Electroencephalographic (EEG) data is a high-dimensional, noisy time series capturing the continuous evolution of neural response patterns. Dimensionality reduction methods such as Principal component analysis (PCA) and T-distributed Stochastic Neighbor Embedding (tSNE) have been used to analyze such data; however, they group time points by amplitude similarity rather than temporal order, making them unable to preserve the sequential structure. Existing time-aware approaches, such as Temporal-Potential of Heat-diffusion for Affinity-based Trajectory Embedding (T-PHATE) and Brain-dynamic Convolutional-Network-based Embedding (BCNE), address this by explicitly encoding temporal autocorrelation into the embedding, yet both have only been demonstrated on continuous recordings such as fMRI and not on epoched ERP-structured Electroencephalography (EEG) data. This study addresses this research gap and evaluates time-aware dimensionality reduction methods on Event-related Potentials (ERP) structured EEG data, investigating whether temporal awareness yields more meaningful embeddings than time-agnostic approaches. ERP components were simulated using the UnfoldSim package, and T-PHATE and BCNE were evaluated along with standard methods under two approaches: condition-averaged input and single-trial projection via grand average. The results show that T-PHATE and BCNE recover temporally ordered trajectories with condition-specific divergences, while time- agnostic methods yield fragmented embeddings. These findings suggest that time-aware methods offer a more faithful representation of ERP data and can be used for exploratory analysis of complex experimental designs