Although our experiment in monkeys revealed that some part of lateral PFC is important for top-down attentional control, it was not possible to more precisely localize the region(s) responsible for the deficit. In addition, because the lesions were confined to the PFC, it was unclear whether other brain regions, such as parietal cortex, were involved in top-down attentional switching in the task. In this second experiment, we sought to answer these questions in humans, using the same tasks as before, but modified slightly for the fMRI environment.
In the main experimental condition (color cueing), which was guided by top-down control, a central cue indicated the color of a peripheral grating on which the subjects performed an orientation judgment. The central cue was a red, green, or blue square. For switch trials, the color of the cue in the current trial was different from the color in the previous one. For non-switch trials, the color of the cue in the current trial was the same as the color in the preceding trial. Switch and non-switch trials occurred in random order. Because the cue was present in every trial, the contrast of switch and non-switch trials was expected to reveal brain regions involved in the updating of cue-related information. In addition, based on the results in our monkeys, we expected such endogenous cue updating to engage different regions relative to those that would be observed when targets were determined in a bottom-up fashion. To test this prediction, we included a pop-out condition in which the target grating was defined by color contrast, i.e., target selection was guided by bottom-up saliency. Specifically, the target grating was the color singleton appearing within an array of like-colored gratings. For the pop-out condition, the central cue was thus uninformative. Again, both switch and non-switch trials occurred.
The trial structure for both color cueing and pop-out trials was identical. An initial central fixation square was shown for 500 ms, followed by a 500-ms display containing a central cue and a peripheral array of five gratings. Potential targets only comprised the central three locations; the two additional extreme gratings were fixed and were included to equate sensory stimulation for the two tasks. For color cueing trials, the fixation square turned color and indicated the color of the target grating on which the subject was to perform an orientation judgment. For pop-out trials, the fixation square remained white and the target on which subjects performed the orientation judgment was the grating whose color differed from that of the remaining four gratings. On all trials, subjects indicated with a button press whether the target grating was vertical or not. The trial then ended with a 1,500-ms blank screen. The target location was randomized on each trial, irrespective of whether the trial was a switch or non-switch trial.
Color cueing and pop-out trials occurred in a blocked fashion and only one condition occurred during an fMRI run. Within individual runs, each block consisted of 40 trials, and during each block, five to ten switch trials occurred at random. Twenty normal, healthy subjects performed alternating runs containing blocks of either color cueing or pop-out trials, and fMRI data were collected using a 3T scanner. Analysis of fMRI data employed standard multiple regression methods; condition (color cueing and pop-out) and trial type (switch and non-switch) were fixed factors and participant was a random factor in a mixed-effects analysis.
First, we investigated the main effect of task on brain activation by contrasting color cueing vs. pop-out conditions. We found stronger responses during color cueing trials in several bilateral fronto-parietal regions, including the inferior parietal lobule, FEF, MFG, and inferior frontal gyrus (IFG). We did not observe any regions in which responses evoked during pop-out trials were stronger than during color cueing trials. Next, we compared switch to non-switch trials (pooled across conditions; see ). Stronger responses evoked during switch trials were observed in the left IPS and left MFG/IFG. These activations were quite extensive; smaller foci of activation included the right IPS and left FEF, as well as the left middle occipital gyrus (MOG) and bilateral inferior occipital gyrus (IOG). Finally, stronger responses for non-switch trials relative to switch trials were observed in the right insula.
Fig. 3 Results from Experiment 2. a Group activation maps displaying switch > non-switch trials across both color cueing and pop-out conditions. b Group activation maps displaying switch > non-switch trials for color cueing > pop-out (more ...)
We also probed for task by trial type interactions (), i.e., regions for which the difference of switch vs. non-switch trials was greater during color cueing trials relative to pop-out trials. These included the following frontoparietal sites: left IPS, left FEF, left MFG/IFG, and right IFG. Interestingly, several sites in visual cortex also showed this interaction effect, including left MOG and IOG, left inferior temporal gyrus, right fusiform gyrus (FG; not illustrated), and right middle temporal gyrus. No other significant interactions were observed (e.g., those involving non-switch > switch trials).
A central goal of this experiment was to determine brain regions in humans engaged during target selection based on the updating of an endogenous cue (i.e., during the color cueing task). To probe this question, we determined brain activations associated with the interaction of task and trial type. Specifically, we determined regions in which the difference between switch and non-switch trials was greater during the color cueing task relative to the pop-out task. Because both types of trials involved a change in the target grating, such a contrast isolated regions that are important for endogenous updating.
Task by trial type interactions were observed in several fronto-parietal regions, including the IPS, FEF, MFG, and IFG. As stated in “Introduction”, these regions are thought to be important sites involved in the control of attention. Our findings further reveal that these regions are also important for the updating of endogenous cue information. Importantly, the FEF, MFG, and IFG were all removed in the region ablated in our monkeys with deficits in top-down attentional control, but the IPS was not. Thus, in addition to lateral PFC, the IPS may also contribute to this function.
Task by trial interactions were observed not only in fronto-parietal “control” areas, but also in several visual regions, namely, the left MOG, left IOG, and right FG. Therefore, such visual activations were not simply due to the task performed or trial type, but depended instead on the combination of the color cueing task and switch trials. We suggest that these visual areas were the recipients of top-down signals from fronto-parietal control regions that are generated when cue information is updated. For anterior visual areas with bilateral visual inputs, these top-down signals may go to visual areas in either or both hemispheres.
Single-cell recording studies have shown that spontaneous (baseline) firing rates are 30–40% higher for neurons in areas V2 and V4 when a monkey is cued to attend covertly to a location within the neuron's receptive field (RF) in expectation of a stimulus but before it is presented there; that is, in the absence of visual stimulation (Luck et al. 1997
). This increased baseline activity, termed the “baseline shift”, has been suggested to reflect top-down signals that feed back from higher-order control areas to lower-order visual processing areas. Such a shift in baseline activity in visual cortex would presumably “sensitize” neurons with RFs at the attended location, so that when a stimulus subsequently appears at that location there would be enhanced visually evoked activity. Similar effects have also been observed in neuroimaging studies (Kastner et al. 1999
), in which “baseline” effects have been observed in human areas V1, V2, V4, and TEO. Increases in baseline activity are not only spatially specific, but also appear to depend on the type of visual feature attended to (Chawla et al. 1999
). If our interpretation is correct, then top-down signals influence visual processing not only during sustained directed attention to a stimulus’ location and features, but also during the updating of the information conveyed by endogenous cues.