallese, Kaseda and Sakata1996) observed neuronal activity in the anterior intraparietal cortex (AIP) during a delayed hand manipulation task in which monkeys had to manipulate the remembered object in the dark. The results indicate that monkey AIP, as part of the dorsal action stream, is involved in working memory maintenance of spatial characteristics that guide hand actions in relation to objects. Studies using delayed saccade tasks also indicate that the human PPC is involved in representing spatial information for prospective guidance of actions. Activity in the PPC was modulated by the length of the delay period, which suggests a critical role of this area for maintaining movement relevant spatial information (Curtis Reference Curtis2006). These findings are supported by brain-imaging studies on delayed pointing movements (Connolly et al. Reference Connolly, Andersen and Goodale2003; Lacquaniti et al. Reference Lacquaniti, Perani, Guigon, Bettinardi, Carrozzo, Grassi, Rossetti and Fazio1997).

At first glance, these findings seem to be at variance with clinical studies demonstrating a performance deficit in a patient with visual form-agnosia (ventral stream lesion; Goodale et al. Reference Goodale, Jakobson and Keillor1994) and performance improvement in a patient with posterior parietal lesions (e.g., Milner et al. Reference Milner, Dijkerman, Pisella, McIntosh, Tilikete, Vighetto and Rossetti2001) when a delay was inserted in a grasping task. To explain these results, it was assumed that patients used information processed and stored in the ventral stream and that the dorsal stream does not support memory functions. However, the grasping tasks realized in these studies needed visuomotor integration, and the results cannot be immediately generalized to purely haptic memory tasks. The aforementioned studies by Stock et al. and Fiehler et al. employed purely haptic tasks without any visual feedforward or feedback component. Therefore, the results clearly support the conclusion that the areas identified by D&dH are not only recruited for online perception and action control in the somatosensory modality, but that they are also functionally involved in working memory maintenance and long-term memory retrieval of action-related spatial information. Moreover, these findings substantiate the general claim that areas specialized for online processing of specific types of information are recruited for short-term and long-term memory storage and retrieval (McClelland et al. Reference McClelland, McNaughton and O'Reilly1995).