**1. Introduction**

It is known that visual attention can be consciously directed towards a selected location in space (spatial attention [1,2]), a stimulus as a whole (object-based attention [3–5]), or specific features of a stimulus (feature-based attention [6–8]). While both behavioral and neural correlates of spatial and feature-related attentional processes have been extensively investigated [9–11], attentive selection for objects requires further consideration. For instance, a processing bias (i.e., increased accuracy and faster response), together with the enhanced engagemen<sup>t</sup> of fronto-parietal regions (i.e., frontal eye-field, posterior parietal cortex), was reported for stimuli occurring within a target location [12]. Several studies have also shown that attention directed towards single (or combined) stimulus features (i.e., color, orientation, motion, spatial frequency) leads to increased activity within the visual areas sensitive to the target physical trait (i.e., V4 for color; MT+ (middle temporal) for motion [8,13]).

In the case of attention on an object, a few studies have shown an enhanced engagemen<sup>t</sup> of the extrastriate visual cortices for target stimulus processing [3]. At the same time, the neural mechanisms and relative brain regions underlying this top-down modulation of the occipito-temporal regions are not ye<sup>t</sup> entirely clarified [4,14]. In two classic fMRI studies by O'Craven and colleagues [3], the participants were presented with stimuli depicting a transparent face superimposed over a house, with one stimulus moving and the other stationary. They were instructed to press a button when a target attribute (i.e., face, house, or motion) was consecutively repeated. Attention directed towards faces, houses, or visual motion resulted in increased activity in the FFA (fusiform face area), PPA (parahippocampal place area), and MT (middle temporal), respectively [15–17]. This top-down bias signal for the attended attribute also led to enhanced engagemen<sup>t</sup> of the areas associated with the processing of the task-irrelevant attribute. This evidence suggests the occurrence of attentive selection processes for the whole object. Similar results were reported in a subsequent study by Serences and colleagues [4], in which two streams of superimposed faces and houses morphed into the next stimuli at a changing rate of 1/s. Target stimuli required the participants to maintain their attention within the current stream or to shift it toward the other stream. Attention directed to faces and houses led to enhanced activity in the right lateral fusiform gyrus and bilateral medial fusiform gyrus, respectively. Additionally, transient shift-related activity was shown in the right superior frontal sulcus/precentral gyrus and the medial superior parietal lobule, suggesting a role of these regions in non-spatial attentional control processes.

More recently, in a MEG (magnetoencephalography) study by Baldauf and Desimone [14], the participants were shown two streams (spatially overlapping) of faces and houses tagged at different presentation frequencies and instructed to attend to one of them (during target detection). Selective attention enhanced the functional connectivity between the stimulus sensitive visual regions (FFA and PPA) and the inferior frontal junction (IFJ, identified using an attention-related fMRI localizer) at the specific tagging frequencies. Increased phase coherence was also visible in the higher gamma range. The phase lag (20 ms) between frontal and temporal areas indicated the IFJ as a potentially key region for top-down modulation of object-based (and feature-based) attention [18,19]. Using a different approach, Valdes-Sosa and colleagues [20] showed evidence for non-spatial, object-based attention selection at an early stage of vision. The authors presented the participants with two superimposed sets of red and green dots. They were instructed to detect coherent motion displacements (in cardinal directions) of one set of dots, ignoring the other one. The two sets could also be stationary or in rigid rotation around the central fixation point in both the same and opposite directions. In the latter baseline condition (perceived as two transparent superimposed surfaces), unattended (rather than attended) stimuli resulted in a suppression of the posterior N1 and P1 components related to the motion onset. This effect was reduced or absent when the two sets were perceived as a single object, while the attended (than unattended) stimuli elicited an increased selection negativity (SN) response.

Another issue that needs more in-depth analysis relates to the roles of the left and right hemispheres in selective attention processes. A right asymmetry has been shown during tasks involving overt and covert attentional shifting to selected spatial locations in healthy and clinical (i.e., neglect syndrome) populations [21–24]. A greater engagemen<sup>t</sup> of the right hemisphere has also been shown during tasks requiring sustained attention [25,26]. At the same time, evidence from several studies seems to sugges<sup>t</sup> left-lateralized neural substrates underlying focused attentive selection [27–31]. A case in point is the ERP (event-related potential) study by Proverbio and colleagues [30]. The participants were presented with images of familiar objects and animals that were associated or not associated with the prototypical color. They were instructed to recognize either the shape or color of the stimuli, ignoring the other trait. Target stimuli characterized by the prototypical (vs. unassociated) color/shape combination elicited a larger N2 component (or SN) over posterior sites, during the attention to color condition only. This effect was found over the left but not right hemisphere, as also confirmed by the topographic map of voltage distribution computed on the difference wave (associated minus unassociated targets). These results likely showed specific involvement of the left occipito-temporal cortex for conjoined color and shape processing of real objects.

Moreover, Milham and colleagues [28] showed hemispheric differences in attentional control based on the response (vs. non-response) conflict level. The participants were engaged in a Stroop task during fMRI scanning. The incongruent color words could give rise to either an eligible (color in the response set) or an ineligible (color outside the response set) response. Greater engagemen<sup>t</sup> of the right fronto-parietal network (i.e., anterior cingulate cortex (ACC), superior, inferior, and middle frontal gyrus, superior parietal lobule) was reported in response to incongruent-eligible (vs. neutral) stimuli. At the same time, both incongruent types of stimuli (relative to neutral stimuli) elicited greater activity in the left hemisphere (i.e., middle frontal cortex, precuneus, and superior parietal lobule), which was likely associated with attentional control at non-response (semantic and phonological) levels.

These findings also appear to be consistent with the left-lateralized brain network reported for local (vs. global) stimulus processing (i.e., Navon stimuli [32–37]) and perception of high (vs. low) spatial frequencies [38–41]. In this vein, Martínez and colleagues [41] reported a larger SN component in response to target (vs. non-target) spatial frequencies (black and white checkboard patterns). The discrimination of high frequencies (5 cpd, cycles per degree) elicited an enhanced SN over the left (compared to the right) hemisphere, while the opposite result was reported for low frequencies (0.8 cpd). In a previous study by Yamaguchi and colleagues [36], the participants were instructed to recognize a target letter (within compound stimuli) at the hierarchical level, indicated by a pre-stimulus cue. The local and global targets elicited an increased N2 response (250–350 ms) over the left and right hemispheres, respectively. The attentional shift elicited by the local (rather than global) cue also resulted in a larger negative response (starting at 240 ms after the onset) at left (vs. right) scalp sites, and vice versa. These findings are in accordance with attentive-selection-related brain asymmetry reported by Fink and colleagues [37]. In that study, greater engagemen<sup>t</sup> of the left inferior occipital cortex was found during attention directed to local stimulus features, while globally directed attention activated the right lingual gyrus. Moreover, the number of attentional switches between hierarchical levels (in a divided attention task) co-variated with the activity in the left posterior aspects of the superior temporal gyrus and in the right temporoparietal–occipital junction. This evidence suggests a role of the temporoparietal regions in attentional control for local/global processing, consistent with previous evidence from unilateral brain-damaged patients [42,43]. Finally, partially conflicting evidence was reported in the ERP study by Johannes and colleagues [44]. The authors presented the participants with images representing hierarchically composed non-linguistic stimuli during a divided attention task (detection of a target stimulus at both local and global levels). Target stimuli elicited a larger posterior negative component (Ne, 250–500 ms) which was larger over the left than the right hemisphere for both local and global targets. Global non-targets (distracters) led to an earlier Ne onset during local target processing, suggesting di fferent mechanisms for local/global analysis. At the same time, the asymmetrical distribution on the Ne possibly indicated a predominant role of the left hemisphere in hierarchical stimulus processing.

The present study aimed to further investigate the neural mechanisms underlying object-based visual attention [3,4], under the assumption of a left-hemispheric advantage [28,44]. The EEG (electroencephalography) technique was used, due to its high temporal resolution. Several subprocesses occurring during a target detection task have been previously revealed, indexed by di fferent ERP components modulated (in amplitude and latency) by selective attention. This includes the frontal N2, occipito-temporal N2 (or SN), and centro-parietal P300 responses. These components are interpreted as an index of response inhibition [45], attention allocation [46], and stimulus categorization [47], respectively. Evidence of an earlier e ffect of visual attention on stimulus processing has also been reported, opening a debate that remains unresolved. While a few authors have claimed an impact of visual attention starting at the level of the extrastriate cortex (i.e., P1 component at 70–75 ms [48]), other authors have reported evidence of a prior modulation of the striate cortex as well (i.e., C1 component at 40–60 ms [49,50]).

Here, 3D graphics were used to create images of three visually comparable (by percentage of non-empty pixels, spatial distribution, etc.) categories of stimuli (wooden dummies, chairs, structures of cubes), which lacked detail. The participants were presented with the stimuli individually displayed at the center of the screen during EEG recordings. At the beginning of each run, they were told which target category required a motor response (button press with the index finger) when perceived between non-target stimuli [51]. Modulation of the amplitude of the N2 (decrease) and P300 (increase) components in response to correctly identified targets (relative to non-targets) was expected over frontal and centro-parietal sites, respectively [45,47]. Moreover, a left asymmetric distribution of the occipito-temporal selection negativity (N2 to target minus non-target) would sugges<sup>t</sup> a predominant role of the left hemisphere in selective visual attention towards objects [30,51]. This hypothesis was also further investigated by applying the standardized weighted low-resolution electromagnetic tomography (swLORETA) inverse solution to estimate the neural sources in the SN time window.

#### **2. Materials and Methods**
