The Mechanisms Of Spatial Attention To Visual Working Memory

In the face of enormous visual inputs,our attention is very limited and can only focus on a small amount of information.Therefore,it is particularly important to use cueing to direct attention to items that are relevant to the current task.Spatial attention can be guided by two types of cues.One is the exogenous cue which is presented in the peripheral visual field,the other is the endogenous cue which is presented in the central visual field.Research over the past decades has shown that these two types of spatial cues involve different mechanisms in the selection of perceptual representations.Except for the external world,individuals should also direct their attention to the internal world,such as mnemonic information.To this end,it is still unclear whether there is a dissociative effect of the two cues on working memory representations.To address this scientific question,this dissertation adopted the retro-cue paradigm that insert spatial cues in the working memory delay period and combined multi-modality neuroimaging techniques to investigat the underlying mechanisms by which endogenous and exogenous cues affect visual working memory representations.The results will help in revealing the similarities and differences between mnemonic attention and perceptual attention.The dissertation contained the following three parts.The first part was a literature review and research proposal.First,this thesis systematically reviewed the development of research paradigms regarding visual working memory,the mechanisms of the retro-cue effect,and the main factors modulating the effect of spatial cues on performance;second,this thesis pointed out the shortcomings of the current research and the specific scientific questions that need to be solved,and putted forward targeted solutions and research hypotheses.The second part was the main body of this dissertation.We investigated the mechanisms by which spatial attention affects visual working memory representations through seven experiments combining kinds of neurotechs,such as eye-tracking,M/EEG,TMS and computational modeling.It was divided into four studies:In the first study,we investigated whether there was a dissociation effect between the effects of exogenous and endogenous retro-cues on visual working memory representations.In Experiment 1,participants were asked to memorize two or four Gabor patches with different orientations.After a short delay,they were supposed to reproduce one of the orientations at the probed location as precise as possible.In the retro-cue trials,either an exogenous/peripheral cue or an endogenous/cental cue was inserted into the delay period,always correctly pointing to the probed location(100% valid).Gaze positions were simultaneously monitored by the eye-tracking system.Our results showed that at load 2,exogenous cues were more effective than endogenous cues and the no cue condition,whereas at load 4,there was no significant difference in the facilitation of working memory performance by the two spatial cues.To investigate the boundary of the dissociation,Experiment 2 additionally included conditions with memory load of 1,3 and 6.Results showed that the dissociation effect between endogenous and exogenous retro-cues was reliable in the condition with a memory load of 2.In the second study,we aimed to investigate the neural mechanisms underlying the dissocation at load 2.In Experiment 3,we simultaneously recorded EEG signals using the same paradigm as in Experiment 1.Compared with endogenous cues,results showed that exogenous cues evoked stronger and earlier ERP lateralizations at frontal electrodes(i.e.,P1 ac and ADAN).At load 2,exogenous cues induced stronger contralateral theta synchronization than endogenous cues at occipital electrodes,and this difference negatively predicted the behavioral difference between the two.These results suggested that exogenous cues may facilitate the cued-item more effectively compared to endogenous cues at the load.Experiment 4 used MEG and multivariate pattern analysis(MVPA)to track the dynamics of the cued-item’s representation.Interestingly,we found that the representation of the cued-item was significantly stronger than that of the no-cue condition around 100 ms after an exogenous cue onset,which was also true around 300 ms after an endogenous cue onset.Consistent with these time windows,the results of Granger causality analysis showed that information flow from the dorsolateral prefrontal lobe(DLPFC)to the occipital cortex(LOC)existed within the first 200 ms after the exogenous cue onset,whereas a similar effect did not arise until 300-500 ms after the endogenous cue onset.These results implied that an early top-down control might be the main reason for facilitating the cued-item’s representation in the exogenous cue condition.In this experiment,we also set a condition in which 50% of the cues were valid and found that neither exogenous nor endogenous cues improved working memory performance,nor was there a significant difference between them,suggesting that the retro-cue effect required the involvement of executive control and that DLPFC may play an important role in it.To further investigate the causal role of DLPFC in the current task,single-pulse TMS and the retro-cue paradigm at a memory load of 2 were applied in Experiment 5.Both exogenous and endogenous cues were 100% valid.Another two brain regions(i.e.,IPS and Vertex)were also targeted for stimulation.The stimulation was given at 100 ms,400ms,or 700 ms after the cue onset.Our data showed that the intervention at DLPFC or IPS could disrupt the advantages of exogenous cues over endogenous cues,suggesting that these two areas contributed to the priority of exogenous cues.More importantly,we found that if the intervention at DLPFC was given at 100 ms after the cue onset,the retro-cue benefit caused by exogenous but not endogenous cues would be significantly disrupted.This finding provided a causal evidence supporting that the early activation of DLPFC led to the exogenous retro-cue benefit.In the third study,we investigated the properties of spatial cues affecting visual working memory representation.Self-confidence was closely related to voluntary control,Experiment 6 aimed to investigate whether exogenous retro-cues relyed on voluntary control to become effective.The experimental design was similar to Experiment 1,with the only difference being that participants should evaluate their confidence on a 1-9 scale after each reproduction test.Higher scores represented higher levels of self-confidence.It was found that exogenous retro-cues enhanced rating scores which was unlike perceptual attention,suggesting that the process by which exogenous cues affect working memory representations was not purely automatic but an active control process existed.Experiment7 aimed to investigate the timings of spatial cues leading to the retro-cue benefit.The cueto-probe intervals were set to 100 ms,200ms,and 1000 ms.The results showed that exogenous cues only took 100 ms to become effective at load 2,a time window similar to that in which exogenous cues function in perceptual tasks,whereas at load 4,exogenous and endogenous cues only had a facilitative effect at the interval of 1000 ms.Besides,we conducted a meta-analysis based on the behavioral data from the above seven experiments.The robust dissociation at load 2 across experiments suggested a convincing consistence of our studies.In the fourth study,we investigated the relationship between modeling parameters and neural activities.To achieve this,we fitted our behavioral data using the Standard Mixture model to extract two parameters: the guess rate and 1/precision.The smaller the value,the better the working memory performance.In two datasets(Experiment 1&3,Experiment4&5),we consistenly found that at load 2,exogenous cues reduced the probability of guessing but not improved precision relative to endogenous cues or no-cue conditions.More interestingly,the differential ADAN amplitudes(Exepriment 3)and the DLPFC activations(Exepriment 4)between the two spatial cues were significantly correlated with their guess rate but not precision differences.Further,the DLPFC-targeted TMS given at post-cue 100 ms specifically interfered with the beneficial effect caused by exogenous cues in the guess rate but not precision(Exepriment 5).Together,these results suggested that the advantage of exogenous cues over endogenous retro-cues can be explained by the rapid activation of the DLPFC to reduce the guessing probability.Finally,we also found that when subjects were asked to do a self-confidence rating on the reproduction task,the two spatial cues were no longer significantly different at load 2.Meanwhile,endogenous cues also reduced the probability of guessing relative to the no cue condition(Experiment 6).We suspected that self-confidence ratings made the activation of the DLPFC similar in the task for both spatial cues,thus leading to the disappearance of behavioral differences.The final part of this dissertation was general discussion and outlooking.Based on the existing experiments,we not only illustrated the theoretical contributions of these findings to the current study,but also suggested research directions for subsequent experiments.In summary,we combined behavioral experiments with neuroimaging and computational modeling techniques to conclude that:(1)under low memory load,spatial retro-cues improved working memory performance by “facilitating the memory representation of the cued-item”.Its neural mechanism was the activation of a top-down control pathway that caused “transient” firing of neurons that represent the cued item;under high memory load,spatial retro-cues might improve working memory performance by “suppressing interference from irrelevant stimuli”.(2)Compared with endogenous retro-cues,exogenous retro-cues were more effective under the low memory load,which was an outcome of the “ahead start of retrieval”.The behind neural mechanism was that exogenous retro-cues activated DLPFC and the representation of the cued item much earlier.(3)Mnemonic attention was more dependent on the activation of DLPFC than perceptual attention;the "noise level hypothesis" was expected to integrate the findings of mnemonic attention and perceptual attention.