Electroencephalographic (EEG) recordings allow to capture temporal activation patterns associated with the current level of workload or emotional states [1-4]. Decoding mental states from these activation patterns and reacting to them accordingly can increase performance, safety, and user experience during human-machine interactions, e.g., in medical surgery or autonomous driving. In such naturalistic environments, it is particularly important to integrate context information and identify the current locus of attention to achieve robust mental state decoding. When combining EEG signals with information regarding the eye movements acquired via eye-tracking, the analysis of neuronal temporal dynamics can be related to the fixation on or saccade towards a stimulus [5-9]. Multivariate pattern analysis (MVPA) receives increasing attention since it allows to distinguish subtle differences in temporal dynamics between conditions [10-11]. MVPA has mainly been applied to distinguish different sensory processes with rather low-level neuronal representation (e.g., [8,12-14]), especially in functional magnetic resonance imaging (fMRI). However, because of their high temporal resolution, magnetoencephalography (MEG) and EEG are particularly suited to unravel fine-grained temporal dynamics [10-11]. In a recent MEG study on elementary arithmetic, Pinheiro-Chagas and colleagues  used MVPA to successfully distinguish between successive additions vs. subtraction. Few studies examined temporal dynamics associated with emotional processing [16-17]. To the best of our knowledge, no study combined eye-tracking with EEG to investigate a fixation-related temporal decoding of higher cognitive processes. Therefore, we here investigate spatio-temporal dynamics of different emotional states and workload levels within a MVPA approach on fixation-related EEG recordings. We are interested whether we can distinguish between (a) emotional states when processing images with positive, neutral, and negative valence and (b) low and high workload.