Alterations in working memory maintenance of fearful face distractors in depressed participants: An ERP study

Task-irrelevant threatening faces (e.g., fearful) are difficult to filter from visual working memory (VWM), but the difficulty in filtering non-threatening negative faces (e.g., sad) is not known. Depressive symptoms could also potentially affect the ability to filter different emotional faces. We tested the filtering of task-irrelevant sad and fearful faces by depressed and control participants performing a color-change detection task. The VWM storage of distractors was indicated by contralateral delay activity, a specific event-related potential index for the number of objects stored in VWM during the maintenance phase. The control group did not store sad face distractors, but they automatically stored fearful face distractors, suggesting that threatening faces are specifically difficult to filter from VWM in non-depressed individuals. By contrast, depressed participants showed no additional consumption of VWM resources for either the distractor condition or the non-distractor condition, possibly suggesting that neither fearful nor sad face distractors were maintained in VWM. Our control group results confirm previous findings of a threat-related filtering difficulty in the normal population while also suggesting that task-irrelevant non-threatening negative faces do not automatically load into VWM. The novel finding of the lack of negative distractors within VWM storage in participants with depressive symptoms may reflect a decreased overall responsiveness to negative facial stimuli. Future studies should investigate the mechanisms underlying distractor filtering in depressed populations.


Fig. S1
Grand average ERPs (averaged over P7/P8, P9/P10, and PO7/PO8) elicited by memory arrays under (A, D) the non-distractor condition, (B, E) the fearful distractor condition, and (C, F), the sad distractor condition for the (A-C) depressed group and (D-F) control group. Blue lines reflect the activity contralateral to the lateralized items, and red lines reflect the activity ipsilateral to the lateralized items. Gray shades indicate the analysis time window used to calculate the mean CDA amplitude. The waveforms were time-locked to the onset of the memory array (y-axis on time zero). contra = contralateral, ipsi = ipsilateral

Exploratory analyses for Ppc and N2pc components
A visual inspection of the difference waveforms suggested that early contralateral positivity and contralateral negativity were elicited in all conditions and that their amplitudes in the sad distractor condition were at least numerically different from those in the non-distractor condition. Early contralateral positivity is referred to as the positivity posterior contralateral (Ppc) component (Corriveau et al., 2012), or the early distractor positivity (Pd) component (Feldmann-Wustefeld & Vogel, 2019). The Ppc (80-160 ms) can be elicited by targets or distractors (Corriveau et al., 2012), and it reflects the initial processing of stimuli . The contralateral negativity is referred to as the N2pc component (Eimer, 1996;Hopf et al., 2000;Liu et al., 2016;Luck & Hillyard, 1994a, 1994bZhao et al., 2011) and has been widely used to investigate the deployment of attention (Liu et al., 2016;Luck & Hillyard, 1994a, 1994bZhao et al., 2011). A relatively common practice is to interpret the N2pc (180-320 ms) as an index of the deployment of covert spatial visual attention (Kiss et al., 2008) or of the onset of attentional engagement (Zivony et al., 2018). Therefore, based on our visual inspection and similar to previous studies using CDA (Corriveau et al., 2012;Feldmann-Wustefeld & Vogel, 2019), we conducted exploratory analyses to compare the Ppc and N2pc components under different conditions. The preprocessing and calculation of the amplitudes of the difference waveforms of the Ppc and N2pc components were conducted essentially as described for the CDA component.
As shown in Fig. S3A, we chose 82-132 ms after the onset of the memory array as the time window for the Ppc, which was defined as ±25 ms from the most positive peak between 80 ms and 160 ms in the difference waveform (the contralateral and ipsilateral waveforms are provided in the Supplementary Materials, Fig. S2). We also investigated the relationship between the initial processing and memory storage of the distractors by examining the correlations between the Ppc difference scores and the CDA difference scores for the different distractors. The results showed that the Ppc difference scores were positively correlated with the CDA difference scores (r = 0.353, p = 0.035) in the sad distractor condition, whereas no correlation was found between the Ppc difference scores and the CDA difference scores for the fearful distractors (r = .283, p = .095).
Similar to the calculations shown in the main text, we also calculated the mean Ppc amplitude difference score (i.e., Ppc amplitude in fearful/sad distractor condition minus Ppc amplitude in the non-distractor condition). This allowed us to examine the correlations between the mean Ppc amplitude difference scores and amount of symptoms (i.e., BDI-II and DASS-A scores) to determine whether depressive symptoms (BDI-II scores) or anxiety symptoms (DASS-A scores) affected the initial processing of the distractors. The Ppc difference scores with a more positive value indicate a larger Ppc in the distractor condition than in the non-distractor condition, suggesting that larger initial processing is elicited by the distractors. No significant correlation was found between the BDI-II scores and the Ppc difference scores for fearful distractors (r = -0.041, p = 0.814) or sad distractors (r = -0.106, p = 0.540).
No significant correlation was found between the DASS-A scores and the Ppc difference scores for fearful distractors (r = -0.030, p = 0.862) or sad distractors (r = -0.062, p = 0.721).
Follow-up pairwise comparisons showed that, for all participants, the N2pc amplitude was significantly larger in the sad distractor condition than in the fearful distractor We also investigated the relationship between attention allocation and memory storage of the distractors by examining the correlations between the N2pc difference scores and the CDA difference scores for the different distractors. The N2pc difference scores correlated positively with the CDA difference scores for the fearful distractors (r = 0.665, p < 0.001) and the sad distractors (r = 0.683, p < 0.001). We again calculated the mean N2pc amplitude difference score (i.e., N2pc amplitude in fearful/sad distractor condition minus N2pc amplitude in the non-distractor condition).
The score was used to examine the correlations between the mean N2pc amplitude difference scores and amount of symptoms (i.e., BDI-II and DASS-A scores) to determine whether the depressive symptoms (BDI-II scores) or anxiety symptoms

Discussion
Very few studies have systematically analyzed the Ppc component (Corriveau et al., 2012;Feldmann-Wustefeld & Vogel, 2019;Jannati et al., 2013), and its functional meaning is not clear. The Ppc component may be related to the stimulus-driven P1 (75-125 ms), which is sometimes larger in areas contralateral to the cued array than in those ipsilateral to it. This lateralization of P1 has been attributed to low-level sensory processes (Luck & Hillyard, 1994a). In our study, the Ppc may be elicited by both the targets and the distractors in the pre-attentive phase. Our results indicated that the Ppc was smaller under the sad distractor condition than under the fearful face distractor condition or the non-distractor condition. The differences in the initial perceptional processing between sad distractors and fearful distractors may also be related to the allocation of more attention to the cued hemifield (enhanced N2pc) by the participants under the sad distractor condition than under the other conditions. We also found a positive correlation between the Ppc difference scores and the CDA difference scores for the sad distractor, suggesting that the change in initial perceptional processing is associated with difficulty in filtering sad face distractors. Future studies should investigate the functional meaning of the Ppc component in the processing of distractors.
In addition to examining CDA and Ppc, we conducted an exploratory analysis of the N2pc component. Even if a well-documented negative bias toward sad faces exists in depression (e.g., Armstrong & Olatunji, 2012;Gotlib & Joormann, 2010), the CDA results did not indicate that task-irrelevant sad faces consumed working memory resources in the depressed group. The reason for the efficient filtering of sad faces in the depressed group is unclear. One possibility is that sad faces attracted the participants' attention, but they did not subsequently store those faces in VWM. The exploratory analysis showed that the N2pc was enlarged at the whole sample level in the sad distractor condition, but no group differences were found for the N2pc.
Although our study is the first to investigate facial filtering in depression, our results suggest that the negative bias toward sad faces in depression (Armstrong & Olatunji, 2012) seems not to influence the maintenance phase of VWM. However, further studies are required to confirm this finding with a larger group of participants.
One notable result of the present study was a positive correlation between the N2pc difference scores and the CDA difference scores, which is in line with the findings reported by Salahub and Emrich (2020). Our results demonstrate an association between an increased likelihood of holding a facial distractor in VWM (as indicated by the CDA) and an increased allocation of attention to the face distractor (as indicated by the N2pc). Nevertheless, since the exploratory analyses of N2pc and Ppc were conducted post hoc (i.e., after observing the ERP waveforms), these findings should be interpreted with caution.