This paper, presented by Xie et al., examines where, when, and how recurrent processing contributes to visual processing. Backward masking, in which a salient mask is shown immediately after an image to interfere with recurrence, was used to isolate recurrent from feedback activity to better examine recurrent processing. This backward masking paradigm was combined with a multivariate analysis of EEG and fMRI human brain activity data from participants who underwent a 2-alternative forced-choice (2-AFC) task to identify objects from neural signals under two masking conditions: backward masked or not backward masked. Through comparing neural data to feedforward convolutional neural network (CNN) data, further analysis of how recurrence shapes visual representations was conducted. The researchers discovered that visual object representations are typically impacted by recurrent processing after around 100 ms in the early visual cortex (EVC), and two phases around 175 and 300 ms with strong effects in the lateral occipital cortex (LOC). Using CNN models, it was found that recurrence can alter the feature format in the LOC from primarily mid to high-level features. Recurrence was also determined to be mediated through four spectro-temporal neural components by comparison of the two masking conditions with clusters found in the theta to beta frequency range. These results support the notion of the role of theta, alpha, and beta frequencies in mediating recurrent processing, cognition, and vision. However, they are unable to classify whether results were influenced by nonvisual regions as the fMRI data exclusively examined the ventral visual stream, thus highlighting the need for further studies assessing the entire brain.