Building Blocks of Visual Working Memory:Objects,Features,or Hybrid？*

SONG Chao　LIU Wan-yi　LU Xi-qian　GU Quan*

(Department of Psychology,Zhejiang University,Hangzhou 310007,China)

?

Building Blocks of Visual Working Memory:Objects,Features,or Hybrid？*

SONG Chao**LIU Wan-yi**LU Xi-qianGU Quan***

(Department of Psychology,Zhejiang University,Hangzhou 310007,China)

Introduction

Visual working memory(VWM)stores a very limited set of information(for reviews see Brady,Konkle,& Alvarez,2011;Eriksson,Vogel,& Nyberg,2015;Fukuda,Awh,& Vogel,2010;Luck & Vogel,2013),yet plays a critical role in many cognitive activities(e.g.,Cowan et al.,2005;Unsworth,Fukuda,Awh,& Vogel,2014;Vogel,McCollough,& Machizawa,2005).Consequently,extensive studies have investigated the mechanisms of VWM in the last 15 years(see Baddeley,2012;Luck & Vogel,2013;Luria,Balaban,Awh & Vogel,2016;Suchow,Fougnie,Brady,& Alvarez,2014 for reviews).However,as a fundamental question,the unit of object representation in VWM remains unresolved.Currently,mainly three distinct views exist that claim that objects(object-based view)(e.g.,Vogel,Woodman,& Luck,2001;Xu,2002b),features(feature-based view)(e.g.,Fougnie & Alvarez,2011;Vul & Rich,2010;Wheeler & Treisman,2002),or both objects and features(hybrid view)(e.g.,Brady et al.,2011;Shen,Yu,Xu,& Gao,2013)form the building blocks of VWM.

Early studies favored the object-based view of VWM.The seminal work of Luck and Vogel(1997)found that memory accuracy for multi-featured objects was comparable to single-featured objects,and the memory accuracy for bi-colored objects was comparable to single-colored objects.Therefore,VWM performance is largely determined by the number of memorized objects,and not the number of features(Luria & Vogel,2011;Vogel & Machizawa,2004;Vogel et al.,2001).Although later studies showed that VWM performance is constrained by both the number and the complexity of the memorized objects(e.g.,Alvarez & Cavanagh,2004;Eng,Chen,& Jiang,2005;Z.Gao et al.,2009;Luria,Sessa,Gotler,Jolicoeur,& Dell’Acqua,2010;Xu & Chun,2006),these studies all implicitly or explicitly agreed that visual information is stored in VWM in an object-based manner.

However,some studies challenged this object-based view and proposed that features are the basic information unit in VWM.Researchers revealed that the number of features contained in the memorized objects influences VWM performance:Increasing the number of to-be-remembered features for the memorized objects significantly impaired VWM performance(e.g.,Oberauer & Eichenberger,2013;Xu,2002a).Moreover,the critical finding that bi-colored objects yield a similar performance as single-colored objects could not be re-established by later studies.Instead,researchers consistently revealed a cost for memorizing bi-colored objects compared to the same number of single-colored objects(e.g.,Olson & Jiang,2002;Parra,Della Sala,Logie,& Morcom,2014;Wheeler & Treisman,2002).Some researchers argued that the colors in bi-colored objects were stored as individual features(e.g.,Parra,Cubelli,& Della Sala,2011;Wheeler & Treisman,2002).Finally,according to the object-based view,the storage precision in VWM of two distinct feature dimensions contained in an object(e.g.,simple colors and orientations)should be correlated.However,such predicted correlations have not been demonstrated(Bays,Wu,& Husain,2011;Fougnie & Alvarez,2011).

It is of note that neither the object-based nor the feature-based view can explain all the available findings in VWM.For instance,although there are significant costs in memorizing multi-featured objects compared to single-featured objects(e.g.,Oberauer & Eichenberger,2013;Wheeler & Treisman,2002),performance is better when memorizing multiple features of the same object than when memorizing multiple features across different objects(e.g.,Xu,2002a,2002b).To reconcile these seemingly inconsistent results,researchers have attempted to propose hybrid models of VWM,arguing that both features and objects are the building blocks of VWM.Wheeler and Treisman(2002)pioneered this approach,proposing a dual-storage mechanism for VWM.They suggested that each feature had its own storage buffer within VWM;if the task requires,the binding information between the features would also be maintained,but it would not compete for the limited storage space for the constituent features.Recently,Brady,Konkle,and Alvarez(2011)proposed that the storage unit in VWM is a two-leveled feature bundle exhibiting a hierarchical structure.The top level of the bundle is an integrated object representation,while the bottom level contains distinct and relatively independent features.Finally,Cowan,Blume,and Saults(2013)recently suggested that attention at the encoding phase of VWM determined the representation format in VWM.If attention allowed a complete encoding of the to-be-memorized objects,the multi-featured objects were stored as integrated objects;if the encoding was not complete,then at least one feature was kept in VWM.Therefore,feature and object representations could co-exist in an equal status,exhibiting a flat structure as opposed to a hierarchical structure.However,all these hypotheses are post-hoc explanations for the available data.To our knowledge,there is no direct empirical evidence demonstrating the co-existence of features and integrated objects within VWM in one study.The current study attempts to address this issue by taking advantage of an object-based encoding(OBE)phenomenon revealed in our previous VWM studies(e.g.,Z.Gao,Li,Yin,& Shen,2010;Z.Gao,Yu,Zhu & Shen,2016;Shen,Tang,Wu,Shui,& Gao,2013;Yin,Gao,et al.,2012;Yin,Zhou,et al.,2012;Zhang,Shen,Tang,Zhao,& Gao,2013).

The OBE of VWM addresses how perceptual information is extracted into VWM.Our recent behavioral(T.Gao,Gao,Li,Sun,& Shen,2011;Z.Gao et al.,2016;Shen,Tang,et al.,2013;Yin,Zhou,et al.,2012;Zhang et al.,2013)and event-related potential(ERP)studies(Z.Gao et al.,2010;Yin et al.,2011;Yin,Gao,et al.,2012;Zhou et al.,2011)have revealed converging evidence that,like object-based information selection in visual perception,VWM encodes the basic visual features(e.g.,color,shape)via an object-based manner(but see Woodman & Vogel,2008).A typical characteristic of OBE in VWM is that for a multi-featured object,even when only one feature has to be memorized,the other irrelevant features are also automatically extracted into VWM.This automatic OBE of VWM is fairly robust:It has been revealed for avariety of memory loads(e.g.,2,6,and 8 objects)(Shen,Tang,et al.,2013;Yin,Zhou,et al.,2012),feature dimensions(Z.Gao et al.,2010;Yin,Gao,et al.,2012),encoding times(100ms vs.1000 ms)(Shen,Tang,et al.,2013),probability of irrelevant change(50%,20%,and 16%)(Shen,Tang,et al.,2013),and age groups(6,22,and 70 years old)(Zhang et al.,2013).

In the current study,we used OBE to examine whether both feature-based and object-based representations exist in VWM.We used oriented color bars as the stimuli of interest,and orientation as the target dimension.To preview the results,Experiments 1 and 2 revealed that there were feature-based representations for orientations in VWM;Experiment 3 further demonstrated that both relevant orientation and irrelevant color were extracted into VWM and stored as an integrated object in VWM.Overall,the results support a hybrid view for the VWM unit.

Experiment 1:Feature-Based

Representation Exists in VWM

According to the OBE of VWM,when only the orientation of a colored bar has to bememorized,the irrelevant color will also be extracted into VWM.Whether or not the orientation representation could be stored independently from the color representation after the completion of OBE in VWM has not been examined in previous OBE studies(e.g.,Z.Gao et al.,2010;Shen,Tang,et al.,2013;Yin,Gao,et al.,2012;Yin,Zhou,et al.,2012;Zhang et al.,2013).A positive answer would provide direct evidence for a feature-based representation in VWM.

To explore this question,we probed a newly revealed similarity effect for storing single features in VWM(e.g.,Johnson,Spencer,Luck,& Schoner,2009;Lin & Luck,2009).The similarity effect indicates that VWM performance for remembering a minimum of two similar visual objects(e.g.,squares with different shades of red)is significantly better than that for the same number of dissimilar ones(e.g.,red,green,and blue squares).This finding is surprising because it is contrary to the common knowledge that similar stimuli at encoding usually decrease memory performance(e.g.,Conrbad,1964).To understand the underlying mechanisms,researchers have demonstrated that the similarity effect is invulnerable to manipulations of the tested materials(e.g.,colors,orientations,shapes)(Johnson et al.,2009;Mate & Baqués,2009;Sanocki & Sulman,2011;Shapiro & Miller,2011),and the display manner of the to-be-memorized stimuli(simultaneous vs.sequential)(Johnson et al.,2009).Although the exact mechanisms for the similarity effect are still under debate(e.g.,Johnsonet al.,2009;Shapiro & Miller,2011;Wei,Wang,& Wang,2012),it is generally accepted that this effect has a tight relation with the representation interaction at the maintenance phase of VWM.Currently the most successful explanation has been offered by Wei et al.(2012),given that their model could predict when and how the visual similarity improves as well as impairs the performance.These authors built a continuous network model for VWM retaining single-featured objects,and demonstrated that the representations of similar features in VWM merged into an intermediate representation at the maintenance phase of VWM.

Because the visual similarity is measured in terms of value distance between items in the feature representation space,the similarity effect can be taken as a property for visual features and previous VWM studies investigating the similarity effect hence only used single-featured objects.In the current experiment,we presented the participants three double-featured objects(i.e.,color and orientation),and examined whether irrelevant colors,which are automatically selected into VWM,could impair the similarity effect for orientation in VWM.As previous studies implied that the display manner did not affect the similarity effect(e.g.,Johnson et al.,2009),we presented the stimuli simultaneously.If orientation was stored independently from color,then the similarity effect for orientation should not be impaired.However,if color and orientation were stored directly as integrated objects,then the similarity between the stored representations should be significantly diminished by the extra colors.

To examine the influence of extra colors on the similarity effect,Experiment 1 contained a multi-color block and a single-color block.In the multi-color block,the participants were required to memorize the orientations of bars in different colors,whereas in the single-color block,the participants were required to memorize the orientations of black bars,which served as the baseline.

Methods

Participants

Twenty undergraduate or graduate students at the Zhejiang University(11 male;aged 18 -25 years)participated in the experiment with signed consent forms.All had normal color vision and normal or corrected-to-normal visual acuity.Ten participants(5 male)took part in the single-color condition(i.e.,black bars);the other ten participants(6 male)took part in the multi-color condition(i.e.,colored bars).The study was approved by the Research Ethics Board of Zhejiang University.

Stimuliandapparatus

For the black bars(1.6° height × 0.4° width of visual angle),we used the stimuli adopted in Johnson et al.(2009).The bars spanned the interior of a small white circle with a diameter of 1.6°(see Fig.1).We defined the horizontal direction as 0°,and chose the orientations starting from 2° with a 5° step,resulting in 36 different orientations.Six distinct colors were used:black(0,0,0),red(255,0,0),green(0,255,0),blue(0,0,255),cyan(0,255,255)and magenta(255,0,255).

The memory array consisted of three bars with different orientations.In the single-color condition,all items were black;in the multi-color condition,each bar had a distinct color that was randomly selected from the color pool(black was excluded).The three bars were displayed at least 90° apart on the circumference of the virtual circle,which was centered at fixation and had a radius of 2.5°.In one half of the trials,all three to-be-memorized items had similar orientations,among which the two closest orientations were 15°,20° or 25° apart(e.g.,2°,17°,or 37°).To ensure that the orientations from the memory items would not be reusedby the probe when a change took place(see Procedure for details),the differences between two pairs of nearest orientations could not be both 15° at the same time(e.g.,2°,17°,and 32° would not be presented simultaneously in the memory array).In the other half of the trials,the three memory items had dissimilar orientations,each differed by at least 45° from the others.

Procedure

A 2(similarity:similar vs.dissimilar orientations)× 2(color-condition:single-color vs.multi-color)mixed design was adopted,in which similarity was the within-subjects factor while color-condition was the between-subjects factor.A change detection task was used(Fig.1).To prevent verbal encoding of the memorized items,each trial began with three randomly generated digits presented in the screen center for 500 ms,and the participants were instructed to repeatedly enunciate the digits during a trial.Then a 1000-ms fixation stimulus was presented,followed by the 500-ms memory array and a 1000-ms delay interval.Finally,a probe was presented at one of the positions occupied in the memory array until a responsewas provided.Participants judged whether the orientation of the probe was changed compared to the one that had appeared at the same location in the memory array,by pressing“F” on a computer keyboard if the orientation changed or “J” if not.In 50% of the trials,the orientation changed by rotating the memorized orientation of the item at that location by 30°(clockwise or counterclockwise)while its color remained constant.Participants were instructed to make unhurried responses to ensure accuracy,and were required to focus on the orientations while ignoring the colors in the multi-color condition.The inter-trial interval was 1000-1200 ms.Participants completed 96 trials in both single-color and multi-color conditions,in which 50% of trials consisted of similar orientations and the other 50% of dissimilar orientations.Similar and dissimilar trials were randomly mixed.Participants practiced at least 16 trials before starting the formal experiment to ensure that they understood the task.The formal experiment lasted for about 10 minutes,and a 3 to 5-minute break after 48 trials was provided.

Figure 1A schematic illustration of a single trial in Experiment 1.The color of bars in memory array here are blue,green and red.And the blue bar changes its orientation in probe.

Analysis

A two-way mixed analysis of variance(ANOVA)was conducted on accuracy.Accuracy was determined in terms of signal detection theory,which allowed us to disentangle the sensitivity(d’)to the change from the response bias(criterion).

Results

Figure 2The results of d’(A)and criterion(B)for Experiment 1.The smaller the criterion was,the larger the tendency was for participants to respond that the orientation changed.The error bars signify standard errors.

Discussion

Experiment 1 re-established the similarity effect in VWM using a simultaneous display,adding extra evidence that the display manner did not affect the similarity effect(e.g.,Johnson et al.,2009;Lin & Luck,2009).Moreover,the similarity effect was not impaired by the irrelevant colors,implying that orientation is represented independently in VWM.However,it is possible that the irrelevant colors were not extracted into VWM in the current multi-color condition(Woodman & Vogel,2008).Since no study has directly examined the OBE of VWM using the stimuli in Experiment 1,Experiment 2 addressed this issue while re-examining the similarity effect.

Experiment 2:Irrelevant

Color is Stored in VWM

Experiment 2 examined whether the irrelevant colors were automatically extracted into VWM when participants were instructed to memorize orientations while ignoring colors for the stimuli used in Experiment 1.

To address the involuntary selection of the irrelevant colors,we probed an “irrelevant-change distracting effect” in VWM,which is commonly used in exploring the OBE in VWM(e.g.,Yin,Zhou,et al.,2012;Zhang et al.,2013).This distracting effect is usually demonstrated in a change detection task,wherein the change type of irrelevant features(no irrelevant-change vs.irrelevant-change)of the probed stimuli is manipulated.It has been demonstrated that when participants are required to remember one feature dimension(e.g.,shape)while ignoring the other basic features(e.g.,color)of multiple-featured objects(e.g.,colored shapes),a change(to a new feature value)of the irrelevant feature at the probe phase dramatically impairs detection of the target dimension,suggesting that the irrelevant features were automatically selected into VWM(e.g.,T.Gao et al.,2011;Yin,Zhou,et al.,2012).Moreover,in line with the behavioral evidence,ERP studies showed that the irrelevant feature change elicited a more negative ERP component of the anterior N2 relative to the no-change condition(e.g.,Z.Gao et al.,2010;Yin,Gao,et al.,2012).

Based on the multi-color block of Experiment 1,Experiment 2 further manipulated the change type of color at the probe phase:in 50% of the trials,the irrelevant color changed to a new color.If the colors were automatically selected into VWM,then the change of colors would lead to an irrelevant-change distracting effect.Meanwhile,we predicted that the similarity effect revealed in Experiment 1 would be obtained again.

Methods

Twenty undergraduate or graduate students at the Zhejiang University(8 male;aged 18-25 years)participated in the experiment.Stimuli were the same as that used in the multi-color condition of Experiment 1.

The procedures were the same as Experiment 1,except that both the relevant and irrelevant features of the probe changed independently(with a probability of 50%)relative to the memorized stimuli.In the irrelevant change condition,the color of the probe changed to a new color not previously used in the corresponding memory array.

A 2(similarity:similar vs.dissimilar)× 2(irrelevant change:change vs.no-change)within-subjects design was used.There were 32 trials in each condition,resulting in 128 total trials,which were presented randomly.The experiment was divided into 2 blocks with a 3- to 5-minute break between blocks,lasting about 15 minutes overall.

Two-way repeated ANOVAs with similarity and irrelevant-change as within-subjects factors were conducted ond’andcriterion.

Results

Figure 3The results ofd′(A)and criterion(B)for Experiment 2.The smaller the criterion was,the larger the tendency was for participants to respond that the orientation changed.The error bars signify standard errors.

Discussion

In line with previous findings(Shen,Tang,et al.,2013;Yin,Zhou,et al.,2012),the results of Experiment 2 demonstrate that the change of task-irrelevant colors significantly affected participants’ judging criterion,indicating that the irrelevant colors were encoded into VWM.However,consistent with the findings of Experiment 1,Experiment 2 re-established the similarity effect.Therefore,the finding of Experiment 1 was not due to that the irrelevant colors were not processed into VWM.Experiments 1 and 2 together suggest that although their relevant colors were extracted into VWM,the independent representations of orientations still existed in VWM.

It is of note that the current irrelevant no-change condition was actually the multi-color condition of Experiment 1.However,as demonstrated in Experiment 2,the change of their relevant colors did not affect the similarity effect,providing further evidence that orientations have independent representations from colors,and the similarity effect has a close relationship with the maintenance phase of VWM.

Experiment 3:Integrated

Object Exits in WM

Experiment 3 examined whether VWM also kept integrated objects containing both colors and orientations when the participants were only required to retain orientations.We explored this issue by probing the “irrelevant-change distracting effect” as used in Experiment 2,while using the settings of the dissimilar condition.However,unlike in Experiment 2 or in previous OBE studies(e.g.,Z.Gao et al.,2010;Yin,Zhou,et al.,2012),the irrelevant color of the probe was always an old value which had been used in the corresponding memory array.We specifically manipulated whether the color-orientation binding changed relative to the memorized objects(binding-same vs.binding-change).If there were integrated representations in VWM,the change of binding would significantly impair the orientation judgment.

Methods

Twenty undergraduate or graduate students at the Zhejiang University(10 male;aged 18-24 years)participated in the experiment.

The procedures were the same as in the dissimilar condition of Experiment 2,except for the aforementioned changes.Since the binding between orientation and color always changed when the probed orientation changed to a new value(the overall accuracy for orientation-change was 64% averaged across trials),only the data in the condition where the orientation remained the same between the memory array and probe were analyzed.A pairedt-test was conducted on the accuracy(percentage of correct responses)between binding-change and binding-same conditions.

Results & Discussion

The pairedt-test on the accuracy(see Fig.4)revealed that the accuracy was significantly lower in the binding-change(66%)condition compared to the binding-same condition(79%),t(19)= 2.67,p<0.05,Cohen’sd=0.45.This finding suggests that the color-orientation binding was maintained in VWM.Therefore,integrated representations existed in VWM,despite the fact that only the orientation had to be remembered.

Figure 4The results for Experiment 4,when theprobed orientation was an old one in the memory array while the binding between color and orientation was manipulated.The error bars signify standard errors.

General Discussion

The current study explored the building blocks of VWM by examining whether both feature and object representations co-exist in VWM.We presented participants with two-featured bars(i.e.,color and orientation),and required them to memorize only the orientations.In line with the hybrid view of VWM(Brady et al.,2011;Cowan et al.,2013;Wheeler & Treisman,2002),the three experiments of the current study provide empirical evidence demonstrating that there were independent orientation representations and integrated representations containing color and orientation in VWM.

Recently,researchers showed that when participants were required to remember both the colors and orientations of colored bars,the memory precision for color and orientation were independent from each other,favoring a feature-based instead of an object-based view of VWM(Bays et al.,2011;Fougnie & Alvarez,2011).The current study on the one hand adds new evidence supporting that there are independent feature representations stored in VWM for multi-featured objects,by using a new method(i.e.,probing the similarity effect).As the irrelevant color was involuntarily encoded into VWM,the similarity effect would be at least weakened if only integrated object representations existed in VWM.However,Experiments 1 and 2 demonstrated that the similarity effect for the orientation dimension was still observed and that it was not weakened.On the other hand,the current study also suggests that the feature-based view can only explain a fraction of results.Although the color was an irrelevant dimension,the binding between color and orientation(i.e.,an integrated object file)was also kept in VWM.Consequently,the participants were prone to make an incorrect “change”response when the binding changed but the probed orientation was the same(Experiment 3).Therefore,the current study provides clear-cut evidence showing that both independent features and integrated objects co-exist in VWM.

Although the current study did not directly examine which hybrid view of VWM is more accurate,certain parts of our findings were not congruent with the prediction of the dual-storage(Wheeler & Treisman,2002)or flat-structure views(Cowan et al.,2013).Inparticular,both views suggest that binding information will not be extracted into VWM until the task requires the participants to keep such information.However,as demonstrated in the current Experiment 3,the binding information was maintained in VWM regardless of taskdemand.In addition,the flat-structure view suggests that once all the features of an object are stored in VWM,these features will not be independent from each other,but rather will be stored as integrated objects(Cowan et al.,2013).With this view,it is difficult to explain the results revealed in Experiments 1,2 and 3.Specifically,to observe a similarity effect in Experiments 1 and 2,we need to have at least two orientation representations in VWM,which would leave at most only one integrated object in VWM.Given that in Experiment 3 the probed item was randomly selected and the interested binding-change occurred in 25% of trials,it is hard to explain why the binding-change exerted such a large effect(13%)on the orientation performance.However,these results do not necessarily support a hierarchical-structure view of VWM,as we have not provided any evidence to support the existence of a hierarchical structure in VWM.This should be examined in future studies.Indeed,a dual-storage view without any attention constraint on binding storage could also explain the current findings.

The present study also contributes at least two new aspects to the storage mechanisms of VWM.First,previous studies explored how perceptual information is extracted into VWM(e.g.,T.Gao et al.,2011;Z.Gao et al.,2010,2016),and we demonstrated that objects containing basic features could be encoded into VWM in an object-based manner,in which even irrelevant basic features are encoded.However,how the information is maintained after the relevant and irrelevant features are extracted into VWM has not been addressed directly.The current study suggests that both independent features and integrated objects are kept in VWM.Yet since the current study only examined one type of stimulus,additional studies are required to verify this finding.Second,the current Experiment 3 provided new evidence supporting the hypothesis that the basic features,which are processed in parallel at the perceptual stage,are stored in VWM in an integrated object manner.Previously,this hypothesis was examined by investigating whether the VWM performance for actively maintaining feature binding(e.g.,color-shape binding)was as good as maintaining constituent single features in a change detection task(T.Gao et al.,2011).Although this hypothesis was confirmed,the uncontrolled comparison difficulty at the testing phase between binding and feature conditions may have contaminated our previous finding(Awh,Barton,& Vogel,2007).Experiment 3 probed the irrelevant-change distracting effect,and provided direct evidence that even passively formed bindings can be stored as integratedrep resentations in VWM.However,it should be noted that our previously proposed object in VWM is a conventional concept(i.e.,the integrated object);the current study implies that this narrow object concept may need to be broadened as a(hierarchical)feature bundle.

Finally,the current study for the first time demonstrated the similarity effect under a multi-featured context in which the irrelevant information is also processed.Moreover,we found that the similarity effect was not affected by the change of irrelevant colors,suggesting that it is robust and its decision process is not affected by the irrelevant colors.Future VWM models explaining the similarity effect needs to take these aspects into consideration.

To conclude,the current study suggests that both feature and integrated object representations exist in VWM,thus supporting a hybrid view of VWM unit.

Reference

Alvarez,G.A.,& Cavanagh,P.(2004).The capacity of visual short-term memory is set both by visual information load and by number of objects.PsychologicalScience,15(2),106-111.doi:01502006 [pii]

Awh,E.,Barton,B.,& Vogel,E.K.(2007).Visual working memory represents a fixed number of items regardless of complexity.PsychologicalScience,18(7),622-628.doi:PSCI1949 [pii] 10.1111/j.1467-9280.2007.01949.x

Baddeley,A.(2012).Working memory:Theories,models,and controversies.AnnualReviewofPsychology,63(1),1-29.doi:10.1146/annurev-psych-120710-100422

Bays,P.M.,Wu,E.Y.,& Husain,M.(2011).Storage and binding of object features in visualworking memory.Neuropsychologia,49(6),1622-1631.doi:S0028-3932(10)00556-7[pii] 10.1016/j.neuropsychologia.2010.12.023

Brady,T.F.,Konkle,T.,& Alvarez,G.A.(2011).A review of visual memory capacity:Beyond individual items and toward structured representations.JournalofVision,11(5).doi:10.1167/11.5.4 4 [pii] 11.5.4 [pii]

Conrad,R.(1964).Acoustic confusions in immediate memory.BritishJournalofPsychology,55(1),75-84.

Cowan,N.,Blume,C.L.,& Saults,J.S.(2013).Attention to attributes and objects inworking memory.JournalofExperimentalPsychology:Learning,Memory,andCognition,39(3),731-747.doi:10.1037/a0029687

Cowan,N.,Elliott,E.M.,Scott Saults,J.,Morey,C.C.,Mattox,S.,Hismjatullina,A.,& Conway,A.R.(2005).On the capacity of attention:Its estimation and its role inworking memory and cognitive aptitudes.CognitivePsychology,51(1),42-100.doi:S0010-0285(05)00002-2 [pii] 10.1016/j.cogpsych.2004.12.001

Eng,H.Y.,Chen,D.,& Jiang,Y.(2005).Visual working memory for simple and complex visual stimuli.PsychonomicBulletin&Review,12(6),1127-1133.

Eriksson,J.,Vogel,E.K.,Lansner,A.,Bergstrom,F.,& Nyberg,L.(2015).Neurocognitive architecture of working memory.Neuron,88(1),33-46.

Fougnie,D.,& Alvarez,G.A.(2011).Object features fail independently in visual working memory:Evidence for a probabilistic feature-store model.JournalofVision,11(12).doi:10.1167/11.12.3 3 [pii] 11.12.3 [pii]

Fukuda,K.,Awh,E.,& Vogel,E.K.(2010).Discrete capacity limits in visual working memory.CurrentOpinioninNeurobiology,20(2),177-182.doi:S0959-4388(10)00043-7 [pii] 10.1016/j.conb.2010.03.005

Gao,T.,Gao,Z.,Li,J.,Sun,Z.,& Shen,M.(2011).The perceptual root of object-based storage:An interactive model of perception and visual working memory.JournalofExperimentalPsychology:HumanPerceptionandPerformance,37(6),1803-1823.doi:2011-23772-001 [pii] 10.1037/a0025637

Gao,Z.,Li,J.,Liang,J.,Chen,H.,Yin,J.,& Shen,M.(2009).Storing fine detailed information in visual working memory-evidence from event-related potentials.JournalofVision,9(7),1-12.doi:10.1167/9.7.17 /9/7/17/ [pii]

Gao,Z.,Li,J.,Yin,J.,& Shen,M.(2010).Dissociated mechanisms of extracting perceptual information into visual working memory.PLoSONE,5(12),e14273.doi:10.1371/journal.pone.0014273

Gao,Z.,Yu,S.,Zhu,C.,Shui,R.,Weng,X.,& Li,P.,et al.(2016).Object-based encoding in visual working memory:evidence from memory-driven attentional capture.ScientificReports,6.

Johnson,J.S.,Spencer,J.P.,Luck,S.J.,& Schoner,G.(2009).A dynamic neural field model of visual working memory and change detection.PsychologicalScience,20(5),568-577.doi:PSCI2329 [pii] 10.1111/j.1467-9280.2009.02329.x

Lin,P.H.,& Luck,S.J.(2009).The influence of similarity on visual working memory representations.VisualCognition,17(3),356-372.doi:10.1080/13506280701766313

Luck,S.J.,& Vogel,E.K.(1997).The capacity of visual working memory for features and conjunctions.Nature,390(6657),279-281.doi:10.1038/36846

Luck,S.J.,& Vogel,E.K.(2013).Visual working memory capacity:From psychophysics and neurobiology to individual differences.TrendsinCognitiveSciences,17(8),391-400.doi:10.1016/j.tics.2013.06.006

Luria,R.,Sessa,P.,Gotler,A.,Jolicoeur,P.,& Dell’Acqua,R.(2010).Visual short-term memory capacity for simple and complex objects.JournalofCognitiveNeuroscience,22(3),496-512.doi:10.1162/jocn.2009.21214

Luria,R.,& Vogel,E.K.(2011).Shape and color conjunction stimuli are represented as bound objects in visual working memory.Neuropsychologia,49(6),1632-1639.doi:S0028-3932(10)00511-7 [pii] 10.1016/j.neuropsychologia.2010.11.031

Luria,R.,Balaban,H.,Awh,E.,& Vogel,E.K.(2016).The contralateral delay activity as a neural measure of visual working memory.Neuroscience&BiobehavioralReviews,62,100-108.

Mate,J.,& Baqués,J.(2009).Visual similarity at encoding and retrieval in an item recognition task.QuarterlyJournalofExperimentalPsychology,62,1277-1284.doi:10.1080/17470210802680769

Oberauer,K.,& Eichenberger,S.(2013).Visual working memory declines when more features must be remembered for each object.MemoryandCognition,41(8),1212-1227.doi:10.3758/s13421-013-0333-6

Olson,I.,& Jiang,Y.(2002).Is visual short-term memory object based? Rejection of the“strong-object” hypothesis.Perception&Psychophysics,64(7),1055-1067.

Parra,M.A.,Cubelli,R.,& Della Sala,S.(2011).Lack of color integration in visual short-term memory binding.MemoryandCognition,39(7),1187-1197.doi:10.3758/s13421-011-0107-y

Parra,M.A.,Della Sala,S.,Logie,R.H.,&Morcom,A.M.(2014).Neural correlates ofshape-color binding in visual working memory.Neuropsychologia,52,27-36.doi:10.1016/j.neuropsychologia.2013.09.036

Sanocki,T.,& Sulman,N.(2011).Color relations increase the capacity of visual short-term memory.Perception,40(6),635-648.doi:10.1068/p6655

Shapiro,K.L.,& Miller,C.E.(2011).The role of biased competition in visual short-term memory.Neuropsychologia,49(6),1506-17.doi:10.1016/j.neuropsychologia.2011.02.017

Shen,M.,Tang,N.,Wu,F.,Shui,R.,& Gao,Z.(2013).Robust object-based encoding invisual working memory.JournalofVision,13(2),1.doi:10.1167/13.2.1

Shen,M.,Yu,W.,Xu,X.,& Gao,Z.(2013).Building blocks of visual working memory:Objects or Boolean maps?JournalofCognitiveNeuroscience,25(5),743-753.doi:10.1162/jocn_a_00348

Suchow,J.W.,Fougnie,D.,Brady,T.F.,& Alvarez,G.a.(2014).Terms of the debate on the format and structure of visual memory.Attention,Perception&Psychophysics.doi:10.3758/s13414-014-0690-7

Unsworth,N.,Fukuda,K.,Awh,E.,& Vogel,E.K.(2014).Working memory and fluid intelligence:Capacity,attention control,and secondary memory retrieval.CognivePsychology,71,1-26.doi:10.1016/j.cogpsych.2014.01.003

Vogel,E.K.,& Machizawa,M.G.(2004).Neural activity predicts individual differences in visual working memory capacity.Nature,428(6984),748-751.doi:10.1038/nature02447 nature02447 [pii]

Vogel,E.K.,McCollough,A.W.,& Machizawa,M.G.(2005).Neural measures reveal individual differences in controlling access to working memory.Nature,438(7067),500-503.doi:nature04171 [pii] 10.1038/nature04171

Vogel,E.K.,Woodman,G.F.,& Luck,S.J.(2001).Storage of features,conjunctions and objects in visual working memory.JournalofExperimentalPsychology:HumanPerceptionandPerformance,27(1),92-114.

Vul,E.,& Rich,A.N.(2010).Independent sampling of features enables conscious perception of bound objects.PsychologicalScience,21,1168-1175.doi:10.1177/0956797610377341

Wei,Z.,Wang,X.-J.,& Wang,D.-H.(2012).From distributed resources to limited slots in multiple-item working memory:A spiking network model with normalization.TheJournalofNeuroscience,32(33),11228-40.doi:10.1523/JNEUROSCI.0735-12.2012

Wheeler,M.E.,& Treisman,A.M.(2002).Binding in short-term visual memory.JournalofExperimentalPsychology:General,131(1),48-64.

Woodman,G.F.,& Vogel,E.K.(2008).Selective storage and maintenance of an object’s features in visual working memory.PsychonomicBulletin&Review,15(1),223-229.

Xu,Y.(2002a).Encoding color and shape from different parts of an object in visual short-term memory.Perception&Psychophysics,64(8),1260-1280.

Xu,Y.(2002b).Limitations of object-based feature encoding in visual short-term memory.JournalofExperimentalPsychology:HumanPerceptionandPerformance,28(2),458-468.

Xu,Y.,& Chun,M.M.(2006).Dissociable neural mechanisms supporting visual short-term memory for objects.Nature,440(7080),91-95.doi:nature04262 [pii]10.1038/nature04262

Yin,J.,Gao,Z.,Jin,X.,Ding,X.,Liang,J.,& Shen,M.(2012).The neural mechanisms of percept-memory comparison in visual working memory.BiologicalPsychology,90(1),71-79.doi:10.1016/j.biopsycho.2012.02.023 S0301-0511(12)00045-2 [pii]

Yin,J.,Gao,Z.,Jin,X.,Ye,L.,Shen,M.,&Shui,R.(2011).Tracking the mismatch information in visual short term memory:An event-related potential study.NeuroscienceLetters,491,26-30.

Yin,J.,Zhou,J.,Xu,H.,Liang,J.,Gao,Z.,& Shen,M.(2012).Does high memory load kick task-irrelevant information out of visual working memory?PsychonomicBulletin&Review,19(2),218-224.doi:10.3758/s13423-011-0201-y

Zhang,Q.,Shen,M.,Tang,N.,Zhao,G.,& Gao,Z.(2013).Object-based encoding in visual working memory:A life span study.JournalofVision,13(10).doi:10.1167/13.10.11

Zhou,J.,Yin,J.,Chen,T.,Ding,X.,Gao,Z.,& Shen,M.(2011).Visual working memory capacity does not modulate the feature-based information filtering in visual working memory.PLoSONE,6(9),e23873.

傳統視覺工作記憶(VWM)的研究認為視覺客體以特征或客體方式存儲。最近研究者提出特征和客體可能在VWM中并存，但尚未有研究直接對其檢驗。本研究采用變化覺察范式，以三個實驗對該問題進行了探討。實驗中要求被試記憶有顏色的線條的朝向而忽略顏色。結果發現，無關維度顏色得到加工，但并不影響朝向特征的高相似效應(特征間相似性越高，記憶績效越高)；同時顏色與朝向的綁定關系存在。該結果支持特征與客體同時在VWM存儲的假設。

The basic unit of object representation in visual working memory(VWM)is a fundamental yet unresolved question.Recently,researchers have proposed that individual-feature and integrated-object representations co-exist in VWM.However,no study has directly examined this hypothesis.The current study explored this issue in three experiments,wherein we presented participants with three two-featured bars(color and orientation),and required them to memorize the orientations and ignore the colors in a change-detection task.Experiments 1 and 2 examined whether independent orientation representations exist in VWM by exploring the similarity effect,which takes place on feature representations and shows better performance in memorizing similar features than in dissimilar ones.Experiments 1 and 2 observed the similarity effect when only the orientations were memorized,although the irrelevant colors were also extracted into VWM.Moreover,the similarity effect of orientation was not weakened by the added colors,suggesting that orientation representations exist in VWM.Experiment 3 examined whether there were integrated objects containing both color and orientation in VWM,by manipulating the color-orientation binding while the orientation was kept constant.We found that the change of binding dramatically impaired the memorization of orientation.Taking together,these results support a hybrid view for the building blocks of VWM and suggest that integrated objects and independent features co-exist in VWM.

visual working memory,object,feature,similarity effect,binding

表征在視覺工作記憶中的存儲單位：特征、客體或二者并存?

宋超劉婉祎魯溪芊顧全

(浙江大學心理與行為科學系，杭州 310007)

視覺工作記憶,客體,特征,相似性效應,綁定

浙江大學本科生探究性實驗教學改革項目。

B849

A

1006-6020(2016)-02-0112-15

**co-first author

***通信作者：顧全，男，浙江大學心理系博士生，E-mail:448360993@qq.com。