Concepts on the organisation of human memory are mainly based on the long-held dichotomy of short-term (STM), respectively, working memory (WM) and long-term memory (LTM) (Atkinson and Shiffrin
1968; Brown
1958; Hebb
1949; Peterson and Peterson
1959). Throughout the neuropsychological and neuroimaging literature, two complementary organizational aspects of the human WM system are consistently acknowledged. One is the presence of a “central executive” forming a capacity engaged in encoding, recall and manipulation of WM content independent of a specific modality (‘amodal processor’). The other aspect is a striking modularity of WM with differentiable function and neural correlates of various aspects such as verbal or non-verbal WM or remembering spatial (object location) versus feature-based (object identity) properties (Ragland et al.
2004; Shen et al.
1999; Thomason et al.
2009; Veltman et al.
2003,
2005). The latter distinction (object location vs. features) is particularly interesting, as it relates to a broad distinction between “what” and “where” pathways, which have been described in several functional systems. The idea of two different pathways or streams (what and where) has been discussed in various functional systems in the brain but is possibly best known and widely acknowledged in the visual system. Here, ventral aspects of the occipital and adjacent inferior temporal cortex, in particular along the fusiform gyrus, have consistently been shown to respond preferentially or even exclusively to objects, faces and similar shapes such as letters. In contrast, the dorsal occipital and superior parietal cortex are more preferentially engaged during visual-spatial tasks (Milner and Goodale
2008; Ungerleider et al.
1998; Ungerleider and Haxby
1994). Furthermore, Ungerleider and colleagues were among the first to present evidence for an extension of these functional preferences into the frontal lobe, prompting the notion of dorsal and ventral pathways or streams. While most prominent in the visual system, the concept of discernible streams devoted to the processing of “what” and “where” features has likewise been proposed in the auditory system. In particular, there is solid evidence for separate processing of spatial and non-spatial information along different pathways in non-human primates (Rauschecker and Tian
2000), which seems to be also present in the human auditory system (Hafke
2008; Krumbholz et al.
2007; Loui et al.
2008). Converging, these findings suggest a predominant processing of non-spatial auditory features (such as fundamental frequency or pitch -envelopes) in a ventral auditory stream including inferior frontal and temporal areas, while spatial features such as binaural time- and amplitude differences seem to be processed mainly in a dorsal stream including the superior frontal sulcus and inferior parietal lobe (Arnott et al.
2004). In addition to these streams of visual and auditory sensory processing, the idea of pathways has also been discussed in the context of language processing. A ventral language stream connecting the middle temporal lobe and the ventrolateral prefrontal cortex has been proposed to map sound to meaning, while a dorsal stream, connecting superior temporal lobe and premotor cortices seems to be more involved in converting sound to articulation (Rauschecker
2012; Saur et al.
2008). Finally, there is also evidence from invasive tracer and electrophysiological studies in non-human primates for dorsal and ventral pathways between parietal and premotor areas (Luppino et al.
1999; Matelli et al.
1998). In particular, dorsal premotor regions seem to connect predominantly to the superior parietal cortex and subserve reaching in space while more ventral aspects of the premotor cortex interact more strongly with the intraparietal sulcus for object manipulation, a finding that has more recently been corroborated in humans using diffusion tractography (Tomassini et al.
2007). In summary, a large body of evidence thus points to the existence of “dorsal” and “ventral” streams in the brains of humans and non-human primates that are preferentially dedicated to the spatial and non-spatial, object related processing, respectively. That is, the distinction between dorsal (“where”) pathways processing spatial codes and relationships as well as accounting for different coordinate systems such as egocentric (eye or body related) and allocentric representations on one hand and ventral (“what”) pathways dealing with non-spatial properties of objects such as its shape and colours, arrangement of local elements as well as, potentially, semantic associations on the other seems to represent a fundamental organizational principle of the primate brain.
In a recent coordinate-based meta-analysis of neuroimaging studies, we could demonstrate a similar distinction in the context of working memory (and hence not motor-related) tasks. In particular, there was a clear distinction between brain regions that are reliably activated in tasks requiring the subjects to remember object identity and object location, respectively (Rottschy et al.
2012). While memory for object identity as compared to object location was significantly more likely to recruit the bilateral posterior inferior frontal gyrus (dorsal to, but overlapping with area 44), significantly stronger convergence in tasks requiring to memorize object location, as compared to identity, was found bilaterally on the posterior superior frontal gyrus and the adjacent precentral gyrus. That is, we found a consistent (across paradigms and experiments) distinction between the neuronal correlates of presumed “what” and “where” aspects of working memory in the posterior frontal cortex. Interestingly, the locations of these segregated representations of “what” and “where” memory moreover closely resemble the distinction between ventral and dorsal premotor areas in humans and other primates (Geyer et al.
1999; Rizzolatti and Luppino
2001); (Schubotz and von Cramon
2003; Tomassini et al.
2007). It is important to empathize, however, that in spite of this topographic similarity, our seeds reflect differential activation of the frontal cortex in relation to spatial and object-centred working memory processes and can therefore not be assumed to necessarily correspond to the dorsal and ventral premotor cortex, respectively. The aim of the present study was to delineate the functional connectivity of these regions for “what” and “where” aspects of working memory in human posterior frontal cortex. Moreover, we aim at investigating whether functional connectivity analyses may delineate more generalized streams of “what” and “where” pathways extending into the visual system based on these frontal, WM-related seeds.