Machine Learning (ML) has recently been used for decoding analyses in electroencephalogram (EEG) research to identify segments of time where neural activity is significantly related to the cognitive processing of particular stimuli. The cognitive neuroscience community has thus far primarily used Support Vector Machines (SVM) for decoding analyses, but this may be suboptimal with particular EEG datasets. Computer scientists, on the other hand, often analyze data with diverse algorithms, especially when it is not known a priori which algorithm would perform best with a dataset. Alternative models may offer comparable or better performance while enabling convergent evidence. In the present study, we demonstrate the utility of conducting multiple ML approaches by re-analyzing EEG datasets from experiments by Bae & Luck (2018; 2019b) using a heterogeneous ensemble of K-Nearest Neighbors, Naïve Bayes, Bootstrap Aggregating and Adaptive Boosting in addition to SVM. We investigated whether models that are more computationally demanding might improve on the original method. While no single model emerged as consistently superior to SVM based solely on decoding accuracy for these datasets, converging results from multiple models provide more robust evidence for drawing inferences by shielding against false positives. Specifically, heterogeneous ensemble analyses can provide greater confidence through utilizing diverse methodologies, ultimately contributing to a more comprehensive understanding of the neural data.