REVIEWS We pr viously disc that of the ed ofa elevance T eaonG or infrequent stim its function tsn which thes nctionsotthe tstudy sho he right TP] arget at山 of the TP]during th that in sistent with t activation o y tas ugh this righ nd a res set (fo with e ask se For the of Hier he TP The a e stin he ent set.are elto th ne no ated pote s with prefr en range of of the ortex ventral Tpl-IFg networks to e nt be ne ch the the ed in c ng system that det studies in h The pariet spati e IPL. uld c TP are preser thar nab locati nlepectedey nt.the square ina fieldo ma n TP UME 3 MARCH 2002211NATURE REVIEWS | NEUROSCIENCE VOLUME 3 | MARCH 2002 | 211 REVIEWS network that has many of the properties expected of a network that is involved in directing attention to a stimulus. We offer several related interpretations of its function. Functions of the ventral frontoparietal network. First, this network might represent the exogenous orienting system, which directs attention to the spatial locations of salient stimuli. This hypothesis is most strongly sup￾ported by the modulation of the TPJ during the presen￾tation of stimuli that induce a task-contingent shift of attention30. It is also consistent with the activation of TPJ–VFC during reorienting to spatially unexpected targets10,79. Although this right-hemisphere network is also engaged by low-frequency stimuli at attended or expected locations80,81, this might reflect its involvement in orienting to segmented objects rather than just to spatial locations. Alternatively, the ventral network could work in conjunction with the dorsal network during stimulus￾driven orienting. Whereas sensory cues produce facili￾tation and inhibition with some spatial selectivity, the ventral network responds equally well to stimuli at dif￾ferent locations10,21,81. At present, there are no data on whether this network contains a spatial map that could direct attention to the location of the detected change. The spatial precision of exogenous orienting might depend on the co-activation of the TPJ with the more dorsal IPs–FEF network. The dorsal network might also be the source of the task-contingent properties of TPJ activation, as the TPJ is not active during cue or search periods in which attentional sets are generated and maintained. This influence might be direct or indi￾rect through the top-down effects of the dorsal net￾work on the visual cortex. Finally, the dorsal IPs–FEF network also has properties, such as sensitivity to stim￾ulus distinctiveness for a range of visual features and task contingency, that are consistent with a role in exogenous orienting. Future studies will need to tease apart the relative contributions of the dorsal IPs–FEF and the ventral TPJ–IFg networks to stimulus-driven orienting. We suggest several ways in which the two systems might interact. One possibility is that the ventral net￾work serves as an alerting system that detects behav￾iourally relevant stimuli in the environment, but is not equipped with high-resolution spatial sensors. Once a relevant stimulus is detected, its precise localization depends on the dorsal IPs–FEF system. A related hypothesis is that the TPJ–VFC system acts as a circuit breaker of ongoing cognitive activity when a behav￾iourally relevant stimulus is detected. When subjects detect a low-frequency or unexpected event, they must break the current attentional set and adopt a new one on the basis of the incoming stimulus. In a recent set of experiments, we found a widely distributed set of endogenous signals that were related to the termination of an ongoing state of readiness for an event96. The pat￾tern of activation in TPJ–VFC was consistent with the hypothesis that the ventral network generated the signal that terminated the task set. We previously discussed how stimulus-driven ori￾enting is modulated by task relevance. Activity in the TPJ is also modulated by task relevance. This region is enhanced when oddball or infrequent stimuli in an attended modality (for example, vision) are detected, as compared with experiments in which the same stimuli are presented and subjects are monitoring for targets in a different modality87. A recent study shows a more direct demonstration of contingent orienting in the right TPJ88. Subjects searched for a red target at the centre of gaze while being presented with irrelevant distracters in the periphery of the visual field. When distracters were the same colour as the target, perfor￾mance on the primary task was disrupted, presumably reflecting a shift of attention to the peripheral dis￾tracters and a resulting loss of targets at the centre. The right TPJ was specifically activated by task-contingent distracters, but not by distracters that did not attract attention. The above studies indicate that the TPJ is differen￾tially engaged by low-frequency stimuli that require reorienting of attention within the current task set (for example, responding to targets on a screen). We suggest that the TPJ is also strongly engaged by stimuli that are behaviourally relevant but require a change in the current task set. For example, the alarm at the museum is not part of the task set of listening to the guide’s discussion of Hieronymous Bosch, but it is clearly a behaviourally relevant stimulus. The TPJ is not well activated by low-frequency task￾irrelevant stimuli that are embedded in trains of stan￾dard and oddball targets87,89. Typically, these stimuli, which do not require either a response or a change in the current task set, are novel to the subjects. Novel stimuli might activate more robustly prefrontal regions89,90, damage to which specifically impairs nov￾elty-related potentials84,85. Patients with prefrontal lesions have problems in adapting to novel situations and stimuli91, and show a decreased autonomic response to novel stimuli92. It is possible that the differ￾ential response of the TPJ and frontal cortex to novel stimuli reflects a differential role in bottom-up atten￾tion. The frontal cortex might be necessary for evaluat￾ing the novelty of stimuli, whereas the TPJ might be more involved in detecting their behavioural valence. Our discussion of the TPJ region has been limited to brain-imaging studies in humans. The parietal compo￾nent of the TPJ response is located on the gyral surface of the IPL, which in monkey would correspond to area 7a. Neurons in area 7a respond more briskly when stim￾uli are presented at unattended than at attended loca￾tions93,94, and when stimuli are behaviourally relevant25. Neurons in area 7a also code the location of distinctive stimuli that differ from the background (such as a red square in a field of green squares)95. These neuronal modulations for unattended and salient stimuli are con￾sistent with the observed fMRI signals in the human TPJ. However, at present, it is unclear whether monkey area 7a is homologous to human TPJ. In summary, the current evidence supports the existence of a right-lateralized ventral parietofrontal