Psychobiology 2000, 28 (2), 187-196 Anatomical and functional subdivision within the primate lateral prefrontal cortex MATTHEW F. S. RUSHWORTH University of Oxford, Oxford, England It has been suggested that there is specialization of function within the lateral prefrontal cortex; it has been argued that the ventral and dorsal divisions of the prefrontal cortex are concerned with the processing of form and spatial information, respectively. The evidence for such a division of function is reviewed. According to an alternative view, the dorsal prefrontal cortex is better thought of as play­ ing a high-level role within the context of the motor system; this is consistent with the fact that it is this part of the prefrontal cortex that is connected withlateral and medial premotor areas, the parietal cor­ tex, and the cerebellum and that it is this part of the prefrontal cortex that is concerned with action se­ lection and attention to action. The ventral prefrontal cortex, on the other hand, has widespread con­ nections with sensory cortices and may be essential for stimulus selection and sensory attention; this may underlie its role in arbitrary or conditional associations involving diverse stimuli. Every year physiologists and anatomists reparcel the to be found within the large expanse of the lateral pre­ primate temporal and parietal lobes into ever smaller di­ frontal cortex. visions. Each region has its own distinctive pattern of neuronal activity and anatomical connections that allows Working Memory and Regional Specialization it to make its own individual contribution to the cognitive Within the Lateral Prefrontal Cortex life ofthe brain's owner. The frontal lobes stand as some­ Any reference to working memory must be prefaced thing of an affront to the subdividers' project, because by acknowledgement of the confusion that has surrounded there is so little agreement about what the basic compo­ the term. The following discussion of working memory nent processes of the frontal lobes might be (Roberts, will be restricted to the more limited sense of retention Robbins, & Weiskrantz, 1998). of information through short delays in order to guide It is less contentious to suggest that we can draw basic subsequent action (Goldman-Rakic, 1987). This hypoth­ distinctions between the lateral and the ventral surfaces esis is distinct from the more complicated set of processes of the prefrontal cortex (see, e.g., Dias, Robbins, & Rob­ that include "executive control," manipulation of infor­ erts, 1996; Roberts, 1996). Perhaps the most influential mation, the control of attention, and short-term memory theory of lateral prefrontal function has concerned its that have been subsumed under the heading of working role in holding (Goldman-Rakic, 1987, 1998) and ma­ memory by authors such as Baddeley and Hitch (1974). nipulating (Petrides, 1998) information in working mem­ According to Goldman-Rakic's (1998) domain-specific ory. In people,.working memory deficits after lateral pre­ theory of prefrontal function, different regions ofthe lat­ frontal lesions can be distinguished from deficits in the eral prefrontal cortex are concerned with holding different affective control of behavior caused by lesions of the ven­ types of information. Whereas the ventral lateral prefrontal tromedial prefrontal cortex (Bechara, Damasio, Tranel, cortex (VLPFC) is concerned with holding information & Anderson, 1998). Neuroimaging studies also suggest about forms, the dorsolateral prefrontal cortex (DLPFC) a similar role for the human ventromedial prefrontal cor­ is concerned with holding information about spatial 10- tex (Elliott, Frith, & Dolan, 1997, 1999). In the monkey, cations. Two principle lines ofevidence have been adduced a high proportion of ventromedial prefrontal cells are con­ in support of this model. cerned with learning about potential rewards and pun­ First, it has been argued that the DLPFC and the VLPFC ishments (Rolls, 1999), and lesions here disrupt a mon­ are distinguished by their anatomical connections with key's ability to learn which one of a pair of objects is the parietal and the temporal lobes, respectively (Petrides rewarded (Baylis & Gaffan, 1991). What is not so clear, & Pandya, 1984). An influential account of parietal and however, is whether there are any functional subdivisions temporal lobe specialization (Ungerleider & Mishkin, 1982) maintains that these regions are relatively special­ ized for processing spatial and object-related information, This research was supported by the Royal Society. Correspondence respectively. It was therefore argued that the specializa­ concerning this article should be addressed to M. F. S. Rushworth, De­ partment of Experimental Psychology, University of Oxford, South tions of the parietal and temporal lobes might be carried Parks Road, Oxford OXI 3UD, England (e-mail: matthew.rushworth@ forward into the DLPFC and the VLPFC. The attractively psy.ox.ac.uk). simple picture of temporal-VLPFC and parietal-DLPFC 187 Copyright 2000 Psychonomic Society, Inc. 188 RUSHWORTH systems should be revised by demonstrations that the only evidence that form and spatial information pro­ anatomical connections are not so clearly segregated; cessing is segregated within the prefrontal cortex is that there are some connections between some more poste­ there may be a rostral!caudal division within the VLPFC rior parts of the VLPFC and the parietal lobe (Cavada & (Figure 1). Possibly related rostral! caudal transitions have Goldman-Rakic, 1989; Schall, Morel, King, & Bullier, recently been observed in macaque single-unit cell re­ 1995). Nevertheless, despite these equivocations, hypoth­ cordings (White & Wise, 1999). Rainer, Asaad, and eses of temporal/parietal specialization can still inform Miller (1998a) have also emphasized a more prevalent spa­ hypotheses ofVLPFC/DLPFC specialization. tial signal in the posterior prefrontal cortex. Human neu­ The second line of evidence is based on claims that roimaging studies have shown that there may also be a monkey lesion and cell recording studies and human dorsal region just in front of the frontal eye fields, out­ neuroimaging studies have demonstrated dorsal and ven­ side of what is classically considered the DLPFC, that is tral specializations for processing spatial and form in­ specialized just for spatial working memory tasks (Court­ formation, respectively. The human neuroimaging liter­ ney, Petit, Maisog, Ungerleider, & Haxby, 1998). What ature will be briefly reviewed before considering the may be an analogous part of the macaque brain, in the an­ results of cell recording experiments performed with terior part of area 8, has also been shown to contain cells monkeys. Rushworth and Owen (Owen 1997; Rushworth, with delay activity (Chafee & Goldman-Rakic, 1998). Nixon, Eacott, & Passingham, 1997; Rushworth & Owen, On turning to the monkey single-unit recordings stud­ 1998) have reviewed this evidence and have argued that ies that have examined DLPFCIVLPFC specializations it is not always persuasive. For example, Figure 1 sum­ for form and spatial processing, it is clear that the ex­ marizes activations recorded in human neuroimaging periments have produced inconsistent results that are dif­ studies of object/form or spatial working memory tasks ficult to reconcile. On the basis of single-unit recording published before 1998. It is clear that there is no dorsal! studies, it has been claimed that there is evidence both ventral distinction between spatial and form tasks. The for (O'Scalaidhe, Wilson, & Goldman-Rakic, 1997, 1999; Wilson, O'Sca1aidhe, & Goldman-Rakic, 1993) and against (Fuster, Bauer, & Jervey, 1982; Rao, Rainer, & Miller, 1997) anatomical segregation within the pre­ frontal cortex. A similar dichotomy of opinion exists when one considers monkey lesion studies (Levy & Goldman-Rakic, 1999; Petrides, 1995; Rushworth, Nixon, et aI., 1997). The debate has been reviewed at length before (Rushworth & Owen, 1998), and rather than detaining ourselves with it further, it may be better to summarize: There may be some segregation of func­ tion between the dorsal and the ventral prefrontal cor­ tices, but if there is, its basis and guiding principle re­ main unclear. The next section considers an alternative proposal for functional specialization within the lateral frontal cortex. A Reevaluation of the Importance of the Lateral Prefrontal Cortices' Anatomical Connections Figure 1. The location of peak activations in spatial working memory tasks (f-.ned shapes, *) and form working memory tasks with the Parietal and Temporal Lobes (unnIled shapes, 0) projected onto a lateral diagram of a human The inference from the lateral prefrontal cortex's con­ brain. Activation peaks recorded in both the left and the right nections with the temporal and parietal lobes to domain­ hemispheres are shown on the same diagram. It is clear that spa­ specific processing within the prefrontal cortex is itself tial working memory tasks activate both the ventrolateral pre­ frontal cortex (VLPFC) and the dorsolateral prefrontal cortex based on a particular conception of parietal and tempo­ (DLPFC) and that form working memory tasks activate both the rallobe function (Ungerleider & Mishkin, 1982). Unger­ VLPFC and the DLPFC. The DLPFC and the VLPFC are not leider and Mishkin argued that the parietal and temporal distinguished by any relative specialization for holding spatial or lobes were the destinations of two streams of visual­ form information in working memory. There is, however,