启动合成人脸的整体与局部加工：一项EEG研究*

(杭州电子科技大学智慧城市研究中心，杭州 310018)

Introduction

Global and local processing is the one of the most fundamental processing in the vision system of human beings.It has been established that the global skeleton of a visual scene is generally available prior the local details (e.g.,Bar,2004;Navon,1977,2003;Xu & Chun,2009).Moreover,using Navon patterns (Navon,2003) as stimuli,studies have demonstrated that the global and local processing can be top-down biased by showing a global-or local-level-specific priming effect (e.g.,Forster,2009;Hubner,2000;Large & McMullen,2006;Robertson,Egly,Lamb,& Kerth,1993;Shedden,Marsman,Paul,& Nelson,2003;Ward,1982).This level-specific priming effect attracts the attention of many research fields,for instance,the research on face perception/recognition (e.g.,Hills & Lewis,2007,2009;Lewis,Mills,Hills,& Weston,2009;Perfect,Dennis,& Snell,2007;Perfect,Weston,Dennis,& Snell,2008;Retter & Rossion,2015;Weston & Perfect,2005;Weston,Perfect,Schooler,& Dennis,2008),particularly since it can help us uncovering whether a specific level of processing is appropriate for the interested stimulus type (e.g.,face) in an ongoing cognitive task.

By examining the level-specific-priming effect,researchers investigated whether the global/local processing affected the holistic processing of human faces (Gao,Flevaris,Robertson,& Bentin,2011) reflected by a composite-face illusion (e.g.,Rossion & Boremanse,2008;Young,Hellawell,& Hay,1987;Rossion,2018),in which for a face the top and the bottom halves were composed of different faces,processing its top half is always affected by the bottom half although it is task-irrelevant.Particularly,in a trial the participants first attended to either the global or the local level of two Navon patterns and made a “match/mismatch” judgment.Then they compared the upper halves of two composite-faces which were presented sequentially and made a “same/different” response.Gao et al.found that relative to a baseline in which the composite-faces were not primed,the global processing of Navon patterns increased the sensitivity to the incongruence between the upper and the lower halves of the composite-faces,yet the local processing did not influence the sensitivity to the incongruence.This finding provided direct evidence supporting the claim that the composite-face was processed holistically in perception.However,the neural mechanism underlying the priming effect,to our knowledge,remains unknown.

One possible answer to the neural mechanism of this priming effect lies in the relative involvement of the left and the right cerebral hemispheres (LH/RH).Particularly,a functional hemisphere asymmetry in the global versus local processing has been revealed by using Navon patterns as materials:the LH and RH tunes to the local and the global processing,respectively (e.g.,Delis,Robertson,& Efron,1986;Fink et al.,1997;Han et al.,2002;Robertson & Delis,1986;Volberg & Hubner,2004;Weissman & Woldorff,2005).In addition,previous studies (particularly using human faces as materials) have provided evidences suggesting that the global information of visual object is primarily conveyed by low spatial frequency (LSF) whereas the local details of faces are mainly transformed by high spatial frequency (HSF) (e.g.,Bar,2004;Cheung,Richler,Palmeri,& Gauthier,2008;Gao & Bentin,2011;Goffaux,Hault,Michel,Vuong,& Rossion,2005;Goffaux et al.,2011;Goffaux & Rossion,2006;Loftus & Harley,2004;Schyns,1998).Importantly,an analogous hemispheric asymmetry in processing the spatial frequencies has also been revealed:the RH is biased in discriminating the LSF and the LH is biased in discriminating the HSF (Christman,Kitterle,& Hellige,1991;Kitterle,Christman,& Hellige,1990;Kitterle,Hellige,& Christman,1992).A double filtering by Frequency theory (DFF) even proposed that the global versus local processing is mediated by relatively LSF and HSF,respectively (A.V.Flevaris,Bentin,& Robertson,2010;A.V.Flevaris,Bentin,& Robertson,2011a,2011b;Ivry & Robertson,1998;Robertson & Ivry,2000).Therefore,it is intuitive to assume that the priming effect revealed in the previous behavioral study was due to the modulation of the relative involvement of the LH and the RH in the priming phase,which further influenced the face processing.

An EEG study provided temporary evidence supporting this possibility.By using alpha (8-12 Hz) suppression as index which is suggested to reflect the elevated brain activation (i.e.,increased involvement of human brain) (Klimesch,Doppelmayr,& Hanslmayr,2007;Volberg & Hubner,2004),Bentin and colleagues found that the alpha suppression during the preparation for the global versus local processing of two Navon patterns was modulated by the selected spatial frequencies (A.V.Flevaris,et al.,2011b).In particular,a stronger alpha suppression was revealed in the RH after being primed by the LSF,yet no difference between the LH and the RH was found after the HSF.This finding suggested that after processing the LSF,the right hemisphere was more activated,which thus facilitated the processing of the global information.These results suggest that the relative involvement of the RH and the LH could be modulated by a pre-conducted priming task.

To this end,the current study tested this hypothesis by using a similar experimental design adopted in Gao,et al.(2011) while taking alpha suppression as the interested neural marker.We predicted that if the influence of the global versus local processing of Navon patterns on the composite-faces was mediated by modulating the relative involvement of the LH and the RH,then the alpha suppression during the priming interval would be modulated by the processing level of the Navon patterns.Otherwise,there would be no difference between the global processing and the local processing on alpha suppression.The result of this prediction will not only promote our understanding about the neural mechanism of the priming effect,but will also shed light on the relationship between the global/local processing and the LSF/HSF processing.

Method

Participants

Twenty-four right-handed students from the Hangzhou Dianzi University (mean age 22.5 years;15 females) took part in the experiment.All the participants had normal or corrected-to-normal visual acuity and no history of psychiatric or neurological disorders.

Stimuli,task,and design

The Navon patterns consisted of black letters presented on a gray background.Each local letter was 0.70 cm wide×1.0 cm high (0.5°×0.82° seen from a distance of 70 cm);the local letters were spatially arranged on a 5×5 grid to form a global letter that was 5.5 cm wide×7.5 cm high (4.5°×6.14°).The letters were black Helvetica bold font D,E,F,and H in all their local and global combinations,making 16 distinct Navon patterns (Fig.1).

Face stimuli were generated from 96 gray scale images of unfamiliar Caucasian faces,50% males and 50% females.The original images were first cut to form an ellipse shape 7 cm wide×10 cm high (5.7°×8.19°) that excluded exterior face features.To create composite-faces,we first divided the vertical dimension of each face into two halves,and then recombined the top and the bottom parts from same-gender faces,between which a 0.19 cm gap was left.The top half of each face was recombined with its original bottom half,as well as with the bottom half of a different face chosen at random without replacement.This resulted in 384 different aligned composite-faces in total.Each composite-face was used only once in the experiment.

In the priming task,two Navon patterns were presented simultaneously,one in the left and one in the right visual field,each with the most medial edge 0.25° from fixation.The participants were instructed to match the two displays,focusing either on the large letters,ignoring the small letters (global level),or on the small letters,ignoring the large letters (local level).

In the composite-face task,two composite-faces were presented one after the other at the center of the screen,and the participants were instructed to match the top halves of the two composites,ignoring the bottom halves.The first face on each trial was composed from the top and the bottom halves of the same face.Since we adopted a full composite-face design (Gauthier & Bukach,2007;Richler,Cheung,Wong,& Gauthier,2009;Richler,Gauthier,Wenger,& Palmeri,2008),the combination of the second composite-face defined one of four pair-type conditions used in the experiment,two congruent and two incongruent.In the congruent pairs,the top and the bottom halves of the two composites were taken either from the same faces or from different faces:top-same and bottom-same or top-different and bottom-different faces.In the incongruent pairs,either the top or the bottom half was taken from the same face,while the other halves were taken from different faces:top-same and bottom-different or top-different and bottom-same faces (Fig.2).The four composite-face conditions were mixed and presented randomly within each experimental block.The composite-face illusion was examined by comparing the accuracy between the incongruent condition and the congruent condition (congruent-incongruent).The bigger difference between the two the larger for the illusion.

Procedure

The stimuli were shown on a 17-in.CRT monitor with a vertical refresh rate of 100 Hz and a resolution of 1024 × 768 pixels.Participants were seated in a dark room.Trial timing was controlled by Presentation (Neurobehavioral Systems,Albany,CA),and is depicted in Fig.2.Each trial began with a black fixation mark that appeared at the center of screen and was presented until the onset of the first composite-face.In the priming condition,after the fixation being displayed for 500 ms,two Navon patterns were presented for 750 ms,and were followed by a 500-600 ms blank interval.In the no priming condition,only the red fixation was presented.The participants were asked to make a “match” or “mismatch” judgment of the Navon patterns and held on their response until the end of a trial.Then the first composite-face was displayed for 250 ms,and after a 250 ms blank interval the second composite-face was displayed until a response or for up to 2s.The participants responded to the composite-face task using their dominant hand,selecting between a “same” or a “different” key.After the completion of the composite-face task for 100 ms,a black question mark was displayed at the center of the screen and the participants responded to the Navon task.Accuracy,rather than speed,was emphasized.Following a blank inter-trial interval of 700-900 ms,a new trial began automatically.

Figure2Illustrationofthetimecourseofatrialinthecurrentstudy

There were 48 trials in each condition,resulting in a total of 192 trials per priming-level block.Each block was split into two sessions of 96 trials each,with a break of 5 min between sessions.Besides,EEG was recorded while participants closed their eyes for two minutes prior to the experiment and two minutes following the experiment.Alpha activity during this period was averaged and based on its topography the electrodes showing the strongest alpha activation during the idle state were selected for future analysis.Before the experiment,the participants were given a two-stage practice to ensure they were familiar with the procedure of the experiment.Particularly,the participants first were trained to get used to responding to the Navon task only and composite-face task only with the corresponding keys in two separate sessions (16 trials each),the order of which were counterbalanced.Then the two tasks were combined together and the participants took part in at least 32 trials to be familiar with the formal procedure.

The entire experiment lasted approximately 90 min.The participants finished the whole experiment in two different days considering the whole experiment was rather long.One time they finished the global-priming and the local-priming blocks,the orders of which were counterbalanced.The other time they finished the no-priming block.The orders of no priming condition and priming condition were also counterbalanced across participants.

EEG recording and analysis

The EEG analog signals were recorded continuously (from DC with a low-pass filter set at 100 Hz) by 64 Ag-AgCl pin-type active electrodes mounted on an elastic cap according to the extended 10-20 system,and from one additional electrode placed at the tip of nose.All electrodes were referenced during recording to a common-mode signal (CMS) electrode between POz and PO3 and were subsequently re-referenced digitally.Eye movements and blinks were monitored using bipolar horizontal and vertical EOG derivations via two pairs of electrodes,one pair attached to the external canthi,and the other to the infraorbital and supraorbital regions of the right eye.Both EEG and EOG were digitally amplified and sampled at 256 Hz using a Biosemi Active II system (www.biosemi.com).The digitized EEG was saved and processed off-line.

Data were analyzed using Brain Vision Analyzer software (Brain Products) and Matlab routines.Raw EEG data was initially 0.5 Hz high-pass filtered (24 dB) and re-referenced off-line to the tip of nose.Eye blinks and movements were corrected using a regression procedure.Remaining artifacts exceeding 100?V in amplitude were rejected.Artifact free data were then segmented into epochs ranging from 1100 ms before to 2900 ms after the onset of Navon patterns for all conditions.In this case,the EEG from the onset of the fixation until ～500 ms after the onset of the second composite-face was included,which allowed us to examine the alpha suppression throughout a whole trial (see Volberg & Hubner,2004 for a similar examination).To analyze oscillatory cortical activity during the Navon-face priming task,we adopted a complex Gaussian Morlet wavelet method using an adaptive time window of 5 cycles for each frequency (ΔT= 5/f ).This procedure was applied to frequencies ranging from 1 to 20 Hz in steps of 1 Hz,and the results were baseline-corrected by subtracting the mean power value of the time between -900 and -700 ms before the onset of the Navon patterns (or -400 to -200 before the onset of the fixation,to avoid the edge effect in the wavelet analysis and thus safely check the possible alpha suppression even before Navon patterns) from the post-stimulus power.Induced oscillatory activity was calculated by applying wavelet analysis to individual trials and averaging the time-frequency plots.

Results

Behavioral results

PerformanceintheNavonprimingtask

Accuracy was fairly high for the Navon priming task (92%).There was no difference between global and local processing of Navon patterns on accuracy [92% vs.93%,F <1] or RT [177 ms vs.177 ms,F<1].The non-significant main effect of priming level was due to the fact that the participants responded to the Navon task at the end of a trial.

PerformanceintheComposite-facetask

Only trials on which the response to the prime was correct were included.Sensitivity to face identity was determined by d’.As shown in Fig.3,across the global-,local-,and no-priming conditions,d’ was consistently higher in the congruent than in the incongruent condition.Because we intended to examine the composite-face illusion in different priming levels,we thus computed the degree of composite-face illusion in each priming level by subtracting the d’ of incongruent condition from the congruent condition (see Fig.3).A t-test was ran under each priming condition to examine whether the composite-face illusion was significantly larger than 0.A one-way repeated ANOVA by taking Priming-level (global-,local-,and no-priming) as the within-subject factor was followed to examine the influence of the priming on the composite-face illusion.As shown in Fig.3b,the composite-face illusion in the global-priming and the no-priming condition was much higher than 0,and the local-priming condition.

Confirming these observations,the t-test showed that the composite-face illusion in the global-priming condition [0.72;t(23)=8.06,p<0.001] and the no-priming condition [0.64;t(23)=4.62,p<0.001] was significantly larger than 0,yet the composite-face illusion in the local-priming condition was only marginally larger than 0 [0.29;t(23)=1.74,p=0.096].The one-way ANOVA revealed a significant main effect of Priming-level [F(2,46)=3.69,MSE=0.35,p<0.05,partialη2=0.14].Post hoc contrasts showed that there was no difference between global-priming and no-priming [p=0.58],but both were larger than local-priming [p<0.06 for global-priming,p<0.05 for no-priming].

Figure3Meand’ (a)andthedegreeofthecomposite-faceillusion
(b)forthecompositefacetaskadoptingthecompletedesigncalculation.
(Errorbarsshowstandarderror)

Alphasuppression

Following inspection of the alpha distribution,we focused on the parieto-occipital sites (PO7/PO8,P9/P10,and O1/O2),in which the alpha power was the greatest during the rest.The no-priming condition was taken as the baseline,the alpha band power of which was subtracted from the global- and the local-priming condition.In this way,we could measure of the relatively pure priming-effect due to the processing of Navon patterns and remove the potential contamination of the already existed hemispheric asymmetry during the idle state on the alpha band.We found that both global-priming and local-priming led to a stronger alpha suppression after the onset of the Navon patterns than the no-priming.This trend initiated from 200 ms after the onset of the Navon patterns and lasted until prior the onset of the first composite-face.Moreover,the suppression in both conditions exhibited a stronger alpha suppression in the RH than in the LH.

A2 (Priming-level:global-priming versus local-priming) ×2 (Hemisphere:LH versus RH)×3 (Site:PO7/PO8,P9/P10,and O1/O2) within-subject ANOVA was performed on the mean power change in the alpha band for the time window 200-1100 ms from the onset of Navon patterns in which the alpha suppression was the most conspicuous (see Figure 4).Confirming the observation,the alpha suppression was significantly stronger in the RH than in the LH [F(1,23)=19.53,MSE=1163.11,p<0.001,partial η2=0.46].However,neither the main effect of Priming-level [F(1,23)=1.11,MSE=8.51,p=0.30,partial η2=0.05] nor any interaction related to the Priming-level [allFs<1,p>0.70,partial η2<0.02] was significant.The main effect of Site was not significant [F(2,46)=1.08,MSE=24.83,p>0.30,partial η2=0.05],but it was modulated by Hemisphere [Hemisphere X Site interaction:F(2,46)=3.38,MSE=11.41,p<0.05,partial η2=0.13].One-way ANOVA elaborating this interaction revealed that there was no difference among the three site-pairs in the LH [F<1],but there was a significant main effect in the RH [F(2,46)=3.25,MSE=19.46,p<0.05,partial η2=0.12].Post hoc contrasted revealed that the alpha suppression was slightly larger on PO7/8 than on O1/2 [p=0.081].

Following Flevaris et al.(2011b),we also examined whether the preparatory stage of the composite-face was modulated by the Navon priming,by testing the alpha power from 200 ms after the offset of the Navon until the onset of the Navon patterns (950-1300 ms).Except for a significant main effect of Hemisphere showing that the alpha suppression was stronger in the RH than in the LH [F(1,23)=11.85,MSE=464.95,p<0.005,partialη2=0.34],the ANOVA did not found any significant effect.

Discussion

The current study explored the neural mechanism underlying the priming effect of the global versus local processing on the composite-faces,by using alpha suppression as the interested index.Adopting a similar design as in Gao,et al.(2011),we found that the global versus local processing modulated the composite-face illusion by reducing the illusion in the local-priming condition,corroborating our claim that there is a perceptual factor in the composite-face illusion.As to the EEG results,both the global and the local processing induced strong alpha suppression in the LH and the RH;however,the alpha suppression was not modulated by the processing level of the Navon patterns,exhibiting significantly stronger alpha suppression in the RH than in the LH regardless of the priming level.These results imply that the priming effect of the global versus local processing is not due to the modulation of the relative involvement of the LH and the RH before the onset of the composite-face.

Figure4Spectralamplitudesintherighthemisphere(electrodeP10,PO8andO2)shownforthedifferencebetweenglobal-primingandno-priming(a&b),aswellasthedifferencebetweenlocal-primingandno-priming(c&d).

We found a significant RH advantage which lasted through the whole blank interval between the Navon patterns and the composite-faces.This phenomenon suggests that the RH is always more strongly activated than the LH.This suggests that the RH is always ready to process the global information regardless of the priming level.Consequently,there was no difference ond’ between the global-priming and the no-priming condition,since our human brain has a RH processing advantage in general (for a detailed discussion see Flevaris,et al.,2011b).

Facing this implication,the possibility that the local-priming led to a higher degree of involvement of the LH is a rather reasonable answer to explain why the composite-face illusion was dramatically reduced after the local-priming.The null effect of the priming level on the alpha suppression,therefore,may be because the current design or index is not sensitive enough to reflect this hemispheric modulation.Indeed,compared to a blank interval of at least 2024 ms was between the two tasks in Flevaris et al.(2011b) which demonstrating the hemispheric modulation,the blank interval of the current study between the Navon task and the composite-face task was rather short (550 ms in average).In Flevaris et al.(2011b),a very strong alpha suppression particularly in the RH,which was greater than the LH,was found during and at least 500 ms after the exposure of the Gabor patches (containing the LSF and the HSF),which did not exhibit any modulation on the two hemispheres.However,this suppression dissipated as the time passed by,and the hemispheric modulation by processing the spatial frequency was found at a later stage (at least 1024 ms after the offset of the Gabor patches).Therefore,it is possible that the short blank interval of the current study with a strong alpha suppression in this period veiled the hemispheric modulation.However,the available evidence did not support this possibility.In the same study showing the modulation,similar to the current study,they find that the global/local processing of Navon patterns did not modulate the alpha suppression during and 500 ms after the offset of the Navon patterns,but found a greater suppression in the RH in both conditions.More importantly,examining the same time period in which the hemispheric modulation was demonstrated in the spatial frequency condition,a null-effect of the modulation of the global/local processing was revealed.Therefore,although we should be cautious in concluding that the relative involvement of the LH and the RH is not modulated by the global versus local processing,this current finding,to some extent,is fairly solid.

As a matter of fact,in line with Flevaris et al.(2011b),the current finding adds extra evidence supporting the DFF theory of hemispheric specialization for spatial frequency processing.This theory claims that there is a direct causal relationship between the spatial frequency selection and the global versus local perception:visual attention selects and is directed to relatively LSF by the RH and relatively HSF by the LH.Therefore,the top-down attentional selection of the spatial frequency mediates the global and local processing;and as revealed in Flevaris et al.(2011b),the selected LSF/HSF could modulate the alpha band (8-12Hz) during preparation for global versus local processing,but not vice versa.In the same vein,since the first task in the current study was a global/local task,a non-modulation result on the preparatory phase of the composite-faces thus fits the prediction of the DFF theory.

In addition,while the behavioral results of current study replicated previous finding that the global versus local processing modulated the composite-face illusion in general (Gao,et al.,2011),we actually found the opposite results on the direction of the priming.While our previous study showed that the global-priming augmented the composite-face illusion yet no difference between the local-priming and the no-priming,the current study found that there was no difference between the global-priming and the no-priming,yet the local-priming significantly lowered the composite-face illusion.The current finding is consistent with Weston and Perfect (2005),in which they found that processing Navon patterns for 3 minutes at the local level helped overcome the detrimental effect of holistically processing the composite-face,but no difference between the global-priming condition and the no-priming condition.Since the result of the non-priming in our previous study actually was the first 16 participants in the current experiment,the factors leading to the different result patterns thus laid in the difference between the priming conditions.One distinguished difference existed between the two studies on the Navon task:the Navon patterns of the current study were always displayed for 750 ms and the participants had to hold the response until the end of a trial;however,the participants in the previous study made a response immediately once they knew the answer and the Navon patterns were disappeared at the initiation of the response.Although the processing time of the Navon patterns was largely equal since the current exposure time of the Navon patterns was set based on the RT of the previous behavioral study,the task load was higher in the current study than in the previous one.To be specific,in the current study the participants needed to hold the response to the Navon task and some participants may even store the Navon patterns into visual working memory,which was required for top-down control and exhausted resources.In contrast,in the previous study once the participants responded to the Navon task they could release the resource devoted to the Navon task and prepared for the composite-face task.Supporting this possibility,on the hand,the task load during the priming phase was also high in Weston and Perfect (2005) because the participants needed to first remember a set of faces before the priming task and hold it in memory throughout the whole priming phase.On the other hand,the overall accuracy of the current EEG study was slightly lower in the local-priming condition (d’=3.12) than in the global-priming condition (d’=3.14),which may because the processing load was higher in the low-priming condition than in the high-priming condition.However,in the previous behavioral study of Gao et al.(2011) a reversed pattern (d’=2.66 for global-priming,and d’=2.81 for local priming) was found.Future work may need to elucidate how the processing load affected the priming.