REVIEWS enthey shareatribute e toa nla area LIPis enhance ing fo drive or has been re beha h th ntion it is atin )ma that are rgenetic This migh with t on)ar rstudies indica tha m this s on target Activity ir nap h um th ed to eot ntio do aiecentyre fg e dorsal the bo d ing both visual search an ade to th 色D it she ome c FE in whi odulation is in ent of the mor the nted with and so e i rch and det arily driv a fo he at in de TP hey ar Ir a ork.O t its ke in the FEF genera ensory stimuli that are outside the focus of processing.I 20 MARCH2002 VOLUM正3 208 | MARCH 2002 | VOLUME 3 www.nature.com/reviews/neuro REVIEWS Similarly, the response of neurons in the posterior parietal cortex is modulated by behavioural relevance. The visual response to a flash in area LIP is enhanced when the stimulus is relevant26. Moreover, LIP neurons show little or no response to static stimuli brought into their receptive field by saccades (producing the same transient input to the cell), unless the stimuli are behav￾iourally significant73. LIP neurons also respond to visual transients26, and so might be sensitive to sensory salience. Therefore, macaque areas LIP and FEF are modulated by both the sensory distinctiveness of objects and top-down contextual information. These areas might be involved in generating the salience (activation) maps that are postu￾lated in models of visual search70, which combine bot￾tom-up with top-down information to represent visual objects of interest (FIG. 5a). Correspondingly, neuroimaging studies indicate that the human dorsal frontoparietal system is modulated during search and detection19,74–77, consistent with the idea that this system maintains a salience map. FIGURE 5b shows the cortical distribution of fMRI activity as sub￾jects search for a coherent motion target in a dynamic noise display, with regions that are involved in search labelled in red76. The response in these regions, which include the dorsal frontoparietal cortex and extrastriate cortex, is time-locked to the onset of the search display and is maintained until a target is detected (see time courses in FIG. 5c, middle, IPs function). This activity does not simply reflect a sensory response. Target detection produces a higher signal than do trials in which the target is missed (FIG. 5c, right). Moreover, after target detection, the signal drops off even though the search display is still present, indicating that the fMRI signal tracks processes that are engaged by search (FIG. 5c, right). Similar results have been reported for tasks in which subjects have to detect changes in a visual scene77,78. These findings indicate that the frontoparietal net￾work is modulated during both visual search and detec￾tion.As this network also codes for top-down signals that are related to visual expectancy or goals, it shows some of the properties of salience maps, studied in the single-unit literature, in which top-down and bottom-up informa￾tion interact to specify which relevant object to select. Ventral frontoparietal network and sensory orienting. Whereas the above studies provide information on the neural systems that are recruited when stimuli are related to a task goal (as in studies of search and detection), sen￾sory stimuli of potentially high behavioural significance, particularly when they are unexpected, as in the case of the museum alarm, can reorient attention.An inevitable consequence of orienting towards unexpected sensory events is that ongoing cognitive activity is interrupted.We suggest that this ‘circuit-breaking’function lies outside the IPs–FEF network and is housed in a more ventral cortical network that includes the temporoparietal junction (TPJ) cortex and the ventral frontal cortex (VFC). This network is strongly lateralized to the right hemisphere: we call it the right ventral frontoparietal network. One of its key functions is to direct attention to behaviourally relevant sensory stimuli that are outside the focus of processing. Its broad spatial extent68. Also, distinctive sensory stimuli attract attention more effectively when they are relevant to the task, as when they share attributes or features with the stimuli for which subjects are actively looking (for example, a red shirt when we are searching for someone wearing a red hat)69. This form of stimulus￾driven orienting has been labelled ‘contingent’ to emphasize its dependence on the underlying task set. Although, in real life, there is no question that sudden unfamiliar stimuli can grab our attention no matter what we are thinking at that moment (for example, the museum alarm), it is also possible that some stimuli attract attention because of some form of contingency that is hard-wired in the brain by learning, develop￾ment or genetics. This might explain why James’s instinctive stimuli (blood, wild animals and so on) are powerful attractors of human attention. Current psychological evidence supports the idea that orienting to sensory stimuli is modulated by both bottom-up and top-down signals. This dynamic inter￾action is central to current theories of visual attention. Top-down signals that reflect our expectations might influence the sensory salience of objects in the visual system. Activity in ‘salience’ maps, which sum the bottom-up and top-down signals for different object features, might determine which objects are selected for recognition and action70. Sensory stimuli also orient attention towards unexpected events of potentially high behavioural significance. Below, we discuss how these different influences on stimulus-driven attention are mediated by the interaction between the dorsal frontoparietal network and a recently recognized right dominant ventral frontoparietal network. Dorsal frontoparietal network and stimulus salience. Neurophysiological studies indicate that the dorsal fron￾toparietal network, which is recruited for top-down selection, is also modulated by the bottom-up distinc￾tiveness of objects in a visual scene. During a visual search task in which a monkey makes a saccade to the location of an oddball target (such as a red square among green squares), the visual response of FEF neurons dis￾criminates between the target and distracters. This sen￾sory modulation is independent of whether the monkey looks at the target, because similar effects can be obtained during covert detection71 (FIG. 5a). The fact that the colour of the oddball is uncertain (sometimes a red target is presented with green distracters, and sometimes a green target is presented with red distracters) indicates that bottom-up sensory signals primarily drive the dis￾crimination. Furthermore, it is possible that the monkey expects some ‘oddness’ in the display, a form of expec￾tancy. Other results indicate that the response of FEF neurons to stimuli is modulated by cognitive factors, such as the relationship between target and distracter stimuli. Distracter stimuli produce larger responses when they are similar to the current target or when they have served as targets in a previous session72. Finally, these effects are observed for both colour and shape, indicating that the interaction of bottom-up and top-down signals in the FEF generalizes over visual dimensions (FIG. 5a)