The present study investigated whether socially anxious people with impaired top-down control directed attention to task-irrelevant stimuli under a working memory load. Because people with impaired prefrontal top-down control have difficulty keeping separate irrelevant information, they will direct attention to the stimuli matching the contents in working memory even if they are task-irrelevant 
. The present study showed that as trait social anxiety increased, greater effects in visual attention by the working memory were observed. In particular, there was a delay in attentional disengagement from the task-irrelevant stimuli matched in the working memory among socially anxious people. Even when working memory load increased, impaired attentional disengagement in trait social anxiety was observed. Although many cognitive models in anxiety propose that impaired top-down attentional control enhances the processing of task-irrelevant stimuli 
, previous studies did not produce an attention to task-irrelevant stimuli under a cognitive load task in anxiety. This study elucidated that trait social anxiety had an influence on visual attention to non-emotional task-irrelevant stimuli that is identical to that of a memory cue under a working memory load.
In the present study, the impaired attentional disengagement from task-irrelevant stimuli was observed in trait social anxiety under not only low but also high working memory load. This is consistent with the attentional control theory, in which deficient top-down control is greater with a high cognitive load 
. Although some previous studies have shown that trait anxiety does not have an effect on attentional control under high load 
, they manipulated not working memory load but perceptual load. The effects of working memory load and perceptual load on top-down control are different 
. Contrary to the effects of perceptual load 
, high working memory load did not exclude the effects of impaired attentional control in trait social anxiety, but maintained theses effects. Manipulating working memory load but not perceptual load is an appropriate way to disrupt top-down control, which inhibits interference from task-irrelevant distractors. The present results suggest that increasing working memory load disrupts top-down control especially in individuals with trait social anxiety, and that they are unable to inhibit the task-irrelevant distractors.
The impaired attentional disengagement under high working memory load in trait social anxiety also suggests that individuals with high trait social anxiety do not necessarily have a low working memory capacity. According to previous studies, the effects of task-irrelevant stimuli on visual attention are not observed under high working memory load because of limited working memory capacity 
. If working memory capacity is reduced in high trait social anxiety as the previous studies showed 
, the effects of task-irrelevant stimuli might not be observed. The effects of task-irrelevant stimuli under high working memory load suggest that individuals with trait social anxiety have sufficient working memory capacity. However, it is still unclear whether individuals with high trait social anxiety have more visual working memory capacity than those with low trait social anxiety as in a previous study 
, because the working memory load in the present study was not too high. Because the average visual working memory capacity is three to four simple objects 
, two stimuli in the present study might not totally deplete working memory capacity. Previous studies also showed that the working memory load of two stimuli was insufficient for the attentional guidance effect to disappear, while the effect disappeared completely with more than three stimuli in working memory load 
. Future studies should increase the stimuli in working memory load to evaluate the effects of individual differences of working memory capacity in trait social anxiety.
Inconsistent with our hypothesis, impaired attentional disengagement was not enhanced in high trait social anxiety under high working memory load compared to that under low working memory load. Because high working memory load disrupts top-down control more than low working memory load, individuals with trait social anxiety might have difficulty in controlling attention under high working memory load. One possibility is that the low working memory load was enough to disrupt top-down control in individuals with trait social anxiety. Even after increasing working memory load, the effects of the load did not change. The other possibility is that the present task underestimated the effects of working memory load on visual attention in trait social anxiety. In the present task, high working memory load diminished the attention to task-irrelevant stimuli. Therefore, even if high working memory load enhances the processing of task-irrelevant stimuli in individuals with high trait social anxiety, the effects might not be observed as, for example, long RTs in invalid trials. Although the working memory load decreased attention to task-irrelevant stimuli in all participants, attentional disengagement index under high working memory load was not diminished compared to that under low working memory load in individuals with high trait social anxiety (low load: 87 ms; high load:82 ms). The invariant disengagement index might reflect the strong impaired attentional control under high working memory load. Future studies should use different visual task to reveal the effects of working memory load on attention in trait social anxiety.
We observed delayed disengagement rather than rapid engagement, from the memory cue in individuals with trait social anxiety. This is consistent with the findings of previous studies that showed that people with trait anxiety and trait social anxiety had difficulty disengaging from threatening stimuli 
. In many previous studies that showed delayed disengagement instead of rapid engagement, a threatening stimulus was presented unilaterally 
, whereas the present study presented several stimuli at the same time. Because an abrupt onset of stimuli captures attention automatically 
, these previous studies might have underestimated the effects of engagement. In fact, when presented with two stimuli at one time, individuals with high trait anxiety rapidly engage their attention to a threatening stimulus 
. In order to reveal whether rapid engagement in trait anxiety, which was not observed in the present study, is specific to emotional stimuli, future studies should use emotional stimuli in a similar task. However, the important point in the present results is that impaired attentional disengagement in trait social anxiety was observed not only for emotional stimuli but also for non-emotional stimuli. This finding is in line with recent studies that have shown that impaired attentional control in trait anxiety was observed not only for emotional stimuli but also for non-emotional stimuli 
, while few studies have shown the impaired attentional “disengagement” for non-emotional stimuli 
. The top-down control mechanisms in trait anxiety might be generally impaired.
We also measured state anxiety and depression for participants, but these did not predict the effects of working memory load on visual attention. According to cognitive models 
, trait anxiety impairs top-down control whereas state anxiety induces bottom-up attention to salient stimuli. Attention to stimuli matched in the working memory is dependent on an impaired top-down control, and enhanced bottom-up attention does not induce attentional attraction 
. Therefore, trait social anxiety, rather than state anxiety, predicted the delayed attentional disengagement in the present study. However, we only measured trait social anxiety, and it is still unclear whether individuals with trait anxiety also show impaired attentional disengagement under working memory load. Moreover, considering that individuals with social anxiety disorders have difficulty in attentional control 
, impaired attentional disengagement might be observed among them.
There was no association between performance in the working memory task and trait social anxiety. Considering the impaired top-down control in trait social anxiety, the performance might be negatively correlated with trait social anxiety. According to Derakshan and Koster 
, people with trait anxiety who have an impaired top-down control can maintain a high performance level (i.e., response accuracy), but at the expense of a reduced processing efficiency (i.e., response latency). Our findings suggest that accuracy rates in the working memory task were not reduced by trait social anxiety.
The present study chose to manipulate the working memory load on top-down control. However, top-down control includes not only working memory but also many other functions such as executive function and effortful control. The attentional guidance from working memory also depends on several other top-down controls 
. Since valid, neutral, and invalid trials were equally presented in the present study, it is difficult to determine whether participants would voluntarily direct attention toward the memory-matching item. Considering that all trials were presented equally, participants might not voluntarily direct attention to the stimuli that matched the content of WM. However, to accurately assess the effects of voluntary and involuntary attention, we need to manipulate the proportion of valid, neutral, and invalid trials. If there were no valid trials, participants could voluntarily inhibit the memory-matching items 
. The goal of attending to memory-matching stimuli might also play an important role for the attentional guidance from the working memory 
. If individuals with trait social anxiety have difficulty in voluntarily inhibiting information in working memory, they might also show impaired attentional disengagement in experiments void of valid trials. Future studies should investigate which specific top-down controls affect trait social anxiety.
In summary, we investigated the effects of impaired top-down control in trait social anxiety on visual attention to task-irrelevant stimuli under a working memory load. The ability of top-down control deteriorates under working memory load and people experience interference from task-irrelevant distractors. We showed that as trait social anxiety increased, interference from task-irrelevant stimuli also increased. Delayed attentional disengagement from the task-irrelevant stimuli matched with working memory was observed with an increase in trait social anxiety. Even when working memory load increased, impaired attentional disengagement in trait social anxiety was observed. Impaired top-down control in socially anxious people might enhance the processing of task-irrelevant stimuli under working memory load and prevent them disengaging from the stimuli.