The sensory recruitment hypothesis posits that frontal brain regions recruit early visual cortex (EVC) for high-fidelity representations during visual working memory (VWM). Strong support for this hypothesis comes from observations that VWM contents can be decoded from EVC with fMRI. However, delay period decoding is not typically found in electro- and magnetoencephalography (MEEG) studies. We hypothesize that this discrepancy emerges because frontal regions recruit EVC via subthreshold feedback that modulates synaptic thresholds but not firing, thereby driving robust decoding from fMRI but not MEEG signals. In this MEG study, we aim to make subthreshold feedback visible by exploiting stochastic resonance – a noise benefit in non-linear systems. Specifically, sub-threshold information (e.g., feedback) can become supra-threshold (e.g., translated into firing) by injecting just the right amount of noise. This supra-threshold signal is patterned by the current state of the system (e.g., VWM contents), making it available for read-out. Here, participants (N=11) remembered an orientation while EVC was activated by noise images shown throughout a 2s delay. Noise strength was manipulated by varying noise contrast across trials. Preliminary results show significantly enhanced VWM decoding when weak noise (<18% Michelson contrast) is presented during the delay, compared to baseline. In line with the idea of stochastic resonance, this implies that subthreshold feedback to EVC can be uncovered with a resonant amount of noise. With this ongoing study, we further hope to introduce a viable method to improve read-out of VWM using MEEG, and bridge the gap with fMRI findings.