Effects of memory load on the contralateral delay activity and induced alpha power in the EEG: studied with a virtual reality setup Felix Klotzsche 1,2, Michael Gaebler 1,2, Arno Villringer 1,2, Werner Sommer 1,3, Vadim Nikulin 1, and Sven Ohl 3 1 Max-Planck-Institut für Kognitions- und Neurowissenschaften, Leipzig, Deutschland 2 Humboldt-Universität zu Berlin, Faculty of Philosophy, Berlin School of Mind and Brain, Germany 3 Humboldt-Universität zu Berlin, Department of Psychology, Germany [email protected]

Introduction Methods 4° Stimulus low (2) Memory 24 participants EEG markers of visual Short-Term Memory (vSTM) 9° eccentricity high (4) load age range: 20–38y processes are well studied in conventional lab settings. 14° tested for color vision Contralateral Delay Activity (CDA) Alpha Oscillations Task: & stereopsis Delayed match-to- sample 720 trials in 10 blocks

Vogel & Machizawa, 2004 Leenders et al., 2016 Virtual Reality (VR) technology offers new possibilities to study these processes (e.g., in naturalistic settings), Experimental setup: Preprocessing: but combining a VR HTC Vive Pro (rf: 90Hz) with  ICA (EOG correlation) headset with EEG Pupil Labs eye-tracker (200Hz)  automated epoch rejection measurements introduces custom Unity 3D scripts utilizing UXF (Jas et al., 2017) new challenges. (Brooks et al., 2020)  rejecting trials with blinks or The weight of the headset 60 active EEG electrodes (10/20) saccades > 2° (Engbert et al., 2015) and its contact with the BrainProducts LiveAmp (500Hz) exclude subjects with >33% rejected electrodes can lead to epochs (n = 3) additional noise. The thereby lowered signal-to-noise Analyses: ratio (SNR) might render the components of interest CDA: Alpha oscillations: untraceable. Multivariate analysis approaches can help  difference wave  induced power (wavelet) from posterior ROI to increase the sensitivity of these measures. contralateral ERP − ipsilateral ERP  difference: contralateral power − ipsilateral power uni- Furthermore, VR offers a from posterior ROI 1A  extract frequency band 8–13Hz 2A variate wide field of view, but vSTM- components in the EEG have ? ? decoding memory load (low vs. high) mostly been studied at small  from voltage data (Adam et al., 2020)  from time-frequency decomposition of the data eccentricities (<5°).  sliding classifier (20ms window)  sliding classifier (500ms windows)  Common Spatial Patterns (CSP ) transformation multi- Research questions:  logistic regression (Blankertz et al., 2008) + LDA on power of the variate  Do we find EEG markers of vSTM when using a VR  „mini-blocks“: avg. across 10 trials 6 most discriminative CSP components headset instead of a desktop monitor?  5-fold repeated (50) cross-validation  5-fold repeated (50) cross-validation  Do they change when presenting the stimuli at  per subject & average ROC AUCs  per subject & average ROC AUCs larger eccentricities than previously studied? 1B 2B Results

1A CDA amplitude varies with memory load but not with stimulus eccentricity 2A Lateralized alpha power varies neither with memory load nor with eccentricity

CDA per memory load lateralized alpha power per memory load

contra- ipsi- lateral lateral power power encode retention cue encode retention

contralateral − ipsilateral CDA per eccentricity lateralized alpha power per eccentricity Memory load: F(1,20) = 17.62, ꞵ = 0.48, SE= 0.11, p < .001

Eccentricity: F(2,40) = 0.72, n.s.

Memory load : eccentricity: F(2,40) = 1.99, n.s.

1B Memory load can be decoded from voltage data for all eccentricities 2B Memory load can be decoded from alpha power for all eccentricities

across all eccentricities per eccentricity across all eccentricities

p < .01 retention

per eccentricity

4° 9° 14° 4° 9° 14° 4° 9° 14°

Top: Avg. decoding performance (ROC AUC) of the sliding classifier (log. regression) over time. Abs. spatial pattern weights of the most discriminative CSP component. Bottom: Spatial pattern weights of the classifier (normalized & averaged across subjects).

Discussion ✓ We replicated the lateralization of induced alpha power References Adam, K. C. S., Vogel, E. K., & Awh, E. (2020). Multivariate analysis reveals a generalizable human ✓ We replicated the effects of memory load on CDA in response to the cue and during memory retention. electrophysiological signature of working memory load. Psychophysiology, 57(12). https://doi.org/10.1111/psyp.13691 amplitude. Therefore, using a VR headset does not Blankertz, B., Tomioka, R., Lemm, S., Kawanabe, M., & Muller, K. (2008). Optimizing Spatial filters for Robust EEG Single-Trial Analysis. IEEE Signal Processing Magazine, 25(1), 41–56. cause a detrimental decrease of the SNR. ✓ This alpha lateralization was modulated neither by https://doi.org/10.1109/MSP.2008.4408441 Brookes, J., Warburton, M., Alghadier, M., Mon-Williams, M., & Mushtaq, F. (2020). Studying human behavior with memory load nor by stimulus eccentricity. virtual reality: The Unity Experiment Framework. Behavior Research Methods, 52(2), 455–463. https://doi.org/10.3758/s13428-019-01242-0 ✓ We show this for stimulus eccentricities of up to 14°. Jas, M., Engemann, D. A., Bekhti, Y., Raimondo, F., & Gramfort, A. (2017). Autoreject: Automated artifact rejection

for MEG and EEG data. NeuroImage, 159, 417–429. https://doi.org/10.1016/j.neuroimage.2017.06.030 Symposium,2021 ✓ Using spatial filters and a multivariate classifier, we Leenders, M. P., Lozano-Soldevilla, D., Roberts, M. J., Jensen, O., & De Weerd, P. (2018). Diminished Alpha Lateralization During Working Memory but Not During Attentional Cueing in Older Adults. Cerebral Cortex, ✓ A more sensitive, multivariate decoding approach could decode memory load from (unlateralized) alpha 28(1), 21–32. https://doi.org/10.1093/cercor/bhw345 Vogel, E. K., & Machizawa, M. G. (2004). Neural activity predicts individual differences in visual working memory

confirmed these findings. power for all eccentricities. capacity. Nature, 428(6984), 748–751. https://doi.org/10.1038/nature02447 MindBrainBody