Studies of parietal cortex in monkeys
In monkeys, the parietal lobe includes both the superior and inferior parietal lobules, which are composed of many different architectonically defined cortical areas (Fig. 1A). The superior parietal lobule (SPL) is composed of area PE and PEc on the gyral surface, and areas PEa and MIP (medial intraparietal) in the dorsal bank of the intraparietal sulcus (IPS). These areas are all components of the classically defined Brodmann’s area 5 (BA5). Areas V6A and V6 (Galletti et al., 1996), respectively in the anterior bank and fundus of the parieto-occipital sulcus, are also part of the SPL. The SPL extends into the medial wall of the hemisphere, including area PEci in the caudal tip of the cingulate sulcus and area PGm (7m). The inferior parietal lobule (IPL; BA7) is composed of areas PF, PFG, PG and Opt on the gyral surface, as well as by anterior intraparietal and lateral intraparietal areas (AIP and LIP) in the lateral bank of the IPS. Because of its corticocortical connectivity (see below), area VIP can also be included in this group, although it lies around the fundus of the IPS. Functionally it does seem to belong more to the IPL than to the SPL. All of the above areas are globally referred to as the PPC.
Figure 1. The parietofrontal network. (A) Parcellation of the parietal and frontal cortex in the monkey. Lateral view of the areas (PE, PEc) of the exposed part of the SPL (top), of the surface of the inferior parietal lobule (IPL; Opt, PG, PFG, PF) and of the intraparietal sulcus (IPS; MIP, PEa, VIP, AIP, LIP), which is shown as flattened in the inset of the brain figurine (bottom). Motor cortex (MI), dorsal (F2, F7) and ventral (F4, F5) premotor areas are shown in the frontal lobe. (B) Parietal areas (PEc, PGm) of the mesial aspect of the hemisphere, together with area PEci and the cingulated motor areas (CMAr, CMAd, CMAv). (C) The inferior parietal lobule (IPL) has been removed to show the areas lying in the rostral bank of the parietooccipital sulcus (V6A) and in the medial wall of the IPS (MIP, PEa). All SPL areas except PGm (7m) are part of Brodmann’s area 5, while all IPL areas are part of Brodmann’s area 7. (D) Lateral (left) and mesial (right) aspects of the human brain showing the areas of the SPL (Brodmann’s areas 5, 7a and 7b; yellow) and of the IPL (Brodmann’s areas 39 and 40; orange).
Download figure to PowerPoint
In recent years the connectivity of the parietal lobe in monkeys has been mapped extensively with anterograde and retrograde tracing techniques. The anatomical afferents and efferents of PPC are primarily composed of reciprocal connections to the frontal motor and premotor cortex and temporal and occipital visual areas, as well as the prefrontal and cingulate cortex. There are very few if any connections with the amygdala, the hippocampus or the ventromedial prefrontal cortex. Thus, there are few if any limbic inputs to these areas. However, some inputs come from orbital cortical areas 12 and 13. Describing the complete set of connections between the parietal lobe and all other areas with which it is interconnected would be highly complex and would not necessarily clarify the routes of information flow into and out of its constituent areas. Therefore in attempting this task we will mostly refer to a recent statistical study of the connectivity of these areas (Averbeck et al., 2009). This approach first clusters together sets of individual architectonically defined areas, based upon their inputs. Following this, one can look at the ‘anatomical fingerprint’ of a cluster of areas, which is the proportion of inputs coming from different sets of areas.
This hierarchical cluster analysis shows that clusters in parietal cortex are composed of spatially adjacent areas. Specifically, there are four well-defined clusters, each forming one branch of a bifurcation in a hierarchical tree (Fig. 2). A dorsal parietal cluster (PAR-D) includes areas MIP, PEc and PEa; a somatosensory cluster (SS) is composed of the first (SI; a ventral parietal cluster (PAR-V) is formed by areas PF, PFG, PG and AIP, and a mediolateral parietal cluster (PAR-ml) consists of areas PGm (7m), V6A, LIP, VIP and Opt.
Figure 2. Hierarchical cluster analysis of the parietofrontal system. Clusters of architectonically defined areas based upon their inputs. The identified clusters were represented in different colors on the basis of the dominant connections (arrows) that each cluster of the parietal (frontal) lobe has with groups of frontal (parietal) areas. Left, clusters of frontal areas. [PM-D (dorsal premotor cluster), red; MI, blue; PM-V (ventral premotor cluster), orange; CING (cingulate cluster), green; PFC (prefrontal cluster), yellow]. Right, clusters of parietal areas. (PAR-D, red; SS, blue; PAR-V, orange; PARml, yellow). Arrows indicate connectivity, and numbers indicate the fraction (percentage) of the total input to each area that corresponds to each input. The main information streams revealed by the cluster analysis are indicated by arrows in the brain figurine at the bottom of the figure. (Modified from Averbeck et al., 2009).
Download figure to PowerPoint
Given these clusters, we can analyze the inputs which characterize the areas belonging to each cluster, as well as the inputs to each cluster from other parietal and frontal areas or from areas outside the parietofrontal network. The strongest input to each parietal cluster from parietal cortex comes from other areas within the same cluster, which shows that connectivity tends to be stronger locally, i.e. cortical areas tend to receive strong connections from spatially nearby areas. The strongest input from frontal cortex to the PAR-D cluster stems from the dorsal premotor cluster, the major input to the SS cluster comes from the primary motor cortex (MI), most of the input to the PAR-V cluster originates from ventral premotor areas, and the strongest input to the PAR-ML areas comes from the lateral prefrontal cluster (PFC). The connectivity between parietal and frontal motor areas is topographically organized. It is also reciprocal, as the strongest input to each corresponding frontal cluster tends to originate from the parietal cluster to which it provides the strongest input.
Thus, parietal areas tend to receive strong inputs from the other parietal areas within the same cluster as well as from topographically related frontal areas. However, many parietal areas also receive inputs from outside the parietal–frontal network and in fact these inputs can be more substantial than those from frontal cortex. Specifically, 31, 10, 7 and 23% of the inputs to the parietal clusters (PAR-ML, SS, PAR-D and PAR-V) came from outside the parietal frontal network. Thus, both the PAR-ML and the PAR-V clusters receive more inputs from outside the parietal–frontal network than they do from the frontal areas. The external inputs into PAR-ML come from extrastriate visual areas, prefrontal area 9 and areas MT, MST and STSd in the caudal tip and dorsal bank of the superior temporal sulcus. The external inputs to PAR-V originate from STSd, MT and MST, as well as temporal visual areas TE and TEO, areas PFop and PGop in the dorsal insula, and orbital areas 12 and 13.
The reciprocity and overall pattern of the parietofrontal connections clearly define the existence of privileged, although not private, routes of information flow between parietal and frontal cortex (Fig. 2). More specifically, the mediolateral parietal cluster and its prefrontal counterpart are involved in the control of visually-guided eye movements and in the detection of saliency in the visual scene (Colby & Goldberg, 1999). Most areas in this cluster, such as Opt, V6A and PGm (7m), are also involved in the early stages of the eye–hand coordination for reaching (Ferraina et al., 1997a,b; Battaglia-Mayer et al., 2000, 2001, 2003, 2005, 2007) and provide the oculomotor system with the visual information necessary for eye-movement control. PAR-D, together with the dorsal premotor cluster, is responsible for the combination of visual and somatic information necessary for visual reaching (Georgopoulos et al., 1984; Kalaska et al., 1990; Colby & Duhamel, 1991; Lacquaniti et al., 1995; Johnson et al., 1996; Battaglia-Mayer et al., 2000, 2001; Hamel-Paquet et al., 2006). PAR-V cooperates with the ventral premotor cluster in the visual control of hand–object interaction underlying different forms of grasping (Taira et al., 1990; Rizzolatti & Matelli, 2003). Furthermore, it has been suggested that areas PFG and AIP represent the parietal node of the mirror system (Fogassi et al., 2005; Rizzolatti & Sinigaglia, 2010). Within this cluster, recent studies (Battaglia-Mayer et al., 2005, 2007) have shown that neurons in areas PG and Opt are involved in directing reaches towards objects mainly located in contralateral space. In these areas, neural firing rates are higher when the hand moves toward the fixation point, as compared to any other possible form of coordinated eye–hand movement. It is worth stressing that this is the most common form of visuomotor behaviour in our daily life. PAR-V is also involved in both the processing of visual information and the preparation of movements in the context of more complex visuomotor tasks, such as interception of moving targets (Merchant et al., 2004). Closer to the motor output, neurons in the somatosensory cluster encode, among other variables, information related more directly to arm movement, such as limb position and velocity (Georgopoulos & Massey, 1985; Prud’homme & Kalaska, 1994; Averbeck et al., 2005; Archambault et al., 2009), and convey this information to frontal cortex via direct projections to MI.
All the above information is crucial to understanding the fine-grain architecture of the parietal–frontal system, which is characterized by the existence of trends in the functional properties of neurons along the rostrocaudal dimension of both frontal and parietal cortex. The gradient-like architecture mostly refers to the arrangement of visual, eye and hand-related signals, as revealed by quantitative analysis in SPL, dorsal premotor and motor cortex (Johnson et al., 1996; Burnod et al., 1999; Battaglia-Mayer et al., 2001, 2003 for a discussion; Ferraina et al., 2009). In the caudal and rostral poles of the network, respectively in the parietal areas V6A (Galletti et al., 1995, 1997; Battaglia-Mayer et al., 2000, 2001), 7m (Ferraina et al., 1997a,b) and dorsorostral premotor cortex (Johnson et al., 1996; Fuji et al., 2000), visual and eye-related signals predominate over coexisting hand information (Johnson et al., 1996). In contrast, hand information dominates over visual and eye signals in the rostralmost part of the SPL (area PE; Johnson et al., 1996) and in the caudalmost part of the frontal cortex (PMdc/F2, MI; Johnson et al., 1996). In intermediate parietal (areas MIP, PEc, PEa) and frontal (PMdc/F2) lobe regions, eye and hand signals coexist, with different relative strengths depending on the cortical zone considered (see Battaglia-Mayer et al., 2003, for a review). Similar trends of eye and hand information, as well as of preparatory and movement-related signals, exist in the frontal node of the parietofrontal network, across motor cortex, supplementary motor area (SMA) and pre-SMA (Alexander & Crutcher, 1990; Rizzolatti et al., 1990; Matsuzaka et al., 1992; Hoshi & Tanji, 2004; see Nakev et al., 2008 for a recent review). Throughout the network all these signals are directional in nature (Georgopoulos et al., 1981; Kalaska et al., 1983; Caminiti et al., 1991; Johnson et al., 1996; Battaglia-Mayer et al., 2005).
Superimposed on this rostrocaudal dimension is a second gradient concerning the relative strength of different motor-related signals within the network. In fact a transition from preparatory (set- and memory-related) to genuine motor signals occurs moving from caudal to rostral in the SPL; the opposite holds true in the frontal cortex, as one moves from dorsorostral to dorsocaudal premotor cortex toward MI. However, although in different proportions, cells encoding eye and/or hand position information are ubiquitous at all rostrocaudal levels in the network, so as to form a matrix of position representation in which preparatory and movement-related signal are embedded and eventually selected for movement on the basis of task demands (Johnson et al., 1996; Battaglia-Mayer et al., 2001). This is not surprising if one considers that information about eye and hand position is essential under virtually all conditions for early planning of combined eye–hand movements, such as those typical of reaching to visual targets or those necessary to manipulate objects of interest. In fact, if the modulation exerted by eye position on visual parietal neurons (for a review see Andersen & Buneo, 2002) might favour the transformation of target location from retinal into body-centred coordinates, hand position signals are essential to compute the corresponding hand movement trajectory. As such, they exert a profound influence on encoding movement direction in the motor (Caminiti et al., 1990), premotor (Caminiti et al., 1991; Burnod et al., 1992) and PPC (Lacquaniti et al., 1995; Battaglia-Mayer et al., 2000, 2005; Ferraina et al., 2009) areas. Similar trends of functional properties exist across the different architectonic areas (Pandya & Seltzer, 1982; Rozzi et al., 2006) of the flat exposed part of IPL, as gradients have been reported for eye-related signals across areas 7a and LIP (Barash et al., 1991), the first containing mostly post-saccadic neurons, the latter containing mainly pre-saccadic cells. A gradual transition of functional properties of neurons across the areas of the convexity of the IPL was first observed by Hyvärinen (1981) and recently confirmed and extended by Rozzi et al. (2008).
An additional and crucial feature of the network emerges when considering that frontal and parietal areas displaying similar neuronal activity-types are linked by reciprocal association connections (Johnson et al., 1996; Chafee & Goldman-Rakic, 2000; Marconi et al., 2001), indicating that the parietofrontal association system probably both reflects and imposes functional specialization on cortical regions within the network. As a consequence, in the parietal and frontal cortex different forms of visuomotor activities involving the coordination of eye–hand movements might emerge as a result of a progressive match of spatial information representing the positions of the two effectors and their relation to visual targets. This match of signals could be based on a recursive signalling operated through ipsilateral association connections and refined locally by intrinsic connectivity, i.e. by short intracortical connections (see Burnod et al., 1999 for a theoretical frame). This interpretation is consistent with a number of experimental observations and with the predictions of network modelling. Experimental results (Johnson et al., 1996; Chafee & Goldman-Rakic, 2000) indicate that association connections are likely to confer common physiological properties on frontal and parietal neurons during behaviour. These connections are also likely to play a major role in shaping network dynamics that are a product of both input activation and previous learning, as a Bayesian collective decision process (Koechlin et al., 1999). Furthermore, populations of units that combine retinal, hand and eye signals, and are linked by recurrent excitatory and inhibitory connections, are necessary to shape the directional tuning properties of SPL neurons (Mascaro et al., 2003). When lateral inhibition is removed from this recurrent network, such spatial tuning properties are disrupted.
Studies of parietal cortex in man
In humans (Fig. 1B), as in monkeys, the PPC is composed of both SPL and IPL, which are divided by the IPS. According to Brodmann’s parcellation, the SPL is coextensive with areas 5 and 7 while the IPL includes areas 39 and 40, i.e. the angular and supramargynal gyrus, respectively. In von Economo’s view (1925), the SPL is composed of area PE and the IPL of areas PF and PG, roughly corresponding to areas 40 and 39 of Brodmann. Thus, critical scrutiny of the various architectonic parcellations available in the literature supports the conclusion of Von Bonin & Bailey (1947) that, in spite of certain differences that will be highlighted later, there is a basic similarity in the organization of the parietal lobe in human and nonhuman primates.
Very little is known of the detailed corticocortical connectivity in man. Recent developments in MRI, such as diffusion tensor imaging, offer preliminary information on parietofrontal connectivity. This information will be of crucial importance in the near future, as ongoing parcellations of cortical areas are performed largely on the basis of corticocortical connectivity and less on the basis of architectonic criteria. Gaining a better understanding of cortical connectivity of parietal areas in humans is likely to have a positive impact on our understanding of the evolution of the parietal lobe and of its pathology across species. So far, these studies have shown that the pattern of parietofrontal connectivity obtained from monkey studies is very similar to that of man (Croxson et al., 2005). Furthermore, the lateral premotor cortex of humans has been divided into two distinct regions (Rushworth et al., 2006; Tomassini et al., 2007), a dorsal one corresponding to monkey dorsal premotor cortex (PMd), having the highest probability of connections with the SPL and the adjacent areas of the IPS, and a ventral one corresponding to the monkey’s ventral premotor cortex, with the highest probability of connections with the IPL, in particular the anterior part of the angular gyrus and the supramarginal gyrus. Interestingly, when the locations of these anatomically-defined subregions of PMd were compared with dorsal and ventral premotor areas as defined functionally by functional magnetic resonance imaging (fMRI) studies (see Mayka et al., 2006 for a meta-analysis), a clear overlap was found. By adopting a similar probabilistic tractography approach, the medial premotor cortex of humans has been subdivided into SMA and pre-SMA based on the pattern of corticocortical connectivity (Johansen-Berg et al., 2004). Furthermore, a high degree of anatomical similarity has been found in the division into subcomponents of the superior longitudinal fascicle in monkeys (Petrides & Pandya, 1984, 2002; Shmahmann et al., 2007) and humans (Croxson et al., 2005; Makris et al., 2005; Rushworth et al., 2006).
From a functional point of view, the parietofrontal network appears to perform different computations related to sensorimotor control, such as converting sensory representations from one set of sensory coordinates into another (intersensory predictions), spatial representations from sensory to motor coordinates (sensory–motor transformations or inverse models), and spatial representations from motor to sensory coordinates (forward models). Thus, coordinate transformation for visually guided eye and/or hand movement during reaching could emerge in the operations of the parietofrontal segment of the network, while the frontoparietal connections, by providing information about the sensory consequences of motor plans, might contribute to the composition of forward models of movement.
In conclusion, the functional architecture of the parietofrontal network as described in monkey studies, and its similarity with that of man derived from fMRI and tractography analysis, provides a reasonable background to attempt an explanation of some of the disorders of parietal patients from a neurophysiological perspective.