Neural Processing Of Salience And Behavioral Relevance In Visual Attentional Networks

ObjectiveVisual objects that are salient or behavioral relevant can render them more attractive. Despite extensive research, the question of the relative contributions of salience and behavioral relevance of the stimulus in the attentional capture remains highly controversial. Theeuwes claimed that a salient stimulus initially captures attention irrespective of its behavioral relevance. On the other end of the spectrum is the contingent attentional capture hypothesis, which assumes that salient stimuli in a visual scene can be ignored when it does not have the same properties with the target. However, these two conflicting viewpoints rely on two fundamentally different experimental paradigms and the two factors are not manipulated simultaneously in both paradigms. The same argument also exists in the functional neuroimaging field. Earlier fMRI studies proposed that ventral attentional network (VAN) is sensitive to stimulus salience regardless of its behavioral relevance, but more fMRI studies suggested that the VAN was activated only by task-relevant stimuli. However, these studies did not distinguish stimulus salience and task relevance, the capturing stimuli were actually both physically salient and behaviorally relevant. Although with high temporal resolution, previous event-related potentials (ERP) studies focused on a few components which have not settled the controversy. Therefore, the neural correlates and their timings of the processing of the salience and behavioral relevance of the stimuli in the attentional networks still need to be clarified.Preserving both spatial and temporal information of neural activities, spatiotemporal patterns of ERP were used to clarify the neural processing of salience and behavioral relevance in visual attentional networks. Applying the spatiotemporal analysis of ERP to the flanker paradigm, we tried to fully dissociate the processes of salience and behavioral relevance of the stimuli. We presented either salient/non-salient or behavioral relevant/irrelevant distractors in the flanking locations, and further divided the behavioral relevance into facilitory and inhibitory. Thus, we can explore the interferences of either the salience or the behavioral relevance of the flanker (distractor) to the central target by their interactions to see how these two factors influence attention selection.MethodsTwenty neurologically intact college students (12females) aged from20to27(22.95±2.16) years participated. They were all right handed and provided written informed consent.Each stimulus picture contains3symbols in the black background (RGB:0,0,0; HSL:160,0,0). The central target was a grey symbol (RGB:192,192,192; HSL:160,0,181) of either "<" or ">" or "X", each presented in1/3trials. Two distractors of the same size as the target,2.3°high, located2.86°to the left and the right of the center of target (fixation), which consisted of double symbols of either "<" or ">" or "X" or "H". Their colors were either red (RGB:255,0,0; HSL:0,240,120) in half trials or grey as the target in the other half. Therefore, the red was defined as salient as it is more prominent in lightness and the grey is non-salient. Since a behavioral response was needed for both the symbol "<" and the symbol ">" in the target location while it did not for the symbol "X", we defined the symbol "<" and the symbol ">" at the distractor locations as facilitory, the symbol "X" as inhibitory, and the symbol "H" as irrelevant because it never appeared in the center. Each stimulus picture was assigned into one of12conditions according to the response type of the target, the flanker silence and the behavioral relevance of the distractors.Each subject was seated in an armchair facing a computer screen (75Hz frame rate) at a distance of1.0m in a quiet environment. Each of the6stimulus conditions that needed response consisted of160pictures and each of the6conditions that did not need any response consisted of80pictures. Each picture lasted200ms with the interval of black screen between stimuli of1000ms. The participants were required to response according to the direction of central target:pressed the left button or the right one to the symbol "<" or the symbol ">" respectively and inhibited response to the symbol "X". The reaction time (RT) and the accuracy were recorded. The procedure was divided into10sections. Nine breaks of30s were arranged between sections. A practice of10min. was arranged before ERP experiment. The EEG was recorded by means of an ERP system developed in our lab withthe bandwidth of [0.5,100](Hz) and the sampling rate of1000Hz. The international10-20system of19recording electrodes (FP1, FP2, F3, F4, C3, C4, P3, P4,01,02, F7, F8, T3, T4, T5, T6, Fz, Cz, Pz) was used with linked earlobes as reference. The electrode impedances were kept below10kΩ. Each epoch of event-related EEG was from100ms before to1000ms after stimulus onset. Trials contaminated with ocular, muscular or any other type of artifacts were inspected visually and rejected. Sweeps exceeding±70microvolt in any of the channels and those with incorrect performance were excluded from the ERP averaging offline. The behavioral data (RT and accuracy) was submitted to a two-way MANOVA of repeated-measure:2(the distractor salience:salient, non-salient)×3(the distractor relevance:facilitory, inhibitory, irrelevant) with SPSS13.0. The simple effect analysis corresponding to the salient and the non-salient conditions separately and a paired, two-sided t-test as post hoc for each composition of relevance conditions were further performed. Greenhouse-Geisser correction was used for the relevance factor as it has more than two levels, and corrected P values were reported. The significant level was0.05.The ERP data of each channel at every moment was entered into a three-way MANOVA of repeated-measure:2(the target response:yes, no)×2(the distractor salience:salient, non-salient)×3(the distractor relevance:facilitory, inhibitory, irrelevant). In order to distinguish the facilitory and inhibitory effects of the distractors, a reduced three-way MANOVA of repeated-measure were carried out three times for each composition of two level of the distractor relevance factor.The resultant multichannel time series of F-value was used to generate spatiotemporal patterns by an interpolation method of generalized cortical imaging technique, which was referred to as statistical parametric mapping of F-value, SPM(F) in abbreviation. Conservative lower-bound epsilon (0.5) was applied to adjust the degree of freedom of F-value for the relevance factor. The significant level was0.05.ResultsBehavioral PerformanceFor the reaction time (RT), the interaction effect was significant:F(1.85,35.21)=13.015, P=0.000. The significant simple effects of the distractor relevance existed for both the salient level [F(1.68,31.99)=40.479, P=0.000] and the non-salient level [F(1.39,26.38)=157.511, P=0.000]. The followed LSD analysis showed significant difference for each of all pair-wise comparisons between the conditions (P=0.000). Only under the facilitory condition, significant difference was found between the salient (411.91±40.39ms) and the non-salient (420.42±38.58ms) levels:t(19)=4.858, P=0.000, which suggests the contingent capture.For the accuracy (A), the main effect of the distractor relevance was significant: F (1.08,20.54)=15.388, P=0.001. The followed LSD analysis showed significant differences between the facilitory and the inhibitory conditions (P=0.001) as well as the facilitory and the irrelevant conditions (P=0.001). No significant effect was found for either the interaction or the distractor salience factor.Spatiotemporal patterns of ERP and SPM(F)Under the condition of2(response of the target:yes, no) x2(salience of the distractors:salient, non-salient) x3(behavioral relevance of the distractors:facilitory, inhibitory, irrelevant), significant target response effect occurred in occipital regions (125-150ms), frontal-parietal regions (150-200ms), bilateral temporal-occipital regions (200-250ms), frontal-parietal-occipital regions (250-350ms), temporal-parietal-occipital regions (350-600ms). Significant salience effect occurred in right occipital regions (100-125ms), bilateral parietal regions (125-150ms), bilateral frontal-parietal-temporal regions (150-200ms), bilateral frontal-parietal-occipital regions (200-300ms), right frontal-parietal-temporal regions (300-400ms), central parietal regions (400-450ms), bilateral parietal regions (500-600ms). Significant behavioral relevance effect occurred in occipital regions (150-200ms), right frontal and occipital regions (200-300ms), extensive temporal-parietal-occipital regions (300-600ms).significant interaction effect of target response and salience occurred in occipital regions and right temporal-occipital regions (100-175ms), bilateral temporal-parietal regions (175-200ms), central frontal-parietal regions (200-350ms). Significant interaction effect of target response and behavioral relevance occurred in bilateral frontal-parietal-occipital regions (200-350ms), left temporal-parietal-occipital regions (450-600ms). Significant interaction effect of salience and behavioral relevance occurred in central parietal regions (250-300ms).Under the three-way MANOVA of repeated-measure for each composition of two level of the distractor relevance factor, the comparisons between conditions for facilitory and inhibitory, significant differences occur in frontal regions(150-175ms), occipital regions(175-200ms), right frontal and occipital regions(200-300ms), extensive frontal-temporal-parietal regions (300-600ms); for facilitory and irrelevant, significant differences reveal in central parietal-occipital regions(100-150ms), bilateral parietal-occipital regions(150-200ms), right frontal and occipital regions(200-300ms), extensive frontal-temporal-parietal regions(300-600ms); for inhibitory and irrelevant, occipital regions(150-200ms) and bilateral frontal regions(350-400ms) show significant differences.ConclusionOur behavioral results support the contingent capture hypothesis which confirms that salient distractors cannot capture attention unless they own some behavioral relevance.Based on the SPM of ERP that reveals the activation of visual attention networks, we can divide the processing into five stages:the early perception, the late perception, the evaluation, the executive and the post-response. Although with less behavioral effect for the salient distractors, the salient information has accessed the VAN as well as other attentional networks, and lasted to the medial frontal regions after the response. On the other hand, the time course and neural mechanisms of the attentional capture by behavioral relevant distractors are not known clearly. Our study suggests that their processing is suppressed in the early stage of perception, indicating the role of the top-down mechanism, and the significant interaction effect of the target and the distractor relevance in parieto-occipital regions (300-350ms) may be the neural correlate of behavioral attentional capture.