**4. Discussion**

The present ERP study investigated the time course and neural correlates of object-based attention, under the assumption of left-hemispheric dominance. For this purpose, healthy, right-handed participants were presented with 3D graphic images depicting the shapes of different categories of stimuli (wooden dummies, chairs, structures of cubes) which lacked detail. They were instructed to pay attention to and detect one given target category (singularly and centrally presented) among non-targets by emitting a motor response (button press). As visible in Figure 3, three main ERP components

(N2, SN, and P300) were shown to be sensitive to selective attention in di fferent time windows, likely highlighting di fferent the cognitive processes involved in recognition of the target objects.

The anterior N2 was the first potential that was modulated by the attentive selection (225–265 ms). The amplitude of this negativity was reduced in response to the target images when compared with the non-targets. Similar results have been previously reported during go/no-go tasks. A larger N2 has been found for non-target (relative to target) item tasks that require response inhibition (no-go) in terms of both actual [45] and imagined [57] motor acts (i.e., button press). This interpretation is supported by source reconstruction studies that have localized the neural generators of the N2 in the anterior cingulate cortex (ACC [58]). The ACC has been associated with cognitive control, as shown by several imaging investigations in both healthy and clinical (i.e., Huntington's disease) individuals [59,60]. The N2 has also been proposed as an index of conflict monitoring [61] and is modulated by stimulus novelty [62], category [63,64], and mismatch [65]. In the present study, the identification of non-target images, which represented two-thirds of the stimuli, required no actual finger movements. Hence, the relative decrease in the N2 amplitude can be considered a correlate of motor inhibition [66]. For instance, in the study by Proverbio and colleagues [66], the participants were presented with four gratings of di fferent spatial frequencies briefly displayed in the four quadrants of the visual field. They were instructed to respond to a target combination of spatial frequency and space location. When compared with the non-targets, the pseudo-targets (stimuli close in spatial frequency to the target and falling within the attended quadrant) elicited both a larger frontal motor N2 and a larger negative prefrontal potential (370–430 ms). This evidence was interpreted as an index of the response inhibition and top-down cognitive control required for irrelevant information suppression. Finally, no hemispheric di fference was found here at this stage of stimulus processing (N2 time window). This result is consistent with the increased power in the theta frequency band that has been previously reported at midline frontal scalp sites during target detection tasks [67]. The medial prefrontal cortex has been proposed as a possible neural generator of this e ffect (i.e., ACC [68]).

Moving forward at the temporal level, the analyses of selection negativity (or posterior N2) response (240–280 ms) revealed an increased amplitude over occipito-temporal areas elicited by target relative to non-target stimuli. This e ffect is illustrated in Figure 5, which reports the grand average ERP waveforms recorded over posterior sites. The SN was typically obtained subtracting the posterior N2 elicited by non-target stimuli from that evoked by target stimuli. It is considered an index of visual attentive selection processes [46], as it shows sensitivity to several target stimulus features (or a combination of them), including color [69,70], orientation [71], spatial frequency [41,72], and shape [73]. Thus, in the present study, the modulation of the SN may have indicated non-spatial attention allocation towards the specific stimulus shape required for object recognition, consistent with previous evidence [74]. Furthermore, the maximum amplitude of the SN was reached over occipito-temporal scalp sites. The location was compatible with the modulation of associative visual cortex previously reported in several imaging studies on non-spatial attention [3,4].

More importantly, the modulation of the SN was specifically visible over the left but not the right hemisphere, as shown by the topographic maps of voltage distribution depicted in Figure 7 (see also Figure 6). This evidence is consistent with previous ERP studies on shape [51] and color detection [30], as well as local (vs. global) stimulus information processing [32,41] and illusory contour perception [31]. For instance, in a previous study by our research group [51], the participants were presented with images representing upright and inverted bodies (wooden dummies) and structures of cubes. They were instructed to attend to one of the two categories of stimuli (by button press), regardless of the orientation. The occipito-temporal SN was shown to be sensitive to the orientation of the human body shape, being larger in response to the inverted (than upright) body targets. This result likely indicated increased attentive processes for body recognition when presented in non-standard orientations. The negative response was also overall larger over the left than the right hemisphere. This evidence possibly suggests a predominant role of the left hemisphere in shape-related attentive selection processes. It also extends previous findings [30] on conjoined color and shape processing. These results are also consistent with those reported by Zani and Proverbio [72] in their attentional task (relative to spatial frequency). In that study, the relevant/target (relative to the irrelevant) stimuli elicited larger ERP components at both occipital (N165 and P3b) and frontal (LP, long latency positivity) scalp sites over the left but not the right hemisphere. In another study, larger negativity (N2) was elicited by the perception of illusory contours of a Kanizsa square over the left occipital regions [31], consistent with the idea of left-sided non-spatial feature selection and local (vs. global) stimulus processing [36,37].

Furthermore, the swLORETA inverse solution was applied in this study to the difference wave target minus non-target to estimate the neural sources of the EEG signals in the SN time window (240–280 ms). Several active dipoles underpinned the left-sided SN within a fronto-temporo-limbic network, associated with attentive selection processes. This included the ACC (BA 24), the right medial prefrontal cortex (mPFC, BA 11), and uncus, bilaterally (BA 28/36). Several pieces of evidence have linked the ACC and mPFC to cognitive control [75], with the former region specifically involved in performance and conflict monitoring [60,76]. Both the mPFC and uncus (included in the parahippocampal cortices) are part of the affective system of the human brain [77]. Thus, their engagemen<sup>t</sup> may also sugges<sup>t</sup> an affective response [78,79] to target stimuli that were correctly identified. More importantly, the swLORETA showed a selective engagemen<sup>t</sup> of the superior/middle temporal (STG/MTG, BA 22) and inferior frontal/precentral gyrus (aka IFJ, BA 6/9) in the left hemisphere. A negative correlation between activity in the left STS and response variability has been reported during the perception of oddball (vs. standard) stimuli in healthy (vs. ADHD) volunteers [80]. At the same time, participants with attentional disorders have shown reduced response to oddball (vs. standard) stimuli in the superior and medial temporal lobe (along with the insula and basal ganglia). The STG/MTG is also considered a key node of an amodal semantic hub, together with the temporal poles [81]. Its involvement may be explained in terms of accessing the knowledge (i.e., name [82]) of the target stimulus category (i.e., bodies, chairs). Finally, previous evidence indicated a role of the IFJ (inferior frontal junction) in the top-down modulation of the inferior temporal cortices when a target object was attended [14,19]. The engagemen<sup>t</sup> of left frontal regions (i.e., BA 6) has also been reported during executive (conflict) control, together with the ACC [83]. Overall, these findings seem to support the hypothesis of effective engagemen<sup>t</sup> of the left hemisphere in selective attentional processes required by feature recognition (i.e., shape) for objects centrally presented in the visual field [4,27,64,84].

Lastly, the effects of attention were also visible at later latencies (350–450 ms) over centro-parietal sites. An increased positive (P300) response to target stimuli was found when compared with non-target stimuli, which likely suggests the recognition of the target object. The P300 response is generally interpreted as an index of item categorization, updating of the mental representation of stimulus context, and visual awareness [47,85–87]. The P300 is typically maximal in response to stimuli that are identical to a target. Its neural generators have been estimated within the parietal (i.e., inferior parietal lobule, posterior parietal cortex) and inferior temporal regions [85]. At the same time, non-target stimuli that share some visual features with the target stimulus elicit a gradient of increasing P300s as a function of enhanced similarity [88]. Moreover, the maximum peak of this ERP component is concurrent with the RTs when an accurate but fast response is required. In our study, the average RT to targets was 425 ms, included within the P300 time window considered (350–450 ms). A strong tendency towards a faster response with the right (compared with the left) hand was also found. Previous evidence has shown that simple RTs are not often affected by the hand dominance [89]. This result is also consistent with the hypothesis of left-hemispheric dominance for action selection [90,91]. Hence, the right-handedness of the participants can only partially account for the right-hand advantage reported here. It is important to report the case of target detection during spatial attention modulation. Faster RTs have been found for stimuli occurring in the visual field ipsilateral (relative to the contralateral) to the response hand [92–94]. This difference in RTs has been ascribed to the interhemispheric transfer time through the corpus callosum. In the present study, the faster button press obtained with the right hand may have suggested a faster intrahemispheric (vs. interhemispheric) transfer between the contralateral (left) motor cortex and attention-related areas within the left hemisphere. This interpretation, which certainly requires further investigation, is consistent with the assumption of a left-hemispheric dominance in attentive processes for object selection.

A few aspects need further consideration and sugges<sup>t</sup> circumspection in the interpretation of the present results. Firstly, our experimental design likely prevented overt spatial attentional shifting (i.e., central stimuli presentation, stimuli with equal dimensions and number of non-empty pixels, and comparable distribution in the four hemi-quadrants of the visual field), and EEG epochs reporting eye movement were discharged. However, it is also true that covert attentional shifting may have occurred. The introduction of a secondary task would be useful in order to entirely disentangle the contribution of spatial and non-spatial attention [95,96]. This would also strengthen our object-based interpretation of the attention-related results. Secondly, despite a clear left-lateralization of the SN component in our experimental group, individual di fferences between participants may exist. It is, therefore, desirable for future investigations to account for such variability. Finally, it is necessary to point out that the EEG technique is characterized by a non-ideal spatial resolution [97] compared to functional neuroimaging (i.e., fMRI). This issue can be partially overcome using by high-density caps and state-of-the-art source reconstruction algorithms [98]. Many studies have shown good reliability of swLORETA, since the estimated dipoles were consistent with brain activity found in previous fMRI investigations [99,100]. Caution is still advised when reconstructed neural sources are considered, even when they are in support of main ERP findings, as was the case for the present study.
