In this paper, we present multimodal computational models of interactional trust in a humans-robot interaction scenario. We address trust modeling as a binary as well as a multi-class classification problem. We also investigate how early- and late-fusion of modalities impact trust modeling. Our results indicate that early-fusion performs better in both the binary and multi-class formulations, meaning that modalities have co-dependencies when studying trust. We also run a SHapley Additive exPlanation (SHAP) values analysis for a Random Forest in the binary classification problem, as it is the model with the best results, to explore which multimodal features are the most relevant to detect trust or mistrust